From c1273557743d21f758b6a51e531ea4d6c17cbdba Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 9 Apr 2021 18:40:55 -0500 Subject: [PATCH 01/77] Add new feature proposal draft for cpgfunction A new feature proposal draft for a fast and accurate g-function calculation is included in this commit. A C++ library, named cpgfunction, with the ability to compute g-functions has been developed by Jack C. Cook, a student of Dr. Jeffrey D. Spitler at Oklahoma State University. The cpgfunction library is a low level implementation of Massimo Cimmino's g-function methodology. The code was developed with the aim of increasing performance (compared to Cimmino's open source Python implementation, pygfunction), in both memory and time to be used in the generation of g-function databases on high performance computing clusters (HPCC). The resulting program makes it more practical to compute g-functions for larger ground heat exchangers on a desktop computer. More information about the g-function calculation can be found in the NFP. This proposal specifically aims to address issue #6651, to provide EnergyPlus with a 3rd party g-function calculation. --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 290 ++++++++++++++++++ .../F1_memory_comparison.png | Bin 0 -> 27193 bytes .../F2_timing_comparison.png | Bin 0 -> 31555 bytes .../F3_memory_reduction.png | Bin 0 -> 54159 bytes .../images_cpgfunction/F4_C_configuration.png | Bin 0 -> 46280 bytes .../F5_g_funtion_C_configuration.png | Bin 0 -> 44148 bytes 6 files changed, 290 insertions(+) create mode 100644 design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md create mode 100644 design/FY2021/images_cpgfunction/F1_memory_comparison.png create mode 100644 design/FY2021/images_cpgfunction/F2_timing_comparison.png create mode 100644 design/FY2021/images_cpgfunction/F3_memory_reduction.png create mode 100644 design/FY2021/images_cpgfunction/F4_C_configuration.png create mode 100644 design/FY2021/images_cpgfunction/F5_g_funtion_C_configuration.png diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md new file mode 100644 index 00000000000..0f74d2175c3 --- /dev/null +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -0,0 +1,290 @@ +New Feature Proposal Draft + +# Fast and accurate g-function calculations for modeling ground heat exchangers + + +**Jeffrey D. Spitler and Jack C. Cook, OSU** + +- Original Date 2021-04-09 + +## Justification for New Feature +In order for EnergyPlus to simulate a ground heat exchanger (GHE) used with a +ground-source heat pump (GSHP) system, a dimensionless response function, +called a g-function, is required. The dimensionless response depends primarily +on geometry – locations of boreholes and depths – though the borehole thermal +resistance and flow rate can also have a minor effect. Users of EnergyPlus +currently have three possibilities for obtaining the g-function: + +- EnergyPlus can calculate the g-function with a boundary condition known as +“uniform heat flux” – that is the heat flux per unit length is assumed uniform +over the entire fields. This assumption works reasonably well for small fields, +but causes significant errors as the number of boreholes increases. +(Malayappan and Spitler 2013) +- The g-function can be input by the end user after determining it for a +specific configuration with a 3rd party tool, GLHEPRO. (Spitler 2000, OSU 2016) +This has the disadvantage of requiring the user to use a commercially +available, non-free, software program. GLHEPRO relies upon a library of +g-functions that were pre-calculated using the uniform borehole wall +temperature boundary condition, so is not able to calculate g-functions on the +fly for larger fields. (GLHEPRO does have an implementation of the uniform +heat flux g-function calculation, but based on the findings in Malayappan +and Spitler (2013), limits its usage to a maximum of 36 boreholes. GLHEPRO +does write a snippet of IDF that can be pasted into the user’s IDF file. +- Another tool, pygfunction (Cimmino 2018b) could be used to calculate a +g-function for any configuration. As shown by Spitler et al. (2020), this +feature can be a powerful aid in optimizing the system performance. Pygfunction +requires the user to be familiar with Python programming, and to write the IDF +description from scratch. For larger fields, the computational time can become +excessive (e.g. hours). A more significant constraint is that the memory +requirements can vastly exceed what is available even on highly-resourced +workstations. + +As each of the above methods has significant limitations, it would be highly +desirable to add the capability to calculate g-functions with a method that is +as accurate as that available in GLHEPRO and pygfunction, but which allows +user-specified (non-library) borehole configurations, without excessive +computational time or memory requirements. + +## E-mail and Conference Call Conclusions + +N/A + +## Overview + +Starting with the methodology utilized by pygfunction to calculate g-functions +with uniform borehole wall temperature, a C++ implementation +(Cook and Spitler 2021) referred to as “cpgfunction” has been developed. +The solution has been considerably restructured to give an 8-fold reduction +in required memory. The reduction in memory depends on the number of finite +line source segments, as shown in Figure 1. This reduction in memory comes at +the expense of some computational speed, but for irregular configurations, as +shown in Figure 2, cpgfunction increases speed by a factor of four when there +are more than about 5000 sources and by higher amounts with smaller amounts of +sources. + +![F1_memory_comparison](images_cpgfunction/F1_memory_comparison.png) + +Figure 1 Comparison of pygfunction and cpgfunction memory requirements + +![F2_memory_comparison](images_cpgfunction/F2_timing_comparison.png) + +Figure 2 Comparison of pygfunction and cpgfunction computational speed for +irregular borehole configurations + +Beyond the improvements described by Cook and Spitler (2021), this new feature +will include an adaptive discretization algorithm to maximize speed and minimize +memory requirements while maintaining acceptable accuracy. + +## Approach + +The recently developed “cpgfunction” implementation (Cook and Spitler 2021) will +be adapted for use in EnergyPlus. In this section, we give an overview of the +methodology. Cpgfunction and pygfunction use a discretized representation of +the ground heat exchanger, with each borehole divided into multiple segments. +The effect of each segment on every other segment is determined with an +analytical solution known as the finite line source (FLS). When applied in +this way, with each borehole represented by multiple finite line sources, the +method is sometimes referred to as stacked finite line source (SFLS). + +The algorithm has three major steps: + +1. Discretization of the boreholes into finite line segments. The actual +discretization is simple – boreholes are divided into equal length segments – +but specifying a minimal number of segments that achieve sufficient accuracy is +more difficult. We are currently finalizing an algorithm that we call the +“adaptive discretization scheme” that ensures sufficient accuracy. The +methodology for this is similar to a grid-independency study. Past work in the +field has not included rigorous examination of this question. +2. Analysis of similarities and calculation of segment-to-segment responses. +For, say, a system with 200 boreholes and 12 segments per borehole, there are +24002 segment-to-segment responses to be calculated. This can be quite +computationally time-consuming and, therefore, as proposed by (Cimmino 2018a), +an analysis of similarities is done first. By similarities, we mean pairs of +segments with identical or near-identical horizontal and vertical offsets, such +that the segment-to-segment responses are identical or near-identical. +Therefore, an analysis is done first, and segment-to-segment responses are only +calculated for the unique cases. +3. Superposition of the segment-to-segment responses. This step may seem +simple, and it is simple if the uniform heat flux boundary condition is used +and all segments have a constant and uniform heat input. However, as shown by +Malayappan and Spitler (2013), this choice gives inaccurate results for larger +borefields. The uniform borehole wall temperature has been considered the +reference method for some years, has been widely used to size ground heat +exchangers and has had the benefit of some field validation. +(Cullin et al. 2015) Arguably, the uniform inlet fluid temperature +method has even better physical justification. Cpgfunction currently +implements the uniform borehole wall temperature boundary condition, and +as will be discussed below, this is advantageous. + +The project team is currently finalizing research into what we call the +“adaptive discretization scheme”, and it takes advantage of the fact that using +the uniform borehole wall temperature approximation with a reduced number of +segments can give a quite good match to the uniform inlet fluid temperature +boundary condition. This gives us the possibility of significant reduction +in both memory and time requirements while also obtaining a highly accurate +solution. The memory reduction is illustrated in Figure 3 for a borefield with +about 18000 segments. Cpgfunction using the same number of segments reduces the +memory required from 435 GB to 50 GB. Applying the adaptive discretization +scheme then gives a further 10-fold reduction in memory requirements to 5 GB. +The reduction in computation time depends on the amount of similarities in the +field. With an irregularly shaped field with few similarities, we would expect +a 4-fold reduction in using cpgfunction with the same number of segments and a +further 20-fold reduction using the adaptive discretization scheme. + +![F3_memory_reduction](images_cpgfunction/F3_memory_reduction.png) + +Figure 3 Reduction in memory requirements. Downward vertical arrow corresponds +to reduction with cpgfunction using the same number of segments as pygfunction. +Other arrow represents savings of adaptive discretization scheme. + +## Testing/Validation + +G-functions are specified as a series of points `(g, ln(t/t_s))`. A proposed +testing algorithm would involve calculation of g-functions for several borehole +configurations and comparison of the results with pre-calculated g-functions. +A sample C-shaped configuration is shown in Figure 4; the dimensions are in m. +Figure 5 shows the g-function corresponding to a 96m depth with the 27 discrete +values marked as points. The testing algorithm will compute the g-function, +also in 27 discrete values, and the RMSE between the EnergyPlus calculated +g-function and the pre-calculated g-function. An RMSE of less than 0.1% should +indicate a “pass”. + +![F4_C_configuration](images_cpgfunction/F4_C_configuration.png) + +Figure 4 Sample C-shaped configuration + +![F5_g_funtion_C_configuration](images_cpgfunction/F5_g_funtion_C_configuration.png) + +Figure 5 Sample g-function showing discrete values + +## I/O Reference Documentation + +EnergyPlus can already calculate g-functions with the uniform heat flux boundary +conditions. It would be possible to simply replace the algorithm for computing +g-functions and make no changes to the GroundHeatExchanger:System object. +Borehole locations will be specified with GroundHeatExchanger:Vertical:Single +objects. +However, for purposes of testing, validation, and research it is desirable to +retain the uniform heat flux boundary condition calculation as a non-default +option. Therefore, we propose a new key for the GroundHeatExchanger:System +object, the “g-function calculation model name”, with values of: + +- UHFcalc for the existing uniform heat flux boundary condition calculation +- UBHWTcalc for the new calculation method that utilizes uniform borehole wall +temperature boundary conditions. + +Sample IDF and IDD follow with changes marked in red: + +``` +!!! IDF example using GHE:Vertical:Array input !!! + +GroundHeatExchanger:System, + Vertical GHE 1x4 Std, !- Name + GLHE Inlet, !- Inlet Node Name + GLHE Outlet, !- Outlet Node Name + 0.004, !- Design Flow Rate {m3/s} + Site:GroundTemperature:Undisturbed:KusudaAchenbach, !- Undisturbed Ground Temperature Model Type + KATemps, !- Undisturbed Ground Temperature Model Name + 2.5, !- Ground Thermal Conductivity {W/m-K} + 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} + , !- GHE:Vertical:ResponseFactors Object Name + UHFcalc, !- g-function Calculation Model Name + GHE-Array; !- GHE:Vertical:Array Object Name +``` +UHFcalc, !- g-function Calculation Model Name + +``` +!!! IDF example using GHE:Vertical:Single input !!! + + GroundHeatExchanger:System, + Vertical GHE 1x4 Std, !- Name + GLHE Inlet, !- Inlet Node Name + GLHE Outlet, !- Outlet Node Name + 0.004, !- Design Flow Rate {m3/s} + Site:GroundTemperature:Undisturbed:KusudaAchenbach, !- Undisturbed Ground Temperature Model Type + KATemps, !- Undisturbed Ground Temperature Model Name + 2.5, !- Ground Thermal Conductivity {W/m-K} + 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} + , !- GHE:Vertical:ResponseFactors Object Name + UHFcalc, !- g-function Calculation Model Name + , !- GHE:Vertical:Array Object Name + BH1, !- GHE:Vertical:Single Object Name 1 + BH2; !- GHE:Vertical:Single Object Name 2 +``` +UHFcalc, !- g-function Calculation Model Name + +``` +!!! IDD Modifications !!! + +GroundHeatExchanger:System, + \memo Models vertical ground heat exchangers systems using the response factor approach + \memo developed by Eskilson. Response factors are calculated using a finite line source + \memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, + \memo or uniform borehole wall temperature if UBHWTcalc is specified. + \extensible:1 + \min-fields 9 + A1, \field Name + + ... + + A6, \field GHE:Vertical:ResponseFactors Object Name + \type object-list + \object-list GroundHeatExchangerVerticalResponseFactorNames + A7, \field g-function Calculation Model Name + \type choice + \key UHFcalc + \key UBHWTcalc + A8, \field GHE:Vertical:Array Object Name + \type object-list + \object-list GroundHeatExchangerVerticalArrayNames + ... +``` +\memo model assuming uniform heat flux at the borehole +wall if UHFcalc is specified, + +\memo or uniform borehole wall temperature if UBHWTcalc +is specified. + +A7, \field g-function Calculation Model Name + \type choice + \key UHFcalc + \key UBHWTcalc + +## Engineering Reference Documentation + +The engineering reference documentation will be prepared and will reference a +conference paper (Cook and Spitler 2021, currently under review) and a journal +paper (in preparation.) + +## References + +Cimmino, M. (2018a). "Fast calculation of the g-functions of geothermal borehole +fields using similarities in the evaluation of the finite line source solution." +Journal of Building Performance Simulation 11(6): 655-668. + +Cimmino, M. (2018b). pygfunction: an open-source toolbox for the evaluation of +thermal. eSim 2018, Montreál, IBPSA Canada. + +Cook, J. C. and J. D. Spitler (2021). Faster computation of g-functions used for +modeling of ground heat exchangers with reduced memory consumption. Submitted to +Building Simulation 2021. Bruges, Belgium, IBPSA. + +Cullin, J. R., J. D. Spitler, C. Montagud, F. Ruiz-Calvo, S. J. Rees, S. S. +Naicker, P. Konečný and L. E. Southard (2015). "Validation of vertical ground +heat exchanger design methodologies." Science and Technology for the Built +Environment 21(2): 137-149. + +Malayappan, V. and J. D. Spitler (2013). Limitations of Using Uniform Heat Flux +Assumptions in Sizing Vertical Borehole Heat Exchanger Fields. Clima 2013. +Prague (Czech Republic). + +OSU (2016). GLHEPro 5.0 for Windows - Users' Guide. Stillwater. + +Spitler, J. D. (2000). GLHEPRO -- A Design Tool For Commercial Building Ground +Loop Heat Exchangers. Fourth International Heat Pumps in Cold Climates +Conference, Aylmer, Québec. + +Spitler, J. D., J. C. Cook and X. Liu (2020). A Preliminary Investigation on the +Cost Reduction Potential of Optimizing Bore Fields for Commercial Ground Source +Heat Pump Systems. Proceedings, 45th Workshop on Geothermal Reservoir +Engineering. Stanford, California, Stanford University. diff --git a/design/FY2021/images_cpgfunction/F1_memory_comparison.png b/design/FY2021/images_cpgfunction/F1_memory_comparison.png new file mode 100644 index 0000000000000000000000000000000000000000..6894bfad752c1958bccf54d88805c3d1e0925369 GIT binary patch literal 27193 zcma(32UHYW7d48uqN1V_RB}dCBq~v)0hJ^{BuNgUfCOnoViOe+5CO>`3Mxr*5?V3{ z2uPBgX~~%;=kQkH`M*2fc<7t2_104SJ||Y<$|- z6TT7B!IJWWh8{ysoG%p4XPuQ`d35XoXB6jUUE!I_!gm*ju1J*RQnrSM&Uae0G#yh` z5*}nnU(@dIyEG|J)wA}S7kep|qUZft8*ibTy0Y1s3QwfCi9*j_9=5zdTzIfqJuOnz z==w7Hk-HevmB}ZY%fq+$<7j+@vj~&2ebv>kFGV&tH&4cS?9b!yl)=yX3o>+Zncu$2 z!40S?>nw+3g<_iK9WVnwmPUgl+}Aoa8P3Vd$`W@nqEMeZ6oeRq%<4nfdwY9p0~tnZ zPM}aJzsNLAdJA7nc0BX1q4=7ZWL|!-M`-zYliy>17vnlgcR7?}s-2(ODc}eSl{jRT znwECUZt8v2@FL@>U_l252TCfc9mW(mGBelHj>mlXAT^d4X<%EZ$IZ=M(gs6#yvfgp zxV5u$TzAsjW_RQ;f2d(pZU8~LtFB%h}Jk>*; z&a1Ditn3A*YvughX!vx?QG9=MRCv&?-7L;+z4;bFKuD;oql51v6$)2 z6%T$$q6zp#rm(Q3>XNd>U}a&UlWkN9_t#W>427z`DcIaJZFkuwC3&V`VZpk5n(E5^ zk7VR|y>tDlX2Hq3wH~{&j~_q&yW~&Bdml>=lfrUIH6?jj4A(#Jxck$agf-V;zRx^v z>x&QLb{DQ}-0?xvx`zD|b^Ri{X~*)NLCx~*&bc|Oq2TA^@S_*WLk68*ERWs8_3s5h zjQpa*BnXcXGnu%CJ#vuvy~|^J`AJIh>iYUN!}RQI%<#yC`z4DMvm4R!3JTH*-@mH|h;|4xndTMh@#OW=i|7-Zp)57RYesWUL*49>6AA7bn@S;UKLa`nXc}hk`K;3U=#-Pc8 z!WSjnaQZ}p>W?>H{K9-q+i2J!l;d$7oL8?>Gcp$FWbi~b6gwHn%KBhpqHRP!8E*a7 zbkc>d$d~%|{&@b_!{Z{D9AfVfiVH zk6Q7m>EBewI8vVp&R~EQX4*<$1aEq(k{*5p{_rI|fxKe0{R`cetNZ=nG=7=4Kz`v< z7&b2@D=VnxAqrJ-p=5`&ZSRFXw~ImWphwu5$6QPdi9`03+&b!phFR8{>gs=3U_9QL zu($0$*5{_0sQu9#-?3sQKwC;@|6t|_%5%|{5azvgql2z&+^wZ%zg1;M3E$JBCDr_5 z{CD`x7ZdM-f@Ei2`-Z_kr53%{`C-L;|1Ka(g9@p1YpM6&cx|4?C$Ubnc6BXjv@KBq z7C^5z36&V5vebR2wHb(=Ic4#!v5^{oJB(g^b2DWysl+ay{AA1T-$k&t%U2V5RaTBj z8n>UHYm!NjO%XPUkB>KNi4B;$+1}RnPA2~Jdos1+E5WYKB*(c!c$K_+cLpnx zsrSR?w$wyR9MclNVJUVEuC1!7Dw5Zbk%mV5hYd?5BdSqu2S5FZP9;}C-kofVm`t=M zCQ(jV>yC43d90w)`Cw^(N^Y%eW>$Mf)piJ#8~pq)UcBgxgyekF?I?{T37`kFeu3X-KmogJUVnj974jGhVi2QwzZrXa`q zT8l8Q5UUoJBIt?hUXYW$81D?H#E@E(vFP1y!olQq(qE7Ix6a~ z$x%;p(N5M#p4XU&@_TJ;`aa)`Fl`M#a{N}q4vJL_u7>Xto1)v|JF2+;2q?|pVPPE~ zANTh5&QrBq7${Ve8cF_r%=1#~-iw>t688DKZl$it9LuE4J-($qQrMxlFKP+wRn(-s zCleWs=RY|Pkq0x}a_H#lYVPWL{UJ$ENa#caRwh2Dq~v4wNmLSHg?uosX5neL_RC5G zE;cl~_>7o*X)oVG@OG1J(V&Y_VBJ{pbt&_kgx+=KgH>hL<;~Hc+F{y`%b|WBNl~T= z+46+Ui>-9El+<%IH`@I9!?4{n^z^PHepkM_FqLL)bvb|eexUU*QkIZKrquoccDQQO zJbMR+g8ckXV$Qj6eEU|{|7+UL!QQ?pPQs(=%~^Me%(8sDY1Q&|iM?rM>9JUs1?_To zH%G^fXnwQPBqZFF{=Rk&4zV|1B;jzHQU)G-YooP6{&l|s=%f=;QtDl<1=Yr=V`YHFI8jDHZcyqrOccU@^}PgZucxA&{dzzr2y4~llSP1}|2E;pnG z;Ih>=7F?*xl9!#td`OuV78V?e2XRB#Sq8^};cgZ+H!IZp?>FA5=5o5`Q+1}KxK|p? zv7zNlH|NiJvGDWr7i~PBb;IiNGzQJ+w$D4HIwwL- zdAXs~wQJWPsyb6j@`Ei1drL;7Jyt^Hi$6)q@+14_3wo_GGBV0Yd#rIA8ymV@5t*4h zgdi79W%Z4PQWCss%e38)eYZZO;0-Heb1zwXgHv9W5JJhhV_%g%RaRO%yRItIcAipq zDLxa=0YN4kq}BANpnv{hT0Ufz`1mS|d{O$qkrCtM_7qX60DYI~#K_&>QW|n8iuc>U z7cUekg!{o--P_v}6cmJOaj)qrbKBY8+0i)FGcYiKM~=AAQ2qd7Jp+SGMjL16!iMp_ ze9K5fhNJ~70IPF8kL~O_hHd70a_+i0j{XX0YHAv~qt=VV;1f2+mz>l;nF+_M|+J?{!Id{OA5oZ%1qNx+^t77HmxReBL?up}C1z$|E-fsZtr=4o?9%*@OL1_oBh_ZjCzDPAVm z&bJ7!G0xHAYX4EDt)|xg=g()x?4k>^HX;H_YWQbQpW2Afv)&LBE5YE)UcEXR->UWC zK{f_29)6BC;Fm_)!<1hUmL2h@SGQ-g(!>aJy?FpOpWM*U)J$6u77|K*rFHO_2<6^h z!_&~S{7P(_u6Oy)+wu!x*ymf5fz@`cH9k!4JsS*Jm2Oeq{xOVp}MB>sBW@ z#JhZHEt7{H!%gSq1Yfbuzg}gNTPqR=D_7Bfi5Z6dpzV)dq&0P#;lMVFs?X)=i zuvDoY#e(ecVj!=@@bncEq`G=%C#Z3E)J^&_LBo0bw6}krp=WPqq zA6nkN{nBD1SeYbolH z3<;Z}9B(Iu4a0I=S#0^#yJ@yKm)C;zLc_9MP9-8fw^_h(^=dYP#uT_|CqKRaBDOUa z$~s~rvOd=v&$A4*$#&EFA2la?N5?ce0Cutx+pVjyfmlAC$TXzta9DK`8{+L0xpPWR z!0S`fu+3((yoy3gRFf(G>3P(M%EG6(fghI?W4`goj_kw)$_1(NjvqVWlDE`SF>Et7 zKHh5u;U5thIcX!(Xu5*mZNhH~uAFOLfh*DCqST=6n_q8iZ0ty~ag2 zJ^}~q7-*OEMyMDL_V-NloN1M7cZ^a~Qzx1~6$9vb^X7zo^NY!5=a|Fd1J?fU-@mPd zboKQI6U-V~S~N9Z`}_NAW*O*fYT};Mhoq)nL*VNg-2_a0UC!9S*>=`c&NF7zChcE% zPs_7HX4e)L5!sx#G`5Q@^H^w9&CQ^!u(YyT&3x$!V^uM}P;x4Q@cK;Fd_OKIFc69s z7=y$5YN2$ORwS9_!Lw)2+}uigt*oCt)2MtTD_aZcE_vf!0R}%bFpz#tx_Hn)MdcB$ z9{@PUqyVZ?n4WJn5fJ)BkxEP0t?|gt-1>7q@u{iu8O)GG=Wx!nsOsb(e$6+h*NwA( zX0P4uoi=hS?hO&Te*HGDQ{mtOj81C5#iN_}eVA3{{moj5A=vYvfC>Ej$-r%=_QlZF zE%IPlkMj-4mM}3s?i?52KD`Seawp7WX#HhO!q)PF(X2yp=NY|xhG@(c%MTHdw5Lbv zKYjWHr&(Wkon-!@{)wvNBxGc+Nf#mVRskrTU1GM-6DyM~Y26KkWLKs|D2)Ms)<0P; zCdrrky}Pwi+9bX+mltDQ#IdN}?qsFnrpAujzxR;?Ct~=6uwb-QLT@ zU)N$ubohC4uf$cEZjR@Uhb1=XtmNA^@wR+}r6zxNv~hX7M6(_T$%@HUBWt`F?HZ;V;7Sn0$lAoX7 z-m`rM71CbevOIz=(&6C@CJ0C`XA)ejLy7}uuA(2zu&@XS3TmjUf9vNhS-j(0q;%F~ z#ob&e#?Wcl`z)Q%A0O1BcjaTO?zuSMMg7U4X{I@H*zH&fqZO03x3{yK03pHl7|Ekw zWZLwB%s#fswExHR`2KnE=OfjT<-6lZ6-gIBj7piEC2eRuNJ$&tE)6bsEH?^*MM_%V z^i*Ib`fjS(kC=_ufKtpJqvWJ50y4;4K66I*hLm$GZ!FNkFD9O$)IEY?e)DSn?X>0I z(8=cdAAyc*4`IEaQ7AI!N{5=&x2g%rV?!MHod;zbEdKzrZ{VLwgH={^jgvn$W?1@j z2UMD^C8H8);uF@o-La|EMNL`#YwE$a_THf?iA-i@leE0O z17oEpfQ{raqqG1W^oHDm=*?_@EgFXJ_X}kA~AbJ^&qZAsCX7 z0Xd4b`&DVI=4ymV@W`N;mmFyrR_+`?(Li5-Ht_`o4h0T%odOv-IoGp0_-O+G8r&$h9sdpKP;EGRzg=Q4)K?7nl01z`Z=)O0 zbz7bt7e#yETa)>==s{BX(kDU!0;_3;d)N5+SNyKT$|)-c?GZVxGzlxmi4S&mva=$9 zwg`|@y7pr}Y5^rcQWWGdOd>HiLdE0~#00wBk_xJZw| z<|pC4`@`32;D?G7EY}Dg{nY!jkQy$2>r0EfNRB6dvAenCIOLq1mX?;AtBO|NQ~v$` z265*+sycx`fq?E2^=ub%$y<+;-aD!o0WiJTiVO)w3wfoTD+y<;Zct7Yyos_zYHtJ&(H5(Rshzl zA}lg;=hK6HO$oY@%BZG^i3!s^<%UC!fL0EdPtt$A* z%$1zuZ`)$7749uKmb;7w2n#61bB?@CG6!h=t2fW={Eqo9LH5d=@Xmpwv%r_Qb~@w@ zI?N)a(sod;kv#*yjahF+MaAMGm4aiz_wPUz)GiM};OXk>NSYGB?2CsV6P1tshDo+`r$f z&XhN3&CAQHNt#qrGB9pC-vgJsvxcRPE*Kh)OGL$*6=9?lhtVv(4;RcARt12j+};(Q7Bq* z#_0C#+eU+G1_leIUCt2nuyw;(pulkCQ`H#?5xaYbU(kp6aX~;y`nt0rKK~ z<;=}f2=9xt+~zH3-q4=>b^t?jrmT`Nn2~}hWyoqdU^PtZgp0+S>I5p*Jy-a~>Rwwh zb*yj}c4r=USk^8*GZmoi>>R2SpM#2_yeKa}e+>vlH8)l4H;RYh1!aPIkXV3xgbgH& zk)fp20k$W?y_>0& zXl}q7S2S!}CxgfY>f6WqezcAZ+Gh)K+?(IC zB`7~hKFnBAWfgMDDwPIj7(G9X^mTPn79YhG&jywFvRiGqcA~bxh_?MV-IE? zeRw5}mA86@7W^>(x6Im@T@w>no6R!K*T z$0Kx)#ziKX&*T7yFROY&d$!Q2yFn) z8w%vk>%C^Bm9g#9>d{>Vuk5019| zHZpr5LJ!DiXaxcMtj0nyo!c#_uU9vk7b_Y6`%ydiQRbdJEv{6(wA55&ABaV}(9+TZ zMnnidNK{(|b|UmJp8$MK^B~>dq(7I}yDRLmBK$Hrw|_o5Iyy}=i-(7Yk&&^Bzr6aX zK&mtWmPNzw-yK9fK(YV*Swrrs8Y7IM3@#}t$<4hH)AUeNQ+T>}d1XbT0iZE#58a)e z+}gPb02QSyEG@O~9QXCkMk11o;y}#w_LQ8mUz7R(AXQ%7g1MYp#utTp4#oNm0x?bQ z(uyE;Jb#{}qhDi@?|!g@_gJhX-YeMcgYum%NK92%*JbhV?L%N&232i-pWNT?CnNyS zSazj5&!i-mr13b~*@+$Ou9VFdO0C}m9Q&;D)o~>yrJ>f>{WwhAs+Gh}ZwsusL2E}5 z`qVo9m?i)?ukb@pC4aa6+w`+HD#U^VSeFS$xS93>#)fY0-BPU_dhnivK?*A=Vm|aq zfSJ|J%?+W9U4u5)Tdu^pF4s!zFHOW?3`(4qz7vd%jQ-Y88FsV-ow!neCGK5Q=8yZy zD}XL%oSm=eiX5_Vm@Dvii}((g)!@Wnbr*0nG(R9u4!tNEv{q`jwY9xGNMDeb_hMea+=1A?|Lnog=v zQyR^YURMgY81gvSgSR5X!<%A+gNC96v&4k99CAXj+o3dr-G?YJ)Wfo!B}cq2jr>Dj zLd^9m_YPDdA|fU8n<;oiJ_`LZ}nClJ2npDnkR$7yJ2 zPOWN==Cz0;oXc?%%GI?sL{s1{(A3qvW$sDLoa0@VIm5}CMuy)QxKzm)zY3)O=6Bn1 zc|32HLFx8%a$IwaFz?IL#rK?L^Elp_6NL;Uj7P@FJ>#O((x#X`5{8Xp39TP1{m2mU z{};8sCuY7kb!wgUM^2QX$-M+(NFHxodCg|6Faoqy=Yojdi`^l134XhJlv9BfEh*du zaIrCc_w?mg!Lhhpb>CIg;AZksRFw9Xpm{b`uUw0+cHchEw{98Jx`YBgmwEeDK##vu zLUOvEawL)3&*K9e4-^L6;JVXq1;`^bGK4pL$DQFzuoxAh)ft;H@+c*G= z523YVFlEt0l1EACbEpyzKVx^xvY9@A>t(|@Edv8k<=5O^O#==5!)a;c@88ec%L<0u z6mGAoClFA5Wa)JdRxQt{@~Bcfk6{^7U!L6WDhF5!40*?by)V<~=qT$5ciq=ZoSgoV zjKRrnmt@Pc(#A_$-Y!K%59u#_(o`~x8Qu3f8O$h|5?S|71`!k_fc^*$4n7%y{Q$g< zf*~X)9yGmSIcL19edkjF9-!e>>GX-ZZ*IGlx^##hDDAIHn7R8zUJG>HEx*>*P0*G? z#+p6z`}glazC$2mhJShkQ@<6?scGtdKb(BXwt*WrU`v9g*KGcHlMR8FmenmEV`IfZ zF}V4H`u59j+Ri~ks)WE=))zrEE8^L%8|yKY?He$N%N?%dH#?y!;e|T}M&EU-!7%CZ z=Ko8#Y~e(_hfHqu_@}#5k!EYP~E>=-*~S5 zjU1_*gI>jP&)yfQ(meVn@m5>7B>sC3kGo_&JKN_eM~c)NQaR0*o0j_A0hfE7twc?< zr^sOgl|23TC=aSJgw8%|G&Dk}X_LDesU8p|iA?(Cj^FWkvsco5l`_a2g<(;iQ_F+O zTv3fy^>~53;wXV0wl<=xYbcaT@|7Ola?Lg~p-fRe<~{j?Js zN<}!eJxC*ieT!GuV?{HEzn{?u(&(MRT@1?C`%mRiBOAI#`0%@@QVvz9RoKpT2Jb5< z#NS0#-vJQJ8XOy9{r+I`>{h2?JudFbzu&%|@T&BS!NDkmKZsGwxPyCUL@94bexS&R zk0y`E@U;<1rVyi|lwPu}v(il;k<)MuL`jN_GETd3yXMhNHvC>|FQ>vIgMoruVV0Q@ zZkTq^=O*m=r8^sc*ciEG44_Z)wapPK)X&xdtgbtUrHteE>G#x1xqNX*q%)$b-^R{2 zT%yp_r|jI-C!Gso>`c7sUh=a&>;!tg>sR zPJ_IMdS7wp*#Ehf9Pg9suy|prC4a^o4mMfsCjSLVTbh|Yd93(~3qoR8F~f`IRf7YFcp)giACk zDmc4xNRbuIJIcK(aQMPmmbZn`ve?7h%HxpArV56!fPpmz|#OIZEdu77LOBNc@DM(ilxz#0$_A z`1`lKPt|o%(bP9x-iwP>r2cXMstgU|noM{8dXb)^7hz@v0^0oruNY%X>Bi8kZ6W@;o#T+;Wr+v5}!X;JUzoN-8^lVx=8r*s;SAg#T6vE00gLlfBlkKGqJQx^X-{m7i0S8@?bYe#hEtIma{?i zv%Gg>ZXi&Sw`_CdaS&Q)8t3YRlIbL7as@obv>5pA?sXSZBYCLbxkXvdEgdTT^Pu$0 zqr)cwy7pTM;Ya^7Fr=4#{FUFA<-V2%Xw#ioLlN0@o=RjbEj^EH2oDbz1h>G@>hp1c zM>)=LPQ-!%y#0WcP%=oNbW2`AH^hIF#W9)qsn4W{@!Gh&BcbB)twZQbw+WyJDo)X- zb`*~W-uk*%!edY}Wm_1*tqVLMcoWV^Fg;#mtOzSP*x#9KiQ~O`^>czRr)$`f%A3vI zs|qD(8y6Rsy+A+^OMx)RC&T1Y479bQeW7{;Tg9!LqfiO3s%mT3H=L@jOQ9^ZV{y&% z-hxasbQG4CzT#yyb8~Y`OH24uO)a#>xK-fUY!_lJuo5D^-;E;65X~CSS`*qI@-Z&) z7AJadAtNIL<>%G4vOm4OTd-yOw}^-ls{J9B@G5pK=#7tuB{c-nEr&S~*UjvPaf)D2 z<$UMdlgw#fgn8O{*t=<98XBz%pNlh`U>@IR6}M0Q^5s!)t_cWBdEzg*c;rj@&6=Gw zYt%{%5!6?fXmI;><>F#CoVieMXM>!UY#BqGRSwcOpV0S30t2jpf>pXs{V4QxsA=Ad52=XOoT$s z?Keu-$}u)JzJqKJqkE?nt@Z6UReAY!mxWt3P^wB83;`@XMT2c^A#rgM#gztnVW9oC zVs!hnVFZmXo@AC;J zC+Ma*@)+k-W{yNf{HcuHm&Fj0ZV6E-Ux5<3_w4%BwahAG&Wes>fju!636UO=HUe_(MjdwE@esAm)6TBf8to$i1;r4qXJlO z2nY!9@sZ6JlaNnyT-;&LDSz&B^qa za8Jy-4i-j;6^&9GhRb|P?DB{WZBpgS3LDAhOX!Sp2x2)N6DBc=xYq&F6#HnUpg+&ihp`J`KHJ!wxqoU3s(w zF8!)fjeee~(!+;cP(J)D&?|Fwp1U%vI-~DnzxKXyAw6s%Zq`CO*V2LUkrA@-TsrPk zMIMtVa~baH+X)v=&CGZG&1UIC_8mpo)1@z;od3Zm&zGlRv=xA$l#rjCStZMTmRWnN z$r1sTyX}7-FB|h+j8x?v%E@F5(aX+gbZxSJIezP6B%m`_Ha0di8ch+pmCVA5W*4$N zj3zW_&%NB)CCEU%zW}SQ1Q%bL{W10r&v7ppPUadA_C~%`^mVZZ?VJ8yxr3@^!XzY1 z3HFdcL~>9g=is<(mrZr0Tk!>9%=>7FC-OFLk0k~Z&knGO6-LU8QYSLn1(+AWF4YUyv<{YZXvFa0B;FXdJIgO$`YG+R66 z2Z#T~wHX++lX11FeMMiS3d7z;JRv|@U%GjSx+EB_uyA78PP(beuBLhyeRydv9)HqIb79;+B%CNT4wgcvT7; z*(YwokI{1mk<>;x7s2x~05DwnqX@ZENpiB;V3B>We*7h2;SCr3&Jb9fdLq&T?f^QG zTn4qOp{~w%M*qyYb0>Bl>gx8xUcl<&2--^D?D}m9p1(&?e3th=iRJ^t7%v+nL*Ri5 zhu&DxKs|Zn6j?I4n*zqD~Dy4(XUtBUKC5@F5r<(zfOH7y892EuWTiv(HNQD;qtnZ z4I0nbqZ`q*nO}s-f-awhYT&*6@m|P*0?qVTlgF<>=Q-jVUV-q^DUYnI<|ZW$wpu*) zS8{T4d^O*H(GYl#H;>W!kJqBr0l@$o4xDq?jpMaP7bBV#3*{*O&P3ydpZ z<{1aTD90d1{`YnQ-B^E%f9oZZ6~EQ@th*s4Cr`L+3v;slJGpkDLD^31Zjc9C+udu6 zkUimDkm1^TRU!6BF44Q+L!D}HB60J)aNYb35Oo@=sy9|bbu+p^6HrmEtTi=(iA6F0 zM3PWxX=!#g9~;}_3CA4ZPqu05-$E(p`Ge9}%b+@W1ZYlhuHMfEK5tH4?=UJq1L9_N zZ_X}csD#p=809O9>mO`yZ}0A=s(kS#aP`8gf90$`EEn)KtviXqEkZgD7R_j`4mA2# z8>;}3@7Pp`xE4ODDJLY8%^jAzPnY_&(vy>~uvQL*b#Xcr>XndHBxq^29E2vOr1-g` zPA4QM@E8)7(-KhDu$4wzRaOf`iE`m#;_pD%Ko_F|5X2SK43VX?$7l$0N+qyUm1l8RI5D^g?2W~-E}Puo6NP4p{F`60^>#!fP~ zflt1k2eFMREPS$~X1%JVBveK20+_hF7d&Un*HC%B$SE#7ytflnC0tCnS#+BG!pM!; zGqQ6ThUEsC_+l$pSI}xe9tYy31vd_$b zqf$$wpIk-`QhiF+AKJ?qkpM238}o+CCZ~+qD$PEq(p{o%7MW z559U_{}*JQzP?YH>g%Ex4+6};+`oYoxjwTBnb8;yhnWtJPu!sQyNNQ%88=oOU!|D| z_wp{%q0KXy1mB{_mLpN}2*vd9Hejvc2;=QYkuCr7O_FUUxf<1$L*NTmKPpPf4~m!L zTjk7-6ulm#uEGSsASgOqw11=fmL@<(DDF73JC%)~Qh%OVP?IT$X${TIU_6Gn1AGrO zJK#{Em<({SI94h$dvi5*7YjbC&-@%5`;X)Zh2YVBd&E~OWztUcnUJojY0{uI#>z^a zxXJVd_|no7Y_++Jk!wZd! zOukkEl_l88K%WCd1ql!oT17=gJs?}l%ge!7my{&a=?Z1`*|TTum{cKHhP_H5-G@91 zyK=ycpTFF87gjq<4-wf*)fQ`94AKR#4PYHgZm@#+*BJBdnp4MB8f9y7NlQzE<^iW9 zK1a_N(6|7GXB?awzhSlCbpnSbd{sk}CreQiW4c}%_(E~&Eec1%-}*@AcL(*Yp+ zLA8u?o}eir<+x&(oy^}xlzEC_n&X%Y%2a0thEQp(R=rtvgXJQSGOMy!Ih%7nqfGn5 zOr~XC(Xr$Hr`A!8-#ffC4Gj(Z1XoX7rs4vWs6YdWCaZM4Z)$1^S*Q?j=5ezB^2@(r zMl%L&DEv=^2ZIk#^a+GOFbQ`WR>ctAR-!%h9AJh8MMWXrZq2WJ6mwQJYlw}FjqV5o zVx@X?2^f(~gHlmmUJUqTVYz_hGi_&TW~Qm30a8oxU~FWh8T25KRE|9spncZ8I3fAq zCB(47Zvbi2)6+}ty=`qb=2@+P>V_b}>hS0myo)8~LmMjQpqpU3ZW>z`&$ zEt6a8_{{-TCb;1jrlub3JQ1M}gxCO!B6tCr;EjrxM6AmlYRava750bM2WTZM8YX}& zUR60G4*Z4Uq4nzO>Z_wn9yzxsflet=VsaiMCx2e3w@S-A!VgRxuZf~4+V~CII*NHF zG4=yyxR|(4xNPYx`SqBl%c1oye?VRGzcBaqgs<=aWzjHr8zs=Dc5)q6ZJ&}n043~D zf9g-&VuuXFSpbIM@CgVcl9GCd%N~`UIhiwZs*=(37(#nizCKK@sHJITy216ewXKYg zlMcm1L4Qnug6ffeXLbYHIE+qD?URyk5daUZE?Zk$FafL7o|~1& z^-w48PL@4+2a%qf7+Yf;9d+8j=1%PqIc9!(>0}SxIv`Qn`%xorU-C%)zWL8bJ%L-J zq3D7vHE6won6FJuOq2*`pcuS`U47bPDii;FZ+p#71ah9y`k9g|XFW3Vp8a}e|M&$a zsfA+aI1}~M!UAQm$Nr?py8d~0c|d*uUZ)gMe!e+ax$DB*ZT&a+ciz6t4?b@ILV3~k z!>Uoz2PG$Hn1KD)^}reu8v2;e^?CM}FO^wqz^DjY^+;y1yfons?=O&R;T+b@#NTRd zQU6^n#K*@cATXx8u5_!1$g`SfO-Dy(p_m{Jp}9V%P1pc~?@^D><>hhaG;&t6sa26! z#CLOVJ7zlL^#s7W!#c2yAEjDLMP=hS-^Y$Hc0+Q?Ftg*`VG$8;>lOx!rv)zwU7YhS z05HcNBABx!jSZ#rfc~5_tRr@!;IZwr$nUjs#~Z-pZrsSM9o5c#at~>8@z!UqglbDY z!RMH;xrS^uF>_~QuqDau@UXDgDW}iA$+zr2-5<|#3O?hzN`X%ZHt?k&LxmJ}+=4a- zU+BEKY&HAy=bgh?x%*RB6TS7EwKRe3YwY7|Jf2~kYp+21h+5U*Zm49VIfj5S3PWEx zUMp~2-qV@^|F@wOLPBL{05Fq!0s@>JeaLO3DviAG^Pw#)0oAk)U6B6~Vfm_^T2So@ z;we4ppTT?&%!N|e=SHrsyB#TN?fXlRJDY9c(vmwG>g!=_&iSydcJw93Vt^7s8%F~k&m*g}(%GCac_(o|$YBhF%OE`?pH5DxthPldUQv9XP| z$};WEV?5{z5?h~^0O>YmgGy1I@vDQd_xRVqRj8K<+cEHqcn9d^DJT%K2{2(Eh6ube zbdj93RC`jmG4o#)2kks+qW7$=c{nc-Ra{DpJ!C(3?vkQcyL2(&ji{b?z!emLJu^lz zNzk$jY(1?{1=yWpr_IvAX(LmUW1Nuw2qT%;?s;_8`kI zR8=s<@&3>tlxP4RO@xw#PAxD({rQ7*+W}pzJH#O2UIN%CJ)Mh}_utM7PHd4}(wfmF zk*rgLs_oyQsqoF4H_$7?pDiYcD8hb`LhJ(@A2Ur!R@gW0V>zw82l2*O%ef;cDg6pq zc);kTXJq8}&%bO)KbD%3g6QY0A}qMf%mpB$&-qljT#~xn-P^0Jr#H9Z1l+tcw;DKV zbNLM4VA<5jjU%Rd@vjhgfqv0(@J{A2SZF|r`w$b;(9l5SD3Pa=0faC3<~1|*ptCS= z?kDH}FU_tQ)NYM~Dr89~>07o5eXxORd!Y$8GNRu))XtoMJ`Ctwp$%lX1*X#0L)(7V zuTt>FrvvE;Z>oJL{i0@Kl2|oN&A>2EAdn8K>I?Gl@)Ble^LniivKssqQiu@&il1h6 z5T~L$UirR;f(+UGcRx3CTS|`sl#P!&VkDCo%LE7}N6VORA>)vpXSPzEvJ=(Vbc7bg zPA%@8^jx&iq>?_%+ncPVQLtSj-2e#fXRhwD%}dS&9+1ovo+?qEqvR2W!f75$^&60go5Ed$>stfMhED z#Kvav4W0GpgHh-pd>$B->vTRZD7RW+4V-QTM8Y15;o&m>4%&IO- z&TXsM$PdS=bUpegcYD1YJ$Bg$k>J-JJNOv)eudm=%Xpnp?hk$;SvhVU4Mw-;AooHN zdy3B&mY32Sj_0yoQlX49UQ~&YON{-!QdC^L_iOBv$n!f=;N_-Lt~ZLB8Slfm&FWe( zm>XI+=7WZ#W)}a4xSnArEdMwos z^e^a}lX?QSws!Hn50+x)DUqR}=jSKwVJCqU4VH!KzhxtR53+Wbtw90q&0X0DA?JIs zl6B7Yx-sUF`#J+Yuw(LZYB^6Ow90KtV}C>MqQ}Odqq2GRP}mQ>Oi1zoNudr4(qw@y zOUlVa#w>*rzh2qikFRGx3>h5wMSwr0hAirUcq|_$yUy{9T?kP z7#lqa)%}uyl-hMwXi#KZh-*7Z$y=xF()ytklHv;+e`@V#b8sJnr^ht!rcKcailAnb z#nWkb+=Q#-WM{x=pQITI)Hu{=DBttXuRqMn%Uju+jH8Ns?lj~_1$qz^4nqdg4q7`=#%5om-U0D(Ve-a%vW;)M&<6SraGy51wI_>$x@ zY-CX9zv-=w9eu0yNf3M({!hIG*QZc|GV_|9Wf z8E*svE3bZ$6o#mRt7zeiINgRjTX66L84N|wkIKC7N=BpLtLbGW`R*bnSE9C*C) z+c(tFcGYl>*GL?Pz3Ix(D-5wd=J}e;LLZ&)`oh6MXz;ZbiQZ#?FbMDMU9*T{W|xN+ zmUh^99zKkSj@H)D0Dm0x2*DDrAH~y zM1b_5Gl<&N4cmN+hZ@Lfcd7PqL67{dQBL^g2m{a*R4l*;A711sg;L1OFl1Yl&K9S} zjp#l@06?dB&U=?+tJ#o7Zz*grh3H`+!`v&6HU#NyW8n_@8MWu+x7PUZJ{gWp4dYSs08X69k_az6g^ZDHZO1Ip$1@cUepwu|m00K4e zmb*eyiEJ20JN>PA35{c~6(l6ZsNtZcRmEwa0B_6BI7;1CfehHvq0bn|5vb3`ZQ%(6 zfNzoJ!$MGRQ&v1hTA}HrwCPtq%9AfKHpv{kjnJEu{_WcXqGgA{c@) z(qm@X1j-I5m(5KPG!g3L6EbpKe}+TdRk(C8WOip6w&+}w8~}6dg>@VX_4?XciH@$W zW`<5aI8l|2SLSKFZL`3RZW${4U1(B4{&E{i0q@Y`QLfyGk8Br7n(UX32dQ zx&6+bJNHnpDu(T7gonb7Jo|jEUg-p?he%9c<3TYirX*9Jc>@|b07GDDkd7XsEXZF! zemDU+0Rz|}z^z^U7?8$!AR>G{?Yz?QX=x^pE|y-@f&6)R4%d`P69HO4I_`5l*THzH zrWSv00bh=vcW8h9htorC(*}cD^}#?eTB$OwY z@jweezRsunW^h1*F|KZwK)^hG`iJHYs{0guPUO$vSVz%kppw+pog|z?N!lrtB0ZJk zwkg=mq@-U~LWeEH&>bcX05X*|hy@GoXCq#3HZ*?vJ2Y{X+yaM}F&WZrSx^nF1)4Pj zP-!Dl8rbd0p~axL)CE`hp5hiw_H`Y*tw5+Qiwp z=1l;daB6>w+T;se;?g`})D)EiQ$K7@+`0vS zR8XELKis84<&d-fTZ_K(i23w#(knUAs8mFg0v=v+OaYB*r4;(Ngo114PAe*Gw2aST zs)YIXqtiVun@sj&p-=Hjn_6=$x?$`LJS2kAw55&)+Su2G3=FPub|*Xez=JwuncP-! z<$D>i5kL!-)F2L=C^RI4#yQ?~VUh+_DIY&MYvkbBg9kiZ|26Yy#&tuOgxg!V=s5EH zj|C^hwX`a8bt~o)|1H#tXO6xQVkplv{c+Bk|{n3=ec?m;MoU2-hzLLg$B6HM6jrXwgSB6d%L4v zr5*oJ8S>R&!|ceA;8-Bp;VSO7EkQrF;J+b-_92NzM$z3YFDBgtc6 zCp5t$0@8r=aTi2t;Xt95Jk?46NdMzl;deZbAOO8#KZezEly9Kj6s8$iK*!X7?)trtAxjTyjT3;=|8(xmV_JM;k{JLZ*6JW zE&$EaaxKT;T?!KLIKt);=QQFMw(#l}XX4IekX8yY9XCrM;&%y&1VAIaxN1^VdM!tpd zaSt{ap{Aol%=^Bs`&*Oj?Cd2SFr_nz34gxm|CGTFjjL!~1p9z|=>Fp0$47U2;Nl+aWQ)(-klvzulD>1MEmB?L*th84YAOlsSgrk z->#NW6f^^W?c8uXY2+wM^*ohLR-SRqTR%&Ro#@xpIg@nL>>{72%xK_tkd=H&$hd{K zO+gFfSJ;ck@dtZ&XW7G0l!AB)x!R_ntlaIaEd-4V1^w2ekcH%Kzw1=60KaYsdz%`F zj?}NgXmK(Dd#{PNN8u~jNnr?wx5-`VFHVDnPQEm35)sDXsU+wpVjQ2xh;xmi$1u=5 zco@-j+stk#l;mj;<>7%iiHu|~EP}_>lJB>I$pdDhw**)P5QIT}0>yvz&?za$2w88) ztp#|cX|lsB6$5sq1P3Vu*tAokH;2e;+p{ z0e=LLxX2H_{I3#4zWNP}56H?xl8W4aSOz%_Bl_PGgv+I9tPj0nUqBC(6b;Rn+QVP} zS=Z8HSYQ<$8v%g{jOHpe3iXI)>S(p!e|ZX2W)!&;^5_;s5I;2PBEK!I37v~GooV1O z6=Njjg?NX2PXYJ@Ffh1f;bhQyuMdQ{nc0;ryIoiT@#m4x#3XqUlfuI%L1My;(;-E*as)7E>|H+upJhHpH3+V*e52yV6)w9yl8gR`d!|*7S)MjYA zflGgwrXe-w@c(J+O2DDsyT3NK+@eCtRw-r6maVL}hGgGm8%0F+vJQrFWeMeyVXRrl z*q6wj3fcGFP{}rEu4U}Y|BUYYcHjHIo}L~v=C_^i`JVN2KI3U(68|viMl~2%(hJ-$ z6bUN_igI&uT0K3Tog-l}eIvQZz!=3w+W=JyA0MCTD4-t74}?j{RWf>hXFaX4Y!jYgEoa&nZ8EWY?G&E4KoVHZU(DwHU zpk>nvJx2XuWqw$<1X#RMfttd>*m+~w`h-EHyO6-j93@YnD414qPxoyL7rm_BJgyNa z>;)d+~V29Q%O!^#-`g;PO) zG-Ct;asg!koh2voWq88d7s3_tGwWU+C~bJ__~of_jqRe2^ag+~WHxYPwJ^qfg2JX% zoY)U#!oQ~JrMaiVDi@o0?+v?4C0H~^WybV4$*azW)p}BCFYauTfPSYsJKJ`&(qma+ zOY!@Q!r$v}83g-8a*h^Yo@ z<&luPerMR&(kNgS=;iOvo!NCD?c5xGOOg$PxQ>~+u$Ik~lV{<2g^MzafIc-ZwPZ)C z=2fnJMQIH8R}(v)bxP=oujtc^x$`tDo#D@#;Pc}3-ynOa!$s`mFk zl-MpVDLG6a^ItxNl~b%yVICVrdC()= z3>FoJ;^pP7$GHbDjPGo3wrV;+6?n!OAZ|K5qSvlvj)ovOCauLT#HawlSA^{;ATPt| zi-O7oA(E%_Xns)vn2xyp$S2@J)4@YWzoMcdSh(>+kY10{zV>Fhq!{ zTA~#LZ|>}@*7Un)W(4;g}CO-{I( zS*yoj_#qvRgsDY*KXe<``)hXOW5M3`+>8jUGn+g+L-7K*fJ^tjcWF=m=l7Za{{BpV zAT6TFTHRptoHt4%HRvW5#sB<6PdS_Co0%?pD0SG$GHG#v^!+6rSz?F-MqU|f4mRCs zmpMSGu6eEf3G9>A-y=cE;2(?seu{8{kA}PoL5a@p9iL;(+zQ6>gVX-f+E?Xfd6p)y zbiTNLT&_t)-%98+gXv`bx!yu$cnT4ku6N$--piei64}TXAt565H<*wGw=l5a=ZK7t zDhNQycI*^6^>eUvIfMnifsXr2Z%5}1q)-k%PR52h)v;5~{+Z?eQ_F(7E4yUb#j(W3 z5*cP#Uan=LXn26}Hz~?ueyE==SG%yctn?ccY#qQfH`IML=v6_{H0e|)Myou1U~CKV z4eAc2=ylFxpm~~ft>%!z$F}18f$32M=phkR)O|1M6tZH#zk>d?tu4STnKPt8ay>S8 ziM3WVIsZ|UqyoX*QEz5)t(7QPN^PBnQOGFj(;?KmnWRrw$LKo4{o}Zs=Db5`8IE=E z%m`_b5-k%H4I2a)$H^VICg=ZtD1~S2hnm4ng+BPXqE=F_<;x7$L0`@hU)OGJj_sCQ zOUs=V0n$m@hT~0d)q#u%HaGlasH(#$(DQ!XPZsVqs^ zX-N8Vr*VbnvKz3ELB4tbzaJXHq6mF`3VtB}XZ36p?m`486Z{$elLHG$k<Ftqn4%Il zFF-xTtv3ViC#;U`ZO#OAuQuxh(7%DKJ2<9cwi8wj^v@SZiS2%xxu0+Tt9*aH`L_dE z+vMq^3D73KEW3VW9!?jCY8`+_@${7Z@k;g6_=1YMdR9L$yitT~Wo4WNm;HINe_htU zUE!}~{onij$I}1vZ2$I5|9sG&isW1L7VGfJ*(5PO-o((b>Ao-o@)Yb>%bxwX2tX)k zpM{(af*qe^A^PRW2sE{wuUD0nD7b*QpL@zxq1L_7E-t!N00!h2jmf2O$QT+J%*@Ss zeJ`5=I?6*!U{nBm)&&aLBSnPrX*@8HvcF&?ICJf}pw2a0G5eSd`Su!p0_zTAqduXhP$C$oiF3 zS75FzIB<+4>q3kio<{@vnZij6@fG~Q%yEar`B;@p<^^{$>40k@JsB=W$4`WkES?Ey zaC$_5nTWiV8F!h3A2SQRSYiulZZ13g3lO=je>bFvRqjmK+XMP8J=R#uviPr*5LQ@R z`~mG_n>0<}Hm17zOK4VqfByrIjMCE5r|m*x^RbQh|7c8&teOYg06zLAek|9~>Q!5t z+6lml#Uvz@>{>$wyk7nZ03L`g>oT$?{mSvCJJ1q)ShWG;pJlL1!SXQwzsQ{TM$K8CdhKkSJs^PupjRM5T)-SvDeI z+WiF}=?ptfbxWXxMy2*S5SOvB1TjC~s;+SrN*7UyB-QIRP7F=Y3^5_y&$H4RjY>V9 z1m$gqYNbupMvTg225`Cnj4jmY&gbJq_N4cA2guNxBRAz;07e5DOM&4x#M>UUj4K#f7^+;gDLd=|9`{K*%f-k zs3`Q4->*IoBbgT+T;Y{hZCza7A`(-|s8?N|KE9b7DVv%ja7)hY~>hB-EG%4M!(dy|rPUA14 zTC`-RP2kbU#`o0isOdD&h=saK*>eX6Lg)()VQNk|r>tf`AT;8IB%%ssFNeRlw;x$p zw7D3Po$x2gUNVWh=y1iXFoE{$uxR|^6db_&2i-7h3ItFiZa^@kI=R-L}w(dzPhY%)L z5LZUuTky1X9$HdEI~yFT~SsaeN)P$U!3G~_R?=H}`e4E?^NsydAIn2}pcy+lB8s5Zq!;jhWg;{dfohPW4u|i0D^fdC@%} zh^iVrW~ag>_evdf)vjk7vxX;rgYT`LoC;6Q$Em|_TCE)e{P3vf@_msU9b-bk}&VRR7ACIXcLDlp*YpVFca!n9d( z8SMC>51NQnaDLO?R4&$zD;VK4r#aZM5FT|DpSA@9`>$b;$taB~d~ZMl+=h^of}Phl_15L7o~EQ5n*5@9r+a9EyaBP`A1)p~h$Txj;2A{39+#G{(vLprLC9 z_yo8JENze086TPJ*B@Rhvt5I-5saHZzrzRMWJdrEG-nu`nuKsnPdjQRpm}Nu-0P2+ zQXRNN=+J-qVUVcCN=4Y2kaTb?!oV4_+RGpeZ1b@ARiG50Tp$4i%%C`?!pnTfAU+v-q8TMZZ}u@nO+r7JNdOsAX9n>>l)NEk zc%GKp55h$dDQKj*DK%8^A!Aj*B=pZRhC52wZt<|vW?d-W5DHg}_7PdzMlYSkHzF>c zK8mgJs6;wkj~7GR2(}lrB-&t41tD6r)4Qz`+aIhK%bk0vmeK5E_=P7rLx2A6RJi<5!mjs4Xr5~P=^{Pw6?&fg*g63Qsq@8k&xCg9%gf8c=Bc;u;5GpJWM(^rbus!yLn-jk9 zN-9KL7s!I6CEtb6iIn>X^Q&+_^p^j~N(kW;7f3anaPce(R)`C2v%lJEAETM6Vp$~A zoEO$=m!01+d8lpTkvt|WV7>rv)uUIX*i@fee->L{k=f9^h51k&k+P;EN#MG`**wV}LA=!sDNr z$^H0Q?#7%_PA0Y!DvnKgYvrK^Xcv58{OMzlAk;hNO>vY61cCDa#YSB0EfANHL6Tqd z>R_~h=&yC=inI#U<%#}E>_o&Aob`%NZ-0MAy+}!=!xV6&#YBC(m)Zi`viRw7bK@sW zFB%&5CbEFzT(!4EH_;?hRQKuz6!uKfb^UfNi&^GHVT`=HpHKA?jPAH4i3$zS`2GR* zu=gD=Y%wTv`_nV9+v_H3^m0brUgEBSZyrEYX*g_9s;B;tG4jxwX3H#nppz~Bmi7zYy+#`h3Gto)t^ffP(&gU0T8TpZVY89j8YWTC zd-<3g=7Aa<0eCB9-8FntQkx;Ovxz?M8eroq7I*Vaa%*dAfl}6#aDD;CD(sypcOaS) zaalbIdp9N2z}clqL;nyAI#m>Ytbs6tsD7AYA3dG8^ND~uX0er9DVMwh- ziohwlltnwy{eDddni^eypu*dyD5b7GE1^;ERg>}H9Y_{7%5S`WiJE#KyBQOhs+Bo2 zWBr&~OX$OkYf6{#yYsBP>^0kqaw*PSho7Yz`Z;c_L4<3g&@+vR}_Y>UUOG)I0v z`fgJ0^rKzDvrpvYs%DnfLw@_rXD)zY{+2yW#Tm6=%vYpWBTV~cQV1|AeIWyvIiAR! z1H9>t^Da*v2>#r4(HQ1?ZKs6>gtziqYxR1Be}QjR01l3TP&)CQj>$Q(?m2(Ew5VI& zVFedmnI98+x$929%$pX|bGAAy+U^<1U))V|HdO3M_cBPAnVNLdWXk5dDD+n*0t!7qAmy1Kmbz7)~vLKt7b@ur2y790EJT`Fo(>?cF@ zAi;m49oWWz4ewV6?KfVJcwTJ)fD!Ok_IwJR>P5G=d*M$2M-AS&*XhE&tI4ql=87Sn z;uO*ce{J@LnIiZA23x3NK4cu9Jc{0&2W z7?qXfWr}-}J_Dr>=&VMGKRI;^Du7ZN<+-aXD-eU(irMT=26@Ts0;$dB_Re3t$dplL z?Mxjm2I21vZBt8Q>uqjYnL5|wCs&N#y|X(*<@-z1ylbM2Hv}y{3w2B@yD;G5f8)HU z)e_9(qe_~~JG%>ShC;}K&jx$&SnGk>>z2e>G&sC?7q^Pr6Kqj(uQxl4sQ$f7c* TXg3P)QQekTL*(AL`}F?+hl@3v literal 0 HcmV?d00001 diff --git a/design/FY2021/images_cpgfunction/F2_timing_comparison.png b/design/FY2021/images_cpgfunction/F2_timing_comparison.png new file mode 100644 index 0000000000000000000000000000000000000000..8e17066f4159180d9b403bac43e255c4eec82765 GIT binary patch literal 31555 zcma&ObzIa@_bxgp$`HbkA|jO z@Ee_>9(VXc=ycn_6@j=&j{Aq#&bBFwKrkUx6mIHy4KI)R>*?**$sUji{{E_T^=I%@ zme^h#-3#XZ<^<&D=CIj4C)F$l-%C<+<5HD(BIY&vM)gatQ{8)FyVhh%b<0SrQ$jpM zd?22-jl$%{pt?g4Wy(Q5>4O3Sb2*Z;mRI6GhW|#bGEP)VE+-H93TZ7TdpGFo|Kj%> z^6l|Ff4nz75t{V(uQTppgE-!Y*Czh#?RD9j*BN+}m!Dr|$UANie?i2)R~%PcTv@r1 ztnOup2fv`|Rahj>FD`DWA%>q&*3_AO3(IS3k7Bod>J-#4b9|%JVV3>p6ejI^(zq2ap7SvWhlj>Z1Tld}6fs)pqi6>~E)hI$sxsczJnU=ekX5y-LO$9^qwH zensi@x2|2vTFXz^`S6bV*HE%Cbhu}cp|x<~f(C{5heVYJd2 z<|DOh=a-S=RQh?iY8W~q?p@HTT~}wSXl$4|8|7sluf6rDuc2h6j+aQDUsaB=dh|$* zamYWFlEv1>=C-=}(N0A}v=$+pn@eBw%C^k#Y6FYaXh#h-w|-@SxQtA9^|NWQkko%yUrKMNMjE?zQ3JVFjVv*UjqH6(rI%y_9@{?K~$=Pj)lBeiLHj z?P-W5y*`>$ximXe$nFeU3O(wM8t!3PLUWKkls#I%#UjscQ#L%)k*NHHY_!UCZaP+0 ziTw@beRjXi)CqM})retDS~|LeC9~5AP3>O7tRSA3-14tUhgUqXn-z`|Y*{z#e(m{L z#Y~UhD>7&Bna@Pk3S;JSG*WlE90Oc;mWOl>!yl!iQb_L==HDxRzO0)wt9e^PBYIYo z_R^(=rKMZ{ZpLHq$?D6~b2=l0uCtw0{)ZwG%=qc30MBSiNlC>XH+T1Qi<3o?3pndm z2zSau4ip+GE6_y3JJpb)3JMEZvwC~)oNAeu87A@?J&!$yJMNW}VXSt3KMfTXO6TTH zyj01i&04I@=OKF-B@g)$3ZJ_AIT=4OIFv~>m*pxIO(6MqB$2DKs+yXst82*XP|_=c zv?Mf=pFx-ldMak$Z!#%j(fPI2ZATg(PTv+!V{g*OZtl>gZ93 z6cQ2&eg2%}{If1=N}4P74h|kqaBBlGb&s5ehQ_BxZK?jv8~yw%w6q^bRzy;tbf<|2 zM{jvy9BphGTW||OoBVmox8Ho@Hf4q(`C3}-jg4~7&H^$t@TZ^%eji?wYBxW>H0hY| z@SCCVJ2~N`=hxvYLQWBh`=WEfGn%~-!wTQR;lt)V6#xE4yE2GjU3T{Ik&wVywKlnu znWwsS&r2*8i@^8{e|`)vNxUrZwfpnqzN{RA_gWb-`^*0>VLS$g5y@9y>3>+>w?Oh! zz|z9v<;#}`hle>Dk>gb|1(Zc5H8j$E;>^tEtD{w=)?M;Fc>{cMGU7vwN= zYhyK{o~xra-ADs=SW{;G1K5lDk7Zh?W4-VXMJl> z_cv#|F4NOnG4jGoe0bV!-D_cKi6sn*h3mgz!HqLQ4~_djC<4{^$fKt2{d+cUZf-U< zCECgl=k@dZN7Nzs1Rs(yzZEeX-qSf{F3~RWVv{KXzh2=&>t=16l}j;FJcgT#tH~E`oQuphR8?)SS*!10 zAXxe}l@z~Ney74)p80pl!xPi?)5m|@H`W4AeOY`?_UGnAy*c}2v{Gkl zdf6zFhL@u36nvG2f|ktwpM&kCk>Oz*kzR1$Y9)lxvu<){XNOE??@d`5oF*(OIb!MTJ=>YQy1M%7 zmp+SOz2E+DnVo@Je*eL2s^cdS`QsMTuq94ovKcoCtu!(0uV2;`BZ+)w^>A68Q1zW* z7{i|r>1&_QSNV}QIzKCmWk5GfLN%u^d>GnJLo%RiHvntI{o`j7#INI7$P{?5B?iHIp z5lkjIf6K(gL`zHS;lqczx|DWTbBl^r7Z$*KT3TMsKX;Mm{^Zwl7pbXBh`|Ow4cWsY zpnhey{IkY$t^7XE*RNk`A|HaYSgwJYi;j*i+O~mfxx1IOxQeHD_vhamu@_0#ym#-O zsIc%i+xqwK-=|A@9ymS0Sui@niR({7TZKyZo6jw`T*V=*Ctf#chjH%es&iYQ$RFHU z8EL5p$f2r0KuYg){9<_6P~Xv+URmMp-MagzEk{57X`CHK8Bk0uI>i!1K)CzSCOCnvW#8(Ujj zySv}0IqY;EjiBbfZ+Yy$oHv>-=GR^N|^WC*C`quek4p*&Oby z6t#HoO@%&l`=TX#{CWL;HufXMYYlU@EC5lIFTOlj*Z-qIBx!i_X0YAN!H-*HB?R_y zSsQi`H^8Z1e{0!j6-1yxSup`8B(54l@a)hiYu6CM| z$G!-8m3u3Etk?g~3*|Fj!aC{fKf8G7A^gfI(&`Kn^{fBTn&o99c_MdPA?u4%hTjIE zZ==#!$;%fS9bVJXq6)66v@}U|o86LG-BevXFFZD9HLyZF8L2!T0&_hXWVh!&j?`x* zDk%W;yP7BkP8dOL{^tR=p3qI%u1sb&#!t1gzx zp6>T72lp);?e_$(j&zt`v@9L6wY4=?Q^(z&Gs@X5gNLW-S4_!$o+3)SK|?99lp7rX zMhPk797}kM0rkm+U0OOl8l$eI1$HWtzDgjw7o3~S@ASi6lQ4T=$>5xUjt)5^W5%mj zuM!hsVZC*-4tKie^Y!_KrC)_34S}Zsfyo$Umg}#7rf6zTc=bvHu=(l9@%Hxia?0RK;@5br&@nI+6%_njaHENQmz#UL za&Ae6_p1rSrKREqOhUrXpFh7@w9d}X;*jD_gvZKo|KMONQF}y0#P8p~BO-_ZzZ*8b zN=Tp=cR?3E1phQT@PfQ{e>TeS0Pkthm)5}<2es}`-Xa9 zcj~t@^e7|XU;&`3uCDG`Z?;Xowj(Zb|Lum1^!H?A>gV$NQ!tq8wMq~pnlGB&)^n6U zgtQ^tUVZS@phHDz)pu!W>85McJ5oKH#56o8>U61|i5k!1(o&%tH(r%J;)OIzGQDrX ziaBEmJ4gnwH#2!Y#c{Oq%QzoatIE_VFi^I|*|6}}2q)Uw%i{>bQpI$B4}Z9E?zg5W9L}F->rUl zM;KrCdX9ZT&VKi;5o--aIgWLGe!{`)Y%uW*k(V)ZL!0vHEwdi;ZW37b`}zH<>dUQN zUE9CzWMjQ*Z^um6_8ZyDQY25{C4Tjh4B4vis}X^T+?G?!J>xu@zH+)&=eukDVPnw? zv$nb#FqiJB&K9qkL+MdL`Pjxr%xitJJ6)2JlCq|{+Sbl)h$RHPuFSG70bC;&c~g;& zjt-K*q3Hp$K$EOm%dg95K6i;RnCE)uX{9Y@IEqVNo3l{g2!iAY4?BBKQIP{SOL1su zh+^gq)`o|N=kv%4MC5?UV1hfkx?{>jIK0c%#Lv`87IvP#g)^Bq---j{Bo{MHC4x(* z^w3-n{{!j+;3xPIt5!U)3o*LI{vFO8z;ASZU3tkJh5P=$fJ^jV4>o;dqwBTQ%~8Mq zpfBO=l?%vhtFKUtMNmXatjfk7B1l&-o3>k!Z0{h8(ztXA?%qg8ONCc}>#BjLhZB_8%^q3VyEAX_3-u#^q7Y0{>!Ke#$_6$% z`hWXz1xQjUW-6%X9^lU8A|WL`8EH5kSsp44Tz|%>o+@(gs$@@^c;DFA*zmBbLI+G) zjmHxYk8-8#DewBs%-}F^A!=$J`2grl9^ z(#lF#TN`$18=PfbU0r&5y24hslzz?_8QGZPja$c;=-!CGHbXT1u_@U%Il%?oHqWnz4-U(?(Qxq zDA2|XJ5*eb{tnoDfF1C&kB`rmPu-XU-}UQkSr8$?N9~xzw^{Min)URn(B=*EaL}9tG!i$PDiw78`ef0K zAD{clgYfCYeM%pEdD>>`3V;m)QB4w9WNKHr=bGsZU*5-$A4^Jfa+Q-87j0wbaQK}b zqJg`+`>k8Iz-<+DLjIUT_U|1E>JnNxi5Ehuv?NGFgW}@IHN`U}%-!!kdX)3!0kZ4E z($W$y4KX?ROaOK}R@T<;l)>G*&F$}jkSTf8-qA5o zWRjYg$WE`|%Sc{$zvL-vpiFK-0S|z;*wWHc4Yv4Ziz#mmtH&q4i0MioRz9|T8gTmk zjqo+8X+uDINFty~94`CvfInHiVXv7_K{>7-{3gG4aha;<9fQ%^_xbsDcX0D@v!HG?M23k>JRsHi9~ga-}xJv_HxC^P8v@)ns9Jv&|YkUXTRDf@>7 z$%rC#LZdb=zB{tXl8Ej9#ugvWnvFobh;2~GuJtR8D?FMr^i?S!mS)9c-kkIqc71HS zsAI-_HBAuiLHQ{u?wko=zQiXB65;4D$Q&tm@|?4|$Wwai zDXQZDZ$f19k0>^)t1A%P9vVfQ@2nz=vJCY(T&a+ziBT24aYH;^bKuO28P_jfI_;sf zK7J-rJ*9SP>gwOvFCkbaR%G7s)^ufdBijybnm9?2Un#koK1}g4S;03BpCBGu&r5#3CsL!p|8bdG6B9YK+G1L$!d)aufx0fX)mc=N;|LAW^)48Lg zIa#vtNQ(1f5uda??N?|cn|t91`nhP)aVC~@&6d|NSz%5cqMT2XWfhm6RJEXW8Bkd) z_^fZi0`UbhO^YAhH8C(40uZ>nHeqIF27vGc%yM^kx2F8_eV&Dd1*L4Pjqv-N93vi@ zbj{soj7N>f+TauZ{{0K!4S{?hU!MfxQX>7j7oRPFOoapt zrru#pA{%>OS9kiaO>%PbvWT0hsj0rczJ^9uQwVWKdwW=JULKKMJ3>w`Y&zk!k1llw zJS!5(P9~)GKFCe3^IrU6GLGV)`=Ez2;cs!E)XFP2x-lAktU zBR<1^qTb}Q>}*)(ec-j2KSa zswpPtO?LmIG(o`zu0Cs9OYwm+LUPo>HETaV`nt&~I}3}j*0MmE!{t(N6K=2`z~9%O z4)1JVkqt0b>Db%byJ+SofFBSih}?mexgyFScd4NBlIC{!su& zYP&`c?Km&jc+{lN?Vls2sdU`y*gOhl45;+mcQGH4z{~;W%>M&9B=NpNr2M9oW5q~_ z^(VgeW81^SL-CkCxPvCM@Zs-BKHx&8S8C#`MI0Hq%4J;=CSo0cQr6HgDYL!%bEmPj zwbgg6E>u&F{iCMW7FKzt!ZWWc^X5a#Xqvgc2H`!I8?$EZR3V2l7nH92nwSWjYC5Nw z{y_Zi_S@-6<2I^0gM_Q@9dDVgl9Wk$u5#D|7cC;)hI99Hz4}aK z>(~101#DL+^O#=?+2`d{)E5FqVY=w7>h;yJ8cRz{U=v-&xF5E?WQBk_ zwNVWal$1fL5^4l6!4RCmD#~x0-@W_Fmmb)w;I~$?qAU?wSy8+)n5WTfTl!RaR->jn z#CtqI)RtyxZ;|wD0(X%LZ}j=GxzvC%?cEbce2ZAQ~pghrrP zQuXui)~9Ct-@sLxlm;IW^d>Hv@R$&J$x(orZ6}Kb-8x&HlhL$8-_k}pb`}waf21LC zNU6x{pE@pS@JS#u13dasIF~%mp#zhhH89enQ(>Uf@gh}5b1j`GJn6@2P#mO#ddm6gbMNW4DDIM9 zpMiHsR}C z7q!U`3hL|YKYYN^$HB`5kNXB`sDRY=^4iNUu5Thh`U!#A0A0(^-&mh^!;ak7XJ zb+%uM#~4*QnculXfhT9<=gr;7drJsr%i_jXPY#Fpvy*+QGxFNEsTmW1u8eHXdi&7V z*VojPfqcQO8vNNa0y0CM*N_H-A)C~Al!!d)P7_F0hlKP?<v zHPiysk|RBwI6a;q4ZR8_iSJRDbPf%y(m(p`Z`gPN@+*PENtZN;@tS(4-~HS%ddEw> zK})IWEtZ&eleN(6D|07T6GRmlz3Wf_bf{1xKm%G~%7Tw&ON=vph5zBs#L=uMkb>~R z@mej;Qa!S}qpr#^^n^RzmkNYQCh^wB>n<}gG7efkPX~x8v)*`iuM9uvfy#Xa?h2h2 zBf&0xhitTB#A(Rf<5E=A_%o;Mh%gpWQBfhGc#Yw>2W5r^39aonS$b`RfjNNs3Xb`0 zJaH@?f^|JuEd&hAHI{5F#K)(iKug&?@qRE|Nz}Dv8uWlDC-BSX@^T7b%TWC>MimZF zW`ug32wk+exOf&JE0B^v#)D(_D#b&{$*`sA$+d=Az-ogc2tF8~LJyBgwVqeOz>?x_ zqz0QV|NA81sVWVPuHPls6k$3#;A2-`JEw=vbUE01%;I8ZXFXo)U9bU#)ZLX}jS#VQwW&bEklvPBj6_753oH>VJ%C%@&r7;e8{C?AwHTtyjsPSSc&4KrK1J9KuzJ@<1M^)5!&ANa5 zz7p)T^;1{NJO51rGAAnT-Lof8o@7O=LmH*B ztjvK8($vt?v*PR2{VJD-!o@OQC76Y;6d3}AfgH-w>`IS|i~H#H_qR>=SqTY=tUOAK z)#%R8enlVuw!eFvU41%rs-nD_mNDjU-6Rw#TGYBvQ07n3+t&x2L-x+V)UQG#u>W~9 z%1vJEE*f|Pg+BqlyMXu=K3jLuWyCJa9BAhvfuTk{O|&i@q9;{WRaO08Pq7h!O6T|8 zS+-e4LwXkz6O+4yKB|lGpj7F;ZMa?6nD&=3P^BM0+tBG=#}Fa$&m9#gz(#GzQbNl2 zgOGV1y>7%*QCh0IrGD?;Mb4zbZ~vnO0E@cc!KwSxzArN<{e^}E%^rU^1koG;cJhyho^(3J}iBR!idsEi==Ag^p*Iw2dF|-s)me_6cH5nGMex#ElN;cJuC>7h2i>6 zUjF$ry|hn)@&CVK`}BKU=bTy1P1rx~;_$Ca+xp}Jr)0_{AnhO*lpHIyIefIVVp|-mij@odZ3)hw2h84OoeJf@Wy5FjzuhDN#QZHm z1bzH6@zyRd!BNDOLAuhkekeRp->r9w86xKJj?bi74A8r`@t6_0c`SOBwr_H;yYbN+ zKl)$wU3Bztzp&KX+?nTQIy=OIPmD$9UhNd7fALX-%y{w~8X>1JRe7MUy|8_NIs9=M zfjTeNd@6!@#WcZYN!-ldCQ58;F(V#bj<0eLr3-(OctgZ@&;4xplO!@p;iK~phi?YzJ>8KPw8sz~ttgKAn>vwXr z?^I=fVO;#)*|TT)YS+&k;31j_BMAv4>qN8}P=c3?=0~|P>;-6^lE_qlhllPd9;;uG zNaoQ>{opHCuGkF}Y23c8efsd~)wKIecEW#mGer?h_P6mL#}gqhg(vZ&!^gbT3tYcT z)mETPA#>e5Lr!fz_`|kr`@csM3@Vn?QzEqDTJu?D7hov_EExna>^hsBu zZkQ6~0VPVX$p-xXn+>z#k`e%lB`t-8g}N>mdMlpy%ch_%P2s}`H#axW69jP_R!nNp z&CH~-@@y>e*t?4CUhy4^ay2|0?L2L+C`IB>8#85o=3L5rcJ}A;GmMnlr9c`1g}gIp zFEaP%4_b^ZffO0?Qm0FR8S1mIubW7KWP_pQ+qXBj5G-aB%9N|0Ra+<2_GP!1jKv?S zay4OZ$X;%kT9V8C-12xvWYH_K-t*@~kaq4f(IX&ddV0uaZg3<^W^G{6 zm3AZu?32lty{~8QJ?>l>sJgGt!7KCFet-8!qN99ZVa9H2(g&uM)%a{vaiBrhEY$Sl zOoCnpX?suMrtbll)TdjIJC5q`ZjZUNThSHB%Jm(|5Q{Q8@+c7Q2E`H1*ffDUgUfyh zixm(NdEbrsW@3p5bBGAtZM*1tk`Z&^q((BL;+!1we-FN!uRjLm`nj3VtAR^K6Ddv2 z&e|}ut*?AOQJbHbMod2qs4jj)kj;Xx{mn?O6>TqUZg0Q*{KDm?Iw9u)uQNG)kL$<) zPx2;^tP$UqLfiM#x{ONp6wgXN|dO8xDR0TIm;Ar`=TEn36fQF?sOvt6ve>EFAjWnS{`8j}@@5)x`e5{aXLc_XZjK4vELx`2d) ztw?H4?O?UXN(w9vqHS32(~K<#_tj>zxHIzD1r$o2KY*Qz=lSom16)o- zYi&4aL}cVbA9}e(b+U&Efed_>u9up<%oloX*cEH{Vy^!F12}Wajn6E5>ZWRUqc1Yp z9rQBg<>kCa<+Ol28GhYhq9Xc+TD#bCjh(&WXrn__RXS!`zP3exOCEvfpANO6tGrG_ z`!i^Lsi9!jGLRf87Wx&c-bxDKd2q=tVu^Wo2(Q=bOnQ0R+0?56L@$ zRQnfm619A}c+7G?DjhJF8#4n|@yr{w@N$v;XSnswqVqLEscz`r`Qv}s@_Fjs=N5{s zAQThwpSKEKe~1V@@XkN_*hI|LghW&4;WgbIp5741Y^rH2p=%Earl85xWu~t>kM4an z`qqmWdDe?4$&glA49_StMOsM*iQ@T3(vR0C`z*<`u--8i<@L63v0N6p0^vpVMx)7t zFt;o)%vNpLg%n33Wy&*LO_Oe+O3%tZ;63fYk@*RGCtNIxtOvw8eiU_kZP#8uG#8k9 zkTbXT{6Pw`QO}hO8RFBOap_Ahp1GVRo-V0pG#=CE*>{{z+8(plsb+k`c?i^k)OmDN z#;xOoQL67YU{eT0e5O@nhP}7w(6g&d=t+=vfVya9X(=_2c%25tRlF#5hXQrrO@ylq z=s=JLZ2dTvV0jY!aO?N)-%pMwPW_#n*7yES>r%W@!AD2+TpBk!*|-?O+hx5uI7ll) zf(r);Iuw)#D3p(+*co|beRsE}dFmKvcsr0ef!FP4+Ft=FqiQ%iH`nt0`|q!)_xz>! zi1$zb)K0z-k9>4UgYuEWo%abKKR7o+v*!W73+?6RdqSM)cc&RoY1dr7-BLm#bZ-i> zEq9Dke*E_)t{0HWs1`J%|JIA9*R+cS<4gbt9NmYM1v?H2F~`sNEKh2`p)D#*1y zxJ-tdc=DSfY+LM1ZnEcK8kl#rogldfZi|Tx?F4s4V9h&K=61+WMfYD%42v`8#v61S{ECUe!N!9_-UqAz zLoa{JN60fN%eMGf_;u1sFvz3zqll=pr1-5Sw3( z&&8%!?^;6UOMe!9eTLlFKtuv8Rws)WNI5xmXGwl*wZ3(f^y)@)M_3;Wl4z*sQrRBJO zT}f8%L6rEysLbzEwh5Q13Xa1N6~>8?FgaG&V3R-^JZ3M?|2?Wqo>bokt7`I42OHD=PO&UszK}P(TmUG5_VT7zO zm;%gl3QZGA2#38?y_%Hgp?{37cjx@BpuX$A8XrTBl9%%~o~p{uwpC_WytwMqR4yC z;i^M@@7L-WHU&8(9&5?DQSfA3e2fAqDA3Z#Q4PA?))B%6F!vf;LWq@9Pyo@{jnpon za!=?LJcz|++{as6!?7<@xhRf4@a#pTO({Iu)85DJ;dTD#HSUJGG5T_~&FxwEr2F#&v2h4^r^ zI5jWp*NL=s-QH|Me;P9O-%GKEN%OQzn1DT1JQg%g(uKfvC=z9P>es=5T@n0v~6)C zq2XSw#AngiWt8l+0B+t+{)+7nK#Khzr&VrXY)s$cJnxOrc2|Vzj_dNEKx8wE6i~OT zh0bbfYNn=`B_Sx~LY*}0WeKkSKkhYsA+_uMQ4x^>+wQTv)M%;gV~9G>;B6rXV4(l#g}ptJ=@hAp1(*@rK7W$f#)){1b8LMe1oqvO>W z8>w$FLT*bR4W)4?YJIXaF3{(Wq=>yD`JK*AI0Z+|zKt7W^LW>9D+-E#P($K`0zTwf zFYLj22oeb}pK0Q7f+9LRLdCcbH!VTpm<+xEGeW^pF&A~1Q~E!IB`Z$t=X=CiGgj2M zVBMJ%EGjvy^3xCEW&Dm9I1O+eXPUx9MMQR1Yt|@qfEIvC-UG7hmr;U@fky)4NRZWW zeD%i*d8d|G9Z&`AFeqvB`-ee(uB&S)m<;%ee?iQJ!@zDByEH|>Y9T-5+g`x&XEyur zDrV_v4-7)EbY;_V2*@jsMi*!i4?Z=0(6r!ub7&|8)ag~0__lBQEK<(Dzk#e_sKQOc z!2=tw(ZxVc&K-qwJpsiz32nvu^g8lvd(1x*)@o8j5s>hX{y+kiF|KQ-@_iu~X z36Q{2$*?Y#d^(*-9GuYY-AM5;X?Ysx5qOe%c8#Z^PH4)E?zyF939TMNK>lvbiL$bn@H zi8PzX&%}w#uuxi5c*4H5yK7eIgqMzbs)=mar%6x$tziw~F709E=i+9j>>Y&3N*`j| z@1c9`z9F*-1)lP|h||Xe9uEXY$HoS(R?Q_W6SLFZoag!o(HilSGmWbW-l8P11?cdB z)Fw!^74!@Z7f(6SD{D+!zm@E@jJ-pO_WYTK6yHvZfkC_1Y7Ftc8M)dIQXy~5oVV#z zy;G5H)YuIXk=F?cZ|>nSJq_tw6f!I<2iZy(XkK9LaZiHEzRS9A!EG7yDdYV4^AT%b zL$7{Ib0))|l~bj8F*}=0AN8Q9--GGJ3>b`nd%sB{=U;6)f^=PGOsL~7xeLn>4 zkg*Yigj-k`4~-f~AE8wV2pU_#WL%k=i;0``I3L*rO?u(NC6e>0 zU2HrOgQKI%!{?h4ZCF0u<2kn_mV&D3RFvNJzu9T+s;LA(=gWd&) zx}yGpxG$g)vdOnFh&OT8&I}F?zG&LG{)uzO*@*m}JW*x2u`xLdQPR>~J)7h8Q1Eo#d3zqws8A z8k`mqg%6eH(IaZ9LjqM_V`Da-D8eyaIcA7NwB25qK+s<>2JD$X`!d<%eC>?9($p|y z;yN~T)rhCRdy2WAh?jTe8&fZ0kqHuS{TK z0k%BodD+-DDF0VwqwAUn4Gk24*=RG3xvOGgNGB|CaT};4I4)Ca*?@w-gicW)v!&*1 z6uhnZQ}Xt@mzxml^j>&EZ`qa=a1DE7b3_JvcIP)7Q%qhI1*lQ5L``wl!`Re(#i_|_$&gY+KOa% zl1K-6={$j=lJ{@Nc2VD~wDGZ-Des{m6idpMaU5iAeshu&H~H}3)@gM5`R+rX9?lbZ zA8GikIR9SR;C*Bl-WOrSMS^5@QR!jY-|Ex4LaHn;e@21`!6fViL|&cgge#-Zt@ep( zs#lPbTjU)1|7IsgJ*DybwgCWA-8gzkbC;NAJ7Rv-V?p{(BlS6URkC(rKEC!ZY_;Tw1d1s39RE z^YrjI@S~r}SGfuc0n98f=NcQ8{%abxD_5CFj3C@KO#?%+PfuxD=X2M*113}U>uqR# zLJFxhlBJSj|FXZMLvpLN*v>a=}C#84(?=ICmB|Tke(uQ*RP4hh-JcS=`_t_?igGN1CMP^{ZFEhDPuL@eS!S z#RVvBg4h+bi6IUI2;`;F8<_vJDQ!Jxp~xmI**1QQItMfq&cn*x1xZ}o%a?cV-D_dE zc=2NPtv}w8go=2~xfDH{(7Tt=wLf}R{`(xx8h@S#CY!$H(%LOYju{-P%k=?Xq_}Qp zcr9SJ1|}ffn&B)iGeF3FlT$*jvNkGjB*{FtV(ahjV@>`}5C2?DVgT=RXBEwH%v z^(CYFfXQkR+=ZD`PYjKWUI}luW|YBe!3xWL)g-O>Qix;aZ$ltw)W4@)ZOa(4_h8iE z0DwcQfyCuTEIRZ#9Bfn?94fwPs)oQJA(8z(mK$>nlqp@-Aq2Hve_4D$_e;&JKtGhE zO@$)Ta@oXsbItMYyT%f-zE2I~aD-0EMRYPNB?u6KNF_O5cUQLw=(jr?9`eDQD=In~ z#E7BOe|Do@)6#LIxTU>*y;GWTOyWoMJim8Fk*NriC#0$yY$vU!kM>?S1m6Exwoc;T)-NDsEtMk<5nPQ2<`^ zI_Kwe`OjH7IejcH?lDqhbqytltAPC^wOWG1)PAV&0i!^aYvp`XUj5%N$f;8S@YlNp zwSd1glw8Kfaj5^CfTC;$In~|omY6vK;xqxv=S`Me$sgfnb_buLDSR#dwhJnnyPBTH?-Pop1+kX(B%7u10EB4?uqcErP5kZ2d*=rE#{1K)z=4 zp%?sbfV6*G!4F5o6(d(XKwIJK?_bw%JLAnDO+9{1gcGCJj|{lzic?Ye?i7G0s<(Qx z@eo&zOVOyb>)iB%9tB_yKmqUD2P%lPw6sSU@AJfi&soh4ZpNZe&lT+K?2L>qlRvbw z+LLMm70PO;?9o)2-GG&mMb#zfAT(C|j~1An3m>xI6v*LmtKQ6AkqM~26 zW%Kn;EYd09x7Mr1SGHZzj4jURpW&H6Q+lvqbQmO`85m4AR|B*;LlTL>U?9J^Ktcjp zZ)|KVt`#9C2b6JDpFX{O^;|_o1^PMW>-nb(mv70@QWW*w}LN65d==k|3z}-vRaD|zf8ED-CEeCeN4RjMqyCIhxaiqujj6*k6!9|$K z#XKa`d7)$r@arYsQ~DMl3zk4?gHy!*@Skd_+r!;-b*UNWA`sg@P&y6{t8W9JnxS#J{DVKEciVZ50zgBrZohd#Zf0zeIpUKGX8K`51?G5I6MEbKU7` zT3W@0h18fKs0}(hmo!^|ZU!`tNfrgBL>)DMNQRG2fpNkz5T9v6#f>|Y{&TI-)KiQn z>jW%52MakviBdg`GJZ*6kh8%jY#x-Ba>XAKUPh~g+4M0-^lR@BCwxTu)7Yzeq}@_S zUOumK?SHfY2t-~!a_ATU_{CB_gX0R&YT&e$bkKng>SWOENoc+Uozx!^RbTpJD#8EM zq5e%$1ihq36iyG^T|ZjuEky3W(Z)V6_6o|!qF3bEL!w>=#ofd$DVPXomki6TsYwlk z@_DlbbJp_mGRFd`NXc`;fdQZeS(r$8XEL(Be7PC=HJITn^q#Vq`ddR12httc18gkm z6^X6I{9+27hY#O<3{(L%19?}n5c|_pBC~RP1Fwej;m466aLg<&Ub{T8_x$DE(R1J) z+Pb@Ey-mw!K+8!!^}IQ|J40{_`;ry1eSJe7 znnL{!473RVBMUl+A|nQ6!{7q!{2K9e;*>Max5^@?kAO6c9mh4v@&rriD_+Ls;Lh3e z`=g_yusMR8ffowr5|4ij42X(}wK8}Dx`J+GPcK;JP3pHUDzNE*i~T{C72X&I-h@ez zy9wf`x<~j;%-qPx2-ZK3OU*yN+4@97V8fEf_i<=* zg{H5TmX^}w);Ffk^p{TKf5*=U-pCa4n{|FC2lOWb6->=Mf++%FJSS(HjqtC5fr^TX zn(ObZ%M9T@VB65No`}Ji>gd3J3?9pa!K=D81Ot7sr9sdh-S=!LFDxX(LYMeWr(>`8fR)4oY@pkVT9m7k<;+^q>Jp$x*F&RRRB@^;7r+lgXG@2qBqSEc<)Gr>=H|986?+}p z=HM8^UR;KPy9zM7brNpd(T;w%rba((ZEZoPdDcCc=uz|Fmrf7_<0Yj7FXCm=>k!(A zq{6a-!jo~vA2A1r3aU*F&}j|>f;$1w{o?1(pVP#gpU%jEJoM;fw}J6-yrp3IkE(l8 zI2+*rE|}x|VMEA8c(MUZbn{{P|l;3vALseet5>@Mq|je+&(S z<8(%a`q0_|X+F|#=^uyd0y)``4|r@QI&S%zv3jSeCXG~)D>8nAR)U0qx-RthjE#@7 z8IL(8FJgy>ZA0+Vq5gj^P9$U5 zd*MIN-b5I1vYGLp?qLIiHr4CUwFlcbfGq!dUZF?C=Xrg>2a3wY$klq+s zzdfg$H^0E{3Lt-1QhL7RNlZ)}-s(S($#8LXgVdQcPUAd|8(QB^bEhVfG^fo-bm3DUbWcjeg4| zV60eDz#6GcOE}7}W&LCR+SFC;#Z*{^Q=V_GTSpgP%QvuQoc-+U2A2X!R%*}E05dE`N3UXf4+N7|U zE0Q~%LOo#)-56Lxub$~j3i3%x8f|q#ceQatLOSMmw{4lvWm{01Gd{ftJ2=QI0P2HH z7Q+7RN221s$9TG@JoWd3Zce%V5XB`E@)WJ~c3BeaC?O=)im*=zTC>u9NniN%aigis zzYRjB88AS>$28l66jda(3*ZrSq(j>{ZljK$gM*dO3dUp--EaFalH4b13)X}yY{i1> zue*1zMkWx{^I*oH#SIR|o9OoaKV+1FqPaUg?LZ~g$duX-ecIde+5(WRkeP$=i8!j{ zTr@X$4YVcn?nE{>t&;a0AAVGD`_uYj^3}6X_a)|K`u6`{$>I^{Cx7<$hhunf?PSV~ z{=)ctg*7!JQ%(iORe~u(^1le%B}Z>CSwWuB;J;flabZcb5}K=9W!!C7L3E7o(ZFJR zTGoff2LDs?-`3K4)F+YM2XFpg)gnF_b*6qkG)UcnmON~>GSo+SY1F{9VdPEy{wmg) z%FZ134h?boK;utGkLBXF&u-#(t|njz(UXcp_9C!}#>@kBoBh9slfmk?Sp6D$s|?NgeRc4yr5m`Q#4+L6?4x^ z^+2_dot>Sy=QhYaws7PVg2{}0L#l_189bf;p+{_m#o2lW&_!leDB_RxKpEkGS4GmX z)!>iyy8_YUR|=4^JM8Viij&x&>Mb0LtUU#!775h<5Qqk@Ch$BgdH>ql$c7X_(`k6Y z=(T?Sh41o1nmM}e`I)BD%G;-JKmV=Q;*5z*(#p|3BBaLrr+^3B9p$rFoh88aO3yEo z=J!|I4+*-P^r<)3+-_l4ydyqu2nUaYb-rD1Io-;U&BSe9qJ3Lxzzaa7!q5MwvM&#( zvVGfCsa8l!q)0MEB143fafwu>21&?}slgCR$`p%`G9(ctBBBs7N2tt6l4MH82$_cr z`+U^<{=RST@7Vj;>yL`ode-yY!!?}Od7XD#8|Q=8n8tx?L&vHl1izBc zw2d^CU%HdOy0*5qvht`svCikNLhfMtBop!hq(RV;NQjGr*p%%2cs*2sm|p6o87nD< zU0oJ`Qhm>t9Ss)q%lvOeNx$tJXM&@R=Y)fO?R+xkerFDHx~~_q(6#AR95keQ$w4%> z_=)n-_FLl~LB;*`7)vD{&AaH7&>swR)zjAgCLiyVKIU1frmK6e-y%#5Y%&&FDnr;^ zEDn;2$~0QGu}8d$e}R3!5OagmKP$g9sV(D8vsFnSYLYvJaNA^t@A>a8Ohh!Ez9tOM zLARPdUGDIv2E(ZAgJnlRNWAb7c{dI13m(w2WR|)q3?HiwJzm2r#8q%@D&^<$PXqu$ zLmv0HL(;XMh}~n)e=3*b4|wKbXKbu^P=W|CX^Nkdu)C04YVg-D%>RCT`N26Ihg)lS zMYupTJ#>n3>GeGW?<7C-PlMZsaDnfaj}p5cB7jvJybW}CxHaM9VMlyMh-({|$Bj=ViLW)J;r`InBqJ-U zzTYCl+0l_@3&Na7dHz8)To(C`NR!LT{l4YCBQv1DjLeX^5CUW>@S!w`bVKY~uLWS$xXHNsa7LG@g>s5`R0cd0y>2SGs!EEX}%3e3gJ zVgyel-fs9XyTZ{vK|SMcvw%tP)w~o`eDaCcK3-m4VPR_^q!t-^73-eF#eJIsCaEaA zp#v}oOPl0v<8yj3rw*zori0tM$i8IpN5{sH;G4ONiHI=r;@vdIk287tm%pR-)04c* z{*5j`P`I)ko3CW~XWrw<0m3ZNT@(4|KwyKUPQmr1Q7yLJ$v>|4mO!~E1vmcws?C&g z>94u;5R2uGB`RRQJjG3_JFO@UtEmBE_sO=rc+Oc3ULL5sG!CHAE>HPV%Th1(pE>Qo z4PEt9B^rIpjdv4EP>@0*H-?M3EO9LU{7gN$)^%8b>BvHx*gtCl9aLINq<_J@XGrQ` zffp0{b9w$L94OAGSys>@!bVf{&yi2Ic0UC*4JC@vJ0F|Y{q5VM(vH^)gkBdev#RW( z>f!$~SG<0_rm#4+IO29`$6zTYY8ANlHFxLtLyIVc>!D)~RLw=Z7N)GIy;p zrQnLlu?%Hx)IABHnSfiF0{3ksQ1+#dP)vJGm$m?ww;;H8f5z^=t#y3918Kuj?x(n| z_#q$jm%&NEaQmINgOQ;Qp`k1Sx$|ZCQ>7@KQn4ofV`-y!_+AgN^MvVlYInUy$ej7T z*VIOMj7k)ui0Lo5BS4sHW2m$){Lpeg`f+q$`wU|qp>~KneFW5Vn>+q}WaYnugP$;w z#GiO#;xB@Qi5GZncmaqd{vzm~OG8ZjMbMe?`u9V@*4R9t^)udBVMu&*e;o7}-z*`U z`udTrHnkNMO)t-HY_kZ)O=lx7eg51U#18Z<_AM=RYVebon3$ZLlhf-r#QY&R83*d? z>l2o4JFa&`Wce`rMpjnT5>a{WY@O(XGcAFT4Dw%pd19pO1uIDpG%X=Gat%F9e5Xcb z-O)qa*475`EjcLrL4;%?A1n@z8`+bKw!;f&9_PGQ4cvOqn^tPKN2ZA|N7~bvWo2Gp z`7fQ?q$?$&bmQJ#GB&p8N<@2sxCqS<>1;NV}Ln&c0qv=d+FW<4@(9 z$L&(0nhkX|y+Us9iYqtSC}!H^x-#x3#WZ!`ciE0H#ta47tg?amk@>lo`#6WIoh>G( zMnH zS6^S+QO#v|T|1V*>)1@>(TXg-6Y2?{iZr&ZkI@!jp?q{~a%2N%m zH{0+*tTBLWiZUC;Hzh@6=CY$>CeeWxOCi-YHYU6hv53G8niGR?;-F?t7YCiU*<_e~ zsw^kW%e~s(u`9A{Zif|8WBA8tZxo^T|ME~v#OML{+4|_w`AzOXHbMQ~o$;jp$4+74 zZhee>+sx*}=M19YB$hkaCmsNECTorLs+4)>cd zVSS6a3YbQ`c1V=`9?BP`*8rVi3Rl-QG?d%8`WotJ2qCq(#TUY5O}0Gjw}#qYMb^U_u-tWF zKmQXVZzA`|CEfNFYs|l+p3=GIMk&JY)5jjh@8@C4_q&5=ki{=5k!jWP>*(;LYnAi- z`Xw0f>HHQq=ztw#QPF7tauF652BP7OQWNesZLh9bOVNLYiptfla;BNQxH#$Qpl6OWLKH_TwnujvtfrxndBD_j8={=iwcCx5Ay}NbGt$T5nl>5r9hDq#ZfwAp z;xNLOeY7N9f7og>)e;)7;pPzCBw3VJI+uVF@Wj;Ad0W(UWO87X1H^QnFZiF-lbkpg z8O?O_$wd=GY|V6awKPm0n@NiJG5%-PQj5+$%VYLOz~IJ}x>B=!%R148)zsGCpK4%~ zce&!F#97^N?Ah;xkx^8&sw5En-H*CX^d>pAO9U_^B!{nGJ{z%b)H}r$DnNvYD^$Z;Ox@nGw~wc$i{8Ae^ZnzmqIP{luFy?H)v%l- z5s~`h-C}3t?>EDR3&V}|JIHk0)hwiV!W#zQ*@Bp!@RzYyHS+QCjis#~W#Yf*lc3MV zLLo}|&Qq;UHP2>%hI*YrB8D&jvN-Ghy=c*-zwWY+nf=5$LAXZaRh#reuCacx2x7$j z1QD*zgSaxD0Do}tMm@)jheY~f6a>}>NEWI9d#mleliKTSijkucFL5@3f#qDshH(%O zG?2vmh;{Y(Gyh))@Xl#t0~?o^%Aj3$4=bfrW(yVA7^|+H%s(n=vA4#7e<; zZiM?;vg7#m;Yxol!u*O!5YH{k36-q(BM;QEEh5a;p{Z~$H?RE%GEzF63fEl ztch90oZn*hbl+m-Js}a=x7}6rnx~m7c)e|n_&+ObtHv^0wFN^UUg9DaOukVlXv#EW z^PGI%n`~MCY%Y2W#d$XeW8~sAsv_ce0{EL|)-Tl77<1~elUb)oF95v;mO0GNPJpC1 zP_L!1fpXzXnh7jxMKS`>k2cxNXO=!SR+y_bH#Bf8H7otlvju;bT(|kFIy&Zi1*m%< zPRSc6OA*GnRj=ng+*8}1>U&q`$@;QLOwVmjxuivIFEz{=o5{m^^J`Ck&`~MAC`g{% z`D9VXp$7jUULq@P9wBz--vJ*p@DcAgzLMf6j-`V=<-k@+<){1O0`UaBN5j_C;K3`g zFg*f3i2Lp5aXdDel{YKYcrT?Uh~EMeLxdPKY%!9Q811(jO)W#Y{jJ#r(VG8$#q%TR zxi8oe*?jAH&Qxdf@Iv%d{my?sw(^3M{)~oRTjtd%_xaxq&sUZZP0p3eV?MRFE9}}; ztb}!GMxE)$#%A7J$oG?RBzntiw^N&QLWNfO=v_E=ZLupU7a7@p2P#W7$DK556g9}T z?+$dIJ0megas#;l3Gym(eRk9a^%2q^2x2{CN^~&dG{j{651SC5_}}*v-r5y~ zwNFf7wJ{&X!}$>K zH>^@o#2B6W)kjbL2xv1oDT#IG$=A7v^`C_>OO^n-b=@0Y4E0#`)ekfiYJ1^dNhm^q z6Dy({#8(P+kU`z90c_BMeNz882Nnxe;P-J!=XSEvukwm$Y=M#|xdi?F0RdNJA9j`n zX1{#7h8v}t&XF9l>C9-i&)hJm9TZU5xCL7Iu zyyxTNb9uO}wV|PV;;Q+en30*l-=I6#f|^d(XSau;>YlJZKeeyFKg1SN=$^&K5KD;_ zfH`w3Y8P8CGlc+kX%BhmTyV}VtH`Lh(2`S^7FSSEkdfhliV&hW>0?_MXqQ8r0RhHT z+ci=8TX`-s>10zYD=Wx$BkZa*;0iT;aS}SPZLM2n{={Vm$Pgtrs8$ME{K-V?$>n7S zz#Y+3rc#dE{9dJUP4egt-q40()UX4~9vg`qVBYuOeds{s2)2&0&@Og^96RpwYeNrQ zb#(mHHLI+w?CsQMDe;Y#+p-l|F&7z^guE8I<%DRjb5&pZd$q0`p#~s(wd@s|a3r!K zLO$5U1gbw|Q)n3JS=~=cg5WGR0?6_sVAMi*til8E{D$gW5wD;J9=*>XhG9PVtg4#} z@chyD+}fFEgW89;lW3JzbpQ!OHSX$D{uUY<_B!H>0O7A(2Ds20KW18PdZqHBnW&$)wnVH?Jt>A(Z?Pw3DP=q3vAA+?)8)$ zgA{FnZB&FioPpap823HUJp9lInS-Fs_bwO_6`t+qs9&{>204|if2tu@S~vlT^~oLn zu}SA3=Z2aAszKdW0Z;}xCEjW5?3_EfO)lh*2a!XCR+3?&q0Oc_OCy>I-LeOGm{g9G zbo{!$Vs#g#sNJ@KhCZ_9#Oc%T&%z3qlQN#i#3&_Kvu&&8>a^PfrIAgyRM7wf2<_lE z+zTByZ?)Jb=`9F+-SU*ikfAv_NhiQYaq;dvy-}3V`P{}xn4q*PIwwwOYHC&zVGa=+ zuz1yyVekG}OBTr#_I6LpFyiMD(~WfA(2Y2NWDt&L`F~a)CdGP4I>AdLq=SCV)M=LX$@7D#(ho6NQ3p-B4U>`Kb(pLZ> z-FNXjmHfRtm=}^89H^)St(48^VvW(ZLPz&l&I*LTqbfUDH*<2HVDa|CUH7m{mgtAx zySLi5a^W+25>UU6dnE_qVIoWlTKS1&yY>dL0OXDI6)yh_Ul3mU{gzK82V4ez$H)7R zgEnE3abIw$+DEO;w2?cu&t0P-qszs6c~$_#ER|S9XfR2?*}>%+_ivhCH_qD zZ=|-znz6+3?|UgoS!Cups6Vv-#NI2*Am3yZ>K;!`|8F7q8`+uZ z&Oqeq+xhuP*dOKm56?$IH=f@O=Q5DE!GU=JrH3Kd_eHcA6QM9d zYrbGLYN^TK;Oo~kyYL&mV;;J<;G~_LroMeM-pvm8-1_po?$GcU*aG9W5aj^ZZKLq1 zH)}vlD<@gT0sw)iv$^lLHaj zmfzpmKw|cL1G=M&Avh5>FR#chph+uA?%HKQ9#ie^MWr3##aS31AZGVgh@o8F-X1M- zERH}?U>iUi81C!S*3r=qNx8a_L<^?x#NmR|!yxd{{iZS7w7;YkBgwU|XD$h_y@&i3 zbxR1IJ@EpItaogOKM&Okt%>B)UCy6JQprKB}++#O7MRiyu@@AzyvyPN=y(L2*kNmhsrl^ zs1eI!#?dG1@C4aH?j2@{<5hyjqXrUHVZzK5LT>Z&<@<}7NVNui-WC?h&+K(rxr;L} zAwSU-((*FrCVJ(kauM8U<$fioYiN{m$`9TLa@E#bsyAce*$na=&uZ4@xvVKET9`xp$G(BU$OdaNPcVsa-S&MEd_Zx0_^`#IihH3qz0#}r?0n_<)K^)c zT!$*$Z<0SbaIa{NZ!sS$dwDI%{F0rW-3PJVT|^{BN~o84#ou&+jlWp)2NuZqjEj#q zI_eH#ZZ4}{f;{XZP7okYJdw*dm`m@)KpDL*AZY)MMQOTUx9I*xt!>e~bynjj$jL7O zOyC>0 zDwFga^BMh!3?%^{+xgH}d`lOSk4_Eu;*wZKz(rjj%0nRZQ5q5)%#;KWg2q93yLbu8 z?sVM8btD;v_c-Tr3XM3%)GV@O0oz6lai1EBeu8!(VrWo6kqt_uNS0l;}i)+PD?7B@L+%c zN71?^H(W(R(8(L@8!OOZH~sqh3j3J$VbPu!D)^G7BjSzg+^fW9GDY* zI3t=)t3X$!>9ShN$~mUku$p11u<8DBAlNB??2n3nMRc#8Ea&1kuXX6J$Orr3_b|~P z-gQPOwyuvzyCD@KqM!Bi`f5(KhsUi}4NB%(*sRi)zj}ZaY`>nbDl+W65@HcBme|FB5&VR()qhKtHLn_401AmNdeWHIfABY%s@$q~!o0%9^(TkH{4}3U{CtDx946dM@tfnC#GHGo2X%7tkJc-6IYQ|Tu|G#K0@!BMoL8FGLEfV8+n4} zCbA07=Cabxc-ZGEoKDq!xZY`sBb!=rB`#Zv) zZ)CceSgdXxJ!wAEz9PCdL4l7U;lQ%hO22s%MCV{IawTTPO^RuYho_R0*l)f?sPCh2 zeQaW)RufNN?4qmpS}Cm{xw@X~WV6nxO|;@AGcgZDb_6EZ_ErsiK6~lx^GiL==k04+ z7#_6zxX1pOkbp zPve8S5eZ6ZS;?7FqyWi>TS4s^{cW6c%Mq9-#TWJb0-IK}v01d7zH`3Em6C7 zE;s4O79Cb)^0R)yvE8K4m+yXFg`$x>G9uH->5jrCQ!JY=x3|gBvAxSC~mbS?gM31hQX!7Z}6Gx>ir^iP4$38 zaZ<|1uUtV)?-TCdXD*Tf(E^i&qeF`AvnPbhP%S>Gjx<(q(V=XO` z!UsIL@+0l`{B+zq@3si8!_-)BZPn&Pv17AM7>?s~IRbNd9xL`V*GI;S=e%Zqz#9O+%@=$-9abll@-M#qG<94FxGd;<|@ zNwX!(p}A7~G%d<{y9`*Pnv}9-Pc*%-ZkC$)zFlGFAe=23$8k8za>^urf7y6i$cGCB zMIrt<^FLXhWt78k3tF)K0~Uib7^aK8a@Y45&$-Gtenp@={hAICId zBS_cA;=DBV#mOHD(LSgKmG6Gr5y89+NK5k) z`@OVFj-OP&eAjym?fLx50{RjI$Gc%GuRmSZ>wWpt+Kh&(%IEFFwU4s`0t2h;vvCRa zQYo*=`vwvT%li(;$~{yoygQ}<+luGg$$DqsnJa?#IIbnv6vGts4GwOg>_NJIX9RL| zEOumZ_198HuyHr|#*VvK8A}VL2nE$IPm*a6ympaF4T} z#^t|eRCX-}GyabmoxR}#N~=}2dr#wv!cwtj85JzF9D zN@gm9oJU7$9brPrtP)D$Cv^1^mBMePh-(px_N)&15dZe=Y#qCXOU03DBYBq4`t~pq z+4ATYJ`E6zN1xX(S!xajW1irkvD)}%!tEaW>vk{heNd4R!+Mb1HT50Q7D6K5pgrJKF_NS5xDE$8fpdw&KCZ}eYR2tA$cipt;;au= zHw!MZz>Xcx=s6~dimi}s+~kpRASdG5%=`l6E=!|-f&7u(+}hL>Q4^3x)Jub_pGv^Y^blFk0|^-p9B;1NwY!!b+d<6*~7uO%B#D(`wd$->p3dT(JMePNuCFX z3&w7fW8>q^U0eWi-Ep)#1WNn{%9U5A&(|vr-idad)bnIOzL$moOuD9L$dAJ*ar6HulCGOrE>^esw1(q?BXbB|=m}WbW_#tk3X%7ccBE1TxGn}V& za~4iY0E5EuTKxP~KsX)#Sv^p+mIW~m^d)QSL7dxXGrha?eZk_wZ~GbsomcyLE`sa| z=Z1^GtF@qdgsZ2y-9u2fPgZs(l{kP0;Gx;kkeCfI+KCAXmf)F@a4uUR!hePF56Y*4 zBy4TonP@c~6+S`ny@rW7Z}j^V4vtx{vt)b{(G%y0vk#qbS1@zUe1Lx>j`4xlN2=vx zcs~Rcy{oVAgcw(%#@}vk%+9~;8~<}f99-f}mXFr@9gK>SdwF-Wu~?#e^2PktS)iHP{ejfVb|8J@C~ ze@1h3Y7u%Y+`KrcPka%Yu86n7%&7L}Iyyi`V+F(;aU1@zhy_vn5d7x2_ZUCjDb~dK zJnoqwxZ$(s&jUPmD>CZYJ${TH4Ul+2$_1p%ARoy-~`wYBM zu}z<(EI%|i3+X>6aOazEP1761Ys52hBfRBu<4@5*lA-H;V$#=-A7kIvtBXlUZ(lLf z;gh!cJ+t#8^T#YI5uatAxw05pB1nIIL4q{E*~}KisOm zO2WZG{o$F4)u)O|F$z1Twn+?TKdIJO*0(F2R`DE;J&GmAqH8B4a(t_E7t`Pb=wn^< zrC&UeHXkQ(+a8km0tkJ9ScD>ffaZ0avI0VzjfImhcR1w$_jiJGTPd?exZNNxta2@INt}mSYetMgsdeb6N?w!6c&}fL#hs@` z95vRHLsud-_)|OV!fkgMb3(o56QCTkEo$%kb{z&GxWfI%g!_Uw8OiCXN>ga?t?O&Y z0d8$t-88qmotau&aqAhGibNutSJ9v z2t4!$pzDeVtI45OeLcO10?Me7xt>Y7fdI=#BTLy!{h==08}kfOKm-MMnBMY6So92z zgxi#!329F@%gtyOM;8o;#y`J(`vwpgah2V-*Vhp?29`-YC)UTPM~sy3W6vYTN9$f6mQ)+j^WO zDiK!&;?=0P&V(9~%kBq~U2s=p^ScUVI+DX44#b3`Hfp6Ceu2qH z;y$XmVGS3vkNyfV842H z(lRnn&2LbAmy%Mn$n&QPYUWD(uxbvG7(2@>zS%-oO6GS!izRyX#%5gPqbEDQir#N& zxlrCf=wzNc6J-c*guAe?R!#GH(4)M6>#A&`dp1Hjjzrt zaMJ|I@P4x7o^^yZZRs8RvDF}^M(~2OYNAHSc4q$Q+IzB0j<&Y8m6mQ+4m}#uyvT^9 z$4G1Mu$hb=Q-qC{ZcBO#;p#oXNHXHK%LyRy11vd}>Rw)%* zo_pU!dE|;L`>lRylj#X<0Y^}_x|ZeoXaPD|#Ogz*ef*$mrIy6Oh0jqmy75>R z-`IgQ+)@|51liB;kd}^2+rD)Ov+?ZIOTpIrY@TP+Wp=<#I>2WkpT9P4m$xM>f!gB8 zMMhGF1DPJuz)8l&qD(k&<1PRyXqsof` literal 0 HcmV?d00001 diff --git a/design/FY2021/images_cpgfunction/F3_memory_reduction.png b/design/FY2021/images_cpgfunction/F3_memory_reduction.png new file mode 100644 index 0000000000000000000000000000000000000000..de1814a06745aa23bf361c60fcb0d656bb7e4faa GIT binary patch literal 54159 zcmZ_01z43`+pWFmRzResy9HDP>F$v3F6k7c1nKT>q)U+Q5|EPa?hfhvCp>R_|F`#E zhq2&*JLf&Gah+q#6(}boiiV7j41qw<#KnXZAP^WW2n5;{5gL3)v%7^A{D)*Cre+U; z;9xxdh5AkPmlpydfrtz7D}J8YYx2}ooJ@Uq5P=WTz7=Nu{6ffhjDm)hf`A-@B?(d= zSY6D^bIxT@U80nX@1gpsrXZI%_mAm>lQ?$gl-5K&@h%Hm39`npNOx#88-?kNUoh@l zVH5O_)ew52+RI(&Mc!wc%+AIKntpn?OLuNZ-TkTdL+>y8Z8qPO*$>{l8W|mh$wmq8 zs<2*c4(GW>gj8EhN5#d(MMXg)dXs=Z0TiU!vN$^qoOz>(iHR<|6D8!LVJxTyZTnx7FH&goOP3{Fxbbw@M6HesoxV3OU$VzfWdnq~S_*u|nc;M!YwBmASdO z1t0Tx9&Z0id>>g|r^UJ?BL8PAxLT?T3JU7#zgbW{h1$D*sxlN56tt6HU0u1jxR5QY zs;MDC5S|nLoZZipNo{w=f7}JSP?BxipTM0R^GuqUjMCDu0D+`jG^dE)&ISGI<|MlzmKESYE%5 z)Wa=N)=*cMkjLt_ZMj`%wzso`WMpUiWYI)&Hrgz=2fd^R7fv>vC)E1w)3(IQcLj)& z2dFYpu2h6EBLakd+A>DBHHT*E9Y5LH{wOjRdOTWK{%;`!{O%>CrSOnY5pspkpC*c- zyjHuz8@T(n9Y)x>J#Ih$=j!3Q%sQ^Ft_q88^e2+L)jRG^NGHFQPGT=;YPy@Nvlq#J z&C5&tj_O}mk4Q_5kbg3t{bQaC93*>KRaqGxvcA6F)z!5|nUwqY@825h#jt;!c+b|b z@3)*Bis49?zZ*#D+ASXQ+YS{BlJYS^rh=ptj%i!EMn*=4hHEQb!N(u}IllK|Dsfjvb91sMCxbq?E@kU#8<4749HFz+`#@4HRe>xmBc_8h(tKHVF`)7McBq_)VZ zGI?8{H8cnyji)1$Jd?Mf)^4lo@jBSAD88fzZV_@GEHr&tZioA>fQN(QPu#uKxomSg z^l&@$cZm0)hdg}A6qefadNdsZnZL{Mo!d(DxZcjsry&eo%E`=Z+jdw9wGvv~5*p3R z*1^Ht+1(Wu#U~&@P@@)>Ps^G@b(Gp092_iCDH9PDH8|SOFdE4bx3sjB<-I%Y418hB z?j6nCd`Sj*DCr@wcBnKNN_%YpA0}ecUa?neyLu$z|E@j~E02nt90r2xhoRi<0<$Vj!zX_j2(?1u@>!U4nJgTZ))^wiozSgeHcI z5ZTw$m+I-3_lW5cK7)FCi|AdC{?8BdYX$ssR)}&-=#TG4G@D${iprW#u1>j}52h@t z17A>AS67GqK6bg@&gkEm{NemIMkJ@Wb5I+j@B=*OBMOuEy~ zZ~Jh6*QRo^)aGNm`s=Fbts%|U&JO-Nt2rNp=eIXRzToxk+tLx34MRQ~H$z5OmqnT` zqUaaUP9*uYICR1Io?Td2n3(uwJGb;{_uwEbC*8PJb$9gRc)Ovp$PBb@ zZEcmSm%OuHI1Q$O^@@l<56U8PS;HWlUAjBYk(0xEgEdl8RaI3`u>C8PfTQPevr;(J zl337u{xD2L7{nBDne6=5*wEL$FzXW~8bUN5uD7zIsb!UE;JLi!=jXx9@eBopz?6!Q zLU*pkv!$}q{$#D!;}oT#p@EBw>(#UFrY{%j8X9@B>0xoC-XuLXn1x>7xz;2)Omys- zZzvc#B6Xj(e#c~@h7AF?^0=p`r!82(u%Fdxt~MYbAd&4$Jl$fQJwi9QKZXh7gON%W z+s_c4P1eOU{K(rFU{6tm#k!gI_Vy^AQ^}+R82DZ;FUn@(@_O8YU0W?ShZ55nvg>`? zMO({F;#4}toSdJxyK!=~w?D+Ol?wTU2eGMOxLNwTYrEbV7_%q#HK#;U#$axI&>&*oc#Z$3)b<9>H`pdmq7x6Ka_yh^E5&NFZw z*Rp*{vVH0a-m$mbQo77#PX=qF-AoS&iJAt@oZSB9!eLfMv?|efI6_$M3NDQ$P%fLG zr0CiH71(lTX<+m{)kwJejlBF}P9AeIce1sNmg1Ugz0>_|tSomG6OZ=Scek&F?ZR-h zwt5*TKF=h45a<^e^VMJLWJX0p>-s8i3+Is%ArC(}-?TT`62}x>s%-MBOIRksW_K|4 zw;z1;cp>i>CmF+>q9sX{qS(7h?QWSwvb^VIggh1%>Ve2}w{!N9FQ_1V(vp2@$)UC! z$}zEOHC8VZy{{zW2LnT3y;gPH^cp`$-ge0(MzPvcHJ1qXgu1FNYeC0Rbru%{6gp#jy))hKP?m6udf@V~u(~dHb)}g(JDU zx;BS!(s>tu82JBW2WkRB(cOu9imB;L3C4aL%PS0U>XkNL0|MG`$d1c=H!R#dHYjBB zIAIp`94y6O=}meoZ6O~i?_b2-e?CQPUGDIQ(jCo}=Eud=EKtny2_S*-P3Pxnbghzd zmge&0fgFALc&=4T62y>jcKcV@^Wnmfts>fpH@|H0teba6?R?23-H2I}MxJfjR^Tm( zB#r#!xc}|KW_Yo3me57QKw3LJ**zG_xVn`QBo(OcTY{GXQU-qY<-|ZOdfuH)`d5cu zXE)NEwof~g$fJzv`{$E0^eHPVLwm{ORjM+W45pZwn)dwuEg>%MvIR9|F+M&Hw3qwY zxJuW#*MPE?)++)6db)>&FNaXAC+mIjj2g-xKaP%##iXPZD-=Ll^?$Ixc{4sa3Df$+ zpnGR$2SUQe_5wmjPk&X&5k?*r8M7&i<5XK(dH}Q@*tVLQnz*>Q73Bt~ucexG3R+rd z5L;VYGc&WCoB?Alhb?s4jv~I1`tm>zoiBtc%F0eaQss>TNmZlU20iSB`qLVbHHn0+ zrw#>&N-CY{Z%oCTPmrZQg-YpMsfP&z6f_|=MHPQ|6crS~b6c4!ts$>my&LWsGvRAU z@32~;VHKM%cOHm-;OHheT=DKdF7d2ESN?ay}4T#LpF6$%QAimpblXkv35n~^e^eok8i z<1jmM5Z)F0o}PkI>>;&wfyOUFZ1`{(9nPb!tRfnAR1rex;yA%@Q8Q<|ikU0Rd^a;0 zV5>ShLWaR$Grczr(KS))8!y7_ig`f|&DSxIY!yob*$N{(#u60G7y28=@ZN44bJfO8 zPKUWf2!oeI5ARv9niyi9Es(sf7#?G zcKpg|i?_*d@X09#y$?tCdyo1H- z+OfGC&R&!CJhwQMEz1_2B<9}PlO)^0ru&eR;Ie+E5ZsTo`fC^q4kE4NPASh|_6rqq z*<)^WxJbPAvLx}{bBM<%;kxKcOr^wEuj5GB%2Mncx;K>+m0D!e@=S(_g(rK8%=G zH9jjXDH-VLVM&w%qLKokvZ5kHWVKYQp`x<#>f`z0qO2m$>waVulzk^I=~19JNBiz;i(XyuV zq@!&=Gy;fM`jND_c&6EmRgDHi+(I%(H-=U%P>Ob-1A>A8%|>0knP`%Rm-Vbny=rHf znZVZ@H^MNP4MNE4K?q^TM`K1tfGM!-jiPKf@&=PU<@25u;VXB+2>C!JQvTD6fli@i z$=hf88XCzo-_R!K@A_nU!97&O;^o>?@9(YP<^F;VF%3$3W0@pml7Dp6zTOAL=Xt(2 z5lt=p<9XeIRbE4vFWJ=1h2&tm3E?wvVnzc;~?z&r2V`pav9&OQyhw-tosePwVk?21)$%onDJkVb2eK&Vw z*)&1@5;sj9NO#T{R#W9|hOAX*?!gi6j8@bx?p|Y55^*JRG+Y0=%%PRXApi@P!MQA4Da<0ZWJX;-!PTfWt-19cM{#N{u`{mDt0=rj<(K2 zl)7~LgyOq*f0Jc$c}@bhiU%IfbPPSkn>yOw<)!j^Ivj|`$v`nsy!Og$5Q-yB0s^rc zeI&6CoRAiKS1Vlm`m!)jJCxMg+vI+RbpD=3 zr1mx501%G`&txCspoYGiR!)BlUIPHe{=|z{OkNR`dt2M%HXnqR<4|iv=akb0lPdM> z;m>?X>o)1ZjsxI=S^cNj=fapAja@aw zd}nNiQ3KLCG2yg-zQ5HQ&8g(@R?EN zbij*h=Ri#H1wLS2_|zU{B5n_8-s=(5PqT-}9)g!J|G!AuIK$eF$y zX|F6Xo3^pBv5=4uZ*TAN^754x{iIy?WB6-Mpj2OAV=K$a$tf#~>{ytYEe8Vy4-XkL zcN7uvlgXo}rM1=20F&&r$=ymYGZht8Tue-7XD4_%Abfgyq$273=$%}I2pTos>S*uZ z|H0YG#EC{abrqBL4MDY0lVn3(fE_AN@I~Bd*Im6n+toQF=aXh~#l^Bw?H7QsBo9Dz z+3~Mh#{fepou2({Emgbvjhc(AAv^oAZsnT!>HSn^7$_;l(x?ytG)l0uLqY`~6x@WxmPkT}44sm$)y_)DREdsfCE)7nYJ>GT zMP_rN8Lo%IE8WJe`Ff{_x#iXlG|$7s!}sPt8X9gE3cQih8xbH!sn1~wP%PgjtCwt) z4I5Wh)znCuax^!7l4-o>4f#?@z?CXfL7Zrq{^A*HUj>_SzfCg7l>QGPQPIDFhpKS^ zu;#oETMbyIMy4F7C?4Hj3l6+?+n;%J;^Bg~nX5`sd@g-zf5ancZcs_C%6961zn#eTvxp--IOYUxY+=d*qSSe4J;B zDK3I>8KruDb^P{Hevu~u!r^qGHg)&ya_P3J{2b!3Cps6(L_r~xqu~b0JG$XNSweEu zgVbKkId_zK;$P3b^Y3CTo; zhr=K|6ASoIaU13-F?#wH20`2XOo>nsmj4%_qvOIPmh0{N1sKoz=aRp^3?=N~{HuCI z{s>ek00VQuQ=E#L0M&15Y6^f^B>DW&ZB{n6RGH21lJU7xiI}fm?VG_pvX!mPRA+d{ z>*+7$2Zzu#wfRG=Ya91I(2p~G&(yapn)h7h>kPxI^f@#V#(J>)5ySnW#Awy<1H5ph zHOp9kE4~)q40XrjWHI2b>OS!}*kgKz-PPX$q&lI`4dV}GbC?G&MG?~j4i|6{JaHcp zUxiJ%#e66MPs*j&syg8N9UbHcb9mA)c73DknsqDwGrV_LHtui#VT4E`xl(mdHfx>_ zll~v;N=pg$_ea~?sTP{JX8b7Weh;*581s(J{qZ8^gmnbxgNW$>=7ZM7ho+&tMWC|y zE%*1PD~a7thvnb%=;`U*rQKlNY4rb&y=)Vd!+G=@_`BPEQ2;~S`nwH=GL|s+#hiT> z%|$kLi$zOtmwOxH6Ow?%GcWakZc%3{${E@gA02Ho#8;$FcST;Lp*ULdSo5$UG}(m_ zg#on5LE#MMAs*|;=i>}lvC*aXIy8V+qN2Y|mYzolgMYl@6zezl%(VJ_ar|dTtx#_9 z=wrSe;S%2~&c9~A7UD=9jEaf^=K#3d`>$0nS4AWgk1yCoa#+0e&N9CMn!_%#EE12Z zaA}%Y67tNLFF(~oh&YBz<=gqqdYE=dp&h5b%+7}X%$BY8t+88`4yZ^NBMzY5LZQs=y zn?(^>sRDG+@)o?G`FBJ6yhgdw$rKj0M8P(NxBH!5ZKWk5e|b;R0XImF00+kZApo2x zDz2ADdXOo=By9o(kYXf^)YR~j0>Fky;xb`<*ioP$88^eC=`b{LDz(o@ke4<_Wml7Z zP|3G&p7DZC?t|;;R*rad^UeIBpPwImGYqv%N-aRxj*f32on2kuzkdhSjus0!&WN6# z9*Qq-RE;5y&18_Bo!!vL=zC7i=fdB{l}#55Um)dfS0_e&aX=h`o*5{mB1O93u92}Z zMID{zknXToMMXt0Uj6-2ko@A}kkC-daXUg$3UXT6K{$w>3Fxf{Q@K>Mv{GYZh3n}a zvMk=tW(xX?;b}uhP54S97Oij`)Bq}{rS;h7tgftJGpHV!KYZtf*(U(Q?dvzAt>=~H z7x2*KFG*BHdYvIDm$@VzGS63_CXdZNeQI#nR*;koCf3;A*_ml@E(1km_YAZoP|YI# z9g%5i2iFCF&T?~e-)h#n>p{`ALYkS+4OSsSac-Xuu z3t8O7u2jgxGzpZZC{KBKXZkkrUAGrmyGcCm_4#g~u!xHKmxuH>*Rf&|7rqt~etf}& z3I1uXumJDu?d|`Z6`%VM-5I{~tRlYre(mr463=rDyLa&vx*t3{baZs8ab4lNJx#+` z!jsD%PFU>EcE*$6Sx0pPXASwt!p_yTzSj`iYihM+OW_3IsY`A{*wdnSaEG+-Pm z(!nAH^Z`pJd0#~^HX|#3C^#ow7S)moaDvgjhM3r_PR5wEmLJ=4UGInTeU_31mJ3BaxO@gjDsg+$zS@{Z zEK)8RF=ZFYw!Ju*&%)2S@gouYv+QJ0!K;i-)`R%i)vfJ}7d;bg`Of(DCu%q`?=uK~ zsK_>B;r*wwCMvgjy$*N?iH0i@8X6j;+-jav?}tg6i!n`ollgd|Kkp|jWPQFU%R!iW znk&ImP`gp$k-(si|3DIqvku@zgszYwUY1$_P<(zU9s<$+s(xS1QrXv>coaLMC?&ti zUS;?&^hF_ni4_CFptHcdNFyvSxa>|vK~Wnt2Jq=u?U-BA>xr6X$`AQ+G(9uj0MkI7 zJ#F~sWB&jqYgUFws|Vbk(ry@Iyr`=QlXUlqPvXPuG%YlLw)f^gh*D`7=<5qnpO6;A z_}vCa>CE0<{T1#GWTj%F^(G-FT3{8bL$sk~ zgnM;d#Uz4GnE(>7j*N`F6c>q*Z~v(}rGPf7Z%YBD_#T!ez}{hhvCdUn5N9uwr6k!N1^r)z6=ZlB}rN%0-Txz-s+iG>+Qw zhyAYU(0!P$z7Z04B!NkPu={%(`Ydq;^8;begsy%{#xT~!!u8(ed*$d_2D-6|(WR$G* zYY+asQDB9^vfAc9VR9t!I{M7{KT4$h+hTbnrW{=OoPqX*SzSJTJYy4+p`hkZv6_ArT7`q8#O@=u3rZQ$CJ9qCJP%uipsod<}5=b`@P#``DBAF3*eVE zR`bZ-ir+gB`zmhXB}$aJu;XI0nf62t3|7IdeV-%U^BM&S%%VZz;lw*3y=6f2pY2Xs zzMZZ5G?|s1-KO<(6I{%+G>{dTSXf}P)ySlwV`jDn;#UZ?sjxtgcgEAw(akh{si7GN z6v!ko&gs$*3=5-r^?9ct&*^YsV2E`n;kC~Y4J8E|+o*X}P#mgj6U}#pL?DJzQwMA{ zB_&ZJL#&eJXs~#%dS&hG?Ln|(pkfacV$E(7K6gR%HBvK|3Ww#}s@fCljJ3_pp+vU4 zA3q*FPN#-cy#!EawOTy5FTV^{FuaBns*lK`sku`0$Erx8TSv8|S;)c-JC+APFW{ zAjV3Rfk7VJMbW>QY(nhu&8aem2 zx9h-kZ)&=AdAur237q{`L_|Caex(%^c0+00pxlLpgsgdHLbpy`h5C{NFn_K>9cDP~ zX7(&m1`PIMe|EDFMK*=QcKvtk;=}#Lj~_pJdwW6anpA1;BnBqkR=T z!1e|P2I6;w-H}`^UHWF`<`Ob8*ms|)LZftX{=xL%?2&!jjVrY>NjR7H9TU!fe0E*Ap zm`Yf_os$u}Pv_x!5)9+DHNy%b@p|4YrvGTDV$C{Ma&n(qQklHJr`x05;lw(1c9dqX zGl=7fI$=szM*(u#9{CP@XWr@&uvJhahleq~k+%+o{-*7}DgMANt)iq9>Mh&;kGtDY zU+{7j@={609rPDSsMYAg?|TWJ@PQDm8&b~`^ZDaI)xHt`W@8w@}K3c2-dmw zu9ezVcF*a1&8zCI6r^NjH(QjhDti!n9{lXMOr$xtt7V#_NA4mIqB@psv|>BM(GM$eer2qI&T?!nJP0mJgirx4w!hqapkzVrIppM0X#%jeO>GJmJhV_=KYyc zTPM5s%7JG3>?n6o-L{5$Lvf56LatqgtY+*<;KYI9)BJaki@E7k0mOJjvX^UE>5|MMUvYc$@ZnQJ$LJ-5g+24r6H`Ri?>W|BD{;qq`O>AnR5%zL1VTpM z`FHTg#Pu554fqvK?pCRnJVO3~CpI4~W|&6vk6>z7R#J*g3&8FM-bJX0VDpI+;BXoD zM_T&0KsADK*)U%^x{b}Ko-=IPBNsgiym_J?SrnNgQX-7O!afFB`It#ccSLVk^2vMX zSI^Wa%TjzBEcv)MYH}%+@m78d=sd=DTx@PHkAZbnQEoC`sN@C}+$B&AudJf-5$kL_ zFa7-VbZui}FpYcGswSP&G3d6OnVFedAg``2QK~U6Hnv2wZlvbBG63m0^8OudWhAm< zVoR~a!ty{pk%u>VBu6wuK^AKOc~>e@1~Yi}a4j(D_0XP@TusDzpuv5j@hz+iC2Z}? z_UZ4e(`E+|cPUh;TX?fZpA+*G51UF+y?7$~w>=TUYHyPnbr(yvp87vXrOv?s#$GJ# z>!a>;gm$1QY|x8Ue+ZNS*JCFog-{A`E0V_gJF&x+k-$h^B3K)Ze+gArd zInculgoL200yhRkF+7fzpwx2Xl9GNGac5ycc|D)$e?d&ljJhNh)YaJ;5Y+S-#ySle zJKw7ux?aZ@w=9)+6yGGk$MT4_JCM~m z^?74OsFwbzW|oVS6YR}a3VjjuX4`vvYxWA#2`o_Eut>;js(ylc*qFIp?K=tyRan~7 zKGwdTq?sj9x}c1&^;BEV_Dll1Hc_r0gS(r)1*FBuE*Y>n)9G`lpVrO{kkZRcjK zTNvqny;BDfMT?kKj(b{R`Of*%93B_34A5{~Hror2qK2SrUVxWxT;iLrb$$p8lPPx z@F|bPQ_ftQ#?99Se-x`#g7|5bdTLr481h$&WMn{IiV@+A)PRoraHc9)w(8|2drsQ~ z@tE&dudIAP!Def$YpScSCGE=@isWcMd?ym{dlj|{%_QlT1`5qsx@IhUiQHWH0G z=fee^`;+)?qAz>B)ajkCpUYS+G+qJhuzz@eaHwgodUHQOfX`{WBE%ke2d8$+u8?ac z0?{Je3$nAuJGdP|(3+Z_cDX*?-c0)<6T_rGW>7gjX8s1!bCop?g$soW-{FOcTx*F1 zG44-TPI`zpQmc3)ghesE?U44~{+Ds3k+Crly`#~Ul*QnC6%2r&XrTZvacj$Tq-?b zW?qA3@M(027pT5Jts;@GeVdC!TBbe#DGK^v7%$q7xl8%*)cgDU;LG5E06v4~NbGyk zYI@Khz;^oCvwFApaRR3xiNy6W{BFk$M#pA{T++|Y~+VLvcZT9iWKN8di779FwsQ*-D%$4AIikG8{t%V&(kIoYQsw(eZ7Ar z9`WAB?R#gRu=*+C+*E3<@HAhNU-mutoHq~Tn1a3@7ZTqD^0%<`$3>)M=D0IWlg}XN zgWtfG_L>B9d9`QAE zL6>P8d1Z6{>H-e5>TWAX%N>w%v(a2VEMH8T*x!-)LvlFqaD?5C4h~BkB#0}#yGv9l zZljWk1$2k?a{5=gLRg(;@Q?&-l)sbp@qT|NL8lJ!S|bdvM;~aHKe9DYi(-`i|MNp6 z`v?$51=1L#&FiW5EzQlZcXb{>GWh?U$zi~j{P^*syE`|8kjtsKwDcJS)Gpc}GVW!7 z2*dJ4?2zAUjV;&LgK9LWq@rSCWCZwSol@fefi)we-X`z=$rQ{%ErS-{ZKo1iyadUD zH*em6TnZDDRhu_lI~1O2_qUGq0&2S%VJ%$E5BHa=z-fLECW`*x12l$@{zF!q>sQ#o zb)Qa#J;seTR;j*S_ob(YMG>^2V48jo2@MT(ZAv3#f5dMacNdG)BY^)as?%j3((f{aPioJ5}vKsXgy4b?{`Xt^|MMTJd(=^zO6|?*8GawNU zx`CdAp2Gyhz_cDS=czW1H;;lTa7UO|k=2ut;kNWf)vr-a4DjRK?M?=YS~N}{4FV>he{wI&v^pSHB}(iL3js?IAgvX9S;D~qxxhA2E_%buyJ*OY6ssmLA31JL zssE#}@HenIOD6Exj4vnjKthNUMC$gPZo*!139SUOrzR)&u}F!EB6P5T;k-89-``(Y zNT;2E859&WSKVbxh+)DmO@Imy*ll{c?CAk8g%D<5V_RaLY|z4P4j5=LID zvzCs5fdV627j2g2N}__m*x>MGpcFAjxkn$wcw-@gchN z^&Z3Y=`I_Iq&r82N=295$iCi)Y89h+kg$KK=db870`#0}FcLGrrE)nlx$t2#|K6MW zQx*J}GB=D_8E2(98T>fGb$==c=XMcZ{{1?0&;AXHmWIP@{h!CfVXrXkjq>1yz1BxQ z5t#(y8V+)EzE3W@ySqC(GqW+2o*ogw7Z&k;9pvs%b~9-tvWHQH{jZ#S1*%I*LJx@` zm!B$P$3&9Q&3Rax7N(|XS_}jynVx1+`5=|Ph#syQwl1<7W<{vcRZ_<1SsQJqpWhj0 z<_p&t>y;_=e9d>^*^3Y%G0&}z770}Lo#5lkB%nzMx1T;CoTS+0WCD4lhey`aB|_x$!v5OjLP;6L}3h+bAVz23mHnc zE9j*_d}Q$&m)%-`hCA(x=-y7-6geu)A&ZQN!3YnD_(+}Ptr68YO84N95W=;kFFU<* zHUHe+wWWzG`*{wspQuI$Mk44q|6!>d3Xd!`PFzgvk10DqQJu4WJw52n^_H`5A-cG^ z85v&vZ!{RNhn=wB&*bOk`h5Ulwf&mK^L;xplRc*|fLh9#+7Wj1u21r~&QD1~gNOt9 zzj71O?Wd^lvfOw(PERcyDuN2{*G0?Id*>6;KW9fM{Y^$&^|qi&Kd3?}Vihx;Jyy+k zxPJ0Qze&M2-n(llfxvBRvSd-Hs`}ze~NiiJh+Qb8*j;~+826@h9 zo1Qu6_Mwk?xw*!MhFDL4_9YI^DEofL`cWzXNJuCs^fWXe)l2SJ{qyHh7f`7%5c4Wd za&p1E(b$-n+7d=m(yvldU<3alFjZ0L1seg#2pIr-}!F5D|N zCz1%|wS9c#Gt?yef;_;a-`L*VoDNAU8^kQ#{tkfkLvGw~ zfad}5huUFUHItP^Rv|X%`1czd8!T1=r0xWs`f=4y z%tw%t2n-1+E-qff;Rg`6q2#gT0oDa-k{)hWe)8IYvq^1&J@f9| zp&O0%+C@X?^Y6<$3V(_TP^{7Gh9B+UW}|l8x+#mdRqALEyJj&5hzP`|jYMK~*C3di zd(G_2mxgbCt9=j0;hv3-yK#90uhQqIwjHAGdMIuPUB(~fFqgqER9TD?Mr=t32ln*K zFeA0^3gV}P-L{=aFY00$`oJ(~PRJjcEY%7Q3v=R@oLc^uME;&fb*T^JZ8ViMX48&f zcjLK`OoLAyFxG084&}j38UlWvpy81N5e0zUjX))So;k^uUAxhIxkq~F{z>! zN4evV<7T*tAU!G&SOzMnxRh{Isrj)Obi7nvLmOKv2pI` zcB8}gNDS>qK+7qyG02{3u6GESiX`glZIaZLUlTJ3_H%@wzQ7XhovtP`2_(`X6_)33 zz>5=m~QEMGx@wEPj*&opct3!xv z@m51j3=wJ19~f`whEf6yf=HKp_G1QCcB6Ha=A}2QMGsAjAK6hNU&HrR09*4hKSU~F zXLn*__A%=8`C=r$_g}CeC&Z?{c1vR%0_kt^M_5&MO_9Kci@{#c{oDE5)0f_)i$HFp z%y-H#q2*i;4IHCE$iJNagCqDihH7y+lb|;< zPB^SC)|3|P+^%yv$*A_V2ey@ludqn}r{b0K1@#;Fz2HYun*hR~MsdIUjEEmyA~gN~a`5i2ax43>xzCwjdHuc$D!S z2*0c>ZzDCrAd>IftH39(13r0e*;>uw@#o(+!9>^%lgCGwj09nKURDkuTV}A|4m^QF zmB0{Q)@d^qT=?nyj%6`fS?9ZYAc@ z@?QP~B;!v`dRG5UoBJxjgi%Kmlm!&MDAAXF5f_)bc`F33jSXW~+LZYC8uM{Fkl19$ z2dIXzn7`g28}k&j?|9D&sFxp(JulQCUN*mnZ1cD?wf8nMX79RakRx^1fwu(77&iTU zw-z|5{YCon^pnmFKIHbiriD@#u+&F2?AXpa^xHif+zokF$Nv-!X<8g_x1icE@#{=Ztcz8+l-}q zeZw4nr?0hx{`FUl%xYwAg~O(<5BwmdkpHxVJC6AIkrnw)0IUNCIsv4azzg`@d82{O$goIc2bFft z=#^pGeJt-qm9Kh-55kMEzwpLeFv;}oI-~nwkddR2&7=W&d&t{xJ@idPSU3`4O)|Rw z=daFy(W|2_@4s1VlR)@F_{i*DWIHghAv@}so9}?o-6TMRV7^aDNdZfG~n7S#o$yF(#*pM_bnpomV^H4RY^u{4W7Yk5?(>LbRLb?y?% zlaa2V$D?BCuuZX93{=r{#{)+Hym0T=ez;&JG>C+xq>E!fy^W&Y+hj179><03v3H3pF|?Md_@FvE;(iev^9`Q+LA%oBX{e} zp0*{MASl`HlkRzc<;7|6uWZ5MMh37oZ@>FnF^3wdjnn{GJLcO4-ECXVlOtKd@acyt zAmK789Lv}C--L$OJTHKha=ih=%($`E6Q}jxS%3{@c$@~A5!S|{gp>836c_qoh!SRM zv?)Ak*eDMUc=y|^NxnT|hL5RY-I`fR35jn>A*&qAA0LCr&mSy5fgG~uuUVQLB0oK3 zWaOTdOv9gh&bBw9J0tm zuW+)fxJfguur^0tq06{L#tMM=6v4*p45E%2M7+R!#X?}LLG#JT$f&BS^7HcpRu29J z??H^JoT7u}zbWPUsVNS}oiVWPJ#c4u_Y4CABQP)!z~*QEK8tB&Xu|RowYuPpL`2}X z`G7t5iQSA>8|$i6rDz-v505joG$*Ii(h_744D~s|?64;o5Euvw4{60`EQAa^@%r-m zdiC5IO%)YUZ|X>!m^Bz+%*{97Q8-abj{*zhX<*=t#AVJ_xu^Zq5>hNv+oGfKRTzmb z#@c_wAAup`20==1#8U^Qd&9mpM!{xi4L7hE85ud*DZOa4yf(NA`9i3QHOWo|%bI5aQFOM1wru%2g@!q*Q zr*FW@mX*)%0G*qGx!gc~vnB$x`OsAys==Cm<+N+dm?7506RY-V83PtNNbaixwT+m? z&AX={``l$_VIeId;s_X!+6_(v`=Mb&R(lSdy&i@|E-kZ*YX+dSvO2{Y#`z2HWl1fDDANJotQxk1yPQoS&bIM^goApJf_AdF5Ed zC2XGuTOkEq`rZcdZ6~k-v*=cpGz%Sqg)9tiC+LbhSRT_N517o4YwLB|b%C5xm1(9` z!3sk;tE~=K^%Jnf!GJq1YEvICo0@?+`c3rCzODDYzE*%SDXXh|fueS_rx?~!rR$M* z_}=Yf`r#0yAGBLmw4Mb0tBglMA7wA!ZqEH?c~%nhmHl6~^ZUD_qa%!qA9&)nC4tny z{}ej>4N@_z=5PN%*`pt*>60ODGZ}H27ZyE2HQzruLGp-L|;Ed?^79^`o$XR ze|$s+rc~Bhd1<Zh*vDtE8dZNkQDT-vbbB}{KNL@@R(W6l1#wrl$CWJU zp_j+rZZZ4QWGUQI%S*hqP^+1rW(JlYTZ(3mB*15il)FV)15pIHs9wu3z<*&vMa=;G ze3a#0x98SzJBD)s6-}5k*)y=!R#QwYU^gKcvvscNi^|5kvl5M(Z&`mViQULqz~^%0 zpa4PGI-j;)UHukv^qvkAl9Q8Ddmu+zZrA)x{FI`_DSKm)smPR}gedt0NLSuml#!Cn~~1G$*-@yKpK z+F`iwM!rh|=W`{!RHKH8CHaF`O44%uaOC?%aP(lg#@`)FV4*edZ#H>8@Th%;hX5)7 zI=)-1@*~PL;PY97ZX64FWnw~kQa_$qN4wf0adJZI?^)nAi7c?KobCahwUc|yNQa6@sYs`U zw4i`=C>;XQ@m}Niobx+xto6b1?>*uL&TJ$2=+GvRZSg%Am~ILP z$ccr2Uchek-n)O}whVCRU}(c059>?FyngwUA&70<*0Kn<{p8J-wSf5reHEghxcEgM z7k>|oCK6l1dPO<#=Otyhs8a;h`IWR#sG#>X@sYGr0yMwNo(Hg_D<%0t-9IJYdyda- zJBIEdsQE#LfMtremds8RrwE4>a+h3~+!%(gnzXP-7p8W`=a_%xz zy^a)#e=KN2(1Vf+&U=Y8&oM#ohwuOZ83tjA;Fk@c!a#9926(?M(Wy@e|JJuJ(X=;> zU_P>(5@Z~f#n=MlYw#x-fAMc%5^A)3Z(5zNDZ!c^7b)oY=n*>~uiCQ0vWcaYI77zE zm!VwHEJ4>b?rmyns%c#Yz+Hgdi+U(bj%Bwu!&slj+`MmoO7TJB?@65>iSDNw9-dwy z_IF0hIJIcvo3aXNz1?A0*){6@n{-U}Y%Qe(SD(;O;@Z+{P+y^V%LOBwP$O=Et%5#9 z=~fXn{3w$rfYy9}^!z08G4YN3=)Amop=7a)vvaM^JLsX;nR_-2N@K)<+4w}9A<|B2(Dq&a06!;iWkm!8 z?j+s0`;(tD6Ywf&X`}uQTbPHppk{DKV`k-{9UE_LBf zVDSYZ@F7vC!0`=o+pYu~`r3={M`}&Q@4F);v%-!ZTji^-<3#Z*0S1!LVaea7^N@y8 z*&veVE6HSsJI_n261yZZbw1$1R9e&5d^6u-X*)nHrh5Cwl zvzj1cFVKX{T{M3loHh)5Z=Wj;=-mpb)CsAt9}9ZHh8l(eLy08Nc>F=Ebrs4Pj$gpu zTjgOOGin$n+SjVe>}3lRedg!VJup_8UXw70%2RGBq!lvvUFv6&*82_`nLPnRN%T-7 z7#E)LTd3}t!|&Va7~b9mm{i`@EAed7c&$8){}KoZ^P;#qc(L)owIW zLz&x+egDoi%wG@t8%S=>8e4_L#rKl$o6|F8n~WgmX9%HAk|OOg%>AftAVTS)-DuaP zbsok$n~^;DrXe>UplIS|$S{o~@?#u9NeQYNdwY6vpj<$pO)rmqc*`H&bXgs9Pc} zZQrrah%%a2m4E#BHyZ7Ck*^w0dc`X(bAuM{K}=j+%0n9evo(vWRka}PN=fr4AtRfq zeO|(7Mh7%dSK6m}f{eJrcS%IUK0i0w?>n~%JTD=>f6iwry(X)}yh#mJ)FOBqJf_*X zfdmmQl0YV9UR>K(A^Z$JN*%mHUG6y6iGe)fpGt(Kq3w6w zYwLV(X-JPv!TP{vr#u7MO3WgQr~B~9lVti5N2N+SorBKLOMcwB*_x>}rE9`ON!Q zdwAcNW0o;(B!}hovWMV7spg{KoouC@j>N7$lBrkJN$S1jPkn1H(`oSAjd;hZaiE4P zD~FnwE!|?fGt6myqs2k|_u?Y7g;1OWQeYF#s?dIabTQ>a7S$=8M{10xd_Ak$a-cn1 z@Mop#XXIgbz7#sA*Uy8~D+k%XYjP9~*9o$F@wPrFr(bfoy zjVV#n+O|984q7R5*4;w9vLC0uv9zT;8rzo!W(a!_c#{ZMmBV>iRP_66GT&9LtVJ@7 zG$i&1;?DPin(FHG%MaTX82#E5=4mOn+k#?*ftSW!g_`{#n#oKibgri1K=^%e&yyOi z75nd6C@n$PKl5Lp0@)dROs??nI)GXyu|A=%h3V=oW~(oCx{5C(0rg6V9TX)lFyht} zR9wutUxf-x`RDYbA`0EG$v?xCe+ESE)<(w ze-So5wuzeNQef-^!B_pCj>|CZmiP8P1~M9W@&W6K2kH}$Sv~p~PRd02^qvYi=Vw-p z{%b6ep+;)6!`a*RY(nB%4LD4s2ZWRgsG#OzIW{dUM;b~&^ZL)w6UED&8+vYoYXZ=- zavxq77f%8m{0Q3X&g^n@ZFzmY>1V7_TT!58Z;e;sfqeoLgOB4FrCWt#h8#?5Tnq0O zw*@JoX20C)_PRzy@R%&{_hsP|JD%J~SboHD9{xKo;$u715i_Sl@t96Dl(InX`1tr_ zL*fLA>N2ysl!OEkAbR=6cZfFZc5%{K0ZX-Q4%EyHvGxEX5Vrv-fAq(X-@kr+|M3Hg zb=oi<)*HdBSk2FOMoWykc1J`dZ$u-X4WnM6^YwicJnQnFKL zRt~z{#1{`~G>H1jwy+Rt-b}D9&^}C?Ekb&V~DRBHksBZ!F7eV)>I{^rFwf}W(JMJ0&q}exi+ghn}$=M5o z>|M=LE||Z2gBo#=%oP&~`9Y;7XV#1w8R%tB5nqcc2fg6ANJ0n@uoK5ZBI5B7I_$58 zLjS2R;Tg5fH^PGv`T5F@WnXK!90^OC!Yx#gRYh~uCDwqf($93K6^1=UkcguBE)5KF z4k=q08{y9~Qu%P1Lh;32LI|DsQoKaUmsaA#r$qYZ-;t?QiH<%y3?6z&S zb4ULrQC7p0vn`hM?Z+&>vkOP=H`XT!Zry5pbw_8!qjC4u?$Usd06i7e?#RW!^ilY= z=S5x@B`Ob!S`1#{wiR+4KEAgo^~(J=QLjlUMkk9?`p>v3g7>3e^k+yhtxWAlZ{3jp zOyf1>^@cC!!{fUC>xt_?m}UAx;~ae>l;-gSm*txKyU4stFX9X{Y>F9HPdYh@!(U{F zGUCsjo0_N2Ep|VD29F?N%8Pz>D$wfS2v+ABQ)rL?U3h0t7g(sYHDZrd>UK4ZDe4B>zyIzc zg=Qs#a>DfEqf1ms#_g>@?`5-qJqi*bg<2H)8Y&p}Rnh?AU%zf___7R#UmkC}3bj7F z_mIKX`mm<1aqyR`S*;sEGgJ+iqUwOPx`G)A)IyLrB&x*11?V!iWtu?eQdudQdz{KJ zhHHz#CivWUpFdVY#JaW|CLRF2gOZ@G&cqYl_@ChbaVYrN3Bhq1`t93+DF&S=aQz8Q zq%6r$oO<}hug9p&Goqhg5K99Wlh-T04PA!Tq!*{3qs!G02@oRPz{*B4H3x8?!TunJ(dQJ1;$YY3vO4kxJgCCv@b-Ev&RqdLamyopAqcLhn1L{B zX(|Ou?42uA1G@=ti3k*F1RJCEe6SwXSgS_xF$`E-a_G4=V3n9C%V$mOy7ogrf^2_c z<{XUgjsW$c=uEQ3!fCT6kCOgiPZH=cBOn#U4$&R zwc|Gsy}%X=CqB&YtvK+KuvR#9Xx67XnX9z2Nt@@KJw11Y%goz*rnzNbK8Lz3D9@|xqS?6cmd7w$Dk<}v|HU_pMTAh4_@VmUgEVg@WCU>au%|8qAH7w_- z1*y1&E1`2qf)p#={vW-!19#a;FnXurEd3+QhH}`%toq7irDA2&chykt3 z21LxYM&|szb+-&qv>w8B2M<3If3ctIbi)3*k*RG(2r2;MD6bvh;hM&zv3mto& z2W|Nu1hqtMf&9{Kp+l9l^Ka8BPV@T7W`@gFgUeHb*H1>vZhvamNnG6?MS2WePSz+M zP7qQfubdsi-=hm#9~C85{8nQU2H)ID#lm6+C~LAe$2j?c8xCyP`6EFRZT1J6>Nm78 z#iJ!w{~kjz!g=LN+z8$8rlX=aB|UyV8!xZb2C3RLdWt-sk^PgB_jr_8>*l4|IEM4R z;-$4UA5ZhjSJ82TqYmF#Jmp3FWhN+_9;cr4e2k_k&60F_q8L0FgRg2P)d6N7Ux|e3 zzW7VBl zGhBorb=-Zm=)K^+)4tz9VeiXZR|7m(TERY{#d}I-AX_fLA3K1o`x+zREij9GMWc-H z5T<>KtsV6!QP+6%t~K^?m(g;YD6#RkNC72LqDbKdI5O{-nd5ACTy}Dtu9rrw`;ms{ z7Zs^c>Q&+x@?~rc5hUHV))0RQ(>~~i${BhawB$B?-n&obDj_~Z(`*2p(M0;8E^PW2 zJ9FW7;0xn%>iL7re^z#Z|M#!q;&kpLPtoY4xPVNB=+8?rOE|$_rHkzggtW zy;etxY&x%3R;?_O37LI{89M?u#H#tHp%8wMnQ>{#rKA{6;JI!WaVAJ&i^x+m8)B_> z2QoaUwyJ&Y5Y&#p=Ro7_$Yal)xT)IIdQikJnB1$#Uf+1n&kstD^@`yY84!?fxz*TL z8Byxx;SkzI^SEF<$FkADY;p=?`?NMH;C=&fg%E?gBi`a*HpI!acI098=*7Ura{Y-c zY1Y6Ef5xmUML5qt|L`fK38K9iy?870{44l$y|*a-6*kxAi2y4HL{xNiettg4tA`vw zgZv?qcK|Ae{{Fl9Luqi1u&`K|nu6)xXsL;6n`(?1CN&b#=v;LPX&UBFl)E!BQB zLqN#cAT0luQkf309D1PSo&x!5Y;<&KI%eqz5KoY3gSG~}RbQILj@gM&#M36!#1=id z#n$Np9|3fJg1o$h4|JAiy+$QxJbmAm)?G%>G70>_$@W@iL#WCfwX0U&SsK97=1Z1s z&$d4GY%0N(JKD`24@p0mzP%K1n1cXEa?m<~bQ8YMl-S5;8(}(_4s*+-)75Eed-fmA z-_;jq_Wat`bMYIW^hx{62J=3}+`>9LwcRG=b~Xtj`puNw|B%T`%A02cZ^|cTKE%I` z+mQt9!PeGRor*9G#?{<_^@@$2lo~CPtID?n++~HIzn#fSe`LzQ!s1pR5wbJ`Ch)Dl z?GZ28_LRB`i7h$i9Sx1hbU$EG*dh?{(`-I*N*|V1%N=xthh=N;;GL3%mXf zm6n#gyNUdb#CM{5_B6PL*PZ$)ZJGBtR0yoc64H=pZ^_WmP}uynHPxM8Cu`8~n|iJ4 z|9Yd>->Rs6QKp8h&ETZ_p{2ND0`NGt_wdT((eA0cdGts`Heen!md6j^z8UuQ`o zdU(?MdU|evTpSgK`jTYZl@eA?<$38f^%E-*E}&HAcae7=l{EeQyTDRqX^S^-ao(F= zd2w9EoHcVWb0I(=e)pxc-p)^)13_zv)c@uJ%xqN$(73c>qQ1n4(Q&@isyx=lPw=03 zWEcEg_3>j@M}7f8Vhc0wAdP8?P@z-fFj;F2T%;wvw-`}zf5#rS=}jtar``KQs;Z%v zHtpvCW1a!=vY-y7tN@DR0Yj@w&Y(b%Efhuhe`LAup5*EBK4bB5>>qhk64L>nO-O-O zw7C528&{;;$>Ht*#%sTyc|UJO!|hvoZ}x1apBcYWD_BbTXUp!jVlCS05A9n&?JmBT z>3mKKl*ORJMuG2Nw3o`+pcFx*!N~_Q@ekiDa zpp}93e}9SS%FBfVU(50nIE@sF662S zP3BuMuHOoxz(10@`okiNoS=Hb^6~G>;l*E&->7DPl&&D{ z<+4#HIPutR_$F%W#K@TD^YW?k@|R`;17v%Cd3lrmel3xD<0q*{C$+xS!Gz5<8! z$oR_PDj(ufPwL4X?$@_9Wjr>FU2floE#&!we*W0_2M-_rtgS=lP^Ijs^mc%ZaC(ux z&P_qC8mA>}5h5gTr>Jem#te8dB)%Wr7?7gtVrE(`nM?DU==U@(g zFELE`GKz^8m@1AmqQT|FW`g&9e1OYOP-5LXh;@1Kjdjd(j}47pB5=@uycH8J z6Kp;*O4fa&%PY6GR_C&yGFZ3UAoP+a4%@72_Y_dVGK+)qM4rk!p27HG>Kdn0#m?p= z+(y0H_?%jE2Su}dLz1bziq@!zKE;k4ex?7Zlt*}UuqFce$G?&P3_O#sa3Jb*fkIXP zb26)94iS*~IFq@#z~e~mJ5pv~baZt1p#pZujz9NSs-IAhVHlr3)~;N@L>OFI_;Us_ zMxMV7f+u!%TwOFHak&$uiwdwEXr2H>1l&6{P0fP%MD%6N1V;^z3bTF!1 z&K3Y?4>V2J1@HspSiuzOB9hK|i&*a2*T#@&oT975d-fw{?+JY+PFxjp_@>A}{~g z(}6u4nq_{I)=sBogP+h3pdT6M1Y^aAdNB@MfS!5szL{0I?XG3;Pd_9!BrA+1y5G1 z*nl2k^0lXllm|yW>^Hbb!z~aOHEwlKr`cN-Jo{H&6691=h%8=+7&`d%-Qgu2V1 zxE*5zaV<|g^j|@OeEaCAu%M9Z1Cu|10cg|sk)BkgLz|EYY;-8h%7hv@#$2B$@nmfV z8iu;~HmDjtOof_JNH5`FhLQ+B*AC2h(cpu+k|Ak954~Qck0k>bS!95yoGt#m=E7g* z+_TH)xp~bFTz{IDMb2tonZX7&Pz7S<_>?)z(W9e}#AohmdBWdZB#E}ByTTpmYfaw+ zT~_^uXgys0~AC-Zf@4w?BvYe%efXb)Llcn;Vk}#;}5b0%xd?td^QQ% zB>jOc@~v2|wRR9JNI!lAP;?W@fd61B95&gk1)ojALd;-1AzFY8v0;w4XX9|_K`kAD zcoHnH{v^5Jpx7poDoDl%^rHZ3sD|GLFhSmY;RvD*05Y%KNz7k@4O;cqtvl(WTr$tu zL=tP^uLD0SiYU9P6Cm2WJjLSSq1P1cW)MrFi`+ce?lw8MI2=>yUCk!gv7*;a`jdv< zcko+9#s72UTg?56<9UcDK@!6Q98q9x6~V(p$5gw0RqlwO$C^yOS}c85ZMw5V4x0Mhc~oAPD-L`-Wx+YmtTdiVWk#owwry=GZ{=%ZE1 zWCnUf^5@(5ZLuBDcNp{tOgG#kiepIRC0O-Z$eCZ$(dp&szBxb2zDO_1!e`G+_$}n< z;9!{t`BIaOeiJY%{r%hB%?%VJAU8#z2PLr#B2h<&v4Kuq;h2xEEQ-cNbu35+xy#

F9zM$Cira+`_QWf*vwD>rKP8LE3@G-VmuKBX_&;aTK`Amc#RV!&cHBX| ztvapHPZ$-VB&R(&KG><9zC;qo|Iz{w5B$4;%^M$2iy8Yj5N8=-@Kvw49d>Ko6Rg

fkdLtaikq+MfO%X2E&s2fqI0nx;7p~-r$7yVShbjh{ag}N9Hj~?yN zSyY0PGXk1`BJ^jF-O<5|JF_9X$$hVDt9fMHa4(bPA-4z-{eRvVJzVqzW?)4k(Y?DZ z;0&0oaVf$Af41>t6KimM+{oNqFpg&k!6n+6l(ctW#D4i7;H7=ZCOCrJ46Mtra{#{t z0$JiD%tF#j@8o)4ZC}3P+~cghefG-^4t(Z5#V|`~6r{Y#dlgmi1=M;)OXNo8I&4xwI_}2cP4eQ5#Bhl(bze@m8+5uY=;Jz-#vXQ7zusHhT7Im zLgf4_k7=*@$g2RHf^QTu^cr!gK^;s}7_M25aD>}&$MNjj(;o{H(eHlyQzAFCBxa0C zLlR5#($+!Z91z{xQym1f@w>$R)-YECnXrb*8HmHc*X`4B1ZzZ>W&KGJm}o(B0=f`j zSqnY^k43PT~%k#uANk_>*~YT`%>m^ne!;pq5|v#6a1 zfZE35v_v>>qj=aNL<)!qf}l3}>F*%U&Jtt4dmE3aPD$}y3S#l%ez`tjYFdFM zDc7mQ*UcZp2^&#N%X`s9FX*Wob8_TOi#acIqe|~a{sR;q;!DR(RNCDJ-9SwUwut+4 zD^t@J%dJDG{poxwF4IYP8!;jp%xM|CaSP0boNZb{ATJds@|v* zjjOmE*)+kz&(%y5wu+E;@XRWMtTBl1)1tqRps&0`KN0vU?^g&yDsH0cftIc=r2q>j zzM?#Xhbu_sK-9fk-}wX5@?^8PBGhcU<>U`o?ti$t8I!TS(P4K>TZO|LMkx6>pc)Fe zIG>)Lc9x=}(WAU{i7S$)hTTLnlHQQGJz6>UpK}E8Z>aSY$wNb1%qW1ShyZk2q`ltF zv{u}*XJmYAkF@ayQiCp@(A&p{jOjY|>-mcey%X%5L;{6rt z5}DcThh8pXQ$Dz(E2AEBZFG)0uq+mNb}fU|H@mfh5o5vvcm}R~m(RD>g6^r+}T7t;7y& zf6!+~mJggAftthBf;mjwD`n6e|7#?3csrmLRHp^o2#Fy`#5o4bLv*Q7fEnprJb>&d zoxNCZ`ek$Z&14syvNet9;=c|XtO~&cm$S&W0r1-c3_G}&a8iJ?+pW~6!RW3O2aH16 ze`$GUc}OwN3t&?SW=e`%i{|T1GpyQjEeoQ?fw6+!Hd9*B?v~D(E@XwAW$VA zj6A?l1Q7U7hWxO{GIsvm+XH4e6sdEqmFyQ!imndMw{C|o9g)-M?pXbUF`h`6fKURy zFxUtaiT@5X$hq8=e06Y)xBz1)Pu>P(cNK;pZg9KK1g#`C8y@p22TWh%7NAKBjfD~o zUa)TC;^Wa)$#;wMgE(3-T|Qbn!EHYp(AmY#bib%G)QzZFb11MYtgb!)LpGdptF|zZ zMvyj0LPN{Zd@t1~-px^C))^yM#L|&BZ{i+t z5C_8(2nhtqDHNK)0G~^N_@JpQUB*k4gpoq<(A;kMiQQZquoi9;iez0`w^erB0$aNl*bk515$$D7+JUP+RGoNO_xZ zNOv>9qT8$d;2WwZynPzYmV7pFlh)0N>Wb#!(LjLz>IhGh0OSgQXAbXl>#i@fvkD49 zpYN(mXtSEUqQuk28i^>1`QOKeLMRf=!V{q*1_QL~ufS93Ie*K?K6}s>f`uJkAkgff zrr|s)py<4i%}Hr6*v?k+YP#RDbe~obaBnX8T5Z9}QB2?u3IO$iyM!36IT-u!9Rx|q z-jtu5<)u-ACJc&V<^R(Ya&$|U>&;1v21x;GpC*Mi?sxO_LSRf4Pcg}Jp z4PQ8Dx;O&~nE*{R5M3Bz_pWi`ThxNBMrCw9I8bWc!(zR5?cV8Z@PGp_#Rku@!Vu)M z4FYk*lAc))XhM`}f!PU;yuEl3PpOXZZfFFNx4{!iW%*htB-$Q*#69=^`c@+b(GzvL zrpQKlG+!rtE}qtaGU67K#SH&dMcik<3kT|#7i`9?c&Im)5kwAn5`T*e(6u`F>fkkM z`@m4(or{hRMTY;z$_fbm@@sn|H3}Kdr6(c%0Wnr?g$0!^zCQYs7;kBY-U!h!k&?cW zW`=%`#wy~lNi<3atN(`kI;w!J7uWXqVbA|L%f67?Xc#sMgHJg~1mbJUb93jwK^x9r zXuxH4nScIcDPbBm`oa-XyElt2Zv4SM#gSGpHJX5@YTS3O6$wtTk*_bg6IueL2E1u) zWVuZ~cmK=;iYgEsXTZnSsM?Vh(ax7J);Tggz3D6kZaEjr`SC!{{&vDK)NKIYcs*ZdO_*eVwR9g6AC&aNT&lbBV6mAog zUH90`9IWcL1@C!u%9e6PBt=GhE7AC3nol?)O^|V(&4`~SYYBNPZyn*FBTMl+=D#s<*Ds6jA9(B=n(=fr*9!Ckd1Mnmgc|5)e=T=~{XoMoO zFlKbn2+WHNNFEf-`q>BRV?zqwy2ipdLRiF(`o*452uCE}2k5f>*yL z1`BwF{IkvifGq09qSX73AE8yRd>?#+6Veub{`~s2W7VzpYL;j4_bNCF*MGqt;D*E6 zN>4Vb0!r2X**+?cVi06Tiw4XOLGKNu+ zR(QDbQspqZ0(bqe1u1ycz%snM7XS%v-r5r%;O2DOii$kzpWrWj%Ypz6g`34B&!+Qo zxZ7+q2At0s8E{Kb#8QwEev1@N9Stna7zP+S$kfTd{=N#7FtCyIIt?Cbrr?N|{J2!n z5xM2{ppekFiRi9V+=hvWYSWa04#8OdH*mvPJI5qsG8AhVzlI0+wt+2cUa@WmfLnp>g zlCz@bT8JT-$1*9m#PAvX@#BhDleN)1ZSJ%uUa%>$=ym)1+Y6e|v?@`)RCk<$&kGM# zZBq(s>wVhFzg3$}nfUCCF86)jhC4qJy1Vgr@UYR)cW9#~``3ia#_5zPKrCNlhF2K% zZJKI9oiss4So7FG#^bAT?lA*v4wTFwowuQ&MkN@{A* zoO%V8z&ZlaBqhp%x9vade5AeixCKNwRbmBgn^2s^JFlN{KnA$cVVcytm!CELU|C51F)aT-E7^o=cD-#&LyoW(F%NdwV(W`e=k-}XZt>D8yl{^6hX7q3RIAjp%d(q+k-P4YhR}x zK(32Kx+b8Z$!F%v5({$eO4ix^AlzQz(~5a&_SyUIBQR$eAD1>xT&Y9neVOwq)e5&# zn1l^eI&MujXhxA-7%_x}E0ebciZiDRh+7q+RC%Cpa-+5x(wj6e{t|EA^@TG|Qxif$c&U(&atNI%L0lUnBt9@teV{1}t!r?FlB|GRoOA?U;?m6t^E z!7zwk4J+?fPzOJeIs~vlPty`i&;RjSAlzHR%aC_NoKsxf=^(BlI7Aw*^!YVtV}x5C zNMs;;-h5_>OndC)MEic=;xPV#`{9-K4`4k|ub0}xbqY3r(s5X`KEDvRNo+t>i1n<- z6wK_ffKX_JxoUNVm@JZ2MV_{@><^e80%~R`TkI(bs$S}!A+O)Rht~EFiF z`ao|A^DBT!KwA|R7bhqnP)O#gnP1P6j#n|Cy?IXm8D>OttQCg*u#i*JR4V8H7pP=m zS%UaJr^FSp4V~+!M+d|!LSMnlw6&5C^q(GZdQCSI= z@3*OWCmx!cPs4YB7?xNSj>r6scVvJfgfRD0|WVHwU~NRyzw2TtUau9b2|jF~4~93Sw8e*B=BeN|LdB?jQ;*`-3mdnM4+ zg6HVa^;VJCvEAK_Qs;ikOeYLL*%=|#;W{Mf1}cS<_5sHZ+-@s3(Vd8|lOs_yPd7?{ z#OmrcuQt!mR%?yZABLCxLmzp75B&{|7RjEq24B>Cj0 zZH3_&a@0~7^os_0q(Bda8Wspwq3=eaXG;9`E%6m0hCH0Oz8PKp{SXR&Tpe&y4SmDh zy-lY_;57pJ0N8EdsMYj8Dz#>pwDMHLW|LgtT-%OOv4ISLP8T3<7?xYMY@3om?Fvyz zHy-pe!j0%vbSl9D8C);r9?U)NL-d>J`&0}6bU(AHU89)Gl6OnK}2lp86S8;6l2bk_IS=leM0-GdS_C_P3RDQR}yUjrq8RYez!4K`TS@VFO z>=FE8aP;`pgBc^&?UXJL1=KRx|ASP-k!2e<9t6@Jw_xuss<|QEO312RI;Y6g)u{Zz?=AE8J zoSK)`>Vsw?>{+dHgx#*06>#4xEUGoBF@dcNEiACBKN|8w`H@5s^*NfX$NBqhKuCp7 z6Ea&BT!5~wKrI@Jw&5`DLD{8I(wgZ0Z=TZsffZHKdMS^|`M`m~<4L^2^UVdWc|#kP zK3IUKUmcK&Y$G>xxT%>Zd?q$ej9BvLd~AKUo^ppTy1c`bxf@KU08qWrt$ko=NfO5= z3na`*-Q8)hj=rCk#}!fMX==xv)WxAgH&Q)tkdg!DT8_CUpu{EDb0uEJjk1mA8T#9q zB&^dFwNaTyS^JC%TcNi6!oJp#J`hG7iL&LED=kR&FK;HGPqZK z_!CqQl>j7y9NjRwjCI$VUVE`<9*UHr6+rY2zEbH~ugCLPRgFL@iluOQ8l}-QJ-B)d zVFB%QQWAAYi?*q$Ou+dmR908sRho#Y4o; z7B;fRlwZH?8CtAR^P;g&q$TqE4iG| zFlGzud^F%}6zI#-3wjVGlw0Mzw-fPXpm*A1d_hk3aQ@?KFYzlOg=J*}saU|p(igD<)=)!fX3=-k6!a-6`ZZ6i*w;F$~S_&U0p!8 zYP8_MP9hNlD&sbFl`lxX5w}wXNopU&C%9?Sem%2?Uu69I_vZsQ4`N1M$6b+q*lzLu z#3`j*h1x)r>hXS_%S1twm*4M^S=F&eA3NIHV-#+pAFG+#ZJ1Rafu`Qz|m)U9xpNT8cJiSdW%3Sv|Q!_04Ww%J*pB62Dn#f-N3K+F+czw!O?Vh8{ ziL1b>&aut5 zG&94TFQd2~XmxE)F+&l^7AU_5xtk_-V~GpOoo)mLM89OT%WfTVA1Wm-h&c%?U{5vf%@tn7VxWT*$V8n8b0^uU-%=r{?+PDr1DMbuz!asPg& zOeBCgC^{?vYkam_@K!$Ht6t%$C>U%Krxv|WrSuzkPhXgccV3_Bw}x8QuHTNx(c`4p8oC{j&eiks%~H7490| z=;e~@ZGsT8a|HDSy4p4kAAhN{Z-^2C4ftJkb&4iaVFrlPoNxRVc=L-n*Yxz@pfdm^ z5P$}ew7~0cmeZbbvT(JWzmFDMoMG#}(O>_VSQqK6h5($T$#GoNK-5Ik0B93zHymC5 zK45XWx@{FYkZ#fP=Fi#D@fefO25e7W5dI+{VAGStna!}h?12K(*vkcG{Zm-k>|iH>ZaA>i4pXvPGHFR=r%vG8x&fB5t%Y*j7u z?t>?C%$-*kw`BqryAmKEaf@^=wd|Eg)YjkKt=1cCGfzA`4iY$2J+oLIF96Jb-FIhy z!s)?-2hmM2jM_w#xSL(O1DI;|qv8yA9zR}#{}a@$lwK@Boh5lKmMC%9L=qpW#zIr? zz?nKI?@cvNN&*1`S}6;me5IXypA1SaHJN@kt&#&O z$jU;D;ELPvO_PyFe<1~AIpvb;srj;Eo(ZO3!_-rVRm{C)Wqr;5Z}1^?5@S4Angs;e z)v~`3Bo>+$Jm4ZckYVeTRV_nH(Soa;2FcN4`^nu8fwv^2Jukn;k*dC&M_6F&=TBZkto6h-x9czzy5eY6 zKG145SUw89l=eHKS9h;Iz9K7UbUSY0?c2AY0C0|b^L-UcYoE(`h?&Ndpo(hn2INOW zHQ>+esM#*6@3{bo&9y2taATfFgscGI1?;SvtHoYhxiv`>HW1h>pX~mvxZz>OO@9x7 zjJ+#r>fQL|8i~Ee)=rt5_bJ{JjYo>g=<>YtfAd0+k?q!QT)<;ULv9@^Ap$NkpkVN7 z0@4Wic_{h^sl!LNUtf#xHv+A47y;7(yo{Zi{T^*?+tY!0AFD{DBaFb=zGlS)9ZD1_A?x4zIQJ^jFiq z8cw871JMt{p1WgXA4T`lTE0o7AoVeA{M&ME<*_%N6yycLnsOTIv|me23jQXT38sz~ zMiAoM=zcvZEV5{)mz{qQTo!QWOUUL61DdGiwYA5Nj?^?HWDQCgeJR=5ESD~U7Q2(? zud6)YAj#raDhEg0Z-EXvITZ#aP_4mW5#V>tQuBGQ=^tc#ctUqazYvB32xcI2W&WOZ z$m$)m3FEw6Rb?e2A|whk)b7jL>gq9AQiR4ry@#WXj$pC6-3!`XTLm_g*V{i#oV?Zo z%AG7`VX%h-VUL?xRQD&vI$ivC83C);tmIm*rW(B8=4fw)fI@yV-2>~?*RO55P9>ir zel-|}sNPIB%j-9Ng}!AIJ5;Iasq;e-t!Af34t;O$n>G3x6h0Y`ReNnf+(Eg$!>n`9+sWBYIjDZI(XNR}(C^@(K%YQPQ@4@_F$z z#QS7Z7l;yLwxa3itaOQ|libpsh66C9+dE;Ns=wTS|b-0RUuw4f!t1ho%YHsv+g= zbMY&TmNyRkSgW(483fG>ghExE5Q0tz6X^7t+y&ecL<(g@U{boO`QxXPsuuzT)@%hIj{ab8~1L{rzPJcIUHT6SulJkIKlV+?)7w zQffOq1EC0ve;M1C)-JoBZYFH4cYXQJAK9hkEkVR%!?~2F{UHZDvV+4zc1DQt(Bx7F zsg2zA7rm`PO@DEKGsxY8GQ(~c3o~X_?H3P?mCt>Bm!NK=)S}5hX5cnpH2*}EFs=%X zg^k8y7K)L=3-AkEy!!hjc;6H8hs>y(HR081EBT7f(Fkwc!48UT6brC1*omDt&s3L`qqho5n^$% zYW|SpRPD9TRd(8)0TAok{kjqdp|ECbt~*2XQFf3W|yWAOHB6wr+iEtU_s0Os^v3KRcyqg*Bwy!UG9Ey@v-k){!sP z4AWz93sLCi3OQn#@Fi=Q$+@|9skJgF-#U3%RY2ZT zjW=>R9m(Eq&(`y-$KVJW+sWGJJsD#9Af1IhCq12~J&6&kHi&)PTEp#3gv;@38dr^_ z2$#*XuVb%{c9(8H-GCYj(E3FCci0ZkoulX|iI%3#qRC@)J z9TmmoaXJiE;LM5t^76CL&xv|3u|QZ7fSeNqKrEi@KIQDhfivaHKIR2}uyW!I*Ecg0 zc)H?RKT8NL@$lZ zmvOmx?<%=oxnO<_l)CpGSS)nJN&^ol61*OT77T*e77mcE@Fo9;l&g0KBC$b<9sSz|JH5^U<|`aVpR&u*R+lL#d2;eS$84cPyX;9}L4 zMQ`;5?ZL5#B=+4-6#MqHnn)3En&>$`Z7r>Zug+Lm^??t>^Tl72@U^8U-rlHVN%Ik* zEP8#)oCtV!ir@8!YN{?_jrLk>g)ZD3_!6&F%{))`pETXmRHIcX5M=atVEKnJwdeWD zH?Bfm9)4v@9CDBKu>=&V9D>p+s+J!BLgVObH`N2&~ z4j~~co0813N=6Zt9U)0VkupO>L=ssgd$04n>+?O|-#X_!&i#kP?R{VG>wR6X>ouOw z=ga+fPu>1#8g?nDIMtAv($KQ8HPL7IZmjhAL;mB2$bV+ZJkz6p{)+cZMhWuZ8XC1Vf3^(9e!145lO3_asVr7M7^@tPFd8?RTOYE7~jX zoJtw$5IsgD$l_RGrMdwF{hV7?*7TK9>htIB5xM1ijd?1gE}IWuGb4%hU+-`tON?ML zMxS14BE|?#I8-M_G(>CdtI?VZO(M6&s-d7N$KCv9Qf`Z;h~Rc}$Z;zv-Pm0cR@)I& z9jd^5V_trtc5fv4l0%5oz~2lL^#&!|sC=ngs_5_QPYibmlTS)DE2Pqw4(!dk!AqyL%3*t@RSI?I|2;LzPxIs!c zq8`$Q-HucErA{@YrMl567aeNs6 zOHGmnHO@g|kLTh#%jLy^NMC`YtTolCnhJ`Ft+qlwf(?REIyYi(zo>~d7`eIfyDV~e zX(Kad^8>P5Ar)T^5^@Z=W3?}A{!c|cUsWBUTU2P0VdcmT-cMi5y?ge;3DEFPiH`@e zq$fvTY4u2S7aw`Hyc8W3=t)^56OZe4=gUR}xg;XV8ut#okL*J$=h)#-0X1qQ^{M0r zs;9QK@ODGg3C9O*R^_huB)q|gZ*D#w(OVWeI6*AvEPkn6`vU3>I*wB1TGVA*p`M*U zc8i=9OofZ{UJkJQ+8500WX;Z8&sT>W0i{w}sS%JTd>r<-`EDx&Y|Xx^&sl8BnEh!C zQ0=(oWUxvA8dDr&CZcB|#^iVIi~D-5*IQqalY(cjJXHx(8w2eTw)zzz=}mgg%DW=A zCjrt9EoB7Fs8s@{OS&$v%SkDhV~-@peK`qEv5!Y~5_<&EZn4}30Tn?@z*d1-1x~5p z?G#hEqQ1z7F+-HP4Mh~LCy#E^{7U0NaoFGt%aquj0SJzx(Tk&_qpZu2>bHE0ErkS@ zLDdsBq)eOdDKi2mQ`Y;nyE=kDXm7OX9}gcxfOp~ z`jK8Wh&;_#U0}`x#z3KJI{|_bIC)44P$@ru`LYl}T615umPEkLlK2=nzPXSt^5dW< zJ+8n`1gky;%<~p&(Ii6cs`D9btRYswVxd{#CYyjTO^(Pv#ga5lBKkHCeNpeLp9E@_ zoQrG9>+HSZR#y+K#~VHVc3~w>gd!L-!sPGhWz5)G9F8ff?Ux(bV2uD=g2*_L1zDyiVD{iCsNE5s;MZ9Z+>tlD^9@l zcQQ-iU+hJ0695=@`OXbZY5#-{is!T?IG^h3w@M;)Mh;0xbgjJ)a0+T4b^YHN2jg-t zH+T={LV>IRnClk1^h*@?9I3BO7-N^25R0bL8>=%gYQOTL?%esYo=Bc&0vU@PVnQdW zLg)2bd{ta^eYX$m&@@$S5ZFEJUdLTO690e@a7smL!SK=7;R2TfhK8KWqjS7b^lX>H z!@UNQ5op&oabiq(%s42Le3aAcAEgaKRP-_g1zqIso*dE0tL#ht$dmW z2DZh9L6PtJop2g5dF(^tme--0hmX(dwAQA0)&0eWYiRcr+8KTNB?1H@VzGR^hS`)c zZwYC7<2}vyj~y6ovM7L+vpf7ycH`ouOHagieNWj~9R`K1uMAK5mFpf#$qu_>0Ao1< z_rK{WbqV^sGwd5>OwI+T)rrk^o}-MCWcz69*F+KoD10_bSjiP=*=s^Lp!%!1sm%ehjNnrg0zwPc|Xm3U19M7uJj7PV{S$%9hQZqqF5j9Oh`W>#X~b=2b9p@IO* zH*W$f&sCZ{1%}H7m`lR;r*1aEGiDos$UwE;zWkZk#h#Iw8H!e@Mhuy7Ywi4Xv>FRw z^QhXHl&B{&48rP9^HBKMR}5FcI_*dqWToaDUyg(R^ev zImkx%4L?^^)Z4cTXr3m>LeS@pgZ>~8**L3`hYdE1&8$pT)6n>u=H`IuST~Pa%@f_9 zSJ*>Y<3K~1Zyn-H2tM^MEC4h)Pjq2G)q(BFboB>tca;6s^gVlf$>rBld% z;nR}uH{X(4L?hk)`7^eA&`lX(NxPsG(Wv$f4>7@q8&j%sn|L)%z)yMYC_JYuPo%|%{$)Aq{pv+ zyS8`f*dJuyvR@NqBW%BsZ;(2$Doxf`(a7$$*edH(>Nh)+(gz(W`KT;#8flF9J5j{^ zJb*-Yk*6e4p}ESAFShdFFXFzh7vWL`iyTT&2lnlIx)p=Y%gsd>`K-LWbx0WRc?LVA zszz2Jka`#~+d)6*0wKY=yt}ncL)I5zfWejHvt>A) zPQ7s7US#dLK7WX@>WL^IB=M?@`9BwUxMYEOVsa9+s%K0*Nqr6(7qBsbOo7GRv>Pet ztW^YzlpJ1&nM`E=f**MYX3g|g$c@p`6)y*NIuoK#7VJ@0yj&-!f5%in4rYF0!`&i~z#u=%tB>4pvm57%f9mH+2N*(pg! z-9VMqsULsp+QB=r=ndfqar%7?FU!u&Wt9XE)oa!oCnhi}8xsn7XUj1*E?L0ujgF}Mj z_n%U)!U7_Xwt7NgvN90KA2qGLkn&jq#rieKJ8FQLjsL@)3PS(HWf99bo< zyr*rTWJ8)=>lJUPG*%h7^fB$)bYvKWv_%8#SmYXqkspU~eqS#=o%NpP8Fb8BeM95N znPu?)UR3H%zu&&}=Z8u7oIQORn3;zQ9+$BD@UkNL|CQd#`)AX@>1$uG(abX_{ikD7 z$laO@46~pZ3daU^7EEo!CqR$+wNOHJML}m2tpXL{S-()3toywi^@M>ITUP@(Om>i zTgz*@KIb9YZqsd9iP(j;HkaJlR6abaA3bMl8;mNDv&A4eF`@0L?Zi5W$7CWjx*tQ0{O#4aSCxoobz zIQ{Klu3q`<`1v1ebdw!5)~RKxmX(YHQg_JvWdZ|d+S2>=(hMdkLx$DpWQ-&j|PHu*VFo^X(VoLZ_Y|<x*F|rw98`))H~lC50|GBrQAz$rKqJy|13|-XO>3lox4&W zP(A!J)1y^Dy%j26M}|!G^l7vUhY5!vNAqgHy0&1B7+v#s4oJOnXc1uDW!-Pw^{b*@ zP>8O$IPhorbEey)cZqjl@`wWDuU9i4j%i3uDWG|ii;K(5YsOuMPuvl#irSrzCc1KC zUu1*YMK%drH_9IuR@ofuCb{px-o2Zd!c+1jsORP$#Ox!EDCDO+0fLIa&;v9DKU5lu zMl}|#HlpaYiXKNQJ8o)u34B}Z+qyAbV96vbl`Bk2%N)GW_r5{kVQ%A$^NHe}U!M0f zS3V%pM##C0T=x)y70J$f0%N^KuJ`-hUcIwt!O3Np_G&+uFarX^@6QE$A?KEt$A2w&FF*f> z?q+@2BV;>tzu#0Ar6iX92^wX1J!+}B8QA~5_+tPW4f z%K^h5FFMXrUGH*xKL5Q}g{s)1)UAMtzJ{$n%`Q&8Q$MGufWG&tCc}j`>7%#e?1LYO z>>8!M{*dQ&GeT$bN^R`^PCB)fcAuvhQkJzL+< z047cIFlBHx6LnOxw@zez=!SHXi(1Ec%*6VW9_O;2%X;faC>1SR(-dJKQ(8}uZQVMP zMPj;v0~T1I{Un5?1jRVA^vJ9nSo#n-*tZ@T zQX9u~D6x^Ia_-6Za}IT*Ivy4Ik(WsWq?&_pr0s|XRJQkE#+h&huFAz40fk4oxe--A zLa+L(o``g6lfgwSPleugtZlXKfBTm7n9?S~i&M>>uvggA2brXuzR3x)JnORG&;Rkx z(e=I`qEe^E*$pgX5;>G@b@)3T7U4?@?^@$pMGHXi7%g1i`j>E7?zu29j?_yeHeP;!p1qo>En*jVk8^EtzLw$3ZH zml_lm>=p)Hg=PeX^7G zm^5Q=wlS6)s+jO-lPUe--nG6^Y7!cMxck2N<+SYVu!PJ_gh6?s1XZZxtQX0y6v2^n zzj7tGx*rZ+a5efEpu*@n<6)IMpXJ?w5Z&d5$ne2%ERMJ=R{rQ*r@bL3Ylt!iWE*h)Ce@5K7L2baR zo)KGNR&lDKVxY!4p~ZPEph|Xda92{0O1j|cu?lbP#+ad_=8b8PSf zxt%88^JU^T*qd5oZw=IlQr>A~x3bC4eA@H*j;Zg;Y^;hBM!hmMQ0`Z$a#aH!A)Ei~ zNiJbQ$!=!yMfud=W5?FK7v70(`66~H`_%w7-RSzbWAB9znf^JBmE{hf4vFn2e%|{N z{Zl7vHoi(#T$fCD;&(B<5PQ8zo;&a)EBo5l1?z&-3&k`tnyG^6rI`*_sHtUEW>VhE zj)l(s{B%(wCx@s>iL_a8pZ`WeaQ`^b*QK1NxP$8{7E`%c zDOM#uAMU-vy~KXug4tKy!Xy?Oj}pu}K~(Jq!tn4T8D0(E^pC6{~eF!wPTh#AJ2vxNcO8rYEIK z+f9~KuH1hqbm;Mwq_9^ zBqDBIW|d>ro9C)j9c&s`W9_-hM;EQQG$WDl+#u6(jORp7SMG;{GOOz=m?M9mzKZr% z44mssnez`L;mq$|v#pG3Cvinu32r{VYRA=5pIMDFhTJ(7cP@l#60gDQ)bMn4eT8PN zE+sQIx;k;bEj5+dZ>ixyI@>W{uZ7Jtk31eo4k+9Y*!?8)3E%#in%%ycDJ!d#m8RUS zcb8W#>g8GTpCCSH0dxe`H{_mUD;)BkE}j^=7+ux>^bI!*-wU%+K`(cf@UB-YSGWE) zeEzHd!w2F=w3oININV=pOz3Zx39mHmZ8>imG2)U^o#;LoNMpFNe2AKjVr}Fkll9ZB z1ZVRIInHl$US(xv1O9Gk?iF4mKVI0n``ithN?tDK=I2s*E#Q4jp%d0lgTg}(kLi^7 zj$h=WZx?M|nmIVc_^ih(C&XvkbYq8$KiuxI8_2l#wv8VxiErZDb+p2BK39Nod9+1y zY##-IC)WrJ(j^4s;7wXaN;rMP6FJ?KXT?%9A)cSmG^DC_PP&M1COxUW;Kp81Zr$pb za}FMkg&bW_U%r|uOQnU0{ISlwxghzIApIDT69%h(imF|-v7x0yH23*@BsALhjquY_ z6m_YdyRqfzggCS3hZMaP|D4k*XLxlt=JInEJtp-k?YdrE=FeXo>MAd#l|EtRR?)pg70Ko6kx$d5`#KMB=@u>JSv8!I>OeyK%~lS3kDpIE zDk}Qc))#;U6!c0Xz2M2jTQMlbeGbS%uP~d+a6O$bi<2~_8n{7D54+ZnJXWqg0KYglB@$Kk_L>G!FEIYf6 zv{KWbOqslRkm7pAr-4r>(rbCVQOmha6=y)XLZP9bgdy_C)P?s1qBV+9|rB~=A`Zti@URi`i!^|WI9p2fGz zb8`+Q0*uZn5)8g+q33&FVq?9{kUsL}<_9}54xh>6Ljt4(=l<;o%MWo8mC)NpA4y8s zcgOWVXCC}IzEqG0OtkD$od&3y(5Yhb+do#CbFSDfyI5yjO=cH$ezAVz(LSlRIt!TT zjq(C51rmf^tE)3RN_yjdom%x?P}&d=?<65Gs_=tT0g5 zTUXawSZK=J{oHfI^=0i*UlW<-4AM&DGh8~6LrpI$Al#RjPHgACmpT75^QVtVD-J!= z-@mtOR3%#wY$XlPuxA4wQgs9=tNhF%ym?0IzChO+|hUG7T51z%|dBas74;{>poqp+K zSUO^z>Wpr8fbbYe)zcCJBEowkbBD4#kv`1H$=MdPWL99Xv%JvZuGEzK`V%A9B26v6 z6F1lh0iuId0O{%}IFyZxsl>{v zF)=QlnrkuirJ;cuga@ykQ{NAGf?b9pn~#j?s^glUKXSq*aRU$mRJbedcJ#P-+62(%ydiC1f<-4)4AA(A=TC zAFD>#ju7yRx$Lx!lappp6$oE<8y-6lnO;YahKDp+C*8Z(*lCTqA(T$i01+X5XA4?E zK#)W6QDNb9YA`$C`%%rqGfO`3d^7CuzNHjOG^`Dgny%eD6|SqN_r?A>93Hhcp4|k3 zT=KBI71qM?IPyMg^L?vLeId2LhW_h!;Wb=bAg?ZukNgZJI?D=>qX^neAk@wnEq)86 z39E%tjpAEWFYU!rnoGdChos2`&}8ruOH%uxh>st1;YaQEVaTs9B^qlU@E(D0n^|J_ zTSU)gn($hrjMUKXnJ}$!?dP{qhQ&~2L^RlDQtZuuMQ#A^oKa8^ZM|ox8F7@tKJPrE ztpvH_!&TQov>+mi3Iy}f9o__zQ)u`VTwMTQA_WQIHOau4j9Mm>{4406%t!&Ga2Tc{ zYP?jCmK?%Ea-4GgQ6>xw3?Mg;&4c8}Z(5d0J8s1s|Lcrmu4|>IgMTE`4F5BKs7H zg^L%A2r{O-#OyV%NbbG`b}T5J+D9JgsTOHUprt3^Sv=GP0_BibB?1Qyy;Kx_o)s5o z9ZqcT?99=Gn!~D}KNQbWX>FPNc+c9fk`SV<8nJ{bAZQ86ob3Ggk*H8n)|>a+{^CVq z()85z_ygN8qLX;%)0R?ANG_U=<1ii>5APLPHv3|NL;vOt8tf$w{Z#>z{xv-4Tt+kPUak zrc+aB?0tsK%MiPXpg&@iIg}o{`&{VG%1GzCv8B?i@F%AebJRqepr;vJ1R zSQCUKBu>%c0UX*-cThkeB{4CLNnaV=Y|*g5-rl~O8}tfWCftu)DQ#^Cw${gwXU4F6 z8)qYBv1MIV{UKL3%H)c*c{R6>Z%$Y}hnGGF+RK!gHuCycp|{rqK{ z=)OEV8XX-Cw8+XtWuHZX!p8GI6w&Mp$eSjKU6S-!bmFA_*6oXNbtMz=h^T*I+0S~P zG4ih&n6m%;bkA}!vW1Wx8X>lz&+W_0$}*4NL?EQ!*|8E`ZsO?5qiRkx0rJ={&ZYKZ zNg=?zLN_aO3p}t8^e-Nx(JCo75eO`QjU^-|By^mOI0ql(kWgwK-o?SOu(FnQ*9R~v zMevXjY#qD*0X)2?$dD-ji1N@jF(i|6R4L;%0Ks9kuzRe(f<%RcrUn4iM-Lw&U1TS= z4CPG_jdJqx#FCn_m?=p~vY&X&j%XD>t+VVd_bwURi3CKJw}PdOjk13`42Oax`V>&j z%TIfi<0Qs0_;=^Y@vHi`)6yk1F~xGi`9H{o9M9OvvfkgA5?jho53$5per*vE5^C-2 z%y3ANaJfO;cY{@)S^!DMVooYg02%3MYz}2gi z1$Q-U$KeX-pbfnlK4?udy?tX#>1g^A|Bw_b8`~PTKNwtIKcAFL--*88?n)}Lyy-CM zSom*p*H+o=#$hQJ*YMcr1$3zB zrj~q!_^PojLX?rms!lLyGdX*MxU)H#(}Qa)3cieUK^}bb)qoJ2#35J7z&C&y{*bIq zxbxB=vqhMcz?o!KGlh*fj819Mndi68iQEEo2X=dw+thE#3M7-6B}j0CN!;f#u-z@L z8Xd3Ngio07qsOiMI4?v=mM`ZExpfplwD^geJoIKn#7Gg%q2)-Z284VFf=X4YX03J6 zmcTbSOLC5_tgJk|n-DI?>9o*$Wu_qhAWr)qD2KrD(7IcIjXYSTk8GJ=rEGux_X`D} zlc5U<1z2KcPUx2~dVfbgILQ77?9>%lV2^Bl%#Y;nQ``)j5+BfIf2(wG>KTxeK9;H{ zBvkjOn?L#X@K&6{dvgM1}D&M>T+Ld0R#o!Yp<+J1 zHUVwJnl~_zW`?^>Jpy}@Dm9C#npa)2t&sX7z?oM4F}NCZ&hX5Ss%K)92lnr0LhpK; zRxX)Cm@f{5!E#NenS+f<>b|$lmlK0v7i3&VS|a^d2SP))5#JO=IddCRit)nQ6DLkUNr+2WOpMr$25_<|tm_AY#MKLk zr?%vQNo(~YXgEC*zC7ytaZ!k&*wL1A{RXRz{WgLrs_{_)`lX$z z|NdV2KYXwFkHbNHkNm&CpY9{N6MvV-V5#jtju!Tq3+gyxFDy+>Iew5}hlac=2BYP#--TY{8>1*G zjXga}`zG{CKvCtvA8|4>*-5CTL%W#xZVKAr`}pUNkMXZ*#{d8Ac>MqWP}$DHA}v4P z(abF146fX+dYiM-v-~Sm6X^^pe3r6xa)`rFzTsysqunI$Rt7@UHGHJA`Bj{-%i}1b zf#9djKzx3@;;sKo6-J$gP>vWR5gxtvJaF`d>*z-szsW!;{pdyORA|M3XjCf)%jPjl z=@?PBzbz(za5n7iKieq4$Oau=M^{yT=P4rA`m@LbjP!Y8j{M!3_?{LU2s{k~X`852gPXgxL^-1ZnM7=IT8!%|7R*WHH(r61p1{@3dhg>Uv|X*`%wifnblY7d=!0>lXDSA_-<+xks%qC z^p67x{OG5amIoWk{GTjaq9$f#Hnt05-jSA7s1R1~`AmVB0_>m)tlv{v8P-nsz)zxdfR zL1m1X((LQPUhnd+BbxDz-@ZMd2trZ!hchlg2U5@Y$cR}K-o*6!)dD5XR)~&V=`_9) zRNW8gf@Qcg+U;?Hy#U`683zvUUj>~e-j|wDu$=$piPZbtTGds#*FWd>djqLn7Xnjl zY0ICRlcRkU7}r88lB)q@2eUPKxc0}|0@Gu+)fKI;*%x1#IL?;78A+mLrS7L)Cv*?PDvDE_Vb;~rsp-)+Fpvlz4wYGufZ!d;rc<8uaJmVF z+|1dx%vN(nA?3mK-M#GCv12!7zl@Az`f;9jbj*Kw|7B99tzltr4=AKluku!aX?*|g zUB&A3>7kA~!&FThd;4F&Z#Uuel|NF5HT{{y{z?1*b1%eBmjTn#y!p!&LO@R9jm&JO zt_JM_xlqO~y#Q^|B9CMUEL8T%J`5%j>wM8ZoPOQTeF8>+bJ$Xm|Lpy$t5>haGo1X8 zEPa?uH^)tEiV-7z=Et>$K%Ws^I6#~VN0y29 zT4(_t{>%nki-4f$9i>jFy>fk@S+Y=H7k3^E34siekUD03OEe`4>%9)lbO;#J@3e|G zjk<#Ck@IP%zYGO_|Iy)V^RX!psr*;dVG70*3bLh-va(Lek`Yy%pLf~Iq~t>|2I?a> zkA5X53*>SJ(EunSaZKBw5C3jfRyf_Q3nQ{9T%KtDpWHJE&+XXZ=8R1XKL*aIvR*3C z|9-%Q`d=QHQ7ZNoXa1j4&ACeLHU@->4ZcKC;uGYEyY8f|%g6^E|yUH$NM_d4hAHR4(#J6p`nA}y* zyvV}3`vL2vla!LTwQHl+*Vp4ykyv@iYdh#9PVYd^fA?a?pKGc>9kIz;WbsEV@rqQp zaK-dRpNjZR^qAJKZEdmAcLWB75?cwZeYcoAc!S7tM(KpN)1dwOFZ1TD4i%?<&} z_WOkvaB#nF>@5$cYa4#yCTN$ibMNz9)yt(uc5~nlp#=amoIcsaO@q} zZ0U2%a4~fsv-ORgRKe9kGBPp{0|Ef^>({&+Cwl{3*P>@u)<$>Un&sEELBQiH)3>8Veqy#KRbkoRZ4z`G)~}qic5)udyfRMxfUDhI(VvKx<>_MU=?4)-Jl=jwarX zDsVlG(r_Ao2MLzJEd-hDC&e zrIQo7ChI5=(?=0{$wI>7a_>qWMo@8=3Z6$X(Bij2nLD4=DRhb2??&+%-jRH&CZ2uw zDkM0?e)cO*Si}!>yvZ!E+A+-nqYG?qXh?Ox$4F{$;iLaLsyU)TS%LB0wPw1tQOnr) z{yN8A=!HVyY)rUfjbu>sHmf+#(DOhIK|(n zFR*Vn_4^()NZ3Kw1^x_o;(@VB2c&lgrW#$vKbEl!5toy_L6DSEYr;V9mA)ehs;qq> zEPoeJeJydOtZi&~hLREf)eGkX4Y(}OT3cKDXEmDNI=A)VyEvVI^Ok}xmYCqyU-IT* z<2I?mF7k`Aj0t^E31mLxpihUL@u;l4596qbC4hCK2_OM%Jac|))g(9 zJdf%f@N})`M8%1hH4-O2vW8lg0xPQ%L-Fo+M0Eb^X7IZyPlsl59gz!zUM!<&dD^x` zU0`q^!9H8uI3Qr|;DlMxPaU0@(9_#EYpTj6JU%DK>@3c5B=G#(xI0t%N6OcQ(vsx< z+g8}mNKtbyrAGbT?0QhRY#u1p+=mO_5TE^)_;!Y{4QP)2%dEIlN4+)H;?pMXeqk9o zk#6Gd4~LN!w~;K`nu)p-mJs^rXy1H>ObVAX(lAJBLPc(IN$SE;+-R4OdUF`G+v~8Jy96MC3>Ogj25Z0P8s~^(VtF@pm;PUk}gA-Iw2x1PW%i+ z7({e*i71@W9`rEk?)IAe_6A~P&#lbfr4zm#_9|+zD-H{3`0>&IhjVFrip$)?zKO_n ztzjqeG>%LrhG?WNM;zJVH&6u4_Kl^P3e{AjkcG!T8-tXLL&I-)3#em6_&*={k>Oh^5ohdww#rbkl3{PnJ$JX5P|ut;KXt$z zt!G((PfTb?=Mv7W?RY>+uy9UAY<3VsLoc3mIFCWTGN%6il`tcr=$O!Lk6H;#(OFI$ zwc7@+C2yGvey>35mH{bJM@OVX@jBF{kT1Vlw6x}z(tKC)9iZUw=5Pxgs@C*xpXkYq(0xffk;{QwSia-nS)%w(5x*&J|^?GUZarH*pmM`;2Kh2 z;`q*``&MJQWpgn3`XsG?V5q2T2bu05mq5j$MU@0++q!dI$Jz3zChKsSR|1J|EYeNx zGY0Er@2VS$4qd+o{=^&y!R=j1>%$?n{X5J9-*BE264Zt*3Pm85l^m70a{d;w%L5x< zPX-g{px4zJcsc@7?MfT^=g#FgWHx4hSOo{(^)I!&m~Aow#DIGF_ckVxxv@t-T+*0{WHJ%27_bM! zLnOD8!&J(7WyYCfLv~+S(`?FyVcuyZlYT`wS!!7;P$tC@$NA|v-mbDM`PrAcP5HYLvzeFQ9j%D=s>$FZJ|aNE zr&sg9dk+6xhs0ojhDU`+BHZ`{bmNhCqyJ4wuTugMd#c8fd7 zDHxtR$ME|rkl>#-sB&JVr5uVj5!(8>X1CLG)I#Yz`-7N2E`}Y_$-JK-sn*s`@%8e1 zwxW%R7j%6g9U$fLY=gAU6W?a{S;ks}^^xNwVOO8e-S$|_gN zJiT@i3;urf@|?7q#?R*G-QAi^*MFFXDGRco6=ejq>LHO)ntRdUi-v}v0O}#NOu_!Q zWSR3a?_Gq4RC4@6I!OGi{-vdNjq`O*o;*3b5~{#8b~3O=wOsOG<`(NKlYOX1UMlKB zWzn`kAF1kgtr($|w}89waBa(oBNdcZuOl;aqNXdoeB(A)5pEe!<2djoJc0IDm!F0{Ig!U&JDYAGPY{imB(S$DJec+x2)f4(EJDKa-*8t*ld4QFXh zltQo`X4Z&*t!{4XPM^}!+tL-fV{4hit$m@<7V2e^bt4QM`&T2l@9CT@x8Y%o$qaR# zU0@aAxM_G7^B{J?Ow4}^&!&L(QbEaZh2ZCPb*Es}d|N4f`M*CKmiIj)y(0-!$<^6b zu&uV`Gp98d9BnfaM~ZuJdKa0C`RWdZU+&(nLn4bW?5hTva8qfCO*d_um z52~}vKl6|6z6IDntN-mfOSCoC6R|xiDw>%y(Bsx=yi9_sAs7_yrAF7~%{oRi@LA<8^yH|<-eVc|x&+sMHeduq!Ys9z~tdt~wi*dp&ti1J3#WmTe; oFch*@3rCMO_1_uI^=})SpT^&%rfuxY!>vNjAMK!UqVfCP62x8T7E?rwtrlA8uj$#a(`7ahA5iQrb&f_oSOnAFq~)(F=XTTnA;Go z&aHvQ!UFn|ntORArcQ0Y{hpyzt+|zNu6Qox#6}o2-0th0{=418cv`<|&OlC9mJey0 zc@`HJ*O7vPqNB!Lg3(^jpar22LyK)bQaEBmtw}q4D)8}It1htYCC&xOc*S{-R@Q@t zDD4<>cb=im7U`c}#JGZm?fSRyFVbnoLy^hF%LAWK%{jomkK~_S3hyxd_<^`DkYqRa zrb

u)fJ&E)r(eX(~}4)2U)*7v*FFg2Q%^iFys2`DwD)I4S#*2Z3X}Z){;IJ-cOh zntefWw?eh6Nq7?!wvi&#SJ}5V< z9p#{0h3Atkvh%!j%skSjT(p$5L8{L?V{T`bL zAz4`CX)*1OV;CY`&T+I1jNs)<=Wa1gf9fGAvz8@x(6#g=3W+6!{CVr{Lo*wouQz*5 zV&T@L)ytA~DGHr8j7H~6-^7ZhP!S=a``S95!NnJ`C4r#u1^!q;AF`&pA9s{+U_NXk zT67&-y`NTRYM^S6FgqwlG8@UIJsQi`AMd%KhcRRqTir}>U)`KCfy|O?s=9CG-W?Bw z#xd|(#E$+Wi<8K}Z9d&thJ^xQ;wJ%P$8^OrJK=u>GLIeIs1bplhn{8A3|p{roT>4{x53;WoKwwDf=}ix z-&<X5iF;J)`(O-VXh#O`DlEuv@w4*DP0C}Hq z0y6ZLY{_!4NIK35(l9Dt3aYct&*GZAdWMRF`euc!3@FG^rDX@YSFdKaibV|7S5tXZ zO45z>S(GOQIIdH4)C@>%H)9B06!aL8T)!!OOeZ)=PN!p1PD%k%98DE78Nb<1uA(@c zEM_`9O5AL(sHayZW4KR=%c<8|1eETY+4+d#P?6ejkJdtL4V$kgEy<+RQz|9Py5)OY zcW333PIqsC&Y=%8DIlg*ax|C(ep$&7gCnS(j=99E)5bUl08GP{Z-yRc8wopt@|dRH zR=tjbo7UIO$9O&9lmnR&Z~y&@+tkz5iA}K|!4ehHh0R=cN+;dr!-VZ*Aj*3x9fLzq zj%PsOfOIV3B7&t8I`gHs=1z2LCvp1##i+a3N<^h(I@9A>huq@gyE2iUg9Mv9rCqxH z+0VQ?fXt+`}%B~lKox_oo~|{0&CPrr+Ky5z&g5%-_a3zur@AjY8q6hrUN_r zB2g>!0iE}ft{Ew~fu2qCStUoU*Ej@;E360=A?6kp?N*1mcr1)reJ1$w+#hK!VC|Jb zUopvA?X!zR#@q`Y{T7-UqX&%~aU!Dq==!X}jY^}|Oux6E->V_0IFf`T1#Zg^b=E>? zW7qH^pG2p`o(wQG>rK-dbf8LvKhyfxWNp(W<0R}P z+eK+{dhGCl8~Uf94WIbO&#wtfV`BiVhsnUY>t#g3soz<5XTQm1OLESD=SGPQemf#N z%ZZVZk$|mvf?ah(NOMXzG7EH`U!lpYy9;f>h)SXVm$y2pnM-$cNeqb)apZLGdhoX7 zsG!$JqM|P3tlmfGp(KD!h;?W;@{-qYM6Z_l^$Yo(mFJ+Ev0CCCBPlDFZZe^W4Au~? zH1&0OIEX#|hEvrn@9&Lb)ycXdVdm7TqL=XRckob=klATEbrU`bY38r{mQ4rjrf1j0 zCH{W*Uv4V|z2W2`Sy!+e&tbAT;>ETh3h`lYBBNo;1>Uy55r)4Kv#Cq>Q{q09v{3A? zOC)w+w}QP1^1{wu z(IRq?@&d+LWr=w#o53{laQM&GGTvTgvom6h2*&V7Wd5uQie6InWJ(-XfbAc6n#RpcYn@nMeEl)|IedF;-V-$*wB zHacI5YHO6-=_pL6aenwKiT#>m7>r+pndoyc-N%|=WpWO!!X^F=pOwnl`>PGk_WB0* z!T_6FKX?cAlk;yi#1xj)G&I^whva<8eH=b7XeJ!-xev?&60RB)m);YOg3qg0u29Zh zT@`MZFh%V|E@2fI50&3Zj6kk~Q5Vbr!zZ>YINF+wRXkJ7y6GQ}bS^FyJ%SySGC5D% z{y9(Jpf72=&d=j@eEgY`Fv(Zv3fY+&g8NP%hwGZZYr#GoPbWM-nQ1Ss)dyG(4S1da z634HAOo6R=a&aZO9eGPc#n%sCVnr5(3oxMbR8N_K4-JgB-kO=Xfm_Mf+&Kw|2ia>ymuh}_dndsj+bz)wtYtDN4giKBC8B_A^ zPnXuJks{%?II|}@1M;h$;;an#UvIUdZSm8hTtTo_>56jJKEG-}d8yD_iY?nNa)uq; z;3P9QmXkb0UAK+@r#|4t^jL1eii%8dQpyXK@*XI_1&ORrC-Pg@_*dwXCh40; zwrL;hXxJC~I|j@u%9N`;H;E-}l8k5+B>Scg1$+W=(Z;)itu^PIEKaeG7xJiS?P7`H z_wff2V@AHFD{0E}1SD_c@C(XP#7t9QI_2eq!&UM(jcp!xQ~BlnXjxKqEOwbCmo2G& zpWOIco$CCp>npxXNh zuiUhA_G`HH^Qd<6odNb_!N+o|?o`H1xm5SEulZR?bh|A5qKZv28S3H@;&Tb}kfsKiS04XunFGL;}SxP+T1?sk%y-Sv$@$mL9>R~x?#e>X1 zW{1Wvwd)U??BTqh!{R)KMgi3|(K~^zdUv?t4i9H(u;6$o0M^R^~s=vxHwAW>?O65C3 zxC#={3TC{x8{_5P#9){Vl@1I7G8v72N` zB2+BLuW!Y$VXqK3>Ow$tL((EO*TxL0S~|;c=;nI^7_tTRjbA<8|FjKjzvPs>YkfTN zm1QZP{}t14+-_CZ6L9AH&?4fn4TNK>D+fzKT3T9+T(Vq45L)YYtt%X;mKBm1T^%h& z++AYoXn&Nge7%&c1>BiYzLtB+{&K&1m;-()ohNM$RFPHL3&@!|%=pN=Ldf0jhmoTI z))!N4b)O;_+)Z@ye(=kl8r>~G2`kd^G^0-YAfsy*r75i)7S}l*iJA7~NwStXpCkcm z`OtP9&nWCPM_bm@vEO?x#x0$H*inMV9~Gq;N(ldKWWRUp{tO`?dtl%tPH7>wfl1yb zPAU35MB0*o1@4D3W+E^f10##;FPJpn(q7&?f$w#1S9pN}1D9q4?&rPH&Kxy0c<*c< ztq~n=Z;764Hh+98e+yqsXB2&r*d~=3J&D++3Nq;w$!K*%Q`K%&iMma*h3KvV@z+N|x%QHi+*U7xbi4rK!fXNQN;vx@6* z!xszXNAMXUS!4X+*T^gkK5eZ^w(iAT=Z)XjGvPnDN!(tD!b#CC$r=ATmA-PZ!(cGB z^C|RV;q468+Hl~Q%1B1#jVP1 z%W+4%=ZHAn?i*UrhO@?kf;{fIE&-*r~0_``JfuMf?w2p!Q#(sv57z0Nz0t0LOl z+lyi(NE4^hv{YDCeVXKz3TewB(#+YFPwBlPfZKftZr?`YMN0y#oW4_j&{&z2A8tLVylJZ;rpbyBv+@hh}2X;IN?l^R1S zZ$*yOD!`-f^Z=EAheaFTG{(I~!1m)wO=DTypklqQLo4HyMcLyT4&F z?hd-M+ElAb8M6e@6K<)E&bfoMwYJpq688YlWu!6cx0A+lw{J8Xl4k{$INL+E3hM^}2H3h=v-oj?fP?3n?q0PFrHh4=S+A=}cQM85#W$=& z?qY(Dlqq!W$Hfz91`-c%^{fc3Eafz0TN;&;MS)^=sVY=GWpb^sDU4&#ZB5%4u;!nu z1HWjz?qG%(F4q$##Xk$J{Q3FjX7=0ZbVG1{T?MiqSitR(q`Xo}5(A4OuCIJT?^DE9B5A?C%FZ&D9IVZ+I~dNT5PL2`2bc%-V!R6h_R4Xwvu3j#VVKdyO=>KY_=dbPjYfzRrL47T8RGIv%I4$Au z#B*N+N8$^Z^x^QKql$~;kF$e*D{KEvy&H6^rIOP+f!vha`FSh{2ZtndY$g1S%P@B` z5R>#*Gr2ahlLRph*ilj}rCryI(8cvYZN#XWeebc+es{xCmqv+tm2^eykwRrWpyx*% z;-BL+q%0&pGxNEz?`l)n<9-GOfie7FJNTU+)V!wm!^POj@` zQ72&}0-5>t0~7Mwlt9dLHTr(M?pNCaQp6lX+Ot&X{uadV_S5taBZeGcoVvqP#~6I; zYApxqFhb@$xg)0x;qP2xhr(#ovQY6!nyc)ekGG1n&S>oy)@Vr0b0X$fG)IN-Fs!xLE(CB4nw-;nqbqwKeOO!LT= ztr>83ua7|`Lh@u@^xPB@1&m^1Dy!~cwmh1B?)AQYmj`cVx{)eT&!7%;8pN;#u1Ksyfs@uh z&dgEhI#Y>Z9}`?PlZhj2uzF{%BVO6dbX!I_Xry#XSz)t29$P-hf&56pL{_i0b!BWJ zo>OUspR*dA6v?qD(4_>$i`ky5q~m^cwG}*u1F}30i(gKYxXC~jL#*!aGd9Y_ zf-YA~^v)vA+@mK9!Cbye?~;L@)IO2#OiZ{kt)152J@|;sTY{LfI82}S80e>BjeN15 zjs2{-l^gX2T%P4E8$?^^xAhBJV^*h1-k`XiR+%6^w#G@-vegzQ6r&oZJmq%YzDP<+ zLg6_Oxe(twIS#@8)#*(W0$fxTqJA28ur)VF~g6wBb~aIU6YK z%(_~ZUgEz1De;&+3La+bjaq_|%Ek(_tpA*A>=oaxrO@pa1BK6AT7dem37OgfqXQI5 z{K^ABxoV)MObAe)zHm?E(Zqp7uiF&lnT$6@K{!b=CRbiQ2^gU2n@h$NH<2wb&!;cI zTE7QsVHV@ElzniBjeTUVt|?B3zu7B-Klwfp`8dzj>p}+is`{~_Zyzd$P^KbNwE$LA zg0XwHA_!KY?d|YHu)4NxoN#`D(`&l=SBrs&qt;4_ceeR~Kr<8(*-Tn?S)M2LNOtO1 z=+d#0&`CehiedoL5CLzJwViTZrVB|xSWc{}{U=QK3?G7bXk@B96*hWGc-jP?X*Nj} zd5V*%cU~HOrOmAdge(yaoFC1tn^y6Hia}uEcq6o=@VFw}MwLF4aZR$bJOzHL&@3g5 zt|_W(u&e?6h;Y}Ohb9Ua9g zu>?>#QJN@!u<$FB2NcN(CWCE4z2u@`$4_5>^a$nYR@Y3LmiSmrzE$BkG_dF0E3R&w zDb9i}bPKHjaw{7ohjnLuW0z~J@~rr+mhHT2R|4u)1;MzjR65U_-1qPv9v)(gxn@#7 zq$DwXb0l>ZpnZEi_0Y1H#z2StX|eUqqjG!f?~B?*`gX$e-P^t0i>U|La4kG33cJ9a z>`NbTv5Et0-)#>bmk4|+!T&EZ{eR!upFbdH;}N4zq88GETIM-8*Df&;MJZ_;Xsu4W zn8+~_g-#b8O{|ntI{M)(4q&2=n@e5J)e7^Qa@zd4B(MELoxtVMv?1oN_3{tfDP#(% zAs`+=JU3Jr{tZh1BEx^ssUG3g*1KnU-h~VYe(Dt>Fnvmd7uyVt=}XQT}Qn378M0_H1aA z0erN3ii@!vcIP%U>n}SY)r(T!L_YjCWdCnID)Iq+p0|sGwCVEQ#KD_Z=l#SDG+~yF z%EMnQS_HJ0pr#Fx;%nxzo4+#+c%1c8GXxirf5HDK!f!*~_A24uLjJ|u>UY+5qYgP# zpM`MYnTh}VFySun?v&@3fOc15UOSh^o)2T7y(R6?cQ~Y9&ZEf2i2zDk9W6LA^EqX* z#)*I0Aca;5Ghn(ob0t|7euvzB-+dKG^*SzspCVMXyo_JvZlreshW|b4gEtd%IdqKB z*tk;uWDct|2VnRE!b8WB6z2M$m3sR6P@4HGTt1o&tX*kKLDEVHk56hpDjjy$O278t zIZo2+8osFOk*YFH!AM4=rm3iZ5rlt;&|m)a&j5Z9QYO&+mBQXhPPi^=V&UqcBhF>k z!=fDuf{(*z+KL9;r`dN)ExjXuzVzP3)4A6EK_Q)jxV$I*U#RNe@QU{S#Z~?-Wp#ih zVbPG3G`(e0j|yhfo5%LsH-^?o_Ya(pgS_1+Dn9ahjXbNuyh00{~u4`|B-P>_R%13ki{t#X%?g zKhTtnWJwE_iU>M-Z`Yk$N`sU)f2Z<#yC2QeF+Sa$aIKc@Bm4T^J1`mgzN#`CgR86sb>8=>9~R-pBru^bNG7y~m!Lhelzw1y}2Q zhlz5P4u?#YaF+7+inOFu#?-?~{Qr5l|A5$=6v0zQ75cOMX8LRT~ODPft$tWvC0Ct zS!nK++`o~ExA_(fGHb2c4!@Txfp<*B4q_ab7u>@6UQrAx zu@Bz~$YxHp{e0lsMZVa{CTGXr24HzE{v(n@^^5q}a_Q337oN1-(SIrU|FOKkgFe%+ znTW#>XuFv4K4OXGKnTyL>dF+^n=uZ0_D{r9x5)11E+;07eJS(KAIFlJ*@|lt#Ar zv`aJ+rCRG<55HOt5*Bf$SK`z*AwVL34`g2~@xGQMt@W(62~!ZM0mu4yvTC7R9GP(4 z9B{Go6;>QxniQp(w=SvDRKO%|Xj^PjrPBnN;BE0}Xz#0-$W*69)5E7ZHYk;(rF-_E z2P6q1(m&jwT#TDm(quSgls+}cAZOSWaL@==gZ19s@aQHKub#O%^@Lx@gHM(k$4#x# zujc!(|HPGK)(VTosbm;HCr1_u!Sw87UZ70jt*O4gV(`w>)KB-RWj~T!C(!Wco=N44 zNAc^At6OP!*!Jza?mUsj8cIC%+r8c8E`0fwDE5+c$91Po`A6 z>*6O?ATIm`5z`!_`1Zom4`ep{^cQ9@2mr*CTjf!?%9A$yB#ogZKtWG&*#J`kFeLH@ zULignnBNJ%h>DF@rPEG1W5Xz!phE(}_DUoq#mu*hxo;`S)jSi4b6ium>8kQ$(K4h5C1ssnbjm z#X#Y=qCd=Ue+@By9f!{#&iow(25Tr~S)b?M9t z8Su8%OaSx-ZK?UZt(5D0o$iPs&m!WF$|3Fad|@U%Bw;I1Y94IdX@m=TZYe zD{hk}o!pjqTwwbvH78^gRWo-ec{veUr#Z z+4^(T&s6dKi(krqHU2=hmkr)nVX*kvCGw&5z%3{cEL4!?!Sy@z9N^uFK_%b+LBz>& zfxDTWpm+3+6kg5>c7CwQRN`pu(HHvqPQh4@&8S$xU)#bTtaHL$jILfq*o1p20M#6w;(4q8N>(`LfU9xVayE=Xn+5Z72x^i-IQWl4O%CtjV zeJysBR8*_RBsW82W2bxO($dmzRW&s={fK90XNj(^+L&l@aX)JRI=EbxQ<9c0e2Jn9 z4@$_KU3j>-OhJfgj7YHY=l8paCUlIaJc`6J4L`F2&E``sNooSnAE+L~5HGMSMmLin zBh5d3Si#({D+ZJ!5;hDDakt4ZAcbUiHi3B;1LT|m6U;Y)V*M~&3oeICZLiD{e6SI8%5s1@1{z6>$mV?y@XXkbzF^sq0fKbw_Q${x;v>M?p`CCE$j|J*j$w9 zgUtwh;*DY|B$nQYwl;Jv>9*cDUIZpOxc-?ejo^G_d)+A!(ydW!vjvM{jX+G(lp?-U zi>#?`qNE5vCq%PY1Xaj`D=6lXy<5lLM+kM?ILVs$QO~?UoHFRU7P+sIOUOL<1<9am zh}{pTD z4JPnE2V{p$m_TVcP4PCJwTmKDZgfU;sM}C;BA%naRiGE0wF`fiJ+U(QD4amlY7K*) zgo#7*k5f0+O7X){UecVcTC(_f$~u7KZVKZm%~h)+o;SA`H7l4<8p|=L!9B9Ofl9w+ zfQu7cT(7T_@r(D?C<_{1=CyfP2M~W88$eL}k$35s@a@~xnKNwT&s-6|UF_@kC((Q) zum$VG#Tqn!9co$K*k6VASFfV<`$}e-fRcnRpuMB`El(u`oAx@@mq=| z2dgpFW$+fgYf%9q0dy1dDTYzTWEN(@>rc)2O&o2FRTHm~60eROh4lE%r%9g~?{EUV zZNs|YtxLw=W}8X)#vBvaEyQH99Eb6&Rm_-PMWF@$BENG4(_(!}2=B^PC-@!WGIAF8bRfMw3^`v!40*&fKPEq4asESq{aFSFMqsP*8N${2Cv^w|6aU^|FfukAE-SjR7~`g-nd*{-L&9gWDei zZBc9>x)k4~0A^ZFma)HJFq;YqP}gujnkIGj2fTcWW7%5lKFVnKB%Vb<$v5jr{lmZQ zR-QTealWA`%($Ra);F&tT~CWKN)^*LFrwv-z|VsP<9vcksP{y^s>>+*!=vKTPdU;U1g z5$G4~cc_3i6`(~_b&grRy6$+#o6*k1x!}tmYiuXtC6Knq_z)j2f5I|-64`ZkcH+yr z8ch||EJV~@Y-v?{4h;EL;rLZ~R(Rn7LL6?ADzSC>CGiRgs7vjM;7jF==FRpc=d@re z*ChNo2xxo#qzyCdVA~VM5_5jtd0kO>Jxl0Yk`eKX1pNGq2v;m-yv*y%4Qe9F%Xa40 z+s+uqU5lW13Y`A-YIjkeaO2#JUOM?-su@(+GdZAJBG3&nV?f|;63sN|9WAXo0un1O zZqf@9)2PTuf-VF_xQ`5SM>@<5!hPi|NX?6XEH~F1NY95H8{y?>(AZ)BmplO0CJw&9 zkzol*`q~tg!GKP(R?`K)vFXoq4JN3+8Vhy!b5&uv9~Rg*`Zf7j3CUb#01eE3gM|g$ zE=O(lM4YY>d*Q79(unNt=g@AM__6s)B{>rm)4_6jBs#^B=xNv$ zZ&|<$;{|6E(*;aWtqZq-?0{{nZ8u>73enx{4_wC;3f`Q9ow4O`{M-@65zv`zzFb17 zkS{;#S97lSFr)S0bt4h&YOdGEJ;)`x{^CjK^7qUuvz5%HMv)(A7a!t1ljGmOtdu5w z$?Y*aGWF5pubQ{CY3Y$1mQ&wk`}?vBz3RT)s2prsbzx5Y(wZM|IgkU`%59I7wX`h! zG7V+B-BAF!{)N|FX<=5|wdVKrfUS;&&%{rd7sEp~E0qU8DDs0l`qAq7*VOR_6Q>f= zMxOx>Keb*dZLz_SvJVW>$ucHJ(8JY$nsa$7Lm5|BA6CMh$58fnv3|Xf-`!dSEWc|% z-D_bupRlk{`~S!yTUfL`9iF)ND$8%XEl~Afc&8&rekgvaX3Mx@xt*+U{MjJ84JK{d zl*65>+AVU`t7 zPq$uP8NjW%(7Svb1#Gde2G9tDt=0-1dH!a#*24W$8!+z;UMYRLxZSC01wz@LugLaw zzBaNyyTi;tuIQa@v19?~v)csST5qT8ktM~*<+M{$RnOO%@%Zaz51XP9*0h5bToW@i;)L2Cl3oFPq^Zu`I5E!Uqn^wgH3;GMo z>SI-K8{69n5Uy6Ctj}^wtz=Mq_f2KdM%LO#PcZ`qO*ze2m7lv5#-RhIpRTzXQ*@qU z@Ac>4lE$TsIaKQ`v>iib`pic_JUwVdz2zN*i33_s1u}ic!xa&@ zJ$@YspKZuT_BB-Yk1~_TXIeS9girC#TNcdWNnmy7@oYc#WgnW$^9%2J3)!j~NsiGr zzW`yXSPw~=vcugduAJYFoNvc>OX#qRFhN7_Rygyem})N;yw?#s2z+p66k08GeZ76v zzV2SEtX-cTfv?rdvKin6j$Ihr_x{Fj_OR{C+H-rttn>Z1E#GrK0T~(edBxlEhhs9V zT>t1HTX=r;_f#{=8H+} z(vlh_-jSVs1m@Z3kZsL@yB`@ok}Osh)*Wtx7HhNXzVN}5xguUfQ5(-JVdov}aP{d2cGk-D&jeb&1WrWHzFsF(E6ImS-VPOWEb8tB{{H;hV~nGDfZK3E`j1y z$*K^4a{hoRgC|Rz|NU_mp&@=0=3^V2u&HKXxxUjC;JmmMmw(vFph6@PmS_E`X zo_PypJTF($N>*8{eDH3az`*sD`1E;zl=e7&iSvhD@r8HQ4`RP;aQFCl+dAmOav1$* zeh>}Y{!Y4a=8Wts%T1QrzFzW^w0NSOgoky|@;AQzA{}eYN43>eEQ@0I^u~i}oa!&R z8IbFDJfb*U;7rLk!>q39x7i)%o^0fs`<8LrQfoim#d6sjbVWD=#DoPhs9*F+eJG32 z8FTk$CUEPAPGmase4@?$4G#l_2>%Zvj*R?(7>%rZ#=3NsoIuMAMd7D)&hy0|&h=E5 zmKGdu8|TJb^nkmA6?}jcDFu43}_~vrHpz!)j=Q6Z0aJ7GI;oO<%y|i?* z#(7psD0)r}A3j#WM@;9~ho5bV`=7{UUX*)P%$+b$Nu;~qjgsvFAsFh4nl2x+ePv)uW|FxgbtUzQP+h zsW6m|9J$d{NtaNWoM9leyaH1pME3Mh>|IN!@O z<$u}qzraJ{7lg8bu7vbA`R|b($kuQ!e6G5F#$|nrY7OZCukPu+ zBF%||HIckm9BLb2a4Qr{8AKrW{DgGXa@F(&yy zZpz>w0LC#iA#J)Lv34Iy_IOt^*HfgJ&T14&5qA;sfmBRPEOV^aGRUKH8U3h;#;Avc zt)bjDBr3#rLnfDO9%2>UeKPt-iwKG6F=8O^y&q_N!puGoK_(YFsl@B}VJ`!4u{7ja<0YcZD`_-72!FX!cGJ_4b8mne8}V!Evk ze*kZoIvwuu@wx$^t(}O522i&b^jX6DWuq2*UqP@|Vp1 zzv>AxNA#Szo~wp;8sSuhLY{nlYki?a{-nUd{SW6hV@I~S zRMg;qy#SKs+0@@_+5`L{KGX5j|EY1hs@aoKLGpx{Q%wzA@X6h`>J%!M~OmQ zCeg~5TQKH2s1gF(g!0#m93!2C@JG??F#~vDRv2O+AA)LEK7so~GI(;c^8Fe7ihe3~ ze^E|!z(0m!eoWgUEoB=TBz*Z^&q2=+@w+(czw4Hhe)gX5x@Ulhz~1Ba^YXyB9MwkI zU}|mC%{~z&CEMeSEr5t!evFWm6u3wG+csM~yxJ)fdTRj^2y}3CCXbp{0!H|8(I0#Q2o8fvMZ1V}O3*`L}I8j(bd9L$`Z!Fw-u z>F+`rCfzzn2D~lEb`#9X4T@c#GSUPBxrg0@W&|vHq0LOPsr#JdMSWjMPc6if9JTyOu01N_^SG7_1R2egQrF+P1cc+@A7f=6~{UU+ModZweVGu`^G z0TcAGhV$CcBmYBC4!#%DXp{4E+wfm!EU%8LGf^0~%LT7YqNmUwFyWaFxL-*<4taBZ z%=qZpLvR=xCYHIdhL1&0N*ChJQD;YW3KHYR{)V15f5h8EDqrWW;Zz0=SXk9S9imwWI=6eT(Sj;GXy5Er5(#Sr`Oc9UVkjQz`@g%R?Xb zVeaAw+l#bw3Ov;Eyc+6HfD<`d@2{vjAO8%K#OnDHx!t8TszueL(IR^KRMo)$rvh*i zkH(?LiDTg4)x8`fS5S(MjMUs8X!iK!4wKkheVtp$7a2Rd4O(YbcOX8?mmndDLs0WdbryHqjGOQAkce`Wylg@;g-3nY@A9iEO~*GPn4sGpwf(!vc0Y$~ z_IJ>4c45f`TYhbn5O@_??>5%ZpNN4YqVO;C;z`3!{rq~e>)rFl`4NO*>&I=gg^IX$ zNl75*9fEnXkuxjVuOB3f9Y$)}VsUa*K`QAY)S|NPr&3#>)tYl4z`WCa)a_GpYkGGU z;RPJ>Z%bu_nAux|afDhOZ!$Xja$fdkABFjKWNM>(HfU!K4Rou-oLEl>Ywz&HYwT{> zR`p-=*h$e{d>ky@O{mun5xv+S8t!}V-Dku1EMj6Tws}sJ?u*uQa%=Uat>tMgRV`N6 zgWifyQq0$Cr}m_CAftKPJlb_<1bp(`9Fe+KD&{ z`5rwf?#1gA`Hw8kJZVTeZym-FOLErA=C$oq5cm@J?YiyUx$Om?yzm}rEwb9Aik_Ow z?Op1{x)u42Tr+)3(Ml3LDU)5q_B8VQqAD?=X@oK?w?y!Y{3U`~sJp`|NX+#8%%njwZ(uO)DxZp09_w7+1AWHi1t9?~vEeap^Z>8eCvT zGpUVft&59Noe7PfyUKaF1M54zE}C#gAxJj&lcBv2vqo4F!KWb9`_}KqQ(}*h0xt}; zN7lW6bEjT(MVyhjoeIrc1dU4^j9IXG-4NxAdMTO6sx{xnw{;P9=KIXPf|V>p^$txi zGj;qblw(^BtN<=`?kJ(F}fmG<~D;|B;}_ZMxtI)f>&Ok_HR1A z8u)OY{MvzM) z%uJu)7_m2eY~Fh}SMm{Hxzxn;hbI8*e0kwZl!;M)_E@%fe$zZjADR|3y%36za2CMh zhh|dKoR`=hELZL$Wz_~ozXYD8@<)?{+snew?=J}RF{Jz`IECLvJ>#yEqy>K`zz0B~aBBbX% z`?_lK;Ez3a)QC!Ay2lGo7XPS*I1?|I%@ZCp#`b4|`PY+W0?m#}oDEuZW&4+(aGwg0 z&SvF8st>o&Bu_DVf}9@VMJ`MXN8*vgCDPrEn5g949>t57ip8qk%8!4GV+jjZO`p!l z&b5zIPE|shvtEpj4O7gifw=ftpdq z>sCpGeNm!cub8g1Q~s$&yD2Y0r60w9-&DX+Jf0iW#uu*W28+0yKkZ$EjGACsFvLkB z-cQqLBuH2^Y#gc1{J^(zYC$7+y4vD!uCP=5(By)oBh}gZ3=M zxa^OiYx-YH`Ar^$6plkZj$*qJv#M=22p+t55Lb_ch*s7(qL5-9t2b?96V17Wjskcq zh^89zEuwCmU<+VAi-q(~^v(hipMdF?+GGYP0cUhRj-RC0?-0*SA_b#Wo>*`Q2;Q_) z&NuVyI4%Vhfw8)I=ZNMxi^95Z125LG=kHWGgD+m1-deR5?%8KEZ*JvYjy@Sl+jS!- zGUv(ARG>&fE1}VPk4YlAdthqlxBftj+szplA8*{1yK9+2$1jYw@BE`ndv(l)C&Nf^ zJh6=Ku!5vYp}L*sl(fzdLAm}uw45%LDjEiU3)Ab>9-D(H7#}~@ zL9`+virFP}_5C!`&Glh$e-BI~nITp*hO?c>A1~P6r(^NcD`b!|&wTAn1XwnG)s54= z8zq*}Ho?D4F5fKy>D4kolnYrlpba5QJS!?UQWiY6lJp$+9zaTj@JJi@QXn9>{;})> zQ;Fw6O3KN_V+_~5oPonqoy-koW2nf0ML%@3tD1U>se-D&2U|KJ34cT5TjIp6Nu8%# z%H5{D?SZH4v#fWKbNCY%wp3sfJ+f|XV7iWqIQx5o1bC&@;&^s+9^g#j<9YltE)~DH zQZMGK3Yo2;J&mbe#f@}j(Z^BQ1QT{>cc+6N$5GBI^#CgS?kC8cxZIqof#TMYzrp*m zXS~lI?#N%bhkp4$#vG_Idf?HF~#{ljkn#p1lSAM5yb}X zYdt<)K=w|?1W2%^U1L?Jd|=;0ha6(GB~M@p;?8;Ut@-r3Hh+5MTwwO(&(JYB#xg9d zD|##(hKu7+kMAE@xhl?XIh?k1ffZ&J_?;|W`uBFm%$M-mAsW1Ph)WJ5;iVQ0%_|eO zF*9G2c!l<$boci6zJy=H6MA+#DJ#r32&LL5L@*MwW0u(pn{muV{WmV&THsr0lU+** zhO0Cjf(qCBozcRNxhL2%19&P-9c8Y>QVm$Hn=o}72Oab*z8M9sAKxH(E>EDu&lPnH zyha^VF?g4cUw7ksBp29MVx@9wfJphNr}yR-7MRc9E&Ue9f>oc{|4cIlpiJ*^p__k? zQM%50FUCoUF0`8F!liUa?nr2l9L&24>~mA#TLzT}Icc241P})fPSFC(CG*KrhbMST$J2`*-($ZV5KFDO%79k- z-o@xy?1l|X5?9&vkhD)9I2i*h{y*N{GAyoUUDu65aCZw5T!VXpCb+u<2<~nTgb>_9 z&|n=Lg1fr~cXx;2?r@5H-`sQWwdOf%o^ACbSJN5YJw}ZhCGUGbPt{8soRE4geG}(> zmTwiQUl(0INWbZT*T3=k`>}4$f@qW?bJI84P3TQnzi@^A;Hqo3X7A%3FS{9jp}h@zli%_Y8wrxP*GhMllMSy?;hb-|x3`T1mgwAe z_?w|mj3A*;pD9J+^SNT6SpWC4S!#Y0+fjhTqANc%q_b*&YjvvQ6lQS|)i# ze!kDEnw$j3#-T0W zc$ws@J0U)IRugFyRTU3G*E5a+Vre@fnAW8l-fMRrJ}U@cVojpd;aso@C}x!;2@X&?9FO@ z9^N}dSc4~@6)3cE;%@4(Dqz|hWoJn>c3YcYTrlraC~PQZsH5bWjg!y4*5tm{zbNzF z%9K>9HFVyJj_=5MA=am5cEDR&kx8#PLZdmzrdKt8@Mu)>Y-*LEaG03)p0J`^XBWP~ zb%pe({#wwKOSc@Q*l8peub7?rrRha!Q`876`#pQVVG&4GmsjqG3NN~UzX!tr7|KuL z@hdzL8Z_8Z_$QKmqa`1g{T9FuHp8rNC=Reg0yzS2<>r_d)6>Jl?Z1pPZmgTsmJGM- zP}gMsMG*VHPbU9Yv!FLMk2%)IWkPQAsj*M8De8j71_W4{tsFFa<4d0sOvt{zs+SqL z6SXmkVyp91!out`o@dRD59bfaHX&*={^-S7%0al>>#p*#bO;uWDAxz9#`k?8I~IAv z_un3qga~|^RO7eE%P#^sH5x&CHay2iAcr{#f4P&Wt?3q5ruCq308ru9_w4ySrx%H8 zb9p7t{6$hmPP{UfKD+LqF&D(EU(UJ{KE9vo%vd6u51VAJ2~xkjapJ?Zf5P;>)mo)D zxNA-_A$kBP6^d6A=!Os6$Lg3fBXxNen0FF9&Wix?Xj)~u3&9<4NLRmnkU9dvCj(NY zNA4q-GDa&E0fX_x_F_rW;e+h3jKQLsO7%-bT~~Y%#ftT#x{7U{i9U*p9$3t{Uuwx` zHg)-1>5&KFk>dZHHDK@q?Ptq6%a|NuU{PKMuab!2ZZUL)|rW zFmn68wZ+Z1q6PjQp!y|8?~-KmP)9YwcmqIAp7k&kp!urIER%O7FYPY_-s);CoySMW z@}`i!KY91!8_A2yo$pj$cN8$HL-kY<3hh(;Sz$X6bf4VK@idzG$+S+W6u%GJpeo85Za-Il`-2 z71U}ViaGHd?l@NCOob0xwf`C|tEl%>El-im_UfA@@B( zmzH8H5rs!^;USUJWcEjBv8t&!OT|VTk+@o-s{0KhwOzn=VKlzvxGP5rW|Alp=EwU^ zIIhbT1qjg)`W(4+%DdeQ_OY4D7s!HxDM7N+Kfd!-1+@j0Lawn8Ez~I#eX;2Jb*V^x z7~!2ut&v-`6V)Ezu6eAnQ|1KPFW-b_QO5|b)Z5n!>~a<3xNqOtySS%&)>PWwC%VQM zC1wRPkdc$KCie;dRCB9?H0?~``EZDR*;seK@#@b~(+;9qTryth5D@C%oloRY zl^lOBTGIMMEntoty2>VfXM9#P+HiSHd_#csHF`iW3!3i2R3NJ_cWtI%sD7DPM;K2_ z!Vfe_{(Wh5UENf!HdulDc{vFK25W z|K5m=4i_-AMWu_{PY|?9aV2tfM^Kn>wtea~kW%YXY1eU%Es_<_Nk)}&ez|Fit7UlN z2LS&}yp25zpQ2c_lC!ozf~A}_+wutQN|pV(Xrsg`{#TqNuO?>_D$gm{BPD?z)|&pO zwgfKzXVJOX6gd(SLYD%6=%5c$gDXOVA?r9fR^{kg-iwaa<1#FKuVbeXij9@EM>yig zPV7{wN|O@Kj!~+_!kWVZI%hU^UsZ{1RjtgNDvq4puh(T~M!wFNqej!mIgtw+e=s29 z_pRUcoi|>mQ}fijd*DDiB$coKY+PM|*p{<&R=bU3y6m*(jSjD=YB8bT@P=b)wDy8F;c!2!GpiFc zmUflmh9FU2%R*~96i)Il*-bBRdQ@W9tGV5L6Ntyb6ZrY*u-~^=t;Jy&Pp)NzyNY+2 zZ($iY%9wBJo%S)&EBb*I4^FJ^{`yfkXJDtLwE$u`Q?{Hqu|Q}CLB=WxRrkJLIPAbO zfhEg}Iak8RjbumR7=iq4KpT*u2b;1ykM`s(sWr~jB^Lq~Jw0E;5cbpUeDR21D|-}l z;dbpi86T3We6=QL(DeDEYS)z0^5X8s7}a%CI2lz$sxLa zcPwo!-6$@ZSTk}&(Cez?{5B8nG<}&*h5ERAFC6GSy6a<8OSt5W3DM6BtTlMmuWI$` z4-NEhHLg-?ias-5pTWbgmAB^s4uSU-x@_2V@x4I9)k?uvD6D{~Zc84dLMGd9*K00P zr@Y%$-$wIThVjx)0trGBWtb(G>4tUu;SHT@=K#6~XFc{4uj){u``PAe_M7S5iUC)hg4N`794^;~gQfzled{_A<4y+8ya*+li?HZbxRT zIrSV{Ldra!hS$h&6zH5g4Nzsxzo{LEP6g_nSNmKU@A$-O93kvW&Pl%Yp;+rDg(l;H zUdt>q(gr{EgHQ?EL+hqIx)aJ_)16#pbCP8cs$M>yQ1|HW5||KT%hQG%rP=Tv6P$f> z_&F?&;#kc!6-1j?k`!Hwf7kX<>&`x1sf~C3oaoYff48D$YvEn5G30biWFvFnq6o8=geF=@X3hm>y zs5I6g%A0b1>V-hSsxLsa5Lrnl7%^$oi(E%x`2~1yyqwQpHdGPdj@FOe3ef!21M|J| z^{a^zoyS|z>Xr+P)b+1(TWfvORgE02Hhu6#eUrq4 zJ}V#ZFOEQ~LJJl;yz7&eDzyQ)#WG7ZYnlwj8hqX5g-z`|ZLtr1N(!Ui`w|~a#K*|$fA8iX+#H~|rg9oU_ zz9yT!1xfCA5Qy$fpnnD|9^^uX^u&70e2dJGM)ZkAEZ)6@tSPFmXj{su%Q0~j2qIR; zUbN+kEZn)`P!)dpaC*=kLnnu=NYO+IGa<4}UJO{XFi(a)_bWsYJ!IT=ntR8(C;xLY z#c9e3k;9))JQf5RA01v!=S@?+i1GwfDfjQkX={&0~eQLjm@c)?0|BqqIf7aZ8 zgX;hJ5jpy%yUPUw`uj8Z8+cxQl?4e2X+My9wAefvaXFxP0`*jaXYX-QtKm|Yh9j_0nXYVRIQ;iEN$^qlZ=;!=UtV=dAEliCR?>JHBrZWSkrD zTsHS}IP>Qu2L$wH4Ly3i@-t<-*I@Ws$d#xFEK(3+Y0Uxp<8s)Eg@%vsML*uxr=IhW z7nWE%FY9DN*jBH>4O(*i|E6&NtA!JQJI+K!=+_~(7-KhDgL~SXe_#w+?2~Ovt%hTC z;hE@9ywJ?Z25EOI309=R#RW;GBBP0VEsw_cWMto%E}u@^bKgFzkH=~UDqhVZ5p~Ik zM+7I^L;=n0)%$l1{v6g=O8R!-_P(w~;g@1%T+Gr$ELyL!J2!Na2I*tFs+sVXk&vE! zQ)BOM)Z!73woM8}T~>2D$6^_lKPO8E|A9FaQgno*Chs=*HzPx9?fR@J-`a+-o7m)eCs=~jU@@vcwPsJRBgD*7~a~}8a-XI=V zKt-|+g9o$IOS|?1&qHEGo^z`x0Q5ui@@o zxcMK+#Fs6&UeLp98{H(FG5a)QwA98Y~;pPs+XoEG% z9;Q|5x%d}r#l+I}!(x*FsCC%4r3lkTd|N)er>t2kxyd=YK@L+b+(jy`!_th zo%l!!e7~587H&MT81Zo~U-hmrYS(<+B12u1_``ot4}gz+XsMG$B_s^V(km`L z;ii9MCCUUOl+I`yUE8m2Lmeypv~^HYMy?n&8nlqT?(YX}Z{U48bw!s;Kg`JRJpo-$ zot7mkR@%y*9vPWi{>M%FJKCN%D{j&2|3yM^2I;`| zNIC=rY&y0RNQ30czrvdS%^nDJn^j!=@gY24PwtX9kMxBqhyQR7)Z!042UrF)%JMgt zUo*+qeTqbnBnS7%w!{#pSf;R3wQ-?6t{-yu5_CPpC9j$uXsgiI#8ZFKNpB+_)mnYT zLew}SNi}ewjvpXKfwnK!o9X=X+ppH0+xEB}4`zmVsiP*a4TD_iSnwX$Daj%2wy|5S zs9w->PuXh0zLjg~LxM}#{D~jX~(YYImhewmTQ=X!UgSeUmu2}hK>U;u_$y9vW zmeFy95ByZxyH8_V3Ju|*BYLgjoGid;OF}B_;X_?921Wg0-nGmd8G|tl>T`+);3KPE8jLlY`J|+`}dUsN-XS5&O{ZXoc4!^Aj7+Ef?+!!XOG-w{i zgH%G$H6?&Gt-ija+G(pJWVqqWKkoJ#ZeDpb3}9dWd1q%bB2d8G6RCbMV$9@G5_5s( z3fVue_6ni7Ow1G*i>$wf)a@`gH`bv?l7fa)j5#B!)j#q4`05~z4m6r%7oz{9s`xiM zqrYNjA6_@j&#)5JT4s$dJsSCG^v5fMI(@yGdu= zKr*3j8{PkL9XtY~YLGXHj0#}BwvV6quP>hwGL7RoCTlfSZwgVXTBH8QHKZFBU-5k} zh90%RPo61F1GyOmb?q~)=K1;gVbNa^CvNp8vB#av{m>KJ`M8eDG``5k6_8H6cemLm z(&ki>UYvQHcc`Z{{N2kvKz(vsKyYAJ+T6i1_F7vGkW}pQJEQZ+``AK}Kl_J^>ho0| z#%{E=cXdu+>hdqC-G49O2_{1VZrHwCk0<2Hqv+h&>=n&kgW*;G!K?uU)L53&zWv(& zUQ7H2Ad%V)x3o?>UN4b!tBAsWERZb5`kwybA9IviWzVUQoReG;i5BQWz&NY%*U+7S zs}*_ewxC^BsQc~#;*99tx!CZGw_Uh#4SzgN$VuX7JOKt+YJYcBbP%Yhtj&KekxB@7 z-(Vkh9g#M$7zUK5Kp!OffwIhz8$bzKPVjI}nP|j6r>2;;d#>;aOf9cNG&Y$}sddEK zUebz%S6JhGN+=CQj|7#5RWfeI?DgjPgAYaG7n$!m0cC={d6xeex4ElfgoK4N^f)<_ zMS%{Jre3~>S^w4otZhShWFhzXvoeR=NCQL>A>6G%t;5Co^- z_T*;ks}geEWdIXT&~=up;O9)lXV(=X@HFnH{WZIB<6sCPy57&ib*AEK5z3JF1Cm5^|>;LTgA)jhk+KMw@khbqDCUkT4xkQ5oAOA{GVkdxelrq}rX3_m%& z88do7ElF|q@w!*gTNiu0^g~XGO2uyD?d^rZgB0N#@OpUJS2bIapjO;Dwx!E&P2?%i zZ=ym4{i~h*o*smCN9LDe*_&T^hY&4>2gdn*<8iU=L;uAk3nWe|0-DJN`OgDiz#_>- zXU_@PB0U8%axi+qoI^!EY+sngFsj6WDFDy3X$EY8pez0p^L)QQ(|1tk)JH;${B-cY*0qDMS{ zNf0BkK5U#4`{fXDzQu5c^@t_uc!10~+t$6Hf#_Xt87VlHc3PF4v)L$U|H|{Rm`VTr;BM zg_8EjF5QIUyx(t3*1*{7=O{Tg# z$!Pt!Is0iqXs}mrRJ7a=D9$K59n=rHCyS3wl6lxy7XQ7UDm<5=Ax?{Y~D@gSdM7vldz+tDN=?hp!({&r;#Nsbbod*tVsR zTg}6I6-&J~;<=ZIxX48ht$MN@RjUERPr-MM;nRZ(?xe@E%q#0+6yVcnjahdsZl zfIMv!498QKkn@T92dx@RcT%Z2kn2Sb2Bdq&*@GZNu}964<9Hk}-NO2ui;Iad&aHQl zmLqNmo3_Rg-}qM5D<$1|{U{LgXsYH`_QJ&oqSN+!cpU?Wz`xMx2F`jS zWW)YoQ>-EmWE7^y~X_Zy=GyZJ40eu)vMG<6gV!Orn!JK6||_b%Vp;HC7Y#y zj1$kzU8lgWt?L~DnhG2mPPSR1wMsk*YW0$jF{|X{O=j`}MMeFRrwtZX4v9<^(=h_EN6wQ1r@R}>Iq;2XKn^#$-Cxx!& zrd+(d?9Zp`+LIrcNqP+>yG%N4jIsXJotB9@>zpiAJx211&FOTzk%kV+`Llg7TuJkm zDJ}mCL`ZaWEcrA^x$NHq7m z*mrnr1lzd|Y2@vWr@XGo&XT06!LA7^bgSgn<{n#x{%A3I{!Dk78ivO&E~BEO2R=iQ z$!OP=$qTVOM>M1BzCdn3{>}LOaod(oexH8QPhAG+*YLWzoSg0KN&DL(J(f{=I%`go z-;f|28zFy`pfm`}gw&d1Nf9d#JJa)po7ZIV-Vg}{OSM>)jEI3&85MzGFPZPMp0eU) zRvMxz#Hjg8 zy>F<}l(w?LLb@$_Hon{74OnTKs%bUNE^wGmpGU^rs6I-Sl)P~j_w+CDR`@(!2+7$( zY`5Zcr|9IP2-PW(Wmf0*1-q5ERikp)=p6b2W^!sy0>q3Q)GAGAz|)OX?3Xxbk^51S zvjMbm6OUzu(!Ri`Zr0aKFGo5Zg{zkr>Y^ri)xc>d5<3Y!3c`qSZBlbZ`jou^D|$G# z@0@{zN9CUHfH;e}z3E448l&|yXoq}inW!<&q_@Ee1y zC6~78V(j5jxBL zw#Dw-2mqU%uvF#WtTp_#yoQz~wjoXUw_H(kf|R6zH@P5h=sQyAu#mv=JWMs{KfrWT zYSxB0Vn}~xg={w{mH>$MO1_uxejD`*Bzji7`xU$a6y=}l#eEK6zgFQHRVn!zxx*(C zYPQ;X2b`i6sQu$Q35*E6usDCu|BMpnDWkY6&~JgbmVUcAtN+66!Gxm8hY}n;PNJHk z6T)gh)Z4WxN=NsX@av=Rm!L*>;Qd&K9z=FYxVS}j{{?TzUAaDeVhCYCE~hGCX9YVj zexi@5z6~-@y=8P(3fqp6*A^Qm8?CTsB)G4_*OlUr`sl9y0#W&PHie!1U2qIi4#o2c zMO9(>)({M0^n40lX> zJ>Hf;I!*i9y9V6t20$j<#mLAY=^bB}1YZI5-cQDjZG9EuKq_Lw4u@g#1Mix?9%m}rjFCdsP7@&v;S$0z7CO)z9-~&L+ogg zeta-$!!KcJ#^eR55qQ*(zxaD%)~luL`8-YinwM8x*M069{aKTY-$%x%VQy7tD7~9)Ka|!C-zu zLO)6WTy-c!J8TkD4v5(h=B^4vOUpY7r3X%z3&^Z*!xz`qf@{J!U(QWgu3p~CT zB{Udw6q028P*{R1Wlcb!OHr|j1`JZHFb z?Wd5!uZD33^5Dy;-u53XCbbSUd7j^?>o?q@LiQMqNE|xr{5Y9+B6H$!iXPi02Xcm1 zel}QT2*w^*E;BsnkM`UgeuZf50ENv*)#vv5qMKX^uXXslU)OY7sn?tz8egXilQmxN z1|55-oQki7D9C&ht+D}E&79WFl{@Cs)c7|42tc$XqHI7UC}Bp<_&&An*QW%BeqQ(Y zR+~?w*pU5PW$u-_wpmQABp1xG8@#L*g*}}W7@C5-barGl44r99LTU2B?kTqX%z(_W z8*08^uyMok(sIV4^)gCAr2ZCKiL>OS;1x<=me9;x=I?WCSTfY8<}q2^#>N23lN3HQ};ukO{3=*xescM5>CM!H$p<9`@H$2bzlBX zRbpLJ@aQJ}@`tF3O4NaQ8E{yc$G(u>*WVC{f~p0 zIf{kHkYSCPA7+hX@Ape{u4|8FP^KvQbup!!A3q2F^sSvJXxSmL_~DE7fQhC zIk9~j!-xRnsGvl=#Sf+I=HcMrD4cXXJv{XG@$t#%0YC9jqJ%_5o&kNhWv6`r2q%E@ z4r|i;K@d5Wn|XF+GGu38v3u!DnZR%1DuOCFBa!NAjX^)#L;J)%a@FhS(x3S$~x|WU155Bi8%R_ zZ(gIca_9)z2 zR(4yyMOX}0v8%3Bi?D@U{|md!Q(kd*pCnd-3COS5-L-DUh$@@2AZ8q?qOPZn@s6dr zz)v6wsp=B^n{g5l6!ck!&jJT0?8V`WJn8NN-|-;~;+#h8YuwHTH9vMiZP7U7q7r+* z+FWXkqIHx;Qv+Qk2x)1m9{z1^nOA>GAzX~QzJ z{C2^GWFd>@VQS^U=QUXqHg`wlU$)t83WqT;{Mp}lBc3g!?%#Qj9doW6q^X%C9jB=s zKCvlx3%ph>T)qqDn!IFd>omBPALCCl_;|&9Q)=;ZusOtAc?gBYzQdndKOIrUDno9+ zFC?0Y?kKF`mQICnGRVuv;87@pR$bka5k{bKc=AhnF)cseLz(-CP30bhuF|o&J7kOw zkI)1stOK1n8^ViS4vXNnf#a&qSlhF11CeIr28K?31z5q!VDSst{pJDj!j)zA;_J-L zNMUcavtRs%R2)~N{rAooOnLA>+M`@I)ySrlR`yoTMMsJc(;fC$Bx24)iMLysn;Xt zFs~08GUJ=?tjge*7no9!J!4=3qqf8juyiw|p~E!eMvHW`XW36^WNzGt4sT_ix!%g< z%kbCTTJbCWGF~+K?x>?=#*T=<>(cb<6m2FKYA_m=bfLFSSVbQfk~yE#z~J!C(u|?A z!`rD#>^B-cG2DC!a+-{5TGyApxNpmHHWXnV9M6L9E|FPdTn*X-Dwh)JH9bxk_!#(L zC7&iLMzcSmQt=9U?q%7unZv~5EP;So^qO7Jq3XKzW61Sye7QXSb$SaV=x9<)cIhOa z83#6ELx5AMQEGnpA>t|1UHjcVnXj&~PY?;4OoA$997siGyH97_yu1?3yZ%UgH_3t_dCbt6 zeN}LfH;G9^C#LW93r))nAK}_A|G@U1mZrdV40gY{Otx}aKi>YwijC)tg$3UsVR9(d zhhncOOro-bMw|~#KUdzYHK?(a2xxt=6PxeZtYc=}TR)#X&?S*tcnze2?`-z+H& zEE6s&SyDJ{*EdOb?vFKl7pfl@yCEfZJGYOrNXs#+4>WhQldnlTo!ADzdj;AUGp*uq zTjq3U$t@kIcfcgez^d(4#g=ctL7@x<|~KFR@Czg(&WuY z8>{LjiC;OxeY}yZVU^tOrCshz5l{+Anwo3-q@1|W?4R&D=?I6o9PXDFGS_UYc8n}D z_9uD-aR|6dTO4d7=|;vUXZxtlI&+70q~kN)^j89}?8nL$y#u^wn4Xiiq8Fj#)N>sq z)){+_NNqa1D6`b74T*RkaqXCQifq*&gUe&vuj5=@4#nPSgygH-v} zS4WcNjQ?DUD0H|=NqF~2)8^_Urp)~XhSiw4e=6P<(d1Zj(I=kn)qEk9OM<4xIk;Z< z8vm+;TCTHg*Se5WU(JTsn++cj;JJ;DQB;SnWJN?`+L7tO020*N+Pbg;os2h8nSedZ zaWgMImsy!L&@E?DY;i#%-Qr&HQ$L?)k?ajYE0Wk5E(<_;(hc<)iB zfa&pE7xu&&KZE#fL_)P1tll8N{?n@!#jKFNhfg)yPC2XD;t4+&&Hm>Eg*OM7F@3Gs*Hz~t{y#4j8L2&7;v9B=y83ptT|v2YWMC&WJbHlnh>jkZ zs!-Eyk)vL1;@{mE@iu7VH7e&1wOJExuRv3t*0!7;c4g~if9FiW_`%=@(?`2pc}it# z*Q(arpg+_-yt2b;rf)Xq3Ex)lwUxUIG>s3ec1TzTgRv&;`?l+s|R3toD;4buX)@XqiymNADl5c;qK)DYAs6Fc)F(f84G$G1%6$ob7_5|D>}2tB z6&sKCJdA8p(QAiZ341N5 z>gCxMQeq<^VyDD@nuCCz&MZe3vr)l)%lS{`J|Czje21@P8uLY=P?47#21*iv{!XtI zwXp+#iNf3w$aktfOcuK(@Qp9ZaRUt*K6=3E?1!~XI|H}j8p&n3`?3W)4;+Ihh!7fY z>FCF5I9yG?vQ(%?-6$(6IL?Y;=kYlZc0VQxoB!k+zx@4UpKmF8X?ED)`MCrUv#GK_ zeGf~i>&#_c_)^b&tROp-Bd)wW^zhm8u~e4D+YvUB$+QO%2kRMoAqul=53Fg&Hpy*3 z4td2=kh?($N}OK_47TlTzBP;IO=XqGjjmQ`eo!|D{;J0|MjoVRlxfTFElh`o>Q3mu z6ZeFRWY_J+BEK|?%8bcz>F~HB;}@Y7zPbLa7LmxAW{Y?8T+ix^U4BkA)`3hQUC~p3 z%AMd3xp0aM$Ye3Q&*(_)pV1K>T{KnN5Od9JqG4U3auzPM9;-1NDeZ=-lW+hnXdBcZQc^eWTZ(sWDzYpR7@dKYeVLu9jHF)9opLJ(ZkHThz_Zd={SJ7Op(5Mhx6IK*5{zT zu+w(C8jCMztO+)WCspsgvB_(e_D)iHF8AZ-Kr13w_X(qpf2U;8Id1=Gj(OG<^DzpJ z<^`6MUs^rKEfbAMAuMp4^L7s>kLc{f&hSXovo@bPN~}F3M1hL7_Grfm3c0qe8PV`0 z5gS|Ap^m4R=+y@@g@=}}bz=}?`GbHC8t@S$WpX$5s7NmgP)n>iBfR3`hu^S&GXx34 zW{@>eqnFV|@f9;+ZY#k&vnw)jwGs?zigwqtu`PN_Lqb)>J*8I0H?iN?25i%8>?U&SUmFko4EJ#Lp(VJ!Ip9fvDYw z&_Qgm_RXag%gc&`=2B`c%t}^^e7TaZtVi(MUtfV8&!T?ErX@ue?Nz;S>@BgBm9a$7 z_O***)Ss$gmYMA5Vi?T_nVIxm#p2!Kag74IcKE{^!4y8G7X7b4h988bLDv&a<*KRn z#sUF>ySe-G8XQ{82q5NC&^Vq4!|jGm3Mlr+Tuu`?KG&5_3e+W>(HoxDJ9t0txg~t2 zcX#ke#ET6@F%~&wzo=ub`8*OfMYq9Q?Z%dYt4mGyz8>}S;r75=Dy!)9RJ7EBq)-@u zRg2)IEOjGGHZ%my`4i*bNRi)cXbwJ>gi-FUK&ryiWLzA$PsU{oXNH2fcXQoitdjg= zE~8y_;Jl!JA)N9?S+YjY2x>Pqo{gg81dsD26pg(7?Hs%-^Hif$g+-I^i?``Bc?QEU z3B^Z{fLsl?kz!NRx^HXO^A(;S0SHG#{KK+MzHlIq1 zpkhtz+ufsN@E=U~Rl%pVR^S9fs=+1}Y62&r7BfwAMc0%K7q?-}18o&WfJLMyP@tLN zppNyl$Erofw?+uYwrLs;FD+)6bwl$?hFDbbwH(osm-!x-<}id{j4#QI+_#V542g|6 zxHU4nLFAMUO9I+BvP$KWaa#P+Sn3n5QvtMFlEL?&adI@`vDDK{VaQA1w_JKi6JEsx z4-!oE+o1B40w^7w!>-i0b3WXkSig%?^}`P?1>Vi&Y^>i`1k`by?i^K(^UgTt&hyL{ z-_?j^RSD=ZriBVE%n&&DssjhS*k`eW*1Dj5mU;`%sWQUKSoW2VTQJUFTL89m&Z6G7a`-ml9sRhB$y z&gx31_5(3TU-`aoHvLBPG{flVy9b-=1~nfK5!1)HJjvKDYZLFIYgQ!i*KEA~!g^+{ zr}VT?Ra~>&cPLL8X-Dl2;|`*QAzlAkCAeJ-ayX$H5-VmhN1k?yOP?2FzVN6;t!D-dvSOJ^ui~2|s8J-a9tnqJW)O7E6 z66d`1>xM89C^(3TGKnMB!-w6U@Jex;**#TA6kL`cWMsW>uQ(jTLSocIvo#UC4CrFh zcU8H!AyVD1Z<9+Q&rSCive_oHcoR<>TqVYIu?$}@ha$dj)D06GT-kC^EEg^vO(k<- z+`W!)V67km?;;U@-fnnfcM>r);3(vp+z{~gVdEMj5ySRyd}#nn!VUGz+D8$sD6Dgrmp^>YXx`AjJQ%0rBcSns-@B~^eWeI7j4VGA@s3pUKh7w~t_H{vV zO2#t*&Od`ZD3g%;9?@eR+H@tajL8t3y;kKD@VZEmr&=%!cI7CEIVZA^qL`SIz+LWl zTYSk%lSd9;>2-{18DA$Oe`_53Mu;b4M?P*!-m-Tb2u`P+COpSvKw){B{VwNvtvR zTA3VQ(O(hwO;_&|nnfv8EMdy7y_fwMq66Pld(Yz!FB&z+OA8ZvBGTSQ%ApWmf1fBE zc++h;8Wj39r{A0`i4i$GxqW${bwhqou{w0xt1xbJl4@luwb^^5`Qk_F5;=zY+PUIZ zA#ZA3fy%}8sz;qR?X6(~9&dQMZY6q8>HY=3(lJZJ3MJi*bJqoo0@-1igDpw#UL#|g zEtX+;Z(<|bH>-FxZOBT^@c@ELZCy#0mTtW{L|ljL(1B%$F@=v^f_5gQ;r8Y2DXD@< zLOasWIj>~l_SEj^4HX+Hp$%m%;*RqmZg^2Hw?jNet*Y!zQwN;=j`SHICG%wBrzanPXJC?P~3v(`-8eC9&sc_pd8mxhR zeLf5bs(!%9-I+3`V#->;4k}X-n!C~jgVXHkSBH^J^ywCDFCC>a2E^de4DeH?Y&>B5 zg)pDae)8~OyBWOZftBIix*&^uM=r19=Q7C+)Mdim<@}W-otfZhZa71L%s<0n)cO3a zlile)$#WKc&2en`Jgri`pVwf^o)-rPXSFL5wxS}YK2hkS6Re8C#xwDhJ$3b0yx+3I zg&oR}@u`ecj(>_qz=X&hE2t3;Vc(9pk@*!8B%i%1i zdBh(h`q&YRcG>>KLLTiE0!{5kp=>a$+S#sh8Z_9$;gHX*`0e(^ThaD`-Tm=NG=j5P zwiRW;rL8q%MCanPUjHy!>M8fZQj` zH*m+I+y+%=4U#0+=%g;%qwo{6Xk}BL4~=(BEJjt&T6`7>Iri-4^77#fT;TXU1zg)c z9-IO%CIX?(4@#X1Laue-!TXVFdvUxYFCs>;poUdQgn2$Jv%5qUB|eq*vXuwy1vy`+rlv+Q_*k}PT@sd*t|xDv}4K)72roRed%dh{mA%n;lTxiR_O z3w6RqOf!dCaXwaNGL3K)vo+zhpjXih;c`Fr!|$k)^bGR~+-oL*RWqRgr1oWZ6l11| zs{jpVu+K=?OOA0Ds8C&)2L1l8FvCUFZtbC?Y%pcdTd=0A96~9^Jz51tuAR=Cv?_+h$Sg*EMoo?8i+dez0da;Iv zFR4ioT1HAoylbJ`_;pmy&|d(+NIs%{qocC!BN-Yj>;{=ag=v zFXjl*S{J>~e53OV(gw&77Zq^Lyhw5-BHXpKxmRav>#^~#U?QbgTr14M%Y@(OB8f?Mdz_67Wf50c<~R5$57>N~!21udCM z644ji5m*hF-8(r4FyG#3S%Y19LPG^RbT?>Xg!y)rw7l9BU+%*ncg~N)1L1SVMb2gF z?~J*OOIF9@=OBkm8`Z9nz2U~iGbzCt2ah(MCGkotL$Ro2r6$w)3y)%TIsv5>AJZ;3 z&J4deuem8+-~={5O8DEPBiF%cAWNGxk)IHky-+1T#X1)0?+h&<{?FhNiH2wWZ~wXMo~Guyl-axcpMJ7bWMC zWu6+Emb+YO(5yz?rVv_0v$aH9!QbLnE*1{_;C_rf*%njr4zE1ol#HBidCy+7-Sjl? zVlRFayBc9TT&=O)^%bLFu}5z~gPvemprs%(9Dl6V+`(j$HyR8Tac4DQd#G~W*?7Q+gADkz0#d<$HHdAHr?Tkfy#m% z_5SuZhP8<47FIaRlp;$~MkZT#VROS*7RFyBEmK@Ps)u2Jh%8sMhTH$vKMmWSyqn

(7WTB(v;k;jeKq`@~*?sYI}f^#0+whbMQK%qx&hA1`@W;_xNuj zTGis7N)m2?08zDFZi7?GQ9!e@vic7FBn1G%VRmb7Ul$DjK7*ecHMsBF(ww|8eri6v zs@?No&+ubQtfK1b?1V)|j;YSgcR7A`7_I1l%5l@KDnmN|i?r5i{VTDp|96)W5_uUl z=Qbil3Eff(f0tJ5!}!WXp*o+{%pvY9hCEdk6sNu}a-@wfr{Z-HAs5{=Z3JwF6Vma+ zdOzDQn{Q z!)?pf!wYG0@rH|LCdF4p8otTZ9}XV%ANW6#7yFY{`|mPsZ^wuxH`Vy{m+7_P|H`=K zZmJP2FQ3>HjcC+RM?L=eyGZ~pRBWKS6OA~!Zo=&8jSCFE-Y+WX({+^WN3uAYl2{Rc z5_EP&V7=@BRBK=ngXoz&(29%zqw>a6_S%9G&+v;GPf;h6?}wB4QZc~^ghKFO4C-~u z_ipPFD@#-xYfk<73hIs8OXj~uiFF`)bpYrzMomlyZ}hQ|Vi$Fq(ugo7IK2l7&aad zI?}3v5nN{#26k}q1(p_1sFeIG{gX0+L1YFLZYkiAf1zh&8KQ2FyJQn#DoXUY7`FIgB`902mmM5~ml}_} zGfl)fES)y86*19=tb5}GYru0&qf7l)%MJxA`?#IDpf)hIo+jU%;Efy0(mKNoc`XHg zf8FtTWw1Jna>S)}W<~xT#m}p1W7(8H&GH?4{CIyjG>q@{_l~(K7!={u!K1z>Q%bbo zguw?j4Ec-9TnvE|hz8FvK1QyO%(Q!|IGQ!ANaydv6IV8Dw%w_Z24 zWAl2kGKythumMsO=(8gJR=}vSMOUH}y8LcT&gjV|rq`JXy_akKjQCg8d`ZAkwZEJZ%rH8s ztwZya^?t`8#8fq325-z+KjnKr284nIc;Jm@Rh-}q7mbZW5Ce1Nu&6-_z*|R++mV#% z%l?XiU}7u)LgXti;@gq9YX(L+FP$qy0^v#@m(Zykb~~}1ZC3n&$F<%~f;u=CS~N9!mfRGI`RgCS6Nq0{-B?)wiBx~@ssS@PtbP7gxOKJBAOwhM zH}{clx|V{$>!-9ELX6)Q2$qGurYs(AQuJw3AGpV;DD)!XnfpTL01$8wkV#C|N0LY@ zk1S3k5h0I?$v>hvV50wh*G*-aSxTU-;Ne)Q}7W>39h92I<6Ob!)^@LorB`qx3 zvl^azbgNuy&suVOgDxVv6oQ`{Azj&~Jr(CcMi4yZ2XeS_WCEU=wUJWrUwKh~+gg!M zk`l46N=j$^H7bHtujJh?owM`d^{TGzm3wOaOjQpF6%iM#KsZv~Ps>uk_Fo_#5z!~dCBR?jxa zuWbv+_oswZlQOSyKD>2jWHDSB$zd`oZNK*H5shx2rKo+lriCl{I_`|2cwt8eEh=`x zVc8E)-K*$1ceTLFBpg>*TzKT@;`D41wEX({1@}?xb@=)T0TtE9W6gF1)l<|7s$QHD zQDgQJNIcnVUnZ_c6X~gK{dm?mip7^`PPbtJA7_nNkw|=tB6eP^VTW4Iib1zL9((2n zxiUnC;SVaZ>L}>EeBsCPEO?rN>So3p)64!olFI#P}{5| z_w#-iJ-T=^&_n9BY16ONaNqS|=WOy@t*qR|SYM3qY^1>yzGwzp_b95;zw3-|wBNmG z2CU@#jt6AUg5Dnnu~)0Rj`pD$s@owiZR8J#44=!EM(YIw*y{joQi!O!U&u6vj(_SwH`Fbx?B@>a`;4B+Jm_IG~9EA2#*q=a4we@kgQo--jJ z&T8ko+%*KStKmpiIr6VFd$eloKq4W$8Dyac1mQPdUek%?VQ!jzz#{2$gJwKH0DXY}*9cC&( zDyOOn7_KrumXLdLn7hJJZxz#PAY1x{fw?JchDI)=Lk;iI{dr&sA2~NWZ80Ptb`0k2V_x?n5sT6E{VR#llaSuR2PYM$Mr%I0wsPrW4 zWWTo|v#iTMg26n5%kTVPcl00np#B}_KlS2m14eq};aOvy7leplal8(6f~K$~&^9~O zIU@Oc$I=Go6`(en0-2YN)(~n&JarXkBD@bchP!OS0HhQ2iWd9d{mW(oL_6P3y~U@a z597UirF0C6(SZ4+Kj^mtN}%GZqJVMQI+}-*)3^1o>{eZUD7GMjWjV4O) zfT%d!nDrB1>gQJkpgV)w+&8bvYU@8=5CXnHwNr&OkYc;?-h7Yd zb^j^9mHr)@{xj#_zgol)+G{_xQ++8PVBZoMsCL8LG9|Z;lEqx6o%=rzD@gQl`xqmqeYYmLi;m07U_R`2cFst4v^7_Z#c;jo<#8t#I8mV=zMzXl z`@4Z5W783j1uZB}TP8OEKbL$#_M7+dN6GVkk&OX1bf!yJKe)g)4x;8A4h;~b&k{e5 z$mQ+nE~-eub6r$qT=J`n?3nu-4<2fpk7$HphTt;yN5| zJ33J0i3lLoX;`~adzQ87xQ8moO94~g9bGst-y44)65CT)P+aj*>1D44{?j>P#^y5O zQXH$(>85gC#xu?-vcTw$d85%t3=! zWRR46Pr*!rugfi8QAwP=T%*SCgnmnvOCiE*%5&bI*~ zC4jAy#Gv9M1JMeg@g}F^S%5e%JN0 z1!Ts2mg7}j7zJM0xX~8arPGW*C)NDhmtN**{ngb89D_5%i>p`}oXx;BjVkfYPM-H7 z2KdjhXC4yG6S#?1k?uYHP`X8<7)j?{?sNJWeDcwoAD;Q}UNghN+75WSk6>ZXKceUn zl$E#0GoO(By|C@^kKenH7Ua?nOb(dHNfRG&N`d=QhMfN;E8hj*b>~E1h-BRK>+?OQ zDNvc&Ak2B=JL$c9_gZc_p8T$xl(RLDx@GbFi}=L+z2721Ap8J8&AB6Izq_P7|0kuF zZufZ4v$(D2V$mBBj*$;0-($IDQBb#V;Q8By=M5hI`f7Ja@npOoehWm4UR4cD`exBP z^I%_>)845teLBy!;X#?Zb5!sN_ZRWei{EWjCnWv= z+fTbfZK6~IzuouR-;p)Zx_RaKHNeyAHH~h#Ez0(li$!(Ngq#L}A?mdP+|KEK8r8U3 zT*I&WGLPr9eT&BTwR;#~`1prvic#MHz%i+W*h*(Yea~wA&eQ99m*%c*D=}LGvJX<{ z7HYq*Hya%`He&l7r$vJE%c}-o1KS*Drg}893f4W=8NQr0)8mMw?Y_U+z4t!ie!P~o zCJcL@hn97SLRyiS*$lR~O{5W>@cTBxvN^>1&@RF%(rjKaaCkka;&A?X;M4>aLF*iU zyjI9C>e7Lp@VPaWW_HS7EY}(8s9wUQ6mosmy;GB@8}R}-|NJF^Txy#0$vZ?sdHX6hPq6h z=fDj*eBkeKkn?MhCtn7$9Ri)%fTdSCL0QDpr16ZUw$-r0Xd`GD?jb`M`RunsBn&?q zHM(0;B!#+Ynsg*yUPN3s-(P3-c54Auipy`R)O-TqZt#>R>L3%uu<$m3{D5j{!rgkKcD&iQ~&B z>PYb6z`92fkm3xW*D9r_8Ff(Wjy9#JHhp{lZyLg?RUw?q`#rQX)6YnaFzCM<<)~52 zqCqVr=9x@nAua?dZpMrIp(sSeB$@7B;0w2K5}HJO`(K^PoSvPi&ZsapC0-8SF>_d# zAh33YWQrk%YmG>!c^>5LJ<7*xz$#1lAaP;XhWem;aIkB4wsE-G`zZRFwa6>227( z8XRd+53zN>_Z>*c4sHiKe{FUj<|XgoB@ zgHA#v2Wx2Toi6cxuodgnEIU%tDLYEq-Rsn>1A~L_*-&8Uf2C3}CtuCA;S^h>$0qTL z4?;*Esk}A~ObUkBT0Gm_popvs68SJe=PQLRDiZd6Igp9q1PLb}&X;^Bbs8yV98&x9 z#8?;jld|0}0!1Yn*9Yca$CK}OcpiY^ZV4NVx~_-T3lN$)%IQ~65jVS?+j|&UHedQS{9OY?xUz*2Y+;_GKxwN&xjB7kS@dCu27gNodle1?~pFdoz^m$IBZAlSYq)8q6FeE1NaHFc@60F%xwE zG8~2swl7Cm&13Imo)u3yqUi1Pi@{U`a5Bf9N+1oocL>(U1)-zb&$iCDKx9Z)O_D>7 z#PGSKLTgoVPay5W(UWgmvCAgO+ku`=@jLyqvkJ#zqM)rnQykDqRY5QeE*G&Fm*S8jRba{q-Jdi{^keqbvEAKe7Z-sg!5t_OVhl$UeH zmQgT4=VsXsyV2T)mK5YRy$*P1Z4$e9lOV0LfX_ye2aCoD!*993wtkMTUaaaRGJoZz zxE(Uca?(#Vie*Dy*wPMRAV)*u$>q z{U?chod|tm-#t{0T_V5XzikkyOg#^ufGmI$hjgeocJrjWZLz47)R|(v&fY#OGNCVVjz^Mzw6L=EP5`T z?CWzh9f52-7MGn|>FBopjy$;>!7kfFpqOkQH&)Xy;Mx{>u#I$UiH2WWSX8^LlCb<8 z^^+yMXss2d0&mhQVmQG`E?0eUqMi_{q4pS|eERbxhjkY{fp+0o)^QtlFRjHG_E`>S zDZhiccc!#pMvQ{j8WKL}E5y>)d{6$Xk6O6DIppiv?NZpaW)j^dGokka(fF;s_SEM& zf*%b+{6#f(t{>ZPy_73KX@e~7+Edsqgoq)X(>7OgeGSodC7qhbqErpaz+`d^=(wTv zJkDKA{Okum8Dvv04J=8Q^JHXE%rR@5{)B3HRXD(oiy0p7r_|$+onD>cU3CP>T8WMl z#3@zpF+KH`!$a|o)|W^R;)6Ci-CTA0kuO$za1!5o>|D2lJ*S3*v}a9uHhoXWFiWE> zJq}7S%Ji*i4corkXhaXn78Ae<6KH*)&oQr8i`7^?8|pWu>AUD}wdtBPOb8!;OkxE4 z7+`5ASTHlhZ+`TfOm`nPfEV}DiZsf1ki2q-pjS%eLsX_~IOc)B<$@U8oeu+7^WXCd zDUNyY?sU77!V2Flt{{`c-&b2cP9vwxvJ@UikcY(XCxC~=o3m2PHFBe)0XZ_s1tLyHNbAbTgW_H|t6abN$ z^6Df&5O&_uzC&kCSrb=2Sq4GF&*^MJv;o21|_YJ+%dl+9$- zwdqNlmF`Lc9`&v*sjyvb9;57-e=B+>ND!-3bAnFvorB$WYemoIX!;r0wxA=jpz7hk6+gFLv&$2*nhLV@{ujop`*Ah)2*+Cw$+VIYs*wjuX`o2+@xqS&AQ+o9M z6aGroq$IK)RNg~J!IQ0e*O$}`{*EMK;e%adbD1ELBumw3s)-w~E5vLphmVG(o!()E zr5W1A2YX)4>mEO@IPac$O=}&{*sYd;)2Ek*p!8^xOExn*`F(CO2NnN0YiGv~#`dG8 zbIbi5d4Cyjf`zT9QGv}fF*nB2r(B*ET!TZXN=iyk847#@3oF%rczHEMwpYb* zERLc%SbFToT536ul4-++6QZYqEs};Y|rK;#z#Z zxOTv1sA|od@KxdUkUsKtcMoa8&-9%Ls=H3aRJV$K*HzoP?(P}Sl@jG)9w(!>DiUIV z$Qr6oXfVsoaaUq6SjvnKyR}zVHmo9(RwP*}_bHdL%)BuPJbVWB>eXTd_r~0A&J!^OBkOuKhM!@F*1n+0tu6{aWyOvrAXkabL07!7C-eP9D z8)pMH93Sjmng6+fF%F)@4g|586WewAdoQUMOL~Y~Z_%0E7NMz0FEc*trw~}Gj6LH& z*&;9Y5XNSx=}3qDP`BcUAVw&|$bHk-0}+CGzC;jrFDoHVx9AbqZ9FVX%8lS-E)b{M zaT~kr=BZfaR0Jn#Z@g!&6-4oASLC+vQG$*9Y~||Chh)yffk#E>(+65g?$#2p55C|^ znR);)O5@DheS+6?W*$MhrJD^>^vt6lfz3x2z2?ktQ&xVHE{jIXhrE0bHdR&~+lx*) zhd-iDLtulj`lwFR7XHFFX^b0;gRi7@7R)Ph=2n0bwJE8I5(i^rUN)E%og#*Yo8^7j z1p-ah=NQng8-U|V??Tn7>3FVHThESoFWS2cdw*vCa|kBBS~ba(9AK*;a@wjxc`Eyz zZV=q7)AttTTD^3w$L|2zkWm7N1P=qh$u)-jIc(_@E;=3rlocJ1E3h`0fX0uy*Drc4 zh%eYpSP1lH^Yc>wAfZBVbC6r`O$`IP53jB7acA=f>;A*E)FbztUV3HG?4K|)=LE=X zx&IQ6Cpe}_PyZ;jq!(3e<-WVxwmwS_hN81INk~VWA&)0A0HJu6-d#^uh&nPjwJQR&)J8EP; zn=6?~^>b2Z@59>X$&lXCXMG3}FQ!J}-Mv8*^3h9m9GvQWJg5QDCMVpDR5|;)l=$_9 zO&CEMMOw%M+NM7F7GAr-0#{jcu>g+FFskU8dUKPXMr`qPNJmz?^Yho$Ts()=s}6=b z6Pv9sL|{@^-Wnhi3SaqSYWI09da~UgPm#-^Hx9l0BX7OdJJ(8Q!%)eH@!Mi!!AAe0rw1o`m4 z(hhr8Z3g;#N-f+_NA53xbY{WJif|sJqn{K~7@VdC( zurxEt5P5AAc1Tulbn1)yg16NbYgU_Sm?G9Ep;fTja)vjnZ<*;V) zh)x7F8s$kQkHyx0u~jWx&9k=C(CJgL^EVsYzu;xR&s*K53L z4V9{-7p^$%7mo*+k~TFFS2}YwE}CoL)i>!HE2v8=;y|j6+pUt#z(rgk&N3akVeINc zvzxl7tEZoG3`NUb2G1>zB1}1bdTY{#WM=*0&dbhcR?cFm!Us!q-?kbfSXPWxM9jO? zooxxSSxYjdQ}j(7h`j{^o`ti-#Cg_NCIP<~f}P<#*{@&w2@P?cw=(oDVUbH;7~V`~ zU$8G@xyQ&*$2$@qhPJbhVW(h>^$}`Q#Ttmiu4zG9#y2oL5HpYWsOtnwr(5fXp59Ga@%0*LI7A{e)r9+*9&7n!XI zdN$t<#9hj1t+rC$BR0XAk3abJASjl4kCX*m^LP`FkgI!E`#qw*9X(>GD2CISgZ^~G zEiOZLEtm1j!3AZylm7TWIoe3hTHiyKVFBt?Xaf<%QMoA^AC0;K*M6w-@rz+8RjN_Db)RJG1c0lpu<# zAfaOKL!j=*T%)P*Q0`HLJV0IRP1_HbOOCo{d$71uC!^Q*Z zzO%`&<+#BEXRe0vNrw~nswkAA|K*-gvx^XG@%vb2Fr0)mb2Xcx=q-*wGoqDZC@rYz zbrnHp&J~VkjKsI*~u9I1ack<~OYRcm5Zd1}T z#A+A=jtUc4O^fV>ROmse(pyYl$O%R-*{i0j$SG>@UKB#ao-y37 z8J)bEQmk|HdL--~5K>UF_P@O@s;lH8QH%u%6ka+wt=LX-29= zHugRi))JB?ct*&3(FhIae!CsZjz)aK)WXj##x@pD{ILN?du(26XX#qz=*59?;>#i= z5;_?b*_DYq|753zz`XZBT+!k4L|uFEa;W8U+mSgHI+T+W^Z(&zanBOaH4v`BQ#JDn zQtiE$iBpt%h^Ybd`S#Ei$ej5&R08O5U`np&Tj9cz^42BB>6{*XT(P$cbQl<2_yP+g zlh~?B)nAp;S-YwnF6|2*?nldso%wN<8G6ydyZIcpmDy(Es5>%O@FweQdDCF=QQL-m z$RPSpx_doP?h<;GRR@Bo#|2OKR8z08n)&S-n+$Q&$jk3{cKgB4Pcmfe?>Ii2TEgjS z$o1Njy14MKLNFSQ#R6LScQ=}q-iQa{oNgyWRxPoBC@_4Hb;EUFdt0dKbQOd8+1{D_ z6jc8GfEw6Bft=ZF_ah1Ab?pH5%FgB1+2cx$pHQ21Vs(uIcy>lz9c#lE*nb|iI;$G~ zF}l&}>i>V2>_MmX_4Ns9{0?oE&EWv1^c^`$tc)++q8^|R$OzwD*Zl>MtfZnuiJ1P| F{{cb$0+#>) literal 0 HcmV?d00001 diff --git a/design/FY2021/images_cpgfunction/F5_g_funtion_C_configuration.png b/design/FY2021/images_cpgfunction/F5_g_funtion_C_configuration.png new file mode 100644 index 0000000000000000000000000000000000000000..222f6c7b2919d1f53dd818e61d769223835382ba GIT binary patch literal 44148 zcmZ^~V|-=7w=LQ+JGRqNcWk4xV|Q%ZwmP=aNjkP|+sTe?+veMS&ikKtfA_xoWqnvx zwN`P=F~^)0_DfdmD?A?jmoHzwN{9<9eE9-C_2mm#6)ZI92{k?w@a0Q6kA$$Gl3V)O zI*b}(>4zvX?yc-43dfm0nm+jY&TCt9yv zsCpqoy7lDz-spLw{H62t7!%JbRCw;e5LT1@w`ciu8@=IXpo&?t@^FieFy=3FLDR)< z>5JB^BB~SS7(FWse@Z5KH?lWKN3S6gUIzv*u@@4j1l7t{hUHCzq2p?tQoqyy>vIvtBqd~M|{Ju0MUjN%^pFrge-Jw?K44wf zI`&Z;G|r8Jmwei(@3_ppM7PNO{DEYZnQ8eROE6|dTIa_8wFO}EHG?=XCrH#hzUnIP z+kQ>6<7ZiGtU7l>-d94*`ryPie}2|kgg%Z^el`p-V<_*>5AWvI)=~@Cmi)rP-rimS z#p{eHY2B!H)B+Z@+1DKZg;)pGDfD-**jRenn=)_v$+d8yp~2-%cF#%Ps9pT^h<3R+ zs0km~Bw>jrRTWXu;L(R9i&^Mbm#3uf3xW}q2I7a3PX|h@PrO9r3O_n^Hvsl+o`$dB zkAH3<+K2t=kEuFJHK0${9{uPNl2+)sWCD8-fR1C{tE#=te$I^Bv#n4z_?)%+HLo=V9yLy!9WuZ}5 zXG0wzK~*ehjuL1qas+%b8X+t<$vInx$lD3(a@Atf3E<})t0~#^hf8TxjW|GI)pugo z@+uS1W*c8CKf&MKXu$GJG@NllCvop>-uaudGdTTaNAfC1{ve_mLZ11h#0)JzMe$6n zUH0i!uH^a?YUt~Pt0tvTZ_Uvi8EY-aGSwfT{ZMr3(%Xq5miy`u z66y4mT@9z4NbBY_rhs{~zdGW%n${wIh?{4w_8m3>lHg6&Y8!%3vE1v1)QpDIJsK;E zC#P5|rJ?A2J#Al3NPW0~9~YFb=X7>P5O0p>d^&^q-^@H5ZPqUb>ft-pH!5j!j@p?q z>)*PcBNc7ghp`B6{YcVn$MT5@#XRBurC2*l zLv`62NQj>JDq|PFXm%F4R5aIE*GMz~{n;y5K6bi_jfPlB$zxf2#mrhVit#|0F5xJI zy}FZ?@0d_13DbRZpB4gvT9&l(uv~XoET0!V^QzJqtFGv=O5I_77H}wOJ0|INi&(4@ zp-JCbo5m`PjlS>F>Y`SG29*Ok&yotm6!Uuhq6JNS)rKC|`1_2t)-I_>;~YXlMXpfi z$1JFE`hqD(T;PJQrurJPMTXhmN^@JL6ca_Vo2VGTVn^*B*Cpvw1{AZ)+^VZ7!FQZBEaC2JH)Ay^_et1ul@$HNlD z_;J?fV~2oiWp@vYD@y4eE))?_X85+}4XO@g?B~ts-V1(k5m|}T7VaoO>$fpI5g~dA%#_b_3Tb?k3+J#?|q=+v-8n~{q`8C zDDhleo=sN;xqaFcGwqv!jaKkec;UO2L!MX1uUR!0JB!$Jf#xI1GA`V3Vl)}1-q`?r zq9{zyXAl$vNB(`l;btclaJn&DE@ z=hJVEtjjC;zDYY8jW>m?fU^|z)R zYn}bPJ`n~@35H^DmI`C=99iU!i&S4rqjf5by%F~Fg3O%~tK+8*I>`WuuY*Gi3svB& zx#4-Z>@qegg(;)PC*R>Un{R5)_ozsMfR@S!?+Sv4-B`?jy?s}+Uvz4frqn$q;xlLz zwL+d}sCA*}qK1H%XIDU4f$}I=<#M8q3+fZ)hTh3z0VnYhpvZJr!<>{^>?L_qj;Oy%{{6Uy`Z#-;8WkX z%2cMc@?uh}vc+PEqP$-dPr<5U_PDr=WTm3`Z94ros@)0=)HBG;%?*WXBzl;ZxF+Go zRE3#hF7qFTXdDt5+fHntY>FjUIIpr;sKf$cO809b;;XW~DaXZ@DwXk&g|(ZNdS*Y| zh@RnBV&PK@My|#Fm19d2hEUavjKl2cSoh~K8**q3WUA9S<;JQ%!+Y$rt5;z<6Gf;{ zJuESm6)SDCd7yuLn3z%$CkS|fpFA*p_~|7-l!rc3P|Wh{Ie}SufCT>=T7POI3Zg%>eq9GPf z_&Vmt%7SBEPU899Z>T5Fy?ON9np0i6ULhvV)-8To>Q-Q5nrOy!d)T~d3hYNM#fbsp0yV+Mo8Wn<6ds(OXPsryFWCRW@5xpT16(AKczHaotO z+F4MwLNEG*kUCFa<`f*JW>iyZKS)a_>4)=^uZozmy_pj8tx4TXl3(j`Rm};Wgw!fo zctsjF;cb((%IfY|(yM&$f*8TnlOdMI)KD%(wYvkB_6p77H_i@s*0myG6?!Y|ECcF5kHTI8EMgqQql7wd!$f8p{BadE4^0EoIRQ3VcZFLo>~R zA0y&AxeIxc7-ACtT`dP{y(`{c~~-i+Xt=SDGTT{$HkzDLa3nz{7enN z!6=>T&pt8=itchHaCFV;kAx%p)zxXmPvxS;hQ)J+&6B-BCClqZ_kn5y{0KQ9X6vK# zks5zE2QwdKag);t3!UP-GPtPY5N2QdlbVW4^6`aO!1wGcLXiC%F-jPl|e@$hjN)-0pr;!mTDiJw&VV>~k6p`9^Tih4@P8W-wDJ?!^L( zShk0h)xva0TJAH1%rn;M!D-TAr<`l%Tl_^PhP(6} ztWTw=zC6t$X45rnO_nUFuR7n*J)*n^9Gf-dYCPK*iyCYxZ4_3C6KNvV@6?HkMb@PD zfkaKeI>aPZP37}yq|S_}oEm}mu9!K5&74Q}@f$)x`(7nV84HB$Gr>u=8dh9BN^@C1 zboAvw0)EKzf3rjpYh z>#<7ZE#pqDgu1>Hsk>~w$Gf5Wa(wmZe1nLzDJX@{28A+axx1dvq=f$`ApmzJHg(wN z%Y!AkWtZ3)%h~97F1!pPyt0|WuVMgt006UXl5=>zkDMh?tzZKn@eQ z%SfSjLjxDB62M1jYHFBttDtp2{lzdjZx1?+L2ohkF74+^Y6YC0 zs(r4ci2eLKDN-)x$?$*;AnH`FLlioopKSWd0|bj0s*hNHMpA3lkqvUz|Lz>LF!X`7 zU&#Q|=j{GK^57_hbkHiZQT~+qtnf%G5;}@xQh=-=_FXwg^?W7j8GW#f2P;?bW^{bg zMNgl4i$3^g3H?ooxY_~EyuiraA}cJkd3}-E$LFIO`NGA zJ_1?Cdu>s<0nOd%Vio4t}op+H^@D`7rmbjNS8QI-m@CBqggfv5BT^2z^*de zc9-GEuF6g+sbc1mSK07V%!Q$lh^}fl-4I0fC(2iS5p_`XWqt^tDETgx!(G%mH{^}Y z5h1HSY)by|(l2E*vo;>dYDA5{_)N%X&Ecnxw&m+Mv^~bwV-ca40ROsQKZmKIEpR*2 zX}ouC(C8~-&mw@0J3CZ|Ip(vfv^t`TEpN+1NK(~a?td^kHy@VejH+x#7pH0fV`*rC zP==yx_VqhG1p0gqdC<3^; zFtGA)x|g!=k{tgZKo&#SR$u6R0~sS60DP|&aIiV zY&A?&`{FI+LLn)RwR1~>4c{R*F^c7emjUl?TiT9k_o^m#;glB&Rcjs%+V{$5>|)6c^FPV^ zr;e-=0*cXAp?68m>ai&=Sk812q?B{!pKohgCr4EtJDp#toYj#fgsR4AM?q#Zu^2!_ zLD93af(K#|W;d5!b2zQghAz4T8`pH_M(T^Jy;v-heK`h_l@?jDK*&+wn%4pY!Q)|WzW2i zmXMh&tbOu*uP>KU%2^&AKhX_aMpaBzFD)IvdFOc4jv=I8HN+xgQDXu+g)I)1ZqylU zT6UHfHG0?0$uq4uqTiMlG1twZvXm9q7E;16eOJnvKU&3(*CLK+tIE$AtdV#!vKL=C zH_&Rnf-egh^!k@amkxsLld}~T#ru$lb!3&n1+MVQ8J|{=BBsWQ;fEL7=}|!@r8HjOmuH!$Mkx% z88aNQDX#j53846V*4vIyBc`L7!k|I28e|tNsxcCa(_V)JfXDbE{rzcFO)kW+s zKhfG|fV}+o(g2H_S$|u7_+e#ZEN`GdxEM{7Pa;ag#{*pgcb;Ttd{ZE{_W(rBgO2vetP~ zKr70z^giM`UvwLt+k{tpomHTSBJVca{+H70$zI!io4JN$zR_Xjq14Y9M~2)_H-Wpq z?N}XnJUy<{kBmM(ke+=;0&-bmwkG(l^IG}bt^?A#%fAMu1z^6BGy6i_T#qZ(^b>FG9s>K( z;UCTu>ktPo+F{QU2iSvJ`Zsx^TiWCd1_M!-s;$f`Ysa43RhafwO!dA5JC_H45_O1r zcyI#%fR=Y&Zf8hK=2=F80*noN)c+YbPXelXaJ(2Fg2%Vlmq*Z?m z*4Fse=*m7q12``rRYZt>b0h@y)=`hmxeS_J0?#7Uu}8`i2H(Jc@CzR7=5fa6RSi^* zOlmzCH0f``w+Z#8oMuWFMD!CE?V`Gn;DG7!ZByTlfARsKOc=EJ-lo~^S<#cmpU?A{ z^gsA*xPnK9WyhldNSj0Loy$Nvkp6sjtOzi_h(WAc79jO686Qg-ZQ!J3R>d8EUkWZO zA5!&jT$O!Y2G`e2-d)m!yBU3MKd4#`MnCoX9WDKQ4avyDVwjnLeVd+w!{Zaommy$? z7LJUEFY%8Wt}xp2|E6}TvGA@TQXG6SSaEsSjV~%5ABZ!_-{AD`Tlg8gH8FPO zy@6ks`G@?0wigIHJ=OkuF!G756rpw}J{+SgN=9m9ZU!==Y|83(N2nAfV$c(7$!E3Z zj0KWb$KS?Ad72eyz55{Skc!4jM)@~He||}V6Q?$z?L6WhHQeQUn{JFRd<=;L5a8zF z!IT6{S(X#;TwO+-RFP&ElDmBYTcGOnmuF&ec4J5Xit$N zVXC`jfOyo)F2ii%_mASR)O=v_>_7;<9A;H=$nxKjrEI$q4J02CdK!=J&IF9opYL0X zC(vDw*K#lQkmXnw&03BhvpIpXc zs?@R3|xR+-gP&YX>6Ewd##;Fu!T_ICZ98yr&C`;L#_FF%>-7 zc5WKsx?4xkv`+`20)(m^+TtLNE>?fv!hS+mba{77%6M~EXD5z|n{eb;QNeXdO6ut$ zV46Tqpj&m$($z~}DRjj2Sn^OOeaNa&9*ZlEZbvigi&n2?$RYK6k64~zJkkoH%%}Zt zS_wkVe{SBdAa*?jQIEbgt?)1%Yy0gUn7eD8?DN>UXzF?}OnOCmLrE~gKYPTix z4TGa9rm-2e_C&?3a-YPA^m?W_kX%W2^~bu`p@$Gvjf)M29-o2 zN9Us{=gVJ+G8OAxpdTM>i3~?$SAsSsWdkM@257sKa=xZ0M&um%_7PwD)P$%9wc>eO zxyU;6GeF0eCw%wid@o#jH3}zZWTIZBH7F>Z!o|fMuvLmKuHzl8x7BC0@aas`T$@%N ze~-*m6+D$RoXRa*+f4Ftt|8eYZK55q->8-nVL40RQQ7haZbMd#doq`S0r5brZ}tyM#Fp*U-?>pR zl~a|Ys9ELH-p2l>m(+amgfK#{`e;|-@c1boL@9z6QEb?bhe}u zRb&Rud~a|c^7U$vna#3Vw@^l@iidKlI_R~3JF}_-<3JBoMFWA%@xPr@l$D>e9C4n~ z_4uPv1k#?Y5X(vY$=nIvLZwzotLnt8vIa-FLJ9L_f9y|bbU-l}&&zEWdjJYBVrw$d z#P?x%Rnb6Z3G>aJEJ|~Oi{6vK%H_3IVSrz8NeZSs95Z`$8WPUPuNBBDt>3Y-4o%Lx~OpA7ZBM$=X}&>KX6N zTO%sGV(JG+ul{l|N1-)$3>_E*8!GzrmmKkhO9Q9RG-wb*8c;cUHJvo<8-#&9N*vb8 zgSUpw=W2CEdTlO!0GAIipmEAzm)UmNr=Z^m6?6OwDLhAYSmk>LTsr%u7hJdm@BO+t zX2)C_xP)kK{}+uw-Whnx`X4y#GXM z2M-NxqiRYk{B?4}BSw|Ud@bm_&Pr?0@xe*E2Gip)aJrcsa%Lv#m?nqFV%wrfjs0$P zoVpe;)a4%K7m+M4vY;q+c-cLFhxIh@h!mL?J?+AYCvOzgSha)bd2qR0SDQeRk&)5T zbvN5|(e^|F;VP|RIQ5fdRB9+F(>ZH;^XIf}TBFZZZJI@h|HXsqvo)X7Hb0>{xE#KS3k)Nmv09X$

{pBoMCfTz1dK(eju%mr!W>{v9a8IZv?zvZu%>uHcfF+d^u0HnirPN$2t2WP&y1n z^YJ~T8|>D*jS)e6j`-^jC5JTF&2+D*fFa#q#`NGB{sl05?uv$(pOaS+^smys>+vnmz9H_3>J**dw(TdLC=Yo`T^f$$0>z<`Asi&9Dgw zsbg*N*9nR{?utip0u0b z^41xtp4k7yA-_MP^|iH`?Hg)l$p2EcF3{r3F} zImYJMKkOTl>uQd$J~*$j_&n~WYzXjvVzxqh0J$OH5-j$UoMraB-`pug?c0JG%jmd4*PhHJnj6n$&JE0UkrT_e_orq&h*6Bg z@Dn~iSlA{7&XU@DKUzD|$7WHI(u zGCkJH?gh>+tYQEMN`={5;?oRFCxpLL96d8}HZ2Qvghy5U(yYyb-DBt(s8Jo1GO1RsJ#ab1gK0H8YNTnG^w+w5V=^~&jqo?8qkA4t(bLp z%1u^brJOM^+pH*9WKGzIFNN7l75ADYAxzmhhz^RT?upKu3-O1cF)=W&eWsJdoPXIY z2P;%zsHmtK&f9c4Jc2eHY+k>>#fHrZBFYyY(MF}Nf;NzUJ zw+}*IUY>fDfhKJEaUkOZXReXPaiXnxLaN=C=GePjmqohZHR*D-wS>4;iM#T~TY!ff zSv$D0CuLo=#$#QD!HEpcDBZtrOn7X#P{&5CeD9u>N)D>aJ&ZVZQTf2%vrQR&k zYQSSl!OLA4uOhKi0Q$%n5jM0gmFnXmI1fbbjHPvApn+t_{~>5937_BIQroWy*bo8nV=|&G>cXBO-NVtSq(v$EzdV2*)qe?zVrSA9k`t~_j^pE4#YCD_X zlxII8W+2Bew_&fnsbIBEzr8E~8MmE-D&mw@w@>&=r|{6G_HiFw@0yQe)7bj_DmtZ& zKfwKXeETC1u&2i$nl}u>lk?_$L4pRk=_?m6vsw)SUOU&23MFhh>z5RAo{vsb2ThqE z)9U|?9se(Mlqp`apedy*s-6&*G%$9pct8-Alsg=al{E0^yW~_LN^ahFL3BW9bS`8C z(hg^NS`#cXzZ$92qec|3k?6R0{z^-)&OqPo+KV{@yYD?iE7Vy)5#Hd(G|sk&+mn@; zkrL+|V`=mX;d&$NrZ&VkK=Q8F@V7VpXDBR08aESwOpD7oQyNEA#XJroA%$5Idhe_y zJRtX*e(9iC>iIrc*3@^_Q4c7Y3@zmkH*D7@{SY(6-rm7Pcimndn60)K_HSehj*B8S zO|d#TOyy48n(*7F#{s3@CWZy7jy6$yeSCqNWw4u#NY9Nv4o+ZLUboOZeL{L!Y&I7) z+4n~rOxe=J@Ei|OL&(Rv%p)x{t{oYh4rU&~oYx6eTOFjhaXjxBr{B5Ch$+k{#0@`x z7k#1z60Cily(_|Dvfzh~`5O)7w@{o~SnG27k-%EFkz~lr$7m!)&Tk)I(Ol!oyU=wM z5^4-lJvs!@B#fnoNf4Q)(#=b3i}O`IWIv#kd143ViU2oMuBfW-u!3f1>*Nb!dK0QA zhgClflbzeFhngwhV1(wP9-kYbiqGhp92x`=_2Y=wfpqZEsjmy>&$pvLrmzu+LGrw} zT9iTADlyk;$BR&K9KB5T=ft2)u}oVWxtbLB*gMvNd-T&G{Yz~gD>#j>tMybV%JP$O7oU**1pc5nvV`A~_K`EJRh$I8-`*nnjV1wE-RM`MJ8h7f-Rzhe-!#GE z+8mLM2$`p`O^A)DOtCf#!4lFD+?GX}s<)051DS}&xw=g6OyeCN46)wQs#meqkyBd9 zC1e%OZl5G=g_orTj}qBM7AA5}v<0jFIG?yj;48ME$2lC5sEFNtVR5Vco+FtEXxAgwz;31Xrj#7=qT?5`^$P*%r5&(z z^gG#YhgU*lLW=_m(&#HXKIsd0MlZOA+E6KjiNTC3n`3{JX&N9=Szy=@TJ!r`oXBt+Yaiz@*UiF+`U8Vc)^~f^-wz>xnA}bpe zcu(6f7OYelEBt%rp`IQ>+Kvur@UfcL=kzGK#zq|$9==}TFa<2D$_;d?wGGMb@(Ux= zXGI2tH^f{8KJ%6)?Ls9ohRJIzr3kp|zFsYhVCYO&Ni3bkG&={)2Yka+;Yf-o4aO>Ye_NrHNMnNq*YVj}VxL127!BZP!BYRDF8+TQM5iLw$RsHpipO`WO%agi>#2699cf`^i<$5`!)K;Jp4yanIFG; z0-4bu1m!ZzF=jGPJC#e&X6x5jB!tlyi;UoaM`4_X0GUx=>ieZ<9JUY$-NloiY{u+ARdiu*pLW1MAj10Dk^X zh+$VUy{M>{G{LPeh0Sws<^Gm>!&p+fdY;Msz(~O-$v20X9SF7yWkiA-uh!^|2I5b* z7yZm<%4${Hm$ z7$uBlNd-7l_dIv$k$~RsvjsU%td{ORY8R_#tw1p)*X3HFrk2pNWXHvuz469{yx6Nu z2~kX=ApqYZl#rEG58Yoj{D5LDMWIxd5%CXBzLZsuz2>MPFFJ-Np~4PdoR!1o&RxBW zb36}QVWWj~6o2kMR7|G!?)U1nr}8mp@0H54FL}8RvSM;rUH^02IEVX}&dl>}mv=xJd$0eiSDF1VVeVum7za(fk?iguy(v4Q`J^(zt$y=w);x5J>Q4*()j{AIXf&(l zZz6~9b9W?vjgJOMH0ag3*n}Z35p8xkz}j&~8oz&Cvwj!^ybaMF@_7ps-Q78*r0(-Y zg%nHek0$9G8Y!=H;4b^XPJIAX^yk1HEVpWRGU5~ny zWPD_g`PN`>uMd{5W&9xT|fqL{NC>QC|z{E$6>H`5+ zE5eju)-n1c>(uDEL}VO}OI66K=iH@^BLhNpv3i^C{t(*vP+cQ$`?w)Z* zun^Uqn#rt(i*q@RaTQ01 zJekDFW(`9yC5_;+;pOXRoJ$jUZ_|2XwgPB?Sd; zYCyTKNVdB(EuF&+rD}RRMTsfOLH~O90~cW5>D-V(r*CZ2pD|VA12Ci$HeUM2UT?xA zhJO#;NIfsqX>aT#Fkb`jACER>;MkKB0drxJe^5He2b^pL5Yrj41kT}l<6~iAf!0HV zr}NXxJ6QCbZie=s76lux#D5HLDjaKB_FpJ?mQwWW4ylddGQWp`e(n(#GO=BYt~hzV z3E`NDgC0N;%E=h2&H{V+)=>&oH?{D#UqQ!y7HdPBz^S3Q*jHK zE9r_5=*aQS+s`CxEqXVJ7}2GEGuWgpc)UH!u=+{TxSHM)_62SPK#RNlQ$#+#HXBk# zoGbn{AI-4*?V2^Z;`ocAVINVhl5PpP;3Xa(8hT2VdRNsINNm!#<8MI2^q~JWPzgtB zVPw=l1U$?=784*Wh=-h@G%aDob>>^lZN!b;a7d?ik%IjJ;nE)Q;RyUm3%01L&cDI* zYPi|LV_(e#gT!DBjQTF-lpu-bG-W0ry@u_-d%9?#HRX`6y9!ZvajRTo2@wCG z*6r~$`hu2D>2KO^*xAt;=&b-g6XlL0Fb>hEGMXtS?xdWGb~jRN$BT*S^|z_1_K*n` zAbIkVCp-@C0Z%iug2Piq_M*B&_B_(noVo-_T=|A9uqpX-HSM`5<=abvG+b*$2uCZU zFVG4%LTbLVKSZ}v49G+j`m(9~%ZF0t-&6KT;c0l0syOeXNoUkff8y$eN%c0$ur}|?J zj)Kdyn%yxg>_dHQ@Zg{pDuH{6_?OtNAX<{5g#QQk4YRDoL3VHC7poNYICT~PW*bHR z+N;q6?nPk2`G?GOc56dpA1Pj8JS3lsEBPQG{I-jz_b5rAC*Q*ovLd|_^fLu|TI9Bn z1da{Iv9l7go0|P4IF)0g_#3H$4o(H>zk0fXA=IuCp*=kIY$|c->Q)Zy><9X@{e_9? zV_cY&W^e%Hw;&s04*7L@I3_!DT!V7nC*r2@!&{O#@#*YfwiFT8V%ZMHQ69_;i6g(J z2X?#D7mC|!Vg$L%pkh^0^o)Zukr?VZ`OMgMN}JATkaDzrW|n*^CIE0yknndY)r$=K z>muYvgVB}l8&(O%!&q4Ih5rL)+OX^oHIY9`#}Lt=UJP4Zr{*$IvdNqqsR(FZ9+of$ z-`Q+<(X#f`h($+I6CJ}w0A7uI;7z>ex7>3YloSA#Qj}cqztOyX0*kvCi|)Kc`3gV& zNmP~!oz1pLxiIAiljM-h)%q7sycpbueZfK1xtrMcL?9D58amR~5?^PPfziib89x1P zS$bddw^>=Z{tR{dybg6+UvRA9awp)oB4@ye zd1UQ_Jsx+kk23QvBXfAcbCm#n{BK&?zdJAq$R%hmI4lu6Di!U9^mU37H6zcj`*h8x zE1|19yURG~4eey#JX1EEQFBJub;ZV7IsdhvCPe~-qkgP93cr2$YaTFt=Bpn**WA9< zD2jsxAs&m1!1TD17vb6bHePRKoO7!u(JaYmhw{jbb{j0@?jD;y;~Q9&xWc?QIB)m^L<7)>@re z-4&~>C2(nTjnkqjf!}}mQ$p`f&XSi$J|TmwhQWXt$G}ThWxt;D-aGhZ-$Ucosb+wt z4ohX6=nlECdSZ5%;P7(ajy528$K#)X%9z8{+Tb*!ViuT)M`4!xM`30_t*<^DxE@Ax z<+5n8@YG($(#q~SXJ-^xj0CE{0!`Ois&;Ve_GB$2Mvq$cxe1ZpnaZc*pe*Kifiein z?1efHMhO6o@+h#K%1=CB!DhWDr~>t9^{15lG4^y+N3)7AtH$3M5`thr9^Ai3Q<@iN zQj2fb7I6wFXtSG1K#-Tx7&g!tYN^cHn|HLfzFa>3gtN4O_VkvVuP_+7J^ss^_5K;{|vyZ&Db{gian!-~2Gqzwo z%R^A)(kx4c%L9PRU5yEN>-Dqg$qB!|Fjc0f;hcAeQtcp}cztqgw5-ea`*QPO$EwO6 zePHZN4x~1~)V52^p%jO2_4^;UhqosT5Cbh}o+OG`KiI^$FB_qRG)8CS;aKle{V^`` z@QCV&DwPj)5u=@T3)66tG%gPu2w%=(!&|5i2I-USm2~4W&*EdT-y7@dm*ZT9 z@tKPX6jo6Ib46MdM~aJ^hU(2$vdywG1l9ej&o&zWvf>jsmW$x3UbD_ne$yp9p+#^$ zq+xr~y4)n6!amg!ov1YIO?o(%=}&0G$0t(n6iOSIdRCg*juPG~n*>LH?Zmvh;Upv_ z*KCoP(uLId>80TE?fub2tFuzuy=F+3Mnh`q$*?F6etw-*{lpahLAomZ z7d0`5k9j1fXe!o$-+^w`RUX*4Wo|?5AAQ?KYQrX+7$~D(5y<`&kgAAqEDt z38*-91t5cGTv_A%M;FzOJcd0Bj#Za;e<4xU_8^i>Jyf_d^Q~Q_-Q{no7(RyPM{3**LlI!}3(X3ctUCU&=XJ6J?@Fd1vf{zie?rp8_17%nr!^|^s8+022(nmcxU8nmDaF6-?%Iq zlHMC>?J9$?i*z6|oGiy--QS!EvU*ws4Q;e5UK+0Ye7`lb&rL88&h#HPzwbckodgaW zJEEeZj#lVyp9ei}qfX5${UK~3bE#U3B!poXx_N>uyS|qr!W7@g8Dd1ttS(D zL20To+3ZT7R<^DuNn*No+s=DMdm44PZBHOCufOt74)eGoq7Cg#sza=jzSlU%EduUb zd3Al~alQI!7P3>cI}K&snG#&hP){f#k*fc&LShXgZ{c<5hf{t|BB|#roY*J?a;H6z zM~@A;4wce8D%xZKpk+S%&z6uujKzphA?mwExYZ35 zxNut6GIK4+uGvtlrSjL*IX(|+*+v~+=5e?$9!Y8Pg~_)cV*0#%QR8F%3~b_qQ0cbZ z6BQ1eR2otw5&i+PP(&c9+>`EGAGW_Wo2RENz-`YZJ;|xRznsp1qUG{>f|nTb?9clP zLs*o_M-`9`ChA^zmg04)q||*O8T_ukv(WGGr}#FgI}p7OJ!st&yWk0xRCXa$4n(=l znB16zD#6C89{j|qfMr%Q!$#uYH6=`$UYNhd_!e_Q_>eybdAc?0gN|z21X!YU7Yw!c5#WYKVM9}ztX+**|@MG+{eLJb|MltR-Gz-=| zF~6(7*Zd;gD<%n%`IiPEeKhLibBLus{HrH)7TxcU+BIDiSa33yoC7@ml}1Q-0dBKpYhxh z@sQ=}W!-xaA!x+-<5vQnK;agPtYx+1`E*UPXFKHW$5FhX6sf26(p0-eH}-SXf*0&I zwXTddh%Ec%MIo|0Rw5yHVK|K+27$JS{lts9x$3~}qs>NWw}nYN=f=r~`TxFnGfZn8 z5;Ej+r-c)ZitD*u_e_}E`^U(d!oMhlgPkA}u-`AR)$v3Dz-sD`-{-lO-TB>8rkYdf z>;ljED#Hly<^AgT$3%^YGePRj)^^PZPh~zXGc$8{+8vbRsI2AbbOPHb@)60J2mfUC zW_4$A)?cB!zZ)j_N_I`_MHv0&i2s%Uq1S4Y$UW1R61|I|iFKwg;BO!sC>2x@kr<`j z=j-H}cR*`}r`{xnT#C`UVc{RD!XUza;8@Y_xD-s_xpumtFB-Y{@5Y9@-P4tlN&|Ha zU*yvT=^l;Ig&ciWd3Edk&CDNh0$qpBV{_-Ri|lG>MXiAj=f7fI{|{~F99&uV_4$r%yJOq7I_lW&*tTt( z9j9a4>ezPDL3f;VtT}m}-%L%_yi;$@AM@8eRrlO;_dZ)|f7fTN-HC)K5$dG>{hf0J zY9#Rwsk=T`u!4f&(A#*r&NE5Bd^d;^YMsxs>#BUw|8nx&^Va#|-)%Jwq~`n7%6SD+ zC%rmY9)F=DQi|HdruC69R<&mmdGYrU*wXlR5ehi#4#UTSEs6rV zF+lMNFPh*WDuI`Dc9vlzv|t_ii^oG{wWD1^UM#=A+)-c>XQYoDYz>s6;%*x%ZIQ#} z<%9rRk@e;Ti<^l&zhsfom~eUVxLB({~86C}zVN+#kUrq3D|kBY!R_o5wl%Q>)z+sDBMQ zA)p5Z0^(cm!!$ZL;3c<&$Xa`EiJB@7J`jO0Qi?mB1gTy#cvVxLXe&d z6l!_8ZwUfzg)+u4%FXyInAIiSIx$PV?vffJ_TmIKR8L3b8=96$6>I-SUrbwZ9ObiN z%`~R9?(Z?(3qA)LR}To58sz$|(_szPX4;l-iWNv&S>guIb(|Uk`CnqC=4vmrSZ?}Q zb?^F&)F8AeoBAGXdcN=P2>{{Y)EBK?n0JwfK4*o>MP%ZQL1(CFbK|9R-IJSa6kUwyN_=L~Ot6{@1X(WE{zJkiR2_RsDQNuMu3NP-Jg+ zbnQke~6GO2LXIrDwIk$1Hykmli$#HGqSGN)AScs z+k=5&Q8imE%dk{IMtW%Dfw?r_Zu8AUmgP*)A3UY!9j+Fn`FtN!BZSBJ-}ijo80u{s z$9lRPr(FwSLG54xEv#di}TB5P3*B~O8$M;c+R?CAOU(jT z?l^w$&!awq(x6yjnmCBhtE-m@*m>zzy50rorq#_M8-@1OuV%u??4UInZhkQj4VRhi zVEjG!VZNKs-}lDz9F#SlxSS$mQ+k`_a|zMC?^+p&b`_zbs$V2xGOm;MDxO8$+Yc2K z0e@-Q68qJ1g1v#x&C9dWyj4FpQd{{$cn;PfopKMYzUf zOam__x>OHDPYUDs*#)x=kph8g{X&ZCbZH=osC5zg39ZW7K;l@_qa^NxUYowTHzU3i*|C*o+_M^O}fV0lh{w5Bov+| zM2O6e=Aph2zDJy*57i-;xwW8#k)f1G`jC)LGJo9VG(wgp6PI`ogT^BscQV_*@vf;( z4(7Q%GAW@G5V6O2Xz;tDg5zXBM{Q(1XJ9FE7G0V6MjqgjAMX|AiVBHQ6!eQ=RQT$g zhSVqw<|}STK603OIsQ5;ZnyUPmdmc0^L*y@nTEv-i?jJz#r0h&$&w!#C#3}r|HieP z2U-!O>r&Vp6z@Dtnim}ELJ6PtK(lSvfcRDADC|JbZrMkmpdG7`d_9TuzB!M6D`pyY zgSM8rpv<=plnzcnV~0LF}{UG^Zn!g{@&VE+phT6iF63ifAHrc zcOuSrT&*GJX$-YYFKqWtBvC815(T-fl6(0^MwtM&u(ky&ZUF*rbr#@F=Wirm&*xU;vFN@ku1d!AsZ#8oV zV2g4C692tB_5bQ;{rmN>urU2mm8aX9dZ^p43Z>oqGlf@7A6AmY-sRL+>zEZuJ6kZJ z8e(G4w=1Hrmm>zuGCiJ}y7;_Fqkx;R^C~EiGRF*oI{zCm?QWT2ew78}Dd1nm7cF z%Lz-^MZ14B`MzT?*;p=)WI)o``}%wj*Oezmvio-Z$kx(avQVxJ1u8rmin{>#z>^vP zA+m%}o?J=UWd2PZT9IUL`)8_9Y7eCorW%CS4M~yfX%PwE z2<$Hmm_j}5{3~pB8#c+v$n5>jPEXm;Y9o)WHcj`Ze*UC#*l-RHLgxR~;ew?$fg6{G zN;$FKK)|$X=&()|{dX12Z5jewSIm*(DG1npK82fwgy6XyX&r3zfqf_7gKYzwmHe-r z+yB?r_<#9A>HId+1y`xudUdHETQap}(N;(k*svC)iLGFAK`n3L7_J^R*)v1<>hH#Y zxrt5K@SVm*2DdefF60Ah_D8@<@_o!q( z+_;{{_ZWrD-d7;92f%dZ2q1Z8yBmc+!Xve+&Sn<=xadIMz5j)M-K^;Ca)7Yaoz5u5 zXd!w$Vjo8O5%OI(8?MuDBTSahJ13BG;@~Rl!|5Jfq@vh;_SW^N$FGnd{88R8PkO%i zra`vX37fVaB=OnHzkfwWc5j%=d$KSKmE!=&cF&{HTE1WA9*ubChN;M34F6?78UO8s zd@m`u*q6>}K2h%c2LrC$%ry+rEd=jqp}^d7tLBrM@tQk1FPwv z7s^e6)Ouah+n*nOJ>ZIoS`xdc5gBbmeFS@qv<_rtUeT~Foxe@#0xW<}7L9J~-V(k~ zbVl*-)|ykeUW_AObKiY*`D?nF(kS*4+ct=rI2xmM2PZLaO=rH0jEoI^T^c(N$0-Z* zohFf%8vRXJ8Sl+nj%$J3PPu@bVi>)_A`)+3+Cz_wl!vQQzg?}p=Q6yee3)5LddV8e zalVu0$+W5n`-{Ry|71Ypbm4PPnV^T7wvJt!O5M0dD&0nJ^||3`))1^?Zh7J1AP}fv zEUXWLnPV+3JF$kK)JC)Z*`CAytB){^gY08i;P$zijjtLT^&g%vu^$rCiz@BhAy{e{ z&&97ro=nA>Mv7uVG`%3ub2og*FbjuQC+G`h)_vsDotvvN+)QbvyH>fby;4oWhEFdHD?K z7Bg7N`0;QR1~W&~F88!|s9xNZ$vSTpQMr(!w2yDtjPkuOii)Bxk4!Yv+WZ-HcY#oi zgED74zzZjpA-nE?){^J_RpM!o1T?4t-^u-Tsp-70|Ea#BI+K_uS!x8YU#fIw+5w001 z@I>e=Xk`)Oen1KF-kt|9%<(8m@c~nl_;o8C0&m^80|>qgA#L&tQNTmU@?Ipf0>T(g z_PP8v`lTECl8u3`Fe^&UDN_5(PW!KuM?~SpMxC(;t?V6c5FlGmG>8Lg!4=MJ6-5|)h)BuYRK(utN~sroGaIMET*nBnryw_syRo}VE7h65@>&c^@#n^UJU8C6~XM6jSRawnqP#6 zaWw(QH{N{Rk3{&D9{$H5vcZ_*!(@4EcbmK%+(&uL9d2+=;fZ>NMvThm?cLoB?Sn() z?^J$6v*v-WE{in^y9)~{=QRwI-j(n^FI!1;f_B79}5o8(i4-12Gn#2Y~ZH)~;F88_X z-E;zxz)HxD+SloWf8|wC)2%I90NTDFOifbkK>O(5RqRV+9x0_BH4vT-QPxj(bWaqF zJ(Gai$nXy#;<(dt6#nuG(W28;K%Lt&E|j#!u>!bK-B=>KnqW9RXlD_C=B1{T-*m5$ znxf<2*4D;$0Q8~UDLLmTSky-J04rbNosDj;HXGc)X-Xppd+0yfv!zVL?cY>?`8_Tj z;Opca=hjo2@2al&F2sz^*R9;t!jO=@GC5v@1fOG<2VO!G|1h#7^Os8IC6*Gz8}7OcvW&uYtKxX*wM*?)u>0r4?MHT^g2K=Dr;c992-Tekc5$a2Y8;}+Ljv81I(%@vZ z(h`iziNmGqvUA zlXiBluz<-VB>(qhBR&pw1@Qhka3l|~Tg4whPu_~l9iV5?fo12nhx6~y7P5BnTVn@o z^?`x#oZgWnm2NAMa7UJ4RGTm%6(NW)E(YVj!Mc$ZI2PpquIZHFCIhV)8xz7TPy~yy zygGo#uRTAp9{O>kmf?1{4-Iq(vJxs~^ONy31gUo!AGj+1M7*;bt=Q6eEf(+2{FL*{ zDsnQlaF1`!`wCW=8R8MmoyjUYaS?QZ*$DEBE%xVET)56ZPW)^X(H3F(`E@&2=48^M ziY4H$BSe1a^~;%;cgISp01Xq##I}M`j$@>ivR4D)K zvvfuZwXm0$*EnWTMyuy<*g0`cW2|dh^j7{!c-hFp8W*u72ZdEMFqj)PZN3Q*uND}L zINj|kihDN|1xjzHeSypjjPAJX22ZrfnW@UHIV9}e7JM{`mf_-HT=NVMT+B@~^w#q3 z$(70rR_GJ$4YmrmE6*vu$yfgRMtbY2=(;$w5_ju6GZEuMNB4KUzjv-7x+s!7j?;gt z1phYKTG^@D&5AC(@_jldDCY8+1I-b%j1zAlnY9cu%K-Xm!YI1*jNO!`?c+)&A`^RL zP}qpdd3#tXQ31(4iS;61ws(4~P}({lNFq-d*?)FBB+>zYNgbh?DI;i0_Jd1xn;!rh?1*^U{oKDgFpp z)5kb`)6}*J6KxcN*cawB?zR*c$s`OaBapeeX_GvDJD__&F#iEe5Y%}q3BP@N9uPke z!0AfNHN-IUh_T_lpHlKdA|fP3JAFUHYkq8^Dlz=@GW)eqYC}GOFX054YDR+-H85a? zrgcjgOAW_qi2CX&D;4oK&x>KZ-4bJxnj38~dZ>?ZPs-Pu5JoAV1XxJFkcm=FwVV%i zu+fW&ILC}3It=Q*BHI0E5*6oZ7nX{s-8?Ls+c;Svf`n0gb^lnj>5)1{$j$DeD6&Qd4wQfwBWz1`2P`AFZ5h>laOlrfHjLsqL>?_SU=2b0Id-DytA9+ zn$k#sJBvc|kZ_4AA^(b-I&t6ln6H$ za^kS;%zBjm{z5}D%MgX2C{Yxj-RseKXvZ@BcMDX^%RjRZkEQhrJXH5r>7VkW2oXyRa4)xx9%xKf!Hr9e#~0AGN2LC3 zY@vNc*S5T~|d!a~YFZP!@V z0L97uZtEIc_H%!kT1pnBK%Ac!n52Oa^3W3z&Q>Q3*_X!V<^fXAaKL<^iFcwy@W?Y= z+U@1AujNpj4vI8pcKpQ=9Ll&~-OcR=4d?nA8%v65d&iNkekNXoN5(+c;`HrV7Zhj6l8PF75v!#1hbkqDnWXk$Ix)-ZIqcyuE5U;E=HfPcEBd&!2w zR#|x-bvsULCYb8?x?iEgPRsHJ%euf@r9*(AUVZH!zX*4(znt-vh z)QbTi%4B44yL;cN1I!2AUwJa0v*I5>rsyShOO`csvX6smf?|!G2Ggi4`8p>iWC7DA z%}wzy`*~skCL8e=_ZN!d-diU=UE-adBfz?dX^k?4-Z?#ut_@#|WalNlbMzDZGb1g^ zjOqD44F7CWzr5DPZF%`62MZvZdONNZpaS<`pNZE1JtrCO`d4;RIWtj0l-9mzB($$A zs9Bo|o8)EXA#QAJ{L4rAs+8wpYn#{(8UEn=#KzA778(5T3xCU1v<0~WjuPG(9*_OW zO2Ex)%XaT8*63sN$Cj_#=f{z@LU{GZ%_37EH&|yqsXbDh}})4vUC7jEDxxT=f&P; zu=k0jN_}tlDm@z!JTZ+J@QdR8v@OH^-7p|CD#X?somLNI0cFu=z_7XhKuWEfj{HPclE5uhesW@W_*wENY|3^*y-N5igha{OOXXYW|@V-d%H*O z(Tz4tHEIRtKWk_|UhhVI>qHHj)aMrqW%Bw`-nnmFyJE8nM;sdbaX)ac8eK!x@ua@s zF9t^la+AdMw+K>2<8aCw5VL68O+nDf-&OB`;akP)Mp@l^3&yM3hmOlFZ;=*UwY%KK zo0Lz@jtBc?20+@Y?Ih)Y9$vBn3}TpSTO7>0?-co8(vnYv4$t?Is;&3?u{MkfK7$XX9vzEwJP3e!YZsU?>6KJRV zc5}@U^t|SV!E=4V&zWsR>$HWY;-p2z#rvcCVIFQ(4bI(Jg?lM!fvlrnA(CArf&+x-rv!T5UwFUY}tBC=^3nU*#BeEG(Ce0B!gK@#iy zza%o`j6dcbwrQz+z*rW=FA{>g-|TzrxY__hv|+mS)=GNMrZ!^-;>@=9M3$OE>gw3~ zXHCu`H-<+gKHV=Ze!;SDqm)?r;lcID1W|~2#=ydj_6pFAzXo!_9R(!&<)o$t&7v3$ z^1V#5JKwr{lZ5DGA&xhjNuHW(q5+g|#%c-2)=^cBz)Xf>6wiw8?ONJ@_Ga0zv3cM# zjX2DjkgiX9aBpqJhGQv3tj3qGjU^Q2}$L_0dKJgu0PYfC!Z5Ka)n0%Sxu^k+f36CA$NnIU? zY^ayFVF0ovvB_@wa&AAB+9^9Bj8dCfPzY&rLN!{M$GO_x)v|B~ekJSk_c%h|*ZJ7T zGXx_~0)d4fq-wenZ1bS!8w)5=KFPcbyBn zObyvIMm_bi^|(+_$OxwFZ5f9nVvgZ2x;#2~p#-Uw?#>rQ;I6ybU#zl(F_UC{aX+-F zUZ+{lA~rItAi0<(Y+m6h2Prvk9gne^ZLFq0^swy7V<0a2Hl`Z*<~<4lEJIIsz*Q0imJv4fJvg`#FL6PWlP096LWDH=v2`5JWFH9wAxqnP0$eq`1$*Z;tP}{a$ zdj;9|KOj8w544^N#xPn`=p+K&#~vrnauN~})jdxldHYK(JpFSat@=GOpA^zRF~HY> zOJ5Xbn3RgND9FnnzKxZV&tpC5wAffD<4ICd#*iQXS*#ExL$&y`tI`jsEdLPj!@8Lk zzg+?t26c<|__MvKL}WNg!?(9236|p`H@3#czKAB2r+%1J=W`Qc)pg)DHgc?v$ORl{ z!+WvI8XM^!_RU-WA+*8_Th~)utT37pvik~7U7Ge&nJ6Xwh=|HZpwa&k7ymC}=A00` z!nI6vAEH|4F#xJ@JR6}om!5RF{ulFmMbC^YRdBqLl2@QW-6D|}!zu}l5sn-G`KhaQ zryfBLz3pyb5GRmewd5T6%UR8$@#Vn}u(C0kK=ocjF>CCSv-o7Pq2S+8wf6?$&0y4Z z3%tA74Nrt5q-qwt=;YGOo9jqml+i;9UV*h_WS~ntZ^G3I7k$=G>Qme)&3YB=!cn;d zU;^4u`#@5S6VRY@Bke(SQw!v`J6srx_pis6+zqT2S#^C^vMb8OO$4!X45!(Fj2vf+ zJvEYZ=(3Sa;x+-m--1|x3TWek-)kM3lk$ZmT@)Q_ORHNoMhkR7Vf?Ml#u)orqxw#W zSxibT2c6psuEzdUNZUZ8;!ucv75Efs6@xEkFipQuj(E$xchO)DB@ufm1PQ)^PWb)o zXJm4A@}Qihl?D0WQUeLNv!8#j`1o!cKkBWsJ1VyL#?sQtz93FVONdd(1#l*WYUOF` zSw<1Db_@Rt1L)mv&$*ePdjnzjC$*_n*R=0?6@0=n%3HguQAqyfXt>K2$wz7e?h~aE zI*za%5-q=?-+x|fu3+H|6`4^dm6d+}yW?sTre1b5oh#h`Qf1>rWJ75uegj!z-n@nt z<>wP~0G884p!fZk121`=|Jkq2nKU$y1BYdsh4kmt+wnhF*tOhZNZn1xZ(g-jPs&p) zFh_35<$4t_9@igms?GP7f~+%LG8Q=`Z9T}T0vw7TJpL?WIPZqkpNYOmjCB#Xr~?SB z37nDB#_KukOgETb-W%3rmfO6o!BG4Ny*tXCuy5R2J#P58RI=m(Qr z4NkZdlIKFkU0hT@o-g|PkJY)yS2!N)0i?ByNNi5S-Ch0wlb`Pu83PA&+vd5}cS?*8 zo!fYME5y?0p_Jzv68zwyTtdJ&7|7}}xRvTHQa3Rd?LuaO@dD7&G#x{)t zgk5>4Ws(@!#A^|$naKVycx>)ki!B)Z(H18}ts=mlSN?&ac#Ol&j**pCB=4mc8}>HGJp=>ckya zM92PqNyW$3;?$YuNI>XEOZ7Qn)T>+NF6$Uv7)RAe_KcKZzZ0%Tx@1*^yk3L@u%@oA4 zedDIO2%4VBq}zb4#Bo!A3`f2k1feAWI1ych&z#Az>8y!yH;v%Q{Z+3=;ds zKA*$24q#$VAsCG9f0u4BMP{n9&Gn&gbh`*ZJg=AYSYinGy@}NXd>V~_2+s`7CqQJx zoa9bVf0>x0{oS7kTESK;v{&=_dfQ+PwyyzPgJl)V;da;{OfadKoVvEw{;StUFgh?R zGE>RRWfAW=vhcVygQgedl&8Z`f_)#&E0Bp59b1i3O6TPqsV8{eJ!nX6JCBXV!2Gehpj~ftP^*OA&A_ z1@jdQDD?mNz=G!@2d)lQWMMO7;?(P2PUf%_RTJmDcJ#as*hYe!F)6+GXpATl2S$Wi zO95P3Fo94sg>^*?kr90~APIi<<0JN4yFK2tk%HA_zuimzBvbj7k)z7q8G3ZS!@um>E(oC%ii40a=hY@DZ-3501GZN72XKMkiXq`7WkP z&ogBoni`TlKOz=wfA{;riU$Q|r=@7E=b~BJ_t!hka;YQ*qt_w>ot8hbkJ_&@RMQ=C zX6kq`maDs0q@)WrDg+!n&wT|0H4zUOvGn@kqW7$QEq37w0(t%@)$&Y}-f@)b3Og(J zcvmC4W5+G#2z)v*O$1fGXx)Q$cxsP_r6en{?<%b4K{+1#fFXn z>gj+T$e8u~^&@gdJC}29{vs&_8L;Q6Lrsm|0)KkhY~21$fS|L;-TlQS5-EvbUO+<9 z`DjI}7OW@S8H^MKJ5^=`N7m^Z)5^wNuS|nPVM9Uc?g^ES*(L~6Fip^}PD!-;TNGsl zseHl;MWb_v*FLYlu9F+5&8(GSpZz z&)7mgX>M=~RgfomH9&B$N)t;Kp)6W|$tUKyT(fDUmHDN4LKNw_A5r_tz{U!U^zbmc zrSx)BP?xz$8w1@hn;Wb~KGf(WK-oP*uQyC^8nilZdZbgq>ZrQm`W&_Yi~Yq>q*+gn zNTUULt4gDg=U=!^IaE*%dM49VDvkGA*!6c`Fb(x!R<~yvG1e9t4>kK64ga=ROR7DU zPSZplY_?|hTzsCuEDN==(^GI6Mf{u0AyVCb^cEkP2m0(#Th-)eNK(1(FuN9}n$b&lM`sQnb$VU`9vqcT^tHP(Os^ z)kYb{@I`vSRO;A^?TBRYU2giE0NPP=Y+v(@2=?MLy3>(!P}L8=J2b(*vtXc;*`d=p zp7A1Jxjh+=|0b$)Oi_(RA~}3~Yr<@}+ZKZCf%Rn6fw>43|8{_4Rc9OwUem1|+@{YT zFxV{_AZRDzj7R)daVhEjrdy=H9i@5lE|I-}X+wGL-{Y0rT{Wk&0jnL}G{QOX?(7u3 zitJy_jh9L88w;U$&Q9%Ac}hV|S!bgR7p;X8+xumVaPP41$GB`I=T;0xhr<4KcgE%F zNHs^G+%4wuTm_P6TSwn4*~M$L;rR+Q7^r}#7w5bY>mSyKfUX?SoaFb3W@e zMjRKUf|zlJ#Eko|N{Ga?ZNu5lnp(3pbnFgK16?sel`#i1iEX0}gQK zrcqt_y;83G?XPs3I0;9kw$!Nan}PniSao)He4YB;xOvNT?a_19=fe{SlnLf4aH2pa zpcyR>SYg~SG(1z|iyslKZ>uzw*gEv@H!7#66xAChJ#4T$bJhGmDz34@S$Dav}#_k?|SU`8tPodfTzAmZ#Ca#pfx2;d^rNG9VjxQO04J=A3 zf&5&(1jI}n0n)kCq=XIfdb8-9Ff|6t!7*GZnfJ$#MDlrb{S-A%Gdk0sK(bGk#*&;h zP0N<^DWc`K1CJlk_{;>}TL)>|14zT{NOWn0Qo7N0HC6UCG18Gji4b$evV(0t#|{;B z64BA+Lg0GIiS%d(JnY`O;6FX_@Wj~U7Q)Ls;5Q*#RevKZ3w&!Z&Ci;PG2<4%=an7A zJnUZ@i(Ywdb$&vB<8}b@KXil}dzlE5dNJtwcdJ4~K6A01Mo&SU0haq38%y8s1|0Xd zV``S&=$9G}Bv%aD!)Ngj{WcCu&8thYgt$LdT8F%eV+}6iJ|(qw0%*HK^Xl=v6mB_3 zFkdSy3PGPDt_{|Am7U+;V7vWpVMqMB0{Zr~(Q_w;kghiS=B47V4RkATS*8evzB@f~ zCS|G{<$@_g$!GTl1?ZLOi_K!ak@_%4dTk@o4|J=l-m~562~%H}{6K$xqdD;^2>wew zf|3}5J0k|+UN&QesP=`!=%DZ2Fm>6j?05~i=`KDTz);SR!&G9>!z5%>TVWg z>a-gYiCQwQM~$=DLZ1ACv%f|_te49hWft_~dusi*6Ywr%2xNNfRF}H;2)~fKd`I_C zPKjJCme@vnO5TSa6#>>T+&#`#?91gDuvVW(2{8`;U$LO<#`*>9#}xXqS6 zeB;BKMAsyIJboz5U4p{mE&@dfcv|cYgM#lA`)9>b?su9t3k>;+#4%mrH%LRL-)R@@ zR$4Ae?2%oy+h|k72_c}4*5sidr| zCz>1_j6_0lzt!74n!D4Ud;Nfnoz*)rMi`4xk#Aafcf7iu^Q@_z620+Mp#SP;f4QaX z-2i0ipS%&4f#P`nl5qD2w0jdPGVj8V>Q!OD;|hO<)&OLSQ7^*yonSCpdYS24G|>}B z4LIw!omW4~E0<&9YT!yEdQNK$q{CFq`ndL@Quhb7a0yj$caIaC30#h!E-cEv zCFbmge~7)82X&bON^5w~rC4cdKGlvB>m4L4 z>m%cj5Ce^%0vm;?x!UfLM7i2z$bK*t)QVn@)zpoAo?!&hBloW7LMP(q53!4V37uK> zT_NEJksH!7Un=K>oY;&pWJENax;741TQtsWH+ooO^1j*yk@tQFlgHNxCfaB1ZE+S_ zY?qroyg68z|3Nc>p^yj_mJ6amy%Ke%Xm+(|P2^2pD>M|U%QRR`!K2Db+s@}Me~gP9 zUA-s9_i(TN;_&n-l^|BaC=;Dx31eq(4Lxs)FwRzv7So4K39S*GcKSQsVW?{m*KZhx zms?JGeV#~N!69Pmd1$p*hO+H`G&n5hkSx*I#*qf$G=F~kju)_k?q^IVpL^0AGstHe zmzg4CMtJOc@D9Nzm~kwL-8xuf)pMAP{LMij_=J~0(TcnMyC|-2(3M~7B#gaTOz@wY zoC4-<`Yx?!P4`^!v*RPvuzUxm-@W7LN8Ct>#X*fr5>(37eAsLR5%EBA2g;9k9qQdi z*>rkza3UFtsofoZgTKAqgUS=Z3?spmBjaiJ%@%Xu-loG}5LWFb_89;0ysJ{UNSZs8 zb2;iz>Y*G{j3osJHf1!QV_wk=^*6;M6^-F>bP-n1S3rQAl=Dc)e?%I&sRTt2fHwO z!v)CkJKtxH>O>+SpE{K<*`O}D(0+T-DwMcf$ARf7a3!u4?}`g`z%nV@<320}GNtuv zv~?`t#KxcuL~YlW`DiSQgbAODuX?-_g|fD_#nH&1R*H zTH+#d@VMXhNPQrF5d~XG#hlNn)fpS|9q~)5p#$?3-OijX(=ZEG%C2UCR+jIjt+Eh~ zh1GU(nOCu24fgln>6_vB$}5}rGfu)>o`;&c&7GRp-XltHuc~AB!F~VPU?`dLjclbi zF+rSgcn8Ht1&(Q#xFr@c-*38rWBh85ccVZreR<&-OT+NVB09bCk_i z@(K_!f09F=cQMiitMf!LI>70YZk&3n7 z7J-af<&cK4E5UsJn;HL0sor3zRjT3Fp?72fD-8K%{J5ryAcK*l%?tSlZ&~d9z`CrG z!Ulw<0+?Bje?3%)b=QYf8u-t-D85`nG?79 zhAM(3n0)h~scT@PJuIZ;ChOWXKC!}mrWQaPoj@4YipoOXG&MB*o-o4P3h7Lfq@maC z4FhC3PS4D=$}Y;V0mcV|0az}l}nx&tZLe4eT!gU;6p>3%Qf>vWZThCpjlVxs;1xo&G)8*ung z`{{H9dilnJ4MATj=Q^abgM*@%=-LkMjH9*q;zCROb?lf{Ls=(#81Mn#(^=84>4+S4 z=o3T(%s71-V>$m^B^Ifzpebv zHC7FiKi;p(fQBI+`~LsCPXfBn|GQFj$-zL-(c>5C#h4gRo!krL?%34kF$C~vjZBzE zf?WCklv;}W?tT%wlE+99;Ma0)Pv{pZFU?Dv&A7}FG;;arF${)i2bVg$Fv6!!P z8}_h)YOk|3Rjij!puRk=Fu!eH>pZ=!G;Y!-#ae3p{z&jy@#E|MxM}mcb|0>FvkAgn zr0bnOPx@?mxu4Opa{Gz~EuN@zLO&7B0CJG6MEFc-(VT`>vaIC~bJWc;m6r&JH3TA8Gf-?#S zzHfNUgo&LHoJxf`B1*%>!#83hG#2E%!56vY*O<{KysY>L@WZ=@U$W@+>q^X=2c#hp zM>gI(mg;pXo3Xd(|7v$~JF5d#+S@u6gR_1HUJtsh@Q7kHU#vwnBb}`{|DNT+?rwHG zOjAKjGrWWs`gH~0q>a~>^}rpkzS{<1nZcQo8fWMTH3COD^JSf?+j9d5`LBk9<* zJ4-hry*Z{Y~7DrN=}q%)0E)HY0MHM4{U#W1(T>go}=5UHP2vLKuX8kg$;89CDd zR4fB`S`)q*fOMy9B!#UEf9Oz$2&=fdh0KSoRAGsy-QeuJ*2F=Uq6Y3~&J8 z4?}%L29dw{;okBhwom37@vk_t&Iqi&!Eh)zWkMy^;-BbW3~kB!5*Z1049JIHpT*to zsThIUobBy_!ok7Wgc;5#DJ!>;$(mH*sBC45np9~X{OBBF=4$XFzuTw+*$(Nj&v$%_ zGc?q~HLyiAevF)uD7EQ7-NJU>KAfXoxq!zW@AA1GQZX-}PaDxM!-DIVaZG}U;W^7>z zLOz=wIvFy{cb4m25m(g2n6>0!hh^Q2 z$pjyvBzvCW)Cna>an$T>XAtwEw_|*oj=}rqejzBVJYmoKoYq8IANMmtjKnZfB%Fu*BOU=*NGby1 zQe|rQoAho8o3;yy29mM7-QfG_dxW*3Q_H451Av4CaIC@X+#FWz%~1+3_uSY7G-!#i_czUT?-qn}_z5CKqWfIR{ps~ z#2)|iM|MxKSyW|L*)gT9O~7gNgq>d`j3UENys}e?Lo;CVcy}2EsL`Y?jOhOUh>MLO zU_<6+7W;Sd3t!UpOAoJk^NrxVU^VzQD7wGKu_5hp9l`tXvN!VCutW1Q;~iI}WjJyc zq0I&0IQcx1`X_hP7f12sn<*@P2#dV`^OhS;|4lARz3th&eHB-_P1BY41@Sa}l;i`G zL+ScKy`-H9(>lumtw{dwq%TH$c;2RLu+J?>sI}10x8@MzaApYV`hJ;=$EX=gKu~C5 z10*8&isq^|EPfh>y!b!=h=hV!xBTn}Pi6qYT*g_!Y(XwTpc#j{ok1{q(7$b?7)3PI z+4o&gZ_mew!Ywe*pYmj#uiXz}kP-$Zjf&2d2*FCjX`f*lZ}-QWKa3*kzmsf%l2{LT z)fPAS81_GvGCA*&VyL~!POI}rN)QVqc$3>T5G6H6eg!lIz)%Vi^Oe3 z06-nHBQq2K5dZg?R3`iCNLMG1$KEHa!`pal1NjQqIdJ!WR3_3k@F* zM|QgP!hd>$Kj)7VZ`U+oJLW!eLtaRKcr&dO=tExiPeNa&Co7?p$0fw5GU=G%ug5Z0+E8K8#`y~%oSG? zETRB+du;Cr*Zp_}!sFR4k{@~ru*3sTqDTqpiC&ILG#C?~(Z3)+6D!}brMIqdu<^p@ zbj2{s#z1WN2ryx}!XMeGvxTN`vonSpvaaCfCM0}6Lx#`oj4rQa4;s^UDyw9#=uB;4 zJ0tF1Z^WULg#;oH_3Tl|_lKUvq&!p&CS*h{En{PbK+`opNnbfgnc6Zj92g(coh(#B zNy%UhINc)H{x}Wn^vGl1y+fxj?S))U^I#l7AzLQ-r^Thvt1p5BVv7 z1~8BAQimUIMP#pl9yJCMWFz#SRE|KhG&q%|O!0#%Zs5b@@_=5hH|3SK?lw<_?RXw( zAnP*0dmsG`WSzbrlzq)LKs!oNgF^r!s(G^i%l+H&VkPJt;Y*P52+_dR&HpIutD~Zd z-bD%N?gjw?>1OCo5m358O1ewBJ4K|s6zOJY7((eDdI0Gbq~jg@z4g|8YrS{x`448! z?7hGJbxMxx9&7FwjaQ7kT}WQfwlr!D5u)zkczhw{>5hfLYu*EcY=M7Q!_Ve?|NMcT=$fj?Z9E;wks&uk5OA0rG` z>_3DC394nfU-@(k&$(LCm!60A2{hRk*wx!s>}{c;?isv6OQOqss)$=Z=4HIZY85*}hueU)wn?aDZz{bSyuzevaGb^m{INh26BErQ@VSTiIISdSU`bgUBq~50K$5v1vST>N=B5eB3!pO&`eosSd&(ZfkRnB=917<^%Dv-)+h$twe0`YLG0BQ&- zac+;fr@aHppb0=~`IIIc+H`u8z{kfJp3`GQ90YipwQ75{6%Z;j5WZ*7l>PB+%& zPd6xz{p~$BQU2dnESQls*JW;L$@Yjc^OXru93qtfhG`wI$QC~TYdA`U&3b_U(%S8E zK^n+7hKPDDI+eobBs2FcUAsr|Xq?@zvo}~zQhIuNW};`|h86+m`Ds9DAdJ$bnH@Hp zC1y(B83FsQ&>CRK*6D`hw|WR-Gx;mg;zU`sqWcVJc2kP+hwbd(;t6B2x`F!SAQUDM z4M1d8xAf11C6cGIB_$V=j*GVwNa{L`lGyZB{81*s7Zi*|GpZ@ z_XO74aSaHV+;|~hN4|2)P^}^g*J=B%1eL)^P1?gsm3Kc^HO*^*-e7z3i28q!=L1B7 z%fjzjQXl`Ik9GbCR#Whc==rC0m`OrY5K%_NFy7NrWpCLnoaKlBX^X&Yt(c{3s!Q5yAy+bm0_qPln$epJ`ZA2Q0#@(Z#v@lonNm1+wbPY(~15`LdF6NJJUKQ1)i} zS8b=GEc*O|;FUlVv=27)$<_=nB#u4`z9h)Qbn6SlCaP;n>OJn)dZoCLS`0S(+O6g(=$Fwp;2i5XX)UgOs|w1}aib@{P$#<%jgmj-BWjU6ZjCx1sHK~L*&DXtF8BMGI}R>GeItcVYv=my1X~f5_{%0F}>^o091BxM28z_{c%tk7u>)R z2k~+?y%i@_GsaQV5zw&MIMY$CRk*L6YUAVQ4+oxGGe-3PBD7v_jgFzVJS+_gdTv*m z83CoKy5i-uY@v@Pid9na7b&%QeKH=oAs^i^`=;eufZS#ZTthq)>FaBH zl-xDxYm6iK=k2==E%PXyCv?obihn#k$#04WqU!^r=2%mES8*^|PG^Pv1A|`5bH@^6 zkOh{hK6Z{bCvjq|Z+1U6VQ9D5auwVR4q*)Y5(6-iPpK15KZymO{NSrUmsp#!ONu2n zqTaKjvMNN-W#07K16{|Y>PziJ)BzpnRT}7I`q@?CsP8~5Vh;Njz|VcSS+1WDhg--2 zo(v*h>ACW)?x}E4XaJ=IDuW`FL4o$irL#i-|4BkH0!mjH3cQjtIVh?ZMr4EJR{u`U zjT#Rp1N~-I0P2*-np|6=zN#|p1}$A~OpTT);vz5-?H5?0c)7beMKaDhxWCc`cCoe4 zTxM)Of2Bsqgsfc$>ff)8e^*)9>WN4nW*s-ypoH{QYvSAa%VQp%L7(-ztZRN-2G88U zXiX^taugsVDQYph9WQnwFK>PH4l{~DJw5Yu5*#enxvQUQFi!39)o#x=vR@p_)JDF~ z3|D=1R}oP^W(L&kymv2W*y2sroXhx#KSO}}e-~xW) zT506uv z(##<{i(efK+qCO@zcQ<|6zl`r24#NhJ%cq;~ILir#p`3ye6yZH|1r4giKpe&a*SVr)Q& z2Lr;r_;s06$6&;h*vRbs0?;^Qc>Din~YxcAK;sGQ5}(`O{tq#T^>?}V1ASI{f>M6m&U`y@N~(bN(W2-x*XwFx<= z76o4ZeBT@^`vJ6$kOUBSfSubJcBNbAi126!C7o3vIJ=%hw1DdL7Ks5}9yP;uPjJ>N z96*9)Yu%Gcyj(UC1o7X`0O|fp85rC6AhC5^_yH)TNPpx%wPE^L#mNg;bs+#{(n`jw z>Al|}RN5oob3z|! zdtYFuEQE%$g}Gtt+ArVEjR|vK6KN??e8UR+OMSDzKm|}j<$&nx5C$|)F8k++kv(t} z9qlfL{Xkk}{IEC`XcWE9!REQ!I(fTXEp$$#dZ-_ngu#t_cF=1s^AhMc{Bh;A-0}jI zE~PtF(=p;iTPwZ#NnUs2U|#-I4MX}`7rDPPK_TLv6xD38U${b5cFE9pc!fsDOF;oQe@cMPyCA?Ai~VPl4KoT$#q|LZ z`cHAV8E^aMN754qW>qHI!jV`O&|2{LRM(XIKp!=dXxYGLhO<&Z%_EPCGrTfIZ+^A} zK6OUqhTmFrs#yJcNyV*pycs~Z6fP4Jrnz1H3OUi~5e4Od(u7bJ=Tuk;o!|o{HTQUq z$*ytiC&1)UO0!>i_Ht(AA->1n!v7ZuAR!?E`V;{}HABBJak1vp4{=fFIjhsV8t z0|f~8xV2$t5a;qObe&H$yP5P84;!}G_aN+tQqLgv`G~RZj?kW#QKtm%)s>0vrV9la z0X{x11;y~JgNTTTmME6`i(d?xK&GDx7xZ8FVN_X##M~G@ey6vB@+e|}f259_jM$Lg zd=x6bkENbfTzlptwtM6qont&~=XPw@~5uhVuopE4_vw&|VqN z4Je>s{P%U{dYxJu=);3Xb{m1tVN-%=iMcFYMkN)B&R7cUz>Noj5cJ|o-*m?lhL<}m<-9N5dpga>op0xS9dk0G;WDJr zyt>|uJT|?x4S9oIni-%ED1i9?6xZHQw*o{dO@=$<;7@1i+BF$yG#ski-lsf5^A4$0 zBV46{Snf94>306GA!OTlfO9qgSv3H6J_VF<``gcy5(tpmJyHg5x`s_^|J zzK_Kx0mBowu>N^~V1v9)Ulyy%i-d|tx@_|1-@C~ItGE3lUgW5wi1n^bWa`ehKfSVQ zgCboJ?Vh79+29*$ye_dFuoK}RP;3QE$-Qp3)A9%d7?#gJhnSgZLXLlIr2;jPPun}T z{sbo)HxeN?4R~QtWsnBQjP^5}di0kpfh=GYkvxN2adC8O3i?i>^XUA12GCoA;u-8w z!3$T@8YNA<{M&sUN=PH;?phTXK>PSBkgzynqH&4jP(!JmW|QRTFTPlE+OVO2o=Hqd zV`TL8{)x`=rRy=MuJ1+BM2<)jG9!#lF8_R;z}n zn?XTKs~5G#A-j1Oxb~8fn|I#xuw__)cmRm6jDT44CA&j^#1a9vRWBf_Pc12Pwz<>JN(1|%R4x6*|%cj z;W<}Jaj=IduG_&tgfE0CUMe65M zC5dtX`gv#G2JSg6F%Zh4IId)v?*4nELDiUpA^s2az)F>FjFJZ23uC)#l zMcj80mdFEFJG^5fIldpqvUgl}a(H)gQ<9ifqBizyqX!q;y1OzQDm&Ha=*p>)t?3HQ zL~xkQAF80P!~M>F7xxa71RP25oPRH6@BdO}<08~UdkyitJ`#a%MBbz%G}u@%HPY#% zl2$}sZrcaqup6;la^yy+6`{Xj|I$-<{>t84g!HRCD=A*8X%HhF{;RTWZPHxjX`-jA z8_xOSJG0l0vk>{ro@xaV67QC8`hT#a?zn{ETM|_>*H_>|H9JhahA|e*>^JO8}ybfN!Xp*!Im9e@w zL2o{y=N4IfIW@?iNqw08>Fipi_G9>r^u13Y=f%l(Os_N*z6M{E>X5XiVUu$XyNa5> zgc)3Q#Yx@Jk_v;^&UugHsjEP!er46V2p!(lQJuZdC?vwTgm{<7nUC&8%GG7}ML%bO z+-w)qR}FAg>(iXr@lu54kdMFuoQiCzD0zvuga+o3Z|-M#qN@`{6dzdXE;8%f0--1r3<)8h{3IJHJ_$eLOCrZ5Qu){x%hHh2z z#q$|1t@Uo1nSFH^(J(rNVZ}yI8v=u}Hanu>of|nNe)We&X^FU{bK(;atZ!{uUVsDA zXCnQVc4wGsKMd&0e!7(J$_;Q%Eup*Zc06p`2%#+?z&p%;5;xK}HlA!asj%xa^UtCN zHYKZhx2o$QkKS~3`*=k$wM)p8ff;N3mlbrLqhHkDXdtAXXYMkUbWI@?rSeN&UppYx zcwV0&jS)AxE8rgXq;WLORm9&7o){RjAoS6xZUBhYJSJffA zbxTCGv_C5>Z!ES5O&^&^oCO%EnzXlT)PI#Yl4bhk@Wn(-En|w1c&1X*0~Ex1|Ap38 zG9Z4eCt*m#@^5xuU9oA?8$>F5eG0SYRotx zAQbG!T^Ja|uEx2LvO8HM+h@CvWfi>8xj82A8ZMC0_Y^la3c!C5J7I(*RW*+n@@^L^NRDJX+A}jaTy=a*;~{# zRu=K|yrm7knfM${>)*v}U{Av8)Lwrga*XC92e$J2_6;s)H;}xNe+E_f(HC?f%{ROj zs9S`C@|Vt9O6;(R@<2K|I&;&YQe@8GMXxEloXeJ7YiTMCQLA|)n`KWDIQV_AYEy4i zwR7F44BqHZ{;duy0a(k)iYn`iTP2$8?{fz+QZoz2@H)N>b#8y0w|tKM=dTevW&UiDz)dnD1-w@q5XYr&aC@*jE9yXHSa~sgUuok}WAWC2>zC zcC|S)9?6i3@E!Ao$evpE1+4o_x8jLe!*-pxE!z4|Lrt@`VvoS16*kZ8&9%2SQ^&0& z0Ulm*0BOF;GCS?-7@0sIpX{#1K*?YKU~2NbnfqjO_YDn68-M%dDl?y*MYPAXuwo^| zY>24O-WDt`T%dy4Y&MkY_%UkU3-P2QxSa(O(Kmu@je5 z-h~WB&QVM_x1^?o*ueF$^ujrWc%n+VOy^#s)3B0#Uk?Q#!nVo-Sxcbhtez&gZM=qU z8PTk0t&)dW7F#h)6+|9(-bo=L(uSR1qnC@-%6fIX{>V%!rDL zvPD5`h{hX)iBTJMN;pSm6=|^?;wK7-o6P zzGWiB24#b6BEg0#A2Q~k_#4G!8>rZcUhe&nkN&pTj_@LOZuU9TdSLbhOTSxJ@P(zM zeaZ3iFDWoUlyv1Pm=lav6Gha}$8)XaYLn3FHh#0u^|co565s~U3cr3^+|bJ^xfSQl zCS-$4M%aA)_bI+HIu}D26XMG0ebKQ8&qxN8p+tB^9RO6+MGu^V*?{O#JU06w9++8w zf9xf)mWmWDWhFbzx?0k(j$AJj!9I)Pb^P)>9SC>iM@&0fqvsdjq9Hb*?Vm=8d2${x zt`L7rBxn=R9U6sgSuw{CNXY`v>>!-O283W)-*<_;jr93HjG^$E&+jVYU`{;$E$j2F z2+);-+*|xFWJV0UkS}`>C!0etW=Ni5SMI9vCns*wdGpaETB7&iAF1&1@MZ}XciB-5 zNa8Q2=Rf$}ZMYn^A#Hvi?1W^up4TjXV1GP$Pcbvu{!Ug??%xuBLx`i_PqxrIg~N5# zrAYqlSS*a61_}BdXdCPl@h8wNEsG{|ae6DQC}I35;?@_qjNInF3768&Jw zS$967{}#FK^LfEEycYkmjh=%}!&uFsUC7qhr`1-v-_VNf7w#9jTP{C{;VC$`5 zS{BsMHkv5ecy~=Ho9opF`*hAV6gka2(_v@`C5~Z`&@-L*EJr$1TuOhc{t2nYW1qtB za@46MDJjHI%e>+Ekxjmx6qB>0cG!yLTF6>(IWKT9guni=6kmLhnJybjIV>Jhp(D2E!T;tG0=ga)4al2x0 zr&+)sq4W8Cx3NcSRRaaVH4;3ssFml3m(z=@@2O?)495bA;%fa zv3c>qiuuk6-);g@fm08=m68*HfR@?gLp;Y6g1+61CuQQSYtEBF%pKtKXTdFNeViVXO-i%ags!g>c_% zU0RHw#n)M&y`n-ecGf{~bzh)-z}0tM{#xa;@k(I}O1<#rGF92Z?{-|w?<^teAVZ>) zB{llLMYysMkHKuL;R$tSN+m%Ir^;`?Y7NUeqn_A|A|8JA=(d9;+HWk|BLM7pyPS8B z&Vd+c&oy=imU>}rIiYEfT|V~J)%)hIXFY3E(2JI4WpQQ2t#N*ge!odbztLuri=`)u zrTi(xEsJX_l46&O0Y63g*0#YWR&~|2{oz>N&-ELk)kU49gteHK<~-Oy;-!IIg2G}m z`>f7;WIrxF@b#U z5ACj+xNSSB6_p@End2DP{^7vQ>`qUYqubLMnineH^RNIHw)uCkKHjt45f;d+|4qY@ zj$I99lq%j?f`$zAIa%{{WU4v6;5Un;yTpMIv^`WKZZz5+#EG&FURMq(h}WIj`X4jW z)%_2Bu=iNo4#wTz9GUQH9T|J8#S@Jap|qegufv7EtULyoeW%ozK?e; z2)5dAYtv{Jd2~IJ9y>+MtOfSs^vdvL5M@68V7kw4iJX5Sfr1rvR}W8i<-O94*ns=R zX!!%f>P33!C?bo#0&*Fw85hsc_WeZB?p2Hwx;5A$0o8{W!IC^h*j@HC2yK8R87-x! zWM>IFQ9o0dZz{^aY>nizoBl~OHat(GE5^b|QefTp{-qk5(8!aGgF`0Vx_nJ<4#C)XrwZ8-jo2FN z8qn`l@nCW$fswwi_h~g|_AcI4f0{~$C@0KqUpq#GncB37h^PM*q=XWx$l@>qyQK>q z9{i4#QoFxDo&NNdbIiY$pes9}F9*TSyU7=yen2#`30qD{kD?yKL;^TxOP}$b2j-qLp&XfyAZ`hE8GZdq@fk56wqTRG}SD zkC+mVWU(`%NZJ0v8)rQ-XKxM43X|Bc2HZHW1~`YiAZa^3W`=MQ(owtXkRlqq=MN6+4njqdHX5~i;cJTe#&R8;7qyoxM?pxJMrm%p;>Ec?Qt@4_oE zWIwz=wSn%y`L6p7oA3mtQx#>1)t@`3AH(8tr#~s2TkkL{q7UKZ7%T;xR`B4zqTu>- zP@|q!hU(MOl|Mb-Xk7mYl1dizi!ieqyxPe?5{a?QpHb;XOf(mts*f!U3 z$KHE(a!fsO_?$}W<#}nV6|6pa%KoWJR<6q=bjL~ryyZ5W+ZQ*zRTp-{U*0hX-?xiC zpabY6{k+CLyAZi=YoB3CP|GWJ!0DzR!|`@GJp3e2Vcu(1MxwdgQ?n^ZCodyvSHlk^ zL3Sseo*12MOiDx@6Hp~LUerdFcN=vKz+tLpSMKUgg%R$eXA#GMNg2i z*Ly=B=DwjV6NL-j4GY+(@yjgsi@s2!j#mZK>o&rCTjETjlPg)MS&9lB+Hh>*Xwr_Q zjVB~84i1s}kER7(^{@_ta#7ux(bkb#P9JbuD)wWV!6Kv^M+A7Y%FiU>Q8D{%7AjDF zfsf&yjT?PLaBQRjg!s>EN=){&mP}_FK01rDA^b3?hK#JLs;A*(Jea57oS9pK#$yM* z3W>bqa)&QPS<(_Vb_-uZM*yV;tY4z~nYn+z6SkmP3!1$PUadEl$=_eCjE7jvYBy5JSxl02^Ks3;{du0Qdlsy39&A_FD zj$(DP$ZO$vP#P$Bu#e(&U(glahW!5&m;aWw0hO3ejO%~Weg4l+JfiT23jQgcJSm3* O9wj+d*(zz%kpBhkS~i{l literal 0 HcmV?d00001 From e042488924ea5942adaf6526f7e6e2d471d56624 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 9 Apr 2021 20:32:51 -0500 Subject: [PATCH 02/77] Refactor NFP mark sample IDF and IDD changes The NFP mentions proposals for new keys for the GroundHeatExchanger:System. This would result in new lines in IDF and IDD sections. Sample IDF and IDD are presented with proposed changes. It is helpful for the reader to quickly be able to distinguish the new lines. The code blocks in markdown do not allow for highlighting (e.g. bold, underline, etc.). Additionally, html font highlighting did not appear to render in the .md file on github. The changed lines were placed below, and were intended to be highlighted in red (rendering md in atom offline looked fine), but were not. Therefore, the proposed changed lines now begin with a ">" to show distinction. --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 27 +++++++------------ 1 file changed, 9 insertions(+), 18 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index 0f74d2175c3..03cde3d1ee0 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -173,7 +173,8 @@ object, the “g-function calculation model name”, with values of: - UBHWTcalc for the new calculation method that utilizes uniform borehole wall temperature boundary conditions. -Sample IDF and IDD follow with changes marked in red: +Sample IDF and IDD follow, new lines are marked with ">" at the beginning of +the line: ``` !!! IDF example using GHE:Vertical:Array input !!! @@ -188,10 +189,10 @@ GroundHeatExchanger:System, 2.5, !- Ground Thermal Conductivity {W/m-K} 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} , !- GHE:Vertical:ResponseFactors Object Name - UHFcalc, !- g-function Calculation Model Name + >UHFcalc, !- g-function Calculation Model Name GHE-Array; !- GHE:Vertical:Array Object Name ``` -UHFcalc, !- g-function Calculation Model Name + ``` !!! IDF example using GHE:Vertical:Single input !!! @@ -206,12 +207,12 @@ GroundHeatExchanger:System, 2.5, !- Ground Thermal Conductivity {W/m-K} 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} , !- GHE:Vertical:ResponseFactors Object Name - UHFcalc, !- g-function Calculation Model Name + >UHFcalc, !- g-function Calculation Model Name , !- GHE:Vertical:Array Object Name BH1, !- GHE:Vertical:Single Object Name 1 BH2; !- GHE:Vertical:Single Object Name 2 ``` -UHFcalc, !- g-function Calculation Model Name + ``` !!! IDD Modifications !!! @@ -219,8 +220,8 @@ GroundHeatExchanger:System, GroundHeatExchanger:System, \memo Models vertical ground heat exchangers systems using the response factor approach \memo developed by Eskilson. Response factors are calculated using a finite line source - \memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, - \memo or uniform borehole wall temperature if UBHWTcalc is specified. + >\memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, + >\memo or uniform borehole wall temperature if UBHWTcalc is specified. \extensible:1 \min-fields 9 A1, \field Name @@ -230,7 +231,7 @@ GroundHeatExchanger:System, A6, \field GHE:Vertical:ResponseFactors Object Name \type object-list \object-list GroundHeatExchangerVerticalResponseFactorNames - A7, \field g-function Calculation Model Name + >A7, \field g-function Calculation Model Name \type choice \key UHFcalc \key UBHWTcalc @@ -239,16 +240,6 @@ GroundHeatExchanger:System, \object-list GroundHeatExchangerVerticalArrayNames ... ``` -\memo model assuming uniform heat flux at the borehole -wall if UHFcalc is specified, - -\memo or uniform borehole wall temperature if UBHWTcalc -is specified. - -A7, \field g-function Calculation Model Name - \type choice - \key UHFcalc - \key UBHWTcalc ## Engineering Reference Documentation From 49df1a03669964eae50878a0fe1390428e275ab0 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Tue, 8 Jun 2021 23:20:37 -0500 Subject: [PATCH 03/77] Append design documentation to NFP related to #6651 --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 103 ++++++++++++++++++ 1 file changed, 103 insertions(+) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index 03cde3d1ee0..a702c88ced9 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -6,6 +6,7 @@ New Feature Proposal Draft **Jeffrey D. Spitler and Jack C. Cook, OSU** - Original Date 2021-04-09 +- Design added 2021-06-08 ## Justification for New Feature In order for EnergyPlus to simulate a ground heat exchanger (GHE) used with a @@ -279,3 +280,105 @@ Spitler, J. D., J. C. Cook and X. Liu (2020). A Preliminary Investigation on the Cost Reduction Potential of Optimizing Bore Fields for Commercial Ground Source Heat Pump Systems. Proceedings, 45th Workshop on Geothermal Reservoir Engineering. Stanford, California, Stanford University. + +## Design + + + +### Current design + + + +EnergyPlus currently generates vertical borehole g-functions internally using a simple, non-optimized line-source model which uses the uniform heat flux boundary condition. For the vertical borehole and slinky ground heat exchanger model classes, ```GLHEVert``` and ```GLHESlinky```, respectively, the access point into these methods is the member function, ```calcGFunctions```. Inside of the + +vertical borehole GHE ```calcGFunctions``` function, three different functions are called to generate the g-function values. + + + +- ```calcShortTimestepGFunctions``` - which uses a 1D, radial finite volume borehole model to compute the g-functions from a time of 500 seconds (~ ln(t/ts) = -13) to when ln(t/ts) = -9.0, which is likely on the order of 1-2 days. + +- ```calcLongTimestepGFunctions``` - which uses the non-optimized line-source model mentioned above to compute the g-functions from ln(t/ts) = -8.5 up to the maximum simulation time as defined in the EnergyPlus input. + +- ```combineShortAndLongTimestepGFunctions``` - which combines the g-functions generated from the short and long time-step routines into the vector objects the rest of the EnergyPlus GHE code requires for computations. + + + +### Updated design + + + +The updated design in ```GLHEVert::calcGFunctions``` will wrap the calls to these existing functions in an IF block, so that when the new input option "UHFcalc" is selected the existing g-function generation code is called. However, when the "UBWTcalc" option is selected, cpgfunction will be called to generate the g-function values with the uniform borehole wall temperature boundary condition. + + + +```C++ + +if (UHF) { + + calcShortTimestepGFunctions(state); + + calcLongTimestepGFunctions(state); + + combineShortAndLongTimestepGFunctions(); + +} else if (UBHWT) { + + gt::gfunction::uniform_borehole_wall_temperature(...); + +} + +``` + + + +### Optional Build + + + +This can be an optional feature built based on new or existing CMake flags. + + + +### Dependencies + +Currently the tool is relying on nlohmann json, boost, BLAS and LAPACK. The following sections aim to communicate the reasoning behind the use of each dependency and give the current outlook on future dependency. There are also requests for input based on the information provided. These outlooks are not yet set in stone, though a decisive "code freeze" and a path forward for the August deadline needs to occur in the near future. + + + +#### Nlohmann json + +The [single include](https://github.com/nlohmann/json#integration), [json.hpp](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp), is made use of. The file sits inside of a `third_party/` folder. The location of the file and the linking of it in cpgfunction is still up for discussion. Beyond that, the input structure in the input JSON files, containing borefield coordinates, is no more complicated than this: `{"x": [1, 2, 3, 4], "y": [1, 2, 3, 4]}`. This is basically just csv data. In the foreseeable future, there will be nothing more input that these coordinates. This dependency could be eliminated. + + + +#### Boost functions + +The boost functions made use of are [Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). The quadrature integration is performed to compute the integration of the finite line source. The thread pool is used in many locations, though is being removed in many due to `cpgfunction` becoming based more on linear algebra routines via `BLAS`, rather than simply multi-threading. After a complete transition to linear algebra in the applicable locations, the goal is to use `OpenMP` to try to match the multi-threading speed occurring in the calculation of finite line source segment responses. Following the removal of boost's thread pool, a Gauss-Kronrod Quadrature integration could be developed. We would like input as to whether or not this development, the removal of the boost dependency, should be made a priority to be accomplished by August. + + + +#### BLAS and LAPACK routines + +BLAS and LAPACK are Fortran libraries. In the current version, these are expected to be installed as system libraries. Last summer, during the initial development of cpgfunction, it was found that [OpenBLAS](https://github.com/xianyi/OpenBLAS) performed LU factorization for solving systems of linear equations ([gesv](https://icl.bitbucket.io/lapackpp/group__gesv.html#ga8f2d5c8af99b6aee06f0650d723f8e2f)) 5x faster than [netlib](http://www.netlib.org/blas/)'s. The speed of OpenBLAS was nearly identical to the solution solved in Python's [numpy.linalg.solve](https://numpy.org/doc/stable/reference/generated/numpy.linalg.solve.html), which makes use of the same LAPACK routine. More recently, it has been found that linear algebra with BLAS is both significantly faster than multi-threading, and that BLAS has low level functionality that helps with the goals of cpgfunction. (cpgfunction was developed due to the need to compute g-functions faster and with lower memory on high performance computers for the development of g-function databases (Cook and Spitler 2021).) BLAS provides for "fast slicing" of vectors or matrices represented in 1D format. The functions take in pointers to a starting location, and a number of elements to be operated on. Additionally, the BLAS routine function [symv](https://icl.bitbucket.io/blaspp/group__symv.html#ga496ee8fe24db5f3dd003b09cc2bec5a4) gives tremendous speed improvement for performing linear algebra on the packed symmetric matrix discussed in Cook and Spitler (2021). The packed symmetric response matrix can remain packed at each point in time, where inputs to the function instruct whether or not it is the upper or lower triangle. The consumption of memory is also minimized by storing the results in the input vector `y`. This allocation of memory can occur once, and that vector can be re-used across functions. This results in significant speed improvements when compared to multi-threading, with minimal to no additional memory consumption. However, while many of these things are good, and a 2x speed improvement has already been made by initial transfer of computation from boost's thread pool to BLAS, the dependency is programmed in Fortran. Therefore, if the dependency is to continue to be made use of for the August release, there will likely need to be optional build flags for this third party tool. We are currently considering making use of the CMake find package functionality to locate BLAS and LAPACK libraries already installed on the system. Mac (brew install) and Linux (apt-get) both have seamless access to OpenBLAS, and we are investigating ease of access on Windows (choco install). + + + + + +Beyond August 2021, we would like input from the development team on how to handle these dependencies. If anyone on the development team has experience in solving a problem similar to this, feedback and experience would be appreciated. Specifically, should they continue to be expected to be installed as system libraries, incorporated into the E+ repo, or move away from Fortran dependencies entirely for a pure C++ project. + + + +### Git + + + +The code is currently maintained in a private repository on [Jack Cook](https://github.com/j-c-cook)'s account. The plan is to make the code public and open source following further documentation and development. The library can be incorporated into the `third_party` folder as a git subtree or however the development team prefers. + + + +-------------------- + +Cook, J. C. and J. D. Spitler (2021). Faster computation of g-functions used for modeling of ground heat exchangers with reduced memory consumption. Accepted for publication in Proceedings of Building Simulation 2021. September 1-3, 2021. Bruges, Belgium. + + From 30160c1bee1cacbaf977e3e8aa7efe1ed93933e6 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Wed, 9 Jun 2021 10:24:38 -0500 Subject: [PATCH 04/77] Document IDD snippets with the diff language marker As noted by @jmarrec, there is a way to highlight lines that conain - something old + something new https://github.com/NREL/EnergyPlus/pull/8708#issuecomment-823278771 --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 26 +++++++++---------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index a702c88ced9..db36bd4ded8 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -174,11 +174,11 @@ object, the “g-function calculation model name”, with values of: - UBHWTcalc for the new calculation method that utilizes uniform borehole wall temperature boundary conditions. -Sample IDF and IDD follow, new lines are marked with ">" at the beginning of -the line: +Sample IDF and IDD follow, new lines are marked with "+" at the beginning of +the line (this makes use of the `diff` language marker): ``` -!!! IDF example using GHE:Vertical:Array input !!! +! IDF example using GHE:Vertical:Array input GroundHeatExchanger:System, Vertical GHE 1x4 Std, !- Name @@ -190,13 +190,13 @@ GroundHeatExchanger:System, 2.5, !- Ground Thermal Conductivity {W/m-K} 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} , !- GHE:Vertical:ResponseFactors Object Name - >UHFcalc, !- g-function Calculation Model Name ++ UHFcalc, !- g-function Calculation Model Name GHE-Array; !- GHE:Vertical:Array Object Name ``` ``` -!!! IDF example using GHE:Vertical:Single input !!! +! IDF example using GHE:Vertical:Single input GroundHeatExchanger:System, Vertical GHE 1x4 Std, !- Name @@ -208,7 +208,7 @@ GroundHeatExchanger:System, 2.5, !- Ground Thermal Conductivity {W/m-K} 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} , !- GHE:Vertical:ResponseFactors Object Name - >UHFcalc, !- g-function Calculation Model Name + + UHFcalc, !- g-function Calculation Model Name , !- GHE:Vertical:Array Object Name BH1, !- GHE:Vertical:Single Object Name 1 BH2; !- GHE:Vertical:Single Object Name 2 @@ -216,13 +216,13 @@ GroundHeatExchanger:System, ``` -!!! IDD Modifications !!! +! IDD Modifications GroundHeatExchanger:System, \memo Models vertical ground heat exchangers systems using the response factor approach \memo developed by Eskilson. Response factors are calculated using a finite line source - >\memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, - >\memo or uniform borehole wall temperature if UBHWTcalc is specified. + + \memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, + + \memo or uniform borehole wall temperature if UBHWTcalc is specified. \extensible:1 \min-fields 9 A1, \field Name @@ -232,10 +232,10 @@ GroundHeatExchanger:System, A6, \field GHE:Vertical:ResponseFactors Object Name \type object-list \object-list GroundHeatExchangerVerticalResponseFactorNames - >A7, \field g-function Calculation Model Name - \type choice - \key UHFcalc - \key UBHWTcalc + + A7, \field g-function Calculation Model Name + + \type choice + + \key UHFcalc + + \key UBHWTcalc A8, \field GHE:Vertical:Array Object Name \type object-list \object-list GroundHeatExchangerVerticalArrayNames From 72b63cdcdd10b334f751e2110dcd9a534b984f11 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Wed, 9 Jun 2021 10:30:56 -0500 Subject: [PATCH 05/77] Add diff as markdown language for IDD snippets --- design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index db36bd4ded8..c02923faf54 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -177,7 +177,7 @@ temperature boundary conditions. Sample IDF and IDD follow, new lines are marked with "+" at the beginning of the line (this makes use of the `diff` language marker): -``` +```diff ! IDF example using GHE:Vertical:Array input GroundHeatExchanger:System, @@ -195,7 +195,7 @@ GroundHeatExchanger:System, ``` -``` +```diff ! IDF example using GHE:Vertical:Single input GroundHeatExchanger:System, @@ -215,7 +215,7 @@ GroundHeatExchanger:System, ``` -``` +```diff ! IDD Modifications GroundHeatExchanger:System, From 69b797d3acf5c9ff608cc11f1e5cd447662a2ef4 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Wed, 9 Jun 2021 10:33:17 -0500 Subject: [PATCH 06/77] Refactor diff Documentation placements For the diff highlighting to work on github, the !'s, +'s and -'s must be at the beginning of a new line. --- design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index c02923faf54..168be616dea 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -208,7 +208,7 @@ GroundHeatExchanger:System, 2.5, !- Ground Thermal Conductivity {W/m-K} 2.5E+06, !- Ground Thermal Heat Capacity {J/m3-K} , !- GHE:Vertical:ResponseFactors Object Name - + UHFcalc, !- g-function Calculation Model Name ++ UHFcalc, !- g-function Calculation Model Name , !- GHE:Vertical:Array Object Name BH1, !- GHE:Vertical:Single Object Name 1 BH2; !- GHE:Vertical:Single Object Name 2 @@ -221,8 +221,8 @@ GroundHeatExchanger:System, GroundHeatExchanger:System, \memo Models vertical ground heat exchangers systems using the response factor approach \memo developed by Eskilson. Response factors are calculated using a finite line source - + \memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, - + \memo or uniform borehole wall temperature if UBHWTcalc is specified. ++ \memo model assuming uniform heat flux at the borehole wall if UHFcalc is specified, ++ \memo or uniform borehole wall temperature if UBHWTcalc is specified. \extensible:1 \min-fields 9 A1, \field Name @@ -232,10 +232,10 @@ GroundHeatExchanger:System, A6, \field GHE:Vertical:ResponseFactors Object Name \type object-list \object-list GroundHeatExchangerVerticalResponseFactorNames - + A7, \field g-function Calculation Model Name - + \type choice - + \key UHFcalc - + \key UBHWTcalc ++ A7, \field g-function Calculation Model Name ++ \type choice ++ \key UHFcalc ++ \key UBHWTcalc A8, \field GHE:Vertical:Array Object Name \type object-list \object-list GroundHeatExchangerVerticalArrayNames From 9b082d9b4c28bfb8b5d10c53c1cc3f4afb44b1d5 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Tue, 29 Jun 2021 11:50:09 -0500 Subject: [PATCH 07/77] Document the design document in regards to dependencies relates to #6651 --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 177 +++++++++++++----- 1 file changed, 126 insertions(+), 51 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index 168be616dea..3234d75dba0 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -283,33 +283,37 @@ Engineering. Stanford, California, Stanford University. ## Design - - ### Current design +EnergyPlus currently generates vertical borehole g-functions internally using a +simple, non-optimized line-source model which uses the uniform heat flux +boundary condition. For the vertical borehole and slinky ground heat exchanger +model classes, ```GLHEVert``` and ```GLHESlinky```, respectively, the access +point into these methods is the member function, ```calcGFunctions```. +Inside of the vertical borehole GHE ```calcGFunctions``` function, three +different functions are called to generate the g-function values. +- ```calcShortTimestepGFunctions``` - which uses a 1D, radial finite volume + borehole model to compute the g-functions from a time of 500 seconds + (~ ln(t/ts) = -13) to when ln(t/ts) = -9.0, which is likely on the order of + 1-2 days. -EnergyPlus currently generates vertical borehole g-functions internally using a simple, non-optimized line-source model which uses the uniform heat flux boundary condition. For the vertical borehole and slinky ground heat exchanger model classes, ```GLHEVert``` and ```GLHESlinky```, respectively, the access point into these methods is the member function, ```calcGFunctions```. Inside of the - -vertical borehole GHE ```calcGFunctions``` function, three different functions are called to generate the g-function values. - - - -- ```calcShortTimestepGFunctions``` - which uses a 1D, radial finite volume borehole model to compute the g-functions from a time of 500 seconds (~ ln(t/ts) = -13) to when ln(t/ts) = -9.0, which is likely on the order of 1-2 days. - -- ```calcLongTimestepGFunctions``` - which uses the non-optimized line-source model mentioned above to compute the g-functions from ln(t/ts) = -8.5 up to the maximum simulation time as defined in the EnergyPlus input. - -- ```combineShortAndLongTimestepGFunctions``` - which combines the g-functions generated from the short and long time-step routines into the vector objects the rest of the EnergyPlus GHE code requires for computations. - +- ```calcLongTimestepGFunctions``` - which uses the non-optimized line-source + model mentioned above to compute the g-functions from ln(t/ts) = -8.5 up to + the maximum simulation time as defined in the EnergyPlus input. +- ```combineShortAndLongTimestepGFunctions``` - which combines the g-functions + generated from the short and long time-step routines into the vector objects + the rest of the EnergyPlus GHE code requires for computations. ### Updated design - - -The updated design in ```GLHEVert::calcGFunctions``` will wrap the calls to these existing functions in an IF block, so that when the new input option "UHFcalc" is selected the existing g-function generation code is called. However, when the "UBWTcalc" option is selected, cpgfunction will be called to generate the g-function values with the uniform borehole wall temperature boundary condition. - - +The updated design in ```GLHEVert::calcGFunctions``` will wrap the calls to +these existing functions in an IF block, so that when the new input option +"UHFcalc" is selected the existing g-function generation code is called. +However, when the "UBWTcalc" option is selected, cpgfunction will be called to +generate the g-function values with the uniform borehole wall temperature +boundary condition. ```C++ @@ -329,56 +333,127 @@ if (UHF) { ``` - - ### Optional Build - - This can be an optional feature built based on new or existing CMake flags. - - ### Dependencies -Currently the tool is relying on nlohmann json, boost, BLAS and LAPACK. The following sections aim to communicate the reasoning behind the use of each dependency and give the current outlook on future dependency. There are also requests for input based on the information provided. These outlooks are not yet set in stone, though a decisive "code freeze" and a path forward for the August deadline needs to occur in the near future. - - +Currently the tool is relying on nlohmann json, boost, BLAS and LAPACK. The +following sections aim to communicate the reasoning behind the use of each +dependency and give the current outlook on future dependency. There are also +requests for input based on the information provided. These outlooks are not +yet set in stone, though a decisive "code freeze" and a path forward for the +August deadline needs to occur in the near future. + +Cpgfunction was originally written making use of Boost and BLAS/LAPACK. The +BLAS/LAPACK library is based on Fortran code. While it is quite easy to make an +executable in Linux and Mac environments that combines C++ code with the +BLAS/LAPACK library, there are some difficulties with doing this for Windows. +Discussions with the EnergyPlus development team have led us to minimize the +use of Boost and replace the BLAS/LAPACK with native C++ code. Eigen may be +used; this has not been decided yet. + +The only libraries made use of from boost are asio for the thread pool and +math for the gaussian quadrature integration. We have therefore reduced the use +of Boost substantially. The full Boost library contains 14,787 source code +files; this has been reduced to 704. The full Boost library is 139.9 MiB; the +reduced set of files uses 8.6 MiB. + +Three BLAS functions were used (copy, axpy, spmv). As of June 28th, copy and +axpy have been replaced with native C++ code; conversion of spmv is in +progress. We plan to replace gesv from LAPACK with either native C++ code or +the Eigen library. #### Nlohmann json -The [single include](https://github.com/nlohmann/json#integration), [json.hpp](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp), is made use of. The file sits inside of a `third_party/` folder. The location of the file and the linking of it in cpgfunction is still up for discussion. Beyond that, the input structure in the input JSON files, containing borefield coordinates, is no more complicated than this: `{"x": [1, 2, 3, 4], "y": [1, 2, 3, 4]}`. This is basically just csv data. In the foreseeable future, there will be nothing more input that these coordinates. This dependency could be eliminated. - - +The [single include](https://github.com/nlohmann/json#integration), +[json.hpp](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp), +is made use of. The file sits inside of a `third_party/` folder. The location +of the file and the linking of it in cpgfunction is still up for discussion. +Beyond that, the input structure in the input JSON files, containing borefield +coordinates, is no more complicated than this: +`{"x": [1, 2, 3, 4], "y": [1, 2, 3, 4]}`. This is basically just csv data. +In the foreseeable future, there will be nothing more input that these +coordinates. This dependency could be eliminated. Although, this is a +relatively small dependency, should effort be put into eliminating this +dependency by August? #### Boost functions -The boost functions made use of are [Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). The quadrature integration is performed to compute the integration of the finite line source. The thread pool is used in many locations, though is being removed in many due to `cpgfunction` becoming based more on linear algebra routines via `BLAS`, rather than simply multi-threading. After a complete transition to linear algebra in the applicable locations, the goal is to use `OpenMP` to try to match the multi-threading speed occurring in the calculation of finite line source segment responses. Following the removal of boost's thread pool, a Gauss-Kronrod Quadrature integration could be developed. We would like input as to whether or not this development, the removal of the boost dependency, should be made a priority to be accomplished by August. - - +The boost functions made use of are[Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). +The quadrature integration is performed to compute the integration of the +finite line source. The thread pool was being removed and replaced by `BLAS` +linear algebra routines. Although, it was found that 1) `EnergyPlus` has a goal +of removing all Fortran from the library, and 2) building `BLAS` on Windows is +not a straightforward task. It is possible that `OpenMP` could be made use of +as opposed to `Boost`s thread pool for multi-threading, though initial speed +comparisons significantly favored `Boost`. A standalone [Gauss-Kronrod](https://people.sc.fsu.edu/~jburkardt/cpp_src/kronrod/kronrod.html) +integration could be made use of, but the speed and accuracy are unknown. We +would like input as to whether or not this development, the removal of the +boost dependency, should be made a priority to be accomplished by August. #### BLAS and LAPACK routines -BLAS and LAPACK are Fortran libraries. In the current version, these are expected to be installed as system libraries. Last summer, during the initial development of cpgfunction, it was found that [OpenBLAS](https://github.com/xianyi/OpenBLAS) performed LU factorization for solving systems of linear equations ([gesv](https://icl.bitbucket.io/lapackpp/group__gesv.html#ga8f2d5c8af99b6aee06f0650d723f8e2f)) 5x faster than [netlib](http://www.netlib.org/blas/)'s. The speed of OpenBLAS was nearly identical to the solution solved in Python's [numpy.linalg.solve](https://numpy.org/doc/stable/reference/generated/numpy.linalg.solve.html), which makes use of the same LAPACK routine. More recently, it has been found that linear algebra with BLAS is both significantly faster than multi-threading, and that BLAS has low level functionality that helps with the goals of cpgfunction. (cpgfunction was developed due to the need to compute g-functions faster and with lower memory on high performance computers for the development of g-function databases (Cook and Spitler 2021).) BLAS provides for "fast slicing" of vectors or matrices represented in 1D format. The functions take in pointers to a starting location, and a number of elements to be operated on. Additionally, the BLAS routine function [symv](https://icl.bitbucket.io/blaspp/group__symv.html#ga496ee8fe24db5f3dd003b09cc2bec5a4) gives tremendous speed improvement for performing linear algebra on the packed symmetric matrix discussed in Cook and Spitler (2021). The packed symmetric response matrix can remain packed at each point in time, where inputs to the function instruct whether or not it is the upper or lower triangle. The consumption of memory is also minimized by storing the results in the input vector `y`. This allocation of memory can occur once, and that vector can be re-used across functions. This results in significant speed improvements when compared to multi-threading, with minimal to no additional memory consumption. However, while many of these things are good, and a 2x speed improvement has already been made by initial transfer of computation from boost's thread pool to BLAS, the dependency is programmed in Fortran. Therefore, if the dependency is to continue to be made use of for the August release, there will likely need to be optional build flags for this third party tool. We are currently considering making use of the CMake find package functionality to locate BLAS and LAPACK libraries already installed on the system. Mac (brew install) and Linux (apt-get) both have seamless access to OpenBLAS, and we are investigating ease of access on Windows (choco install). - - - - - -Beyond August 2021, we would like input from the development team on how to handle these dependencies. If anyone on the development team has experience in solving a problem similar to this, feedback and experience would be appreciated. Specifically, should they continue to be expected to be installed as system libraries, incorporated into the E+ repo, or move away from Fortran dependencies entirely for a pure C++ project. - - +BLAS and LAPACK are Fortran libraries. In the current version, these are +expected to be installed as system libraries. Last summer, during the initial +development of cpgfunction, it was found that [OpenBLAS](https://github.com/xianyi/OpenBLAS) +performed LU factorization for solving systems of linear equations ([gesv](https://icl.bitbucket.io/lapackpp/group__gesv.html#ga8f2d5c8af99b6aee06f0650d723f8e2f)) +5x faster than [netlib](http://www.netlib.org/blas/). The speed of OpenBLAS was +nearly identical to the solution solved in Python's [numpy.linalg.solve](https://numpy.org/doc/stable/reference/generated/numpy.linalg.solve.html), +which makes use of the same LAPACK routine. More recently, it has been found +that linear algebra with BLAS is both significantly faster than +multi-threading, and that BLAS has low level functionality that helps with the +goals of cpgfunction. (cpgfunction was developed due to the need to compute +g-functions faster and with lower memory on high performance computers for the +development of g-function databases (Cook and Spitler 2021).) BLAS provides for +"fast slicing" of vectors or matrices represented in 1D format. The functions +take in pointers to a starting location, and a number of elements to be +operated on. Additionally, the BLAS routine function [symv](https://icl.bitbucket.io/blaspp/group__symv.html#ga496ee8fe24db5f3dd003b09cc2bec5a4) +gives tremendous speed improvement for performing linear algebra on the packed +symmetric matrix discussed in Cook and Spitler (2021). The packed symmetric +response matrix can remain packed at each point in time, where inputs to the +function instruct whether or not it is the upper or lower triangle. The +consumption of memory is also minimized by storing the results in the input +vector `y`. This allocation of memory can occur once, and that vector can be +re-used across functions. This results in significant speed improvements when +compared to multi-threading, with minimal to no additional memory consumption. +However, while many of these things are good, and a 2x speed improvement has +already been made by initial transfer of computation from boost's thread pool +to BLAS, the dependency is programmed in Fortran. + +However, there are more than one sticking points with the use of BLAS/LAPACK +being a dependency. A fork has been made in the project so that `EnergyPlus` +can be delivered what is requested. The name of the fork is `cpgfunctionEP`. +Therefore, BLAS/LAPACK will not be included in the dependency for +`cpgfunctionEP`. The speed of `cpgfunctionEP` is slower than `cpgfunction`, but +for reasons previously stated, it is necessary to remove all Fortran code +from the library for delivery to EnergyPlus. + +#### Eigen + +The `Eigen` library will likely become a dependency to be used in solving a +system of equations using LU decomposition. As previously mentioned, +`cpgfunction` is faster than `cpgfunctionEP` due to more linear algebra being +made use of. It may be possible for `cpgfunctionEP` to someday match the speed +performance of `cpgfunction` by depending more heavily on `Eigen`, though the +library from its conception has been geared towards BLAS/LAPACK. The re-write +would be substantial, and the amount of memory consumed will grow. + +A re-write with a dependency on `Eigen` would more than likely provide a great +increase in speed performance of `cpgfunctionEP`. This re-write would +take time, and is not feasible to be accomplished by August. ### Git - - -The code is currently maintained in a private repository on [Jack Cook](https://github.com/j-c-cook)'s account. The plan is to make the code public and open source following further documentation and development. The library can be incorporated into the `third_party` folder as a git subtree or however the development team prefers. - - +The code is currently maintained in a private repository on the account of [Jack Cook](https://github.com/j-c-cook). +The plan is to make the code public and open source following further +documentation and development. The library can be incorporated into the +`third_party` folder as a git subtree or however the development team prefers. -------------------- -Cook, J. C. and J. D. Spitler (2021). Faster computation of g-functions used for modeling of ground heat exchangers with reduced memory consumption. Accepted for publication in Proceedings of Building Simulation 2021. September 1-3, 2021. Bruges, Belgium. - - +Cook, J. C. and J. D. Spitler (2021). Faster computation of g-functions used +for modeling of ground heat exchangers with reduced memory consumption. +Accepted for publication in Proceedings of Building Simulation 2021. +September 1-3, 2021. Bruges, Belgium. From a416e1a8afd0d4766d47cfe84299aaf16a1d973d Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 2 Jul 2021 15:10:35 -0500 Subject: [PATCH 08/77] Modify design document to reflect outlook On the EnergyPlus call Wednesday, June 30, 2021, a path was laid out for the future of cpgfunction. The project has been forked. There now exists a fork by the name of cpgfunctionEP. cpgfuncitonEP (a fork of cpgfunction) will contain neither Fortran nor Boost dependencies. --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 44 +++++++++---------- 1 file changed, 21 insertions(+), 23 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index 3234d75dba0..928dcb9eb96 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -330,7 +330,6 @@ if (UHF) { gt::gfunction::uniform_borehole_wall_temperature(...); } - ``` ### Optional Build @@ -339,34 +338,35 @@ This can be an optional feature built based on new or existing CMake flags. ### Dependencies -Currently the tool is relying on nlohmann json, boost, BLAS and LAPACK. The +Currently, the tool is relying on nlohmann json and boost. The following sections aim to communicate the reasoning behind the use of each -dependency and give the current outlook on future dependency. There are also -requests for input based on the information provided. These outlooks are not -yet set in stone, though a decisive "code freeze" and a path forward for the -August deadline needs to occur in the near future. +dependency and give the current outlook on future dependency. Cpgfunction was originally written making use of Boost and BLAS/LAPACK. The BLAS/LAPACK library is based on Fortran code. While it is quite easy to make an executable in Linux and Mac environments that combines C++ code with the -BLAS/LAPACK library, there are some difficulties with doing this for Windows. +BLAS/LAPACK library, there are some difficulties in doing this on Windows. Discussions with the EnergyPlus development team have led us to minimize the -use of Boost and replace the BLAS/LAPACK with native C++ code. Eigen may be -used; this has not been decided yet. +use of Boost and replace the BLAS/LAPACK with native C++ code. A small portion +of Eigen has been introduced to perform an LU decomposition of a system of +equations for the g-function. -The only libraries made use of from boost are asio for the thread pool and -math for the gaussian quadrature integration. We have therefore reduced the use -of Boost substantially. The full Boost library contains 14,787 source code -files; this has been reduced to 704. The full Boost library is 139.9 MiB; the -reduced set of files uses 8.6 MiB. +The only libraries currently made use of from boost are `asio` for the thread +pool and `math` for the gaussian quadrature integration. The `asio` portion of +boost is being replaced by `OpenMP`. The Gauss-Kronrod quadrature integration +of the finite line source is being replaced with the QUADPACK methodology +written in C/C++. -Three BLAS functions were used (copy, axpy, spmv). As of June 28th, copy and -axpy have been replaced with native C++ code; conversion of spmv is in -progress. We plan to replace gesv from LAPACK with either native C++ code or -the Eigen library. +Three BLAS functions were used (copy, axpy, spmv). These functions have been +replaced with native C++ code. The gesv function from LAPACK has been of loaded +to the Eigen library. #### Nlohmann json +This dependency is not necessary for EnergyPlus. This dependency exists for +running unit tests in the `cpgfunctionEP` repository. The unit tests will not +be built and ran, and neither will this dependency. + The [single include](https://github.com/nlohmann/json#integration), [json.hpp](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp), is made use of. The file sits inside of a `third_party/` folder. The location @@ -375,13 +375,11 @@ Beyond that, the input structure in the input JSON files, containing borefield coordinates, is no more complicated than this: `{"x": [1, 2, 3, 4], "y": [1, 2, 3, 4]}`. This is basically just csv data. In the foreseeable future, there will be nothing more input that these -coordinates. This dependency could be eliminated. Although, this is a -relatively small dependency, should effort be put into eliminating this -dependency by August? +coordinates. #### Boost functions -The boost functions made use of are[Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). +The boost functions made use of are [Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). The quadrature integration is performed to compute the integration of the finite line source. The thread pool was being removed and replaced by `BLAS` linear algebra routines. Although, it was found that 1) `EnergyPlus` has a goal @@ -393,7 +391,7 @@ integration could be made use of, but the speed and accuracy are unknown. We would like input as to whether or not this development, the removal of the boost dependency, should be made a priority to be accomplished by August. -#### BLAS and LAPACK routines +#### BLAS and LAPACK routines (deprecated) BLAS and LAPACK are Fortran libraries. In the current version, these are expected to be installed as system libraries. Last summer, during the initial From 40af248bfa5edc2abcb310454cadb02ec05d8971 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 2 Jul 2021 15:44:57 -0500 Subject: [PATCH 09/77] Update design document to reflect current outlook --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 20 +++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index 928dcb9eb96..f2739040b97 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -338,7 +338,12 @@ This can be an optional feature built based on new or existing CMake flags. ### Dependencies -Currently, the tool is relying on nlohmann json and boost. The +The current outlook for the future of this library and its interaction with +EnergyPlus is to only be dependent on Eigen. The project is dependent on +nlohmann json for running unit tests, but the local tests will not be built +by EnergyPlus. + +Currently, the tool is relying on nlohmann json, Eigen and Boost. The following sections aim to communicate the reasoning behind the use of each dependency and give the current outlook on future dependency. @@ -346,6 +351,8 @@ Cpgfunction was originally written making use of Boost and BLAS/LAPACK. The BLAS/LAPACK library is based on Fortran code. While it is quite easy to make an executable in Linux and Mac environments that combines C++ code with the BLAS/LAPACK library, there are some difficulties in doing this on Windows. +Additionally, EnergyPlus has a goal of ridding the repository of Fortran +dependencies. Discussions with the EnergyPlus development team have led us to minimize the use of Boost and replace the BLAS/LAPACK with native C++ code. A small portion of Eigen has been introduced to perform an LU decomposition of a system of @@ -430,17 +437,18 @@ from the library for delivery to EnergyPlus. #### Eigen -The `Eigen` library will likely become a dependency to be used in solving a -system of equations using LU decomposition. As previously mentioned, +The `Eigen` library has been made a dependency to be used in solving a system +of equations using LU decomposition. As previously mentioned, `cpgfunction` is faster than `cpgfunctionEP` due to more linear algebra being made use of. It may be possible for `cpgfunctionEP` to someday match the speed performance of `cpgfunction` by depending more heavily on `Eigen`, though the library from its conception has been geared towards BLAS/LAPACK. The re-write would be substantial, and the amount of memory consumed will grow. -A re-write with a dependency on `Eigen` would more than likely provide a great -increase in speed performance of `cpgfunctionEP`. This re-write would -take time, and is not feasible to be accomplished by August. +It is unclear if a greater dependency on `Eigen` could provide performance +similar to the BLAS/LAPACK of OpenBLAS. A re-write with a focus on `Eigen` for +linear algebra could be done, but would take time, and is not feasible to be +accomplished by August. ### Git From 8c9f71ae846977890a862ddfbb4e1c8154a6a0d4 Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 2 Jul 2021 16:35:21 -0500 Subject: [PATCH 10/77] Refactor design document to only include "bottom line up front" --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 167 ++++++++---------- 1 file changed, 70 insertions(+), 97 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index f2739040b97..e34dcd01af0 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -338,117 +338,90 @@ This can be an optional feature built based on new or existing CMake flags. ### Dependencies -The current outlook for the future of this library and its interaction with -EnergyPlus is to only be dependent on Eigen. The project is dependent on -nlohmann json for running unit tests, but the local tests will not be built -by EnergyPlus. - -Currently, the tool is relying on nlohmann json, Eigen and Boost. The -following sections aim to communicate the reasoning behind the use of each -dependency and give the current outlook on future dependency. - Cpgfunction was originally written making use of Boost and BLAS/LAPACK. The BLAS/LAPACK library is based on Fortran code. While it is quite easy to make an executable in Linux and Mac environments that combines C++ code with the BLAS/LAPACK library, there are some difficulties in doing this on Windows. -Additionally, EnergyPlus has a goal of ridding the repository of Fortran -dependencies. -Discussions with the EnergyPlus development team have led us to minimize the -use of Boost and replace the BLAS/LAPACK with native C++ code. A small portion -of Eigen has been introduced to perform an LU decomposition of a system of -equations for the g-function. - -The only libraries currently made use of from boost are `asio` for the thread -pool and `math` for the gaussian quadrature integration. The `asio` portion of -boost is being replaced by `OpenMP`. The Gauss-Kronrod quadrature integration -of the finite line source is being replaced with the QUADPACK methodology -written in C/C++. +Additionally, the EnergyPlus development team has goals of ridding the +repository of Fortran dependencies and eliminating the use of Boost. -Three BLAS functions were used (copy, axpy, spmv). These functions have been -replaced with native C++ code. The gesv function from LAPACK has been of loaded -to the Eigen library. +Therefore, we have replaced the calls to BLAS/LAPACK with native C++ code and a +call to the Eigen library, and (to date) minimized the use of Boost. Based on +the further discussions with the EnergyPlus development team, we plan to +eliminate the use of Boost also. The following sections describe the +dependencies, past, present, and future, along with some discussion of the +performance implications of the changes already made or to be made in the code +for this feature. #### Nlohmann json -This dependency is not necessary for EnergyPlus. This dependency exists for -running unit tests in the `cpgfunctionEP` repository. The unit tests will not -be built and ran, and neither will this dependency. - -The [single include](https://github.com/nlohmann/json#integration), -[json.hpp](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp), -is made use of. The file sits inside of a `third_party/` folder. The location -of the file and the linking of it in cpgfunction is still up for discussion. -Beyond that, the input structure in the input JSON files, containing borefield -coordinates, is no more complicated than this: -`{"x": [1, 2, 3, 4], "y": [1, 2, 3, 4]}`. This is basically just csv data. -In the foreseeable future, there will be nothing more input that these -coordinates. - -#### Boost functions - -The boost functions made use of are [Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). -The quadrature integration is performed to compute the integration of the -finite line source. The thread pool was being removed and replaced by `BLAS` -linear algebra routines. Although, it was found that 1) `EnergyPlus` has a goal -of removing all Fortran from the library, and 2) building `BLAS` on Windows is -not a straightforward task. It is possible that `OpenMP` could be made use of -as opposed to `Boost`s thread pool for multi-threading, though initial speed -comparisons significantly favored `Boost`. A standalone [Gauss-Kronrod](https://people.sc.fsu.edu/~jburkardt/cpp_src/kronrod/kronrod.html) -integration could be made use of, but the speed and accuracy are unknown. We -would like input as to whether or not this development, the removal of the -boost dependency, should be made a priority to be accomplished by August. +This dependency exists for running unit tests in the cpgfunctionEP repository. +This dependence will not be used in the EnergyPlus build. #### BLAS and LAPACK routines (deprecated) -BLAS and LAPACK are Fortran libraries. In the current version, these are -expected to be installed as system libraries. Last summer, during the initial -development of cpgfunction, it was found that [OpenBLAS](https://github.com/xianyi/OpenBLAS) -performed LU factorization for solving systems of linear equations ([gesv](https://icl.bitbucket.io/lapackpp/group__gesv.html#ga8f2d5c8af99b6aee06f0650d723f8e2f)) -5x faster than [netlib](http://www.netlib.org/blas/). The speed of OpenBLAS was -nearly identical to the solution solved in Python's [numpy.linalg.solve](https://numpy.org/doc/stable/reference/generated/numpy.linalg.solve.html), -which makes use of the same LAPACK routine. More recently, it has been found -that linear algebra with BLAS is both significantly faster than -multi-threading, and that BLAS has low level functionality that helps with the -goals of cpgfunction. (cpgfunction was developed due to the need to compute -g-functions faster and with lower memory on high performance computers for the -development of g-function databases (Cook and Spitler 2021).) BLAS provides for -"fast slicing" of vectors or matrices represented in 1D format. The functions -take in pointers to a starting location, and a number of elements to be -operated on. Additionally, the BLAS routine function [symv](https://icl.bitbucket.io/blaspp/group__symv.html#ga496ee8fe24db5f3dd003b09cc2bec5a4) -gives tremendous speed improvement for performing linear algebra on the packed -symmetric matrix discussed in Cook and Spitler (2021). The packed symmetric -response matrix can remain packed at each point in time, where inputs to the -function instruct whether or not it is the upper or lower triangle. The -consumption of memory is also minimized by storing the results in the input -vector `y`. This allocation of memory can occur once, and that vector can be -re-used across functions. This results in significant speed improvements when -compared to multi-threading, with minimal to no additional memory consumption. -However, while many of these things are good, and a 2x speed improvement has -already been made by initial transfer of computation from boost's thread pool -to BLAS, the dependency is programmed in Fortran. - -However, there are more than one sticking points with the use of BLAS/LAPACK -being a dependency. A fork has been made in the project so that `EnergyPlus` -can be delivered what is requested. The name of the fork is `cpgfunctionEP`. -Therefore, BLAS/LAPACK will not be included in the dependency for -`cpgfunctionEP`. The speed of `cpgfunctionEP` is slower than `cpgfunction`, but -for reasons previously stated, it is necessary to remove all Fortran code -from the library for delivery to EnergyPlus. +Three BLAS functions were used (copy, axpy, spmv). These functions have been +replaced with native C++ code. The gesv function from LAPACK has been off +loaded to the Eigen library. + +In the previous version, these were expected to be installed as system +libraries. Last summer, during the initial development of cpgfunction, it was +found that OpenBLAS performed LU factorization for solving systems of linear +equations (gesv) 5x faster than netlib. The speed of OpenBLAS was nearly +identical to the solution solved in Python's numpy.linalg.solve, which makes +use of the same LAPACK routine. More recently, it has been found that linear +algebra with BLAS is both significantly faster than multi-threading, and that +BLAS has low level functionality that helps with the goals of cpgfunction. +(cpgfunction was developed due to the need to compute g-functions faster and +with lower memory on high performance computers for the development of +g-function databases (Cook and Spitler 2021).) BLAS provides for "fast slicing" +of vectors or matrices represented in 1D format. The functions take in pointers +to a starting location, and a number of elements to be operated on. +Additionally, the BLAS routine function symv gives tremendous speed improvement +for performing linear algebra on the packed symmetric matrix discussed in Cook +and Spitler (2021). The packed symmetric response matrix can remain packed at +each point in time, where inputs to the function instruct whether or not it is +the upper or lower triangle. The consumption of memory is also minimized by +storing the results in the input vector y. This allocation of memory can occur +once, and that vector can be re-used across functions. This results in +significant speed improvements when compared to multi-threading, with minimal +to no additional memory consumption. However, while many of these things are +good, and a 2x speed improvement has already been made by initial transfer of +computation from boost's thread pool to BLAS, the dependency was programmed in +Fortran. + +There is more than one sticking point with the use of BLAS/LAPACK being a +dependency. A fork has been made in the project so that EnergyPlus can be +delivered what is requested. The name of the fork is cpgfunctionEP. Therefore, +BLAS/LAPACK will not be included in the dependency for cpgfunctionEP. The speed +of cpgfunctionEP is slower than cpgfunction, but for reasons previously stated, +it is necessary to remove all Fortran code from the library for delivery to +EnergyPlus. #### Eigen -The `Eigen` library has been made a dependency to be used in solving a system -of equations using LU decomposition. As previously mentioned, -`cpgfunction` is faster than `cpgfunctionEP` due to more linear algebra being -made use of. It may be possible for `cpgfunctionEP` to someday match the speed -performance of `cpgfunction` by depending more heavily on `Eigen`, though the -library from its conception has been geared towards BLAS/LAPACK. The re-write -would be substantial, and the amount of memory consumed will grow. - -It is unclear if a greater dependency on `Eigen` could provide performance -similar to the BLAS/LAPACK of OpenBLAS. A re-write with a focus on `Eigen` for -linear algebra could be done, but would take time, and is not feasible to be -accomplished by August. +The Eigen library has been made a dependency to be used in solving a system of +equations using LU decomposition. As previously mentioned, cpgfunction is faster +than cpgfunctionEP due to more linear algebra being made use of. It may be +possible for cpgfunctionEP to someday match the speed performance of cpgfunction +by depending more heavily on Eigen, though the library from its conception has +been geared towards BLAS/LAPACK. The re-write would be substantial, and the +amount of memory consumed will grow. + +It is unclear if a greater dependency on Eigen could provide performance similar +to the BLAS/LAPACK of OpenBLAS. A re-write with a focus on Eigen for linear +algebra could be done, but would take time, and is not feasible to be +accomplished by August. + +#### Boost + +The boost functions currently made use of are Gauss-Kronrod Quadrature and +thread pool. The quadrature integration is performed to compute the integration +of the finite line source. Replacements for both functions are under +investigation. It appears likely that OpenMP could be made use of as opposed to +Boosts thread pool for multi-threading, though initial speed comparisons +significantly favored Boost. A standalone Gauss-Kronrod integration could be +made use of, but the speed and accuracy are unknown. ### Git From 063c86cdef2e5c150404543c98005058b293266c Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Fri, 2 Jul 2021 16:52:05 -0500 Subject: [PATCH 11/77] Add in links to design document --- .../FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md | 22 ++++++++++++------- 1 file changed, 14 insertions(+), 8 deletions(-) diff --git a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md index e34dcd01af0..9a4eed1ab6a 100644 --- a/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md +++ b/design/FY2021/NFP-Fast_gFunc_Calc_cpgfunction.md @@ -355,7 +355,7 @@ for this feature. #### Nlohmann json -This dependency exists for running unit tests in the cpgfunctionEP repository. +This dependency exists for running unit tests in the `cpgfunctionEP` repository. This dependence will not be used in the EnergyPlus build. #### BLAS and LAPACK routines (deprecated) @@ -367,8 +367,12 @@ loaded to the Eigen library. In the previous version, these were expected to be installed as system libraries. Last summer, during the initial development of cpgfunction, it was found that OpenBLAS performed LU factorization for solving systems of linear -equations (gesv) 5x faster than netlib. The speed of OpenBLAS was nearly -identical to the solution solved in Python's numpy.linalg.solve, which makes +equations ([gesv](https://icl.bitbucket.io/lapackpp/group__gesv.html#ga8f2d5c8af99b6aee06f0650d723f8e2f)) +5x faster than [netlib](https://www.netlib.org/blas/). +The speed of [OpenBLAS](https://github.com/xianyi/OpenBLAS) +was nearly +identical to the solution solved in Python's [numpy.linalg.solve](https://numpy.org/doc/stable/reference/generated/numpy.linalg.solve.html), +which makes use of the same LAPACK routine. More recently, it has been found that linear algebra with BLAS is both significantly faster than multi-threading, and that BLAS has low level functionality that helps with the goals of cpgfunction. @@ -377,7 +381,8 @@ with lower memory on high performance computers for the development of g-function databases (Cook and Spitler 2021).) BLAS provides for "fast slicing" of vectors or matrices represented in 1D format. The functions take in pointers to a starting location, and a number of elements to be operated on. -Additionally, the BLAS routine function symv gives tremendous speed improvement +Additionally, the BLAS routine function [symv](https://icl.bitbucket.io/blaspp/group__symv.html#ga496ee8fe24db5f3dd003b09cc2bec5a4) +gives tremendous speed improvement for performing linear algebra on the packed symmetric matrix discussed in Cook and Spitler (2021). The packed symmetric response matrix can remain packed at each point in time, where inputs to the function instruct whether or not it is @@ -415,13 +420,14 @@ accomplished by August. #### Boost -The boost functions currently made use of are Gauss-Kronrod Quadrature and -thread pool. The quadrature integration is performed to compute the integration +The boost functions currently made use of are [Gauss-Kronrod Quadrature](https://www.boost.org/doc/libs/1_71_0/libs/math/doc/html/math_toolkit/gauss_kronrod.html) +and [thread pool](https://www.boost.org/doc/libs/1_76_0/doc/html/boost_asio/reference/thread_pool.html). +The quadrature integration is performed to compute the integration of the finite line source. Replacements for both functions are under investigation. It appears likely that OpenMP could be made use of as opposed to Boosts thread pool for multi-threading, though initial speed comparisons -significantly favored Boost. A standalone Gauss-Kronrod integration could be -made use of, but the speed and accuracy are unknown. +significantly favored Boost. A standalone [Gauss-Kronrod](https://people.sc.fsu.edu/~jburkardt/cpp_src/kronrod/kronrod.html) +integration could be made use of, but the speed and accuracy are unknown. ### Git From a795a7e11e5240543225776ccaf643c24d84f32f Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Thu, 8 Jul 2021 16:40:44 -0500 Subject: [PATCH 12/77] Add cpgfunctionEP-0.1 to the third_party folder --- .../.github/workflows/linux_build.yml | 47 + .../.github/workflows/mac_build.yml | 38 + .../.github/workflows/windows_build.yml | 59 + third_party/cpgfunctionEP-0.1/.gitignore | 7 + third_party/cpgfunctionEP-0.1/CHANGELOG.md | 104 + third_party/cpgfunctionEP-0.1/CMakeLists.txt | 87 + third_party/cpgfunctionEP-0.1/CONTRIBUTING.md | 21 + third_party/cpgfunctionEP-0.1/LICENSE | 29 + third_party/cpgfunctionEP-0.1/README.md | 88 + .../cpgfunctionEP-0.1/include/blas/blas.h | 26 + .../include/cpgfunction/boreholes.h | 96 + .../include/cpgfunction/coordinates.h | 32 + .../include/cpgfunction/gfunction.h | 54 + .../include/cpgfunction/heat_transfer.h | 52 + .../include/cpgfunction/interpolation.h | 24 + .../include/cpgfunction/statistics.h | 20 + .../include/cpgfunction/utilities.h | 23 + third_party/cpgfunctionEP-0.1/src/blas.cpp | 44 + .../cpgfunctionEP-0.1/src/boreholes.cpp | 281 + .../cpgfunctionEP-0.1/src/coordinates.cpp | 191 + .../cpgfunctionEP-0.1/src/gfunction.cpp | 522 + .../cpgfunctionEP-0.1/src/heat_transfer.cpp | 285 + .../cpgfunctionEP-0.1/src/interpolation.cpp | 91 + .../cpgfunctionEP-0.1/src/statistics.cpp | 25 + .../cpgfunctionEP-0.1/src/utilities.cpp | 85 + .../test/borefield_definition.cpp | 32 + .../test/compute_UBHWT_gFunction.cpp | 138 + .../cpgfunctionEP-0.1/test/configurations.cpp | 51 + .../Poisson_Disk_120_30_101.json | 208 + .../test/gFunction_minimal.cpp | 59 + .../cpgfunctionEP-0.1/test/interpolation.cpp | 23 + third_party/cpgfunctionEP-0.1/test/packed.cpp | 28 + .../test/time_definition.cpp | 23 + .../cpgfunctionEP-0.1/test/validation/L.json | 60 + .../test/validation/OpenRectangle.json | 60 + .../test/validation/Rectangle.json | 60 + .../cpgfunctionEP-0.1/test/validation/U.json | 60 + .../test/validation/custom.json | 60 + .../third_party/eigen-3.3.9/.gitignore | 37 + .../third_party/eigen-3.3.9/.hgeol | 11 + .../eigen-3.3.9/Eigen/CMakeLists.txt | 19 + .../third_party/eigen-3.3.9/Eigen/Cholesky | 46 + .../eigen-3.3.9/Eigen/CholmodSupport | 48 + .../third_party/eigen-3.3.9/Eigen/Core | 542 + .../third_party/eigen-3.3.9/Eigen/Dense | 7 + .../third_party/eigen-3.3.9/Eigen/Eigen | 2 + .../third_party/eigen-3.3.9/Eigen/Eigenvalues | 61 + .../third_party/eigen-3.3.9/Eigen/Geometry | 62 + .../third_party/eigen-3.3.9/Eigen/Householder | 30 + .../eigen-3.3.9/Eigen/IterativeLinearSolvers | 48 + .../third_party/eigen-3.3.9/Eigen/Jacobi | 33 + .../third_party/eigen-3.3.9/Eigen/LU | 50 + .../eigen-3.3.9/Eigen/MetisSupport | 35 + .../eigen-3.3.9/Eigen/OrderingMethods | 73 + .../eigen-3.3.9/Eigen/PaStiXSupport | 48 + .../eigen-3.3.9/Eigen/PardisoSupport | 35 + .../third_party/eigen-3.3.9/Eigen/QR | 51 + .../eigen-3.3.9/Eigen/QtAlignedMalloc | 40 + .../third_party/eigen-3.3.9/Eigen/SPQRSupport | 34 + .../third_party/eigen-3.3.9/Eigen/SVD | 51 + .../third_party/eigen-3.3.9/Eigen/Sparse | 36 + .../eigen-3.3.9/Eigen/SparseCholesky | 45 + .../third_party/eigen-3.3.9/Eigen/SparseCore | 69 + .../third_party/eigen-3.3.9/Eigen/SparseLU | 46 + .../third_party/eigen-3.3.9/Eigen/SparseQR | 36 + .../third_party/eigen-3.3.9/Eigen/StdDeque | 27 + .../third_party/eigen-3.3.9/Eigen/StdList | 26 + .../third_party/eigen-3.3.9/Eigen/StdVector | 27 + .../eigen-3.3.9/Eigen/SuperLUSupport | 64 + .../eigen-3.3.9/Eigen/UmfPackSupport | 40 + .../eigen-3.3.9/Eigen/src/Cholesky/LDLT.h | 673 + .../eigen-3.3.9/Eigen/src/Cholesky/LLT.h | 542 + .../Eigen/src/Cholesky/LLT_LAPACKE.h | 99 + .../Eigen/src/CholmodSupport/CholmodSupport.h | 639 + .../eigen-3.3.9/Eigen/src/Core/Array.h | 329 + .../eigen-3.3.9/Eigen/src/Core/ArrayBase.h | 226 + .../eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h | 209 + .../eigen-3.3.9/Eigen/src/Core/Assign.h | 90 + .../Eigen/src/Core/AssignEvaluator.h | 935 + .../eigen-3.3.9/Eigen/src/Core/Assign_MKL.h | 178 + .../eigen-3.3.9/Eigen/src/Core/BandMatrix.h | 353 + .../eigen-3.3.9/Eigen/src/Core/Block.h | 452 + .../eigen-3.3.9/Eigen/src/Core/BooleanRedux.h | 164 + .../Eigen/src/Core/CommaInitializer.h | 160 + .../Eigen/src/Core/ConditionEstimator.h | 175 + .../Eigen/src/Core/CoreEvaluators.h | 1688 ++ .../Eigen/src/Core/CoreIterators.h | 127 + .../Eigen/src/Core/CwiseBinaryOp.h | 184 + .../Eigen/src/Core/CwiseNullaryOp.h | 866 + .../Eigen/src/Core/CwiseTernaryOp.h | 197 + .../eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h | 103 + .../Eigen/src/Core/CwiseUnaryView.h | 130 + .../eigen-3.3.9/Eigen/src/Core/DenseBase.h | 612 + .../Eigen/src/Core/DenseCoeffsBase.h | 681 + .../eigen-3.3.9/Eigen/src/Core/DenseStorage.h | 570 + .../eigen-3.3.9/Eigen/src/Core/Diagonal.h | 260 + .../Eigen/src/Core/DiagonalMatrix.h | 343 + .../Eigen/src/Core/DiagonalProduct.h | 28 + .../eigen-3.3.9/Eigen/src/Core/Dot.h | 318 + .../eigen-3.3.9/Eigen/src/Core/EigenBase.h | 159 + .../Eigen/src/Core/ForceAlignedAccess.h | 146 + .../eigen-3.3.9/Eigen/src/Core/Fuzzy.h | 155 + .../Eigen/src/Core/GeneralProduct.h | 455 + .../Eigen/src/Core/GenericPacketMath.h | 590 + .../Eigen/src/Core/GlobalFunctions.h | 187 + .../eigen-3.3.9/Eigen/src/Core/IO.h | 225 + .../eigen-3.3.9/Eigen/src/Core/Inverse.h | 118 + .../eigen-3.3.9/Eigen/src/Core/Map.h | 171 + .../eigen-3.3.9/Eigen/src/Core/MapBase.h | 308 + .../Eigen/src/Core/MathFunctions.h | 1421 ++ .../Eigen/src/Core/MathFunctionsImpl.h | 101 + .../eigen-3.3.9/Eigen/src/Core/Matrix.h | 459 + .../eigen-3.3.9/Eigen/src/Core/MatrixBase.h | 530 + .../eigen-3.3.9/Eigen/src/Core/NestByValue.h | 110 + .../eigen-3.3.9/Eigen/src/Core/NoAlias.h | 108 + .../eigen-3.3.9/Eigen/src/Core/NumTraits.h | 248 + .../Eigen/src/Core/PermutationMatrix.h | 605 + .../Eigen/src/Core/PlainObjectBase.h | 1037 + .../eigen-3.3.9/Eigen/src/Core/Product.h | 186 + .../Eigen/src/Core/ProductEvaluators.h | 1138 ++ .../eigen-3.3.9/Eigen/src/Core/Random.h | 182 + .../eigen-3.3.9/Eigen/src/Core/Redux.h | 505 + .../eigen-3.3.9/Eigen/src/Core/Ref.h | 284 + .../eigen-3.3.9/Eigen/src/Core/Replicate.h | 142 + .../Eigen/src/Core/ReturnByValue.h | 117 + .../eigen-3.3.9/Eigen/src/Core/Reverse.h | 211 + .../eigen-3.3.9/Eigen/src/Core/Select.h | 162 + .../Eigen/src/Core/SelfAdjointView.h | 352 + .../Eigen/src/Core/SelfCwiseBinaryOp.h | 47 + .../eigen-3.3.9/Eigen/src/Core/Solve.h | 188 + .../Eigen/src/Core/SolveTriangular.h | 235 + .../eigen-3.3.9/Eigen/src/Core/SolverBase.h | 130 + .../eigen-3.3.9/Eigen/src/Core/StableNorm.h | 221 + .../eigen-3.3.9/Eigen/src/Core/Stride.h | 111 + .../eigen-3.3.9/Eigen/src/Core/Swap.h | 67 + .../eigen-3.3.9/Eigen/src/Core/Transpose.h | 405 + .../Eigen/src/Core/Transpositions.h | 368 + .../Eigen/src/Core/TriangularMatrix.h | 985 + .../eigen-3.3.9/Eigen/src/Core/VectorBlock.h | 96 + .../eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h | 695 + .../eigen-3.3.9/Eigen/src/Core/Visitor.h | 273 + .../Eigen/src/Core/arch/AVX/Complex.h | 451 + .../Eigen/src/Core/arch/AVX/MathFunctions.h | 439 + .../Eigen/src/Core/arch/AVX/PacketMath.h | 637 + .../Eigen/src/Core/arch/AVX/TypeCasting.h | 51 + .../src/Core/arch/AVX512/MathFunctions.h | 389 + .../Eigen/src/Core/arch/AVX512/PacketMath.h | 1305 ++ .../Eigen/src/Core/arch/AltiVec/Complex.h | 430 + .../src/Core/arch/AltiVec/MathFunctions.h | 322 + .../Eigen/src/Core/arch/AltiVec/PacketMath.h | 1061 + .../Eigen/src/Core/arch/CUDA/Complex.h | 103 + .../Eigen/src/Core/arch/CUDA/Half.h | 675 + .../Eigen/src/Core/arch/CUDA/MathFunctions.h | 91 + .../Eigen/src/Core/arch/CUDA/PacketMath.h | 333 + .../Eigen/src/Core/arch/CUDA/PacketMathHalf.h | 1124 ++ .../Eigen/src/Core/arch/CUDA/TypeCasting.h | 212 + .../Eigen/src/Core/arch/Default/ConjHelper.h | 29 + .../Eigen/src/Core/arch/Default/Settings.h | 49 + .../Eigen/src/Core/arch/NEON/Complex.h | 490 + .../Eigen/src/Core/arch/NEON/MathFunctions.h | 91 + .../Eigen/src/Core/arch/NEON/PacketMath.h | 760 + .../Eigen/src/Core/arch/SSE/Complex.h | 471 + .../Eigen/src/Core/arch/SSE/MathFunctions.h | 562 + .../Eigen/src/Core/arch/SSE/PacketMath.h | 895 + .../Eigen/src/Core/arch/SSE/TypeCasting.h | 77 + .../Eigen/src/Core/arch/ZVector/Complex.h | 397 + .../src/Core/arch/ZVector/MathFunctions.h | 137 + .../Eigen/src/Core/arch/ZVector/PacketMath.h | 945 + .../src/Core/functors/AssignmentFunctors.h | 168 + .../Eigen/src/Core/functors/BinaryFunctors.h | 475 + .../Eigen/src/Core/functors/NullaryFunctors.h | 188 + .../Eigen/src/Core/functors/StlFunctors.h | 136 + .../Eigen/src/Core/functors/TernaryFunctors.h | 25 + .../Eigen/src/Core/functors/UnaryFunctors.h | 792 + .../Core/products/GeneralBlockPanelKernel.h | 2157 ++ .../src/Core/products/GeneralMatrixMatrix.h | 495 + .../products/GeneralMatrixMatrixTriangular.h | 317 + .../GeneralMatrixMatrixTriangular_BLAS.h | 145 + .../Core/products/GeneralMatrixMatrix_BLAS.h | 124 + .../src/Core/products/GeneralMatrixVector.h | 619 + .../Core/products/GeneralMatrixVector_BLAS.h | 136 + .../Eigen/src/Core/products/Parallelizer.h | 166 + .../Core/products/SelfadjointMatrixMatrix.h | 527 + .../products/SelfadjointMatrixMatrix_BLAS.h | 295 + .../Core/products/SelfadjointMatrixVector.h | 260 + .../products/SelfadjointMatrixVector_BLAS.h | 118 + .../src/Core/products/SelfadjointProduct.h | 133 + .../Core/products/SelfadjointRank2Update.h | 93 + .../Core/products/TriangularMatrixMatrix.h | 472 + .../products/TriangularMatrixMatrix_BLAS.h | 317 + .../Core/products/TriangularMatrixVector.h | 350 + .../products/TriangularMatrixVector_BLAS.h | 255 + .../Core/products/TriangularSolverMatrix.h | 335 + .../products/TriangularSolverMatrix_BLAS.h | 167 + .../Core/products/TriangularSolverVector.h | 145 + .../Eigen/src/Core/util/BlasUtil.h | 499 + .../Eigen/src/Core/util/Constants.h | 547 + .../src/Core/util/DisableStupidWarnings.h | 94 + .../Eigen/src/Core/util/ForwardDeclarations.h | 298 + .../Eigen/src/Core/util/MKL_support.h | 130 + .../eigen-3.3.9/Eigen/src/Core/util/Macros.h | 1053 + .../eigen-3.3.9/Eigen/src/Core/util/Memory.h | 993 + .../eigen-3.3.9/Eigen/src/Core/util/Meta.h | 568 + .../eigen-3.3.9/Eigen/src/Core/util/NonMPL2.h | 3 + .../src/Core/util/ReenableStupidWarnings.h | 31 + .../Eigen/src/Core/util/StaticAssert.h | 218 + .../Eigen/src/Core/util/XprHelper.h | 838 + .../src/Eigenvalues/ComplexEigenSolver.h | 346 + .../Eigen/src/Eigenvalues/ComplexSchur.h | 462 + .../src/Eigenvalues/ComplexSchur_LAPACKE.h | 91 + .../Eigen/src/Eigenvalues/EigenSolver.h | 622 + .../src/Eigenvalues/GeneralizedEigenSolver.h | 418 + .../GeneralizedSelfAdjointEigenSolver.h | 226 + .../src/Eigenvalues/HessenbergDecomposition.h | 374 + .../src/Eigenvalues/MatrixBaseEigenvalues.h | 158 + .../Eigen/src/Eigenvalues/RealQZ.h | 654 + .../Eigen/src/Eigenvalues/RealSchur.h | 553 + .../Eigen/src/Eigenvalues/RealSchur_LAPACKE.h | 77 + .../src/Eigenvalues/SelfAdjointEigenSolver.h | 871 + .../SelfAdjointEigenSolver_LAPACKE.h | 87 + .../src/Eigenvalues/Tridiagonalization.h | 556 + .../Eigen/src/Geometry/AlignedBox.h | 392 + .../Eigen/src/Geometry/AngleAxis.h | 247 + .../Eigen/src/Geometry/EulerAngles.h | 114 + .../Eigen/src/Geometry/Homogeneous.h | 497 + .../Eigen/src/Geometry/Hyperplane.h | 282 + .../Eigen/src/Geometry/OrthoMethods.h | 234 + .../Eigen/src/Geometry/ParametrizedLine.h | 195 + .../Eigen/src/Geometry/Quaternion.h | 832 + .../Eigen/src/Geometry/Rotation2D.h | 199 + .../Eigen/src/Geometry/RotationBase.h | 206 + .../eigen-3.3.9/Eigen/src/Geometry/Scaling.h | 170 + .../Eigen/src/Geometry/Transform.h | 1542 ++ .../Eigen/src/Geometry/Translation.h | 202 + .../eigen-3.3.9/Eigen/src/Geometry/Umeyama.h | 166 + .../Eigen/src/Geometry/arch/Geometry_SSE.h | 161 + .../Eigen/src/Householder/BlockHouseholder.h | 103 + .../Eigen/src/Householder/Householder.h | 172 + .../src/Householder/HouseholderSequence.h | 470 + .../BasicPreconditioners.h | 226 + .../src/IterativeLinearSolvers/BiCGSTAB.h | 228 + .../ConjugateGradient.h | 246 + .../IncompleteCholesky.h | 400 + .../IterativeLinearSolvers/IncompleteLUT.h | 462 + .../IterativeSolverBase.h | 394 + .../LeastSquareConjugateGradient.h | 216 + .../IterativeLinearSolvers/SolveWithGuess.h | 115 + .../eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h | 462 + .../eigen-3.3.9/Eigen/src/LU/Determinant.h | 101 + .../eigen-3.3.9/Eigen/src/LU/FullPivLU.h | 891 + .../eigen-3.3.9/Eigen/src/LU/InverseImpl.h | 415 + .../eigen-3.3.9/Eigen/src/LU/PartialPivLU.h | 614 + .../Eigen/src/LU/PartialPivLU_LAPACKE.h | 83 + .../Eigen/src/LU/arch/Inverse_SSE.h | 338 + .../Eigen/src/MetisSupport/MetisSupport.h | 137 + .../Eigen/src/OrderingMethods/Amd.h | 445 + .../Eigen/src/OrderingMethods/Eigen_Colamd.h | 1843 ++ .../Eigen/src/OrderingMethods/Ordering.h | 157 + .../Eigen/src/PaStiXSupport/PaStiXSupport.h | 678 + .../Eigen/src/PardisoSupport/PardisoSupport.h | 544 + .../Eigen/src/QR/ColPivHouseholderQR.h | 653 + .../src/QR/ColPivHouseholderQR_LAPACKE.h | 97 + .../src/QR/CompleteOrthogonalDecomposition.h | 562 + .../Eigen/src/QR/FullPivHouseholderQR.h | 676 + .../eigen-3.3.9/Eigen/src/QR/HouseholderQR.h | 409 + .../Eigen/src/QR/HouseholderQR_LAPACKE.h | 68 + .../src/SPQRSupport/SuiteSparseQRSupport.h | 313 + .../eigen-3.3.9/Eigen/src/SVD/BDCSVD.h | 1277 ++ .../eigen-3.3.9/Eigen/src/SVD/JacobiSVD.h | 804 + .../Eigen/src/SVD/JacobiSVD_LAPACKE.h | 91 + .../eigen-3.3.9/Eigen/src/SVD/SVDBase.h | 315 + .../Eigen/src/SVD/UpperBidiagonalization.h | 414 + .../src/SparseCholesky/SimplicialCholesky.h | 689 + .../SparseCholesky/SimplicialCholesky_impl.h | 199 + .../Eigen/src/SparseCore/AmbiVector.h | 378 + .../Eigen/src/SparseCore/CompressedStorage.h | 258 + .../ConservativeSparseSparseProduct.h | 352 + .../Eigen/src/SparseCore/MappedSparseMatrix.h | 67 + .../Eigen/src/SparseCore/SparseAssign.h | 216 + .../Eigen/src/SparseCore/SparseBlock.h | 603 + .../Eigen/src/SparseCore/SparseColEtree.h | 206 + .../src/SparseCore/SparseCompressedBase.h | 341 + .../src/SparseCore/SparseCwiseBinaryOp.h | 726 + .../Eigen/src/SparseCore/SparseCwiseUnaryOp.h | 150 + .../Eigen/src/SparseCore/SparseDenseProduct.h | 320 + .../src/SparseCore/SparseDiagonalProduct.h | 138 + .../Eigen/src/SparseCore/SparseDot.h | 98 + .../Eigen/src/SparseCore/SparseFuzzy.h | 29 + .../Eigen/src/SparseCore/SparseMap.h | 305 + .../Eigen/src/SparseCore/SparseMatrix.h | 1404 ++ .../Eigen/src/SparseCore/SparseMatrixBase.h | 405 + .../Eigen/src/SparseCore/SparsePermutation.h | 178 + .../Eigen/src/SparseCore/SparseProduct.h | 169 + .../Eigen/src/SparseCore/SparseRedux.h | 49 + .../Eigen/src/SparseCore/SparseRef.h | 397 + .../src/SparseCore/SparseSelfAdjointView.h | 656 + .../Eigen/src/SparseCore/SparseSolverBase.h | 124 + .../SparseSparseProductWithPruning.h | 198 + .../Eigen/src/SparseCore/SparseTranspose.h | 92 + .../src/SparseCore/SparseTriangularView.h | 189 + .../Eigen/src/SparseCore/SparseUtil.h | 178 + .../Eigen/src/SparseCore/SparseVector.h | 478 + .../Eigen/src/SparseCore/SparseView.h | 254 + .../Eigen/src/SparseCore/TriangularSolver.h | 315 + .../eigen-3.3.9/Eigen/src/SparseLU/SparseLU.h | 773 + .../Eigen/src/SparseLU/SparseLUImpl.h | 66 + .../Eigen/src/SparseLU/SparseLU_Memory.h | 226 + .../Eigen/src/SparseLU/SparseLU_Structs.h | 110 + .../src/SparseLU/SparseLU_SupernodalMatrix.h | 301 + .../Eigen/src/SparseLU/SparseLU_Utils.h | 80 + .../Eigen/src/SparseLU/SparseLU_column_bmod.h | 181 + .../Eigen/src/SparseLU/SparseLU_column_dfs.h | 179 + .../src/SparseLU/SparseLU_copy_to_ucol.h | 107 + .../Eigen/src/SparseLU/SparseLU_gemm_kernel.h | 280 + .../src/SparseLU/SparseLU_heap_relax_snode.h | 126 + .../Eigen/src/SparseLU/SparseLU_kernel_bmod.h | 130 + .../Eigen/src/SparseLU/SparseLU_panel_bmod.h | 223 + .../Eigen/src/SparseLU/SparseLU_panel_dfs.h | 258 + .../Eigen/src/SparseLU/SparseLU_pivotL.h | 137 + .../Eigen/src/SparseLU/SparseLU_pruneL.h | 136 + .../Eigen/src/SparseLU/SparseLU_relax_snode.h | 83 + .../eigen-3.3.9/Eigen/src/SparseQR/SparseQR.h | 745 + .../Eigen/src/StlSupport/StdDeque.h | 128 + .../Eigen/src/StlSupport/StdList.h | 106 + .../Eigen/src/StlSupport/StdVector.h | 131 + .../Eigen/src/StlSupport/details.h | 84 + .../Eigen/src/SuperLUSupport/SuperLUSupport.h | 1027 + .../Eigen/src/UmfPackSupport/UmfPackSupport.h | 506 + .../eigen-3.3.9/Eigen/src/misc/Image.h | 82 + .../eigen-3.3.9/Eigen/src/misc/Kernel.h | 79 + .../eigen-3.3.9/Eigen/src/misc/RealSvd2x2.h | 55 + .../eigen-3.3.9/Eigen/src/misc/blas.h | 440 + .../eigen-3.3.9/Eigen/src/misc/lapack.h | 152 + .../eigen-3.3.9/Eigen/src/misc/lapacke.h | 16291 ++++++++++++++++ .../Eigen/src/misc/lapacke_mangling.h | 17 + .../Eigen/src/plugins/ArrayCwiseBinaryOps.h | 332 + .../Eigen/src/plugins/ArrayCwiseUnaryOps.h | 552 + .../Eigen/src/plugins/BlockMethods.h | 1058 + .../Eigen/src/plugins/CommonCwiseBinaryOps.h | 115 + .../Eigen/src/plugins/CommonCwiseUnaryOps.h | 163 + .../Eigen/src/plugins/MatrixCwiseBinaryOps.h | 152 + .../Eigen/src/plugins/MatrixCwiseUnaryOps.h | 85 + .../third_party/eigen-3.3.9/README.md | 3 + .../nlohmann/adl_serializer.hpp | 49 + .../nlohmann/byte_container_with_subtype.hpp | 166 + .../nlohmann/detail/conversions/from_json.hpp | 403 + .../nlohmann/detail/conversions/to_chars.hpp | 1105 ++ .../nlohmann/detail/conversions/to_json.hpp | 374 + .../nlohmann/detail/exceptions.hpp | 357 + .../nlohmann-3.9.1/nlohmann/detail/hash.hpp | 117 + .../nlohmann/detail/input/binary_reader.hpp | 2459 +++ .../nlohmann/detail/input/input_adapters.hpp | 454 + .../nlohmann/detail/input/json_sax.hpp | 695 + .../nlohmann/detail/input/lexer.hpp | 1623 ++ .../nlohmann/detail/input/parser.hpp | 501 + .../nlohmann/detail/input/position_t.hpp | 27 + .../detail/iterators/internal_iterator.hpp | 25 + .../nlohmann/detail/iterators/iter_impl.hpp | 637 + .../detail/iterators/iteration_proxy.hpp | 178 + .../detail/iterators/iterator_traits.hpp | 51 + .../iterators/json_reverse_iterator.hpp | 119 + .../detail/iterators/primitive_iterator.hpp | 120 + .../nlohmann/detail/json_pointer.hpp | 975 + .../nlohmann/detail/json_ref.hpp | 76 + .../nlohmann/detail/macro_scope.hpp | 296 + .../nlohmann/detail/macro_unscope.hpp | 23 + .../nlohmann/detail/meta/cpp_future.hpp | 62 + .../nlohmann/detail/meta/detected.hpp | 58 + .../nlohmann/detail/meta/is_sax.hpp | 149 + .../nlohmann/detail/meta/type_traits.hpp | 396 + .../nlohmann/detail/meta/void_t.hpp | 13 + .../nlohmann/detail/output/binary_writer.hpp | 1595 ++ .../detail/output/output_adapters.hpp | 123 + .../nlohmann/detail/output/serializer.hpp | 947 + .../nlohmann/detail/value_t.hpp | 81 + .../nlohmann-3.9.1/nlohmann/json.hpp | 8791 +++++++++ .../nlohmann-3.9.1/nlohmann/json_fwd.hpp | 78 + .../nlohmann-3.9.1/nlohmann/ordered_map.hpp | 171 + .../nlohmann/thirdparty/hedley/hedley.hpp | 1911 ++ .../thirdparty/hedley/hedley_undef.hpp | 143 + .../third_party/qdt-master/.gitignore | 28 + .../third_party/qdt-master/CMakeLists.txt | 3 + .../third_party/qdt-master/LICENSE.txt | 674 + .../third_party/qdt-master/README.md | 67 + .../third_party/qdt-master/common/method.h | 211 + .../third_party/qdt-master/examples/qdt-pi.cc | 28 + .../third_party/qdt-master/examples/test.h | 135 + .../third_party/qdt-master/methods/adaptive.h | 395 + .../qdt-master/methods/constant-step.h | 102 + .../qdt-master/methods/gauss-kronrod.h | 110 + .../qdt-master/methods/monte-carlo.h | 41 + .../qdt-master/methods/rectangle.h | 26 + .../third_party/qdt-master/methods/romberg.h | 87 + .../third_party/qdt-master/methods/simpson.h | 27 + .../qdt-master/methods/trapezoid.h | 26 + .../third_party/qdt-master/qdt.h | 8 + 396 files changed, 147273 insertions(+) create mode 100644 third_party/cpgfunctionEP-0.1/.github/workflows/linux_build.yml create mode 100644 third_party/cpgfunctionEP-0.1/.github/workflows/mac_build.yml create mode 100644 third_party/cpgfunctionEP-0.1/.github/workflows/windows_build.yml create mode 100644 third_party/cpgfunctionEP-0.1/.gitignore create mode 100644 third_party/cpgfunctionEP-0.1/CHANGELOG.md create mode 100644 third_party/cpgfunctionEP-0.1/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.1/CONTRIBUTING.md create mode 100644 third_party/cpgfunctionEP-0.1/LICENSE create mode 100644 third_party/cpgfunctionEP-0.1/README.md create mode 100644 third_party/cpgfunctionEP-0.1/include/blas/blas.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/boreholes.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/coordinates.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/gfunction.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/heat_transfer.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/interpolation.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/statistics.h create mode 100644 third_party/cpgfunctionEP-0.1/include/cpgfunction/utilities.h create mode 100644 third_party/cpgfunctionEP-0.1/src/blas.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/boreholes.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/coordinates.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/gfunction.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/heat_transfer.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/interpolation.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/statistics.cpp create mode 100644 third_party/cpgfunctionEP-0.1/src/utilities.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/borefield_definition.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/compute_UBHWT_gFunction.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/configurations.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/custom_configurations/Poisson_Disk_120_30_101.json create mode 100644 third_party/cpgfunctionEP-0.1/test/gFunction_minimal.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/interpolation.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/packed.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/time_definition.cpp create mode 100644 third_party/cpgfunctionEP-0.1/test/validation/L.json create mode 100644 third_party/cpgfunctionEP-0.1/test/validation/OpenRectangle.json create mode 100644 third_party/cpgfunctionEP-0.1/test/validation/Rectangle.json create mode 100644 third_party/cpgfunctionEP-0.1/test/validation/U.json create mode 100644 third_party/cpgfunctionEP-0.1/test/validation/custom.json create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.gitignore create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.hgeol create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Cholesky create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CholmodSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Core create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Dense create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigen create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigenvalues create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Geometry create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Householder create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/IterativeLinearSolvers create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Jacobi create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/LU create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/MetisSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/OrderingMethods create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PaStiXSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PardisoSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QR create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QtAlignedMalloc create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SPQRSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SVD create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Sparse create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCholesky create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCore create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseLU create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseQR create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdDeque create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdList create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdVector create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SuperLUSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/UmfPackSupport create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LDLT.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/CholmodSupport/CholmodSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Array.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign_MKL.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Block.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ForceAlignedAccess.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/IO.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Map.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Product.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Random.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Select.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/TypeCasting.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/Complex.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/NonMPL2.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexEigenSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexSchur.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexSchur_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/EigenSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedEigenSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/HessenbergDecomposition.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealQZ.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/Tridiagonalization.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AlignedBox.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AngleAxis.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/EulerAngles.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Homogeneous.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Hyperplane.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/OrthoMethods.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/ParametrizedLine.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Quaternion.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Rotation2D.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/RotationBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Scaling.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Transform.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Translation.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Umeyama.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/arch/Geometry_SSE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BiCGSTAB.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteCholesky.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteLUT.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IterativeSolverBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/LeastSquareConjugateGradient.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/SolveWithGuess.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/MetisSupport/MetisSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Amd.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Eigen_Colamd.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Ordering.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PaStiXSupport/PaStiXSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PardisoSupport/PardisoSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/CompleteOrthogonalDecomposition.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/FullPivHouseholderQR.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SPQRSupport/SuiteSparseQRSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/BDCSVD.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/SVDBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/UpperBidiagonalization.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/AmbiVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/CompressedStorage.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/ConservativeSparseSparseProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/MappedSparseMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseAssign.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseBlock.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseColEtree.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCompressedBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseBinaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseUnaryOp.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDenseProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDiagonalProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDot.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseFuzzy.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMap.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrixBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparsePermutation.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseProduct.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRedux.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRef.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSelfAdjointView.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSolverBase.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSparseProductWithPruning.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTranspose.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTriangularView.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseUtil.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseView.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/TriangularSolver.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLUImpl.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Memory.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Structs.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Utils.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_bmod.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_dfs.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_copy_to_ucol.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_gemm_kernel.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_heap_relax_snode.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_kernel_bmod.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_panel_bmod.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_panel_dfs.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pivotL.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pruneL.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_relax_snode.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseQR/SparseQR.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdDeque.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdList.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdVector.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/details.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SuperLUSupport/SuperLUSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/UmfPackSupport/UmfPackSupport.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/Image.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/RealSvd2x2.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/blas.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapack.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke_mangling.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/README.md create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/adl_serializer.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/byte_container_with_subtype.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/from_json.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_chars.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_json.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/exceptions.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/hash.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/binary_reader.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/input_adapters.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/json_sax.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/lexer.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/parser.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/position_t.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/internal_iterator.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iter_impl.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iteration_proxy.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iterator_traits.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/json_reverse_iterator.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/primitive_iterator.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_pointer.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_ref.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_scope.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_unscope.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/cpp_future.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/detected.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/is_sax.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/type_traits.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/void_t.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/binary_writer.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/output_adapters.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/serializer.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/value_t.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json_fwd.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/ordered_map.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley_undef.hpp create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/.gitignore create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/LICENSE.txt create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/README.md create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/common/method.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/examples/qdt-pi.cc create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/examples/test.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/adaptive.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/constant-step.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/gauss-kronrod.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/monte-carlo.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/rectangle.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/romberg.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/simpson.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/trapezoid.h create mode 100644 third_party/cpgfunctionEP-0.1/third_party/qdt-master/qdt.h diff --git a/third_party/cpgfunctionEP-0.1/.github/workflows/linux_build.yml b/third_party/cpgfunctionEP-0.1/.github/workflows/linux_build.yml new file mode 100644 index 00000000000..e015e29b3e0 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/.github/workflows/linux_build.yml @@ -0,0 +1,47 @@ + +name: Build and Test (Linux) + +on: [push] + +jobs: + linux_build: + runs-on: ubuntu-20.04 + + steps: + - uses: actions/checkout@v2 + + - name: Set up Python 3.7 + uses: actions/setup-python@v2 + with: + python-version: 3.7 + + - name: Init Submodules + working-directory: ./ + shell: bash + run: git submodule init && git submodule update + + - name: Install Apt Dependencies + run: sudo apt-get install libboost-dev libblas-dev liblapack-dev + + - name: Create Build Directory + run: cmake -E make_directory ./build/ + + - name: Configure CMake + working-directory: ./build + shell: bash + run: cmake -DCMAKE_BUILD_TYPE=Release ../ + + - name: Build + working-directory: ./build + shell: bash + run: cmake --build . -j 2 + + - name: listFiles + working-directory: ./build + shell: bash + run: ls + + - name: Run Tests + working-directory: ./build + shell: bash + run: ctest --verbose diff --git a/third_party/cpgfunctionEP-0.1/.github/workflows/mac_build.yml b/third_party/cpgfunctionEP-0.1/.github/workflows/mac_build.yml new file mode 100644 index 00000000000..93e865512e5 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/.github/workflows/mac_build.yml @@ -0,0 +1,38 @@ +name: Build and Test (Mac) + +on: [push] + +jobs: + mac_build: + runs-on: macos-10.15 + + steps: + - uses: actions/checkout@v2 + + - name: Set up Python 3.7 + uses: actions/setup-python@v2 + with: + python-version: 3.7 + + - name: Install Python dependencies + run: | + python -m pip install --upgrade pip + pip install aqtinstall + + - name: Create Build Directory + run: cmake -E make_directory ./build/ + + - name: Configure CMake + working-directory: ./build + shell: bash + run: cmake -DCMAKE_BUILD_TYPE=Release ../ + + - name: Build Package + working-directory: ./build + shell: bash + run: cmake --build . -j 2 + + - name: Run Tests + working-directory: ./build + shell: bash + run: ctest --verbose diff --git a/third_party/cpgfunctionEP-0.1/.github/workflows/windows_build.yml b/third_party/cpgfunctionEP-0.1/.github/workflows/windows_build.yml new file mode 100644 index 00000000000..87716d015a6 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/.github/workflows/windows_build.yml @@ -0,0 +1,59 @@ +name: Build and Test (Windows) + +on: [push] + +jobs: + windows_build: + runs-on: self-hosted + continue-on-error: ${{ matrix.allow_failure }} + strategy: + # fail-fast: Default is true, switch to false to allow one platform to fail and still run others + fail-fast: false + matrix: + arch: [x64, x86] + include: + - arch: x64 + allow_failure: false + vs-generator: x64 + package-arch: x86_64 + - arch: x86 + allow_failure: false + vs-generator: Win32 + package-arch: i386 + + steps: + - uses: actions/checkout@v2 + + - name: Set up Python 3.7 + uses: actions/setup-python@v2 + with: + python-version: 3.7 + architecture: ${{ matrix.arch }} + + - name: Install Python dependencies + run: | + python -m pip install --upgrade pip + pip install aqtinstall + + - name: Create Build Directory + run: cmake -E make_directory ./build/ + + - name: Configure CMake + working-directory: ./build + shell: bash + run: | + set -x + cmake -G "Visual Studio 16 2019" -A ${{ matrix.vs-generator }} \ + -DLINK_WITH_PYTHON=ON -DBUILD_FORTRAN=ON -DBUILD_PACKAGE:BOOL=ON \ + -DDOCUMENTATION_BUILD="BuildWithAll" -DTEX_INTERACTION="batchmode" \ + ../ + + - name: Build Package + working-directory: ./build + shell: bash + run: cmake --build . -j 2 + + - name: Run Tests + working-directory: ./build + shell: bash + run: ctest -C Debug diff --git a/third_party/cpgfunctionEP-0.1/.gitignore b/third_party/cpgfunctionEP-0.1/.gitignore new file mode 100644 index 00000000000..a4122fb56eb --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/.gitignore @@ -0,0 +1,7 @@ +cmake-build-debug/ +.idea/ +cmake-build-release/ +build/ +Create_Branch.sh +package-lock.json +Shell_Speed/* diff --git a/third_party/cpgfunctionEP-0.1/CHANGELOG.md b/third_party/cpgfunctionEP-0.1/CHANGELOG.md new file mode 100644 index 00000000000..12191e076de --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/CHANGELOG.md @@ -0,0 +1,104 @@ +# History of changes (cpgfunctionEP) + +## Version 0.1.0 (2021-07-08) + +### Maintenance + +* [Issue 9](https://github.com/j-c-cook/cpgfunctionEP/issues/9) - The project builds on Linux, MacOS and Windows upon + commit [558c63c](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/558c63c9b5c4041282fdb473327463b8fe183fbd). + +* [Issue 6](https://github.com/j-c-cook/cpgfunctionEP/issues/6) - Dependencies are placed into the `third_party/` + folder. The dependencies source code is pasted in, rather than using `git subtree` or `git submodule`. + +* [Issue 5](https://github.com/j-c-cook/cpgfunctionEP/issues/5) - Blas and lapack have been removed, thus deprecating + libraries dependency on Fortran. The following lists what the Fortran function calls were replaced with by commit: + - Commit [a20d3ea](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/a20d3eacec67d5994b65d6716774c50404e26428) - + Replaces the blas copy function with the native C++ algorithm standard copy function (`std::copy`) + - Commits [ec35baf](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/ec35baf21025bb61a105ac9edd06e1e348676702) + and [70ab158](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/70ab1585d0efe5dd53dca90f3cf5685585030ef9) - + Replaces the `blas::axpy` Fortran code with a C++ version. + - Commit [383bf08](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/383bf08f31d020af2cabe63d0d8b86eb4bbb8191) - + Replaces the symmetric packed matrix Fortran call with C++ code. + - Commit [231e35e](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/231e35ea416fe0943514cc82dc79238f3f2b79dd) - The + LU decomposition of lapack is replaced by an Eigen function call. + +* [Issue 3](https://github.com/j-c-cook/cpgfunctionEP/issues/3) - Boost is removed by replacing the thread pools with + `OpenMP` calls, and the Gauss-Kronrod integration is replaced with a header only file by the name of + [qdt](https://github.com/j-c-cook/cpgfunctionEP/tree/MilestoneV0.1/third_party/qdt-master). Qdt is a library + written by Adolfo Munoz who released the code alongside a paper by the name of "Higher Order Ray Marching". + The Kronrod integral of Munoz appears much faster than boost's. + + +### Fixes + +* [Issue 10](https://github.com/j-c-cook/cpgfunctionEP/issues/10) - Vector over flow error associated with number of + time steps being less than 6 is fixed. + Commit [558c63c](https://github.com/j-c-cook/cpgfunctionEP/commit/558c63c9b5c4041282fdb473327463b8fe183fbd) + fixes a vector overflow error that occured when the number of time steps described were less than 6 (`nt<6`). Commit + [94e222b](https://github.com/j-c-cook/cpgfunctionEP/commit/94e222bb0cf670f1a4f3058bd1b7b6c7f8bf8233) now allows the + function time_geometric to have less than 6 time steps. + + +# History of changes (cpgfunction) + +## Version 2.0.0 (2021-05-23) + +### Enhancements + +* [Issue 25](https://github.com/j-c-cook/cpgfunction/issues/25) - Removes all references to the 3D `h_ij` + segment response matrix. See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +* [Issue 32](https://github.com/j-c-cook/cpgfunction/issues/32) - The multi-dimensional matrices, + `q_reconstructed` and `h_ij`, are made one dimensional prior to passage into the temporal superposition + function so that `BLAS` routines can be heavily depended on and the loops completely unraveled. + See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +* [Issue 33](https://github.com/j-c-cook/cpgfunction/issues/33) - It is found that the + packed segment resopnse matrix can be directly made us of in `BLAS spmv`, and that addition + greatly optimizes the temporal superposition function. For now the assumption is made that all + segments in the field are of equivalent length, which is true and fine, but at some point in the + future unequal segment lengths should be made possible again. + See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +### New features + +* [Issue 28](https://github.com/j-c-cook/cpgfunction/issues/28) - + The third party library LinearAlgebra (`jcc:la`) is included and made use of for `LU` + factorization in `gfunction.cpp` + +* [Issue 12](https://github.com/j-c-cook/cpgfunction/issues/12) - + A boolean toggle option is added for multi-threading for computing the + uniform borehole wall temperature (UBHWT) g-function + +### API Changes + +* [Issue 16](https://github.com/j-c-cook/cpgfunction/issues/16) - The `uniform borehole wall temperature` + g-function definition is defined for planned use in EnergyPlus with all arguments. Not all the arguments + currently have a purpose, the adaptive discretization and number of thread arguments are place holders. + +## Version 1.0.0 (2021-05-12) + +### New features + +* [Issue 20](https://github.com/j-c-cook/cpgfunction/issues/20) - + Added OpenBlas as the basic linear algebra subprogram (BLAS) vendor to CMakeLists.txt + +* [Issue 18](https://github.com/j-c-cook/cpgfunction/issues/18) - + Added new borefield interface with API access to typical borehole configurations + +* [Issue 13](https://github.com/j-c-cook/cpgfunction/issues/13) - + Implemented g-function accuracy tests via CMakeLists.txt for a Rectangle, Open Rectangle, U shape, + L shape and a custom (Poisson disk) configuration + +* [Commit 45141fa](https://github.com/j-c-cook/cpgfunction/pull/14/commits/45141fa745d92ac8a08eea2a06801d7a01fac367) - + Create new uniform borehole wall temperature API to consider the new borefield and time API's + +* [Commit f8863ad](https://github.com/j-c-cook/cpgfunction/pull/14/commits/f8863ad6879bdcb43d8bbed48ab1be1701eb56f5) - + Added time vector API and associated test + +* [Commit 654160f](https://github.com/j-c-cook/cpgfunction/pull/14/commits/654160f9b508f57b917fc0630437cff726dc8440) - + Modify API for creating a vector of boreholes (borefield) + + + + diff --git a/third_party/cpgfunctionEP-0.1/CMakeLists.txt b/third_party/cpgfunctionEP-0.1/CMakeLists.txt new file mode 100644 index 00000000000..225772509bd --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/CMakeLists.txt @@ -0,0 +1,87 @@ +cmake_minimum_required(VERSION 3.6) +project(cpgfunctionEP) + +set(CMAKE_CXX_STANDARD 17) +if(CMAKE_CXX_COMPILER_ID MATCHES "GNU|Clang") + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} --std=c++17") +else() + set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON) + add_compile_definitions(_USE_MATH_DEFINES) +endif() + +# Everything will need to access this +include_directories(include) +# Third party includes +#include_directories(third_party/boost-1.71.0) # It depends on boost (stripped) +include_directories(third_party/nlohmann-3.9.1) +include_directories(third_party/eigen-3.3.9) # Eigen for LU decomposition +#include_directories(third_party) # Gauss-Kronrod Integration +#include_directories(src third_party/cquadpack/src) +include_directories(third_party/qdt-master) + +# Create a DLL with an API +add_library(cpgfunctionEP SHARED + src/boreholes.cpp + src/gfunction.cpp + src/heat_transfer.cpp + src/interpolation.cpp + src/utilities.cpp + src/coordinates.cpp + src/statistics.cpp + src/blas.cpp) + +#add_subdirectory(third_party/cquadpack) +#target_link_libraries(cpgfunctionEP third_party/cquadpack/cquadpack) +#target_link_libraries(cpgfunctionEP cquadpack) +#include_directories(${PROJECT_BINARY_DIR}/third_party/cquadpack) + +# It depends on threads +find_package(Threads REQUIRED) +target_link_libraries(cpgfunctionEP ${CMAKE_THREAD_LIBS_INIT}) + +OPTION (USE_OpenMP "Use OpenMP" ON) +IF(USE_OpenMP) + FIND_PACKAGE(OpenMP) + IF(OPENMP_FOUND) + SET(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}") + SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}") + ENDIF() +ENDIF() + +# Create a test binary that depends on the DLL +add_executable(test1 test/gFunction_minimal.cpp) +add_executable(test2 test/interpolation.cpp) +# add_executable(test3 test/configurations.cpp) +add_executable(test4 test/borefield_definition.cpp) +add_executable(test5 test/time_definition.cpp) +add_executable(test6 test/compute_UBHWT_gFunction.cpp) +add_executable(test7 test/packed.cpp) + +target_link_libraries(test1 cpgfunctionEP) +target_link_libraries(test2 cpgfunctionEP) +# target_link_libraries(test3 cpgfunctionEP) +target_link_libraries(test4 cpgfunctionEP) +target_link_libraries(test5 cpgfunctionEP) +target_link_libraries(test6 cpgfunctionEP) +target_link_libraries(test7 cpgfunctionEP) + +# Copy validation files to build directory so tests can open +file(GLOB JsonValidation test/validation/*.json) +file(GLOB CustomLayout test/custom_configurations/*.json) + +file(COPY ${JsonValidation} DESTINATION ${CMAKE_BINARY_DIR}) +file(COPY ${CustomLayout} DESTINATION ${CMAKE_BINARY_DIR}) +message(STATUS "The binary directory: ${CMAKE_BINARY_DIR}") + +# Turn on testing +enable_testing() + +# Test the test binary itself +add_test(NAME RunTest1 COMMAND "${CMAKE_BINARY_DIR}/test1") +add_test(NAME RunTest2 COMMAND "${CMAKE_BINARY_DIR}/test2") +# add_test(NAME RunTest3 COMMAND "${CMAKE_BINARY_DIR}/test3") +add_test(NAME RunTest4 COMMAND "${CMAKE_BINARY_DIR}/test4") +add_test(NAME RunTest5 COMMAND "${CMAKE_BINARY_DIR}/test5") +# Pass variable path into test 6 for json files +add_test(NAME RunTest6 COMMAND ${CMAKE_BINARY_DIR}/test6) +add_test(NAME RunTest7 COMMAND ${CMAKE_BINARY_DIR}/test7) diff --git a/third_party/cpgfunctionEP-0.1/CONTRIBUTING.md b/third_party/cpgfunctionEP-0.1/CONTRIBUTING.md new file mode 100644 index 00000000000..742b3f33069 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/CONTRIBUTING.md @@ -0,0 +1,21 @@ +# Contributing to cpgfunction + +## Changelog + +The cpgfunction library keeps a +[changelog](https://github.com/j-c-cook/cpgfunction/blob/master/CHANGELOG.md) +so that changes upon each release are transparent and easily understood. Prior +to a pull request being accepted, all changes must be marked in the changelog. +The changes should fall under one of the following categories: + +- New features - for new features +- Enhancements - for improvements made to code performance and functionality +- Maintenance - for tidying code +- Changed - for changes in functionality of the code +- Depracated - for soon-to-be removed features +- Removed - for removed features +- Fixes - for any bug fixes + +## Versioning + +This library makes use of [Semantic Versioning](https://semver.org/). diff --git a/third_party/cpgfunctionEP-0.1/LICENSE b/third_party/cpgfunctionEP-0.1/LICENSE new file mode 100644 index 00000000000..1c1a20e0de4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/LICENSE @@ -0,0 +1,29 @@ +BSD 3-Clause License + +Copyright (c) 2020, Jack C. Cook +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are met: + +1. Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + +2. Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + +3. Neither the name of the copyright holder nor the names of its + contributors may be used to endorse or promote products derived from + this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/third_party/cpgfunctionEP-0.1/README.md b/third_party/cpgfunctionEP-0.1/README.md new file mode 100644 index 00000000000..b1040e32bc8 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/README.md @@ -0,0 +1,88 @@ +# cpgfunctionEP +An open source low level (C++) implementation of [Massimo Cimmino][1]'s +g-function methodology. The g-function calculation has to do with the distribution +of heat in a ground heat exchanger (GHE). The g-function, after computed, +can be used to simulate a +ground source heat pump to determine the heat pump +exiting fluid temperatures after a +period of time. It is of paramount importance to accurately predict +the thermal response of the borefield. An improper prediction of the +ground response could result in a system too large (waste of money) or a +system too small (resulting in failure). The g-function is currently the only +known methodology to accurately predict the thermal response of the ground. + +Computing the g-function is a computationally demanding procedure. Historically, +g-functions are pre-computed and stored in libraries (databases) which can be +accessed in GHE design tools. This is done so that the ground source heat pump +can be sized within a few seconds, rather than waiting anywhere from minutes +to hours for a g-function to be computed. This code base was implemented to help +solve a major limitation (problem) associated with +computing large sets of g-functions for borefields containing anywhere from 1 +to 1024 boreholes. Massimo Cimmino's open source implementation of the g-function +calculation +in Python ([pygfunction][2]) has a limitation; at times the memory consumption is +inordinate. +This library was implemented so that computing g-functions would require +significantly less memory to increase throughput on high performance clusters +which contain many "low memory" nodes, and only a few "high memory" nodes. + +The g-function is greatly dependent on the boundary condition used. The following +is a checklist of boundary conditions contained in this library: + +- [ ] Uniform heat flux (UHF) +- [x] Uniform borehole wall temperature (UBHWT) +- [ ] Uniform inlet fluid temperature (UIFT) + +[1]: https://github.com/MassimoCimmino +[2]: https://github.com/MassimoCimmino/pygfunction + +# Build + +Create a build with cmake: + +``` +cd /path/to/repo +mkdir build +cd build +ccmake .. # set settings as needed, shouldn't be any of interest, configure then generate +``` + +On a Visual Studio generator this will create a solution you can launch into Visual Studio. +On makefile style generators, this will create a makefile. From the build directory, just make the project: + +``` +make -j 4 # number of processors to use to build +``` + +Then you can run the test: + +``` +ctest +``` + +On my platform this resulted in: +``` +Test project /home/edwin/Projects/cpgfunction/cmake-build-debug + Start 1: RunTest1 +1/1 Test #1: RunTest1 ......................... Passed 0.79 sec +``` + +This ran what used to just be the main.cpp built file. + +# Acknowledgments +The initial funding of this research project was provided by Dr. Jeffrey D. +Spitler of Oklahoma State University. Further development of the library was +funded by the National Renewable Energy Laboratory (NREL), which is owned by +the U.S. Department of Energy (DOE). The NREL contract requests delivery of +cpgfunction for use in EnergyPlus as a third party tool to calculate fast and +accurate g-functions. Initial integration into EnergyPlus makes use of the +UBHWT boundary condition with the adaptive discretization scheme to approximate +UIFT g-functions. + +The original g-function theory appeared in a Per Eskilon's PhD thesis, +his advisor who helped develop that theory was Prof. Johan Claesson of Lund +University in Sweden. Massimo Cimmino, Assistant Professor at Polytechnique +Montreal an Canada, has built on the work of Eskilson, Claesson +and others. Cimmino has developed a methodology for g-function calculations, +written programs to compute the g-function and then made the program open source. +The openness of Cimmino's g-function calculation is monumental. diff --git a/third_party/cpgfunctionEP-0.1/include/blas/blas.h b/third_party/cpgfunctionEP-0.1/include/blas/blas.h new file mode 100644 index 00000000000..5dc19d916cd --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/blas/blas.h @@ -0,0 +1,26 @@ +// +// Created by jackcook on 6/29/21. +// + +// BLAS functions that have been written in C++ + +#include + +using namespace std; + +#ifndef CPGFUNCTION_BLAS_H +#define CPGFUNCTION_BLAS_H + +namespace jcc { namespace blas { + + void axpy(int &n, double &a, vector &x, vector &y, + int &start, int &n_threads); + + void spmv(int &n, double &alpha, vector &A, vector &x, + double &beta, vector &y, int &start, int &n_threads); + + } // namespace blas +} // namespace jcc + + +#endif //CPGFUNCTION_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/boreholes.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/boreholes.h new file mode 100644 index 00000000000..d4bbcab83b5 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/boreholes.h @@ -0,0 +1,96 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_BOREHOLES_H +#define CPPGFUNCTION_BOREHOLES_H + +namespace gt { + + double Distance_Formula(double x1, double y1, double x2, double y2); + + namespace boreholes { + struct Borehole { + // Destructor + virtual ~Borehole() { + } + + double H; // height or length of the borehole (meters) + double D; // borehole burial depth (meters) + double r_b; // borehole radius (meters) + double x; // position (meters) of the center of the borehole along the x-axis + double y; // position (meters) of the center of the borehole along the y-axis + + Borehole(double H=0.0, + double D=0.0, + double r_b=0.0, + double x=0.0, + double y=0.0) : H(H), D(D), r_b(r_b), x(x), y(y) { + } + + double distance(Borehole target); + tuple position(); + }; + + vectorboreField( + const vector > &coordinates, + const double &r_b, const double &H, const double &D); + + struct SimilaritiesType { + ~SimilaritiesType() {} // destructor + + int nSim = 0; + vector< vector > > Sim; + vector > HSim; + vector > DSim; + vector disSim; + + SimilaritiesType() {} // constructor + }; + + struct Similarity { + ~Similarity() {} // destructor + + // delcare variables + // positive similarities + vector nSimPos; // number of positive similarities + vector > > simPos; // number of sim positions + vector disSimPos; // the distances between each position + vector > HSimPos; // the heights of each borehole + vector > DSimPos; // the burial depth of each borehole + // negative similarities + vector nSimNeg; + vector > > simNeg; + vector disSimNeg; + vector > HSimNeg; + vector > DSimNeg; + + Similarity() {} // constructor + + + void similarities(SimilaritiesType &SimReal, SimilaritiesType &SimImage, + vector &boreSegments, + bool splitRealAndImage = true, double disTol = 0.1, double tol = 1.0e-6); + + void _similarities_group_by_distance(vector &boreSegments, + vector > > &Pairs, + vector &nPairs, vector &disPairs, int &nDis, + double disTol = 0.1); + + void _similarities_one_distance(SimilaritiesType &SimT, vector > &pairs, + vector &boreSegments, const string &kind, + double tol = 1.0e-6); + }; + + } // boreholes name space + +} // gt namespace + +#endif //CPPGFUNCTION_BOREHOLES_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/coordinates.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/coordinates.h new file mode 100644 index 00000000000..ec41f804316 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/coordinates.h @@ -0,0 +1,32 @@ +// +// Created by jackcook on 5/6/21. +// + +#include +#include +#include +#include + +#ifndef CPGFUNCTION_COORDINATES_H +#define CPGFUNCTION_COORDINATES_H + +namespace gt { + namespace coordinates { + + std::vector> configuration(const std::string& shape, int Nx, int Ny, + double Bx, double By); + std::vector> configuration(const std::string& shape, const std::string& input_path); + + std::vector> rectangle(int Nx, int Ny, double Bx, double By); + std::vector> Open_rectangle(int Nx, int Ny, double Bx, double By); + std::vector> U_shape(int Nx, int Ny, double Bx, double By); + std::vector> L_shape(int Nx, int Ny, double Bx, double By); + + std::vector> import_coordinates_from_file(const std::string& input_path); + void export_coordinates_to_file(const std::vector> &coordinates, + const std::string& output_path); + + } +} + +#endif //CPGFUNCTION_COORDINATES_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/gfunction.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/gfunction.h new file mode 100644 index 00000000000..6ce041e185a --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/gfunction.h @@ -0,0 +1,54 @@ +// -*- lsst-c++ -*- + +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_GFUNCTION_H +#define CPPGFUNCTION_GFUNCTION_H + +/** The functions will be listed here in the order which they appear in the associated cpp file **/ + +namespace gt { +namespace gfunction { + /** + * Uniform borehole wall temperature (UBWHT) g-function calculation method + * + * This function superimposes the finite line source (FLS) solution to estimate the g-function of a + * geothermal bore field. Each borehole is modeled as a series of finite line source segments, as proposed + * in [CITE: CimminoBernier2014]. + * + * @param gfunction + * @param boreholes + * @param time + * @param alpha + * @param nSegments + * @param use_similarities + * @param disp + */ + vector uniform_borehole_wall_temperature( + vector &boreField, + vector &time, double alpha, int nSegments=12, + bool use_similarities=true, bool adaptive=true, int n_Threads=1, + bool multi_thread=true, bool display=false); + + void _borehole_segments(vector& boreSegments, + vector& boreholes, int nSegments); + void load_history_reconstruction(vector& q_reconstructed, vector& time, + vector& _time, vector >& Q, + vector& dt, const int p); + void _temporal_superposition(vector& Tb_0, gt::heat_transfer::SegmentResponse &SegRes, + vector &h_ij, vector &q_reconstructed, int p, int &nSources); + void _solve_eqn(vector& x, vector>& A, vector& b); + +} // namespace gfunction +} // namespace gt + +#endif //CPPGFUNCTION_GFUNCTION_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/heat_transfer.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/heat_transfer.h new file mode 100644 index 00000000000..57c0ba9f6f2 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/heat_transfer.h @@ -0,0 +1,52 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include + +using namespace std; +using namespace gt; + +#ifndef CPPGFUNCTION_HEAT_TRANSFER_H +#define CPPGFUNCTION_HEAT_TRANSFER_H + +namespace gt::heat_transfer { + + struct SegmentResponse { + ~SegmentResponse() {} // destructor + + int nSources; + int nSum; + vector < vector < double > > h_ij; + vector boreSegments; + + SegmentResponse(int nSources, + int nSum, + int nt) : + nSources(nSources), + boreSegments(nSources), + h_ij(nSum, vector(nt, 0)), + nSum(nSum) + {} // constructor + + // storage_mode = 1 is the reduced segment response vector + int storage_mode = 1; + +// void ReSizeContainers(int n, int nt); + void get_h_value(double &h, int i, int j, int k); + void get_index_value(int &index, int i, int j); + }; // struct SegmentResponse(); + + double finite_line_source(double time_, double alpha, + boreholes::Borehole& b1, boreholes::Borehole& b2, + bool reaSource=true, bool imgSource=true); + void thermal_response_factors(SegmentResponse &SegRes, + vector& time, double alpha, + bool use_similaries, bool disp=false); + +} // namespace gt::heat_transfer + +#endif //CPPGFUNCTION_HEAT_TRANSFER_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/interpolation.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/interpolation.h new file mode 100644 index 00000000000..019a271534b --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/interpolation.h @@ -0,0 +1,24 @@ +// +// Created by jackcook on 7/15/20. +// + +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_INTERPOLATION_H +#define CPPGFUNCTION_INTERPOLATION_H + +namespace jcc { namespace interpolation { + + double linterp(double xp, double x0, double y0, double x1, double y1); + void interp1d(vector& xp, vector& yp, vector& x, vector& y); + void interp1d(double &xp, double &yp, vector& x, vector& y); + void interp1d(double &xp, double &yp, vector &time, + gt::heat_transfer::SegmentResponse &SegRes, int &i, int &j, int &k); + +} } // jcc::interpolation + +#endif //CPPGFUNCTION_INTERPOLATION_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/statistics.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/statistics.h new file mode 100644 index 00000000000..b9dc497bf67 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/statistics.h @@ -0,0 +1,20 @@ +// +// Created by jackcook on 5/10/21. +// + +#include +#include +#include + +#ifndef CPGFUNCTION_STATISTICS_H +#define CPGFUNCTION_STATISTICS_H + +namespace gt { + namespace statistics { + + double root_mean_square_error(std::vector actual, std::vector predicted); + + } // namespace statistics +} // namespace gt + +#endif //CPGFUNCTION_STATISTICS_H diff --git a/third_party/cpgfunctionEP-0.1/include/cpgfunction/utilities.h b/third_party/cpgfunctionEP-0.1/include/cpgfunction/utilities.h new file mode 100644 index 00000000000..8b7cbc28d86 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/include/cpgfunction/utilities.h @@ -0,0 +1,23 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include + +#ifndef CPPGFUNCTION_UTILITIES_H +#define CPPGFUNCTION_UTILITIES_H + +namespace gt { + namespace utilities { + std::vector time_geometric(double dt, double tmax, int Nt); + std::vector Eskilson_original_points(); + std::vector time_Eskilson(const double &H, const double &alpha); + std::vector convert_time(std::vector &logtime, const double &H, const double &alpha); + void cook_spitler_time (std::vector &logtime); + void convert_time(std::vector &logtime, std::vector &time, double H, double alpha); + } // namespace utilities +} // namespace gt + +#endif //CPPGFUNCTION_UTILITIES_H diff --git a/third_party/cpgfunctionEP-0.1/src/blas.cpp b/third_party/cpgfunctionEP-0.1/src/blas.cpp new file mode 100644 index 00000000000..0bcf4248cbf --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/blas.cpp @@ -0,0 +1,44 @@ +// +// Created by jackcook on 6/29/21. +// + +#include + +namespace jcc { namespace blas { + + void axpy(int &n, double &a, vector &x, vector &y, + int &start, int &n_threads){ + // y = a*x + y + #pragma omp parallel for num_threads(n_threads) + for (int i=0; i &A, vector &x, + double &beta, vector &y, int &start, int &n_threads) { + // TODO: implement upper + + // LOWER + double zero = 0.; + double temp1; + double temp2; + int kk = 0; + int k; + for (int j=0; j + + +namespace gt { + + double Distance_Formula(double x1, double y1, double x2, double y2) { + return sqrt(pow((x1 - x2), 2) + pow((y1 - y2), 2)); + } + + namespace boreholes { + + double Borehole::distance(Borehole target) { + double x1 = x; + double y1 = y; + double x2 = target.x; + double y2 = target.y; + double dist = Distance_Formula(x1, y1, x2, y2); + return std::max(r_b, double(dist)); // max needs doubles + } + + std::tuple Borehole::position() { + std::tuple t (x, y); + return t; + }; + + std::vector boreField(const std::vector> &coordinates, const double &r_b, + const double &H, const double &D){ + std::vector bores(coordinates.size()); + + double x; + double y; + + for (std::size_t i = 0; i < coordinates.size(); i++) { + x = std::get<0>(coordinates[i]); + y = std::get<1>(coordinates[i]); + bores[i] = Borehole(H, D, r_b, x, y); + } // next i + + return bores; + } // boreField(); + + void Similarity::similarities(SimilaritiesType &SimReal, SimilaritiesType &SimImage, + vector &boreSegments, bool splitRealAndImage, + double disTol, double tol) { + // TODO: fork a pool + + // declare the variables local to this function + int nDis; + vector disPairs; + vector nPairs; + vector< vector < tuple > > Pairs; + _similarities_group_by_distance(boreSegments, Pairs, nPairs, disPairs, nDis); + + vector RealSimT(Pairs.size()); + vector ImageSimT; + + // if real and image parts of the FLS are split, evaluate real and image similarities seperately: + if (splitRealAndImage) { + ImageSimT.resize(Pairs.size()); + for (std::size_t i=0; i &SimFrom, SimilaritiesType &SimTo) { + int nSimtmp; + // reserve and insert + auto _res_and_ins = [](vector &SimFrom, SimilaritiesType &SimTo, const int &i) { + // Sim positions + SimTo.Sim.reserve(SimTo.Sim.size() + distance(SimFrom[i].Sim.begin(), SimFrom[i].Sim.end())); + SimTo.Sim.insert(SimTo.Sim.end(), SimFrom[i].Sim.begin(), SimFrom[i].Sim.end()); + // H values + SimTo.HSim.reserve(SimTo.HSim.size() + distance(SimFrom[i].HSim.begin(), SimFrom[i].HSim.end())); + SimTo.HSim.insert(SimTo.HSim.end(), SimFrom[i].HSim.begin(), SimFrom[i].HSim.end()); + // D Values + SimTo.DSim.reserve(SimTo.DSim.size() + distance(SimFrom[i].DSim.begin(), SimFrom[i].DSim.end())); + SimTo.DSim.insert(SimTo.DSim.end(), SimFrom[i].DSim.begin(), SimFrom[i].DSim.end()); + }; + for (int i=0; i > > tmp; + tmp = SimFrom[i].Sim; + _res_and_ins(SimFrom, SimTo, i); + } + }; + // Sim positions + _aggregate(RealSimT, SimReal); + if (splitRealAndImage) { + _aggregate(ImageSimT, SimImage); + } + + int a = 1; + } // Similarity::similarities + + void Similarity::_similarities_group_by_distance(vector &boreSegments, + vector< vector < tuple > > &Pairs, vector &nPairs, vector &disPairs, int &nDis, + double disTol) { + // initialize lists + nPairs.push_back(1); + vector< tuple > vect_w_tup(1); + vect_w_tup[0] = tuple (0, 0); + Pairs.push_back(vect_w_tup); + disPairs.push_back(boreSegments[0].r_b); + nDis = 1; + + int nb = boreSegments.size(); + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + int i2; + double dis; + double rTol; + double diff; + for (int i=0; i (i, j)); + nPairs[k]++; + break; + } // fi disPairs[k] - dis < rTol + // add symmetry to the list if no match was found + if (k == nDis-1) { + nDis++; + disPairs.push_back(dis); + vect_w_tup[0] = tuple (i, j); + Pairs.push_back(vect_w_tup); + nPairs.push_back(1); + break; + } + } // next k + } // for j in range(i2, nb) + } // for i in range(nb) + int a = 1; + } // Similarity::_similarities_group_by_distance + + void Similarity::_similarities_one_distance(SimilaritiesType & SimT, vector > &pairs, + vector &boreSegments, const string& kind, + double tol) { + // Condition for equivalence of the real part of the FLS solution + auto compare_real_segments = [](const double &H1a, const double &H1b, const double &H2a, const double &H2b, + const double &D1a, const double &D1b, const double &D2a, const double &D2b, + const double &tol){ + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && abs((H2a - H2b) / H2a) < tol && + abs(((D2a - D1a) - (D2b - D1b)) / (D2a - D1a + 1e-30)) < tol; + return similarity; + }; + // Condition for equivalence of the image part of the FLS solution + auto compare_image_segments = [](const double &H1a, const double &H1b, const double &H2a, const double &H2b, + const double &D1a, const double &D1b, const double &D2a, const double &D2b, + const double &tol) { + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && abs((H2a - H2b) / H2a) < tol && + abs(((D2a + D1a) - (D2b + D1b)) / (D2a + D1a + 1e-30)) < tol; + return similarity; + }; + // Condition for equivalence of the full FLS solution + auto compare_realandimage_segments = [](const double &H1a, const double &H1b, const double &H2a, const double &H2b, + const double &D1a, const double &D1b, const double &D2a, const double &D2b, + const double &tol) { + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && abs((H2a - H2b) / H2a) < tol && + abs((D1a - D1b) / (D1a + 1e-30)) < tol && abs((D2a - D2b) / (D2a + 1e-30)) < tol; + return similarity; + }; + + string real("real"); + string image("image"); + string realandimage("realandimage"); + + // compare segments is a pointer to one of the lambda functions + bool(*compare_segments)(const double&, const double&, const double&, const double&, + const double&, const double&, const double&, const double&, const double&); + + if (real.compare(kind) == 0) { + compare_segments = compare_real_segments; + } else if (image.compare(kind) == 0) { + compare_segments = compare_image_segments; + } else if (realandimage.compare(kind) == 0) { + compare_segments = compare_realandimage_segments; + } else { + throw invalid_argument("Error kind not implemented yet."); + } + + SimT.nSim = 1; + tuple doub_tup_temp_H; + tuple doub_tup_temp_D; + tuple int_tup_temp_sim; + vector< tuple > vect_w_tup(1); + vect_w_tup[0] = pairs[0]; +// tuple pair0 = pairs[0]; + int i0 = get<0>(pairs[0]); + int j0 = get<1>(pairs[0]); + SimT.Sim.push_back(vect_w_tup); + doub_tup_temp_H = make_tuple(boreSegments[i0].H, boreSegments[j0].H); + SimT.HSim.push_back(doub_tup_temp_H); + doub_tup_temp_D = make_tuple(boreSegments[i0].D, boreSegments[j0].D); + SimT.DSim.push_back(doub_tup_temp_D); + + // values used in loops + int ibor; + int jbor; + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + double H1; + double H2; + double D1; + double D2; + + // Cycle through all pairs of boreholes for the given distance + for (std::size_t i=1; i(pairs[i]); + jbor = get<1>(pairs[i]); + if (ibor > jbor) { + swap(ibor, jbor); + } + b1 = boreSegments[ibor]; + b2 = boreSegments[jbor]; + // Verify if the current pair should be included in the previously identified symmetries + for (int j=0; j(SimT.HSim[j]); + H2 = get<1>(SimT.HSim[j]); + D1 = get<0>(SimT.DSim[j]); + D2 = get<1>(SimT.DSim[j]); + if (compare_segments(H1, b1.H, H2, b2.H, D1, b1.D, D2, b2.D, tol)) { + int_tup_temp_sim = make_tuple(ibor, jbor); + SimT.Sim[j].push_back(int_tup_temp_sim); + break; + } else if (compare_segments(H1, b2.H, H2, b1.H, D1, b2.D, D2, b1.D, tol)) { + int_tup_temp_sim = make_tuple(jbor, ibor); + SimT.Sim[j].push_back(int_tup_temp_sim); + break; + } else if (j == SimT.nSim-1) { + SimT.nSim++; + int_tup_temp_sim = make_tuple(ibor, jbor); + vect_w_tup[0] = int_tup_temp_sim; + SimT.Sim.push_back(vect_w_tup); + doub_tup_temp_H = make_tuple(b1.H, b2.H); + SimT.HSim.push_back(doub_tup_temp_H); + doub_tup_temp_D = make_tuple(b1.D, b2.D); + SimT.DSim.push_back(doub_tup_temp_D); + break; + } + } // next j + } // next i + } // Similarity::_similarities_one_distance + + } // namespace boreholes + +} // namespace gt diff --git a/third_party/cpgfunctionEP-0.1/src/coordinates.cpp b/third_party/cpgfunctionEP-0.1/src/coordinates.cpp new file mode 100644 index 00000000000..d02182299b1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/coordinates.cpp @@ -0,0 +1,191 @@ +// +// Created by jackcook on 5/6/21. +// + +#include +#include +#include +#include + +using namespace std; + +namespace gt::coordinates { + + vector> configuration(const string& shape, int Nx, + int Ny, double Bx, double By){ + // define acceptable inputs + vector acceptable_arguments{"Rectangle", + "OpenRectangle", + "U", + "L"}; + // check if the input string is acceptable + bool acceptable = (find(acceptable_arguments.begin(), + acceptable_arguments.end(), + shape) != acceptable_arguments.end()); + + if (!acceptable) { + throw std::invalid_argument("The shape described (" + + shape + ") is not an available input for " + "gt::coordinates::configuration()."); + } + + std::vector> coordinates; + + if (shape == "Rectangle") { + coordinates = rectangle(Nx, Ny, Bx, By); + } else if (shape == "OpenRectangle") { + coordinates = Open_rectangle(Nx, Ny, Bx, By); + } else if (shape == "U") { + coordinates = U_shape(Nx, Ny, Bx, By); + } else if (shape == "L") { + coordinates = L_shape(Nx, Ny, Bx, By); + } + + return coordinates; + } // configuration(); + + vector> configuration(const string& shape, + const string& input_path){ + // define acceptable inputs + std::vector acceptable_arguments{"custom"}; + // check if the input string is acceptable + bool acceptable = (std::find(acceptable_arguments.begin(), acceptable_arguments.end(), shape) + != acceptable_arguments.end()); + + if (!acceptable) { + throw invalid_argument("The shape described (" + + shape + ") is not an available input for " + "gt::coordinates::configuration()."); + } + + vector> custom = + import_coordinates_from_file(input_path); + + return custom; + } // configuration(); + + vector> rectangle(int Nx, int Ny, double Bx, + double By) { + std::vector> r; // rectangular coordinates + + int nbh = Nx * Ny; + r.reserve(nbh); + for (int i = 0; i < Nx; i++) { + for (int j = 0; j < Ny; j++) { + r.emplace_back(i * Bx, j * By); + } // next j + } // next i + + return r; + } // rectangle(); + + vector> Open_rectangle(int Nx, int Ny, double Bx, + double By){ + vector> open_r; // open rectangle coordinates + + int nbh; + if (Nx > 2 && Ny > 2) { + nbh = Ny * 2 + (Nx - 2) * 2; + open_r.reserve(nbh); + for (int i = 0; i < Nx; i++) { + open_r.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny - 1; j++) { + open_r.emplace_back(0., j * By); + open_r.emplace_back((Nx-1) * Bx, j * By); + } // next j + for (int i = 0; i < Nx; i++) { + open_r.emplace_back(i * Bx, (Ny-1)*By); + } // next i + } // if (Nx > 2 && Ny > 2) + else { + nbh = Nx * Ny; + open_r.reserve(nbh); + open_r = rectangle(Nx, Ny, Bx, By); + } // else() + + return open_r; + } // Open_rectangle(); + + vector> U_shape(int Nx, int Ny, double Bx, double By){ + std::vector> U; // U-shape coordinates + + int nbh; + if (Nx > 2 && Ny > 1) { + nbh = 2 * Ny + (Nx - 2); + U.reserve(nbh); + for (int i = 0; i < Nx; i++){ + U.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny; j++) { + U.emplace_back(0., j * By); + U.emplace_back((Nx-1) * Bx, j * By); + } // next j + } // if (Nx > 2 && Ny > 1) + else { + nbh = Nx * Ny; + U.reserve(nbh); + U = rectangle(Nx, Ny, Bx, By); + } // else() + + return U; + } // U_shape(); + + vector> L_shape(int Nx, int Ny, double Bx, double By){ + std::vector> L; // L-shape coordinates + + int nbh = Nx + Ny - 1; + L.reserve(nbh); // reserve for "emplace_back" + for (int i = 0; i < Nx; i++){ + L.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny; j++){ + L.emplace_back(0., j * By); + } // next j + + return L; + } // L_shape(); + + vector> import_coordinates_from_file( + const string& input_path){ + vector> custom; // custom-shape from .json file + + // nlohmann json input + std::ifstream in(input_path); + nlohmann::json js; + in >> js; + + std::vector x = js["x"]; + std::vector y = js["y"]; + + custom.reserve(x.size()); + for (int i = 0; i < x.size(); i++){ + custom.emplace_back(x[i], y[i]); + } + + return custom; + } // import_coordinates_from_file(); + + void export_coordinates_to_file( + const vector> &coordinates, + const std::string& output_path){ + // Use nlohmann json to export the coordinates to a path + + std::vector x_values(coordinates.size()); + std::vector y_values(coordinates.size()); + + for (int i=0; i(coordinates[i]); + y_values[i] = std::get<1>(coordinates[i]); + } + + std::ofstream o(output_path); + + nlohmann::json j; + + j["x"] = x_values; + j["y"] = y_values; + + o << std::setw(4) << j << std::endl; + } // export_coordinates_to_file(); +} // namespace gt::coordinates \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/src/gfunction.cpp b/third_party/cpgfunctionEP-0.1/src/gfunction.cpp new file mode 100644 index 00000000000..cf84c9a3db3 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/gfunction.cpp @@ -0,0 +1,522 @@ +// -*- lsst-c++ -*- + +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include +#include +#include + + +using namespace std; // lots of vectors, only namespace to be used + +namespace gt { namespace gfunction { + // The uniform borehole wall temperature (UBWHT) g-function calculation. + // Originally presented in Cimmino and Bernier (2014) and a later paper on + // speed improvements by Cimmino (2018) + vector uniform_borehole_wall_temperature( + vector &boreField, vector &time, + double alpha, int nSegments, bool use_similarities, bool adaptive, + int n_Threads, bool multi_thread, bool display){ + vector gFunction(time.size()); + + if (display) { + std::cout << "---------------------------------------------------" + "---------" << std::endl; + std::cout << "Calculating g-function for uniform borehole wall " + "temperature" << std::endl; + std::cout << "---------------------------------------------------" + "---------" << std::endl; + } + auto startall = std::chrono::steady_clock::now(); + // Open up processes here + // Create a vector of threads + //may return 0 when not able to detect + const auto processor_count = thread::hardware_concurrency(); + // TODO: make n_threads optional + n_Threads = int(processor_count); + + if (display) { + // Launch the pool with n threads. + cout << "\tDetected " << processor_count << + " as the number of available threads" << endl; + cout << "\tMaking use of " << n_Threads << " threads." << endl; + } + + // Number of boreholes + int nbh = boreField.size(); + // Total number of line sources + int nSources = nSegments * nbh; + // Number of time values + int nt = time.size(); + + auto sum_to_n = [](const int n) { + return n * (n + 1) / 2; + }; + int nSum = sum_to_n(nSources); + + // Segment Response struct + gt::heat_transfer::SegmentResponse SegRes(nSources, nSum, nt); + + // Split boreholes into segments + _borehole_segments(SegRes.boreSegments, boreField, nSegments); + + // Calculate segment to segment thermal response factors + auto start = std::chrono::steady_clock::now(); + gt::heat_transfer::thermal_response_factors(SegRes, + time, + alpha, + use_similarities, + display); + auto end = std::chrono::steady_clock::now(); + + if (display) { + std::cout << "Building and solving system of equations ..." + << std::endl; + } + // --------------------------------------------------------------------- + // Build a system of equation [A]*[X] = [B] for the evaluation of the + // g-function. [A] is a coefficient matrix, [X] = [Qb,Tb] is a state + // space vector of the borehole heat extraction rates and borehole wall + // temperature (equal for all segments), [B] is a coefficient vector. + // --------------------------------------------------------------------- + + // -------- timings for debug + double milli = 0; + double segment_length_time = 0; + double time_vector_time = 0; + double segment_h_values_time = 0; + double fill_A_time = 0; + double load_history_reconstruction_time = 0; + double temporal_superposition_time = 0; + double fill_gsl_matrices_time = 0; + double LU_decomposition_time = 0; + + auto start2 = std::chrono::steady_clock::now(); + + // ------ Segment lengths ------- + start = std::chrono::steady_clock::now(); + std::vector Hb(nSources); + # pragma omp parallel for num_threads(processor_count) + for (int b=0; b + (end - start).count(); + segment_length_time += milli; + + // ------ time vectors --------- + start = std::chrono::steady_clock::now(); + // create new time vector that starts at 0 + std::vector _time_untouched(time.size()+1); + std::vector _time(time.size()+1); + std::vector dt(time.size()+1); + +// auto _fill_time = [&_time, &time, &dt, &_time_untouched]() { +// for (int i=0; i<_time.size(); i++) { +// if (i==0) { +// _time[0] = 0; +// _time_untouched[0] = 0; +// dt[i] = time[i]; +// } else { +// _time[i] = time[i-1]; +// _time_untouched[i] = time[i-1]; +// dt[i] = time[i] - time[i-1]; +// } // fi +// } // next i +// }; // auto _fill_time +// if (multi_thread) { +// boost::asio::post(pool, [&_fill_time]{ _fill_time() ;}); +// } else { +// _fill_time(); +// } // if (multi_thread); + # pragma omp parallel for num_threads(processor_count) + for (int i=0; i<_time.size(); i++) { + if (i==0) { + _time[0] = 0; + _time_untouched[0] = 0; + dt[i] = time[i]; + } else { + _time[i] = time[i-1]; + _time_untouched[i] = time[i-1]; + if (i < time.size()) { + dt[i] = time[i] - time[i-1]; + } else if (i == time.size()) { + dt[i] = time[i-1] - time[i-2]; + } + } // fi + } // next i + + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + time_vector_time += milli; + + // ---------- segment h values ------------- + /** Starting up pool2 here **/ + // Launch the pool with n threads. + auto tic = std::chrono::steady_clock::now(); +// boost::asio::thread_pool pool2(processor_count); + auto toc = std::chrono::steady_clock::now(); + if (display) { + double milli = chrono::duration_cast + (tic - toc).count(); + double seconds = milli; + std::cout << "Time to open a pool : " + << seconds + << " sec" << std::endl; + } + + start = std::chrono::steady_clock::now(); + + end = chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + segment_h_values_time += milli; + + // after interpolation scheme, get rid of h_ij first + // Initialize segment heat extraction rates + vector > Q(nSources, vector (nt)); + + // Define A and b for utitilizing Ax=b + /** + * A = [ [ ], + * [hb[0:len(hb), 0] + * ] + * b = [ [ ], + * [sum(hb)] + * ] + * **/ + + int SIZE = nSources + 1; + Eigen::MatrixXd A(SIZE, SIZE); + Eigen::MatrixXd B(SIZE, 1); + Eigen::MatrixXd X(SIZE, 1); + + // Fill A + int n = SIZE - 1; + double Hb_sum=0; + for (auto & _hb : Hb) { + Hb_sum += _hb; + } + + // Build and solve the system of equations at all times + + // the loop p=n depends on what occured at p=n-1, so this will be be in + // series however threading will be interspersed throughout to make use + // of as many threads as possible + std::vector Tb_0 (nSources); + // Restructured load history + // create interpolation object for accumulated heat extraction + std::vector> + q_reconstructed (nSources, std::vector (nt)); + std::vector q_r(nSources * nt, 0); + + int gauss_sum = nSources * (nSources + 1) / 2; + std::vector H_ij(gauss_sum * nt, 0); // 1D nSources x nt + int idx; + for (int i=0; i + (end - start).count(); + fill_A_time += milli; + + // ----- load history reconstruction ------- + start = std::chrono::steady_clock::now(); + load_history_reconstruction(q_r,time, _time, Q, dt, p); + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + load_history_reconstruction_time += milli; + + // ----- temporal superposition + start = chrono::steady_clock::now(); + _temporal_superposition(Tb_0, + SegRes, + H_ij, + q_r, + p, + nSources); + // fill b with -Tb + B(SIZE-1, 0) = Hb_sum; + for (int i=0; i + (end - start).count(); + temporal_superposition_time += milli; + + int m = SIZE; + int n = SIZE; + vector x(SIZE); +// _solve_eqn(x, A, b); + /** was _solve_eqn **/ + + // ---- fill gsl matrix A and b ----- + start = std::chrono::steady_clock::now(); + + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + fill_gsl_matrices_time += milli; + + // ----- LU decomposition ----- + start = std::chrono::steady_clock::now(); + X = A.lu().solve(B); + + for (int i=0; i + (end - start).count(); + LU_decomposition_time += milli; + + // ---- Save Q's for next p --- + for (int j=0; j + (end2 - start2).count(); + double seconds1 = milli1 / 1000; + double milli2 = chrono::duration_cast + (end2 - startall).count(); + double seconds2 = milli2 / 1000; + std::cout << "Elapsed time in seconds : " + << seconds1 + << " sec" << std::endl; + std::cout << "Total time for g-function evaluation : " + << seconds2 + << " sec" << std::endl; + } + + return gFunction; + } // uniform_borehole_wall_temperature(); + + void _borehole_segments(std::vector& boreSegments, + std::vector& boreholes, const int nSegments) { + double H; + double D; + int count = 0; + // Split boreholes into segments + for(auto& b : boreholes) { + // TODO: maybe thread this later on + for (int i=0; i& q_reconstructed, + vector& time, vector& _time, + vector >& Q, vector& dt, const int p) { + // for s in range p+1 + int nSources = Q.size(); + + // Inverted time steps + std::vector dt_reconstructed (p+1); + for (int i=p; i>=0; i--) { + dt_reconstructed[p-i] = dt[i]; // reverse the dt + } + // t_restructured is [0, cumsum(dt_reversed)] + std::vector t_reconstructed(p+2); // will start at 0 + for (int i=1; i<=p+1; i++) { + t_reconstructed[i] = dt_reconstructed[i-1]; + } + for (int i=1; i<=p+1; i++) { + t_reconstructed[i] = t_reconstructed[i] + t_reconstructed[i-1]; + } + // local time vector + std::vector t(p+3); + for (int i=0; i> Q_dt (nSources, + std::vector (t.size())); + auto _Q_dot_dt = [&Q_dt, &Q, &dt, &p, &_tsize, &t](const int i) { + for (int j = 1; j<_tsize; j++) { + if (j>=p+1) { + Q_dt[i][j] = Q_dt[i][j-1]; + } else { + Q_dt[i][j] = Q[i][j-1] * dt[j-1] + Q_dt[i][j-1]; + } // fi + } // next j + }; + for (int i=0; i y(n); + for (int j=0; j yp(n2); + jcc::interpolation::interp1d(t_reconstructed, yp, t, y); + + int idx; + for (int j=0; j& Tb_0, gt::heat_transfer::SegmentResponse &SegRes, + vector &h_ij, vector &q_reconstructed, + const int p, int &nSources) + { + // This function performs equation (37) of Cimmino (2017) + std::fill(Tb_0.begin(), Tb_0.end(), 0); + // Number of time steps + int nt = p + 1; + + const auto processor_count = thread::hardware_concurrency(); + int n_threads = int(processor_count); + + int gauss_sum = nSources * (nSources + 1) / 2; // Number of positions in packed symmetric matrix + // Storage of h_ij differences + std::vector dh_ij(gauss_sum, 0); + int begin_1; // integer declarations for where the linear algebra will begin + int begin_2; + int begin_q; // time for q_reconstructed to begin + int inc = 1; // the vectors are of increment 1, they can be completely unwrapped in BLAS + + double alpha = 1; + double alpha_n = -1; + + std::vector::iterator begin_it_1; + std::vector::iterator end_it_1; +// std::vector::iterator begin_it_2; +// std::vector::iterator end_it_2; + + for (int k = 0; k < nt; k++) { + begin_1 = k * gauss_sum; + begin_it_1 = h_ij.begin() + begin_1; + end_it_1 = h_ij.begin() + begin_1 + gauss_sum; + if (k==0){ + // dh_ij = h(k) + std::copy(begin_it_1, end_it_1, dh_ij.begin()); + } else { + begin_2 = (k-1) * gauss_sum; + // h_1 -> dh_ij + std::copy(begin_it_1, end_it_1, dh_ij.begin()); + // dh_ij = -1 * h(k) + h(k-1) + jcc::blas::axpy(gauss_sum, alpha_n, h_ij, dh_ij, begin_2, + n_threads); + } + // q_reconstructed(t_k - t_k') + begin_q = (nt - k - 1) * nSources; + // dh_ij is a lower triangular packed matrix + char uplo = 'l'; + // Tb_0 = 1 * dh_ij * q(t_k-t_k') + 1 * Tb_0 +// jcc::blas::dspmv_(&uplo, &nSources, &alpha, &*dh_ij.begin(), +// &q_reconstructed.at(begin_q), &inc, &alpha, +// &*Tb_0.begin(), &inc); + jcc::blas::spmv(nSources, alpha, dh_ij, q_reconstructed, alpha, + Tb_0, begin_q, n_threads); + } // next k + } // _temporal_superposition(); +} } // namespace gt::gfunction diff --git a/third_party/cpgfunctionEP-0.1/src/heat_transfer.cpp b/third_party/cpgfunctionEP-0.1/src/heat_transfer.cpp new file mode 100644 index 00000000000..f66818e8dce --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/heat_transfer.cpp @@ -0,0 +1,285 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include +#include +#include + +using namespace gt; +using namespace std; + +namespace gt::heat_transfer { + + double finite_line_source(const double time_, const double alpha, + boreholes::Borehole &b1, boreholes::Borehole &b2, + bool reaSource, bool imgSource) { + + auto _Ils = [&b1, &b2, reaSource, imgSource](const double s) { + auto _erfint = [](const double x) { + return x * std::erf(x) - (1 / sqrt(M_PI)) * (1 - exp(-pow(x, 2))); + }; + double r = b1.distance(b2); + double func = 0.; + // function to integrate + if (reaSource) { + // Real part of the FLS solution + func += _erfint(double(b2.D - b1.D + b2.H) * s); + func += -_erfint(double(b2.D - b1.D) * s); + func += _erfint(double(b2.D - b1.D - b1.H) * s); + func += -_erfint(double(b2.D - b1.D + b2.H - b1.H) * s); + } // fi reaSource + if (imgSource) { + // Image part of the FLS solution + func += _erfint(double(b2.D + b1.D + b2.H) * s); + func += -_erfint(double(b2.D + b1.D) * s); + func += _erfint(double(b2.D + b1.D + b1.H) * s); + func += -_erfint(double(b2.D + b1.D + b2.H + b1.H) * s); + } // fi imgSource + double a = 0.5 / (b2.H * pow(s, 2)) * func * exp(-pow(r, 2) * pow(s, 2)); + return a; + }; // auto _Ils + + // lower bound of integration + double a = double(1.) / sqrt(double(4.) * alpha * time_); + // Evaluate the integral using Gauss-Kronrod + double result; + auto method = qdt::adaptive(qdt::gauss_kronrod()); + result = method.integrate(_Ils, a, qdt::INF); + + return result; + } // void finite_line_source + + void thermal_response_factors(SegmentResponse &SegRes, + vector &time, const double alpha, + bool use_similaries, bool disp) { + // total number of line sources + int nSources = SegRes.boreSegments.size(); + // number of time values + int nt = time.size(); + + // Open up processes here + // Create a vector of threads + //may return 0 when not able to detect + const auto processor_count = thread::hardware_concurrency(); + if (disp) { + cout << "\tDetected " << processor_count + << " as the number of available threads" << endl; + } + + gt::boreholes::SimilaritiesType SimReal; // positive + gt::boreholes::SimilaritiesType SimImage; // negative + + auto sum_to_n = [](const int n) { + return n * (n + 1) / 2; + }; + if (use_similaries) { + auto start = std::chrono::steady_clock::now(); + // Calculations with similarities + if (disp) { + cout << "Identifying similarities..." << endl; + } + bool splitRealAndImage = true; + double disTol = 0.1; + double tol = 1.0e-6; + gt::boreholes::Similarity sim; + sim.similarities(SimReal, SimImage, SegRes.boreSegments, + splitRealAndImage, disTol, tol); + + // --- + // Adaptive hashing scheme if statement + // Determine the Segment Response storing mode here + int Ntot = sum_to_n(nSources); + + // lambda function for calculating h at each time step + auto _calculate_h = [&SegRes, &splitRealAndImage, &time, &alpha, + &nt](boreholes::SimilaritiesType &SimReal, + int s, bool reaSource, + bool imgSource) { + // begin function + int n1; + int n2; + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + // begin thread + n1 = get<0>(SimReal.Sim[s][0]); + n2 = get<1>(SimReal.Sim[s][0]); + b1 = SegRes.boreSegments[n1]; + b2 = SegRes.boreSegments[n2]; + vector hPos(nt); + if (splitRealAndImage) { + for (int k=0; k(SimReal.Sim[s][k]); + j = get<1>(SimReal.Sim[s][k]); + for (std::size_t t=0; t(end - start).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " << seconds + << " sec" << std::endl; + std::cout << "Calculating segment to segment response " + "factors ..." << std::endl; + } // end if + + // inputs + bool reaSource; + bool imgSource; + # pragma omp parallel for num_threads(processor_count) + for (int s=0; s(end2 - end).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " + << seconds + << " sec" << std::endl; + } + } else { + if (disp) { + std::cout << "Calculating segment to segment response factors ..." << std::endl; + } // end if + auto start = std::chrono::steady_clock::now(); + bool sameSegment; + bool otherSegment; + +// auto _fill_line = [&h_ij, &time, &boreSegments](const int i, const int j, const double alpha, +// bool sameSegment, bool otherSegment) { +// auto _dot_product = [&h_ij, &time](const int i, const int j, const double constant) { +// for (std::size_t k=0; k < time.size(); k++) { +// h_ij[j][i][k+1] = constant * h_ij[i][j][k+1]; +// } // end for +// }; +// double h; +// double constant; +// gt::boreholes::Borehole b1; +// gt::boreholes::Borehole b2; +// b2 = boreSegments[i]; +// for (std::size_t k = 0; k < time.size(); k++) { +// double t = time[k]; +// if (!otherSegment){ +// if (sameSegment) { +// b1 = boreSegments[i]; +// h = finite_line_source(t, alpha, b2, b2); +// } +// } else if (otherSegment && !sameSegment) { +// b1 = boreSegments[j]; +// h = finite_line_source(t, alpha, b1, b2); +// } else { +// throw std::invalid_argument( "sameSegment and otherSegment cannot both be true" ); +// } // end if +// h_ij[i][j][k+1] = h; +// if (!sameSegment) { +// if (otherSegment) { +// constant = double(b2.H / b1.H); +// _dot_product(i, j, constant); +// } +// } // end if +// }; // end for +// }; // auto _fill_line + +// for (int i = 0; i < nSources; i++) { +// // Segment to same-segment thermal response factor +// // FLS solution for combined real and image sources +// sameSegment = true; +// otherSegment = false; +// boost::asio::post(pool, [i, alpha, sameSegment, otherSegment, &_fill_line] +// { _fill_line(i, i, alpha, true, false); }); +//// _fill_line(i, i, alpha, sameSegment, otherSegment); // could call with no threading during debugging +// +// // Segment to other segment thermal response factor +// for (int j = i + 1; j(end - start).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " + << seconds + << " sec" << std::endl; + } + // Iterate over the thread vector + } // fi similarity + } // void thermal_response_factors + + void SegmentResponse::get_h_value(double &h, const int i, const int j, const int k) { + int index; + switch (storage_mode) { + case 0 : + cout << "Case 0 not written yet" << endl; + break; + case 1 : + if (i <= j) { + get_index_value(index, i, j); + h = h_ij[index][k]; + } else { + get_index_value(index, j, i); + h = boreSegments[j].H/boreSegments[i].H * h_ij[index][k]; + } + break; + default: + throw invalid_argument("The case selected is not currently implemented."); + } // switch(); + } // SegmentResponse::get_h_value(); + + void SegmentResponse::get_index_value(int &index, const int i, const int j) { + index = i * (2*nSources - i - 1) / 2 + j; + } // SegmentResponse::get_index_value(); + +} // namespace gt::heat_transfer \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/src/interpolation.cpp b/third_party/cpgfunctionEP-0.1/src/interpolation.cpp new file mode 100644 index 00000000000..f48cf914976 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/interpolation.cpp @@ -0,0 +1,91 @@ +// +// Created by jackcook on 7/15/20. +// + +#include +#include + +using namespace std; + +namespace jcc { namespace interpolation { + + double linterp(double xp, double x0, double y0, double x1, double y1) { + double yp; + yp = y0 + ((y1-y0) / (x1-x0)) * (xp-x0); + return yp; + } // interp + + void interp1d(vector& xp, vector& yp, vector& x, vector& y) { + int counter = 0; + for (int i=0; i x[x.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + for (int j = counter; j= x[j] && xp[i] <= x[j+1]) { + yp[i] = linterp(xp[i], x[j], y[j], x[j + 1], y[j + 1]); + break; + } else { + counter++; + } // fi + } // next j + } // next i + } // interp1d + + void interp1d(double &xp, double &yp, vector& x, vector& y) { + int counter = 0; + + if (xp < x[0] || xp > x[x.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + for (int j = counter; j= x[j] && xp <= x[j+1]) { + yp = linterp(xp, x[j], y[j], x[j + 1], y[j + 1]); + return; + } else { + counter++; + } // fi + } // next j + } // interp1d + + void interp1d(double &xp, double &yp, vector &time, + gt::heat_transfer::SegmentResponse &SegRes, int &i, int &j, int &k) { + // if the x point is out of bounds, then tell the user that extrapolation is not possible + if (xp < 0 || xp > time[time.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + // if the time value falls in between 0 and the first time value, then interpolate 0 to t1 + if (0 < xp && xp < time[0]) { + double h; + SegRes.get_h_value(h, i, j, k); + yp = linterp(xp, 0, 0, time[0], h); + return; + } + // loop until the value for interpolation is found + int counter=0; + double h1; + SegRes.get_h_value(h1, i, j, 0); + double h2; + SegRes.get_h_value(h2, i, j, 1); + for (int k=counter; k=time[k] && xp <=time[k+1]) { + yp = linterp(xp, time[k], h1, time[k+1], h2); + return; + } else { + counter++; + h1 = h2; + SegRes.get_h_value(h2, i, j, counter+1); + } // else() + + } // next k + } // interp1d(); + +} } // jcc::interpolation + + diff --git a/third_party/cpgfunctionEP-0.1/src/statistics.cpp b/third_party/cpgfunctionEP-0.1/src/statistics.cpp new file mode 100644 index 00000000000..96320c13f71 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/statistics.cpp @@ -0,0 +1,25 @@ +// +// Created by jackcook on 5/10/21. +// + +#include + +namespace gt { + namespace statistics { + + double root_mean_square_error(std::vector actual, std::vector predicted) { + // Return the root mean square error between two g-function curves + + double square_error = 0; + + for (int i = 0; i < actual.size(); i++) { + square_error += std::pow(((actual[i] - predicted[i]) / actual[i]), 2); + } // next i + + double rmse = std::sqrt(square_error / double(actual.size())); + + return rmse; + } + + } // namespace statistics +} // namespace gt \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/src/utilities.cpp b/third_party/cpgfunctionEP-0.1/src/utilities.cpp new file mode 100644 index 00000000000..d331ad83bd9 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/src/utilities.cpp @@ -0,0 +1,85 @@ +// +// Created by jackcook on 7/11/20. +// + +#include + +namespace gt { + namespace utilities { + std::vector time_geometric(double dt, double tmax, int Nt) { + std::vector time(Nt); // create a time vector of size Nt + + double value; + double tmax_calc = double(Nt) * double(dt); + if (tmax > tmax_calc) { + double dr = 1.0e99; + double r = 2.; + while (std::abs(dr) > 1.0e-10) { + dr = std::pow(1+tmax/double(dt)*(r-1), 1/double(Nt)) - r; + r += dr; + } // end while + for (int j=0; j < time.size(); j++) { + value = 1 - std::pow(r, double(j+1)); + value = value * double(dt) / (1 - r); + time[j] = value; + } // end for + } else { + for (int j=0; j < time.size(); j++) { + value = double(dt) * double(j + 1); + time[j] = value; + } // end for + } // end if + return time; + } // vector time_geometric + + std::vector Eskilson_original_points() { + // Eskilsons original 27 time steps + std::vector logtime = {-8.5, -7.8, -7.2, -6.5, -5.9, -5.2, -4.5, -3.963, -3.27, -2.864,-2.577, + -2.171, -1.884, -1.191, -0.497, -0.274, -0.051, 0.196, 0.419, 0.642, 0.873, + 1.112, 1.335, 1.679, 2.028, 2.275, 3.003}; + return logtime; + + } + + std::vector time_Eskilson(const double &H, const double &alpha){ + std::vector logtime = Eskilson_original_points(); + std::vector time = convert_time(logtime, H, alpha); + + return time; + } // time_Eskilson(); + + std::vector convert_time(std::vector &logtime, const double &H, const double &alpha) { + int nt = logtime.size(); + std::vector time(nt); + + double ts = pow(H, 2) / (9 * alpha); + for (int i=0; i &logtime){ + int np = 31; // 31 total points + if (logtime.size() != np) { + logtime.resize(np); + } + for (int i=1; i &logtime, std::vector &time, const double H, const double alpha) { + int nt = logtime.size(); + if (time.size() != nt) { + time.resize(nt); + } + double ts = pow(H, 2) / (9 * alpha); + for (int i=0; i +#include + + +int main(){ + // -- Definitions -- + // Coordinate geometry + int Nx = 3; + int Ny = 3; + double Bx = 6.; + double By = 4.5; + + // -- Configurations -- + // Get x,y coordinates for a rectangle + std::string shape = "Rectangle"; + std::vector> coordinates = + gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + + // -- Borehole geometry -- + double H = 100; // height of the borehole (in meters) + double D = 4; // burial depth (in meters) + double r_b = 0.075; // borehole radius (in meters) + + // -- boreField -- + std::vector boreField = gt::boreholes::boreField(coordinates, r_b, H, D); + + return 0; +} diff --git a/third_party/cpgfunctionEP-0.1/test/compute_UBHWT_gFunction.cpp b/third_party/cpgfunctionEP-0.1/test/compute_UBHWT_gFunction.cpp new file mode 100644 index 00000000000..fdb6c261290 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/compute_UBHWT_gFunction.cpp @@ -0,0 +1,138 @@ +// +// Created by jackcook on 5/7/21. +// + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +void export_gFunction(std::string output_path, std::vector logtime, + std::vector gFunction) { + std::ofstream o(output_path); + + nlohmann::json j; + + j["logtime"] = logtime; + j["g"] = gFunction; + + o << std::setw(4) << j << std::endl; +} // export_gFunction(); + + +std::vector> import_gFunction(std::string input_path) { + std::vector> gFunction; // custom-shape from .json file + + // nlohmann json input + std::ifstream in(input_path); + nlohmann::json js; + in >> js; + + std::vector x = js["logtime"]; + std::vector y = js["g"]; + + gFunction.reserve(x.size()); + for (int i = 0; i < x.size(); i++){ + gFunction.emplace_back(x[i], y[i]); + } + + return gFunction; +} + + +int main(){ + // Toggle whether or not files are being created + bool creation = false; + + // -- Definitions -- + // Coordinate geometry + int Nx = 10; + int Ny = 10; + double Bx = 6.; + double By = 4.5; + + // -- Borehole geometry -- + double H = 100; // height of the borehole (in meters) + double D = 4; // burial depth (in meters) + double r_b = 0.075; // borehole radius (in meters) + + // Ground properties + double alpha = 1.0e-06; // ground thermal diffusivity + + // -- Time definition -- + // Eskilson's original 27 time steps (in seconds) + std::vector time = gt::utilities::time_Eskilson(H, alpha); + // Eskilson's original 27 time steps (logarithmic) + std::vector logtime = gt::utilities::Eskilson_original_points(); + + // -- Configurations -- + std::vector shapes{"Rectangle", "OpenRectangle", "U", "L", "custom"}; + std::string custom_path = "Poisson_Disk_120_30_101.json"; + + for (int i = 0; i < shapes.size(); i++) { + // Field shape + std::string shape = shapes[i]; + std::vector> coordinates; + // Define a path + std::string path = shape + ".json"; + if (shape != "custom") { + // Get x,y coordinates + coordinates = gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + } else { + coordinates = gt::coordinates::configuration(shape, custom_path); + } + + // Define borehole field + std::vector boreField = gt::boreholes::boreField(coordinates, r_b, H, D); + std::cout << "Compute g-function for borefield of shape: " + shape << std::endl; + auto start = std::chrono::steady_clock::now(); + // Compute uniform borehole wall temperature g-function + vector gFunction = + gt::gfunction::uniform_borehole_wall_temperature(boreField, + time, alpha, + 12, + true, + true); + auto end = std::chrono::steady_clock::now(); + auto milli = std::chrono::duration_cast(end - start).count(); + double seconds = double(milli) / 1000; + std::cout << "\ttime: "; + std::cout << seconds; + std::cout << " seconds" << std::endl; + + if (creation) { + export_gFunction(path, logtime, gFunction); + } else { + std::vector> gFunctionValues = import_gFunction(path); + std::vector gFunctionReference(gFunctionValues.size()); + // pull out the g-function value from the tuple + for (int j = 0; j < gFunctionValues.size(); j++) { + gFunctionReference[j] = std::get<1>(gFunctionValues[j]); + } // next i + double rmse = gt::statistics::root_mean_square_error(gFunctionReference, gFunction); + rmse *= 100; + + if (rmse > 1.0E-3) { + throw std::invalid_argument("The root mean square error is more than 0."); + } + + std::cout << "\trmse (%): "; + std::cout.precision(17); + std::cout << rmse; + std::cout << std::endl; + std::cout.precision(5); + } + } + + // A test for a custom configuration + + return 0; +} diff --git a/third_party/cpgfunctionEP-0.1/test/configurations.cpp b/third_party/cpgfunctionEP-0.1/test/configurations.cpp new file mode 100644 index 00000000000..14ed45bacbe --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/configurations.cpp @@ -0,0 +1,51 @@ +// +// Created by jackcook on 2/28/21. +// + +// tests for creating coordinates + +#include + +int main() { + // -- Definitions -- + // Coordinate geometry + int Nx = 3; + int Ny = 3; + double Bx = 6.; + double By = 4.5; + + // -- Configurations -- + // Get x,y coordinates for a rectangle + std::string shape = "Rectangle"; + std::vector> coordinates = + gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + std::string output_path = shape + "_configuration.json"; + gt::coordinates::export_coordinates_to_file(coordinates, output_path); + + // Get x,y coordinates for an open-rectangle + shape = "OpenRectangle"; + coordinates = gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + output_path = shape + "_configuration.json"; + gt::coordinates::export_coordinates_to_file(coordinates, output_path); + + // Get x,y coordinates for a U shape + shape = "U"; + coordinates = gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + output_path = shape + "_configuration.json"; + gt::coordinates::export_coordinates_to_file(coordinates, output_path); + + // Get x,y coordinates for a U shape + shape = "L"; + coordinates = gt::coordinates::configuration(shape, Nx, Ny, Bx, By); + output_path = shape + "_configuration.json"; + gt::coordinates::export_coordinates_to_file(coordinates, output_path); + + // Read in custom configuration from (.json) file + shape = "custom"; + std::string input_path = "U_configuration.json"; // read in the previously exported U + coordinates = gt::coordinates::configuration(shape, input_path); + output_path = shape + "_configuration.json"; + gt::coordinates::export_coordinates_to_file(coordinates, output_path); + + return 0; +} \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/test/custom_configurations/Poisson_Disk_120_30_101.json b/third_party/cpgfunctionEP-0.1/test/custom_configurations/Poisson_Disk_120_30_101.json new file mode 100644 index 00000000000..1116ae02f8a --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/custom_configurations/Poisson_Disk_120_30_101.json @@ -0,0 +1,208 @@ +{ + "x": [ + 82.53830257954516, + 81.87070593878656, + 78.58218265923767, + 70.40664179403504, + 73.77564334480255, + 75.61587205486207, + 71.57274334582024, + 85.63826505674038, + 68.43524190194671, + 91.35627219283978, + 65.03504382864351, + 79.17704150414791, + 88.76650886916337, + 63.299486308930476, + 56.950258121128904, + 84.68854842256398, + 91.20209150048643, + 99.24557904094948, + 96.8147202327192, + 85.7200937717854, + 67.56145828482917, + 95.05729494204986, + 50.82071533270727, + 67.47079409675368, + 107.11408092064968, + 77.01994458218645, + 56.351047926499845, + 100.6539723529873, + 94.091201080575, + 60.24457238575478, + 103.87315284447172, + 52.15192259075156, + 106.09836539345163, + 56.30773092817518, + 99.60166880635227, + 104.63125089570325, + 65.6800406775613, + 51.86536435043161, + 79.31004871963378, + 60.64294400711204, + 102.75836187012716, + 90.0783289851516, + 109.3302764456526, + 98.8561313464538, + 47.77704200885893, + 61.252895791436046, + 72.66038543062956, + 54.52664365159561, + 112.546428455935, + 114.98487710788912, + 80.54609526328501, + 90.09644317231545, + 39.09081213223031, + 46.686003498072864, + 117.21459677056723, + 111.18483042989034, + 117.18301697684532, + 43.91026574205974, + 111.47630832169632, + 42.95385892838383, + 117.48345447101605, + 36.76272500396449, + 47.510208514608266, + 40.69665821587505, + 40.660662917952436, + 117.22486690990571, + 34.43056348688545, + 33.67674149490491, + 84.79749191413725, + 34.44297140821213, + 30.66204645293468, + 113.13762997562056, + 26.825331773060004, + 26.224000400668768, + 28.33033370272318, + 27.11501115770472, + 21.619667058139786, + 14.170668651531908, + 30.741376715045053, + 20.60304999486002, + 7.713332945412799, + 39.683078101882344, + 18.598771117738444, + 23.37180821034542, + 17.28363599611696, + 21.67721240932794, + 10.120152824367713, + 22.547061623520307, + 10.08332226163904, + 14.67549856343755, + 1.4449735794283711, + 2.2907824871296305, + 14.945746756853808, + 0.3916478455688139, + 0.5848249323707975, + 1.3368226032780255, + 8.787351781898822, + 8.650007441421549, + 2.5265361762391647, + 16.415215981328746, + 11.134286712117683 + ], + "y": [ + 10.000105567082112, + 0.2433723101509297, + 13.53914540956573, + 12.803186036901195, + 5.966573718408955, + 0.984313869394854, + 20.50694077071949, + 5.060298877753024, + 2.8969744202921635, + 1.3910388109934322, + 19.092209279347287, + 5.415338509815117, + 9.453508511925266, + 11.874412450400095, + 18.427229965525076, + 14.538881653213538, + 14.593924740501231, + 2.6856051998287143, + 8.759341104784731, + 22.568750327891802, + 24.587902886522212, + 20.569006597580728, + 23.0460915249237, + 8.194046700114004, + 1.4137976263762286, + 22.588781392392868, + 5.8961087405068495, + 12.631108226922409, + 26.46737171630396, + 2.131092969107664, + 25.42979826253254, + 2.7758765871147872, + 11.139488601596081, + 13.038468001309008, + 17.967490737879178, + 19.62484314646731, + 29.93715575470231, + 8.661216189128947, + 28.69496913308712, + 29.173522296023144, + 6.407700892823862, + 19.51269273351043, + 15.61088035056052, + 28.217155001296128, + 17.539202284506818, + 23.301405224514085, + 25.49806095054708, + 26.59713265068969, + 5.209530413496177, + 14.632619934447494, + 18.54153939761814, + 29.484723743372804, + 16.382255579659223, + 29.578790986383773, + 7.088003551538951, + 24.033899402847968, + 22.701628031195305, + 10.810882349439215, + 29.032749649107576, + 22.611754353840873, + 0.5165999417896687, + 22.216888964686085, + 4.763395815415525, + 28.243875060226554, + 2.078129610329672, + 28.35775807053709, + 27.329203186879763, + 7.563693584237157, + 29.495612006219574, + 13.497753838647338, + 20.46309934877686, + 19.334092602057506, + 26.716892587827893, + 17.516974724646563, + 12.725189046668156, + 7.3875004829982425, + 6.2213508084175775, + 6.282618122695393, + 3.2777832922852563, + 11.748498871918153, + 6.378705113938347, + 7.5147859766638785, + 17.392290676628896, + 1.3458634016582471, + 1.0227018640403598, + 29.690076596827605, + 15.575859260423849, + 21.14127167420851, + 0.6225461253923629, + 22.643661051495037, + 12.864037428685322, + 6.654861524266177, + 13.957063337852233, + 1.335535392641794, + 18.802353759560408, + 23.868543596583663, + 28.659784010947934, + 20.885775814123846, + 29.6455367854156, + 29.80533023017508, + 10.501603154205998 + ] +} \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/test/gFunction_minimal.cpp b/third_party/cpgfunctionEP-0.1/test/gFunction_minimal.cpp new file mode 100644 index 00000000000..a874226f6ae --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/gFunction_minimal.cpp @@ -0,0 +1,59 @@ +#include +#include +#include +#include +#include +#include +#include + + +int main() { + // --------------------------------------------------------- + // Simulation parameters + // --------------------------------------------------------- + + // Borehole dimensions + float D = 4.0; // Borehole buried depth (m) + float H = 100.; // Borehole length (m) + float r_b = 0.075; // Borehole radius (m) + float B = 7.5; // Borehole spacing (m) + + // Thermal properties + double alpha = 1.0e-6; // Ground thermal diffusivity (m2/s) + + // Number of segments per borehole + int nSegments = 2; + + // Geometrically expanding time vector. + float dt = 100*3600.; // Time step + double tmax = 3000. * 8760. * 3600.; // Maximum time + int Nt = 6; // Number of time steps + double ts = pow(H, 2)/(9.*alpha); // Bore field characteristic time + + std::vector time = gt::utilities::time_geometric(dt, tmax, Nt); + + // --------------------------------------------------------- + // Borehole fields + // --------------------------------------------------------- + + // Field of 2x3 (n=6) boreholes + int n1 = 2; + int n2 = 2; + // Coordinates + std::vector> coordinates = gt::coordinates::rectangle(n1, n2, B, B); + // Move boreholes slightly off of symmetric + std::get<0>(coordinates[1]) += 2; // move position 1 borehole x + std::get<0>(coordinates[2]) -= 1; // move position 2 borehole x + std::get<1>(coordinates[2]) -= 1; // move position 2 borehole y + std::get<1>(coordinates[3]) -= 1; // move position 3 borehole y + // Create borehole field + std::vector boreField = gt::boreholes::boreField(coordinates, r_b, H, D); + + std::vector gFunction = gt::gfunction::uniform_borehole_wall_temperature(boreField, time, alpha, nSegments,true, true); + + for (double i : gFunction) { + std::cout << i << std::endl; + } + + return 0; +} diff --git a/third_party/cpgfunctionEP-0.1/test/interpolation.cpp b/third_party/cpgfunctionEP-0.1/test/interpolation.cpp new file mode 100644 index 00000000000..075cb07ffd0 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/interpolation.cpp @@ -0,0 +1,23 @@ +// +// Created by jackcook on 2/27/21. +// + +#include +#include +#include + +int main() { + // 1D interpolation test + std::vector x {0, 200, 400, 600}; + std::vector y {373.0, 156.1, 113.6, 93.1}; + std::vector xp {90, 210, 310}; + std::vector yp (3); + jcc::interpolation::interp1d(xp, yp, x, y); + + std::cout << "The y points for linear interpolation" << std::endl; + for (const auto& i: yp) { + std:: cout << i << std::endl; + } + + return 0; +} \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/test/packed.cpp b/third_party/cpgfunctionEP-0.1/test/packed.cpp new file mode 100644 index 00000000000..98d16dfe78f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/packed.cpp @@ -0,0 +1,28 @@ +// +// Created by jackcook on 6/28/21. +// + +#include +#include + +auto sum_to_n = [](const int n) { + return n * (n + 1) / 2; +}; + +int main(){ + int nSources = 5; + int nSum = sum_to_n(nSources); + int nt = 5; + + gt::heat_transfer::SegmentResponse SegRes(nSources, nSum, nt); + + int index; + for (int i=0; i<5; i++) { + for (int j=i; j<5; j++) { + SegRes.get_index_value(index, i, j); + std::cout << i << "\t" << j << "\t" << index << std::endl; + } + } + + return 0; +} \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/test/time_definition.cpp b/third_party/cpgfunctionEP-0.1/test/time_definition.cpp new file mode 100644 index 00000000000..71284ce03df --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/time_definition.cpp @@ -0,0 +1,23 @@ +// +// Created by jackcook on 5/7/21. +// + +#include + + +int main() { + // -- Time definitions -- + // Geometrically expanding time vector + double dt = 100 * 3600.; // time step + int duration = 36; // duration in months + double tmax = double(duration) / 3 * 8760. * 3600.; + int Nt = 10; // number of time steps + std::vector time = gt::utilities::time_geometric(dt, tmax, Nt); + + // Eskilson's original 27 time steps + double H = 100; // height of the boreholes + double alpha = 1.0e-06; // thermal diffusivity + time = gt::utilities::time_Eskilson(H, alpha); + + return 0; +} \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.1/test/validation/L.json b/third_party/cpgfunctionEP-0.1/test/validation/L.json new file mode 100644 index 00000000000..1901db3c139 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/validation/L.json @@ -0,0 +1,60 @@ +{ + "g": [ + 2.24977204749898, + 2.5959317999870635, + 2.8926661310589505, + 3.2438595465527005, + 3.575314245755779, + 4.063380847752393, + 4.732950537825905, + 5.412183691395843, + 6.5555412103892055, + 7.390856496702103, + 8.062671332947247, + 9.128454066781998, + 9.964339347890093, + 12.19328882441035, + 14.548850387738256, + 15.27996956985495, + 15.965728101312537, + 16.660592242566672, + 17.21602210596885, + 17.70225419557096, + 18.12796921804031, + 18.48852164958685, + 18.760130892728313, + 19.071817294093368, + 19.286654236358398, + 19.39307473902638, + 19.567196885090237 + ], + "logtime": [ + -8.5, + -7.8, + -7.2, + -6.5, + -5.9, + -5.2, + -4.5, + -3.963, + -3.27, + -2.864, + -2.577, + -2.171, + -1.884, + -1.191, + -0.497, + -0.274, + -0.051, + 0.196, + 0.419, + 0.642, + 0.873, + 1.112, + 1.335, + 1.679, + 2.028, + 2.275, + 3.003 + ] +} diff --git a/third_party/cpgfunctionEP-0.1/test/validation/OpenRectangle.json b/third_party/cpgfunctionEP-0.1/test/validation/OpenRectangle.json new file mode 100644 index 00000000000..5770bf8271e --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/validation/OpenRectangle.json @@ -0,0 +1,60 @@ +{ + "g": [ + 2.2497720474991847, + 2.595931832104826, + 2.8926745138412797, + 3.2442057136676556, + 3.5778145357455715, + 4.0759296748887905, + 4.774743417675502, + 5.501680312073553, + 6.768069453827016, + 7.728558788249318, + 8.525142081636794, + 9.84281751470179, + 10.930151667680043, + 14.107429738242411, + 17.83709039846818, + 19.04871716419615, + 20.191880505435726, + 21.351950059981075, + 22.27523542207011, + 23.078257698674122, + 23.774334353765784, + 24.357157518186934, + 24.79248294013983, + 25.286316979803633, + 25.624760852813843, + 25.792325747743785, + 26.067477653778386 + ], + "logtime": [ + -8.5, + -7.8, + -7.2, + -6.5, + -5.9, + -5.2, + -4.5, + -3.963, + -3.27, + -2.864, + -2.577, + -2.171, + -1.884, + -1.191, + -0.497, + -0.274, + -0.051, + 0.196, + 0.419, + 0.642, + 0.873, + 1.112, + 1.335, + 1.679, + 2.028, + 2.275, + 3.003 + ] +} diff --git a/third_party/cpgfunctionEP-0.1/test/validation/Rectangle.json b/third_party/cpgfunctionEP-0.1/test/validation/Rectangle.json new file mode 100644 index 00000000000..771e765445a --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/validation/Rectangle.json @@ -0,0 +1,60 @@ +{ + "g": [ + 2.249772047502418, + 2.5959323202948426, + 2.8928019336498303, + 3.249482305161263, + 3.6169412198762445, + 4.28674400929944, + 5.517589963743413, + 7.114917627775926, + 10.471588887585417, + 13.360038045992797, + 15.887615509894198, + 20.156571764794197, + 23.668374529488883, + 33.23748892365322, + 43.222845944566025, + 46.246536600751135, + 49.01363239132643, + 51.74641922470241, + 53.854970449932665, + 55.66288248348939, + 57.19961703669294, + 58.46644012437772, + 59.40724316636071, + 60.46804616919918, + 61.19426853275642, + 61.554647393983906, + 62.15244294767487 + ], + "logtime": [ + -8.5, + -7.8, + -7.2, + -6.5, + -5.9, + -5.2, + -4.5, + -3.963, + -3.27, + -2.864, + -2.577, + -2.171, + -1.884, + -1.191, + -0.497, + -0.274, + -0.051, + 0.196, + 0.419, + 0.642, + 0.873, + 1.112, + 1.335, + 1.679, + 2.028, + 2.275, + 3.003 + ] +} diff --git a/third_party/cpgfunctionEP-0.1/test/validation/U.json b/third_party/cpgfunctionEP-0.1/test/validation/U.json new file mode 100644 index 00000000000..1eb0122d141 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/validation/U.json @@ -0,0 +1,60 @@ +{ + "g": [ + 2.2497720475003415, + 2.5959320064184213, + 2.892719494888639, + 3.2458451962033648, + 3.5864058984434415, + 4.098855790548383, + 4.808982307327865, + 5.534637675700166, + 6.768549325376686, + 7.67976058176301, + 8.419036345832929, + 9.605582548367142, + 10.549765972114066, + 13.158171181450776, + 16.10095147988399, + 17.054369469028504, + 17.96110545322208, + 18.88982407476407, + 19.63730612231681, + 20.29360048893262, + 20.86833113777573, + 21.354184882019617, + 21.71942630171469, + 22.13678596922885, + 22.423992498801336, + 22.566363543286045, + 22.799833176206246 + ], + "logtime": [ + -8.5, + -7.8, + -7.2, + -6.5, + -5.9, + -5.2, + -4.5, + -3.963, + -3.27, + -2.864, + -2.577, + -2.171, + -1.884, + -1.191, + -0.497, + -0.274, + -0.051, + 0.196, + 0.419, + 0.642, + 0.873, + 1.112, + 1.335, + 1.679, + 2.028, + 2.275, + 3.003 + ] +} diff --git a/third_party/cpgfunctionEP-0.1/test/validation/custom.json b/third_party/cpgfunctionEP-0.1/test/validation/custom.json new file mode 100644 index 00000000000..a36ce8398c7 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/test/validation/custom.json @@ -0,0 +1,60 @@ +{ + "g": [ + 2.24977204749518, + 2.595931237286799, + 2.8925347519050164, + 3.2401372599791403, + 3.5627191681590284, + 4.07693414086791, + 4.954537104890203, + 6.069862517433721, + 8.40660843978132, + 10.4222035403255, + 12.193523658544342, + 15.208155891437102, + 17.708489708963338, + 24.6663638120643, + 32.23331471292215, + 34.62714321747046, + 36.863827964418064, + 39.12668205067839, + 40.921714306572376, + 42.493845673922216, + 43.85902042782804, + 45.00527796638447, + 45.86779712837001, + 46.85109277332042, + 47.53087844737528, + 47.87096617108988, + 48.439969465560964 + ], + "logtime": [ + -8.5, + -7.8, + -7.2, + -6.5, + -5.9, + -5.2, + -4.5, + -3.963, + -3.27, + -2.864, + -2.577, + -2.171, + -1.884, + -1.191, + -0.497, + -0.274, + -0.051, + 0.196, + 0.419, + 0.642, + 0.873, + 1.112, + 1.335, + 1.679, + 2.028, + 2.275, + 3.003 + ] +} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.gitignore b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.gitignore new file mode 100644 index 00000000000..fc0e5486c41 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.gitignore @@ -0,0 +1,37 @@ +qrc_*cxx +*.orig +*.pyc +*.diff +diff +*.save +save +*.old +*.gmo +*.qm +core +core.* +*.bak +*~ +*build* +*.moc.* +*.moc +ui_* +CMakeCache.txt +tags +.*.swp +activity.png +*.out +*.php* +*.log +*.orig +*.rej +log +patch +*.patch +a +a.* +lapack/testing +lapack/reference +.*project +.settings +Makefile diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.hgeol b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.hgeol new file mode 100644 index 00000000000..5327df16153 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/.hgeol @@ -0,0 +1,11 @@ +[patterns] +*.sh = LF +*.MINPACK = CRLF +scripts/*.in = LF +debug/msvc/*.dat = CRLF +debug/msvc/*.natvis = CRLF +unsupported/test/mpreal/*.* = CRLF +** = native + +[repository] +native = LF diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CMakeLists.txt b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CMakeLists.txt new file mode 100644 index 00000000000..9eb502b792d --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CMakeLists.txt @@ -0,0 +1,19 @@ +include(RegexUtils) +test_escape_string_as_regex() + +file(GLOB Eigen_directory_files "*") + +escape_string_as_regex(ESCAPED_CMAKE_CURRENT_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}") + +foreach(f ${Eigen_directory_files}) + if(NOT f MATCHES "\\.txt" AND NOT f MATCHES "${ESCAPED_CMAKE_CURRENT_SOURCE_DIR}/[.].+" AND NOT f MATCHES "${ESCAPED_CMAKE_CURRENT_SOURCE_DIR}/src") + list(APPEND Eigen_directory_files_to_install ${f}) + endif() +endforeach(f ${Eigen_directory_files}) + +install(FILES + ${Eigen_directory_files_to_install} + DESTINATION ${INCLUDE_INSTALL_DIR}/Eigen COMPONENT Devel + ) + +install(DIRECTORY src DESTINATION ${INCLUDE_INSTALL_DIR}/Eigen COMPONENT Devel FILES_MATCHING PATTERN "*.h") diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Cholesky b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Cholesky new file mode 100644 index 00000000000..1332b540d83 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Cholesky @@ -0,0 +1,46 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CHOLESKY_MODULE_H +#define EIGEN_CHOLESKY_MODULE_H + +#include "Core" +#include "Jacobi" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Cholesky_Module Cholesky module + * + * + * + * This module provides two variants of the Cholesky decomposition for selfadjoint (hermitian) matrices. + * Those decompositions are also accessible via the following methods: + * - MatrixBase::llt() + * - MatrixBase::ldlt() + * - SelfAdjointView::llt() + * - SelfAdjointView::ldlt() + * + * \code + * #include + * \endcode + */ + +#include "src/Cholesky/LLT.h" +#include "src/Cholesky/LDLT.h" +#ifdef EIGEN_USE_LAPACKE +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/Cholesky/LLT_LAPACKE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_CHOLESKY_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CholmodSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CholmodSupport new file mode 100644 index 00000000000..bed8924d31e --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/CholmodSupport @@ -0,0 +1,48 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CHOLMODSUPPORT_MODULE_H +#define EIGEN_CHOLMODSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +extern "C" { + #include +} + +/** \ingroup Support_modules + * \defgroup CholmodSupport_Module CholmodSupport module + * + * This module provides an interface to the Cholmod library which is part of the suitesparse package. + * It provides the two following main factorization classes: + * - class CholmodSupernodalLLT: a supernodal LLT Cholesky factorization. + * - class CholmodDecomposiiton: a general L(D)LT Cholesky factorization with automatic or explicit runtime selection of the underlying factorization method (supernodal or simplicial). + * + * For the sake of completeness, this module also propose the two following classes: + * - class CholmodSimplicialLLT + * - class CholmodSimplicialLDLT + * Note that these classes does not bring any particular advantage compared to the built-in + * SimplicialLLT and SimplicialLDLT factorization classes. + * + * \code + * #include + * \endcode + * + * In order to use this module, the cholmod headers must be accessible from the include paths, and your binary must be linked to the cholmod library and its dependencies. + * The dependencies depend on how cholmod has been compiled. + * For a cmake based project, you can use our FindCholmod.cmake module to help you in this task. + * + */ + +#include "src/CholmodSupport/CholmodSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_CHOLMODSUPPORT_MODULE_H + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Core b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Core new file mode 100644 index 00000000000..ac7c5b3004f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Core @@ -0,0 +1,542 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2007-2011 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CORE_H +#define EIGEN_CORE_H + +// first thing Eigen does: stop the compiler from committing suicide +#include "src/Core/util/DisableStupidWarnings.h" + +#if defined(__CUDACC__) && !defined(EIGEN_NO_CUDA) + #define EIGEN_CUDACC __CUDACC__ +#endif + +#if defined(__CUDA_ARCH__) && !defined(EIGEN_NO_CUDA) + #define EIGEN_CUDA_ARCH __CUDA_ARCH__ +#endif + +#if defined(__CUDACC_VER_MAJOR__) && (__CUDACC_VER_MAJOR__ >= 9) +#define EIGEN_CUDACC_VER ((__CUDACC_VER_MAJOR__ * 10000) + (__CUDACC_VER_MINOR__ * 100)) +#elif defined(__CUDACC_VER__) +#define EIGEN_CUDACC_VER __CUDACC_VER__ +#else +#define EIGEN_CUDACC_VER 0 +#endif + +// Handle NVCC/CUDA/SYCL +#if defined(__CUDACC__) || defined(__SYCL_DEVICE_ONLY__) + // Do not try asserts on CUDA and SYCL! + #ifndef EIGEN_NO_DEBUG + #define EIGEN_NO_DEBUG + #endif + + #ifdef EIGEN_INTERNAL_DEBUGGING + #undef EIGEN_INTERNAL_DEBUGGING + #endif + + #ifdef EIGEN_EXCEPTIONS + #undef EIGEN_EXCEPTIONS + #endif + + // All functions callable from CUDA code must be qualified with __device__ + #ifdef __CUDACC__ + // Do not try to vectorize on CUDA and SYCL! + #ifndef EIGEN_DONT_VECTORIZE + #define EIGEN_DONT_VECTORIZE + #endif + + #define EIGEN_DEVICE_FUNC __host__ __device__ + // We need cuda_runtime.h to ensure that that EIGEN_USING_STD_MATH macro + // works properly on the device side + #include + #else + #define EIGEN_DEVICE_FUNC + #endif + +#else + #define EIGEN_DEVICE_FUNC + +#endif + +// When compiling CUDA device code with NVCC, pull in math functions from the +// global namespace. In host mode, and when device doee with clang, use the +// std versions. +#if defined(__CUDA_ARCH__) && defined(__NVCC__) + #define EIGEN_USING_STD_MATH(FUNC) using ::FUNC; +#else + #define EIGEN_USING_STD_MATH(FUNC) using std::FUNC; +#endif + +#if (defined(_CPPUNWIND) || defined(__EXCEPTIONS)) && !defined(__CUDA_ARCH__) && !defined(EIGEN_EXCEPTIONS) && !defined(EIGEN_USE_SYCL) + #define EIGEN_EXCEPTIONS +#endif + +#ifdef EIGEN_EXCEPTIONS + #include +#endif + +// then include this file where all our macros are defined. It's really important to do it first because +// it's where we do all the alignment settings (platform detection and honoring the user's will if he +// defined e.g. EIGEN_DONT_ALIGN) so it needs to be done before we do anything with vectorization. +#include "src/Core/util/Macros.h" + +// Disable the ipa-cp-clone optimization flag with MinGW 6.x or newer (enabled by default with -O3) +// See http://eigen.tuxfamily.org/bz/show_bug.cgi?id=556 for details. +#if EIGEN_COMP_MINGW && EIGEN_GNUC_AT_LEAST(4,6) + #pragma GCC optimize ("-fno-ipa-cp-clone") +#endif + +#include + +// this include file manages BLAS and MKL related macros +// and inclusion of their respective header files +#include "src/Core/util/MKL_support.h" + +// if alignment is disabled, then disable vectorization. Note: EIGEN_MAX_ALIGN_BYTES is the proper check, it takes into +// account both the user's will (EIGEN_MAX_ALIGN_BYTES,EIGEN_DONT_ALIGN) and our own platform checks +#if EIGEN_MAX_ALIGN_BYTES==0 + #ifndef EIGEN_DONT_VECTORIZE + #define EIGEN_DONT_VECTORIZE + #endif +#endif + +#if EIGEN_COMP_MSVC + #include // for _aligned_malloc -- need it regardless of whether vectorization is enabled + #if (EIGEN_COMP_MSVC >= 1500) // 2008 or later + // Remember that usage of defined() in a #define is undefined by the standard. + // a user reported that in 64-bit mode, MSVC doesn't care to define _M_IX86_FP. + #if (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || EIGEN_ARCH_x86_64 + #define EIGEN_SSE2_ON_MSVC_2008_OR_LATER + #endif + #endif +#else + // Remember that usage of defined() in a #define is undefined by the standard + #if (defined __SSE2__) && ( (!EIGEN_COMP_GNUC) || EIGEN_COMP_ICC || EIGEN_GNUC_AT_LEAST(4,2) ) + #define EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC + #endif +#endif + +#ifndef EIGEN_DONT_VECTORIZE + + #if defined (EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC) || defined(EIGEN_SSE2_ON_MSVC_2008_OR_LATER) + + // Defines symbols for compile-time detection of which instructions are + // used. + // EIGEN_VECTORIZE_YY is defined if and only if the instruction set YY is used + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_SSE + #define EIGEN_VECTORIZE_SSE2 + + // Detect sse3/ssse3/sse4: + // gcc and icc defines __SSE3__, ... + // there is no way to know about this on msvc. You can define EIGEN_VECTORIZE_SSE* if you + // want to force the use of those instructions with msvc. + #ifdef __SSE3__ + #define EIGEN_VECTORIZE_SSE3 + #endif + #ifdef __SSSE3__ + #define EIGEN_VECTORIZE_SSSE3 + #endif + #ifdef __SSE4_1__ + #define EIGEN_VECTORIZE_SSE4_1 + #endif + #ifdef __SSE4_2__ + #define EIGEN_VECTORIZE_SSE4_2 + #endif + #ifdef __AVX__ + #define EIGEN_VECTORIZE_AVX + #define EIGEN_VECTORIZE_SSE3 + #define EIGEN_VECTORIZE_SSSE3 + #define EIGEN_VECTORIZE_SSE4_1 + #define EIGEN_VECTORIZE_SSE4_2 + #endif + #ifdef __AVX2__ + #define EIGEN_VECTORIZE_AVX2 + #endif + #ifdef __FMA__ + #define EIGEN_VECTORIZE_FMA + #endif + #if defined(__AVX512F__) && defined(EIGEN_ENABLE_AVX512) + #define EIGEN_VECTORIZE_AVX512 + #define EIGEN_VECTORIZE_AVX2 + #define EIGEN_VECTORIZE_AVX + #define EIGEN_VECTORIZE_FMA + #ifdef __AVX512DQ__ + #define EIGEN_VECTORIZE_AVX512DQ + #endif + #ifdef __AVX512ER__ + #define EIGEN_VECTORIZE_AVX512ER + #endif + #endif + + // include files + + // This extern "C" works around a MINGW-w64 compilation issue + // https://sourceforge.net/tracker/index.php?func=detail&aid=3018394&group_id=202880&atid=983354 + // In essence, intrin.h is included by windows.h and also declares intrinsics (just as emmintrin.h etc. below do). + // However, intrin.h uses an extern "C" declaration, and g++ thus complains of duplicate declarations + // with conflicting linkage. The linkage for intrinsics doesn't matter, but at that stage the compiler doesn't know; + // so, to avoid compile errors when windows.h is included after Eigen/Core, ensure intrinsics are extern "C" here too. + // notice that since these are C headers, the extern "C" is theoretically needed anyways. + extern "C" { + // In theory we should only include immintrin.h and not the other *mmintrin.h header files directly. + // Doing so triggers some issues with ICC. However old gcc versions seems to not have this file, thus: + #if EIGEN_COMP_ICC >= 1110 + #include + #else + #include + #include + #include + #ifdef EIGEN_VECTORIZE_SSE3 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSSE3 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSE4_1 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSE4_2 + #include + #endif + #if defined(EIGEN_VECTORIZE_AVX) || defined(EIGEN_VECTORIZE_AVX512) + #include + #endif + #endif + } // end extern "C" + #elif defined __VSX__ + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_VSX + #include + // We need to #undef all these ugly tokens defined in + // => use __vector instead of vector + #undef bool + #undef vector + #undef pixel + #elif defined __ALTIVEC__ + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_ALTIVEC + #include + // We need to #undef all these ugly tokens defined in + // => use __vector instead of vector + #undef bool + #undef vector + #undef pixel + #elif (defined __ARM_NEON) || (defined __ARM_NEON__) + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_NEON + #include + #elif (defined __s390x__ && defined __VEC__) + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_ZVECTOR + #include + #endif +#endif + +#if defined(__F16C__) && !defined(EIGEN_COMP_CLANG) + // We can use the optimized fp16 to float and float to fp16 conversion routines + #define EIGEN_HAS_FP16_C +#endif + +#if defined __CUDACC__ + #define EIGEN_VECTORIZE_CUDA + #include + #if EIGEN_CUDACC_VER >= 70500 + #define EIGEN_HAS_CUDA_FP16 + #endif +#endif + +#if defined EIGEN_HAS_CUDA_FP16 + #include + #include +#endif + +#if (defined _OPENMP) && (!defined EIGEN_DONT_PARALLELIZE) + #define EIGEN_HAS_OPENMP +#endif + +#ifdef EIGEN_HAS_OPENMP +#include +#endif + +// MSVC for windows mobile does not have the errno.h file +#if !(EIGEN_COMP_MSVC && EIGEN_OS_WINCE) && !EIGEN_COMP_ARM +#define EIGEN_HAS_ERRNO +#endif + +#ifdef EIGEN_HAS_ERRNO +#include +#endif +#include +#include +#include +#include +#include +#include +#ifndef EIGEN_NO_IO + #include +#endif +#include +#include +#include +#include // for CHAR_BIT +// for min/max: +#include + +// for std::is_nothrow_move_assignable +#ifdef EIGEN_INCLUDE_TYPE_TRAITS +#include +#endif + +// for outputting debug info +#ifdef EIGEN_DEBUG_ASSIGN +#include +#endif + +// required for __cpuid, needs to be included after cmath +#if EIGEN_COMP_MSVC && EIGEN_ARCH_i386_OR_x86_64 && !EIGEN_OS_WINCE + #include +#endif + +/** \brief Namespace containing all symbols from the %Eigen library. */ +namespace Eigen { + +inline static const char *SimdInstructionSetsInUse(void) { +#if defined(EIGEN_VECTORIZE_AVX512) + return "AVX512, FMA, AVX2, AVX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_AVX) + return "AVX SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_SSE4_2) + return "SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_SSE4_1) + return "SSE, SSE2, SSE3, SSSE3, SSE4.1"; +#elif defined(EIGEN_VECTORIZE_SSSE3) + return "SSE, SSE2, SSE3, SSSE3"; +#elif defined(EIGEN_VECTORIZE_SSE3) + return "SSE, SSE2, SSE3"; +#elif defined(EIGEN_VECTORIZE_SSE2) + return "SSE, SSE2"; +#elif defined(EIGEN_VECTORIZE_ALTIVEC) + return "AltiVec"; +#elif defined(EIGEN_VECTORIZE_VSX) + return "VSX"; +#elif defined(EIGEN_VECTORIZE_NEON) + return "ARM NEON"; +#elif defined(EIGEN_VECTORIZE_ZVECTOR) + return "S390X ZVECTOR"; +#else + return "None"; +#endif +} + +} // end namespace Eigen + +#if defined EIGEN2_SUPPORT_STAGE40_FULL_EIGEN3_STRICTNESS || defined EIGEN2_SUPPORT_STAGE30_FULL_EIGEN3_API || defined EIGEN2_SUPPORT_STAGE20_RESOLVE_API_CONFLICTS || defined EIGEN2_SUPPORT_STAGE10_FULL_EIGEN2_API || defined EIGEN2_SUPPORT +// This will generate an error message: +#error Eigen2-support is only available up to version 3.2. Please go to "http://eigen.tuxfamily.org/index.php?title=Eigen2" for further information +#endif + +namespace Eigen { + +// we use size_t frequently and we'll never remember to prepend it with std:: everytime just to +// ensure QNX/QCC support +using std::size_t; +// gcc 4.6.0 wants std:: for ptrdiff_t +using std::ptrdiff_t; + +} + +/** \defgroup Core_Module Core module + * This is the main module of Eigen providing dense matrix and vector support + * (both fixed and dynamic size) with all the features corresponding to a BLAS library + * and much more... + * + * \code + * #include + * \endcode + */ + +#include "src/Core/util/Constants.h" +#include "src/Core/util/Meta.h" +#include "src/Core/util/ForwardDeclarations.h" +#include "src/Core/util/StaticAssert.h" +#include "src/Core/util/XprHelper.h" +#include "src/Core/util/Memory.h" + +#include "src/Core/NumTraits.h" +#include "src/Core/MathFunctions.h" +#include "src/Core/GenericPacketMath.h" +#include "src/Core/MathFunctionsImpl.h" +#include "src/Core/arch/Default/ConjHelper.h" + +#if defined EIGEN_VECTORIZE_AVX512 + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/AVX/PacketMath.h" + #include "src/Core/arch/AVX/MathFunctions.h" + #include "src/Core/arch/AVX512/PacketMath.h" + #include "src/Core/arch/AVX512/MathFunctions.h" +#elif defined EIGEN_VECTORIZE_AVX + // Use AVX for floats and doubles, SSE for integers + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/Complex.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/AVX/PacketMath.h" + #include "src/Core/arch/AVX/MathFunctions.h" + #include "src/Core/arch/AVX/Complex.h" + #include "src/Core/arch/AVX/TypeCasting.h" + #include "src/Core/arch/SSE/TypeCasting.h" +#elif defined EIGEN_VECTORIZE_SSE + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/SSE/Complex.h" + #include "src/Core/arch/SSE/TypeCasting.h" +#elif defined(EIGEN_VECTORIZE_ALTIVEC) || defined(EIGEN_VECTORIZE_VSX) + #include "src/Core/arch/AltiVec/PacketMath.h" + #include "src/Core/arch/AltiVec/MathFunctions.h" + #include "src/Core/arch/AltiVec/Complex.h" +#elif defined EIGEN_VECTORIZE_NEON + #include "src/Core/arch/NEON/PacketMath.h" + #include "src/Core/arch/NEON/MathFunctions.h" + #include "src/Core/arch/NEON/Complex.h" +#elif defined EIGEN_VECTORIZE_ZVECTOR + #include "src/Core/arch/ZVector/PacketMath.h" + #include "src/Core/arch/ZVector/MathFunctions.h" + #include "src/Core/arch/ZVector/Complex.h" +#endif + +// Half float support +#include "src/Core/arch/CUDA/Half.h" +#include "src/Core/arch/CUDA/PacketMathHalf.h" +#include "src/Core/arch/CUDA/TypeCasting.h" + +#if defined EIGEN_VECTORIZE_CUDA + #include "src/Core/arch/CUDA/PacketMath.h" + #include "src/Core/arch/CUDA/MathFunctions.h" +#endif + +#include "src/Core/arch/Default/Settings.h" + +#include "src/Core/functors/TernaryFunctors.h" +#include "src/Core/functors/BinaryFunctors.h" +#include "src/Core/functors/UnaryFunctors.h" +#include "src/Core/functors/NullaryFunctors.h" +#include "src/Core/functors/StlFunctors.h" +#include "src/Core/functors/AssignmentFunctors.h" + +// Specialized functors to enable the processing of complex numbers +// on CUDA devices +#include "src/Core/arch/CUDA/Complex.h" + +#include "src/Core/IO.h" +#include "src/Core/DenseCoeffsBase.h" +#include "src/Core/DenseBase.h" +#include "src/Core/MatrixBase.h" +#include "src/Core/EigenBase.h" + +#include "src/Core/Product.h" +#include "src/Core/CoreEvaluators.h" +#include "src/Core/AssignEvaluator.h" + +#ifndef EIGEN_PARSED_BY_DOXYGEN // work around Doxygen bug triggered by Assign.h r814874 + // at least confirmed with Doxygen 1.5.5 and 1.5.6 + #include "src/Core/Assign.h" +#endif + +#include "src/Core/ArrayBase.h" +#include "src/Core/util/BlasUtil.h" +#include "src/Core/DenseStorage.h" +#include "src/Core/NestByValue.h" + +// #include "src/Core/ForceAlignedAccess.h" + +#include "src/Core/ReturnByValue.h" +#include "src/Core/NoAlias.h" +#include "src/Core/PlainObjectBase.h" +#include "src/Core/Matrix.h" +#include "src/Core/Array.h" +#include "src/Core/CwiseTernaryOp.h" +#include "src/Core/CwiseBinaryOp.h" +#include "src/Core/CwiseUnaryOp.h" +#include "src/Core/CwiseNullaryOp.h" +#include "src/Core/CwiseUnaryView.h" +#include "src/Core/SelfCwiseBinaryOp.h" +#include "src/Core/Dot.h" +#include "src/Core/StableNorm.h" +#include "src/Core/Stride.h" +#include "src/Core/MapBase.h" +#include "src/Core/Map.h" +#include "src/Core/Ref.h" +#include "src/Core/Block.h" +#include "src/Core/VectorBlock.h" +#include "src/Core/Transpose.h" +#include "src/Core/DiagonalMatrix.h" +#include "src/Core/Diagonal.h" +#include "src/Core/DiagonalProduct.h" +#include "src/Core/Redux.h" +#include "src/Core/Visitor.h" +#include "src/Core/Fuzzy.h" +#include "src/Core/Swap.h" +#include "src/Core/CommaInitializer.h" +#include "src/Core/GeneralProduct.h" +#include "src/Core/Solve.h" +#include "src/Core/Inverse.h" +#include "src/Core/SolverBase.h" +#include "src/Core/PermutationMatrix.h" +#include "src/Core/Transpositions.h" +#include "src/Core/TriangularMatrix.h" +#include "src/Core/SelfAdjointView.h" +#include "src/Core/products/GeneralBlockPanelKernel.h" +#include "src/Core/products/Parallelizer.h" +#include "src/Core/ProductEvaluators.h" +#include "src/Core/products/GeneralMatrixVector.h" +#include "src/Core/products/GeneralMatrixMatrix.h" +#include "src/Core/SolveTriangular.h" +#include "src/Core/products/GeneralMatrixMatrixTriangular.h" +#include "src/Core/products/SelfadjointMatrixVector.h" +#include "src/Core/products/SelfadjointMatrixMatrix.h" +#include "src/Core/products/SelfadjointProduct.h" +#include "src/Core/products/SelfadjointRank2Update.h" +#include "src/Core/products/TriangularMatrixVector.h" +#include "src/Core/products/TriangularMatrixMatrix.h" +#include "src/Core/products/TriangularSolverMatrix.h" +#include "src/Core/products/TriangularSolverVector.h" +#include "src/Core/BandMatrix.h" +#include "src/Core/CoreIterators.h" +#include "src/Core/ConditionEstimator.h" + +#include "src/Core/BooleanRedux.h" +#include "src/Core/Select.h" +#include "src/Core/VectorwiseOp.h" +#include "src/Core/Random.h" +#include "src/Core/Replicate.h" +#include "src/Core/Reverse.h" +#include "src/Core/ArrayWrapper.h" + +#ifdef EIGEN_USE_BLAS +#include "src/Core/products/GeneralMatrixMatrix_BLAS.h" +#include "src/Core/products/GeneralMatrixVector_BLAS.h" +#include "src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h" +#include "src/Core/products/SelfadjointMatrixMatrix_BLAS.h" +#include "src/Core/products/SelfadjointMatrixVector_BLAS.h" +#include "src/Core/products/TriangularMatrixMatrix_BLAS.h" +#include "src/Core/products/TriangularMatrixVector_BLAS.h" +#include "src/Core/products/TriangularSolverMatrix_BLAS.h" +#endif // EIGEN_USE_BLAS + +#ifdef EIGEN_USE_MKL_VML +#include "src/Core/Assign_MKL.h" +#endif + +#include "src/Core/GlobalFunctions.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_CORE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Dense b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Dense new file mode 100644 index 00000000000..5768910bd88 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Dense @@ -0,0 +1,7 @@ +#include "Core" +#include "LU" +#include "Cholesky" +#include "QR" +#include "SVD" +#include "Geometry" +#include "Eigenvalues" diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigen b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigen new file mode 100644 index 00000000000..654c8dc6380 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigen @@ -0,0 +1,2 @@ +#include "Dense" +#include "Sparse" diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigenvalues b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigenvalues new file mode 100644 index 00000000000..7d6ac787bed --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Eigenvalues @@ -0,0 +1,61 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_EIGENVALUES_MODULE_H +#define EIGEN_EIGENVALUES_MODULE_H + +#include "Core" + +#include "Cholesky" +#include "Jacobi" +#include "Householder" +#include "LU" +#include "Geometry" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Eigenvalues_Module Eigenvalues module + * + * + * + * This module mainly provides various eigenvalue solvers. + * This module also provides some MatrixBase methods, including: + * - MatrixBase::eigenvalues(), + * - MatrixBase::operatorNorm() + * + * \code + * #include + * \endcode + */ + +#include "src/misc/RealSvd2x2.h" +#include "src/Eigenvalues/Tridiagonalization.h" +#include "src/Eigenvalues/RealSchur.h" +#include "src/Eigenvalues/EigenSolver.h" +#include "src/Eigenvalues/SelfAdjointEigenSolver.h" +#include "src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h" +#include "src/Eigenvalues/HessenbergDecomposition.h" +#include "src/Eigenvalues/ComplexSchur.h" +#include "src/Eigenvalues/ComplexEigenSolver.h" +#include "src/Eigenvalues/RealQZ.h" +#include "src/Eigenvalues/GeneralizedEigenSolver.h" +#include "src/Eigenvalues/MatrixBaseEigenvalues.h" +#ifdef EIGEN_USE_LAPACKE +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/Eigenvalues/RealSchur_LAPACKE.h" +#include "src/Eigenvalues/ComplexSchur_LAPACKE.h" +#include "src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_EIGENVALUES_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Geometry b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Geometry new file mode 100644 index 00000000000..da88c03bbf5 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Geometry @@ -0,0 +1,62 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GEOMETRY_MODULE_H +#define EIGEN_GEOMETRY_MODULE_H + +#include "Core" + +#include "SVD" +#include "LU" +#include + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Geometry_Module Geometry module + * + * This module provides support for: + * - fixed-size homogeneous transformations + * - translation, scaling, 2D and 3D rotations + * - \link Quaternion quaternions \endlink + * - cross products (\ref MatrixBase::cross, \ref MatrixBase::cross3) + * - orthognal vector generation (\ref MatrixBase::unitOrthogonal) + * - some linear components: \link ParametrizedLine parametrized-lines \endlink and \link Hyperplane hyperplanes \endlink + * - \link AlignedBox axis aligned bounding boxes \endlink + * - \link umeyama least-square transformation fitting \endlink + * + * \code + * #include + * \endcode + */ + +#include "src/Geometry/OrthoMethods.h" +#include "src/Geometry/EulerAngles.h" + +#include "src/Geometry/Homogeneous.h" +#include "src/Geometry/RotationBase.h" +#include "src/Geometry/Rotation2D.h" +#include "src/Geometry/Quaternion.h" +#include "src/Geometry/AngleAxis.h" +#include "src/Geometry/Transform.h" +#include "src/Geometry/Translation.h" +#include "src/Geometry/Scaling.h" +#include "src/Geometry/Hyperplane.h" +#include "src/Geometry/ParametrizedLine.h" +#include "src/Geometry/AlignedBox.h" +#include "src/Geometry/Umeyama.h" + +// Use the SSE optimized version whenever possible. At the moment the +// SSE version doesn't compile when AVX is enabled +#if defined EIGEN_VECTORIZE_SSE && !defined EIGEN_VECTORIZE_AVX +#include "src/Geometry/arch/Geometry_SSE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_GEOMETRY_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Householder b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Householder new file mode 100644 index 00000000000..89cd81b1afb --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Householder @@ -0,0 +1,30 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_HOUSEHOLDER_MODULE_H +#define EIGEN_HOUSEHOLDER_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Householder_Module Householder module + * This module provides Householder transformations. + * + * \code + * #include + * \endcode + */ + +#include "src/Householder/Householder.h" +#include "src/Householder/HouseholderSequence.h" +#include "src/Householder/BlockHouseholder.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_HOUSEHOLDER_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/IterativeLinearSolvers b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/IterativeLinearSolvers new file mode 100644 index 00000000000..957d5750b2c --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/IterativeLinearSolvers @@ -0,0 +1,48 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ITERATIVELINEARSOLVERS_MODULE_H +#define EIGEN_ITERATIVELINEARSOLVERS_MODULE_H + +#include "SparseCore" +#include "OrderingMethods" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** + * \defgroup IterativeLinearSolvers_Module IterativeLinearSolvers module + * + * This module currently provides iterative methods to solve problems of the form \c A \c x = \c b, where \c A is a squared matrix, usually very large and sparse. + * Those solvers are accessible via the following classes: + * - ConjugateGradient for selfadjoint (hermitian) matrices, + * - LeastSquaresConjugateGradient for rectangular least-square problems, + * - BiCGSTAB for general square matrices. + * + * These iterative solvers are associated with some preconditioners: + * - IdentityPreconditioner - not really useful + * - DiagonalPreconditioner - also called Jacobi preconditioner, work very well on diagonal dominant matrices. + * - IncompleteLUT - incomplete LU factorization with dual thresholding + * + * Such problems can also be solved using the direct sparse decomposition modules: SparseCholesky, CholmodSupport, UmfPackSupport, SuperLUSupport. + * + \code + #include + \endcode + */ + +#include "src/IterativeLinearSolvers/SolveWithGuess.h" +#include "src/IterativeLinearSolvers/IterativeSolverBase.h" +#include "src/IterativeLinearSolvers/BasicPreconditioners.h" +#include "src/IterativeLinearSolvers/ConjugateGradient.h" +#include "src/IterativeLinearSolvers/LeastSquareConjugateGradient.h" +#include "src/IterativeLinearSolvers/BiCGSTAB.h" +#include "src/IterativeLinearSolvers/IncompleteLUT.h" +#include "src/IterativeLinearSolvers/IncompleteCholesky.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_ITERATIVELINEARSOLVERS_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Jacobi b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Jacobi new file mode 100644 index 00000000000..17c1d785a16 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Jacobi @@ -0,0 +1,33 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_JACOBI_MODULE_H +#define EIGEN_JACOBI_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Jacobi_Module Jacobi module + * This module provides Jacobi and Givens rotations. + * + * \code + * #include + * \endcode + * + * In addition to listed classes, it defines the two following MatrixBase methods to apply a Jacobi or Givens rotation: + * - MatrixBase::applyOnTheLeft() + * - MatrixBase::applyOnTheRight(). + */ + +#include "src/Jacobi/Jacobi.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_JACOBI_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/LU b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/LU new file mode 100644 index 00000000000..6418a86e192 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/LU @@ -0,0 +1,50 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_LU_MODULE_H +#define EIGEN_LU_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup LU_Module LU module + * This module includes %LU decomposition and related notions such as matrix inversion and determinant. + * This module defines the following MatrixBase methods: + * - MatrixBase::inverse() + * - MatrixBase::determinant() + * + * \code + * #include + * \endcode + */ + +#include "src/misc/Kernel.h" +#include "src/misc/Image.h" +#include "src/LU/FullPivLU.h" +#include "src/LU/PartialPivLU.h" +#ifdef EIGEN_USE_LAPACKE +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/LU/PartialPivLU_LAPACKE.h" +#endif +#include "src/LU/Determinant.h" +#include "src/LU/InverseImpl.h" + +// Use the SSE optimized version whenever possible. At the moment the +// SSE version doesn't compile when AVX is enabled +#if defined EIGEN_VECTORIZE_SSE && !defined EIGEN_VECTORIZE_AVX + #include "src/LU/arch/Inverse_SSE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_LU_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/MetisSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/MetisSupport new file mode 100644 index 00000000000..85c41bf3400 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/MetisSupport @@ -0,0 +1,35 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_METISSUPPORT_MODULE_H +#define EIGEN_METISSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +extern "C" { +#include +} + + +/** \ingroup Support_modules + * \defgroup MetisSupport_Module MetisSupport module + * + * \code + * #include + * \endcode + * This module defines an interface to the METIS reordering package (http://glaros.dtc.umn.edu/gkhome/views/metis). + * It can be used just as any other built-in method as explained in \link OrderingMethods_Module here. \endlink + */ + + +#include "src/MetisSupport/MetisSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_METISSUPPORT_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/OrderingMethods b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/OrderingMethods new file mode 100644 index 00000000000..d8ea3619366 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/OrderingMethods @@ -0,0 +1,73 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ORDERINGMETHODS_MODULE_H +#define EIGEN_ORDERINGMETHODS_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** + * \defgroup OrderingMethods_Module OrderingMethods module + * + * This module is currently for internal use only + * + * It defines various built-in and external ordering methods for sparse matrices. + * They are typically used to reduce the number of elements during + * the sparse matrix decomposition (LLT, LU, QR). + * Precisely, in a preprocessing step, a permutation matrix P is computed using + * those ordering methods and applied to the columns of the matrix. + * Using for instance the sparse Cholesky decomposition, it is expected that + * the nonzeros elements in LLT(A*P) will be much smaller than that in LLT(A). + * + * + * Usage : + * \code + * #include + * \endcode + * + * A simple usage is as a template parameter in the sparse decomposition classes : + * + * \code + * SparseLU > solver; + * \endcode + * + * \code + * SparseQR > solver; + * \endcode + * + * It is possible as well to call directly a particular ordering method for your own purpose, + * \code + * AMDOrdering ordering; + * PermutationMatrix perm; + * SparseMatrix A; + * //Fill the matrix ... + * + * ordering(A, perm); // Call AMD + * \endcode + * + * \note Some of these methods (like AMD or METIS), need the sparsity pattern + * of the input matrix to be symmetric. When the matrix is structurally unsymmetric, + * Eigen computes internally the pattern of \f$A^T*A\f$ before calling the method. + * If your matrix is already symmetric (at leat in structure), you can avoid that + * by calling the method with a SelfAdjointView type. + * + * \code + * // Call the ordering on the pattern of the lower triangular matrix A + * ordering(A.selfadjointView(), perm); + * \endcode + */ + +#ifndef EIGEN_MPL2_ONLY +#include "src/OrderingMethods/Amd.h" +#endif + +#include "src/OrderingMethods/Ordering.h" +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_ORDERINGMETHODS_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PaStiXSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PaStiXSupport new file mode 100644 index 00000000000..de3a63b4d12 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PaStiXSupport @@ -0,0 +1,48 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PASTIXSUPPORT_MODULE_H +#define EIGEN_PASTIXSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +extern "C" { +#include +#include +} + +#ifdef complex +#undef complex +#endif + +/** \ingroup Support_modules + * \defgroup PaStiXSupport_Module PaStiXSupport module + * + * This module provides an interface to the PaSTiX library. + * PaSTiX is a general \b supernodal, \b parallel and \b opensource sparse solver. + * It provides the two following main factorization classes: + * - class PastixLLT : a supernodal, parallel LLt Cholesky factorization. + * - class PastixLDLT: a supernodal, parallel LDLt Cholesky factorization. + * - class PastixLU : a supernodal, parallel LU factorization (optimized for a symmetric pattern). + * + * \code + * #include + * \endcode + * + * In order to use this module, the PaSTiX headers must be accessible from the include paths, and your binary must be linked to the PaSTiX library and its dependencies. + * The dependencies depend on how PaSTiX has been compiled. + * For a cmake based project, you can use our FindPaSTiX.cmake module to help you in this task. + * + */ + +#include "src/PaStiXSupport/PaStiXSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_PASTIXSUPPORT_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PardisoSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PardisoSupport new file mode 100644 index 00000000000..340edf51fe2 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/PardisoSupport @@ -0,0 +1,35 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PARDISOSUPPORT_MODULE_H +#define EIGEN_PARDISOSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +#include + +/** \ingroup Support_modules + * \defgroup PardisoSupport_Module PardisoSupport module + * + * This module brings support for the Intel(R) MKL PARDISO direct sparse solvers. + * + * \code + * #include + * \endcode + * + * In order to use this module, the MKL headers must be accessible from the include paths, and your binary must be linked to the MKL library and its dependencies. + * See this \ref TopicUsingIntelMKL "page" for more information on MKL-Eigen integration. + * + */ + +#include "src/PardisoSupport/PardisoSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_PARDISOSUPPORT_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QR b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QR new file mode 100644 index 00000000000..1be1863a1d5 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QR @@ -0,0 +1,51 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_QR_MODULE_H +#define EIGEN_QR_MODULE_H + +#include "Core" + +#include "Cholesky" +#include "Jacobi" +#include "Householder" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup QR_Module QR module + * + * + * + * This module provides various QR decompositions + * This module also provides some MatrixBase methods, including: + * - MatrixBase::householderQr() + * - MatrixBase::colPivHouseholderQr() + * - MatrixBase::fullPivHouseholderQr() + * + * \code + * #include + * \endcode + */ + +#include "src/QR/HouseholderQR.h" +#include "src/QR/FullPivHouseholderQR.h" +#include "src/QR/ColPivHouseholderQR.h" +#include "src/QR/CompleteOrthogonalDecomposition.h" +#ifdef EIGEN_USE_LAPACKE +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/QR/HouseholderQR_LAPACKE.h" +#include "src/QR/ColPivHouseholderQR_LAPACKE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_QR_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QtAlignedMalloc b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QtAlignedMalloc new file mode 100644 index 00000000000..4f07df02ae9 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/QtAlignedMalloc @@ -0,0 +1,40 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_QTMALLOC_MODULE_H +#define EIGEN_QTMALLOC_MODULE_H + +#include "Core" + +#if (!EIGEN_MALLOC_ALREADY_ALIGNED) + +#include "src/Core/util/DisableStupidWarnings.h" + +void *qMalloc(std::size_t size) +{ + return Eigen::internal::aligned_malloc(size); +} + +void qFree(void *ptr) +{ + Eigen::internal::aligned_free(ptr); +} + +void *qRealloc(void *ptr, std::size_t size) +{ + void* newPtr = Eigen::internal::aligned_malloc(size); + std::memcpy(newPtr, ptr, size); + Eigen::internal::aligned_free(ptr); + return newPtr; +} + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif + +#endif // EIGEN_QTMALLOC_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SPQRSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SPQRSupport new file mode 100644 index 00000000000..f70390c1766 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SPQRSupport @@ -0,0 +1,34 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPQRSUPPORT_MODULE_H +#define EIGEN_SPQRSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +#include "SuiteSparseQR.hpp" + +/** \ingroup Support_modules + * \defgroup SPQRSupport_Module SuiteSparseQR module + * + * This module provides an interface to the SPQR library, which is part of the suitesparse package. + * + * \code + * #include + * \endcode + * + * In order to use this module, the SPQR headers must be accessible from the include paths, and your binary must be linked to the SPQR library and its dependencies (Cholmod, AMD, COLAMD,...). + * For a cmake based project, you can use our FindSPQR.cmake and FindCholmod.Cmake modules + * + */ + +#include "src/CholmodSupport/CholmodSupport.h" +#include "src/SPQRSupport/SuiteSparseQRSupport.h" + +#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SVD b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SVD new file mode 100644 index 00000000000..5d0e75f7f75 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SVD @@ -0,0 +1,51 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SVD_MODULE_H +#define EIGEN_SVD_MODULE_H + +#include "QR" +#include "Householder" +#include "Jacobi" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup SVD_Module SVD module + * + * + * + * This module provides SVD decomposition for matrices (both real and complex). + * Two decomposition algorithms are provided: + * - JacobiSVD implementing two-sided Jacobi iterations is numerically very accurate, fast for small matrices, but very slow for larger ones. + * - BDCSVD implementing a recursive divide & conquer strategy on top of an upper-bidiagonalization which remains fast for large problems. + * These decompositions are accessible via the respective classes and following MatrixBase methods: + * - MatrixBase::jacobiSvd() + * - MatrixBase::bdcSvd() + * + * \code + * #include + * \endcode + */ + +#include "src/misc/RealSvd2x2.h" +#include "src/SVD/UpperBidiagonalization.h" +#include "src/SVD/SVDBase.h" +#include "src/SVD/JacobiSVD.h" +#include "src/SVD/BDCSVD.h" +#if defined(EIGEN_USE_LAPACKE) && !defined(EIGEN_USE_LAPACKE_STRICT) +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/SVD/JacobiSVD_LAPACKE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_SVD_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Sparse b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Sparse new file mode 100644 index 00000000000..136e681a1f6 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/Sparse @@ -0,0 +1,36 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPARSE_MODULE_H +#define EIGEN_SPARSE_MODULE_H + +/** \defgroup Sparse_Module Sparse meta-module + * + * Meta-module including all related modules: + * - \ref SparseCore_Module + * - \ref OrderingMethods_Module + * - \ref SparseCholesky_Module + * - \ref SparseLU_Module + * - \ref SparseQR_Module + * - \ref IterativeLinearSolvers_Module + * + \code + #include + \endcode + */ + +#include "SparseCore" +#include "OrderingMethods" +#ifndef EIGEN_MPL2_ONLY +#include "SparseCholesky" +#endif +#include "SparseLU" +#include "SparseQR" +#include "IterativeLinearSolvers" + +#endif // EIGEN_SPARSE_MODULE_H + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCholesky b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCholesky new file mode 100644 index 00000000000..b6a320c4027 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCholesky @@ -0,0 +1,45 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2013 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPARSECHOLESKY_MODULE_H +#define EIGEN_SPARSECHOLESKY_MODULE_H + +#include "SparseCore" +#include "OrderingMethods" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** + * \defgroup SparseCholesky_Module SparseCholesky module + * + * This module currently provides two variants of the direct sparse Cholesky decomposition for selfadjoint (hermitian) matrices. + * Those decompositions are accessible via the following classes: + * - SimplicialLLt, + * - SimplicialLDLt + * + * Such problems can also be solved using the ConjugateGradient solver from the IterativeLinearSolvers module. + * + * \code + * #include + * \endcode + */ + +#ifdef EIGEN_MPL2_ONLY +#error The SparseCholesky module has nothing to offer in MPL2 only mode +#endif + +#include "src/SparseCholesky/SimplicialCholesky.h" + +#ifndef EIGEN_MPL2_ONLY +#include "src/SparseCholesky/SimplicialCholesky_impl.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_SPARSECHOLESKY_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCore b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCore new file mode 100644 index 00000000000..76966c4c4cb --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseCore @@ -0,0 +1,69 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPARSECORE_MODULE_H +#define EIGEN_SPARSECORE_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +#include +#include +#include +#include +#include + +/** + * \defgroup SparseCore_Module SparseCore module + * + * This module provides a sparse matrix representation, and basic associated matrix manipulations + * and operations. + * + * See the \ref TutorialSparse "Sparse tutorial" + * + * \code + * #include + * \endcode + * + * This module depends on: Core. + */ + +#include "src/SparseCore/SparseUtil.h" +#include "src/SparseCore/SparseMatrixBase.h" +#include "src/SparseCore/SparseAssign.h" +#include "src/SparseCore/CompressedStorage.h" +#include "src/SparseCore/AmbiVector.h" +#include "src/SparseCore/SparseCompressedBase.h" +#include "src/SparseCore/SparseMatrix.h" +#include "src/SparseCore/SparseMap.h" +#include "src/SparseCore/MappedSparseMatrix.h" +#include "src/SparseCore/SparseVector.h" +#include "src/SparseCore/SparseRef.h" +#include "src/SparseCore/SparseCwiseUnaryOp.h" +#include "src/SparseCore/SparseCwiseBinaryOp.h" +#include "src/SparseCore/SparseTranspose.h" +#include "src/SparseCore/SparseBlock.h" +#include "src/SparseCore/SparseDot.h" +#include "src/SparseCore/SparseRedux.h" +#include "src/SparseCore/SparseView.h" +#include "src/SparseCore/SparseDiagonalProduct.h" +#include "src/SparseCore/ConservativeSparseSparseProduct.h" +#include "src/SparseCore/SparseSparseProductWithPruning.h" +#include "src/SparseCore/SparseProduct.h" +#include "src/SparseCore/SparseDenseProduct.h" +#include "src/SparseCore/SparseSelfAdjointView.h" +#include "src/SparseCore/SparseTriangularView.h" +#include "src/SparseCore/TriangularSolver.h" +#include "src/SparseCore/SparsePermutation.h" +#include "src/SparseCore/SparseFuzzy.h" +#include "src/SparseCore/SparseSolverBase.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_SPARSECORE_MODULE_H + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseLU b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseLU new file mode 100644 index 00000000000..38b38b531d1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseLU @@ -0,0 +1,46 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2012 Désiré Nuentsa-Wakam +// Copyright (C) 2012 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPARSELU_MODULE_H +#define EIGEN_SPARSELU_MODULE_H + +#include "SparseCore" + +/** + * \defgroup SparseLU_Module SparseLU module + * This module defines a supernodal factorization of general sparse matrices. + * The code is fully optimized for supernode-panel updates with specialized kernels. + * Please, see the documentation of the SparseLU class for more details. + */ + +// Ordering interface +#include "OrderingMethods" + +#include "src/SparseLU/SparseLU_gemm_kernel.h" + +#include "src/SparseLU/SparseLU_Structs.h" +#include "src/SparseLU/SparseLU_SupernodalMatrix.h" +#include "src/SparseLU/SparseLUImpl.h" +#include "src/SparseCore/SparseColEtree.h" +#include "src/SparseLU/SparseLU_Memory.h" +#include "src/SparseLU/SparseLU_heap_relax_snode.h" +#include "src/SparseLU/SparseLU_relax_snode.h" +#include "src/SparseLU/SparseLU_pivotL.h" +#include "src/SparseLU/SparseLU_panel_dfs.h" +#include "src/SparseLU/SparseLU_kernel_bmod.h" +#include "src/SparseLU/SparseLU_panel_bmod.h" +#include "src/SparseLU/SparseLU_column_dfs.h" +#include "src/SparseLU/SparseLU_column_bmod.h" +#include "src/SparseLU/SparseLU_copy_to_ucol.h" +#include "src/SparseLU/SparseLU_pruneL.h" +#include "src/SparseLU/SparseLU_Utils.h" +#include "src/SparseLU/SparseLU.h" + +#endif // EIGEN_SPARSELU_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseQR b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseQR new file mode 100644 index 00000000000..f5fc5fa7fee --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SparseQR @@ -0,0 +1,36 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SPARSEQR_MODULE_H +#define EIGEN_SPARSEQR_MODULE_H + +#include "SparseCore" +#include "OrderingMethods" +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup SparseQR_Module SparseQR module + * \brief Provides QR decomposition for sparse matrices + * + * This module provides a simplicial version of the left-looking Sparse QR decomposition. + * The columns of the input matrix should be reordered to limit the fill-in during the + * decomposition. Built-in methods (COLAMD, AMD) or external methods (METIS) can be used to this end. + * See the \link OrderingMethods_Module OrderingMethods\endlink module for the list + * of built-in and external ordering methods. + * + * \code + * #include + * \endcode + * + * + */ + +#include "src/SparseCore/SparseColEtree.h" +#include "src/SparseQR/SparseQR.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdDeque b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdDeque new file mode 100644 index 00000000000..bc68397be25 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdDeque @@ -0,0 +1,27 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// Copyright (C) 2009 Hauke Heibel +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_STDDEQUE_MODULE_H +#define EIGEN_STDDEQUE_MODULE_H + +#include "Core" +#include + +#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */ + +#define EIGEN_DEFINE_STL_DEQUE_SPECIALIZATION(...) + +#else + +#include "src/StlSupport/StdDeque.h" + +#endif + +#endif // EIGEN_STDDEQUE_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdList b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdList new file mode 100644 index 00000000000..4c6262c08cc --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdList @@ -0,0 +1,26 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Hauke Heibel +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_STDLIST_MODULE_H +#define EIGEN_STDLIST_MODULE_H + +#include "Core" +#include + +#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */ + +#define EIGEN_DEFINE_STL_LIST_SPECIALIZATION(...) + +#else + +#include "src/StlSupport/StdList.h" + +#endif + +#endif // EIGEN_STDLIST_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdVector b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdVector new file mode 100644 index 00000000000..0c4697ad5be --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/StdVector @@ -0,0 +1,27 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// Copyright (C) 2009 Hauke Heibel +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_STDVECTOR_MODULE_H +#define EIGEN_STDVECTOR_MODULE_H + +#include "Core" +#include + +#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */ + +#define EIGEN_DEFINE_STL_VECTOR_SPECIALIZATION(...) + +#else + +#include "src/StlSupport/StdVector.h" + +#endif + +#endif // EIGEN_STDVECTOR_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SuperLUSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SuperLUSupport new file mode 100644 index 00000000000..59312a82db0 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/SuperLUSupport @@ -0,0 +1,64 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SUPERLUSUPPORT_MODULE_H +#define EIGEN_SUPERLUSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +#ifdef EMPTY +#define EIGEN_EMPTY_WAS_ALREADY_DEFINED +#endif + +typedef int int_t; +#include +#include +#include + +// slu_util.h defines a preprocessor token named EMPTY which is really polluting, +// so we remove it in favor of a SUPERLU_EMPTY token. +// If EMPTY was already defined then we don't undef it. + +#if defined(EIGEN_EMPTY_WAS_ALREADY_DEFINED) +# undef EIGEN_EMPTY_WAS_ALREADY_DEFINED +#elif defined(EMPTY) +# undef EMPTY +#endif + +#define SUPERLU_EMPTY (-1) + +namespace Eigen { struct SluMatrix; } + +/** \ingroup Support_modules + * \defgroup SuperLUSupport_Module SuperLUSupport module + * + * This module provides an interface to the SuperLU library. + * It provides the following factorization class: + * - class SuperLU: a supernodal sequential LU factorization. + * - class SuperILU: a supernodal sequential incomplete LU factorization (to be used as a preconditioner for iterative methods). + * + * \warning This wrapper requires at least versions 4.0 of SuperLU. The 3.x versions are not supported. + * + * \warning When including this module, you have to use SUPERLU_EMPTY instead of EMPTY which is no longer defined because it is too polluting. + * + * \code + * #include + * \endcode + * + * In order to use this module, the superlu headers must be accessible from the include paths, and your binary must be linked to the superlu library and its dependencies. + * The dependencies depend on how superlu has been compiled. + * For a cmake based project, you can use our FindSuperLU.cmake module to help you in this task. + * + */ + +#include "src/SuperLUSupport/SuperLUSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_SUPERLUSUPPORT_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/UmfPackSupport b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/UmfPackSupport new file mode 100644 index 00000000000..00eec80875f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/UmfPackSupport @@ -0,0 +1,40 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_UMFPACKSUPPORT_MODULE_H +#define EIGEN_UMFPACKSUPPORT_MODULE_H + +#include "SparseCore" + +#include "src/Core/util/DisableStupidWarnings.h" + +extern "C" { +#include +} + +/** \ingroup Support_modules + * \defgroup UmfPackSupport_Module UmfPackSupport module + * + * This module provides an interface to the UmfPack library which is part of the suitesparse package. + * It provides the following factorization class: + * - class UmfPackLU: a multifrontal sequential LU factorization. + * + * \code + * #include + * \endcode + * + * In order to use this module, the umfpack headers must be accessible from the include paths, and your binary must be linked to the umfpack library and its dependencies. + * The dependencies depend on how umfpack has been compiled. + * For a cmake based project, you can use our FindUmfPack.cmake module to help you in this task. + * + */ + +#include "src/UmfPackSupport/UmfPackSupport.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_UMFPACKSUPPORT_MODULE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LDLT.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LDLT.h new file mode 100644 index 00000000000..15ccf24f144 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LDLT.h @@ -0,0 +1,673 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2011 Gael Guennebaud +// Copyright (C) 2009 Keir Mierle +// Copyright (C) 2009 Benoit Jacob +// Copyright (C) 2011 Timothy E. Holy +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_LDLT_H +#define EIGEN_LDLT_H + +namespace Eigen { + +namespace internal { + template struct LDLT_Traits; + + // PositiveSemiDef means positive semi-definite and non-zero; same for NegativeSemiDef + enum SignMatrix { PositiveSemiDef, NegativeSemiDef, ZeroSign, Indefinite }; +} + +/** \ingroup Cholesky_Module + * + * \class LDLT + * + * \brief Robust Cholesky decomposition of a matrix with pivoting + * + * \tparam _MatrixType the type of the matrix of which to compute the LDL^T Cholesky decomposition + * \tparam _UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper. + * The other triangular part won't be read. + * + * Perform a robust Cholesky decomposition of a positive semidefinite or negative semidefinite + * matrix \f$ A \f$ such that \f$ A = P^TLDL^*P \f$, where P is a permutation matrix, L + * is lower triangular with a unit diagonal and D is a diagonal matrix. + * + * The decomposition uses pivoting to ensure stability, so that L will have + * zeros in the bottom right rank(A) - n submatrix. Avoiding the square root + * on D also stabilizes the computation. + * + * Remember that Cholesky decompositions are not rank-revealing. Also, do not use a Cholesky + * decomposition to determine whether a system of equations has a solution. + * + * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. + * + * \sa MatrixBase::ldlt(), SelfAdjointView::ldlt(), class LLT + */ +template class LDLT +{ + public: + typedef _MatrixType MatrixType; + enum { + RowsAtCompileTime = MatrixType::RowsAtCompileTime, + ColsAtCompileTime = MatrixType::ColsAtCompileTime, + MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, + UpLo = _UpLo + }; + typedef typename MatrixType::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 + typedef typename MatrixType::StorageIndex StorageIndex; + typedef Matrix TmpMatrixType; + + typedef Transpositions TranspositionType; + typedef PermutationMatrix PermutationType; + + typedef internal::LDLT_Traits Traits; + + /** \brief Default Constructor. + * + * The default constructor is useful in cases in which the user intends to + * perform decompositions via LDLT::compute(const MatrixType&). + */ + LDLT() + : m_matrix(), + m_transpositions(), + m_sign(internal::ZeroSign), + m_isInitialized(false) + {} + + /** \brief Default Constructor with memory preallocation + * + * Like the default constructor but with preallocation of the internal data + * according to the specified problem \a size. + * \sa LDLT() + */ + explicit LDLT(Index size) + : m_matrix(size, size), + m_transpositions(size), + m_temporary(size), + m_sign(internal::ZeroSign), + m_isInitialized(false) + {} + + /** \brief Constructor with decomposition + * + * This calculates the decomposition for the input \a matrix. + * + * \sa LDLT(Index size) + */ + template + explicit LDLT(const EigenBase& matrix) + : m_matrix(matrix.rows(), matrix.cols()), + m_transpositions(matrix.rows()), + m_temporary(matrix.rows()), + m_sign(internal::ZeroSign), + m_isInitialized(false) + { + compute(matrix.derived()); + } + + /** \brief Constructs a LDLT factorization from a given matrix + * + * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. + * + * \sa LDLT(const EigenBase&) + */ + template + explicit LDLT(EigenBase& matrix) + : m_matrix(matrix.derived()), + m_transpositions(matrix.rows()), + m_temporary(matrix.rows()), + m_sign(internal::ZeroSign), + m_isInitialized(false) + { + compute(matrix.derived()); + } + + /** Clear any existing decomposition + * \sa rankUpdate(w,sigma) + */ + void setZero() + { + m_isInitialized = false; + } + + /** \returns a view of the upper triangular matrix U */ + inline typename Traits::MatrixU matrixU() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return Traits::getU(m_matrix); + } + + /** \returns a view of the lower triangular matrix L */ + inline typename Traits::MatrixL matrixL() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return Traits::getL(m_matrix); + } + + /** \returns the permutation matrix P as a transposition sequence. + */ + inline const TranspositionType& transpositionsP() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_transpositions; + } + + /** \returns the coefficients of the diagonal matrix D */ + inline Diagonal vectorD() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_matrix.diagonal(); + } + + /** \returns true if the matrix is positive (semidefinite) */ + inline bool isPositive() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_sign == internal::PositiveSemiDef || m_sign == internal::ZeroSign; + } + + /** \returns true if the matrix is negative (semidefinite) */ + inline bool isNegative(void) const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_sign == internal::NegativeSemiDef || m_sign == internal::ZeroSign; + } + + /** \returns a solution x of \f$ A x = b \f$ using the current decomposition of A. + * + * This function also supports in-place solves using the syntax x = decompositionObject.solve(x) . + * + * \note_about_checking_solutions + * + * More precisely, this method solves \f$ A x = b \f$ using the decomposition \f$ A = P^T L D L^* P \f$ + * by solving the systems \f$ P^T y_1 = b \f$, \f$ L y_2 = y_1 \f$, \f$ D y_3 = y_2 \f$, + * \f$ L^* y_4 = y_3 \f$ and \f$ P x = y_4 \f$ in succession. If the matrix \f$ A \f$ is singular, then + * \f$ D \f$ will also be singular (all the other matrices are invertible). In that case, the + * least-square solution of \f$ D y_3 = y_2 \f$ is computed. This does not mean that this function + * computes the least-square solution of \f$ A x = b \f$ is \f$ A \f$ is singular. + * + * \sa MatrixBase::ldlt(), SelfAdjointView::ldlt() + */ + template + inline const Solve + solve(const MatrixBase& b) const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + eigen_assert(m_matrix.rows()==b.rows() + && "LDLT::solve(): invalid number of rows of the right hand side matrix b"); + return Solve(*this, b.derived()); + } + + template + bool solveInPlace(MatrixBase &bAndX) const; + + template + LDLT& compute(const EigenBase& matrix); + + /** \returns an estimate of the reciprocal condition number of the matrix of + * which \c *this is the LDLT decomposition. + */ + RealScalar rcond() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return internal::rcond_estimate_helper(m_l1_norm, *this); + } + + template + LDLT& rankUpdate(const MatrixBase& w, const RealScalar& alpha=1); + + /** \returns the internal LDLT decomposition matrix + * + * TODO: document the storage layout + */ + inline const MatrixType& matrixLDLT() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_matrix; + } + + MatrixType reconstructedMatrix() const; + + /** \returns the adjoint of \c *this, that is, a const reference to the decomposition itself as the underlying matrix is self-adjoint. + * + * This method is provided for compatibility with other matrix decompositions, thus enabling generic code such as: + * \code x = decomposition.adjoint().solve(b) \endcode + */ + const LDLT& adjoint() const { return *this; }; + + inline Index rows() const { return m_matrix.rows(); } + inline Index cols() const { return m_matrix.cols(); } + + /** \brief Reports whether previous computation was successful. + * + * \returns \c Success if computation was succesful, + * \c NumericalIssue if the factorization failed because of a zero pivot. + */ + ComputationInfo info() const + { + eigen_assert(m_isInitialized && "LDLT is not initialized."); + return m_info; + } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC + void _solve_impl(const RhsType &rhs, DstType &dst) const; + #endif + + protected: + + static void check_template_parameters() + { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); + } + + /** \internal + * Used to compute and store the Cholesky decomposition A = L D L^* = U^* D U. + * The strict upper part is used during the decomposition, the strict lower + * part correspond to the coefficients of L (its diagonal is equal to 1 and + * is not stored), and the diagonal entries correspond to D. + */ + MatrixType m_matrix; + RealScalar m_l1_norm; + TranspositionType m_transpositions; + TmpMatrixType m_temporary; + internal::SignMatrix m_sign; + bool m_isInitialized; + ComputationInfo m_info; +}; + +namespace internal { + +template struct ldlt_inplace; + +template<> struct ldlt_inplace +{ + template + static bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, SignMatrix& sign) + { + using std::abs; + typedef typename MatrixType::Scalar Scalar; + typedef typename MatrixType::RealScalar RealScalar; + typedef typename TranspositionType::StorageIndex IndexType; + eigen_assert(mat.rows()==mat.cols()); + const Index size = mat.rows(); + bool found_zero_pivot = false; + bool ret = true; + + if (size <= 1) + { + transpositions.setIdentity(); + if(size==0) sign = ZeroSign; + else if (numext::real(mat.coeff(0,0)) > static_cast(0) ) sign = PositiveSemiDef; + else if (numext::real(mat.coeff(0,0)) < static_cast(0)) sign = NegativeSemiDef; + else sign = ZeroSign; + return true; + } + + for (Index k = 0; k < size; ++k) + { + // Find largest diagonal element + Index index_of_biggest_in_corner; + mat.diagonal().tail(size-k).cwiseAbs().maxCoeff(&index_of_biggest_in_corner); + index_of_biggest_in_corner += k; + + transpositions.coeffRef(k) = IndexType(index_of_biggest_in_corner); + if(k != index_of_biggest_in_corner) + { + // apply the transposition while taking care to consider only + // the lower triangular part + Index s = size-index_of_biggest_in_corner-1; // trailing size after the biggest element + mat.row(k).head(k).swap(mat.row(index_of_biggest_in_corner).head(k)); + mat.col(k).tail(s).swap(mat.col(index_of_biggest_in_corner).tail(s)); + std::swap(mat.coeffRef(k,k),mat.coeffRef(index_of_biggest_in_corner,index_of_biggest_in_corner)); + for(Index i=k+1;i::IsComplex) + mat.coeffRef(index_of_biggest_in_corner,k) = numext::conj(mat.coeff(index_of_biggest_in_corner,k)); + } + + // partition the matrix: + // A00 | - | - + // lu = A10 | A11 | - + // A20 | A21 | A22 + Index rs = size - k - 1; + Block A21(mat,k+1,k,rs,1); + Block A10(mat,k,0,1,k); + Block A20(mat,k+1,0,rs,k); + + if(k>0) + { + temp.head(k) = mat.diagonal().real().head(k).asDiagonal() * A10.adjoint(); + mat.coeffRef(k,k) -= (A10 * temp.head(k)).value(); + if(rs>0) + A21.noalias() -= A20 * temp.head(k); + } + + // In some previous versions of Eigen (e.g., 3.2.1), the scaling was omitted if the pivot + // was smaller than the cutoff value. However, since LDLT is not rank-revealing + // we should only make sure that we do not introduce INF or NaN values. + // Remark that LAPACK also uses 0 as the cutoff value. + RealScalar realAkk = numext::real(mat.coeffRef(k,k)); + bool pivot_is_valid = (abs(realAkk) > RealScalar(0)); + + if(k==0 && !pivot_is_valid) + { + // The entire diagonal is zero, there is nothing more to do + // except filling the transpositions, and checking whether the matrix is zero. + sign = ZeroSign; + for(Index j = 0; j0) && pivot_is_valid) + A21 /= realAkk; + else if(rs>0) + ret = ret && (A21.array()==Scalar(0)).all(); + + if(found_zero_pivot && pivot_is_valid) ret = false; // factorization failed + else if(!pivot_is_valid) found_zero_pivot = true; + + if (sign == PositiveSemiDef) { + if (realAkk < static_cast(0)) sign = Indefinite; + } else if (sign == NegativeSemiDef) { + if (realAkk > static_cast(0)) sign = Indefinite; + } else if (sign == ZeroSign) { + if (realAkk > static_cast(0)) sign = PositiveSemiDef; + else if (realAkk < static_cast(0)) sign = NegativeSemiDef; + } + } + + return ret; + } + + // Reference for the algorithm: Davis and Hager, "Multiple Rank + // Modifications of a Sparse Cholesky Factorization" (Algorithm 1) + // Trivial rearrangements of their computations (Timothy E. Holy) + // allow their algorithm to work for rank-1 updates even if the + // original matrix is not of full rank. + // Here only rank-1 updates are implemented, to reduce the + // requirement for intermediate storage and improve accuracy + template + static bool updateInPlace(MatrixType& mat, MatrixBase& w, const typename MatrixType::RealScalar& sigma=1) + { + using numext::isfinite; + typedef typename MatrixType::Scalar Scalar; + typedef typename MatrixType::RealScalar RealScalar; + + const Index size = mat.rows(); + eigen_assert(mat.cols() == size && w.size()==size); + + RealScalar alpha = 1; + + // Apply the update + for (Index j = 0; j < size; j++) + { + // Check for termination due to an original decomposition of low-rank + if (!(isfinite)(alpha)) + break; + + // Update the diagonal terms + RealScalar dj = numext::real(mat.coeff(j,j)); + Scalar wj = w.coeff(j); + RealScalar swj2 = sigma*numext::abs2(wj); + RealScalar gamma = dj*alpha + swj2; + + mat.coeffRef(j,j) += swj2/alpha; + alpha += swj2/dj; + + + // Update the terms of L + Index rs = size-j-1; + w.tail(rs) -= wj * mat.col(j).tail(rs); + if(gamma != 0) + mat.col(j).tail(rs) += (sigma*numext::conj(wj)/gamma)*w.tail(rs); + } + return true; + } + + template + static bool update(MatrixType& mat, const TranspositionType& transpositions, Workspace& tmp, const WType& w, const typename MatrixType::RealScalar& sigma=1) + { + // Apply the permutation to the input w + tmp = transpositions * w; + + return ldlt_inplace::updateInPlace(mat,tmp,sigma); + } +}; + +template<> struct ldlt_inplace +{ + template + static EIGEN_STRONG_INLINE bool unblocked(MatrixType& mat, TranspositionType& transpositions, Workspace& temp, SignMatrix& sign) + { + Transpose matt(mat); + return ldlt_inplace::unblocked(matt, transpositions, temp, sign); + } + + template + static EIGEN_STRONG_INLINE bool update(MatrixType& mat, TranspositionType& transpositions, Workspace& tmp, WType& w, const typename MatrixType::RealScalar& sigma=1) + { + Transpose matt(mat); + return ldlt_inplace::update(matt, transpositions, tmp, w.conjugate(), sigma); + } +}; + +template struct LDLT_Traits +{ + typedef const TriangularView MatrixL; + typedef const TriangularView MatrixU; + static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); } + static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); } +}; + +template struct LDLT_Traits +{ + typedef const TriangularView MatrixL; + typedef const TriangularView MatrixU; + static inline MatrixL getL(const MatrixType& m) { return MatrixL(m.adjoint()); } + static inline MatrixU getU(const MatrixType& m) { return MatrixU(m); } +}; + +} // end namespace internal + +/** Compute / recompute the LDLT decomposition A = L D L^* = U^* D U of \a matrix + */ +template +template +LDLT& LDLT::compute(const EigenBase& a) +{ + check_template_parameters(); + + eigen_assert(a.rows()==a.cols()); + const Index size = a.rows(); + + m_matrix = a.derived(); + + // Compute matrix L1 norm = max abs column sum. + m_l1_norm = RealScalar(0); + // TODO move this code to SelfAdjointView + for (Index col = 0; col < size; ++col) { + RealScalar abs_col_sum; + if (_UpLo == Lower) + abs_col_sum = m_matrix.col(col).tail(size - col).template lpNorm<1>() + m_matrix.row(col).head(col).template lpNorm<1>(); + else + abs_col_sum = m_matrix.col(col).head(col).template lpNorm<1>() + m_matrix.row(col).tail(size - col).template lpNorm<1>(); + if (abs_col_sum > m_l1_norm) + m_l1_norm = abs_col_sum; + } + + m_transpositions.resize(size); + m_isInitialized = false; + m_temporary.resize(size); + m_sign = internal::ZeroSign; + + m_info = internal::ldlt_inplace::unblocked(m_matrix, m_transpositions, m_temporary, m_sign) ? Success : NumericalIssue; + + m_isInitialized = true; + return *this; +} + +/** Update the LDLT decomposition: given A = L D L^T, efficiently compute the decomposition of A + sigma w w^T. + * \param w a vector to be incorporated into the decomposition. + * \param sigma a scalar, +1 for updates and -1 for "downdates," which correspond to removing previously-added column vectors. Optional; default value is +1. + * \sa setZero() + */ +template +template +LDLT& LDLT::rankUpdate(const MatrixBase& w, const typename LDLT::RealScalar& sigma) +{ + typedef typename TranspositionType::StorageIndex IndexType; + const Index size = w.rows(); + if (m_isInitialized) + { + eigen_assert(m_matrix.rows()==size); + } + else + { + m_matrix.resize(size,size); + m_matrix.setZero(); + m_transpositions.resize(size); + for (Index i = 0; i < size; i++) + m_transpositions.coeffRef(i) = IndexType(i); + m_temporary.resize(size); + m_sign = sigma>=0 ? internal::PositiveSemiDef : internal::NegativeSemiDef; + m_isInitialized = true; + } + + internal::ldlt_inplace::update(m_matrix, m_transpositions, m_temporary, w, sigma); + + return *this; +} + +#ifndef EIGEN_PARSED_BY_DOXYGEN +template +template +void LDLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const +{ + eigen_assert(rhs.rows() == rows()); + // dst = P b + dst = m_transpositions * rhs; + + // dst = L^-1 (P b) + matrixL().solveInPlace(dst); + + // dst = D^-1 (L^-1 P b) + // more precisely, use pseudo-inverse of D (see bug 241) + using std::abs; + const typename Diagonal::RealReturnType vecD(vectorD()); + // In some previous versions, tolerance was set to the max of 1/highest (or rather numeric_limits::min()) + // and the maximal diagonal entry * epsilon as motivated by LAPACK's xGELSS: + // RealScalar tolerance = numext::maxi(vecD.array().abs().maxCoeff() * NumTraits::epsilon(),RealScalar(1) / NumTraits::highest()); + // However, LDLT is not rank revealing, and so adjusting the tolerance wrt to the highest + // diagonal element is not well justified and leads to numerical issues in some cases. + // Moreover, Lapack's xSYTRS routines use 0 for the tolerance. + // Using numeric_limits::min() gives us more robustness to denormals. + RealScalar tolerance = (std::numeric_limits::min)(); + + for (Index i = 0; i < vecD.size(); ++i) + { + if(abs(vecD(i)) > tolerance) + dst.row(i) /= vecD(i); + else + dst.row(i).setZero(); + } + + // dst = L^-T (D^-1 L^-1 P b) + matrixU().solveInPlace(dst); + + // dst = P^-1 (L^-T D^-1 L^-1 P b) = A^-1 b + dst = m_transpositions.transpose() * dst; +} +#endif + +/** \internal use x = ldlt_object.solve(x); + * + * This is the \em in-place version of solve(). + * + * \param bAndX represents both the right-hand side matrix b and result x. + * + * \returns true always! If you need to check for existence of solutions, use another decomposition like LU, QR, or SVD. + * + * This version avoids a copy when the right hand side matrix b is not + * needed anymore. + * + * \sa LDLT::solve(), MatrixBase::ldlt() + */ +template +template +bool LDLT::solveInPlace(MatrixBase &bAndX) const +{ + eigen_assert(m_isInitialized && "LDLT is not initialized."); + eigen_assert(m_matrix.rows() == bAndX.rows()); + + bAndX = this->solve(bAndX); + + return true; +} + +/** \returns the matrix represented by the decomposition, + * i.e., it returns the product: P^T L D L^* P. + * This function is provided for debug purpose. */ +template +MatrixType LDLT::reconstructedMatrix() const +{ + eigen_assert(m_isInitialized && "LDLT is not initialized."); + const Index size = m_matrix.rows(); + MatrixType res(size,size); + + // P + res.setIdentity(); + res = transpositionsP() * res; + // L^* P + res = matrixU() * res; + // D(L^*P) + res = vectorD().real().asDiagonal() * res; + // L(DL^*P) + res = matrixL() * res; + // P^T (LDL^*P) + res = transpositionsP().transpose() * res; + + return res; +} + +/** \cholesky_module + * \returns the Cholesky decomposition with full pivoting without square root of \c *this + * \sa MatrixBase::ldlt() + */ +template +inline const LDLT::PlainObject, UpLo> +SelfAdjointView::ldlt() const +{ + return LDLT(m_matrix); +} + +/** \cholesky_module + * \returns the Cholesky decomposition with full pivoting without square root of \c *this + * \sa SelfAdjointView::ldlt() + */ +template +inline const LDLT::PlainObject> +MatrixBase::ldlt() const +{ + return LDLT(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_LDLT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT.h new file mode 100644 index 00000000000..e1624d21b69 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT.h @@ -0,0 +1,542 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_LLT_H +#define EIGEN_LLT_H + +namespace Eigen { + +namespace internal{ +template struct LLT_Traits; +} + +/** \ingroup Cholesky_Module + * + * \class LLT + * + * \brief Standard Cholesky decomposition (LL^T) of a matrix and associated features + * + * \tparam _MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition + * \tparam _UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper. + * The other triangular part won't be read. + * + * This class performs a LL^T Cholesky decomposition of a symmetric, positive definite + * matrix A such that A = LL^* = U^*U, where L is lower triangular. + * + * While the Cholesky decomposition is particularly useful to solve selfadjoint problems like D^*D x = b, + * for that purpose, we recommend the Cholesky decomposition without square root which is more stable + * and even faster. Nevertheless, this standard Cholesky decomposition remains useful in many other + * situations like generalised eigen problems with hermitian matrices. + * + * Remember that Cholesky decompositions are not rank-revealing. This LLT decomposition is only stable on positive definite matrices, + * use LDLT instead for the semidefinite case. Also, do not use a Cholesky decomposition to determine whether a system of equations + * has a solution. + * + * Example: \include LLT_example.cpp + * Output: \verbinclude LLT_example.out + * + * \b Performance: for best performance, it is recommended to use a column-major storage format + * with the Lower triangular part (the default), or, equivalently, a row-major storage format + * with the Upper triangular part. Otherwise, you might get a 20% slowdown for the full factorization + * step, and rank-updates can be up to 3 times slower. + * + * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. + * + * Note that during the decomposition, only the lower (or upper, as defined by _UpLo) triangular part of A is considered. + * Therefore, the strict lower part does not have to store correct values. + * + * \sa MatrixBase::llt(), SelfAdjointView::llt(), class LDLT + */ +template class LLT +{ + public: + typedef _MatrixType MatrixType; + enum { + RowsAtCompileTime = MatrixType::RowsAtCompileTime, + ColsAtCompileTime = MatrixType::ColsAtCompileTime, + MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime + }; + typedef typename MatrixType::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 + typedef typename MatrixType::StorageIndex StorageIndex; + + enum { + PacketSize = internal::packet_traits::size, + AlignmentMask = int(PacketSize)-1, + UpLo = _UpLo + }; + + typedef internal::LLT_Traits Traits; + + /** + * \brief Default Constructor. + * + * The default constructor is useful in cases in which the user intends to + * perform decompositions via LLT::compute(const MatrixType&). + */ + LLT() : m_matrix(), m_isInitialized(false) {} + + /** \brief Default Constructor with memory preallocation + * + * Like the default constructor but with preallocation of the internal data + * according to the specified problem \a size. + * \sa LLT() + */ + explicit LLT(Index size) : m_matrix(size, size), + m_isInitialized(false) {} + + template + explicit LLT(const EigenBase& matrix) + : m_matrix(matrix.rows(), matrix.cols()), + m_isInitialized(false) + { + compute(matrix.derived()); + } + + /** \brief Constructs a LDLT factorization from a given matrix + * + * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when + * \c MatrixType is a Eigen::Ref. + * + * \sa LLT(const EigenBase&) + */ + template + explicit LLT(EigenBase& matrix) + : m_matrix(matrix.derived()), + m_isInitialized(false) + { + compute(matrix.derived()); + } + + /** \returns a view of the upper triangular matrix U */ + inline typename Traits::MatrixU matrixU() const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + return Traits::getU(m_matrix); + } + + /** \returns a view of the lower triangular matrix L */ + inline typename Traits::MatrixL matrixL() const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + return Traits::getL(m_matrix); + } + + /** \returns the solution x of \f$ A x = b \f$ using the current decomposition of A. + * + * Since this LLT class assumes anyway that the matrix A is invertible, the solution + * theoretically exists and is unique regardless of b. + * + * Example: \include LLT_solve.cpp + * Output: \verbinclude LLT_solve.out + * + * \sa solveInPlace(), MatrixBase::llt(), SelfAdjointView::llt() + */ + template + inline const Solve + solve(const MatrixBase& b) const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + eigen_assert(m_matrix.rows()==b.rows() + && "LLT::solve(): invalid number of rows of the right hand side matrix b"); + return Solve(*this, b.derived()); + } + + template + void solveInPlace(const MatrixBase &bAndX) const; + + template + LLT& compute(const EigenBase& matrix); + + /** \returns an estimate of the reciprocal condition number of the matrix of + * which \c *this is the Cholesky decomposition. + */ + RealScalar rcond() const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + eigen_assert(m_info == Success && "LLT failed because matrix appears to be negative"); + return internal::rcond_estimate_helper(m_l1_norm, *this); + } + + /** \returns the LLT decomposition matrix + * + * TODO: document the storage layout + */ + inline const MatrixType& matrixLLT() const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + return m_matrix; + } + + MatrixType reconstructedMatrix() const; + + + /** \brief Reports whether previous computation was successful. + * + * \returns \c Success if computation was succesful, + * \c NumericalIssue if the matrix.appears not to be positive definite. + */ + ComputationInfo info() const + { + eigen_assert(m_isInitialized && "LLT is not initialized."); + return m_info; + } + + /** \returns the adjoint of \c *this, that is, a const reference to the decomposition itself as the underlying matrix is self-adjoint. + * + * This method is provided for compatibility with other matrix decompositions, thus enabling generic code such as: + * \code x = decomposition.adjoint().solve(b) \endcode + */ + const LLT& adjoint() const { return *this; }; + + inline Index rows() const { return m_matrix.rows(); } + inline Index cols() const { return m_matrix.cols(); } + + template + LLT rankUpdate(const VectorType& vec, const RealScalar& sigma = 1); + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC + void _solve_impl(const RhsType &rhs, DstType &dst) const; + #endif + + protected: + + static void check_template_parameters() + { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); + } + + /** \internal + * Used to compute and store L + * The strict upper part is not used and even not initialized. + */ + MatrixType m_matrix; + RealScalar m_l1_norm; + bool m_isInitialized; + ComputationInfo m_info; +}; + +namespace internal { + +template struct llt_inplace; + +template +static Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) +{ + using std::sqrt; + typedef typename MatrixType::Scalar Scalar; + typedef typename MatrixType::RealScalar RealScalar; + typedef typename MatrixType::ColXpr ColXpr; + typedef typename internal::remove_all::type ColXprCleaned; + typedef typename ColXprCleaned::SegmentReturnType ColXprSegment; + typedef Matrix TempVectorType; + typedef typename TempVectorType::SegmentReturnType TempVecSegment; + + Index n = mat.cols(); + eigen_assert(mat.rows()==n && vec.size()==n); + + TempVectorType temp; + + if(sigma>0) + { + // This version is based on Givens rotations. + // It is faster than the other one below, but only works for updates, + // i.e., for sigma > 0 + temp = sqrt(sigma) * vec; + + for(Index i=0; i g; + g.makeGivens(mat(i,i), -temp(i), &mat(i,i)); + + Index rs = n-i-1; + if(rs>0) + { + ColXprSegment x(mat.col(i).tail(rs)); + TempVecSegment y(temp.tail(rs)); + apply_rotation_in_the_plane(x, y, g); + } + } + } + else + { + temp = vec; + RealScalar beta = 1; + for(Index j=0; j struct llt_inplace +{ + typedef typename NumTraits::Real RealScalar; + template + static Index unblocked(MatrixType& mat) + { + using std::sqrt; + + eigen_assert(mat.rows()==mat.cols()); + const Index size = mat.rows(); + for(Index k = 0; k < size; ++k) + { + Index rs = size-k-1; // remaining size + + Block A21(mat,k+1,k,rs,1); + Block A10(mat,k,0,1,k); + Block A20(mat,k+1,0,rs,k); + + RealScalar x = numext::real(mat.coeff(k,k)); + if (k>0) x -= A10.squaredNorm(); + if (x<=RealScalar(0)) + return k; + mat.coeffRef(k,k) = x = sqrt(x); + if (k>0 && rs>0) A21.noalias() -= A20 * A10.adjoint(); + if (rs>0) A21 /= x; + } + return -1; + } + + template + static Index blocked(MatrixType& m) + { + eigen_assert(m.rows()==m.cols()); + Index size = m.rows(); + if(size<32) + return unblocked(m); + + Index blockSize = size/8; + blockSize = (blockSize/16)*16; + blockSize = (std::min)((std::max)(blockSize,Index(8)), Index(128)); + + for (Index k=0; k A11(m,k, k, bs,bs); + Block A21(m,k+bs,k, rs,bs); + Block A22(m,k+bs,k+bs,rs,rs); + + Index ret; + if((ret=unblocked(A11))>=0) return k+ret; + if(rs>0) A11.adjoint().template triangularView().template solveInPlace(A21); + if(rs>0) A22.template selfadjointView().rankUpdate(A21,typename NumTraits::Literal(-1)); // bottleneck + } + return -1; + } + + template + static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma) + { + return Eigen::internal::llt_rank_update_lower(mat, vec, sigma); + } +}; + +template struct llt_inplace +{ + typedef typename NumTraits::Real RealScalar; + + template + static EIGEN_STRONG_INLINE Index unblocked(MatrixType& mat) + { + Transpose matt(mat); + return llt_inplace::unblocked(matt); + } + template + static EIGEN_STRONG_INLINE Index blocked(MatrixType& mat) + { + Transpose matt(mat); + return llt_inplace::blocked(matt); + } + template + static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma) + { + Transpose matt(mat); + return llt_inplace::rankUpdate(matt, vec.conjugate(), sigma); + } +}; + +template struct LLT_Traits +{ + typedef const TriangularView MatrixL; + typedef const TriangularView MatrixU; + static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); } + static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); } + static bool inplace_decomposition(MatrixType& m) + { return llt_inplace::blocked(m)==-1; } +}; + +template struct LLT_Traits +{ + typedef const TriangularView MatrixL; + typedef const TriangularView MatrixU; + static inline MatrixL getL(const MatrixType& m) { return MatrixL(m.adjoint()); } + static inline MatrixU getU(const MatrixType& m) { return MatrixU(m); } + static bool inplace_decomposition(MatrixType& m) + { return llt_inplace::blocked(m)==-1; } +}; + +} // end namespace internal + +/** Computes / recomputes the Cholesky decomposition A = LL^* = U^*U of \a matrix + * + * \returns a reference to *this + * + * Example: \include TutorialLinAlgComputeTwice.cpp + * Output: \verbinclude TutorialLinAlgComputeTwice.out + */ +template +template +LLT& LLT::compute(const EigenBase& a) +{ + check_template_parameters(); + + eigen_assert(a.rows()==a.cols()); + const Index size = a.rows(); + m_matrix.resize(size, size); + if (!internal::is_same_dense(m_matrix, a.derived())) + m_matrix = a.derived(); + + // Compute matrix L1 norm = max abs column sum. + m_l1_norm = RealScalar(0); + // TODO move this code to SelfAdjointView + for (Index col = 0; col < size; ++col) { + RealScalar abs_col_sum; + if (_UpLo == Lower) + abs_col_sum = m_matrix.col(col).tail(size - col).template lpNorm<1>() + m_matrix.row(col).head(col).template lpNorm<1>(); + else + abs_col_sum = m_matrix.col(col).head(col).template lpNorm<1>() + m_matrix.row(col).tail(size - col).template lpNorm<1>(); + if (abs_col_sum > m_l1_norm) + m_l1_norm = abs_col_sum; + } + + m_isInitialized = true; + bool ok = Traits::inplace_decomposition(m_matrix); + m_info = ok ? Success : NumericalIssue; + + return *this; +} + +/** Performs a rank one update (or dowdate) of the current decomposition. + * If A = LL^* before the rank one update, + * then after it we have LL^* = A + sigma * v v^* where \a v must be a vector + * of same dimension. + */ +template +template +LLT<_MatrixType,_UpLo> LLT<_MatrixType,_UpLo>::rankUpdate(const VectorType& v, const RealScalar& sigma) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorType); + eigen_assert(v.size()==m_matrix.cols()); + eigen_assert(m_isInitialized); + if(internal::llt_inplace::rankUpdate(m_matrix,v,sigma)>=0) + m_info = NumericalIssue; + else + m_info = Success; + + return *this; +} + +#ifndef EIGEN_PARSED_BY_DOXYGEN +template +template +void LLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const +{ + dst = rhs; + solveInPlace(dst); +} +#endif + +/** \internal use x = llt_object.solve(x); + * + * This is the \em in-place version of solve(). + * + * \param bAndX represents both the right-hand side matrix b and result x. + * + * This version avoids a copy when the right hand side matrix b is not needed anymore. + * + * \warning The parameter is only marked 'const' to make the C++ compiler accept a temporary expression here. + * This function will const_cast it, so constness isn't honored here. + * + * \sa LLT::solve(), MatrixBase::llt() + */ +template +template +void LLT::solveInPlace(const MatrixBase &bAndX) const +{ + eigen_assert(m_isInitialized && "LLT is not initialized."); + eigen_assert(m_matrix.rows()==bAndX.rows()); + matrixL().solveInPlace(bAndX); + matrixU().solveInPlace(bAndX); +} + +/** \returns the matrix represented by the decomposition, + * i.e., it returns the product: L L^*. + * This function is provided for debug purpose. */ +template +MatrixType LLT::reconstructedMatrix() const +{ + eigen_assert(m_isInitialized && "LLT is not initialized."); + return matrixL() * matrixL().adjoint().toDenseMatrix(); +} + +/** \cholesky_module + * \returns the LLT decomposition of \c *this + * \sa SelfAdjointView::llt() + */ +template +inline const LLT::PlainObject> +MatrixBase::llt() const +{ + return LLT(derived()); +} + +/** \cholesky_module + * \returns the LLT decomposition of \c *this + * \sa SelfAdjointView::llt() + */ +template +inline const LLT::PlainObject, UpLo> +SelfAdjointView::llt() const +{ + return LLT(m_matrix); +} + +} // end namespace Eigen + +#endif // EIGEN_LLT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT_LAPACKE.h new file mode 100644 index 00000000000..bc6489e69a9 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Cholesky/LLT_LAPACKE.h @@ -0,0 +1,99 @@ +/* + Copyright (c) 2011, Intel Corporation. All rights reserved. + + Redistribution and use in source and binary forms, with or without modification, + are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + * Neither the name of Intel Corporation nor the names of its contributors may + be used to endorse or promote products derived from this software without + specific prior written permission. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND + ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR + ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON + ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + ******************************************************************************** + * Content : Eigen bindings to LAPACKe + * LLt decomposition based on LAPACKE_?potrf function. + ******************************************************************************** +*/ + +#ifndef EIGEN_LLT_LAPACKE_H +#define EIGEN_LLT_LAPACKE_H + +namespace Eigen { + +namespace internal { + +template struct lapacke_llt; + +#define EIGEN_LAPACKE_LLT(EIGTYPE, BLASTYPE, LAPACKE_PREFIX) \ +template<> struct lapacke_llt \ +{ \ + template \ + static inline Index potrf(MatrixType& m, char uplo) \ + { \ + lapack_int matrix_order; \ + lapack_int size, lda, info, StorageOrder; \ + EIGTYPE* a; \ + eigen_assert(m.rows()==m.cols()); \ + /* Set up parameters for ?potrf */ \ + size = convert_index(m.rows()); \ + StorageOrder = MatrixType::Flags&RowMajorBit?RowMajor:ColMajor; \ + matrix_order = StorageOrder==RowMajor ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \ + a = &(m.coeffRef(0,0)); \ + lda = convert_index(m.outerStride()); \ +\ + info = LAPACKE_##LAPACKE_PREFIX##potrf( matrix_order, uplo, size, (BLASTYPE*)a, lda ); \ + info = (info==0) ? -1 : info>0 ? info-1 : size; \ + return info; \ + } \ +}; \ +template<> struct llt_inplace \ +{ \ + template \ + static Index blocked(MatrixType& m) \ + { \ + return lapacke_llt::potrf(m, 'L'); \ + } \ + template \ + static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \ + { return Eigen::internal::llt_rank_update_lower(mat, vec, sigma); } \ +}; \ +template<> struct llt_inplace \ +{ \ + template \ + static Index blocked(MatrixType& m) \ + { \ + return lapacke_llt::potrf(m, 'U'); \ + } \ + template \ + static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \ + { \ + Transpose matt(mat); \ + return llt_inplace::rankUpdate(matt, vec.conjugate(), sigma); \ + } \ +}; + +EIGEN_LAPACKE_LLT(double, double, d) +EIGEN_LAPACKE_LLT(float, float, s) +EIGEN_LAPACKE_LLT(dcomplex, lapack_complex_double, z) +EIGEN_LAPACKE_LLT(scomplex, lapack_complex_float, c) + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_LLT_LAPACKE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/CholmodSupport/CholmodSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/CholmodSupport/CholmodSupport.h new file mode 100644 index 00000000000..57197202383 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/CholmodSupport/CholmodSupport.h @@ -0,0 +1,639 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CHOLMODSUPPORT_H +#define EIGEN_CHOLMODSUPPORT_H + +namespace Eigen { + +namespace internal { + +template struct cholmod_configure_matrix; + +template<> struct cholmod_configure_matrix { + template + static void run(CholmodType& mat) { + mat.xtype = CHOLMOD_REAL; + mat.dtype = CHOLMOD_DOUBLE; + } +}; + +template<> struct cholmod_configure_matrix > { + template + static void run(CholmodType& mat) { + mat.xtype = CHOLMOD_COMPLEX; + mat.dtype = CHOLMOD_DOUBLE; + } +}; + +// Other scalar types are not yet suppotred by Cholmod +// template<> struct cholmod_configure_matrix { +// template +// static void run(CholmodType& mat) { +// mat.xtype = CHOLMOD_REAL; +// mat.dtype = CHOLMOD_SINGLE; +// } +// }; +// +// template<> struct cholmod_configure_matrix > { +// template +// static void run(CholmodType& mat) { +// mat.xtype = CHOLMOD_COMPLEX; +// mat.dtype = CHOLMOD_SINGLE; +// } +// }; + +} // namespace internal + +/** Wraps the Eigen sparse matrix \a mat into a Cholmod sparse matrix object. + * Note that the data are shared. + */ +template +cholmod_sparse viewAsCholmod(Ref > mat) +{ + cholmod_sparse res; + res.nzmax = mat.nonZeros(); + res.nrow = mat.rows(); + res.ncol = mat.cols(); + res.p = mat.outerIndexPtr(); + res.i = mat.innerIndexPtr(); + res.x = mat.valuePtr(); + res.z = 0; + res.sorted = 1; + if(mat.isCompressed()) + { + res.packed = 1; + res.nz = 0; + } + else + { + res.packed = 0; + res.nz = mat.innerNonZeroPtr(); + } + + res.dtype = 0; + res.stype = -1; + + if (internal::is_same<_StorageIndex,int>::value) + { + res.itype = CHOLMOD_INT; + } + else if (internal::is_same<_StorageIndex,long>::value) + { + res.itype = CHOLMOD_LONG; + } + else + { + eigen_assert(false && "Index type not supported yet"); + } + + // setup res.xtype + internal::cholmod_configure_matrix<_Scalar>::run(res); + + res.stype = 0; + + return res; +} + +template +const cholmod_sparse viewAsCholmod(const SparseMatrix<_Scalar,_Options,_Index>& mat) +{ + cholmod_sparse res = viewAsCholmod(Ref >(mat.const_cast_derived())); + return res; +} + +template +const cholmod_sparse viewAsCholmod(const SparseVector<_Scalar,_Options,_Index>& mat) +{ + cholmod_sparse res = viewAsCholmod(Ref >(mat.const_cast_derived())); + return res; +} + +/** Returns a view of the Eigen sparse matrix \a mat as Cholmod sparse matrix. + * The data are not copied but shared. */ +template +cholmod_sparse viewAsCholmod(const SparseSelfAdjointView, UpLo>& mat) +{ + cholmod_sparse res = viewAsCholmod(Ref >(mat.matrix().const_cast_derived())); + + if(UpLo==Upper) res.stype = 1; + if(UpLo==Lower) res.stype = -1; + + return res; +} + +/** Returns a view of the Eigen \b dense matrix \a mat as Cholmod dense matrix. + * The data are not copied but shared. */ +template +cholmod_dense viewAsCholmod(MatrixBase& mat) +{ + EIGEN_STATIC_ASSERT((internal::traits::Flags&RowMajorBit)==0,THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); + typedef typename Derived::Scalar Scalar; + + cholmod_dense res; + res.nrow = mat.rows(); + res.ncol = mat.cols(); + res.nzmax = res.nrow * res.ncol; + res.d = Derived::IsVectorAtCompileTime ? mat.derived().size() : mat.derived().outerStride(); + res.x = (void*)(mat.derived().data()); + res.z = 0; + + internal::cholmod_configure_matrix::run(res); + + return res; +} + +/** Returns a view of the Cholmod sparse matrix \a cm as an Eigen sparse matrix. + * The data are not copied but shared. */ +template +MappedSparseMatrix viewAsEigen(cholmod_sparse& cm) +{ + return MappedSparseMatrix + (cm.nrow, cm.ncol, static_cast(cm.p)[cm.ncol], + static_cast(cm.p), static_cast(cm.i),static_cast(cm.x) ); +} + +enum CholmodMode { + CholmodAuto, CholmodSimplicialLLt, CholmodSupernodalLLt, CholmodLDLt +}; + + +/** \ingroup CholmodSupport_Module + * \class CholmodBase + * \brief The base class for the direct Cholesky factorization of Cholmod + * \sa class CholmodSupernodalLLT, class CholmodSimplicialLDLT, class CholmodSimplicialLLT + */ +template +class CholmodBase : public SparseSolverBase +{ + protected: + typedef SparseSolverBase Base; + using Base::derived; + using Base::m_isInitialized; + public: + typedef _MatrixType MatrixType; + enum { UpLo = _UpLo }; + typedef typename MatrixType::Scalar Scalar; + typedef typename MatrixType::RealScalar RealScalar; + typedef MatrixType CholMatrixType; + typedef typename MatrixType::StorageIndex StorageIndex; + enum { + ColsAtCompileTime = MatrixType::ColsAtCompileTime, + MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime + }; + + public: + + CholmodBase() + : m_cholmodFactor(0), m_info(Success), m_factorizationIsOk(false), m_analysisIsOk(false) + { + EIGEN_STATIC_ASSERT((internal::is_same::value), CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY); + m_shiftOffset[0] = m_shiftOffset[1] = 0.0; + cholmod_start(&m_cholmod); + } + + explicit CholmodBase(const MatrixType& matrix) + : m_cholmodFactor(0), m_info(Success), m_factorizationIsOk(false), m_analysisIsOk(false) + { + EIGEN_STATIC_ASSERT((internal::is_same::value), CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY); + m_shiftOffset[0] = m_shiftOffset[1] = 0.0; + cholmod_start(&m_cholmod); + compute(matrix); + } + + ~CholmodBase() + { + if(m_cholmodFactor) + cholmod_free_factor(&m_cholmodFactor, &m_cholmod); + cholmod_finish(&m_cholmod); + } + + inline StorageIndex cols() const { return internal::convert_index(m_cholmodFactor->n); } + inline StorageIndex rows() const { return internal::convert_index(m_cholmodFactor->n); } + + /** \brief Reports whether previous computation was successful. + * + * \returns \c Success if computation was succesful, + * \c NumericalIssue if the matrix.appears to be negative. + */ + ComputationInfo info() const + { + eigen_assert(m_isInitialized && "Decomposition is not initialized."); + return m_info; + } + + /** Computes the sparse Cholesky decomposition of \a matrix */ + Derived& compute(const MatrixType& matrix) + { + analyzePattern(matrix); + factorize(matrix); + return derived(); + } + + /** Performs a symbolic decomposition on the sparsity pattern of \a matrix. + * + * This function is particularly useful when solving for several problems having the same structure. + * + * \sa factorize() + */ + void analyzePattern(const MatrixType& matrix) + { + if(m_cholmodFactor) + { + cholmod_free_factor(&m_cholmodFactor, &m_cholmod); + m_cholmodFactor = 0; + } + cholmod_sparse A = viewAsCholmod(matrix.template selfadjointView()); + m_cholmodFactor = cholmod_analyze(&A, &m_cholmod); + + this->m_isInitialized = true; + this->m_info = Success; + m_analysisIsOk = true; + m_factorizationIsOk = false; + } + + /** Performs a numeric decomposition of \a matrix + * + * The given matrix must have the same sparsity pattern as the matrix on which the symbolic decomposition has been performed. + * + * \sa analyzePattern() + */ + void factorize(const MatrixType& matrix) + { + eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); + cholmod_sparse A = viewAsCholmod(matrix.template selfadjointView()); + cholmod_factorize_p(&A, m_shiftOffset, 0, 0, m_cholmodFactor, &m_cholmod); + + // If the factorization failed, minor is the column at which it did. On success minor == n. + this->m_info = (m_cholmodFactor->minor == m_cholmodFactor->n ? Success : NumericalIssue); + m_factorizationIsOk = true; + } + + /** Returns a reference to the Cholmod's configuration structure to get a full control over the performed operations. + * See the Cholmod user guide for details. */ + cholmod_common& cholmod() { return m_cholmod; } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal */ + template + void _solve_impl(const MatrixBase &b, MatrixBase &dest) const + { + eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()"); + const Index size = m_cholmodFactor->n; + EIGEN_UNUSED_VARIABLE(size); + eigen_assert(size==b.rows()); + + // Cholmod needs column-major stoarge without inner-stride, which corresponds to the default behavior of Ref. + Ref > b_ref(b.derived()); + + cholmod_dense b_cd = viewAsCholmod(b_ref); + cholmod_dense* x_cd = cholmod_solve(CHOLMOD_A, m_cholmodFactor, &b_cd, &m_cholmod); + if(!x_cd) + { + this->m_info = NumericalIssue; + return; + } + // TODO optimize this copy by swapping when possible (be careful with alignment, etc.) + dest = Matrix::Map(reinterpret_cast(x_cd->x),b.rows(),b.cols()); + cholmod_free_dense(&x_cd, &m_cholmod); + } + + /** \internal */ + template + void _solve_impl(const SparseMatrixBase &b, SparseMatrixBase &dest) const + { + eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()"); + const Index size = m_cholmodFactor->n; + EIGEN_UNUSED_VARIABLE(size); + eigen_assert(size==b.rows()); + + // note: cs stands for Cholmod Sparse + Ref > b_ref(b.const_cast_derived()); + cholmod_sparse b_cs = viewAsCholmod(b_ref); + cholmod_sparse* x_cs = cholmod_spsolve(CHOLMOD_A, m_cholmodFactor, &b_cs, &m_cholmod); + if(!x_cs) + { + this->m_info = NumericalIssue; + return; + } + // TODO optimize this copy by swapping when possible (be careful with alignment, etc.) + dest.derived() = viewAsEigen(*x_cs); + cholmod_free_sparse(&x_cs, &m_cholmod); + } + #endif // EIGEN_PARSED_BY_DOXYGEN + + + /** Sets the shift parameter that will be used to adjust the diagonal coefficients during the numerical factorization. + * + * During the numerical factorization, an offset term is added to the diagonal coefficients:\n + * \c d_ii = \a offset + \c d_ii + * + * The default is \a offset=0. + * + * \returns a reference to \c *this. + */ + Derived& setShift(const RealScalar& offset) + { + m_shiftOffset[0] = double(offset); + return derived(); + } + + /** \returns the determinant of the underlying matrix from the current factorization */ + Scalar determinant() const + { + using std::exp; + return exp(logDeterminant()); + } + + /** \returns the log determinant of the underlying matrix from the current factorization */ + Scalar logDeterminant() const + { + using std::log; + using numext::real; + eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()"); + + RealScalar logDet = 0; + Scalar *x = static_cast(m_cholmodFactor->x); + if (m_cholmodFactor->is_super) + { + // Supernodal factorization stored as a packed list of dense column-major blocs, + // as described by the following structure: + + // super[k] == index of the first column of the j-th super node + StorageIndex *super = static_cast(m_cholmodFactor->super); + // pi[k] == offset to the description of row indices + StorageIndex *pi = static_cast(m_cholmodFactor->pi); + // px[k] == offset to the respective dense block + StorageIndex *px = static_cast(m_cholmodFactor->px); + + Index nb_super_nodes = m_cholmodFactor->nsuper; + for (Index k=0; k < nb_super_nodes; ++k) + { + StorageIndex ncols = super[k + 1] - super[k]; + StorageIndex nrows = pi[k + 1] - pi[k]; + + Map, 0, InnerStride<> > sk(x + px[k], ncols, InnerStride<>(nrows+1)); + logDet += sk.real().log().sum(); + } + } + else + { + // Simplicial factorization stored as standard CSC matrix. + StorageIndex *p = static_cast(m_cholmodFactor->p); + Index size = m_cholmodFactor->n; + for (Index k=0; kis_ll) + logDet *= 2.0; + return logDet; + }; + + template + void dumpMemory(Stream& /*s*/) + {} + + protected: + mutable cholmod_common m_cholmod; + cholmod_factor* m_cholmodFactor; + double m_shiftOffset[2]; + mutable ComputationInfo m_info; + int m_factorizationIsOk; + int m_analysisIsOk; +}; + +/** \ingroup CholmodSupport_Module + * \class CholmodSimplicialLLT + * \brief A simplicial direct Cholesky (LLT) factorization and solver based on Cholmod + * + * This class allows to solve for A.X = B sparse linear problems via a simplicial LL^T Cholesky factorization + * using the Cholmod library. + * This simplicial variant is equivalent to Eigen's built-in SimplicialLLT class. Therefore, it has little practical interest. + * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices + * X and B can be either dense or sparse. + * + * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> + * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower + * or Upper. Default is Lower. + * + * \implsparsesolverconcept + * + * This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed. + * + * \warning Only double precision real and complex scalar types are supported by Cholmod. + * + * \sa \ref TutorialSparseSolverConcept, class CholmodSupernodalLLT, class SimplicialLLT + */ +template +class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLLT<_MatrixType, _UpLo> > +{ + typedef CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLLT> Base; + using Base::m_cholmod; + + public: + + typedef _MatrixType MatrixType; + + CholmodSimplicialLLT() : Base() { init(); } + + CholmodSimplicialLLT(const MatrixType& matrix) : Base() + { + init(); + this->compute(matrix); + } + + ~CholmodSimplicialLLT() {} + protected: + void init() + { + m_cholmod.final_asis = 0; + m_cholmod.supernodal = CHOLMOD_SIMPLICIAL; + m_cholmod.final_ll = 1; + } +}; + + +/** \ingroup CholmodSupport_Module + * \class CholmodSimplicialLDLT + * \brief A simplicial direct Cholesky (LDLT) factorization and solver based on Cholmod + * + * This class allows to solve for A.X = B sparse linear problems via a simplicial LDL^T Cholesky factorization + * using the Cholmod library. + * This simplicial variant is equivalent to Eigen's built-in SimplicialLDLT class. Therefore, it has little practical interest. + * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices + * X and B can be either dense or sparse. + * + * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> + * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower + * or Upper. Default is Lower. + * + * \implsparsesolverconcept + * + * This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed. + * + * \warning Only double precision real and complex scalar types are supported by Cholmod. + * + * \sa \ref TutorialSparseSolverConcept, class CholmodSupernodalLLT, class SimplicialLDLT + */ +template +class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLDLT<_MatrixType, _UpLo> > +{ + typedef CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLDLT> Base; + using Base::m_cholmod; + + public: + + typedef _MatrixType MatrixType; + + CholmodSimplicialLDLT() : Base() { init(); } + + CholmodSimplicialLDLT(const MatrixType& matrix) : Base() + { + init(); + this->compute(matrix); + } + + ~CholmodSimplicialLDLT() {} + protected: + void init() + { + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_SIMPLICIAL; + } +}; + +/** \ingroup CholmodSupport_Module + * \class CholmodSupernodalLLT + * \brief A supernodal Cholesky (LLT) factorization and solver based on Cholmod + * + * This class allows to solve for A.X = B sparse linear problems via a supernodal LL^T Cholesky factorization + * using the Cholmod library. + * This supernodal variant performs best on dense enough problems, e.g., 3D FEM, or very high order 2D FEM. + * The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices + * X and B can be either dense or sparse. + * + * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> + * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower + * or Upper. Default is Lower. + * + * \implsparsesolverconcept + * + * This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed. + * + * \warning Only double precision real and complex scalar types are supported by Cholmod. + * + * \sa \ref TutorialSparseSolverConcept + */ +template +class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSupernodalLLT<_MatrixType, _UpLo> > +{ + typedef CholmodBase<_MatrixType, _UpLo, CholmodSupernodalLLT> Base; + using Base::m_cholmod; + + public: + + typedef _MatrixType MatrixType; + + CholmodSupernodalLLT() : Base() { init(); } + + CholmodSupernodalLLT(const MatrixType& matrix) : Base() + { + init(); + this->compute(matrix); + } + + ~CholmodSupernodalLLT() {} + protected: + void init() + { + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_SUPERNODAL; + } +}; + +/** \ingroup CholmodSupport_Module + * \class CholmodDecomposition + * \brief A general Cholesky factorization and solver based on Cholmod + * + * This class allows to solve for A.X = B sparse linear problems via a LL^T or LDL^T Cholesky factorization + * using the Cholmod library. The sparse matrix A must be selfadjoint and positive definite. The vectors or matrices + * X and B can be either dense or sparse. + * + * This variant permits to change the underlying Cholesky method at runtime. + * On the other hand, it does not provide access to the result of the factorization. + * The default is to let Cholmod automatically choose between a simplicial and supernodal factorization. + * + * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> + * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower + * or Upper. Default is Lower. + * + * \implsparsesolverconcept + * + * This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed. + * + * \warning Only double precision real and complex scalar types are supported by Cholmod. + * + * \sa \ref TutorialSparseSolverConcept + */ +template +class CholmodDecomposition : public CholmodBase<_MatrixType, _UpLo, CholmodDecomposition<_MatrixType, _UpLo> > +{ + typedef CholmodBase<_MatrixType, _UpLo, CholmodDecomposition> Base; + using Base::m_cholmod; + + public: + + typedef _MatrixType MatrixType; + + CholmodDecomposition() : Base() { init(); } + + CholmodDecomposition(const MatrixType& matrix) : Base() + { + init(); + this->compute(matrix); + } + + ~CholmodDecomposition() {} + + void setMode(CholmodMode mode) + { + switch(mode) + { + case CholmodAuto: + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_AUTO; + break; + case CholmodSimplicialLLt: + m_cholmod.final_asis = 0; + m_cholmod.supernodal = CHOLMOD_SIMPLICIAL; + m_cholmod.final_ll = 1; + break; + case CholmodSupernodalLLt: + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_SUPERNODAL; + break; + case CholmodLDLt: + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_SIMPLICIAL; + break; + default: + break; + } + } + protected: + void init() + { + m_cholmod.final_asis = 1; + m_cholmod.supernodal = CHOLMOD_AUTO; + } +}; + +} // end namespace Eigen + +#endif // EIGEN_CHOLMODSUPPORT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Array.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Array.h new file mode 100644 index 00000000000..16770fc7b3a --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Array.h @@ -0,0 +1,329 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAY_H +#define EIGEN_ARRAY_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits > +{ + typedef ArrayXpr XprKind; + typedef ArrayBase > XprBase; +}; +} + +/** \class Array + * \ingroup Core_Module + * + * \brief General-purpose arrays with easy API for coefficient-wise operations + * + * The %Array class is very similar to the Matrix class. It provides + * general-purpose one- and two-dimensional arrays. The difference between the + * %Array and the %Matrix class is primarily in the API: the API for the + * %Array class provides easy access to coefficient-wise operations, while the + * API for the %Matrix class provides easy access to linear-algebra + * operations. + * + * See documentation of class Matrix for detailed information on the template parameters + * storage layout. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN. + * + * \sa \blank \ref TutorialArrayClass, \ref TopicClassHierarchy + */ +template +class Array + : public PlainObjectBase > +{ + public: + + typedef PlainObjectBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Array) + + enum { Options = _Options }; + typedef typename Base::PlainObject PlainObject; + + protected: + template + friend struct internal::conservative_resize_like_impl; + + using Base::m_storage; + + public: + + using Base::base; + using Base::coeff; + using Base::coeffRef; + + /** + * The usage of + * using Base::operator=; + * fails on MSVC. Since the code below is working with GCC and MSVC, we skipped + * the usage of 'using'. This should be done only for operator=. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const EigenBase &other) + { + return Base::operator=(other); + } + + /** Set all the entries to \a value. + * \sa DenseBase::setConstant(), DenseBase::fill() + */ + /* This overload is needed because the usage of + * using Base::operator=; + * fails on MSVC. Since the code below is working with GCC and MSVC, we skipped + * the usage of 'using'. This should be done only for operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const Scalar &value) + { + Base::setConstant(value); + return *this; + } + + /** Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const DenseBase& other) + { + return Base::_set(other); + } + + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const Array& other) + { + return Base::_set(other); + } + + /** Default constructor. + * + * For fixed-size matrices, does nothing. + * + * For dynamic-size matrices, creates an empty matrix of size 0. Does not allocate any array. Such a matrix + * is called a null matrix. This constructor is the unique way to create null matrices: resizing + * a matrix to 0 is not supported. + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array() : Base() + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + // FIXME is it still needed ?? + /** \internal */ + EIGEN_DEVICE_FUNC + Array(internal::constructor_without_unaligned_array_assert) + : Base(internal::constructor_without_unaligned_array_assert()) + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } +#endif + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + Array(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible::value) + : Base(std::move(other)) + { + Base::_check_template_params(); + } + EIGEN_DEVICE_FUNC + Array& operator=(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable::value) + { + other.swap(*this); + return *this; + } +#endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Array(const T& x) + { + Base::_check_template_params(); + Base::template _init1(x); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const T0& val0, const T1& val1) + { + Base::_check_template_params(); + this->template _init2(val0, val1); + } + #else + /** \brief Constructs a fixed-sized array initialized with coefficients starting at \a data */ + EIGEN_DEVICE_FUNC explicit Array(const Scalar *data); + /** Constructs a vector or row-vector with given dimension. \only_for_vectors + * + * Note that this is only useful for dynamic-size vectors. For fixed-size vectors, + * it is redundant to pass the dimension here, so it makes more sense to use the default + * constructor Array() instead. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Array(Index dim); + /** constructs an initialized 1x1 Array with the given coefficient */ + Array(const Scalar& value); + /** constructs an uninitialized array with \a rows rows and \a cols columns. + * + * This is useful for dynamic-size arrays. For fixed-size arrays, + * it is redundant to pass these parameters, so one should use the default constructor + * Array() instead. */ + Array(Index rows, Index cols); + /** constructs an initialized 2D vector with given coefficients */ + Array(const Scalar& val0, const Scalar& val1); + #endif + + /** constructs an initialized 3D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 3) + m_storage.data()[0] = val0; + m_storage.data()[1] = val1; + m_storage.data()[2] = val2; + } + /** constructs an initialized 4D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2, const Scalar& val3) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 4) + m_storage.data()[0] = val0; + m_storage.data()[1] = val1; + m_storage.data()[2] = val2; + m_storage.data()[3] = val3; + } + + /** Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Array& other) + : Base(other) + { } + + private: + struct PrivateType {}; + public: + + /** \sa MatrixBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const EigenBase &other, + typename internal::enable_if::value, + PrivateType>::type = PrivateType()) + : Base(other.derived()) + { } + + EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); } + + #ifdef EIGEN_ARRAY_PLUGIN + #include EIGEN_ARRAY_PLUGIN + #endif + + private: + + template + friend struct internal::matrix_swap_impl; +}; + +/** \defgroup arraytypedefs Global array typedefs + * \ingroup Core_Module + * + * Eigen defines several typedef shortcuts for most common 1D and 2D array types. + * + * The general patterns are the following: + * + * \c ArrayRowsColsType where \c Rows and \c Cols can be \c 2,\c 3,\c 4 for fixed size square matrices or \c X for dynamic size, + * and where \c Type can be \c i for integer, \c f for float, \c d for double, \c cf for complex float, \c cd + * for complex double. + * + * For example, \c Array33d is a fixed-size 3x3 array type of doubles, and \c ArrayXXf is a dynamic-size matrix of floats. + * + * There are also \c ArraySizeType which are self-explanatory. For example, \c Array4cf is + * a fixed-size 1D array of 4 complex floats. + * + * \sa class Array + */ + +#define EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, Size, SizeSuffix) \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##SizeSuffix##SizeSuffix##TypeSuffix; \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##SizeSuffix##TypeSuffix; + +#define EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, Size) \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##Size##X##TypeSuffix; \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##X##Size##TypeSuffix; + +#define EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(Type, TypeSuffix) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 2, 2) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 3, 3) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 4, 4) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, Dynamic, X) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 2) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 3) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 4) + +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(int, i) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(float, f) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(double, d) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(std::complex, cf) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(std::complex, cd) + +#undef EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES +#undef EIGEN_MAKE_ARRAY_TYPEDEFS + +#undef EIGEN_MAKE_ARRAY_TYPEDEFS_LARGE + +#define EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, SizeSuffix) \ +using Eigen::Matrix##SizeSuffix##TypeSuffix; \ +using Eigen::Vector##SizeSuffix##TypeSuffix; \ +using Eigen::RowVector##SizeSuffix##TypeSuffix; + +#define EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(TypeSuffix) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 2) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 3) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 4) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, X) \ + +#define EIGEN_USING_ARRAY_TYPEDEFS \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(i) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(f) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(d) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(cf) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(cd) + +} // end namespace Eigen + +#endif // EIGEN_ARRAY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h new file mode 100644 index 00000000000..33f644e2132 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h @@ -0,0 +1,226 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAYBASE_H +#define EIGEN_ARRAYBASE_H + +namespace Eigen { + +template class MatrixWrapper; + +/** \class ArrayBase + * \ingroup Core_Module + * + * \brief Base class for all 1D and 2D array, and related expressions + * + * An array is similar to a dense vector or matrix. While matrices are mathematical + * objects with well defined linear algebra operators, an array is just a collection + * of scalar values arranged in a one or two dimensionnal fashion. As the main consequence, + * all operations applied to an array are performed coefficient wise. Furthermore, + * arrays support scalar math functions of the c++ standard library (e.g., std::sin(x)), and convenient + * constructors allowing to easily write generic code working for both scalar values + * and arrays. + * + * This class is the base that is inherited by all array expression types. + * + * \tparam Derived is the derived type, e.g., an array or an expression type. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN. + * + * \sa class MatrixBase, \ref TopicClassHierarchy + */ +template class ArrayBase + : public DenseBase +{ + public: +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** The base class for a given storage type. */ + typedef ArrayBase StorageBaseType; + + typedef ArrayBase Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + typedef DenseBase Base; + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::operator=; + using Base::operator+=; + using Base::operator-=; + using Base::operator*=; + using Base::operator/=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Base::PlainObject PlainObject; + + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::ArrayBase +#define EIGEN_DOC_UNARY_ADDONS(X,Y) +# include "../plugins/CommonCwiseUnaryOps.h" +# include "../plugins/MatrixCwiseUnaryOps.h" +# include "../plugins/ArrayCwiseUnaryOps.h" +# include "../plugins/CommonCwiseBinaryOps.h" +# include "../plugins/MatrixCwiseBinaryOps.h" +# include "../plugins/ArrayCwiseBinaryOps.h" +# ifdef EIGEN_ARRAYBASE_PLUGIN +# include EIGEN_ARRAYBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_UNARY_ADDONS + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const ArrayBase& other) + { + internal::call_assignment(derived(), other.derived()); + return derived(); + } + + /** Set all the entries to \a value. + * \sa DenseBase::setConstant(), DenseBase::fill() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const Scalar &value) + { Base::setConstant(value); return derived(); } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const Scalar& scalar); + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const Scalar& scalar); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const ArrayBase& other); + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const ArrayBase& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator*=(const ArrayBase& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator/=(const ArrayBase& other); + + public: + EIGEN_DEVICE_FUNC + ArrayBase& array() { return *this; } + EIGEN_DEVICE_FUNC + const ArrayBase& array() const { return *this; } + + /** \returns an \link Eigen::MatrixBase Matrix \endlink expression of this array + * \sa MatrixBase::array() */ + EIGEN_DEVICE_FUNC + MatrixWrapper matrix() { return MatrixWrapper(derived()); } + EIGEN_DEVICE_FUNC + const MatrixWrapper matrix() const { return MatrixWrapper(derived()); } + +// template +// inline void evalTo(Dest& dst) const { dst = matrix(); } + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(ArrayBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(ArrayBase) + + private: + explicit ArrayBase(Index); + ArrayBase(Index,Index); + template explicit ArrayBase(const ArrayBase&); + protected: + // mixing arrays and matrices is not legal + template Derived& operator+=(const MatrixBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} + // mixing arrays and matrices is not legal + template Derived& operator-=(const MatrixBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} +}; + +/** replaces \c *this by \c *this - \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator-=(const ArrayBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this + \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator+=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this * \a other coefficient wise. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator*=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::mul_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this / \a other coefficient wise. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator/=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::div_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_ARRAYBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h new file mode 100644 index 00000000000..688aadd6260 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h @@ -0,0 +1,209 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAYWRAPPER_H +#define EIGEN_ARRAYWRAPPER_H + +namespace Eigen { + +/** \class ArrayWrapper + * \ingroup Core_Module + * + * \brief Expression of a mathematical vector or matrix as an array object + * + * This class is the return type of MatrixBase::array(), and most of the time + * this is the only way it is use. + * + * \sa MatrixBase::array(), class MatrixWrapper + */ + +namespace internal { +template +struct traits > + : public traits::type > +{ + typedef ArrayXpr XprKind; + // Let's remove NestByRefBit + enum { + Flags0 = traits::type >::Flags, + LvalueBitFlag = is_lvalue::value ? LvalueBit : 0, + Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag + }; +}; +} + +template +class ArrayWrapper : public ArrayBase > +{ + public: + typedef ArrayBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(ArrayWrapper) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(ArrayWrapper) + typedef typename internal::remove_all::type NestedExpression; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + typedef typename internal::ref_selector::non_const_type NestedExpressionType; + + using Base::coeffRef; + + EIGEN_DEVICE_FUNC + explicit EIGEN_STRONG_INLINE ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC + inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC + inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return m_expression.data(); } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_expression.coeffRef(rowId, colId); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_expression.coeffRef(index); + } + + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const { dst = m_expression; } + + const typename internal::remove_all::type& + EIGEN_DEVICE_FUNC + nestedExpression() const + { + return m_expression; + } + + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index) */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) { m_expression.resize(newSize); } + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index,Index)*/ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) { m_expression.resize(rows,cols); } + + protected: + NestedExpressionType m_expression; +}; + +/** \class MatrixWrapper + * \ingroup Core_Module + * + * \brief Expression of an array as a mathematical vector or matrix + * + * This class is the return type of ArrayBase::matrix(), and most of the time + * this is the only way it is use. + * + * \sa MatrixBase::matrix(), class ArrayWrapper + */ + +namespace internal { +template +struct traits > + : public traits::type > +{ + typedef MatrixXpr XprKind; + // Let's remove NestByRefBit + enum { + Flags0 = traits::type >::Flags, + LvalueBitFlag = is_lvalue::value ? LvalueBit : 0, + Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag + }; +}; +} + +template +class MatrixWrapper : public MatrixBase > +{ + public: + typedef MatrixBase > Base; + EIGEN_DENSE_PUBLIC_INTERFACE(MatrixWrapper) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(MatrixWrapper) + typedef typename internal::remove_all::type NestedExpression; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + typedef typename internal::ref_selector::non_const_type NestedExpressionType; + + using Base::coeffRef; + + EIGEN_DEVICE_FUNC + explicit inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC + inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC + inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return m_expression.data(); } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_expression.derived().coeffRef(rowId, colId); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_expression.coeffRef(index); + } + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& + nestedExpression() const + { + return m_expression; + } + + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index) */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) { m_expression.resize(newSize); } + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index,Index)*/ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) { m_expression.resize(rows,cols); } + + protected: + NestedExpressionType m_expression; +}; + +} // end namespace Eigen + +#endif // EIGEN_ARRAYWRAPPER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h new file mode 100644 index 00000000000..53806ba33c4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h @@ -0,0 +1,90 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007 Michael Olbrich +// Copyright (C) 2006-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ASSIGN_H +#define EIGEN_ASSIGN_H + +namespace Eigen { + +template +template +EIGEN_STRONG_INLINE Derived& DenseBase + ::lazyAssign(const DenseBase& other) +{ + enum{ + SameType = internal::is_same::value + }; + + EIGEN_STATIC_ASSERT_LVALUE(Derived) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Derived,OtherDerived) + EIGEN_STATIC_ASSERT(SameType,YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) + + eigen_assert(rows() == other.rows() && cols() == other.cols()); + internal::call_assignment_no_alias(derived(),other.derived()); + + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& DenseBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& DenseBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const MatrixBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const EigenBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const ReturnByValue& other) +{ + other.derived().evalTo(derived()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_ASSIGN_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h new file mode 100644 index 00000000000..dbe435d86b3 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h @@ -0,0 +1,935 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2011 Benoit Jacob +// Copyright (C) 2011-2014 Gael Guennebaud +// Copyright (C) 2011-2012 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ASSIGN_EVALUATOR_H +#define EIGEN_ASSIGN_EVALUATOR_H + +namespace Eigen { + +// This implementation is based on Assign.h + +namespace internal { + +/*************************************************************************** +* Part 1 : the logic deciding a strategy for traversal and unrolling * +***************************************************************************/ + +// copy_using_evaluator_traits is based on assign_traits + +template +struct copy_using_evaluator_traits +{ + typedef typename DstEvaluator::XprType Dst; + typedef typename Dst::Scalar DstScalar; + + enum { + DstFlags = DstEvaluator::Flags, + SrcFlags = SrcEvaluator::Flags + }; + +public: + enum { + DstAlignment = DstEvaluator::Alignment, + SrcAlignment = SrcEvaluator::Alignment, + DstHasDirectAccess = (DstFlags & DirectAccessBit) == DirectAccessBit, + JointAlignment = EIGEN_PLAIN_ENUM_MIN(DstAlignment,SrcAlignment) + }; + +private: + enum { + InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime) + : int(DstFlags)&RowMajorBit ? int(Dst::ColsAtCompileTime) + : int(Dst::RowsAtCompileTime), + InnerMaxSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::MaxSizeAtCompileTime) + : int(DstFlags)&RowMajorBit ? int(Dst::MaxColsAtCompileTime) + : int(Dst::MaxRowsAtCompileTime), + OuterStride = int(outer_stride_at_compile_time::ret), + MaxSizeAtCompileTime = Dst::SizeAtCompileTime + }; + + // TODO distinguish between linear traversal and inner-traversals + typedef typename find_best_packet::type LinearPacketType; + typedef typename find_best_packet::type InnerPacketType; + + enum { + LinearPacketSize = unpacket_traits::size, + InnerPacketSize = unpacket_traits::size + }; + +public: + enum { + LinearRequiredAlignment = unpacket_traits::alignment, + InnerRequiredAlignment = unpacket_traits::alignment + }; + +private: + enum { + DstIsRowMajor = DstFlags&RowMajorBit, + SrcIsRowMajor = SrcFlags&RowMajorBit, + StorageOrdersAgree = (int(DstIsRowMajor) == int(SrcIsRowMajor)), + MightVectorize = bool(StorageOrdersAgree) + && (int(DstFlags) & int(SrcFlags) & ActualPacketAccessBit) + && bool(functor_traits::PacketAccess), + MayInnerVectorize = MightVectorize + && int(InnerSize)!=Dynamic && int(InnerSize)%int(InnerPacketSize)==0 + && int(OuterStride)!=Dynamic && int(OuterStride)%int(InnerPacketSize)==0 + && (EIGEN_UNALIGNED_VECTORIZE || int(JointAlignment)>=int(InnerRequiredAlignment)), + MayLinearize = bool(StorageOrdersAgree) && (int(DstFlags) & int(SrcFlags) & LinearAccessBit), + MayLinearVectorize = bool(MightVectorize) && bool(MayLinearize) && bool(DstHasDirectAccess) + && (EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment)) || MaxSizeAtCompileTime == Dynamic), + /* If the destination isn't aligned, we have to do runtime checks and we don't unroll, + so it's only good for large enough sizes. */ + MaySliceVectorize = bool(MightVectorize) && bool(DstHasDirectAccess) + && (int(InnerMaxSize)==Dynamic || int(InnerMaxSize)>=(EIGEN_UNALIGNED_VECTORIZE?InnerPacketSize:(3*InnerPacketSize))) + /* slice vectorization can be slow, so we only want it if the slices are big, which is + indicated by InnerMaxSize rather than InnerSize, think of the case of a dynamic block + in a fixed-size matrix + However, with EIGEN_UNALIGNED_VECTORIZE and unrolling, slice vectorization is still worth it */ + }; + +public: + enum { + Traversal = int(MayLinearVectorize) && (LinearPacketSize>InnerPacketSize) ? int(LinearVectorizedTraversal) + : int(MayInnerVectorize) ? int(InnerVectorizedTraversal) + : int(MayLinearVectorize) ? int(LinearVectorizedTraversal) + : int(MaySliceVectorize) ? int(SliceVectorizedTraversal) + : int(MayLinearize) ? int(LinearTraversal) + : int(DefaultTraversal), + Vectorized = int(Traversal) == InnerVectorizedTraversal + || int(Traversal) == LinearVectorizedTraversal + || int(Traversal) == SliceVectorizedTraversal + }; + + typedef typename conditional::type PacketType; + +private: + enum { + ActualPacketSize = int(Traversal)==LinearVectorizedTraversal ? LinearPacketSize + : Vectorized ? InnerPacketSize + : 1, + UnrollingLimit = EIGEN_UNROLLING_LIMIT * ActualPacketSize, + MayUnrollCompletely = int(Dst::SizeAtCompileTime) != Dynamic + && int(Dst::SizeAtCompileTime) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit), + MayUnrollInner = int(InnerSize) != Dynamic + && int(InnerSize) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit) + }; + +public: + enum { + Unrolling = (int(Traversal) == int(InnerVectorizedTraversal) || int(Traversal) == int(DefaultTraversal)) + ? ( + int(MayUnrollCompletely) ? int(CompleteUnrolling) + : int(MayUnrollInner) ? int(InnerUnrolling) + : int(NoUnrolling) + ) + : int(Traversal) == int(LinearVectorizedTraversal) + ? ( bool(MayUnrollCompletely) && ( EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment))) + ? int(CompleteUnrolling) + : int(NoUnrolling) ) + : int(Traversal) == int(LinearTraversal) + ? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling) + : int(NoUnrolling) ) +#if EIGEN_UNALIGNED_VECTORIZE + : int(Traversal) == int(SliceVectorizedTraversal) + ? ( bool(MayUnrollInner) ? int(InnerUnrolling) + : int(NoUnrolling) ) +#endif + : int(NoUnrolling) + }; + +#ifdef EIGEN_DEBUG_ASSIGN + static void debug() + { + std::cerr << "DstXpr: " << typeid(typename DstEvaluator::XprType).name() << std::endl; + std::cerr << "SrcXpr: " << typeid(typename SrcEvaluator::XprType).name() << std::endl; + std::cerr.setf(std::ios::hex, std::ios::basefield); + std::cerr << "DstFlags" << " = " << DstFlags << " (" << demangle_flags(DstFlags) << " )" << std::endl; + std::cerr << "SrcFlags" << " = " << SrcFlags << " (" << demangle_flags(SrcFlags) << " )" << std::endl; + std::cerr.unsetf(std::ios::hex); + EIGEN_DEBUG_VAR(DstAlignment) + EIGEN_DEBUG_VAR(SrcAlignment) + EIGEN_DEBUG_VAR(LinearRequiredAlignment) + EIGEN_DEBUG_VAR(InnerRequiredAlignment) + EIGEN_DEBUG_VAR(JointAlignment) + EIGEN_DEBUG_VAR(InnerSize) + EIGEN_DEBUG_VAR(InnerMaxSize) + EIGEN_DEBUG_VAR(LinearPacketSize) + EIGEN_DEBUG_VAR(InnerPacketSize) + EIGEN_DEBUG_VAR(ActualPacketSize) + EIGEN_DEBUG_VAR(StorageOrdersAgree) + EIGEN_DEBUG_VAR(MightVectorize) + EIGEN_DEBUG_VAR(MayLinearize) + EIGEN_DEBUG_VAR(MayInnerVectorize) + EIGEN_DEBUG_VAR(MayLinearVectorize) + EIGEN_DEBUG_VAR(MaySliceVectorize) + std::cerr << "Traversal" << " = " << Traversal << " (" << demangle_traversal(Traversal) << ")" << std::endl; + EIGEN_DEBUG_VAR(SrcEvaluator::CoeffReadCost) + EIGEN_DEBUG_VAR(UnrollingLimit) + EIGEN_DEBUG_VAR(MayUnrollCompletely) + EIGEN_DEBUG_VAR(MayUnrollInner) + std::cerr << "Unrolling" << " = " << Unrolling << " (" << demangle_unrolling(Unrolling) << ")" << std::endl; + std::cerr << std::endl; + } +#endif +}; + +/*************************************************************************** +* Part 2 : meta-unrollers +***************************************************************************/ + +/************************ +*** Default traversal *** +************************/ + +template +struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling +{ + // FIXME: this is not very clean, perhaps this information should be provided by the kernel? + typedef typename Kernel::DstEvaluatorType DstEvaluatorType; + typedef typename DstEvaluatorType::XprType DstXprType; + + enum { + outer = Index / DstXprType::InnerSizeAtCompileTime, + inner = Index % DstXprType::InnerSizeAtCompileTime + }; + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + kernel.assignCoeffByOuterInner(outer, inner); + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +template +struct copy_using_evaluator_DefaultTraversal_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer) + { + kernel.assignCoeffByOuterInner(outer, Index_); + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } +}; + +template +struct copy_using_evaluator_DefaultTraversal_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index) { } +}; + +/*********************** +*** Linear traversal *** +***********************/ + +template +struct copy_using_evaluator_LinearTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel& kernel) + { + kernel.assignCoeff(Index); + copy_using_evaluator_LinearTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_LinearTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +/************************** +*** Inner vectorization *** +**************************/ + +template +struct copy_using_evaluator_innervec_CompleteUnrolling +{ + // FIXME: this is not very clean, perhaps this information should be provided by the kernel? + typedef typename Kernel::DstEvaluatorType DstEvaluatorType; + typedef typename DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { + outer = Index / DstXprType::InnerSizeAtCompileTime, + inner = Index % DstXprType::InnerSizeAtCompileTime, + SrcAlignment = Kernel::AssignmentTraits::SrcAlignment, + DstAlignment = Kernel::AssignmentTraits::DstAlignment + }; + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + kernel.template assignPacketByOuterInner(outer, inner); + enum { NextIndex = Index + unpacket_traits::size }; + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_innervec_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +template +struct copy_using_evaluator_innervec_InnerUnrolling +{ + typedef typename Kernel::PacketType PacketType; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer) + { + kernel.template assignPacketByOuterInner(outer, Index_); + enum { NextIndex = Index_ + unpacket_traits::size }; + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + } +}; + +template +struct copy_using_evaluator_innervec_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &, Index) { } +}; + +/*************************************************************************** +* Part 3 : implementation of all cases +***************************************************************************/ + +// dense_assignment_loop is based on assign_impl + +template +struct dense_assignment_loop; + +/************************ +*** Default traversal *** +************************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static void EIGEN_STRONG_INLINE run(Kernel &kernel) + { + for(Index outer = 0; outer < kernel.outerSize(); ++outer) { + for(Index inner = 0; inner < kernel.innerSize(); ++inner) { + kernel.assignCoeffByOuterInner(outer, inner); + } + } + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + + const Index outerSize = kernel.outerSize(); + for(Index outer = 0; outer < outerSize; ++outer) + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } +}; + +/*************************** +*** Linear vectorization *** +***************************/ + + +// The goal of unaligned_dense_assignment_loop is simply to factorize the handling +// of the non vectorizable beginning and ending parts + +template +struct unaligned_dense_assignment_loop +{ + // if IsAligned = true, then do nothing + template + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index, Index) {} +}; + +template <> +struct unaligned_dense_assignment_loop +{ + // MSVC must not inline this functions. If it does, it fails to optimize the + // packet access path. + // FIXME check which version exhibits this issue +#if EIGEN_COMP_MSVC + template + static EIGEN_DONT_INLINE void run(Kernel &kernel, + Index start, + Index end) +#else + template + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, + Index start, + Index end) +#endif + { + for (Index index = start; index < end; ++index) + kernel.assignCoeff(index); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index size = kernel.size(); + typedef typename Kernel::Scalar Scalar; + typedef typename Kernel::PacketType PacketType; + enum { + requestedAlignment = Kernel::AssignmentTraits::LinearRequiredAlignment, + packetSize = unpacket_traits::size, + dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment), + dstAlignment = packet_traits::AlignedOnScalar ? int(requestedAlignment) + : int(Kernel::AssignmentTraits::DstAlignment), + srcAlignment = Kernel::AssignmentTraits::JointAlignment + }; + const Index alignedStart = dstIsAligned ? 0 : internal::first_aligned(kernel.dstDataPtr(), size); + const Index alignedEnd = alignedStart + ((size-alignedStart)/packetSize)*packetSize; + + unaligned_dense_assignment_loop::run(kernel, 0, alignedStart); + + for(Index index = alignedStart; index < alignedEnd; index += packetSize) + kernel.template assignPacket(index); + + unaligned_dense_assignment_loop<>::run(kernel, alignedEnd, size); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { size = DstXprType::SizeAtCompileTime, + packetSize =unpacket_traits::size, + alignedSize = (size/packetSize)*packetSize }; + + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +/************************** +*** Inner vectorization *** +**************************/ + +template +struct dense_assignment_loop +{ + typedef typename Kernel::PacketType PacketType; + enum { + SrcAlignment = Kernel::AssignmentTraits::SrcAlignment, + DstAlignment = Kernel::AssignmentTraits::DstAlignment + }; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index innerSize = kernel.innerSize(); + const Index outerSize = kernel.outerSize(); + const Index packetSize = unpacket_traits::size; + for(Index outer = 0; outer < outerSize; ++outer) + for(Index inner = 0; inner < innerSize; inner+=packetSize) + kernel.template assignPacketByOuterInner(outer, inner); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::AssignmentTraits Traits; + const Index outerSize = kernel.outerSize(); + for(Index outer = 0; outer < outerSize; ++outer) + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + } +}; + +/*********************** +*** Linear traversal *** +***********************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index size = kernel.size(); + for(Index i = 0; i < size; ++i) + kernel.assignCoeff(i); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_LinearTraversal_CompleteUnrolling::run(kernel); + } +}; + +/************************** +*** Slice vectorization *** +***************************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::Scalar Scalar; + typedef typename Kernel::PacketType PacketType; + enum { + packetSize = unpacket_traits::size, + requestedAlignment = int(Kernel::AssignmentTraits::InnerRequiredAlignment), + alignable = packet_traits::AlignedOnScalar || int(Kernel::AssignmentTraits::DstAlignment)>=sizeof(Scalar), + dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment), + dstAlignment = alignable ? int(requestedAlignment) + : int(Kernel::AssignmentTraits::DstAlignment) + }; + const Scalar *dst_ptr = kernel.dstDataPtr(); + if((!bool(dstIsAligned)) && (UIntPtr(dst_ptr) % sizeof(Scalar))>0) + { + // the pointer is not aligend-on scalar, so alignment is not possible + return dense_assignment_loop::run(kernel); + } + const Index packetAlignedMask = packetSize - 1; + const Index innerSize = kernel.innerSize(); + const Index outerSize = kernel.outerSize(); + const Index alignedStep = alignable ? (packetSize - kernel.outerStride() % packetSize) & packetAlignedMask : 0; + Index alignedStart = ((!alignable) || bool(dstIsAligned)) ? 0 : internal::first_aligned(dst_ptr, innerSize); + + for(Index outer = 0; outer < outerSize; ++outer) + { + const Index alignedEnd = alignedStart + ((innerSize-alignedStart) & ~packetAlignedMask); + // do the non-vectorizable part of the assignment + for(Index inner = 0; inner(outer, inner); + + // do the non-vectorizable part of the assignment + for(Index inner = alignedEnd; inner +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { size = DstXprType::InnerSizeAtCompileTime, + packetSize =unpacket_traits::size, + vectorizableSize = (size/packetSize)*packetSize }; + + for(Index outer = 0; outer < kernel.outerSize(); ++outer) + { + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } + } +}; +#endif + + +/*************************************************************************** +* Part 4 : Generic dense assignment kernel +***************************************************************************/ + +// This class generalize the assignment of a coefficient (or packet) from one dense evaluator +// to another dense writable evaluator. +// It is parametrized by the two evaluators, and the actual assignment functor. +// This abstraction level permits to keep the evaluation loops as simple and as generic as possible. +// One can customize the assignment using this generic dense_assignment_kernel with different +// functors, or by completely overloading it, by-passing a functor. +template +class generic_dense_assignment_kernel +{ +protected: + typedef typename DstEvaluatorTypeT::XprType DstXprType; + typedef typename SrcEvaluatorTypeT::XprType SrcXprType; +public: + + typedef DstEvaluatorTypeT DstEvaluatorType; + typedef SrcEvaluatorTypeT SrcEvaluatorType; + typedef typename DstEvaluatorType::Scalar Scalar; + typedef copy_using_evaluator_traits AssignmentTraits; + typedef typename AssignmentTraits::PacketType PacketType; + + + EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) + : m_dst(dst), m_src(src), m_functor(func), m_dstExpr(dstExpr) + { + #ifdef EIGEN_DEBUG_ASSIGN + AssignmentTraits::debug(); + #endif + } + + EIGEN_DEVICE_FUNC Index size() const { return m_dstExpr.size(); } + EIGEN_DEVICE_FUNC Index innerSize() const { return m_dstExpr.innerSize(); } + EIGEN_DEVICE_FUNC Index outerSize() const { return m_dstExpr.outerSize(); } + EIGEN_DEVICE_FUNC Index rows() const { return m_dstExpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_dstExpr.cols(); } + EIGEN_DEVICE_FUNC Index outerStride() const { return m_dstExpr.outerStride(); } + + EIGEN_DEVICE_FUNC DstEvaluatorType& dstEvaluator() { return m_dst; } + EIGEN_DEVICE_FUNC const SrcEvaluatorType& srcEvaluator() const { return m_src; } + + /// Assign src(row,col) to dst(row,col) through the assignment functor. + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index row, Index col) + { + m_functor.assignCoeff(m_dst.coeffRef(row,col), m_src.coeff(row,col)); + } + + /// \sa assignCoeff(Index,Index) + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index index) + { + m_functor.assignCoeff(m_dst.coeffRef(index), m_src.coeff(index)); + } + + /// \sa assignCoeff(Index,Index) + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeffByOuterInner(Index outer, Index inner) + { + Index row = rowIndexByOuterInner(outer, inner); + Index col = colIndexByOuterInner(outer, inner); + assignCoeff(row, col); + } + + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index row, Index col) + { + m_functor.template assignPacket(&m_dst.coeffRef(row,col), m_src.template packet(row,col)); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index index) + { + m_functor.template assignPacket(&m_dst.coeffRef(index), m_src.template packet(index)); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacketByOuterInner(Index outer, Index inner) + { + Index row = rowIndexByOuterInner(outer, inner); + Index col = colIndexByOuterInner(outer, inner); + assignPacket(row, col); + } + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner) + { + typedef typename DstEvaluatorType::ExpressionTraits Traits; + return int(Traits::RowsAtCompileTime) == 1 ? 0 + : int(Traits::ColsAtCompileTime) == 1 ? inner + : int(DstEvaluatorType::Flags)&RowMajorBit ? outer + : inner; + } + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner) + { + typedef typename DstEvaluatorType::ExpressionTraits Traits; + return int(Traits::ColsAtCompileTime) == 1 ? 0 + : int(Traits::RowsAtCompileTime) == 1 ? inner + : int(DstEvaluatorType::Flags)&RowMajorBit ? inner + : outer; + } + + EIGEN_DEVICE_FUNC const Scalar* dstDataPtr() const + { + return m_dstExpr.data(); + } + +protected: + DstEvaluatorType& m_dst; + const SrcEvaluatorType& m_src; + const Functor &m_functor; + // TODO find a way to avoid the needs of the original expression + DstXprType& m_dstExpr; +}; + +/*************************************************************************** +* Part 5 : Entry point for dense rectangular assignment +***************************************************************************/ + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const Functor &/*func*/) +{ + EIGEN_ONLY_USED_FOR_DEBUG(dst); + EIGEN_ONLY_USED_FOR_DEBUG(src); + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const internal::assign_op &/*func*/) +{ + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if(((dst.rows()!=dstRows) || (dst.cols()!=dstCols))) + dst.resize(dstRows, dstCols); + eigen_assert(dst.rows() == dstRows && dst.cols() == dstCols); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src, const Functor &func) +{ + typedef evaluator DstEvaluatorType; + typedef evaluator SrcEvaluatorType; + + SrcEvaluatorType srcEvaluator(src); + + // NOTE To properly handle A = (A*A.transpose())/s with A rectangular, + // we need to resize the destination after the source evaluator has been created. + resize_if_allowed(dst, src, func); + + DstEvaluatorType dstEvaluator(dst); + + typedef generic_dense_assignment_kernel Kernel; + Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived()); + + dense_assignment_loop::run(kernel); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src) +{ + call_dense_assignment_loop(dst, src, internal::assign_op()); +} + +/*************************************************************************** +* Part 6 : Generic assignment +***************************************************************************/ + +// Based on the respective shapes of the destination and source, +// the class AssignmentKind determine the kind of assignment mechanism. +// AssignmentKind must define a Kind typedef. +template struct AssignmentKind; + +// Assignement kind defined in this file: +struct Dense2Dense {}; +struct EigenBase2EigenBase {}; + +template struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; +template<> struct AssignmentKind { typedef Dense2Dense Kind; }; + +// This is the main assignment class +template< typename DstXprType, typename SrcXprType, typename Functor, + typename Kind = typename AssignmentKind< typename evaluator_traits::Shape , typename evaluator_traits::Shape >::Kind, + typename EnableIf = void> +struct Assignment; + + +// The only purpose of this call_assignment() function is to deal with noalias() / "assume-aliasing" and automatic transposition. +// Indeed, I (Gael) think that this concept of "assume-aliasing" was a mistake, and it makes thing quite complicated. +// So this intermediate function removes everything related to "assume-aliasing" such that Assignment +// does not has to bother about these annoying details. + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src) +{ + call_assignment(dst, src, internal::assign_op()); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(const Dst& dst, const Src& src) +{ + call_assignment(dst, src, internal::assign_op()); +} + +// Deal with "assume-aliasing" +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if< evaluator_assume_aliasing::value, void*>::type = 0) +{ + typename plain_matrix_type::type tmp(src); + call_assignment_no_alias(dst, tmp, func); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if::value, void*>::type = 0) +{ + call_assignment_no_alias(dst, src, func); +} + +// by-pass "assume-aliasing" +// When there is no aliasing, we require that 'dst' has been properly resized +template class StorageBase, typename Src, typename Func> +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(NoAlias& dst, const Src& src, const Func& func) +{ + call_assignment_no_alias(dst.expression(), src, func); +} + + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias(Dst& dst, const Src& src, const Func& func) +{ + enum { + NeedToTranspose = ( (int(Dst::RowsAtCompileTime) == 1 && int(Src::ColsAtCompileTime) == 1) + || (int(Dst::ColsAtCompileTime) == 1 && int(Src::RowsAtCompileTime) == 1) + ) && int(Dst::SizeAtCompileTime) != 1 + }; + + typedef typename internal::conditional, Dst>::type ActualDstTypeCleaned; + typedef typename internal::conditional, Dst&>::type ActualDstType; + ActualDstType actualDst(dst); + + // TODO check whether this is the right place to perform these checks: + EIGEN_STATIC_ASSERT_LVALUE(Dst) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(ActualDstTypeCleaned,Src) + EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename ActualDstTypeCleaned::Scalar,typename Src::Scalar); + + Assignment::run(actualDst, src, func); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias(Dst& dst, const Src& src) +{ + call_assignment_no_alias(dst, src, internal::assign_op()); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src, const Func& func) +{ + // TODO check whether this is the right place to perform these checks: + EIGEN_STATIC_ASSERT_LVALUE(Dst) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Dst,Src) + EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename Dst::Scalar,typename Src::Scalar); + + Assignment::run(dst, src, func); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src) +{ + call_assignment_no_alias_no_transpose(dst, src, internal::assign_op()); +} + +// forward declaration +template void check_for_aliasing(const Dst &dst, const Src &src); + +// Generic Dense to Dense assignment +// Note that the last template argument "Weak" is needed to make it possible to perform +// both partial specialization+SFINAE without ambiguous specialization +template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak> +struct Assignment +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const Functor &func) + { +#ifndef EIGEN_NO_DEBUG + internal::check_for_aliasing(dst, src); +#endif + + call_dense_assignment_loop(dst, src, func); + } +}; + +// Generic assignment through evalTo. +// TODO: not sure we have to keep that one, but it helps porting current code to new evaluator mechanism. +// Note that the last template argument "Weak" is needed to make it possible to perform +// both partial specialization+SFINAE without ambiguous specialization +template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak> +struct Assignment +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.evalTo(dst); + } + + // NOTE The following two functions are templated to avoid their instanciation if not needed + // This is needed because some expressions supports evalTo only and/or have 'void' as scalar type. + template + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.addTo(dst); + } + + template + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.subTo(dst); + } +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_ASSIGN_EVALUATOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign_MKL.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign_MKL.h new file mode 100644 index 00000000000..6866095bf8a --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Assign_MKL.h @@ -0,0 +1,178 @@ +/* + Copyright (c) 2011, Intel Corporation. All rights reserved. + Copyright (C) 2015 Gael Guennebaud + + Redistribution and use in source and binary forms, with or without modification, + are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + * Neither the name of Intel Corporation nor the names of its contributors may + be used to endorse or promote products derived from this software without + specific prior written permission. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND + ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR + ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON + ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + ******************************************************************************** + * Content : Eigen bindings to Intel(R) MKL + * MKL VML support for coefficient-wise unary Eigen expressions like a=b.sin() + ******************************************************************************** +*/ + +#ifndef EIGEN_ASSIGN_VML_H +#define EIGEN_ASSIGN_VML_H + +namespace Eigen { + +namespace internal { + +template +class vml_assign_traits +{ + private: + enum { + DstHasDirectAccess = Dst::Flags & DirectAccessBit, + SrcHasDirectAccess = Src::Flags & DirectAccessBit, + StorageOrdersAgree = (int(Dst::IsRowMajor) == int(Src::IsRowMajor)), + InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime) + : int(Dst::Flags)&RowMajorBit ? int(Dst::ColsAtCompileTime) + : int(Dst::RowsAtCompileTime), + InnerMaxSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::MaxSizeAtCompileTime) + : int(Dst::Flags)&RowMajorBit ? int(Dst::MaxColsAtCompileTime) + : int(Dst::MaxRowsAtCompileTime), + MaxSizeAtCompileTime = Dst::SizeAtCompileTime, + + MightEnableVml = StorageOrdersAgree && DstHasDirectAccess && SrcHasDirectAccess && Src::InnerStrideAtCompileTime==1 && Dst::InnerStrideAtCompileTime==1, + MightLinearize = MightEnableVml && (int(Dst::Flags) & int(Src::Flags) & LinearAccessBit), + VmlSize = MightLinearize ? MaxSizeAtCompileTime : InnerMaxSize, + LargeEnough = VmlSize==Dynamic || VmlSize>=EIGEN_MKL_VML_THRESHOLD + }; + public: + enum { + EnableVml = MightEnableVml && LargeEnough, + Traversal = MightLinearize ? LinearTraversal : DefaultTraversal + }; +}; + +#define EIGEN_PP_EXPAND(ARG) ARG +#if !defined (EIGEN_FAST_MATH) || (EIGEN_FAST_MATH != 1) +#define EIGEN_VMLMODE_EXPAND_LA , VML_HA +#else +#define EIGEN_VMLMODE_EXPAND_LA , VML_LA +#endif + +#define EIGEN_VMLMODE_EXPAND__ + +#define EIGEN_VMLMODE_PREFIX_LA vm +#define EIGEN_VMLMODE_PREFIX__ v +#define EIGEN_VMLMODE_PREFIX(VMLMODE) EIGEN_CAT(EIGEN_VMLMODE_PREFIX_,VMLMODE) + +#define EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \ + template< typename DstXprType, typename SrcXprNested> \ + struct Assignment, SrcXprNested>, assign_op, \ + Dense2Dense, typename enable_if::EnableVml>::type> { \ + typedef CwiseUnaryOp, SrcXprNested> SrcXprType; \ + static void run(DstXprType &dst, const SrcXprType &src, const assign_op &func) { \ + resize_if_allowed(dst, src, func); \ + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \ + if(vml_assign_traits::Traversal==LinearTraversal) { \ + VMLOP(dst.size(), (const VMLTYPE*)src.nestedExpression().data(), \ + (VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \ + } else { \ + const Index outerSize = dst.outerSize(); \ + for(Index outer = 0; outer < outerSize; ++outer) { \ + const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.nestedExpression().coeffRef(outer,0)) : \ + &(src.nestedExpression().coeffRef(0, outer)); \ + EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \ + VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, \ + (VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \ + } \ + } \ + } \ + }; \ + + +#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),s##VMLOP), float, float, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),d##VMLOP), double, double, VMLMODE) + +#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),c##VMLOP), scomplex, MKL_Complex8, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),z##VMLOP), dcomplex, MKL_Complex16, VMLMODE) + +#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS(EIGENOP, VMLOP, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \ + EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE) + + +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sin, Sin, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(asin, Asin, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sinh, Sinh, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cos, Cos, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(acos, Acos, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cosh, Cosh, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tan, Tan, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(atan, Atan, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tanh, Tanh, LA) +// EIGEN_MKL_VML_DECLARE_UNARY_CALLS(abs, Abs, _) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(exp, Exp, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log, Ln, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log10, Log10, LA) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sqrt, Sqrt, _) + +EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(square, Sqr, _) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(arg, Arg, _) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(round, Round, _) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(floor, Floor, _) +EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(ceil, Ceil, _) + +#define EIGEN_MKL_VML_DECLARE_POW_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \ + template< typename DstXprType, typename SrcXprNested, typename Plain> \ + struct Assignment, SrcXprNested, \ + const CwiseNullaryOp,Plain> >, assign_op, \ + Dense2Dense, typename enable_if::EnableVml>::type> { \ + typedef CwiseBinaryOp, SrcXprNested, \ + const CwiseNullaryOp,Plain> > SrcXprType; \ + static void run(DstXprType &dst, const SrcXprType &src, const assign_op &func) { \ + resize_if_allowed(dst, src, func); \ + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \ + VMLTYPE exponent = reinterpret_cast(src.rhs().functor().m_other); \ + if(vml_assign_traits::Traversal==LinearTraversal) \ + { \ + VMLOP( dst.size(), (const VMLTYPE*)src.lhs().data(), exponent, \ + (VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \ + } else { \ + const Index outerSize = dst.outerSize(); \ + for(Index outer = 0; outer < outerSize; ++outer) { \ + const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.lhs().coeffRef(outer,0)) : \ + &(src.lhs().coeffRef(0, outer)); \ + EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \ + VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, exponent, \ + (VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \ + } \ + } \ + } \ + }; + +EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmsPowx, float, float, LA) +EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmdPowx, double, double, LA) +EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmcPowx, scomplex, MKL_Complex8, LA) +EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmzPowx, dcomplex, MKL_Complex16, LA) + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_ASSIGN_VML_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h new file mode 100644 index 00000000000..4978c914057 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h @@ -0,0 +1,353 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_BANDMATRIX_H +#define EIGEN_BANDMATRIX_H + +namespace Eigen { + +namespace internal { + +template +class BandMatrixBase : public EigenBase +{ + public: + + enum { + Flags = internal::traits::Flags, + CoeffReadCost = internal::traits::CoeffReadCost, + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, + Supers = internal::traits::Supers, + Subs = internal::traits::Subs, + Options = internal::traits::Options + }; + typedef typename internal::traits::Scalar Scalar; + typedef Matrix DenseMatrixType; + typedef typename DenseMatrixType::StorageIndex StorageIndex; + typedef typename internal::traits::CoefficientsType CoefficientsType; + typedef EigenBase Base; + + protected: + enum { + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) + ? 1 + Supers + Subs + : Dynamic, + SizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime,ColsAtCompileTime) + }; + + public: + + using Base::derived; + using Base::rows; + using Base::cols; + + /** \returns the number of super diagonals */ + inline Index supers() const { return derived().supers(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return derived().subs(); } + + /** \returns an expression of the underlying coefficient matrix */ + inline const CoefficientsType& coeffs() const { return derived().coeffs(); } + + /** \returns an expression of the underlying coefficient matrix */ + inline CoefficientsType& coeffs() { return derived().coeffs(); } + + /** \returns a vector expression of the \a i -th column, + * only the meaningful part is returned. + * \warning the internal storage must be column major. */ + inline Block col(Index i) + { + EIGEN_STATIC_ASSERT((Options&RowMajor)==0,THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); + Index start = 0; + Index len = coeffs().rows(); + if (i<=supers()) + { + start = supers()-i; + len = (std::min)(rows(),std::max(0,coeffs().rows() - (supers()-i))); + } + else if (i>=rows()-subs()) + len = std::max(0,coeffs().rows() - (i + 1 - rows() + subs())); + return Block(coeffs(), start, i, len, 1); + } + + /** \returns a vector expression of the main diagonal */ + inline Block diagonal() + { return Block(coeffs(),supers(),0,1,(std::min)(rows(),cols())); } + + /** \returns a vector expression of the main diagonal (const version) */ + inline const Block diagonal() const + { return Block(coeffs(),supers(),0,1,(std::min)(rows(),cols())); } + + template struct DiagonalIntReturnType { + enum { + ReturnOpposite = (Options&SelfAdjoint) && (((Index)>0 && Supers==0) || ((Index)<0 && Subs==0)), + Conjugate = ReturnOpposite && NumTraits::IsComplex, + ActualIndex = ReturnOpposite ? -Index : Index, + DiagonalSize = (RowsAtCompileTime==Dynamic || ColsAtCompileTime==Dynamic) + ? Dynamic + : (ActualIndex<0 + ? EIGEN_SIZE_MIN_PREFER_DYNAMIC(ColsAtCompileTime, RowsAtCompileTime + ActualIndex) + : EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime, ColsAtCompileTime - ActualIndex)) + }; + typedef Block BuildType; + typedef typename internal::conditional,BuildType >, + BuildType>::type Type; + }; + + /** \returns a vector expression of the \a N -th sub or super diagonal */ + template inline typename DiagonalIntReturnType::Type diagonal() + { + return typename DiagonalIntReturnType::BuildType(coeffs(), supers()-N, (std::max)(0,N), 1, diagonalLength(N)); + } + + /** \returns a vector expression of the \a N -th sub or super diagonal */ + template inline const typename DiagonalIntReturnType::Type diagonal() const + { + return typename DiagonalIntReturnType::BuildType(coeffs(), supers()-N, (std::max)(0,N), 1, diagonalLength(N)); + } + + /** \returns a vector expression of the \a i -th sub or super diagonal */ + inline Block diagonal(Index i) + { + eigen_assert((i<0 && -i<=subs()) || (i>=0 && i<=supers())); + return Block(coeffs(), supers()-i, std::max(0,i), 1, diagonalLength(i)); + } + + /** \returns a vector expression of the \a i -th sub or super diagonal */ + inline const Block diagonal(Index i) const + { + eigen_assert((i<0 && -i<=subs()) || (i>=0 && i<=supers())); + return Block(coeffs(), supers()-i, std::max(0,i), 1, diagonalLength(i)); + } + + template inline void evalTo(Dest& dst) const + { + dst.resize(rows(),cols()); + dst.setZero(); + dst.diagonal() = diagonal(); + for (Index i=1; i<=supers();++i) + dst.diagonal(i) = diagonal(i); + for (Index i=1; i<=subs();++i) + dst.diagonal(-i) = diagonal(-i); + } + + DenseMatrixType toDenseMatrix() const + { + DenseMatrixType res(rows(),cols()); + evalTo(res); + return res; + } + + protected: + + inline Index diagonalLength(Index i) const + { return i<0 ? (std::min)(cols(),rows()+i) : (std::min)(rows(),cols()-i); } +}; + +/** + * \class BandMatrix + * \ingroup Core_Module + * + * \brief Represents a rectangular matrix with a banded storage + * + * \tparam _Scalar Numeric type, i.e. float, double, int + * \tparam _Rows Number of rows, or \b Dynamic + * \tparam _Cols Number of columns, or \b Dynamic + * \tparam _Supers Number of super diagonal + * \tparam _Subs Number of sub diagonal + * \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of \b #SelfAdjoint + * The former controls \ref TopicStorageOrders "storage order", and defaults to + * column-major. The latter controls whether the matrix represents a selfadjoint + * matrix in which case either Supers of Subs have to be null. + * + * \sa class TridiagonalMatrix + */ + +template +struct traits > +{ + typedef _Scalar Scalar; + typedef Dense StorageKind; + typedef Eigen::Index StorageIndex; + enum { + CoeffReadCost = NumTraits::ReadCost, + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _Rows, + MaxColsAtCompileTime = _Cols, + Flags = LvalueBit, + Supers = _Supers, + Subs = _Subs, + Options = _Options, + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) ? 1 + Supers + Subs : Dynamic + }; + typedef Matrix CoefficientsType; +}; + +template +class BandMatrix : public BandMatrixBase > +{ + public: + + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::traits::StorageIndex StorageIndex; + typedef typename internal::traits::CoefficientsType CoefficientsType; + + explicit inline BandMatrix(Index rows=Rows, Index cols=Cols, Index supers=Supers, Index subs=Subs) + : m_coeffs(1+supers+subs,cols), + m_rows(rows), m_supers(supers), m_subs(subs) + { + } + + /** \returns the number of columns */ + inline Index rows() const { return m_rows.value(); } + + /** \returns the number of rows */ + inline Index cols() const { return m_coeffs.cols(); } + + /** \returns the number of super diagonals */ + inline Index supers() const { return m_supers.value(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return m_subs.value(); } + + inline const CoefficientsType& coeffs() const { return m_coeffs; } + inline CoefficientsType& coeffs() { return m_coeffs; } + + protected: + + CoefficientsType m_coeffs; + internal::variable_if_dynamic m_rows; + internal::variable_if_dynamic m_supers; + internal::variable_if_dynamic m_subs; +}; + +template +class BandMatrixWrapper; + +template +struct traits > +{ + typedef typename _CoefficientsType::Scalar Scalar; + typedef typename _CoefficientsType::StorageKind StorageKind; + typedef typename _CoefficientsType::StorageIndex StorageIndex; + enum { + CoeffReadCost = internal::traits<_CoefficientsType>::CoeffReadCost, + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _Rows, + MaxColsAtCompileTime = _Cols, + Flags = LvalueBit, + Supers = _Supers, + Subs = _Subs, + Options = _Options, + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) ? 1 + Supers + Subs : Dynamic + }; + typedef _CoefficientsType CoefficientsType; +}; + +template +class BandMatrixWrapper : public BandMatrixBase > +{ + public: + + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::traits::CoefficientsType CoefficientsType; + typedef typename internal::traits::StorageIndex StorageIndex; + + explicit inline BandMatrixWrapper(const CoefficientsType& coeffs, Index rows=_Rows, Index cols=_Cols, Index supers=_Supers, Index subs=_Subs) + : m_coeffs(coeffs), + m_rows(rows), m_supers(supers), m_subs(subs) + { + EIGEN_UNUSED_VARIABLE(cols); + //internal::assert(coeffs.cols()==cols() && (supers()+subs()+1)==coeffs.rows()); + } + + /** \returns the number of columns */ + inline Index rows() const { return m_rows.value(); } + + /** \returns the number of rows */ + inline Index cols() const { return m_coeffs.cols(); } + + /** \returns the number of super diagonals */ + inline Index supers() const { return m_supers.value(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return m_subs.value(); } + + inline const CoefficientsType& coeffs() const { return m_coeffs; } + + protected: + + const CoefficientsType& m_coeffs; + internal::variable_if_dynamic m_rows; + internal::variable_if_dynamic m_supers; + internal::variable_if_dynamic m_subs; +}; + +/** + * \class TridiagonalMatrix + * \ingroup Core_Module + * + * \brief Represents a tridiagonal matrix with a compact banded storage + * + * \tparam Scalar Numeric type, i.e. float, double, int + * \tparam Size Number of rows and cols, or \b Dynamic + * \tparam Options Can be 0 or \b SelfAdjoint + * + * \sa class BandMatrix + */ +template +class TridiagonalMatrix : public BandMatrix +{ + typedef BandMatrix Base; + typedef typename Base::StorageIndex StorageIndex; + public: + explicit TridiagonalMatrix(Index size = Size) : Base(size,size,Options&SelfAdjoint?0:1,1) {} + + inline typename Base::template DiagonalIntReturnType<1>::Type super() + { return Base::template diagonal<1>(); } + inline const typename Base::template DiagonalIntReturnType<1>::Type super() const + { return Base::template diagonal<1>(); } + inline typename Base::template DiagonalIntReturnType<-1>::Type sub() + { return Base::template diagonal<-1>(); } + inline const typename Base::template DiagonalIntReturnType<-1>::Type sub() const + { return Base::template diagonal<-1>(); } + protected: +}; + + +struct BandShape {}; + +template +struct evaluator_traits > + : public evaluator_traits_base > +{ + typedef BandShape Shape; +}; + +template +struct evaluator_traits > + : public evaluator_traits_base > +{ + typedef BandShape Shape; +}; + +template<> struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_BANDMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Block.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Block.h new file mode 100644 index 00000000000..11de45c2ecb --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Block.h @@ -0,0 +1,452 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_BLOCK_H +#define EIGEN_BLOCK_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits +{ + typedef typename traits::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type XprTypeNested; + typedef typename remove_reference::type _XprTypeNested; + enum{ + MatrixRows = traits::RowsAtCompileTime, + MatrixCols = traits::ColsAtCompileTime, + RowsAtCompileTime = MatrixRows == 0 ? 0 : BlockRows, + ColsAtCompileTime = MatrixCols == 0 ? 0 : BlockCols, + MaxRowsAtCompileTime = BlockRows==0 ? 0 + : RowsAtCompileTime != Dynamic ? int(RowsAtCompileTime) + : int(traits::MaxRowsAtCompileTime), + MaxColsAtCompileTime = BlockCols==0 ? 0 + : ColsAtCompileTime != Dynamic ? int(ColsAtCompileTime) + : int(traits::MaxColsAtCompileTime), + + XprTypeIsRowMajor = (int(traits::Flags)&RowMajorBit) != 0, + IsRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0 + : XprTypeIsRowMajor, + HasSameStorageOrderAsXprType = (IsRowMajor == XprTypeIsRowMajor), + InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + InnerStrideAtCompileTime = HasSameStorageOrderAsXprType + ? int(inner_stride_at_compile_time::ret) + : int(outer_stride_at_compile_time::ret), + OuterStrideAtCompileTime = HasSameStorageOrderAsXprType + ? int(outer_stride_at_compile_time::ret) + : int(inner_stride_at_compile_time::ret), + + // FIXME, this traits is rather specialized for dense object and it needs to be cleaned further + FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, + FlagsRowMajorBit = IsRowMajor ? RowMajorBit : 0, + Flags = (traits::Flags & (DirectAccessBit | (InnerPanel?CompressedAccessBit:0))) | FlagsLvalueBit | FlagsRowMajorBit, + // FIXME DirectAccessBit should not be handled by expressions + // + // Alignment is needed by MapBase's assertions + // We can sefely set it to false here. Internal alignment errors will be detected by an eigen_internal_assert in the respective evaluator + Alignment = 0 + }; +}; + +template::ret> class BlockImpl_dense; + +} // end namespace internal + +template class BlockImpl; + +/** \class Block + * \ingroup Core_Module + * + * \brief Expression of a fixed-size or dynamic-size block + * + * \tparam XprType the type of the expression in which we are taking a block + * \tparam BlockRows the number of rows of the block we are taking at compile time (optional) + * \tparam BlockCols the number of columns of the block we are taking at compile time (optional) + * \tparam InnerPanel is true, if the block maps to a set of rows of a row major matrix or + * to set of columns of a column major matrix (optional). The parameter allows to determine + * at compile time whether aligned access is possible on the block expression. + * + * This class represents an expression of either a fixed-size or dynamic-size block. It is the return + * type of DenseBase::block(Index,Index,Index,Index) and DenseBase::block(Index,Index) and + * most of the time this is the only way it is used. + * + * However, if you want to directly maniputate block expressions, + * for instance if you want to write a function returning such an expression, you + * will need to use this class. + * + * Here is an example illustrating the dynamic case: + * \include class_Block.cpp + * Output: \verbinclude class_Block.out + * + * \note Even though this expression has dynamic size, in the case where \a XprType + * has fixed size, this expression inherits a fixed maximal size which means that evaluating + * it does not cause a dynamic memory allocation. + * + * Here is an example illustrating the fixed-size case: + * \include class_FixedBlock.cpp + * Output: \verbinclude class_FixedBlock.out + * + * \sa DenseBase::block(Index,Index,Index,Index), DenseBase::block(Index,Index), class VectorBlock + */ +template class Block + : public BlockImpl::StorageKind> +{ + typedef BlockImpl::StorageKind> Impl; + public: + //typedef typename Impl::Base Base; + typedef Impl Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(Block) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block) + + typedef typename internal::remove_all::type NestedExpression; + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline Block(XprType& xpr, Index i) : Impl(xpr,i) + { + eigen_assert( (i>=0) && ( + ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && i= 0 && BlockRows >= 0 && startRow + BlockRows <= xpr.rows() + && startCol >= 0 && BlockCols >= 0 && startCol + BlockCols <= xpr.cols()); + } + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline Block(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : Impl(xpr, startRow, startCol, blockRows, blockCols) + { + eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows) + && (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols)); + eigen_assert(startRow >= 0 && blockRows >= 0 && startRow <= xpr.rows() - blockRows + && startCol >= 0 && blockCols >= 0 && startCol <= xpr.cols() - blockCols); + } +}; + +// The generic default implementation for dense block simplu forward to the internal::BlockImpl_dense +// that must be specialized for direct and non-direct access... +template +class BlockImpl + : public internal::BlockImpl_dense +{ + typedef internal::BlockImpl_dense Impl; + typedef typename XprType::StorageIndex StorageIndex; + public: + typedef Impl Base; + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl) + EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index i) : Impl(xpr,i) {} + EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index startRow, Index startCol) : Impl(xpr, startRow, startCol) {} + EIGEN_DEVICE_FUNC + inline BlockImpl(XprType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) + : Impl(xpr, startRow, startCol, blockRows, blockCols) {} +}; + +namespace internal { + +/** \internal Internal implementation of dense Blocks in the general case. */ +template class BlockImpl_dense + : public internal::dense_xpr_base >::type +{ + typedef Block BlockType; + typedef typename internal::ref_selector::non_const_type XprTypeNested; + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(BlockType) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense) + + // class InnerIterator; // FIXME apparently never used + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index i) + : m_xpr(xpr), + // It is a row if and only if BlockRows==1 and BlockCols==XprType::ColsAtCompileTime, + // and it is a column if and only if BlockRows==XprType::RowsAtCompileTime and BlockCols==1, + // all other cases are invalid. + // The case a 1x1 matrix seems ambiguous, but the result is the same anyway. + m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0), + m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0), + m_blockRows(BlockRows==1 ? 1 : xpr.rows()), + m_blockCols(BlockCols==1 ? 1 : xpr.cols()) + {} + + /** Fixed-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol) + : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol), + m_blockRows(BlockRows), m_blockCols(BlockCols) + {} + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol), + m_blockRows(blockRows), m_blockCols(blockCols) + {} + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_blockRows.value(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_blockCols.value(); } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index rowId, Index colId) + { + EIGEN_STATIC_ASSERT_LVALUE(XprType) + return m_xpr.coeffRef(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_xpr.derived().coeffRef(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index rowId, Index colId) const + { + return m_xpr.coeff(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index index) + { + EIGEN_STATIC_ASSERT_LVALUE(XprType) + return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + EIGEN_DEVICE_FUNC + inline const CoeffReturnType coeff(Index index) const + { + return m_xpr.coeff(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + template + inline PacketScalar packet(Index rowId, Index colId) const + { + return m_xpr.template packet(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + template + inline void writePacket(Index rowId, Index colId, const PacketScalar& val) + { + m_xpr.template writePacket(rowId + m_startRow.value(), colId + m_startCol.value(), val); + } + + template + inline PacketScalar packet(Index index) const + { + return m_xpr.template packet + (m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + template + inline void writePacket(Index index, const PacketScalar& val) + { + m_xpr.template writePacket + (m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0), val); + } + + #ifdef EIGEN_PARSED_BY_DOXYGEN + /** \sa MapBase::data() */ + EIGEN_DEVICE_FUNC inline const Scalar* data() const; + EIGEN_DEVICE_FUNC inline Index innerStride() const; + EIGEN_DEVICE_FUNC inline Index outerStride() const; + #endif + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& nestedExpression() const + { + return m_xpr; + } + + EIGEN_DEVICE_FUNC + XprType& nestedExpression() { return m_xpr; } + + EIGEN_DEVICE_FUNC + StorageIndex startRow() const + { + return m_startRow.value(); + } + + EIGEN_DEVICE_FUNC + StorageIndex startCol() const + { + return m_startCol.value(); + } + + protected: + + XprTypeNested m_xpr; + const internal::variable_if_dynamic m_startRow; + const internal::variable_if_dynamic m_startCol; + const internal::variable_if_dynamic m_blockRows; + const internal::variable_if_dynamic m_blockCols; +}; + +/** \internal Internal implementation of dense Blocks in the direct access case.*/ +template +class BlockImpl_dense + : public MapBase > +{ + typedef Block BlockType; + typedef typename internal::ref_selector::non_const_type XprTypeNested; + enum { + XprTypeIsRowMajor = (int(traits::Flags)&RowMajorBit) != 0 + }; + public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(BlockType) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense) + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index i) + : Base(xpr.data() + i * ( ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && (!XprTypeIsRowMajor)) + || ((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && ( XprTypeIsRowMajor)) ? xpr.innerStride() : xpr.outerStride()), + BlockRows==1 ? 1 : xpr.rows(), + BlockCols==1 ? 1 : xpr.cols()), + m_xpr(xpr), + m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0), + m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0) + { + init(); + } + + /** Fixed-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol) + : Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol)), + m_xpr(xpr), m_startRow(startRow), m_startCol(startCol) + { + init(); + } + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol), blockRows, blockCols), + m_xpr(xpr), m_startRow(startRow), m_startCol(startCol) + { + init(); + } + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& nestedExpression() const + { + return m_xpr; + } + + EIGEN_DEVICE_FUNC + XprType& nestedExpression() { return m_xpr; } + + /** \sa MapBase::innerStride() */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return internal::traits::HasSameStorageOrderAsXprType + ? m_xpr.innerStride() + : m_xpr.outerStride(); + } + + /** \sa MapBase::outerStride() */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return m_outerStride; + } + + EIGEN_DEVICE_FUNC + StorageIndex startRow() const + { + return m_startRow.value(); + } + + EIGEN_DEVICE_FUNC + StorageIndex startCol() const + { + return m_startCol.value(); + } + + #ifndef __SUNPRO_CC + // FIXME sunstudio is not friendly with the above friend... + // META-FIXME there is no 'friend' keyword around here. Is this obsolete? + protected: + #endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal used by allowAligned() */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, const Scalar* data, Index blockRows, Index blockCols) + : Base(data, blockRows, blockCols), m_xpr(xpr) + { + init(); + } + #endif + + protected: + EIGEN_DEVICE_FUNC + void init() + { + m_outerStride = internal::traits::HasSameStorageOrderAsXprType + ? m_xpr.outerStride() + : m_xpr.innerStride(); + } + + XprTypeNested m_xpr; + const internal::variable_if_dynamic m_startRow; + const internal::variable_if_dynamic m_startCol; + Index m_outerStride; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_BLOCK_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h new file mode 100644 index 00000000000..8409d8749ad --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h @@ -0,0 +1,164 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ALLANDANY_H +#define EIGEN_ALLANDANY_H + +namespace Eigen { + +namespace internal { + +template +struct all_unroller +{ + typedef typename Derived::ExpressionTraits Traits; + enum { + col = (UnrollCount-1) / Traits::RowsAtCompileTime, + row = (UnrollCount-1) % Traits::RowsAtCompileTime + }; + + static inline bool run(const Derived &mat) + { + return all_unroller::run(mat) && mat.coeff(row, col); + } +}; + +template +struct all_unroller +{ + static inline bool run(const Derived &/*mat*/) { return true; } +}; + +template +struct all_unroller +{ + static inline bool run(const Derived &) { return false; } +}; + +template +struct any_unroller +{ + typedef typename Derived::ExpressionTraits Traits; + enum { + col = (UnrollCount-1) / Traits::RowsAtCompileTime, + row = (UnrollCount-1) % Traits::RowsAtCompileTime + }; + + static inline bool run(const Derived &mat) + { + return any_unroller::run(mat) || mat.coeff(row, col); + } +}; + +template +struct any_unroller +{ + static inline bool run(const Derived & /*mat*/) { return false; } +}; + +template +struct any_unroller +{ + static inline bool run(const Derived &) { return false; } +}; + +} // end namespace internal + +/** \returns true if all coefficients are true + * + * Example: \include MatrixBase_all.cpp + * Output: \verbinclude MatrixBase_all.out + * + * \sa any(), Cwise::operator<() + */ +template +inline bool DenseBase::all() const +{ + typedef internal::evaluator Evaluator; + enum { + unroll = SizeAtCompileTime != Dynamic + && SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits::AddCost) <= EIGEN_UNROLLING_LIMIT + }; + Evaluator evaluator(derived()); + if(unroll) + return internal::all_unroller::run(evaluator); + else + { + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if (!evaluator.coeff(i, j)) return false; + return true; + } +} + +/** \returns true if at least one coefficient is true + * + * \sa all() + */ +template +inline bool DenseBase::any() const +{ + typedef internal::evaluator Evaluator; + enum { + unroll = SizeAtCompileTime != Dynamic + && SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits::AddCost) <= EIGEN_UNROLLING_LIMIT + }; + Evaluator evaluator(derived()); + if(unroll) + return internal::any_unroller::run(evaluator); + else + { + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if (evaluator.coeff(i, j)) return true; + return false; + } +} + +/** \returns the number of coefficients which evaluate to true + * + * \sa all(), any() + */ +template +inline Eigen::Index DenseBase::count() const +{ + return derived().template cast().template cast().sum(); +} + +/** \returns true is \c *this contains at least one Not A Number (NaN). + * + * \sa allFinite() + */ +template +inline bool DenseBase::hasNaN() const +{ +#if EIGEN_COMP_MSVC || (defined __FAST_MATH__) + return derived().array().isNaN().any(); +#else + return !((derived().array()==derived().array()).all()); +#endif +} + +/** \returns true if \c *this contains only finite numbers, i.e., no NaN and no +/-INF values. + * + * \sa hasNaN() + */ +template +inline bool DenseBase::allFinite() const +{ +#if EIGEN_COMP_MSVC || (defined __FAST_MATH__) + return derived().array().isFinite().all(); +#else + return !((derived()-derived()).hasNaN()); +#endif +} + +} // end namespace Eigen + +#endif // EIGEN_ALLANDANY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h new file mode 100644 index 00000000000..d218e98143f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h @@ -0,0 +1,160 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_COMMAINITIALIZER_H +#define EIGEN_COMMAINITIALIZER_H + +namespace Eigen { + +/** \class CommaInitializer + * \ingroup Core_Module + * + * \brief Helper class used by the comma initializer operator + * + * This class is internally used to implement the comma initializer feature. It is + * the return type of MatrixBase::operator<<, and most of the time this is the only + * way it is used. + * + * \sa \blank \ref MatrixBaseCommaInitRef "MatrixBase::operator<<", CommaInitializer::finished() + */ +template +struct CommaInitializer +{ + typedef typename XprType::Scalar Scalar; + + EIGEN_DEVICE_FUNC + inline CommaInitializer(XprType& xpr, const Scalar& s) + : m_xpr(xpr), m_row(0), m_col(1), m_currentBlockRows(1) + { + m_xpr.coeffRef(0,0) = s; + } + + template + EIGEN_DEVICE_FUNC + inline CommaInitializer(XprType& xpr, const DenseBase& other) + : m_xpr(xpr), m_row(0), m_col(other.cols()), m_currentBlockRows(other.rows()) + { + m_xpr.block(0, 0, other.rows(), other.cols()) = other; + } + + /* Copy/Move constructor which transfers ownership. This is crucial in + * absence of return value optimization to avoid assertions during destruction. */ + // FIXME in C++11 mode this could be replaced by a proper RValue constructor + EIGEN_DEVICE_FUNC + inline CommaInitializer(const CommaInitializer& o) + : m_xpr(o.m_xpr), m_row(o.m_row), m_col(o.m_col), m_currentBlockRows(o.m_currentBlockRows) { + // Mark original object as finished. In absence of R-value references we need to const_cast: + const_cast(o).m_row = m_xpr.rows(); + const_cast(o).m_col = m_xpr.cols(); + const_cast(o).m_currentBlockRows = 0; + } + + /* inserts a scalar value in the target matrix */ + EIGEN_DEVICE_FUNC + CommaInitializer& operator,(const Scalar& s) + { + if (m_col==m_xpr.cols()) + { + m_row+=m_currentBlockRows; + m_col = 0; + m_currentBlockRows = 1; + eigen_assert(m_row + EIGEN_DEVICE_FUNC + CommaInitializer& operator,(const DenseBase& other) + { + if (m_col==m_xpr.cols() && (other.cols()!=0 || other.rows()!=m_currentBlockRows)) + { + m_row+=m_currentBlockRows; + m_col = 0; + m_currentBlockRows = other.rows(); + eigen_assert(m_row+m_currentBlockRows<=m_xpr.rows() + && "Too many rows passed to comma initializer (operator<<)"); + } + eigen_assert((m_col + other.cols() <= m_xpr.cols()) + && "Too many coefficients passed to comma initializer (operator<<)"); + eigen_assert(m_currentBlockRows==other.rows()); + m_xpr.template block + (m_row, m_col, other.rows(), other.cols()) = other; + m_col += other.cols(); + return *this; + } + + EIGEN_DEVICE_FUNC + inline ~CommaInitializer() +#if defined VERIFY_RAISES_ASSERT && (!defined EIGEN_NO_ASSERTION_CHECKING) && defined EIGEN_EXCEPTIONS + EIGEN_EXCEPTION_SPEC(Eigen::eigen_assert_exception) +#endif + { + finished(); + } + + /** \returns the built matrix once all its coefficients have been set. + * Calling finished is 100% optional. Its purpose is to write expressions + * like this: + * \code + * quaternion.fromRotationMatrix((Matrix3f() << axis0, axis1, axis2).finished()); + * \endcode + */ + EIGEN_DEVICE_FUNC + inline XprType& finished() { + eigen_assert(((m_row+m_currentBlockRows) == m_xpr.rows() || m_xpr.cols() == 0) + && m_col == m_xpr.cols() + && "Too few coefficients passed to comma initializer (operator<<)"); + return m_xpr; + } + + XprType& m_xpr; // target expression + Index m_row; // current row id + Index m_col; // current col id + Index m_currentBlockRows; // current block height +}; + +/** \anchor MatrixBaseCommaInitRef + * Convenient operator to set the coefficients of a matrix. + * + * The coefficients must be provided in a row major order and exactly match + * the size of the matrix. Otherwise an assertion is raised. + * + * Example: \include MatrixBase_set.cpp + * Output: \verbinclude MatrixBase_set.out + * + * \note According the c++ standard, the argument expressions of this comma initializer are evaluated in arbitrary order. + * + * \sa CommaInitializer::finished(), class CommaInitializer + */ +template +inline CommaInitializer DenseBase::operator<< (const Scalar& s) +{ + return CommaInitializer(*static_cast(this), s); +} + +/** \sa operator<<(const Scalar&) */ +template +template +inline CommaInitializer +DenseBase::operator<<(const DenseBase& other) +{ + return CommaInitializer(*static_cast(this), other); +} + +} // end namespace Eigen + +#endif // EIGEN_COMMAINITIALIZER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h new file mode 100644 index 00000000000..51a2e5f1b6f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h @@ -0,0 +1,175 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2016 Rasmus Munk Larsen (rmlarsen@google.com) +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CONDITIONESTIMATOR_H +#define EIGEN_CONDITIONESTIMATOR_H + +namespace Eigen { + +namespace internal { + +template +struct rcond_compute_sign { + static inline Vector run(const Vector& v) { + const RealVector v_abs = v.cwiseAbs(); + return (v_abs.array() == static_cast(0)) + .select(Vector::Ones(v.size()), v.cwiseQuotient(v_abs)); + } +}; + +// Partial specialization to avoid elementwise division for real vectors. +template +struct rcond_compute_sign { + static inline Vector run(const Vector& v) { + return (v.array() < static_cast(0)) + .select(-Vector::Ones(v.size()), Vector::Ones(v.size())); + } +}; + +/** + * \returns an estimate of ||inv(matrix)||_1 given a decomposition of + * \a matrix that implements .solve() and .adjoint().solve() methods. + * + * This function implements Algorithms 4.1 and 5.1 from + * http://www.maths.manchester.ac.uk/~higham/narep/narep135.pdf + * which also forms the basis for the condition number estimators in + * LAPACK. Since at most 10 calls to the solve method of dec are + * performed, the total cost is O(dims^2), as opposed to O(dims^3) + * needed to compute the inverse matrix explicitly. + * + * The most common usage is in estimating the condition number + * ||matrix||_1 * ||inv(matrix)||_1. The first term ||matrix||_1 can be + * computed directly in O(n^2) operations. + * + * Supports the following decompositions: FullPivLU, PartialPivLU, LDLT, and + * LLT. + * + * \sa FullPivLU, PartialPivLU, LDLT, LLT. + */ +template +typename Decomposition::RealScalar rcond_invmatrix_L1_norm_estimate(const Decomposition& dec) +{ + typedef typename Decomposition::MatrixType MatrixType; + typedef typename Decomposition::Scalar Scalar; + typedef typename Decomposition::RealScalar RealScalar; + typedef typename internal::plain_col_type::type Vector; + typedef typename internal::plain_col_type::type RealVector; + const bool is_complex = (NumTraits::IsComplex != 0); + + eigen_assert(dec.rows() == dec.cols()); + const Index n = dec.rows(); + if (n == 0) + return 0; + + // Disable Index to float conversion warning +#ifdef __INTEL_COMPILER + #pragma warning push + #pragma warning ( disable : 2259 ) +#endif + Vector v = dec.solve(Vector::Ones(n) / Scalar(n)); +#ifdef __INTEL_COMPILER + #pragma warning pop +#endif + + // lower_bound is a lower bound on + // ||inv(matrix)||_1 = sup_v ||inv(matrix) v||_1 / ||v||_1 + // and is the objective maximized by the ("super-") gradient ascent + // algorithm below. + RealScalar lower_bound = v.template lpNorm<1>(); + if (n == 1) + return lower_bound; + + // Gradient ascent algorithm follows: We know that the optimum is achieved at + // one of the simplices v = e_i, so in each iteration we follow a + // super-gradient to move towards the optimal one. + RealScalar old_lower_bound = lower_bound; + Vector sign_vector(n); + Vector old_sign_vector; + Index v_max_abs_index = -1; + Index old_v_max_abs_index = v_max_abs_index; + for (int k = 0; k < 4; ++k) + { + sign_vector = internal::rcond_compute_sign::run(v); + if (k > 0 && !is_complex && sign_vector == old_sign_vector) { + // Break if the solution stagnated. + break; + } + // v_max_abs_index = argmax |real( inv(matrix)^T * sign_vector )| + v = dec.adjoint().solve(sign_vector); + v.real().cwiseAbs().maxCoeff(&v_max_abs_index); + if (v_max_abs_index == old_v_max_abs_index) { + // Break if the solution stagnated. + break; + } + // Move to the new simplex e_j, where j = v_max_abs_index. + v = dec.solve(Vector::Unit(n, v_max_abs_index)); // v = inv(matrix) * e_j. + lower_bound = v.template lpNorm<1>(); + if (lower_bound <= old_lower_bound) { + // Break if the gradient step did not increase the lower_bound. + break; + } + if (!is_complex) { + old_sign_vector = sign_vector; + } + old_v_max_abs_index = v_max_abs_index; + old_lower_bound = lower_bound; + } + // The following calculates an independent estimate of ||matrix||_1 by + // multiplying matrix by a vector with entries of slowly increasing + // magnitude and alternating sign: + // v_i = (-1)^{i} (1 + (i / (dim-1))), i = 0,...,dim-1. + // This improvement to Hager's algorithm above is due to Higham. It was + // added to make the algorithm more robust in certain corner cases where + // large elements in the matrix might otherwise escape detection due to + // exact cancellation (especially when op and op_adjoint correspond to a + // sequence of backsubstitutions and permutations), which could cause + // Hager's algorithm to vastly underestimate ||matrix||_1. + Scalar alternating_sign(RealScalar(1)); + for (Index i = 0; i < n; ++i) { + // The static_cast is needed when Scalar is a complex and RealScalar implements expression templates + v[i] = alternating_sign * static_cast(RealScalar(1) + (RealScalar(i) / (RealScalar(n - 1)))); + alternating_sign = -alternating_sign; + } + v = dec.solve(v); + const RealScalar alternate_lower_bound = (2 * v.template lpNorm<1>()) / (3 * RealScalar(n)); + return numext::maxi(lower_bound, alternate_lower_bound); +} + +/** \brief Reciprocal condition number estimator. + * + * Computing a decomposition of a dense matrix takes O(n^3) operations, while + * this method estimates the condition number quickly and reliably in O(n^2) + * operations. + * + * \returns an estimate of the reciprocal condition number + * (1 / (||matrix||_1 * ||inv(matrix)||_1)) of matrix, given ||matrix||_1 and + * its decomposition. Supports the following decompositions: FullPivLU, + * PartialPivLU, LDLT, and LLT. + * + * \sa FullPivLU, PartialPivLU, LDLT, LLT. + */ +template +typename Decomposition::RealScalar +rcond_estimate_helper(typename Decomposition::RealScalar matrix_norm, const Decomposition& dec) +{ + typedef typename Decomposition::RealScalar RealScalar; + eigen_assert(dec.rows() == dec.cols()); + if (dec.rows() == 0) return NumTraits::infinity(); + if (matrix_norm == RealScalar(0)) return RealScalar(0); + if (dec.rows() == 1) return RealScalar(1); + const RealScalar inverse_matrix_norm = rcond_invmatrix_L1_norm_estimate(dec); + return (inverse_matrix_norm == RealScalar(0) ? RealScalar(0) + : (RealScalar(1) / inverse_matrix_norm) / matrix_norm); +} + +} // namespace internal + +} // namespace Eigen + +#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h new file mode 100644 index 00000000000..910889efa70 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h @@ -0,0 +1,1688 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2011 Benoit Jacob +// Copyright (C) 2011-2014 Gael Guennebaud +// Copyright (C) 2011-2012 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + + +#ifndef EIGEN_COREEVALUATORS_H +#define EIGEN_COREEVALUATORS_H + +namespace Eigen { + +namespace internal { + +// This class returns the evaluator kind from the expression storage kind. +// Default assumes index based accessors +template +struct storage_kind_to_evaluator_kind { + typedef IndexBased Kind; +}; + +// This class returns the evaluator shape from the expression storage kind. +// It can be Dense, Sparse, Triangular, Diagonal, SelfAdjoint, Band, etc. +template struct storage_kind_to_shape; + +template<> struct storage_kind_to_shape { typedef DenseShape Shape; }; +template<> struct storage_kind_to_shape { typedef SolverShape Shape; }; +template<> struct storage_kind_to_shape { typedef PermutationShape Shape; }; +template<> struct storage_kind_to_shape { typedef TranspositionsShape Shape; }; + +// Evaluators have to be specialized with respect to various criteria such as: +// - storage/structure/shape +// - scalar type +// - etc. +// Therefore, we need specialization of evaluator providing additional template arguments for each kind of evaluators. +// We currently distinguish the following kind of evaluators: +// - unary_evaluator for expressions taking only one arguments (CwiseUnaryOp, CwiseUnaryView, Transpose, MatrixWrapper, ArrayWrapper, Reverse, Replicate) +// - binary_evaluator for expression taking two arguments (CwiseBinaryOp) +// - ternary_evaluator for expression taking three arguments (CwiseTernaryOp) +// - product_evaluator for linear algebra products (Product); special case of binary_evaluator because it requires additional tags for dispatching. +// - mapbase_evaluator for Map, Block, Ref +// - block_evaluator for Block (special dispatching to a mapbase_evaluator or unary_evaluator) + +template< typename T, + typename Arg1Kind = typename evaluator_traits::Kind, + typename Arg2Kind = typename evaluator_traits::Kind, + typename Arg3Kind = typename evaluator_traits::Kind, + typename Arg1Scalar = typename traits::Scalar, + typename Arg2Scalar = typename traits::Scalar, + typename Arg3Scalar = typename traits::Scalar> struct ternary_evaluator; + +template< typename T, + typename LhsKind = typename evaluator_traits::Kind, + typename RhsKind = typename evaluator_traits::Kind, + typename LhsScalar = typename traits::Scalar, + typename RhsScalar = typename traits::Scalar> struct binary_evaluator; + +template< typename T, + typename Kind = typename evaluator_traits::Kind, + typename Scalar = typename T::Scalar> struct unary_evaluator; + +// evaluator_traits contains traits for evaluator + +template +struct evaluator_traits_base +{ + // by default, get evaluator kind and shape from storage + typedef typename storage_kind_to_evaluator_kind::StorageKind>::Kind Kind; + typedef typename storage_kind_to_shape::StorageKind>::Shape Shape; +}; + +// Default evaluator traits +template +struct evaluator_traits : public evaluator_traits_base +{ +}; + +template::Shape > +struct evaluator_assume_aliasing { + static const bool value = false; +}; + +// By default, we assume a unary expression: +template +struct evaluator : public unary_evaluator +{ + typedef unary_evaluator Base; + EIGEN_DEVICE_FUNC explicit evaluator(const T& xpr) : Base(xpr) {} +}; + + +// TODO: Think about const-correctness +template +struct evaluator + : evaluator +{ + EIGEN_DEVICE_FUNC + explicit evaluator(const T& xpr) : evaluator(xpr) {} +}; + +// ---------- base class for all evaluators ---------- + +template +struct evaluator_base : public noncopyable +{ + // TODO that's not very nice to have to propagate all these traits. They are currently only needed to handle outer,inner indices. + typedef traits ExpressionTraits; + + enum { + Alignment = 0 + }; +}; + +// -------------------- Matrix and Array -------------------- +// +// evaluator is a common base class for the +// Matrix and Array evaluators. +// Here we directly specialize evaluator. This is not really a unary expression, and it is, by definition, dense, +// so no need for more sophisticated dispatching. + +template +struct evaluator > + : evaluator_base +{ + typedef PlainObjectBase PlainObjectType; + typedef typename PlainObjectType::Scalar Scalar; + typedef typename PlainObjectType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = PlainObjectType::IsRowMajor, + IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime, + RowsAtCompileTime = PlainObjectType::RowsAtCompileTime, + ColsAtCompileTime = PlainObjectType::ColsAtCompileTime, + + CoeffReadCost = NumTraits::ReadCost, + Flags = traits::EvaluatorFlags, + Alignment = traits::Alignment + }; + + EIGEN_DEVICE_FUNC evaluator() + : m_data(0), + m_outerStride(IsVectorAtCompileTime ? 0 + : int(IsRowMajor) ? ColsAtCompileTime + : RowsAtCompileTime) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC explicit evaluator(const PlainObjectType& m) + : m_data(m.data()), m_outerStride(IsVectorAtCompileTime ? 0 : m.outerStride()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + if (IsRowMajor) + return m_data[row * m_outerStride.value() + col]; + else + return m_data[row + col * m_outerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_data[index]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + if (IsRowMajor) + return const_cast(m_data)[row * m_outerStride.value() + col]; + else + return const_cast(m_data)[row + col * m_outerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return const_cast(m_data)[index]; + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + if (IsRowMajor) + return ploadt(m_data + row * m_outerStride.value() + col); + else + return ploadt(m_data + row + col * m_outerStride.value()); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return ploadt(m_data + index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + if (IsRowMajor) + return pstoret + (const_cast(m_data) + row * m_outerStride.value() + col, x); + else + return pstoret + (const_cast(m_data) + row + col * m_outerStride.value(), x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + return pstoret(const_cast(m_data) + index, x); + } + +protected: + const Scalar *m_data; + + // We do not need to know the outer stride for vectors + variable_if_dynamic m_outerStride; +}; + +template +struct evaluator > + : evaluator > > +{ + typedef Matrix XprType; + + EIGEN_DEVICE_FUNC evaluator() {} + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& m) + : evaluator >(m) + { } +}; + +template +struct evaluator > + : evaluator > > +{ + typedef Array XprType; + + EIGEN_DEVICE_FUNC evaluator() {} + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& m) + : evaluator >(m) + { } +}; + +// -------------------- Transpose -------------------- + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef Transpose XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + Flags = evaluator::Flags ^ RowMajorBit, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& t) : m_argImpl(t.nestedExpression()) {} + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(col, row); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(col, row); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + typename XprType::Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(col, row); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_argImpl.template packet(index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + m_argImpl.template writePacket(col, row, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + m_argImpl.template writePacket(index, x); + } + +protected: + evaluator m_argImpl; +}; + +// -------------------- CwiseNullaryOp -------------------- +// Like Matrix and Array, this is not really a unary expression, so we directly specialize evaluator. +// Likewise, there is not need to more sophisticated dispatching here. + +template::value, + bool has_unary = has_unary_operator::value, + bool has_binary = has_binary_operator::value> +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { return op(i,j); } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { return op(i); } + + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { return op.template packetOp(i,j); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { return op.template packetOp(i); } +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType=0, IndexType=0) const { return op(); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType=0, IndexType=0) const { return op.template packetOp(); } +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j=0) const { return op(i,j); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j=0) const { return op.template packetOp(i,j); } +}; + +// We need the following specialization for vector-only functors assigned to a runtime vector, +// for instance, using linspace and assigning a RowVectorXd to a MatrixXd or even a row of a MatrixXd. +// In this case, i==0 and j is used for the actual iteration. +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { + eigen_assert(i==0 || j==0); + return op(i+j); + } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { + eigen_assert(i==0 || j==0); + return op.template packetOp(i+j); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { return op(i); } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { return op.template packetOp(i); } +}; + +template +struct nullary_wrapper {}; + +#if 0 && EIGEN_COMP_MSVC>0 +// Disable this ugly workaround. This is now handled in traits::match, +// but this piece of code might still become handly if some other weird compilation +// erros pop up again. + +// MSVC exhibits a weird compilation error when +// compiling: +// Eigen::MatrixXf A = MatrixXf::Random(3,3); +// Ref R = 2.f*A; +// and that has_*ary_operator> have not been instantiated yet. +// The "problem" is that evaluator<2.f*A> is instantiated by traits::match<2.f*A> +// and at that time has_*ary_operator returns true regardless of T. +// Then nullary_wrapper is badly instantiated as nullary_wrapper<.,.,true,true,true>. +// The trick is thus to defer the proper instantiation of nullary_wrapper when coeff(), +// and packet() are really instantiated as implemented below: + +// This is a simple wrapper around Index to enforce the re-instantiation of +// has_*ary_operator when needed. +template struct nullary_wrapper_workaround_msvc { + nullary_wrapper_workaround_msvc(const T&); + operator T()const; +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().operator()(op,i,j); + } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().operator()(op,i); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().template packetOp(op,i,j); + } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().template packetOp(op,i); + } +}; +#endif // MSVC workaround + +template +struct evaluator > + : evaluator_base > +{ + typedef CwiseNullaryOp XprType; + typedef typename internal::remove_all::type PlainObjectTypeCleaned; + + enum { + CoeffReadCost = internal::functor_traits::Cost, + + Flags = (evaluator::Flags + & ( HereditaryBits + | (functor_has_linear_access::ret ? LinearAccessBit : 0) + | (functor_traits::PacketAccess ? PacketAccessBit : 0))) + | (functor_traits::IsRepeatable ? 0 : EvalBeforeNestingBit), + Alignment = AlignedMax + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& n) + : m_functor(n.functor()), m_wrapper() + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(IndexType row, IndexType col) const + { + return m_wrapper(m_functor, row, col); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(IndexType index) const + { + return m_wrapper(m_functor,index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(IndexType row, IndexType col) const + { + return m_wrapper.template packetOp(m_functor, row, col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(IndexType index) const + { + return m_wrapper.template packetOp(m_functor, index); + } + +protected: + const NullaryOp m_functor; + const internal::nullary_wrapper m_wrapper; +}; + +// -------------------- CwiseUnaryOp -------------------- + +template +struct unary_evaluator, IndexBased > + : evaluator_base > +{ + typedef CwiseUnaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, + + Flags = evaluator::Flags + & (HereditaryBits | LinearAccessBit | (functor_traits::PacketAccess ? PacketAccessBit : 0)), + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit unary_evaluator(const XprType& op) + : m_functor(op.functor()), + m_argImpl(op.nestedExpression()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_argImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_argImpl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_argImpl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_argImpl.template packet(index)); + } + +protected: + const UnaryOp m_functor; + evaluator m_argImpl; +}; + +// -------------------- CwiseTernaryOp -------------------- + +// this is a ternary expression +template +struct evaluator > + : public ternary_evaluator > +{ + typedef CwiseTernaryOp XprType; + typedef ternary_evaluator > Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +template +struct ternary_evaluator, IndexBased, IndexBased> + : evaluator_base > +{ + typedef CwiseTernaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, + + Arg1Flags = evaluator::Flags, + Arg2Flags = evaluator::Flags, + Arg3Flags = evaluator::Flags, + SameType = is_same::value && is_same::value, + StorageOrdersAgree = (int(Arg1Flags)&RowMajorBit)==(int(Arg2Flags)&RowMajorBit) && (int(Arg1Flags)&RowMajorBit)==(int(Arg3Flags)&RowMajorBit), + Flags0 = (int(Arg1Flags) | int(Arg2Flags) | int(Arg3Flags)) & ( + HereditaryBits + | (int(Arg1Flags) & int(Arg2Flags) & int(Arg3Flags) & + ( (StorageOrdersAgree ? LinearAccessBit : 0) + | (functor_traits::PacketAccess && StorageOrdersAgree && SameType ? PacketAccessBit : 0) + ) + ) + ), + Flags = (Flags0 & ~RowMajorBit) | (Arg1Flags & RowMajorBit), + Alignment = EIGEN_PLAIN_ENUM_MIN( + EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment), + evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit ternary_evaluator(const XprType& xpr) + : m_functor(xpr.functor()), + m_arg1Impl(xpr.arg1()), + m_arg2Impl(xpr.arg2()), + m_arg3Impl(xpr.arg3()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_arg1Impl.coeff(row, col), m_arg2Impl.coeff(row, col), m_arg3Impl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_arg1Impl.coeff(index), m_arg2Impl.coeff(index), m_arg3Impl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_arg1Impl.template packet(row, col), + m_arg2Impl.template packet(row, col), + m_arg3Impl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_arg1Impl.template packet(index), + m_arg2Impl.template packet(index), + m_arg3Impl.template packet(index)); + } + +protected: + const TernaryOp m_functor; + evaluator m_arg1Impl; + evaluator m_arg2Impl; + evaluator m_arg3Impl; +}; + +// -------------------- CwiseBinaryOp -------------------- + +// this is a binary expression +template +struct evaluator > + : public binary_evaluator > +{ + typedef CwiseBinaryOp XprType; + typedef binary_evaluator > Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +template +struct binary_evaluator, IndexBased, IndexBased> + : evaluator_base > +{ + typedef CwiseBinaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, + + LhsFlags = evaluator::Flags, + RhsFlags = evaluator::Flags, + SameType = is_same::value, + StorageOrdersAgree = (int(LhsFlags)&RowMajorBit)==(int(RhsFlags)&RowMajorBit), + Flags0 = (int(LhsFlags) | int(RhsFlags)) & ( + HereditaryBits + | (int(LhsFlags) & int(RhsFlags) & + ( (StorageOrdersAgree ? LinearAccessBit : 0) + | (functor_traits::PacketAccess && StorageOrdersAgree && SameType ? PacketAccessBit : 0) + ) + ) + ), + Flags = (Flags0 & ~RowMajorBit) | (LhsFlags & RowMajorBit), + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment,evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit binary_evaluator(const XprType& xpr) + : m_functor(xpr.functor()), + m_lhsImpl(xpr.lhs()), + m_rhsImpl(xpr.rhs()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_lhsImpl.coeff(row, col), m_rhsImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_lhsImpl.coeff(index), m_rhsImpl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_lhsImpl.template packet(row, col), + m_rhsImpl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_lhsImpl.template packet(index), + m_rhsImpl.template packet(index)); + } + +protected: + const BinaryOp m_functor; + evaluator m_lhsImpl; + evaluator m_rhsImpl; +}; + +// -------------------- CwiseUnaryView -------------------- + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef CwiseUnaryView XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, + + Flags = (evaluator::Flags & (HereditaryBits | LinearAccessBit | DirectAccessBit)), + + Alignment = 0 // FIXME it is not very clear why alignment is necessarily lost... + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& op) + : m_unaryOp(op.functor()), + m_argImpl(op.nestedExpression()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_unaryOp(m_argImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_unaryOp(m_argImpl.coeff(index)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_unaryOp(m_argImpl.coeffRef(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_unaryOp(m_argImpl.coeffRef(index)); + } + +protected: + const UnaryOp m_unaryOp; + evaluator m_argImpl; +}; + +// -------------------- Map -------------------- + +// FIXME perhaps the PlainObjectType could be provided by Derived::PlainObject ? +// but that might complicate template specialization +template +struct mapbase_evaluator; + +template +struct mapbase_evaluator : evaluator_base +{ + typedef Derived XprType; + typedef typename XprType::PointerType PointerType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = XprType::RowsAtCompileTime, + ColsAtCompileTime = XprType::ColsAtCompileTime, + CoeffReadCost = NumTraits::ReadCost + }; + + EIGEN_DEVICE_FUNC explicit mapbase_evaluator(const XprType& map) + : m_data(const_cast(map.data())), + m_innerStride(map.innerStride()), + m_outerStride(map.outerStride()) + { + EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(evaluator::Flags&PacketAccessBit, internal::inner_stride_at_compile_time::ret==1), + PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_data[index * m_innerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_data[index * m_innerStride.value()]; + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + PointerType ptr = m_data + row * rowStride() + col * colStride(); + return internal::ploadt(ptr); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return internal::ploadt(m_data + index * m_innerStride.value()); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + PointerType ptr = m_data + row * rowStride() + col * colStride(); + return internal::pstoret(ptr, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + internal::pstoret(m_data + index * m_innerStride.value(), x); + } +protected: + EIGEN_DEVICE_FUNC + inline Index rowStride() const { return XprType::IsRowMajor ? m_outerStride.value() : m_innerStride.value(); } + EIGEN_DEVICE_FUNC + inline Index colStride() const { return XprType::IsRowMajor ? m_innerStride.value() : m_outerStride.value(); } + + PointerType m_data; + const internal::variable_if_dynamic m_innerStride; + const internal::variable_if_dynamic m_outerStride; +}; + +template +struct evaluator > + : public mapbase_evaluator, PlainObjectType> +{ + typedef Map XprType; + typedef typename XprType::Scalar Scalar; + // TODO: should check for smaller packet types once we can handle multi-sized packet types + typedef typename packet_traits::type PacketScalar; + + enum { + InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0 + ? int(PlainObjectType::InnerStrideAtCompileTime) + : int(StrideType::InnerStrideAtCompileTime), + OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0 + ? int(PlainObjectType::OuterStrideAtCompileTime) + : int(StrideType::OuterStrideAtCompileTime), + HasNoInnerStride = InnerStrideAtCompileTime == 1, + HasNoOuterStride = StrideType::OuterStrideAtCompileTime == 0, + HasNoStride = HasNoInnerStride && HasNoOuterStride, + IsDynamicSize = PlainObjectType::SizeAtCompileTime==Dynamic, + + PacketAccessMask = bool(HasNoInnerStride) ? ~int(0) : ~int(PacketAccessBit), + LinearAccessMask = bool(HasNoStride) || bool(PlainObjectType::IsVectorAtCompileTime) ? ~int(0) : ~int(LinearAccessBit), + Flags = int( evaluator::Flags) & (LinearAccessMask&PacketAccessMask), + + Alignment = int(MapOptions)&int(AlignedMask) + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& map) + : mapbase_evaluator(map) + { } +}; + +// -------------------- Ref -------------------- + +template +struct evaluator > + : public mapbase_evaluator, PlainObjectType> +{ + typedef Ref XprType; + + enum { + Flags = evaluator >::Flags, + Alignment = evaluator >::Alignment + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& ref) + : mapbase_evaluator(ref) + { } +}; + +// -------------------- Block -------------------- + +template::ret> struct block_evaluator; + +template +struct evaluator > + : block_evaluator +{ + typedef Block XprType; + typedef typename XprType::Scalar Scalar; + // TODO: should check for smaller packet types once we can handle multi-sized packet types + typedef typename packet_traits::type PacketScalar; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime, + + ArgTypeIsRowMajor = (int(evaluator::Flags)&RowMajorBit) != 0, + IsRowMajor = (MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1) ? 0 + : ArgTypeIsRowMajor, + HasSameStorageOrderAsArgType = (IsRowMajor == ArgTypeIsRowMajor), + InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + InnerStrideAtCompileTime = HasSameStorageOrderAsArgType + ? int(inner_stride_at_compile_time::ret) + : int(outer_stride_at_compile_time::ret), + OuterStrideAtCompileTime = HasSameStorageOrderAsArgType + ? int(outer_stride_at_compile_time::ret) + : int(inner_stride_at_compile_time::ret), + MaskPacketAccessBit = (InnerStrideAtCompileTime == 1 || HasSameStorageOrderAsArgType) ? PacketAccessBit : 0, + + FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1 || (InnerPanel && (evaluator::Flags&LinearAccessBit))) ? LinearAccessBit : 0, + FlagsRowMajorBit = XprType::Flags&RowMajorBit, + Flags0 = evaluator::Flags & ( (HereditaryBits & ~RowMajorBit) | + DirectAccessBit | + MaskPacketAccessBit), + Flags = Flags0 | FlagsLinearAccessBit | FlagsRowMajorBit, + + PacketAlignment = unpacket_traits::alignment, + Alignment0 = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) + && (OuterStrideAtCompileTime!=0) + && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % int(PacketAlignment)) == 0)) ? int(PacketAlignment) : 0, + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, Alignment0) + }; + typedef block_evaluator block_evaluator_type; + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& block) : block_evaluator_type(block) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } +}; + +// no direct-access => dispatch to a unary evaluator +template +struct block_evaluator + : unary_evaluator > +{ + typedef Block XprType; + + EIGEN_DEVICE_FUNC explicit block_evaluator(const XprType& block) + : unary_evaluator(block) + {} +}; + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef Block XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& block) + : m_argImpl(block.nestedExpression()), + m_startRow(block.startRow()), + m_startCol(block.startCol()), + m_linear_offset(InnerPanel?(XprType::IsRowMajor ? block.startRow()*block.cols() : block.startCol()*block.rows()):0) + { } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + RowsAtCompileTime = XprType::RowsAtCompileTime, + ForwardLinearAccess = InnerPanel && bool(evaluator::Flags&LinearAccessBit) + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(m_startRow.value() + row, m_startCol.value() + col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + if (ForwardLinearAccess) + return m_argImpl.coeff(m_linear_offset.value() + index); + else + return coeff(RowsAtCompileTime == 1 ? 0 : index, RowsAtCompileTime == 1 ? index : 0); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(m_startRow.value() + row, m_startCol.value() + col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + if (ForwardLinearAccess) + return m_argImpl.coeffRef(m_linear_offset.value() + index); + else + return coeffRef(RowsAtCompileTime == 1 ? 0 : index, RowsAtCompileTime == 1 ? index : 0); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(m_startRow.value() + row, m_startCol.value() + col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + if (ForwardLinearAccess) + return m_argImpl.template packet(m_linear_offset.value() + index); + else + return packet(RowsAtCompileTime == 1 ? 0 : index, + RowsAtCompileTime == 1 ? index : 0); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + return m_argImpl.template writePacket(m_startRow.value() + row, m_startCol.value() + col, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + if (ForwardLinearAccess) + return m_argImpl.template writePacket(m_linear_offset.value() + index, x); + else + return writePacket(RowsAtCompileTime == 1 ? 0 : index, + RowsAtCompileTime == 1 ? index : 0, + x); + } + +protected: + evaluator m_argImpl; + const variable_if_dynamic m_startRow; + const variable_if_dynamic m_startCol; + const variable_if_dynamic m_linear_offset; +}; + +// TODO: This evaluator does not actually use the child evaluator; +// all action is via the data() as returned by the Block expression. + +template +struct block_evaluator + : mapbase_evaluator, + typename Block::PlainObject> +{ + typedef Block XprType; + typedef typename XprType::Scalar Scalar; + + EIGEN_DEVICE_FUNC explicit block_evaluator(const XprType& block) + : mapbase_evaluator(block) + { + // TODO: for the 3.3 release, this should be turned to an internal assertion, but let's keep it as is for the beta lifetime + eigen_assert(((internal::UIntPtr(block.data()) % EIGEN_PLAIN_ENUM_MAX(1,evaluator::Alignment)) == 0) && "data is not aligned"); + } +}; + + +// -------------------- Select -------------------- +// NOTE shall we introduce a ternary_evaluator? + +// TODO enable vectorization for Select +template +struct evaluator > + : evaluator_base > +{ + typedef Select XprType; + enum { + CoeffReadCost = evaluator::CoeffReadCost + + EIGEN_PLAIN_ENUM_MAX(evaluator::CoeffReadCost, + evaluator::CoeffReadCost), + + Flags = (unsigned int)evaluator::Flags & evaluator::Flags & HereditaryBits, + + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& select) + : m_conditionImpl(select.conditionMatrix()), + m_thenImpl(select.thenMatrix()), + m_elseImpl(select.elseMatrix()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + if (m_conditionImpl.coeff(row, col)) + return m_thenImpl.coeff(row, col); + else + return m_elseImpl.coeff(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + if (m_conditionImpl.coeff(index)) + return m_thenImpl.coeff(index); + else + return m_elseImpl.coeff(index); + } + +protected: + evaluator m_conditionImpl; + evaluator m_thenImpl; + evaluator m_elseImpl; +}; + + +// -------------------- Replicate -------------------- + +template +struct unary_evaluator > + : evaluator_base > +{ + typedef Replicate XprType; + typedef typename XprType::CoeffReturnType CoeffReturnType; + enum { + Factor = (RowFactor==Dynamic || ColFactor==Dynamic) ? Dynamic : RowFactor*ColFactor + }; + typedef typename internal::nested_eval::type ArgTypeNested; + typedef typename internal::remove_all::type ArgTypeNestedCleaned; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + LinearAccessMask = XprType::IsVectorAtCompileTime ? LinearAccessBit : 0, + Flags = (evaluator::Flags & (HereditaryBits|LinearAccessMask) & ~RowMajorBit) | (traits::Flags & RowMajorBit), + + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& replicate) + : m_arg(replicate.nestedExpression()), + m_argImpl(m_arg), + m_rows(replicate.nestedExpression().rows()), + m_cols(replicate.nestedExpression().cols()) + {} + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + // try to avoid using modulo; this is a pure optimization strategy + const Index actual_row = internal::traits::RowsAtCompileTime==1 ? 0 + : RowFactor==1 ? row + : row % m_rows.value(); + const Index actual_col = internal::traits::ColsAtCompileTime==1 ? 0 + : ColFactor==1 ? col + : col % m_cols.value(); + + return m_argImpl.coeff(actual_row, actual_col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + // try to avoid using modulo; this is a pure optimization strategy + const Index actual_index = internal::traits::RowsAtCompileTime==1 + ? (ColFactor==1 ? index : index%m_cols.value()) + : (RowFactor==1 ? index : index%m_rows.value()); + + return m_argImpl.coeff(actual_index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + const Index actual_row = internal::traits::RowsAtCompileTime==1 ? 0 + : RowFactor==1 ? row + : row % m_rows.value(); + const Index actual_col = internal::traits::ColsAtCompileTime==1 ? 0 + : ColFactor==1 ? col + : col % m_cols.value(); + + return m_argImpl.template packet(actual_row, actual_col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + const Index actual_index = internal::traits::RowsAtCompileTime==1 + ? (ColFactor==1 ? index : index%m_cols.value()) + : (RowFactor==1 ? index : index%m_rows.value()); + + return m_argImpl.template packet(actual_index); + } + +protected: + const ArgTypeNested m_arg; + evaluator m_argImpl; + const variable_if_dynamic m_rows; + const variable_if_dynamic m_cols; +}; + + +// -------------------- PartialReduxExpr -------------------- + +template< typename ArgType, typename MemberOp, int Direction> +struct evaluator > + : evaluator_base > +{ + typedef PartialReduxExpr XprType; + typedef typename internal::nested_eval::type ArgTypeNested; + typedef typename internal::remove_all::type ArgTypeNestedCleaned; + typedef typename ArgType::Scalar InputScalar; + typedef typename XprType::Scalar Scalar; + enum { + TraversalSize = Direction==int(Vertical) ? int(ArgType::RowsAtCompileTime) : int(ArgType::ColsAtCompileTime) + }; + typedef typename MemberOp::template Cost CostOpType; + enum { + CoeffReadCost = TraversalSize==Dynamic ? HugeCost + : TraversalSize * evaluator::CoeffReadCost + int(CostOpType::value), + + Flags = (traits::Flags&RowMajorBit) | (evaluator::Flags&(HereditaryBits&(~RowMajorBit))) | LinearAccessBit, + + Alignment = 0 // FIXME this will need to be improved once PartialReduxExpr is vectorized + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType xpr) + : m_arg(xpr.nestedExpression()), m_functor(xpr.functor()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(TraversalSize==Dynamic ? HugeCost : int(CostOpType::value)); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const Scalar coeff(Index i, Index j) const + { + if (Direction==Vertical) + return m_functor(m_arg.col(j)); + else + return m_functor(m_arg.row(i)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const Scalar coeff(Index index) const + { + if (Direction==Vertical) + return m_functor(m_arg.col(index)); + else + return m_functor(m_arg.row(index)); + } + +protected: + typename internal::add_const_on_value_type::type m_arg; + const MemberOp m_functor; +}; + + +// -------------------- MatrixWrapper and ArrayWrapper -------------------- +// +// evaluator_wrapper_base is a common base class for the +// MatrixWrapper and ArrayWrapper evaluators. + +template +struct evaluator_wrapper_base + : evaluator_base +{ + typedef typename remove_all::type ArgType; + enum { + CoeffReadCost = evaluator::CoeffReadCost, + Flags = evaluator::Flags, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit evaluator_wrapper_base(const ArgType& arg) : m_argImpl(arg) {} + + typedef typename ArgType::Scalar Scalar; + typedef typename ArgType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(row, col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_argImpl.template packet(index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + m_argImpl.template writePacket(row, col, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + m_argImpl.template writePacket(index, x); + } + +protected: + evaluator m_argImpl; +}; + +template +struct unary_evaluator > + : evaluator_wrapper_base > +{ + typedef MatrixWrapper XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& wrapper) + : evaluator_wrapper_base >(wrapper.nestedExpression()) + { } +}; + +template +struct unary_evaluator > + : evaluator_wrapper_base > +{ + typedef ArrayWrapper XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& wrapper) + : evaluator_wrapper_base >(wrapper.nestedExpression()) + { } +}; + + +// -------------------- Reverse -------------------- + +// defined in Reverse.h: +template struct reverse_packet_cond; + +template +struct unary_evaluator > + : evaluator_base > +{ + typedef Reverse XprType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = XprType::IsRowMajor, + IsColMajor = !IsRowMajor, + ReverseRow = (Direction == Vertical) || (Direction == BothDirections), + ReverseCol = (Direction == Horizontal) || (Direction == BothDirections), + ReversePacket = (Direction == BothDirections) + || ((Direction == Vertical) && IsColMajor) + || ((Direction == Horizontal) && IsRowMajor), + + CoeffReadCost = evaluator::CoeffReadCost, + + // let's enable LinearAccess only with vectorization because of the product overhead + // FIXME enable DirectAccess with negative strides? + Flags0 = evaluator::Flags, + LinearAccess = ( (Direction==BothDirections) && (int(Flags0)&PacketAccessBit) ) + || ((ReverseRow && XprType::ColsAtCompileTime==1) || (ReverseCol && XprType::RowsAtCompileTime==1)) + ? LinearAccessBit : 0, + + Flags = int(Flags0) & (HereditaryBits | PacketAccessBit | LinearAccess), + + Alignment = 0 // FIXME in some rare cases, Alignment could be preserved, like a Vector4f. + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& reverse) + : m_argImpl(reverse.nestedExpression()), + m_rows(ReverseRow ? reverse.nestedExpression().rows() : 1), + m_cols(ReverseCol ? reverse.nestedExpression().cols() : 1) + { } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(ReverseRow ? m_rows.value() - row - 1 : row, + ReverseCol ? m_cols.value() - col - 1 : col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(m_rows.value() * m_cols.value() - index - 1); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(ReverseRow ? m_rows.value() - row - 1 : row, + ReverseCol ? m_cols.value() - col - 1 : col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(m_rows.value() * m_cols.value() - index - 1); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + enum { + PacketSize = unpacket_traits::size, + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1 + }; + typedef internal::reverse_packet_cond reverse_packet; + return reverse_packet::run(m_argImpl.template packet( + ReverseRow ? m_rows.value() - row - OffsetRow : row, + ReverseCol ? m_cols.value() - col - OffsetCol : col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + enum { PacketSize = unpacket_traits::size }; + return preverse(m_argImpl.template packet(m_rows.value() * m_cols.value() - index - PacketSize)); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + // FIXME we could factorize some code with packet(i,j) + enum { + PacketSize = unpacket_traits::size, + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1 + }; + typedef internal::reverse_packet_cond reverse_packet; + m_argImpl.template writePacket( + ReverseRow ? m_rows.value() - row - OffsetRow : row, + ReverseCol ? m_cols.value() - col - OffsetCol : col, + reverse_packet::run(x)); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + enum { PacketSize = unpacket_traits::size }; + m_argImpl.template writePacket + (m_rows.value() * m_cols.value() - index - PacketSize, preverse(x)); + } + +protected: + evaluator m_argImpl; + + // If we do not reverse rows, then we do not need to know the number of rows; same for columns + // Nonetheless, in this case it is important to set to 1 such that the coeff(index) method works fine for vectors. + const variable_if_dynamic m_rows; + const variable_if_dynamic m_cols; +}; + + +// -------------------- Diagonal -------------------- + +template +struct evaluator > + : evaluator_base > +{ + typedef Diagonal XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + + Flags = (unsigned int)(evaluator::Flags & (HereditaryBits | DirectAccessBit) & ~RowMajorBit) | LinearAccessBit, + + Alignment = 0 + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& diagonal) + : m_argImpl(diagonal.nestedExpression()), + m_index(diagonal.index()) + { } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index) const + { + return m_argImpl.coeff(row + rowOffset(), row + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index + rowOffset(), index + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index) + { + return m_argImpl.coeffRef(row + rowOffset(), row + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index + rowOffset(), index + colOffset()); + } + +protected: + evaluator m_argImpl; + const internal::variable_if_dynamicindex m_index; + +private: + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index rowOffset() const { return m_index.value() > 0 ? 0 : -m_index.value(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index colOffset() const { return m_index.value() > 0 ? m_index.value() : 0; } +}; + + +//---------------------------------------------------------------------- +// deprecated code +//---------------------------------------------------------------------- + +// -------------------- EvalToTemp -------------------- + +// expression class for evaluating nested expression to a temporary + +template class EvalToTemp; + +template +struct traits > + : public traits +{ }; + +template +class EvalToTemp + : public dense_xpr_base >::type +{ + public: + + typedef typename dense_xpr_base::type Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(EvalToTemp) + + explicit EvalToTemp(const ArgType& arg) + : m_arg(arg) + { } + + const ArgType& arg() const + { + return m_arg; + } + + Index rows() const + { + return m_arg.rows(); + } + + Index cols() const + { + return m_arg.cols(); + } + + private: + const ArgType& m_arg; +}; + +template +struct evaluator > + : public evaluator +{ + typedef EvalToTemp XprType; + typedef typename ArgType::PlainObject PlainObject; + typedef evaluator Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) + : m_result(xpr.arg()) + { + ::new (static_cast(this)) Base(m_result); + } + + // This constructor is used when nesting an EvalTo evaluator in another evaluator + EIGEN_DEVICE_FUNC evaluator(const ArgType& arg) + : m_result(arg) + { + ::new (static_cast(this)) Base(m_result); + } + +protected: + PlainObject m_result; +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_COREEVALUATORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h new file mode 100644 index 00000000000..4eb42b93af1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h @@ -0,0 +1,127 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_COREITERATORS_H +#define EIGEN_COREITERATORS_H + +namespace Eigen { + +/* This file contains the respective InnerIterator definition of the expressions defined in Eigen/Core + */ + +namespace internal { + +template +class inner_iterator_selector; + +} + +/** \class InnerIterator + * \brief An InnerIterator allows to loop over the element of any matrix expression. + * + * \warning To be used with care because an evaluator is constructed every time an InnerIterator iterator is constructed. + * + * TODO: add a usage example + */ +template +class InnerIterator +{ +protected: + typedef internal::inner_iterator_selector::Kind> IteratorType; + typedef internal::evaluator EvaluatorType; + typedef typename internal::traits::Scalar Scalar; +public: + /** Construct an iterator over the \a outerId -th row or column of \a xpr */ + InnerIterator(const XprType &xpr, const Index &outerId) + : m_eval(xpr), m_iter(m_eval, outerId, xpr.innerSize()) + {} + + /// \returns the value of the current coefficient. + EIGEN_STRONG_INLINE Scalar value() const { return m_iter.value(); } + /** Increment the iterator \c *this to the next non-zero coefficient. + * Explicit zeros are not skipped over. To skip explicit zeros, see class SparseView + */ + EIGEN_STRONG_INLINE InnerIterator& operator++() { m_iter.operator++(); return *this; } + /// \returns the column or row index of the current coefficient. + EIGEN_STRONG_INLINE Index index() const { return m_iter.index(); } + /// \returns the row index of the current coefficient. + EIGEN_STRONG_INLINE Index row() const { return m_iter.row(); } + /// \returns the column index of the current coefficient. + EIGEN_STRONG_INLINE Index col() const { return m_iter.col(); } + /// \returns \c true if the iterator \c *this still references a valid coefficient. + EIGEN_STRONG_INLINE operator bool() const { return m_iter; } + +protected: + EvaluatorType m_eval; + IteratorType m_iter; +private: + // If you get here, then you're not using the right InnerIterator type, e.g.: + // SparseMatrix A; + // SparseMatrix::InnerIterator it(A,0); + template InnerIterator(const EigenBase&,Index outer); +}; + +namespace internal { + +// Generic inner iterator implementation for dense objects +template +class inner_iterator_selector +{ +protected: + typedef evaluator EvaluatorType; + typedef typename traits::Scalar Scalar; + enum { IsRowMajor = (XprType::Flags&RowMajorBit)==RowMajorBit }; + +public: + EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &innerSize) + : m_eval(eval), m_inner(0), m_outer(outerId), m_end(innerSize) + {} + + EIGEN_STRONG_INLINE Scalar value() const + { + return (IsRowMajor) ? m_eval.coeff(m_outer, m_inner) + : m_eval.coeff(m_inner, m_outer); + } + + EIGEN_STRONG_INLINE inner_iterator_selector& operator++() { m_inner++; return *this; } + + EIGEN_STRONG_INLINE Index index() const { return m_inner; } + inline Index row() const { return IsRowMajor ? m_outer : index(); } + inline Index col() const { return IsRowMajor ? index() : m_outer; } + + EIGEN_STRONG_INLINE operator bool() const { return m_inner < m_end && m_inner>=0; } + +protected: + const EvaluatorType& m_eval; + Index m_inner; + const Index m_outer; + const Index m_end; +}; + +// For iterator-based evaluator, inner-iterator is already implemented as +// evaluator<>::InnerIterator +template +class inner_iterator_selector + : public evaluator::InnerIterator +{ +protected: + typedef typename evaluator::InnerIterator Base; + typedef evaluator EvaluatorType; + +public: + EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &/*innerSize*/) + : Base(eval, outerId) + {} +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_COREITERATORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h new file mode 100644 index 00000000000..a36765e396b --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h @@ -0,0 +1,184 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_BINARY_OP_H +#define EIGEN_CWISE_BINARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > +{ + // we must not inherit from traits since it has + // the potential to cause problems with MSVC + typedef typename remove_all::type Ancestor; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime + }; + + // even though we require Lhs and Rhs to have the same scalar type (see CwiseBinaryOp constructor), + // we still want to handle the case when the result type is different. + typedef typename result_of< + BinaryOp( + const typename Lhs::Scalar&, + const typename Rhs::Scalar& + ) + >::type Scalar; + typedef typename cwise_promote_storage_type::StorageKind, + typename traits::StorageKind, + BinaryOp>::ret StorageKind; + typedef typename promote_index_type::StorageIndex, + typename traits::StorageIndex>::type StorageIndex; + typedef typename Lhs::Nested LhsNested; + typedef typename Rhs::Nested RhsNested; + typedef typename remove_reference::type _LhsNested; + typedef typename remove_reference::type _RhsNested; + enum { + Flags = cwise_promote_storage_order::StorageKind,typename traits::StorageKind,_LhsNested::Flags & RowMajorBit,_RhsNested::Flags & RowMajorBit>::value + }; +}; +} // end namespace internal + +template +class CwiseBinaryOpImpl; + +/** \class CwiseBinaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise binary operator is applied to two expressions + * + * \tparam BinaryOp template functor implementing the operator + * \tparam LhsType the type of the left-hand side + * \tparam RhsType the type of the right-hand side + * + * This class represents an expression where a coefficient-wise binary operator is applied to two expressions. + * It is the return type of binary operators, by which we mean only those binary operators where + * both the left-hand side and the right-hand side are Eigen expressions. + * For example, the return type of matrix1+matrix2 is a CwiseBinaryOp. + * + * Most of the time, this is the only way that it is used, so you typically don't have to name + * CwiseBinaryOp types explicitly. + * + * \sa MatrixBase::binaryExpr(const MatrixBase &,const CustomBinaryOp &) const, class CwiseUnaryOp, class CwiseNullaryOp + */ +template +class CwiseBinaryOp : + public CwiseBinaryOpImpl< + BinaryOp, LhsType, RhsType, + typename internal::cwise_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + BinaryOp>::ret>, + internal::no_assignment_operator +{ + public: + + typedef typename internal::remove_all::type Functor; + typedef typename internal::remove_all::type Lhs; + typedef typename internal::remove_all::type Rhs; + + typedef typename CwiseBinaryOpImpl< + BinaryOp, LhsType, RhsType, + typename internal::cwise_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + BinaryOp>::ret>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseBinaryOp) + + typedef typename internal::ref_selector::type LhsNested; + typedef typename internal::ref_selector::type RhsNested; + typedef typename internal::remove_reference::type _LhsNested; + typedef typename internal::remove_reference::type _RhsNested; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CwiseBinaryOp(const Lhs& aLhs, const Rhs& aRhs, const BinaryOp& func = BinaryOp()) + : m_lhs(aLhs), m_rhs(aRhs), m_functor(func) + { + EIGEN_CHECK_BINARY_COMPATIBILIY(BinaryOp,typename Lhs::Scalar,typename Rhs::Scalar); + // require the sizes to match + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Lhs, Rhs) + eigen_assert(aLhs.rows() == aRhs.rows() && aLhs.cols() == aRhs.cols()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { + // return the fixed size type if available to enable compile time optimizations + if (internal::traits::type>::RowsAtCompileTime==Dynamic) + return m_rhs.rows(); + else + return m_lhs.rows(); + } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { + // return the fixed size type if available to enable compile time optimizations + if (internal::traits::type>::ColsAtCompileTime==Dynamic) + return m_rhs.cols(); + else + return m_lhs.cols(); + } + + /** \returns the left hand side nested expression */ + EIGEN_DEVICE_FUNC + const _LhsNested& lhs() const { return m_lhs; } + /** \returns the right hand side nested expression */ + EIGEN_DEVICE_FUNC + const _RhsNested& rhs() const { return m_rhs; } + /** \returns the functor representing the binary operation */ + EIGEN_DEVICE_FUNC + const BinaryOp& functor() const { return m_functor; } + + protected: + LhsNested m_lhs; + RhsNested m_rhs; + const BinaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseBinaryOpImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +/** replaces \c *this by \c *this - \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_STRONG_INLINE Derived & +MatrixBase::operator-=(const MatrixBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this + \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_STRONG_INLINE Derived & +MatrixBase::operator+=(const MatrixBase& other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_CWISE_BINARY_OP_H + diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h new file mode 100644 index 00000000000..ddd607e3832 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h @@ -0,0 +1,866 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_NULLARY_OP_H +#define EIGEN_CWISE_NULLARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits +{ + enum { + Flags = traits::Flags & RowMajorBit + }; +}; + +} // namespace internal + +/** \class CwiseNullaryOp + * \ingroup Core_Module + * + * \brief Generic expression of a matrix where all coefficients are defined by a functor + * + * \tparam NullaryOp template functor implementing the operator + * \tparam PlainObjectType the underlying plain matrix/array type + * + * This class represents an expression of a generic nullary operator. + * It is the return type of the Ones(), Zero(), Constant(), Identity() and Random() methods, + * and most of the time this is the only way it is used. + * + * However, if you want to write a function returning such an expression, you + * will need to use this class. + * + * The functor NullaryOp must expose one of the following method: + + + + +
\c operator()() if the procedural generation does not depend on the coefficient entries (e.g., random numbers)
\c operator()(Index i)if the procedural generation makes sense for vectors only and that it depends on the coefficient index \c i (e.g., linspace)
\c operator()(Index i,Index j)if the procedural generation depends on the matrix coordinates \c i, \c j (e.g., to generate a checkerboard with 0 and 1)
+ * It is also possible to expose the last two operators if the generation makes sense for matrices but can be optimized for vectors. + * + * See DenseBase::NullaryExpr(Index,const CustomNullaryOp&) for an example binding + * C++11 random number generators. + * + * A nullary expression can also be used to implement custom sophisticated matrix manipulations + * that cannot be covered by the existing set of natively supported matrix manipulations. + * See this \ref TopicCustomizing_NullaryExpr "page" for some examples and additional explanations + * on the behavior of CwiseNullaryOp. + * + * \sa class CwiseUnaryOp, class CwiseBinaryOp, DenseBase::NullaryExpr + */ +template +class CwiseNullaryOp : public internal::dense_xpr_base< CwiseNullaryOp >::type, internal::no_assignment_operator +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(CwiseNullaryOp) + + EIGEN_DEVICE_FUNC + CwiseNullaryOp(Index rows, Index cols, const NullaryOp& func = NullaryOp()) + : m_rows(rows), m_cols(cols), m_functor(func) + { + eigen_assert(rows >= 0 + && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows) + && cols >= 0 + && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols)); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { return m_rows.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { return m_cols.value(); } + + /** \returns the functor representing the nullary operation */ + EIGEN_DEVICE_FUNC + const NullaryOp& functor() const { return m_functor; } + + protected: + const internal::variable_if_dynamic m_rows; + const internal::variable_if_dynamic m_cols; + const NullaryOp m_functor; +}; + + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func) +{ + return CwiseNullaryOp(rows, cols, func); +} + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * Here is an example with C++11 random generators: \include random_cpp11.cpp + * Output: \verbinclude random_cpp11.out + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(Index size, const CustomNullaryOp& func) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + if(RowsAtCompileTime == 1) return CwiseNullaryOp(1, size, func); + else return CwiseNullaryOp(size, 1, func); +} + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * This variant is only for fixed-size DenseBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(const CustomNullaryOp& func) +{ + return CwiseNullaryOp(RowsAtCompileTime, ColsAtCompileTime, func); +} + +/** \returns an expression of a constant matrix of value \a value + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this DenseBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(Index rows, Index cols, const Scalar& value) +{ + return DenseBase::NullaryExpr(rows, cols, internal::scalar_constant_op(value)); +} + +/** \returns an expression of a constant matrix of value \a value + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this DenseBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(Index size, const Scalar& value) +{ + return DenseBase::NullaryExpr(size, internal::scalar_constant_op(value)); +} + +/** \returns an expression of a constant matrix of value \a value + * + * This variant is only for fixed-size DenseBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(const Scalar& value) +{ + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_constant_op(value)); +} + +/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(Index,const Scalar&,const Scalar&) + * + * \sa LinSpaced(Index,Scalar,Scalar), setLinSpaced(Index,const Scalar&,const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return DenseBase::NullaryExpr(size, internal::linspaced_op(low,high,size)); +} + +/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(const Scalar&,const Scalar&) + * + * \sa LinSpaced(Scalar,Scalar) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Sequential_t, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op(low,high,Derived::SizeAtCompileTime)); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function generates 'size' equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * Example: \include DenseBase_LinSpaced.cpp + * Output: \verbinclude DenseBase_LinSpaced.out + * + * For integer scalar types, an even spacing is possible if and only if the length of the range, + * i.e., \c high-low is a scalar multiple of \c size-1, or if \c size is a scalar multiple of the + * number of values \c high-low+1 (meaning each value can be repeated the same number of time). + * If one of these two considions is not satisfied, then \c high is lowered to the largest value + * satisfying one of this constraint. + * Here are some examples: + * + * Example: \include DenseBase_LinSpacedInt.cpp + * Output: \verbinclude DenseBase_LinSpacedInt.out + * + * \sa setLinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Index size, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return DenseBase::NullaryExpr(size, internal::linspaced_op(low,high,size)); +} + +/** + * \copydoc DenseBase::LinSpaced(Index, const Scalar&, const Scalar&) + * Special version for fixed size types which does not require the size parameter. + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op(low,high,Derived::SizeAtCompileTime)); +} + +/** \returns true if all coefficients in this matrix are approximately equal to \a val, to within precision \a prec */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isApproxToConstant +(const Scalar& val, const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if(!internal::isApprox(self.coeff(i, j), val, prec)) + return false; + return true; +} + +/** This is just an alias for isApproxToConstant(). + * + * \returns true if all coefficients in this matrix are approximately equal to \a value, to within precision \a prec */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isConstant +(const Scalar& val, const RealScalar& prec) const +{ + return isApproxToConstant(val, prec); +} + +/** Alias for setConstant(): sets all coefficients in this expression to \a val. + * + * \sa setConstant(), Constant(), class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void DenseBase::fill(const Scalar& val) +{ + setConstant(val); +} + +/** Sets all coefficients in this expression to value \a val. + * + * \sa fill(), setConstant(Index,const Scalar&), setConstant(Index,Index,const Scalar&), setZero(), setOnes(), Constant(), class CwiseNullaryOp, setZero(), setOnes() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setConstant(const Scalar& val) +{ + return derived() = Constant(rows(), cols(), val); +} + +/** Resizes to the given \a size, and sets all coefficients in this expression to the given value \a val. + * + * \only_for_vectors + * + * Example: \include Matrix_setConstant_int.cpp + * Output: \verbinclude Matrix_setConstant_int.out + * + * \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setConstant(Index size, const Scalar& val) +{ + resize(size); + return setConstant(val); +} + +/** Resizes to the given size, and sets all coefficients in this expression to the given value \a val. + * + * \param rows the new number of rows + * \param cols the new number of columns + * \param val the value to which all coefficients are set + * + * Example: \include Matrix_setConstant_int_int.cpp + * Output: \verbinclude Matrix_setConstant_int_int.out + * + * \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setConstant(Index rows, Index cols, const Scalar& val) +{ + resize(rows, cols); + return setConstant(val); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function generates 'size' equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * Example: \include DenseBase_setLinSpaced.cpp + * Output: \verbinclude DenseBase_setLinSpaced.out + * + * For integer scalar types, do not miss the explanations on the definition + * of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink. + * + * \sa LinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setLinSpaced(Index newSize, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return derived() = Derived::NullaryExpr(newSize, internal::linspaced_op(low,high,newSize)); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function fills \c *this with equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * For integer scalar types, do not miss the explanations on the definition + * of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink. + * + * \sa LinSpaced(Index,const Scalar&,const Scalar&), setLinSpaced(Index, const Scalar&, const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setLinSpaced(const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return setLinSpaced(size(), low, high); +} + +// zero: + +/** \returns an expression of a zero matrix. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * Example: \include MatrixBase_zero_int_int.cpp + * Output: \verbinclude MatrixBase_zero_int_int.out + * + * \sa Zero(), Zero(Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero(Index rows, Index cols) +{ + return Constant(rows, cols, Scalar(0)); +} + +/** \returns an expression of a zero vector. + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * Example: \include MatrixBase_zero_int.cpp + * Output: \verbinclude MatrixBase_zero_int.out + * + * \sa Zero(), Zero(Index,Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero(Index size) +{ + return Constant(size, Scalar(0)); +} + +/** \returns an expression of a fixed-size zero matrix or vector. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_zero.cpp + * Output: \verbinclude MatrixBase_zero.out + * + * \sa Zero(Index), Zero(Index,Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero() +{ + return Constant(Scalar(0)); +} + +/** \returns true if *this is approximately equal to the zero matrix, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isZero.cpp + * Output: \verbinclude MatrixBase_isZero.out + * + * \sa class CwiseNullaryOp, Zero() + */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isZero(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast(1), prec)) + return false; + return true; +} + +/** Sets all coefficients in this expression to zero. + * + * Example: \include MatrixBase_setZero.cpp + * Output: \verbinclude MatrixBase_setZero.out + * + * \sa class CwiseNullaryOp, Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setZero() +{ + return setConstant(Scalar(0)); +} + +/** Resizes to the given \a size, and sets all coefficients in this expression to zero. + * + * \only_for_vectors + * + * Example: \include Matrix_setZero_int.cpp + * Output: \verbinclude Matrix_setZero_int.out + * + * \sa DenseBase::setZero(), setZero(Index,Index), class CwiseNullaryOp, DenseBase::Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setZero(Index newSize) +{ + resize(newSize); + return setConstant(Scalar(0)); +} + +/** Resizes to the given size, and sets all coefficients in this expression to zero. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setZero_int_int.cpp + * Output: \verbinclude Matrix_setZero_int_int.out + * + * \sa DenseBase::setZero(), setZero(Index), class CwiseNullaryOp, DenseBase::Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setZero(Index rows, Index cols) +{ + resize(rows, cols); + return setConstant(Scalar(0)); +} + +// ones: + +/** \returns an expression of a matrix where all coefficients equal one. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Ones() should be used + * instead. + * + * Example: \include MatrixBase_ones_int_int.cpp + * Output: \verbinclude MatrixBase_ones_int_int.out + * + * \sa Ones(), Ones(Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones(Index rows, Index cols) +{ + return Constant(rows, cols, Scalar(1)); +} + +/** \returns an expression of a vector where all coefficients equal one. + * + * The parameter \a newSize is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Ones() should be used + * instead. + * + * Example: \include MatrixBase_ones_int.cpp + * Output: \verbinclude MatrixBase_ones_int.out + * + * \sa Ones(), Ones(Index,Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones(Index newSize) +{ + return Constant(newSize, Scalar(1)); +} + +/** \returns an expression of a fixed-size matrix or vector where all coefficients equal one. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_ones.cpp + * Output: \verbinclude MatrixBase_ones.out + * + * \sa Ones(Index), Ones(Index,Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones() +{ + return Constant(Scalar(1)); +} + +/** \returns true if *this is approximately equal to the matrix where all coefficients + * are equal to 1, within the precision given by \a prec. + * + * Example: \include MatrixBase_isOnes.cpp + * Output: \verbinclude MatrixBase_isOnes.out + * + * \sa class CwiseNullaryOp, Ones() + */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isOnes +(const RealScalar& prec) const +{ + return isApproxToConstant(Scalar(1), prec); +} + +/** Sets all coefficients in this expression to one. + * + * Example: \include MatrixBase_setOnes.cpp + * Output: \verbinclude MatrixBase_setOnes.out + * + * \sa class CwiseNullaryOp, Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setOnes() +{ + return setConstant(Scalar(1)); +} + +/** Resizes to the given \a newSize, and sets all coefficients in this expression to one. + * + * \only_for_vectors + * + * Example: \include Matrix_setOnes_int.cpp + * Output: \verbinclude Matrix_setOnes_int.out + * + * \sa MatrixBase::setOnes(), setOnes(Index,Index), class CwiseNullaryOp, MatrixBase::Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setOnes(Index newSize) +{ + resize(newSize); + return setConstant(Scalar(1)); +} + +/** Resizes to the given size, and sets all coefficients in this expression to one. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setOnes_int_int.cpp + * Output: \verbinclude Matrix_setOnes_int_int.out + * + * \sa MatrixBase::setOnes(), setOnes(Index), class CwiseNullaryOp, MatrixBase::Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setOnes(Index rows, Index cols) +{ + resize(rows, cols); + return setConstant(Scalar(1)); +} + +// Identity: + +/** \returns an expression of the identity matrix (not necessarily square). + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Identity() should be used + * instead. + * + * Example: \include MatrixBase_identity_int_int.cpp + * Output: \verbinclude MatrixBase_identity_int_int.out + * + * \sa Identity(), setIdentity(), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::IdentityReturnType +MatrixBase::Identity(Index rows, Index cols) +{ + return DenseBase::NullaryExpr(rows, cols, internal::scalar_identity_op()); +} + +/** \returns an expression of the identity matrix (not necessarily square). + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variant taking size arguments. + * + * Example: \include MatrixBase_identity.cpp + * Output: \verbinclude MatrixBase_identity.out + * + * \sa Identity(Index,Index), setIdentity(), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::IdentityReturnType +MatrixBase::Identity() +{ + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return MatrixBase::NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_identity_op()); +} + +/** \returns true if *this is approximately equal to the identity matrix + * (not necessarily square), + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isIdentity.cpp + * Output: \verbinclude MatrixBase_isIdentity.out + * + * \sa class CwiseNullaryOp, Identity(), Identity(Index,Index), setIdentity() + */ +template +bool MatrixBase::isIdentity +(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + { + for(Index i = 0; i < rows(); ++i) + { + if(i == j) + { + if(!internal::isApprox(self.coeff(i, j), static_cast(1), prec)) + return false; + } + else + { + if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast(1), prec)) + return false; + } + } + } + return true; +} + +namespace internal { + +template=16)> +struct setIdentity_impl +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Derived& run(Derived& m) + { + return m = Derived::Identity(m.rows(), m.cols()); + } +}; + +template +struct setIdentity_impl +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Derived& run(Derived& m) + { + m.setZero(); + const Index size = numext::mini(m.rows(), m.cols()); + for(Index i = 0; i < size; ++i) m.coeffRef(i,i) = typename Derived::Scalar(1); + return m; + } +}; + +} // end namespace internal + +/** Writes the identity expression (not necessarily square) into *this. + * + * Example: \include MatrixBase_setIdentity.cpp + * Output: \verbinclude MatrixBase_setIdentity.out + * + * \sa class CwiseNullaryOp, Identity(), Identity(Index,Index), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase::setIdentity() +{ + return internal::setIdentity_impl::run(derived()); +} + +/** \brief Resizes to the given size, and writes the identity expression (not necessarily square) into *this. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setIdentity_int_int.cpp + * Output: \verbinclude Matrix_setIdentity_int_int.out + * + * \sa MatrixBase::setIdentity(), class CwiseNullaryOp, MatrixBase::Identity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase::setIdentity(Index rows, Index cols) +{ + derived().resize(rows, cols); + return setIdentity(); +} + +/** \returns an expression of the i-th unit (basis) vector. + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::Unit(Index newSize, Index i) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return BasisReturnType(SquareMatrixType::Identity(newSize,newSize), i); +} + +/** \returns an expression of the i-th unit (basis) vector. + * + * \only_for_vectors + * + * This variant is for fixed-size vector only. + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::Unit(Index i) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return BasisReturnType(SquareMatrixType::Identity(),i); +} + +/** \returns an expression of the X axis unit vector (1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitX() +{ return Derived::Unit(0); } + +/** \returns an expression of the Y axis unit vector (0,1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitY() +{ return Derived::Unit(1); } + +/** \returns an expression of the Z axis unit vector (0,0,1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitZ() +{ return Derived::Unit(2); } + +/** \returns an expression of the W axis unit vector (0,0,0,1) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitW() +{ return Derived::Unit(3); } + +} // end namespace Eigen + +#endif // EIGEN_CWISE_NULLARY_OP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h new file mode 100644 index 00000000000..9f3576fecef --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h @@ -0,0 +1,197 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2016 Eugene Brevdo +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_TERNARY_OP_H +#define EIGEN_CWISE_TERNARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > { + // we must not inherit from traits since it has + // the potential to cause problems with MSVC + typedef typename remove_all::type Ancestor; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime + }; + + // even though we require Arg1, Arg2, and Arg3 to have the same scalar type + // (see CwiseTernaryOp constructor), + // we still want to handle the case when the result type is different. + typedef typename result_of::type Scalar; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + + typedef typename Arg1::Nested Arg1Nested; + typedef typename Arg2::Nested Arg2Nested; + typedef typename Arg3::Nested Arg3Nested; + typedef typename remove_reference::type _Arg1Nested; + typedef typename remove_reference::type _Arg2Nested; + typedef typename remove_reference::type _Arg3Nested; + enum { Flags = _Arg1Nested::Flags & RowMajorBit }; +}; +} // end namespace internal + +template +class CwiseTernaryOpImpl; + +/** \class CwiseTernaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise ternary operator is + * applied to two expressions + * + * \tparam TernaryOp template functor implementing the operator + * \tparam Arg1Type the type of the first argument + * \tparam Arg2Type the type of the second argument + * \tparam Arg3Type the type of the third argument + * + * This class represents an expression where a coefficient-wise ternary + * operator is applied to three expressions. + * It is the return type of ternary operators, by which we mean only those + * ternary operators where + * all three arguments are Eigen expressions. + * For example, the return type of betainc(matrix1, matrix2, matrix3) is a + * CwiseTernaryOp. + * + * Most of the time, this is the only way that it is used, so you typically + * don't have to name + * CwiseTernaryOp types explicitly. + * + * \sa MatrixBase::ternaryExpr(const MatrixBase &, const + * MatrixBase &, const CustomTernaryOp &) const, class CwiseBinaryOp, + * class CwiseUnaryOp, class CwiseNullaryOp + */ +template +class CwiseTernaryOp : public CwiseTernaryOpImpl< + TernaryOp, Arg1Type, Arg2Type, Arg3Type, + typename internal::traits::StorageKind>, + internal::no_assignment_operator +{ + public: + typedef typename internal::remove_all::type Arg1; + typedef typename internal::remove_all::type Arg2; + typedef typename internal::remove_all::type Arg3; + + typedef typename CwiseTernaryOpImpl< + TernaryOp, Arg1Type, Arg2Type, Arg3Type, + typename internal::traits::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseTernaryOp) + + typedef typename internal::ref_selector::type Arg1Nested; + typedef typename internal::ref_selector::type Arg2Nested; + typedef typename internal::ref_selector::type Arg3Nested; + typedef typename internal::remove_reference::type _Arg1Nested; + typedef typename internal::remove_reference::type _Arg2Nested; + typedef typename internal::remove_reference::type _Arg3Nested; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CwiseTernaryOp(const Arg1& a1, const Arg2& a2, + const Arg3& a3, + const TernaryOp& func = TernaryOp()) + : m_arg1(a1), m_arg2(a2), m_arg3(a3), m_functor(func) { + // require the sizes to match + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg2) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg3) + + // The index types should match + EIGEN_STATIC_ASSERT((internal::is_same< + typename internal::traits::StorageKind, + typename internal::traits::StorageKind>::value), + STORAGE_KIND_MUST_MATCH) + EIGEN_STATIC_ASSERT((internal::is_same< + typename internal::traits::StorageKind, + typename internal::traits::StorageKind>::value), + STORAGE_KIND_MUST_MATCH) + + eigen_assert(a1.rows() == a2.rows() && a1.cols() == a2.cols() && + a1.rows() == a3.rows() && a1.cols() == a3.cols()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { + // return the fixed size type if available to enable compile time + // optimizations + if (internal::traits::type>:: + RowsAtCompileTime == Dynamic && + internal::traits::type>:: + RowsAtCompileTime == Dynamic) + return m_arg3.rows(); + else if (internal::traits::type>:: + RowsAtCompileTime == Dynamic && + internal::traits::type>:: + RowsAtCompileTime == Dynamic) + return m_arg2.rows(); + else + return m_arg1.rows(); + } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { + // return the fixed size type if available to enable compile time + // optimizations + if (internal::traits::type>:: + ColsAtCompileTime == Dynamic && + internal::traits::type>:: + ColsAtCompileTime == Dynamic) + return m_arg3.cols(); + else if (internal::traits::type>:: + ColsAtCompileTime == Dynamic && + internal::traits::type>:: + ColsAtCompileTime == Dynamic) + return m_arg2.cols(); + else + return m_arg1.cols(); + } + + /** \returns the first argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg1Nested& arg1() const { return m_arg1; } + /** \returns the first argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg2Nested& arg2() const { return m_arg2; } + /** \returns the third argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg3Nested& arg3() const { return m_arg3; } + /** \returns the functor representing the ternary operation */ + EIGEN_DEVICE_FUNC + const TernaryOp& functor() const { return m_functor; } + + protected: + Arg1Nested m_arg1; + Arg2Nested m_arg2; + Arg3Nested m_arg3; + const TernaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseTernaryOpImpl + : public internal::generic_xpr_base< + CwiseTernaryOp >::type { + public: + typedef typename internal::generic_xpr_base< + CwiseTernaryOp >::type Base; +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_TERNARY_OP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h new file mode 100644 index 00000000000..1d2dd19f2b1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h @@ -0,0 +1,103 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_UNARY_OP_H +#define EIGEN_CWISE_UNARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename result_of< + UnaryOp(const typename XprType::Scalar&) + >::type Scalar; + typedef typename XprType::Nested XprTypeNested; + typedef typename remove_reference::type _XprTypeNested; + enum { + Flags = _XprTypeNested::Flags & RowMajorBit + }; +}; +} + +template +class CwiseUnaryOpImpl; + +/** \class CwiseUnaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise unary operator is applied to an expression + * + * \tparam UnaryOp template functor implementing the operator + * \tparam XprType the type of the expression to which we are applying the unary operator + * + * This class represents an expression where a unary operator is applied to an expression. + * It is the return type of all operations taking exactly 1 input expression, regardless of the + * presence of other inputs such as scalars. For example, the operator* in the expression 3*matrix + * is considered unary, because only the right-hand side is an expression, and its + * return type is a specialization of CwiseUnaryOp. + * + * Most of the time, this is the only way that it is used, so you typically don't have to name + * CwiseUnaryOp types explicitly. + * + * \sa MatrixBase::unaryExpr(const CustomUnaryOp &) const, class CwiseBinaryOp, class CwiseNullaryOp + */ +template +class CwiseUnaryOp : public CwiseUnaryOpImpl::StorageKind>, internal::no_assignment_operator +{ + public: + + typedef typename CwiseUnaryOpImpl::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryOp) + typedef typename internal::ref_selector::type XprTypeNested; + typedef typename internal::remove_all::type NestedExpression; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit CwiseUnaryOp(const XprType& xpr, const UnaryOp& func = UnaryOp()) + : m_xpr(xpr), m_functor(func) {} + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Index cols() const { return m_xpr.cols(); } + + /** \returns the functor representing the unary operation */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const UnaryOp& functor() const { return m_functor; } + + /** \returns the nested expression */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const typename internal::remove_all::type& + nestedExpression() const { return m_xpr; } + + /** \returns the nested expression */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + typename internal::remove_all::type& + nestedExpression() { return m_xpr; } + + protected: + XprTypeNested m_xpr; + const UnaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseUnaryOpImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_UNARY_OP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h new file mode 100644 index 00000000000..5a30fa8df18 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h @@ -0,0 +1,130 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_UNARY_VIEW_H +#define EIGEN_CWISE_UNARY_VIEW_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename result_of< + ViewOp(const typename traits::Scalar&) + >::type Scalar; + typedef typename MatrixType::Nested MatrixTypeNested; + typedef typename remove_all::type _MatrixTypeNested; + enum { + FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, + Flags = traits<_MatrixTypeNested>::Flags & (RowMajorBit | FlagsLvalueBit | DirectAccessBit), // FIXME DirectAccessBit should not be handled by expressions + MatrixTypeInnerStride = inner_stride_at_compile_time::ret, + // need to cast the sizeof's from size_t to int explicitly, otherwise: + // "error: no integral type can represent all of the enumerator values + InnerStrideAtCompileTime = MatrixTypeInnerStride == Dynamic + ? int(Dynamic) + : int(MatrixTypeInnerStride) * int(sizeof(typename traits::Scalar) / sizeof(Scalar)), + OuterStrideAtCompileTime = outer_stride_at_compile_time::ret == Dynamic + ? int(Dynamic) + : outer_stride_at_compile_time::ret * int(sizeof(typename traits::Scalar) / sizeof(Scalar)) + }; +}; +} + +template +class CwiseUnaryViewImpl; + +/** \class CwiseUnaryView + * \ingroup Core_Module + * + * \brief Generic lvalue expression of a coefficient-wise unary operator of a matrix or a vector + * + * \tparam ViewOp template functor implementing the view + * \tparam MatrixType the type of the matrix we are applying the unary operator + * + * This class represents a lvalue expression of a generic unary view operator of a matrix or a vector. + * It is the return type of real() and imag(), and most of the time this is the only way it is used. + * + * \sa MatrixBase::unaryViewExpr(const CustomUnaryOp &) const, class CwiseUnaryOp + */ +template +class CwiseUnaryView : public CwiseUnaryViewImpl::StorageKind> +{ + public: + + typedef typename CwiseUnaryViewImpl::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryView) + typedef typename internal::ref_selector::non_const_type MatrixTypeNested; + typedef typename internal::remove_all::type NestedExpression; + + explicit inline CwiseUnaryView(MatrixType& mat, const ViewOp& func = ViewOp()) + : m_matrix(mat), m_functor(func) {} + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryView) + + EIGEN_STRONG_INLINE Index rows() const { return m_matrix.rows(); } + EIGEN_STRONG_INLINE Index cols() const { return m_matrix.cols(); } + + /** \returns the functor representing unary operation */ + const ViewOp& functor() const { return m_functor; } + + /** \returns the nested expression */ + const typename internal::remove_all::type& + nestedExpression() const { return m_matrix; } + + /** \returns the nested expression */ + typename internal::remove_reference::type& + nestedExpression() { return m_matrix.const_cast_derived(); } + + protected: + MatrixTypeNested m_matrix; + ViewOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseUnaryViewImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +template +class CwiseUnaryViewImpl + : public internal::dense_xpr_base< CwiseUnaryView >::type +{ + public: + + typedef CwiseUnaryView Derived; + typedef typename internal::dense_xpr_base< CwiseUnaryView >::type Base; + + EIGEN_DENSE_PUBLIC_INTERFACE(Derived) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryViewImpl) + + EIGEN_DEVICE_FUNC inline Scalar* data() { return &(this->coeffRef(0)); } + EIGEN_DEVICE_FUNC inline const Scalar* data() const { return &(this->coeff(0)); } + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return derived().nestedExpression().innerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); + } + + EIGEN_DEVICE_FUNC inline Index outerStride() const + { + return derived().nestedExpression().outerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); + } + protected: + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(CwiseUnaryViewImpl) +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_UNARY_VIEW_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h new file mode 100644 index 00000000000..c55a68230cd --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h @@ -0,0 +1,612 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSEBASE_H +#define EIGEN_DENSEBASE_H + +namespace Eigen { + +namespace internal { + +// The index type defined by EIGEN_DEFAULT_DENSE_INDEX_TYPE must be a signed type. +// This dummy function simply aims at checking that at compile time. +static inline void check_DenseIndex_is_signed() { + EIGEN_STATIC_ASSERT(NumTraits::IsSigned,THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE); +} + +} // end namespace internal + +/** \class DenseBase + * \ingroup Core_Module + * + * \brief Base class for all dense matrices, vectors, and arrays + * + * This class is the base that is inherited by all dense objects (matrix, vector, arrays, + * and related expression types). The common Eigen API for dense objects is contained in this class. + * + * \tparam Derived is the derived type, e.g., a matrix type or an expression. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_DENSEBASE_PLUGIN. + * + * \sa \blank \ref TopicClassHierarchy + */ +template class DenseBase +#ifndef EIGEN_PARSED_BY_DOXYGEN + : public DenseCoeffsBase::value> +#else + : public DenseCoeffsBase +#endif // not EIGEN_PARSED_BY_DOXYGEN +{ + public: + + /** Inner iterator type to iterate over the coefficients of a row or column. + * \sa class InnerIterator + */ + typedef Eigen::InnerIterator InnerIterator; + + typedef typename internal::traits::StorageKind StorageKind; + + /** + * \brief The type used to store indices + * \details This typedef is relevant for types that store multiple indices such as + * PermutationMatrix or Transpositions, otherwise it defaults to Eigen::Index + * \sa \blank \ref TopicPreprocessorDirectives, Eigen::Index, SparseMatrixBase. + */ + typedef typename internal::traits::StorageIndex StorageIndex; + + /** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex, etc. */ + typedef typename internal::traits::Scalar Scalar; + + /** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex, etc. + * + * It is an alias for the Scalar type */ + typedef Scalar value_type; + + typedef typename NumTraits::Real RealScalar; + typedef DenseCoeffsBase::value> Base; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::rowIndexByOuterInner; + using Base::colIndexByOuterInner; + using Base::coeff; + using Base::coeffByOuterInner; + using Base::operator(); + using Base::operator[]; + using Base::x; + using Base::y; + using Base::z; + using Base::w; + using Base::stride; + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + typedef typename Base::CoeffReturnType CoeffReturnType; + + enum { + + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + /**< The number of rows at compile-time. This is just a copy of the value provided + * by the \a Derived type. If a value is not known at compile-time, + * it is set to the \a Dynamic constant. + * \sa MatrixBase::rows(), MatrixBase::cols(), ColsAtCompileTime, SizeAtCompileTime */ + + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + /**< The number of columns at compile-time. This is just a copy of the value provided + * by the \a Derived type. If a value is not known at compile-time, + * it is set to the \a Dynamic constant. + * \sa MatrixBase::rows(), MatrixBase::cols(), RowsAtCompileTime, SizeAtCompileTime */ + + + SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, + internal::traits::ColsAtCompileTime>::ret), + /**< This is equal to the number of coefficients, i.e. the number of + * rows times the number of columns, or to \a Dynamic if this is not + * known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */ + + MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, + /**< This value is equal to the maximum possible number of rows that this expression + * might have. If this expression might have an arbitrarily high number of rows, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa RowsAtCompileTime, MaxColsAtCompileTime, MaxSizeAtCompileTime + */ + + MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, + /**< This value is equal to the maximum possible number of columns that this expression + * might have. If this expression might have an arbitrarily high number of columns, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa ColsAtCompileTime, MaxRowsAtCompileTime, MaxSizeAtCompileTime + */ + + MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime>::ret), + /**< This value is equal to the maximum possible number of coefficients that this expression + * might have. If this expression might have an arbitrarily high number of coefficients, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa SizeAtCompileTime, MaxRowsAtCompileTime, MaxColsAtCompileTime + */ + + IsVectorAtCompileTime = internal::traits::MaxRowsAtCompileTime == 1 + || internal::traits::MaxColsAtCompileTime == 1, + /**< This is set to true if either the number of rows or the number of + * columns is known at compile-time to be equal to 1. Indeed, in that case, + * we are dealing with a column-vector (if there is only one column) or with + * a row-vector (if there is only one row). */ + + Flags = internal::traits::Flags, + /**< This stores expression \ref flags flags which may or may not be inherited by new expressions + * constructed from this one. See the \ref flags "list of flags". + */ + + IsRowMajor = int(Flags) & RowMajorBit, /**< True if this expression has row-major storage order. */ + + InnerSizeAtCompileTime = int(IsVectorAtCompileTime) ? int(SizeAtCompileTime) + : int(IsRowMajor) ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + + InnerStrideAtCompileTime = internal::inner_stride_at_compile_time::ret, + OuterStrideAtCompileTime = internal::outer_stride_at_compile_time::ret + }; + + typedef typename internal::find_best_packet::type PacketScalar; + + enum { IsPlainObjectBase = 0 }; + + /** The plain matrix type corresponding to this expression. + * \sa PlainObject */ + typedef Matrix::Scalar, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime, + AutoAlign | (internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor), + internal::traits::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime + > PlainMatrix; + + /** The plain array type corresponding to this expression. + * \sa PlainObject */ + typedef Array::Scalar, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime, + AutoAlign | (internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor), + internal::traits::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime + > PlainArray; + + /** \brief The plain matrix or array type corresponding to this expression. + * + * This is not necessarily exactly the return type of eval(). In the case of plain matrices, + * the return type of eval() is a const reference to a matrix, not a matrix! It is however guaranteed + * that the return type of eval() is either PlainObject or const PlainObject&. + */ + typedef typename internal::conditional::XprKind,MatrixXpr >::value, + PlainMatrix, PlainArray>::type PlainObject; + + /** \returns the number of nonzero coefficients which is in practice the number + * of stored coefficients. */ + EIGEN_DEVICE_FUNC + inline Index nonZeros() const { return size(); } + + /** \returns the outer size. + * + * \note For a vector, this returns just 1. For a matrix (non-vector), this is the major dimension + * with respect to the \ref TopicStorageOrders "storage order", i.e., the number of columns for a + * column-major matrix, and the number of rows for a row-major matrix. */ + EIGEN_DEVICE_FUNC + Index outerSize() const + { + return IsVectorAtCompileTime ? 1 + : int(IsRowMajor) ? this->rows() : this->cols(); + } + + /** \returns the inner size. + * + * \note For a vector, this is just the size. For a matrix (non-vector), this is the minor dimension + * with respect to the \ref TopicStorageOrders "storage order", i.e., the number of rows for a + * column-major matrix, and the number of columns for a row-major matrix. */ + EIGEN_DEVICE_FUNC + Index innerSize() const + { + return IsVectorAtCompileTime ? this->size() + : int(IsRowMajor) ? this->cols() : this->rows(); + } + + /** Only plain matrices/arrays, not expressions, may be resized; therefore the only useful resize methods are + * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does + * nothing else. + */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) + { + EIGEN_ONLY_USED_FOR_DEBUG(newSize); + eigen_assert(newSize == this->size() + && "DenseBase::resize() does not actually allow to resize."); + } + /** Only plain matrices/arrays, not expressions, may be resized; therefore the only useful resize methods are + * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does + * nothing else. + */ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) + { + EIGEN_ONLY_USED_FOR_DEBUG(rows); + EIGEN_ONLY_USED_FOR_DEBUG(cols); + eigen_assert(rows == this->rows() && cols == this->cols() + && "DenseBase::resize() does not actually allow to resize."); + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; + /** \internal \deprecated Represents a vector with linearly spaced coefficients that allows sequential access only. */ + typedef CwiseNullaryOp,PlainObject> SequentialLinSpacedReturnType; + /** \internal Represents a vector with linearly spaced coefficients that allows random access. */ + typedef CwiseNullaryOp,PlainObject> RandomAccessLinSpacedReturnType; + /** \internal the return type of MatrixBase::eigenvalues() */ + typedef Matrix::Scalar>::Real, internal::traits::ColsAtCompileTime, 1> EigenvaluesReturnType; + +#endif // not EIGEN_PARSED_BY_DOXYGEN + + /** Copies \a other into *this. \returns a reference to *this. */ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator+=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator-=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const ReturnByValue& func); + + /** \internal + * Copies \a other into *this without evaluating other. \returns a reference to *this. + * \deprecated */ + template + EIGEN_DEVICE_FUNC + Derived& lazyAssign(const DenseBase& other); + + EIGEN_DEVICE_FUNC + CommaInitializer operator<< (const Scalar& s); + + /** \deprecated it now returns \c *this */ + template + EIGEN_DEPRECATED + const Derived& flagged() const + { return derived(); } + + template + EIGEN_DEVICE_FUNC + CommaInitializer operator<< (const DenseBase& other); + + typedef Transpose TransposeReturnType; + EIGEN_DEVICE_FUNC + TransposeReturnType transpose(); + typedef typename internal::add_const >::type ConstTransposeReturnType; + EIGEN_DEVICE_FUNC + ConstTransposeReturnType transpose() const; + EIGEN_DEVICE_FUNC + void transposeInPlace(); + + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(Index rows, Index cols, const Scalar& value); + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(Index size, const Scalar& value); + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(const Scalar& value); + + EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType + LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType + LinSpaced(Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType + LinSpaced(Sequential_t, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType + LinSpaced(const Scalar& low, const Scalar& high); + + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func); + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(Index size, const CustomNullaryOp& func); + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(const CustomNullaryOp& func); + + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index size); + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index size); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(); + + EIGEN_DEVICE_FUNC void fill(const Scalar& value); + EIGEN_DEVICE_FUNC Derived& setConstant(const Scalar& value); + EIGEN_DEVICE_FUNC Derived& setLinSpaced(Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC Derived& setLinSpaced(const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC Derived& setZero(); + EIGEN_DEVICE_FUNC Derived& setOnes(); + EIGEN_DEVICE_FUNC Derived& setRandom(); + + template EIGEN_DEVICE_FUNC + bool isApprox(const DenseBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC + bool isMuchSmallerThan(const RealScalar& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + template EIGEN_DEVICE_FUNC + bool isMuchSmallerThan(const DenseBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + + EIGEN_DEVICE_FUNC bool isApproxToConstant(const Scalar& value, const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isConstant(const Scalar& value, const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isZero(const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isOnes(const RealScalar& prec = NumTraits::dummy_precision()) const; + + inline bool hasNaN() const; + inline bool allFinite() const; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator*=(const Scalar& other); + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator/=(const Scalar& other); + + typedef typename internal::add_const_on_value_type::type>::type EvalReturnType; + /** \returns the matrix or vector obtained by evaluating this expression. + * + * Notice that in the case of a plain matrix or vector (not an expression) this function just returns + * a const reference, in order to avoid a useless copy. + * + * \warning Be carefull with eval() and the auto C++ keyword, as detailed in this \link TopicPitfalls_auto_keyword page \endlink. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE EvalReturnType eval() const + { + // Even though MSVC does not honor strong inlining when the return type + // is a dynamic matrix, we desperately need strong inlining for fixed + // size types on MSVC. + return typename internal::eval::type(derived()); + } + + /** swaps *this with the expression \a other. + * + */ + template + EIGEN_DEVICE_FUNC + void swap(const DenseBase& other) + { + EIGEN_STATIC_ASSERT(!OtherDerived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + eigen_assert(rows()==other.rows() && cols()==other.cols()); + call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); + } + + /** swaps *this with the matrix or array \a other. + * + */ + template + EIGEN_DEVICE_FUNC + void swap(PlainObjectBase& other) + { + eigen_assert(rows()==other.rows() && cols()==other.cols()); + call_assignment(derived(), other.derived(), internal::swap_assign_op()); + } + + EIGEN_DEVICE_FUNC inline const NestByValue nestByValue() const; + EIGEN_DEVICE_FUNC inline const ForceAlignedAccess forceAlignedAccess() const; + EIGEN_DEVICE_FUNC inline ForceAlignedAccess forceAlignedAccess(); + template EIGEN_DEVICE_FUNC + inline const typename internal::conditional,Derived&>::type forceAlignedAccessIf() const; + template EIGEN_DEVICE_FUNC + inline typename internal::conditional,Derived&>::type forceAlignedAccessIf(); + + EIGEN_DEVICE_FUNC Scalar sum() const; + EIGEN_DEVICE_FUNC Scalar mean() const; + EIGEN_DEVICE_FUNC Scalar trace() const; + + EIGEN_DEVICE_FUNC Scalar prod() const; + + EIGEN_DEVICE_FUNC typename internal::traits::Scalar minCoeff() const; + EIGEN_DEVICE_FUNC typename internal::traits::Scalar maxCoeff() const; + + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar minCoeff(IndexType* row, IndexType* col) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar maxCoeff(IndexType* row, IndexType* col) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar minCoeff(IndexType* index) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar maxCoeff(IndexType* index) const; + + template + EIGEN_DEVICE_FUNC + Scalar redux(const BinaryOp& func) const; + + template + EIGEN_DEVICE_FUNC + void visit(Visitor& func) const; + + /** \returns a WithFormat proxy object allowing to print a matrix the with given + * format \a fmt. + * + * See class IOFormat for some examples. + * + * \sa class IOFormat, class WithFormat + */ + inline const WithFormat format(const IOFormat& fmt) const + { + return WithFormat(derived(), fmt); + } + + /** \returns the unique coefficient of a 1x1 expression */ + EIGEN_DEVICE_FUNC + CoeffReturnType value() const + { + EIGEN_STATIC_ASSERT_SIZE_1x1(Derived) + eigen_assert(this->rows() == 1 && this->cols() == 1); + return derived().coeff(0,0); + } + + EIGEN_DEVICE_FUNC bool all() const; + EIGEN_DEVICE_FUNC bool any() const; + EIGEN_DEVICE_FUNC Index count() const; + + typedef VectorwiseOp RowwiseReturnType; + typedef const VectorwiseOp ConstRowwiseReturnType; + typedef VectorwiseOp ColwiseReturnType; + typedef const VectorwiseOp ConstColwiseReturnType; + + /** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations + * + * Example: \include MatrixBase_rowwise.cpp + * Output: \verbinclude MatrixBase_rowwise.out + * + * \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting + */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC inline ConstRowwiseReturnType rowwise() const { + return ConstRowwiseReturnType(derived()); + } + EIGEN_DEVICE_FUNC RowwiseReturnType rowwise(); + + /** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations + * + * Example: \include MatrixBase_colwise.cpp + * Output: \verbinclude MatrixBase_colwise.out + * + * \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting + */ + EIGEN_DEVICE_FUNC inline ConstColwiseReturnType colwise() const { + return ConstColwiseReturnType(derived()); + } + EIGEN_DEVICE_FUNC ColwiseReturnType colwise(); + + typedef CwiseNullaryOp,PlainObject> RandomReturnType; + static const RandomReturnType Random(Index rows, Index cols); + static const RandomReturnType Random(Index size); + static const RandomReturnType Random(); + + template + const Select + select(const DenseBase& thenMatrix, + const DenseBase& elseMatrix) const; + + template + inline const Select + select(const DenseBase& thenMatrix, const typename ThenDerived::Scalar& elseScalar) const; + + template + inline const Select + select(const typename ElseDerived::Scalar& thenScalar, const DenseBase& elseMatrix) const; + + template RealScalar lpNorm() const; + + template + EIGEN_DEVICE_FUNC + const Replicate replicate() const; + /** + * \return an expression of the replication of \c *this + * + * Example: \include MatrixBase_replicate_int_int.cpp + * Output: \verbinclude MatrixBase_replicate_int_int.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(), class Replicate + */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC + const Replicate replicate(Index rowFactor, Index colFactor) const + { + return Replicate(derived(), rowFactor, colFactor); + } + + typedef Reverse ReverseReturnType; + typedef const Reverse ConstReverseReturnType; + EIGEN_DEVICE_FUNC ReverseReturnType reverse(); + /** This is the const version of reverse(). */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC ConstReverseReturnType reverse() const + { + return ConstReverseReturnType(derived()); + } + EIGEN_DEVICE_FUNC void reverseInPlace(); + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::DenseBase +#define EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL +#define EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(COND) +# include "../plugins/BlockMethods.h" +# ifdef EIGEN_DENSEBASE_PLUGIN +# include EIGEN_DENSEBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL +#undef EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF + + // disable the use of evalTo for dense objects with a nice compilation error + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& ) const + { + EIGEN_STATIC_ASSERT((internal::is_same::value),THE_EVAL_EVALTO_FUNCTION_SHOULD_NEVER_BE_CALLED_FOR_DENSE_OBJECTS); + } + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(DenseBase) + /** Default constructor. Do nothing. */ + EIGEN_DEVICE_FUNC DenseBase() + { + /* Just checks for self-consistency of the flags. + * Only do it when debugging Eigen, as this borders on paranoia and could slow compilation down + */ +#ifdef EIGEN_INTERNAL_DEBUGGING + EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, int(IsRowMajor)) + && EIGEN_IMPLIES(MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1, int(!IsRowMajor))), + INVALID_STORAGE_ORDER_FOR_THIS_VECTOR_EXPRESSION) +#endif + } + + private: + EIGEN_DEVICE_FUNC explicit DenseBase(int); + EIGEN_DEVICE_FUNC DenseBase(int,int); + template EIGEN_DEVICE_FUNC explicit DenseBase(const DenseBase&); +}; + +} // end namespace Eigen + +#endif // EIGEN_DENSEBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h new file mode 100644 index 00000000000..c4af48ab699 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h @@ -0,0 +1,681 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSECOEFFSBASE_H +#define EIGEN_DENSECOEFFSBASE_H + +namespace Eigen { + +namespace internal { +template struct add_const_on_value_type_if_arithmetic +{ + typedef typename conditional::value, T, typename add_const_on_value_type::type>::type type; +}; +} + +/** \brief Base class providing read-only coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #ReadOnlyAccessors Constant indicating read-only access + * + * This class defines the \c operator() \c const function and friends, which can be used to read specific + * entries of a matrix or array. + * + * \sa DenseCoeffsBase, DenseCoeffsBase, + * \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public EigenBase +{ + public: + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + + // Explanation for this CoeffReturnType typedef. + // - This is the return type of the coeff() method. + // - The LvalueBit means exactly that we can offer a coeffRef() method, which means exactly that we can get references + // to coeffs, which means exactly that we can have coeff() return a const reference (as opposed to returning a value). + // - The is_artihmetic check is required since "const int", "const double", etc. will cause warnings on some systems + // while the declaration of "const T", where T is a non arithmetic type does not. Always returning "const Scalar&" is + // not possible, since the underlying expressions might not offer a valid address the reference could be referring to. + typedef typename internal::conditional::Flags&LvalueBit), + const Scalar&, + typename internal::conditional::value, Scalar, const Scalar>::type + >::type CoeffReturnType; + + typedef typename internal::add_const_on_value_type_if_arithmetic< + typename internal::packet_traits::type + >::type PacketReturnType; + + typedef EigenBase Base; + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner) const + { + return int(Derived::RowsAtCompileTime) == 1 ? 0 + : int(Derived::ColsAtCompileTime) == 1 ? inner + : int(Derived::Flags)&RowMajorBit ? outer + : inner; + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner) const + { + return int(Derived::ColsAtCompileTime) == 1 ? 0 + : int(Derived::RowsAtCompileTime) == 1 ? inner + : int(Derived::Flags)&RowMajorBit ? inner + : outer; + } + + /** Short version: don't use this function, use + * \link operator()(Index,Index) const \endlink instead. + * + * Long version: this function is similar to + * \link operator()(Index,Index) const \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator()(Index,Index) const \endlink. + * + * \sa operator()(Index,Index) const, coeffRef(Index,Index), coeff(Index) const + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType coeff(Index row, Index col) const + { + eigen_internal_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return internal::evaluator(derived()).coeff(row,col); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType coeffByOuterInner(Index outer, Index inner) const + { + return coeff(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \returns the coefficient at given the given row and column. + * + * \sa operator()(Index,Index), operator[](Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType operator()(Index row, Index col) const + { + eigen_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return coeff(row, col); + } + + /** Short version: don't use this function, use + * \link operator[](Index) const \endlink instead. + * + * Long version: this function is similar to + * \link operator[](Index) const \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameter \a index is in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator[](Index) const \endlink. + * + * \sa operator[](Index) const, coeffRef(Index), coeff(Index,Index) const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + coeff(Index index) const + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).coeff(index); + } + + + /** \returns the coefficient at given index. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index), operator()(Index,Index) const, x() const, y() const, + * z() const, w() const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + operator[](Index index) const + { + EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime, + THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD) + eigen_assert(index >= 0 && index < size()); + return coeff(index); + } + + /** \returns the coefficient at given index. + * + * This is synonymous to operator[](Index) const. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index), operator()(Index,Index) const, x() const, y() const, + * z() const, w() const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + operator()(Index index) const + { + eigen_assert(index >= 0 && index < size()); + return coeff(index); + } + + /** equivalent to operator[](0). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + x() const { return (*this)[0]; } + + /** equivalent to operator[](1). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + y() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS); + return (*this)[1]; + } + + /** equivalent to operator[](2). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + z() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS); + return (*this)[2]; + } + + /** equivalent to operator[](3). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + w() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS); + return (*this)[3]; + } + + /** \internal + * \returns the packet of coefficients starting at the given row and column. It is your responsibility + * to ensure that a packet really starts there. This method is only available on expressions having the + * PacketAccessBit. + * + * The \a LoadMode parameter may have the value \a #Aligned or \a #Unaligned. Its effect is to select + * the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets + * starting at an address which is a multiple of the packet size. + */ + + template + EIGEN_STRONG_INLINE PacketReturnType packet(Index row, Index col) const + { + typedef typename internal::packet_traits::type DefaultPacketType; + eigen_internal_assert(row >= 0 && row < rows() && col >= 0 && col < cols()); + return internal::evaluator(derived()).template packet(row,col); + } + + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketReturnType packetByOuterInner(Index outer, Index inner) const + { + return packet(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \internal + * \returns the packet of coefficients starting at the given index. It is your responsibility + * to ensure that a packet really starts there. This method is only available on expressions having the + * PacketAccessBit and the LinearAccessBit. + * + * The \a LoadMode parameter may have the value \a #Aligned or \a #Unaligned. Its effect is to select + * the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets + * starting at an address which is a multiple of the packet size. + */ + + template + EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + typedef typename internal::packet_traits::type DefaultPacketType; + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).template packet(index); + } + + protected: + // explanation: DenseBase is doing "using ..." on the methods from DenseCoeffsBase. + // But some methods are only available in the DirectAccess case. + // So we add dummy methods here with these names, so that "using... " doesn't fail. + // It's not private so that the child class DenseBase can access them, and it's not public + // either since it's an implementation detail, so has to be protected. + void coeffRef(); + void coeffRefByOuterInner(); + void writePacket(); + void writePacketByOuterInner(); + void copyCoeff(); + void copyCoeffByOuterInner(); + void copyPacket(); + void copyPacketByOuterInner(); + void stride(); + void innerStride(); + void outerStride(); + void rowStride(); + void colStride(); +}; + +/** \brief Base class providing read/write coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #WriteAccessors Constant indicating read/write access + * + * This class defines the non-const \c operator() function and friends, which can be used to write specific + * entries of a matrix or array. This class inherits DenseCoeffsBase which + * defines the const variant for reading specific entries. + * + * \sa DenseCoeffsBase, \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + using Base::coeff; + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + using Base::rowIndexByOuterInner; + using Base::colIndexByOuterInner; + using Base::operator[]; + using Base::operator(); + using Base::x; + using Base::y; + using Base::z; + using Base::w; + + /** Short version: don't use this function, use + * \link operator()(Index,Index) \endlink instead. + * + * Long version: this function is similar to + * \link operator()(Index,Index) \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator()(Index,Index) \endlink. + * + * \sa operator()(Index,Index), coeff(Index, Index) const, coeffRef(Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index row, Index col) + { + eigen_internal_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return internal::evaluator(derived()).coeffRef(row,col); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + coeffRefByOuterInner(Index outer, Index inner) + { + return coeffRef(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \returns a reference to the coefficient at given the given row and column. + * + * \sa operator[](Index) + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator()(Index row, Index col) + { + eigen_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return coeffRef(row, col); + } + + + /** Short version: don't use this function, use + * \link operator[](Index) \endlink instead. + * + * Long version: this function is similar to + * \link operator[](Index) \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator[](Index) \endlink. + * + * \sa operator[](Index), coeff(Index) const, coeffRef(Index,Index) + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + coeffRef(Index index) + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).coeffRef(index); + } + + /** \returns a reference to the coefficient at given index. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w() + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator[](Index index) + { + EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime, + THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD) + eigen_assert(index >= 0 && index < size()); + return coeffRef(index); + } + + /** \returns a reference to the coefficient at given index. + * + * This is synonymous to operator[](Index). + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w() + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator()(Index index) + { + eigen_assert(index >= 0 && index < size()); + return coeffRef(index); + } + + /** equivalent to operator[](0). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + x() { return (*this)[0]; } + + /** equivalent to operator[](1). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + y() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS); + return (*this)[1]; + } + + /** equivalent to operator[](2). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + z() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS); + return (*this)[2]; + } + + /** equivalent to operator[](3). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + w() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS); + return (*this)[3]; + } +}; + +/** \brief Base class providing direct read-only coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #DirectAccessors Constant indicating direct access + * + * This class defines functions to work with strides which can be used to access entries directly. This class + * inherits DenseCoeffsBase which defines functions to access entries read-only using + * \c operator() . + * + * \sa \blank \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + typedef typename internal::traits::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + /** \returns the pointer increment between two consecutive elements within a slice in the inner direction. + * + * \sa outerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return derived().innerStride(); + } + + /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns + * in a column-major matrix). + * + * \sa innerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return derived().outerStride(); + } + + // FIXME shall we remove it ? + inline Index stride() const + { + return Derived::IsVectorAtCompileTime ? innerStride() : outerStride(); + } + + /** \returns the pointer increment between two consecutive rows. + * + * \sa innerStride(), outerStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index rowStride() const + { + return Derived::IsRowMajor ? outerStride() : innerStride(); + } + + /** \returns the pointer increment between two consecutive columns. + * + * \sa innerStride(), outerStride(), rowStride() + */ + EIGEN_DEVICE_FUNC + inline Index colStride() const + { + return Derived::IsRowMajor ? innerStride() : outerStride(); + } +}; + +/** \brief Base class providing direct read/write coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #DirectWriteAccessors Constant indicating direct access + * + * This class defines functions to work with strides which can be used to access entries directly. This class + * inherits DenseCoeffsBase which defines functions to access entries read/write using + * \c operator(). + * + * \sa \blank \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase + : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + typedef typename internal::traits::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + /** \returns the pointer increment between two consecutive elements within a slice in the inner direction. + * + * \sa outerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return derived().innerStride(); + } + + /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns + * in a column-major matrix). + * + * \sa innerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return derived().outerStride(); + } + + // FIXME shall we remove it ? + inline Index stride() const + { + return Derived::IsVectorAtCompileTime ? innerStride() : outerStride(); + } + + /** \returns the pointer increment between two consecutive rows. + * + * \sa innerStride(), outerStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index rowStride() const + { + return Derived::IsRowMajor ? outerStride() : innerStride(); + } + + /** \returns the pointer increment between two consecutive columns. + * + * \sa innerStride(), outerStride(), rowStride() + */ + EIGEN_DEVICE_FUNC + inline Index colStride() const + { + return Derived::IsRowMajor ? innerStride() : outerStride(); + } +}; + +namespace internal { + +template +struct first_aligned_impl +{ + static inline Index run(const Derived&) + { return 0; } +}; + +template +struct first_aligned_impl +{ + static inline Index run(const Derived& m) + { + return internal::first_aligned(m.data(), m.size()); + } +}; + +/** \internal \returns the index of the first element of the array stored by \a m that is properly aligned with respect to \a Alignment for vectorization. + * + * \tparam Alignment requested alignment in Bytes. + * + * There is also the variant first_aligned(const Scalar*, Integer) defined in Memory.h. See it for more + * documentation. + */ +template +static inline Index first_aligned(const DenseBase& m) +{ + enum { ReturnZero = (int(evaluator::Alignment) >= Alignment) || !(Derived::Flags & DirectAccessBit) }; + return first_aligned_impl::run(m.derived()); +} + +template +static inline Index first_default_aligned(const DenseBase& m) +{ + typedef typename Derived::Scalar Scalar; + typedef typename packet_traits::type DefaultPacketType; + return internal::first_aligned::alignment),Derived>(m); +} + +template::ret> +struct inner_stride_at_compile_time +{ + enum { ret = traits::InnerStrideAtCompileTime }; +}; + +template +struct inner_stride_at_compile_time +{ + enum { ret = 0 }; +}; + +template::ret> +struct outer_stride_at_compile_time +{ + enum { ret = traits::OuterStrideAtCompileTime }; +}; + +template +struct outer_stride_at_compile_time +{ + enum { ret = 0 }; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DENSECOEFFSBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h new file mode 100644 index 00000000000..7d6d4e66d4b --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h @@ -0,0 +1,570 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2009 Benoit Jacob +// Copyright (C) 2010-2013 Hauke Heibel +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIXSTORAGE_H +#define EIGEN_MATRIXSTORAGE_H + +#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) X; EIGEN_DENSE_STORAGE_CTOR_PLUGIN; +#else + #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) +#endif + +namespace Eigen { + +namespace internal { + +struct constructor_without_unaligned_array_assert {}; + +template +EIGEN_DEVICE_FUNC +void check_static_allocation_size() +{ + // if EIGEN_STACK_ALLOCATION_LIMIT is defined to 0, then no limit + #if EIGEN_STACK_ALLOCATION_LIMIT + EIGEN_STATIC_ASSERT(Size * sizeof(T) <= EIGEN_STACK_ALLOCATION_LIMIT, OBJECT_ALLOCATED_ON_STACK_IS_TOO_BIG); + #endif +} + +/** \internal + * Static array. If the MatrixOrArrayOptions require auto-alignment, the array will be automatically aligned: + * to 16 bytes boundary if the total size is a multiple of 16 bytes. + */ +template ::value > +struct plain_array +{ + T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +#if defined(EIGEN_DISABLE_UNALIGNED_ARRAY_ASSERT) + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) +#elif EIGEN_GNUC_AT_LEAST(4,7) + // GCC 4.7 is too aggressive in its optimizations and remove the alignement test based on the fact the array is declared to be aligned. + // See this bug report: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53900 + // Hiding the origin of the array pointer behind a function argument seems to do the trick even if the function is inlined: + template + EIGEN_ALWAYS_INLINE PtrType eigen_unaligned_array_assert_workaround_gcc47(PtrType array) { return array; } + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \ + eigen_assert((internal::UIntPtr(eigen_unaligned_array_assert_workaround_gcc47(array)) & (sizemask)) == 0 \ + && "this assertion is explained here: " \ + "http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \ + " **** READ THIS WEB PAGE !!! ****"); +#else + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \ + eigen_assert((internal::UIntPtr(array) & (sizemask)) == 0 \ + && "this assertion is explained here: " \ + "http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \ + " **** READ THIS WEB PAGE !!! ****"); +#endif + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(8) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(7); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(16) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(15); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(32) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(31); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(64) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(63); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + T array[1]; + EIGEN_DEVICE_FUNC plain_array() {} + EIGEN_DEVICE_FUNC plain_array(constructor_without_unaligned_array_assert) {} +}; + +} // end namespace internal + +/** \internal + * + * \class DenseStorage + * \ingroup Core_Module + * + * \brief Stores the data of a matrix + * + * This class stores the data of fixed-size, dynamic-size or mixed matrices + * in a way as compact as possible. + * + * \sa Matrix + */ +template class DenseStorage; + +// purely fixed-size matrix +template class DenseStorage +{ + internal::plain_array m_data; + public: + EIGEN_DEVICE_FUNC DenseStorage() { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size) + } + EIGEN_DEVICE_FUNC + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()) {} + EIGEN_DEVICE_FUNC + DenseStorage(const DenseStorage& other) : m_data(other.m_data) { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size) + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) m_data = other.m_data; + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows==_Rows && cols==_Cols); + EIGEN_UNUSED_VARIABLE(size); + EIGEN_UNUSED_VARIABLE(rows); + EIGEN_UNUSED_VARIABLE(cols); + } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); } + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// null matrix +template class DenseStorage +{ + public: + EIGEN_DEVICE_FUNC DenseStorage() {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage&) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage&) { return *this; } + EIGEN_DEVICE_FUNC DenseStorage(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& ) {} + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC const T *data() const { return 0; } + EIGEN_DEVICE_FUNC T *data() { return 0; } +}; + +// more specializations for null matrices; these are necessary to resolve ambiguities +template class DenseStorage +: public DenseStorage { }; + +template class DenseStorage +: public DenseStorage { }; + +template class DenseStorage +: public DenseStorage { }; + +// dynamic-size matrix with fixed-size storage +template class DenseStorage +{ + internal::plain_array m_data; + Index m_rows; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows), m_cols(other.m_cols) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_rows = other.m_rows; + m_cols = other.m_cols; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index cols) : m_rows(rows), m_cols(cols) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) + { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows() const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols() const {return m_cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; } + EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// dynamic-size matrix with fixed-size storage and fixed width +template class DenseStorage +{ + internal::plain_array m_data; + Index m_rows; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_rows = other.m_rows; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index) : m_rows(rows) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index) { m_rows = rows; } + EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index) { m_rows = rows; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// dynamic-size matrix with fixed-size storage and fixed height +template class DenseStorage +{ + internal::plain_array m_data; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_cols(other.m_cols) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_cols = other.m_cols; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index, Index cols) : m_cols(cols) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows(void) const {return _Rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + void conservativeResize(Index, Index, Index cols) { m_cols = cols; } + void resize(Index, Index, Index cols) { m_cols = cols; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// purely dynamic matrix. +template class DenseStorage +{ + T *m_data; + Index m_rows; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(0), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) + : m_data(internal::conditional_aligned_new_auto(size)), m_rows(rows), m_cols(cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows>=0 && cols >=0); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(other.m_rows*other.m_cols)) + , m_rows(other.m_rows) + , m_cols(other.m_cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*m_cols) + internal::smart_copy(other.m_data, other.m_data+other.m_rows*other.m_cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_rows(std::move(other.m_rows)) + , m_cols(std::move(other.m_cols)) + { + other.m_data = nullptr; + other.m_rows = 0; + other.m_cols = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_rows, other.m_rows); + swap(m_cols, other.m_cols); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, m_rows*m_cols); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) + { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + void conservativeResize(Index size, Index rows, Index cols) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, m_rows*m_cols); + m_rows = rows; + m_cols = cols; + } + EIGEN_DEVICE_FUNC void resize(Index size, Index rows, Index cols) + { + if(size != m_rows*m_cols) + { + internal::conditional_aligned_delete_auto(m_data, m_rows*m_cols); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_rows = rows; + m_cols = cols; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +// matrix with dynamic width and fixed height (so that matrix has dynamic size). +template class DenseStorage +{ + T *m_data; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_cols(0) {} + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto(size)), m_cols(cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows==_Rows && cols >=0); + EIGEN_UNUSED_VARIABLE(rows); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(_Rows*other.m_cols)) + , m_cols(other.m_cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_cols*_Rows) + internal::smart_copy(other.m_data, other.m_data+_Rows*m_cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_cols(std::move(other.m_cols)) + { + other.m_data = nullptr; + other.m_cols = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_cols, other.m_cols); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, _Rows*m_cols); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index size, Index, Index cols) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, _Rows*m_cols); + m_cols = cols; + } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index, Index cols) + { + if(size != _Rows*m_cols) + { + internal::conditional_aligned_delete_auto(m_data, _Rows*m_cols); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_cols = cols; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +// matrix with dynamic height and fixed width (so that matrix has dynamic size). +template class DenseStorage +{ + T *m_data; + Index m_rows; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0) {} + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto(size)), m_rows(rows) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows>=0 && cols == _Cols); + EIGEN_UNUSED_VARIABLE(cols); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(other.m_rows*_Cols)) + , m_rows(other.m_rows) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*_Cols) + internal::smart_copy(other.m_data, other.m_data+other.m_rows*_Cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_rows(std::move(other.m_rows)) + { + other.m_data = nullptr; + other.m_rows = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_rows, other.m_rows); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, _Cols*m_rows); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + void conservativeResize(Index size, Index rows, Index) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, m_rows*_Cols); + m_rows = rows; + } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index rows, Index) + { + if(size != m_rows*_Cols) + { + internal::conditional_aligned_delete_auto(m_data, _Cols*m_rows); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_rows = rows; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +} // end namespace Eigen + +#endif // EIGEN_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h new file mode 100644 index 00000000000..afcaf357567 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h @@ -0,0 +1,260 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2009 Benoit Jacob +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONAL_H +#define EIGEN_DIAGONAL_H + +namespace Eigen { + +/** \class Diagonal + * \ingroup Core_Module + * + * \brief Expression of a diagonal/subdiagonal/superdiagonal in a matrix + * + * \param MatrixType the type of the object in which we are taking a sub/main/super diagonal + * \param DiagIndex the index of the sub/super diagonal. The default is 0 and it means the main diagonal. + * A positive value means a superdiagonal, a negative value means a subdiagonal. + * You can also use DynamicIndex so the index can be set at runtime. + * + * The matrix is not required to be square. + * + * This class represents an expression of the main diagonal, or any sub/super diagonal + * of a square matrix. It is the return type of MatrixBase::diagonal() and MatrixBase::diagonal(Index) and most of the + * time this is the only way it is used. + * + * \sa MatrixBase::diagonal(), MatrixBase::diagonal(Index) + */ + +namespace internal { +template +struct traits > + : traits +{ + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + typedef typename MatrixType::StorageKind StorageKind; + enum { + RowsAtCompileTime = (int(DiagIndex) == DynamicIndex || int(MatrixType::SizeAtCompileTime) == Dynamic) ? Dynamic + : (EIGEN_PLAIN_ENUM_MIN(MatrixType::RowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0), + MatrixType::ColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))), + ColsAtCompileTime = 1, + MaxRowsAtCompileTime = int(MatrixType::MaxSizeAtCompileTime) == Dynamic ? Dynamic + : DiagIndex == DynamicIndex ? EIGEN_SIZE_MIN_PREFER_FIXED(MatrixType::MaxRowsAtCompileTime, + MatrixType::MaxColsAtCompileTime) + : (EIGEN_PLAIN_ENUM_MIN(MatrixType::MaxRowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0), + MatrixType::MaxColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))), + MaxColsAtCompileTime = 1, + MaskLvalueBit = is_lvalue::value ? LvalueBit : 0, + Flags = (unsigned int)_MatrixTypeNested::Flags & (RowMajorBit | MaskLvalueBit | DirectAccessBit) & ~RowMajorBit, // FIXME DirectAccessBit should not be handled by expressions + MatrixTypeOuterStride = outer_stride_at_compile_time::ret, + InnerStrideAtCompileTime = MatrixTypeOuterStride == Dynamic ? Dynamic : MatrixTypeOuterStride+1, + OuterStrideAtCompileTime = 0 + }; +}; +} + +template class Diagonal + : public internal::dense_xpr_base< Diagonal >::type +{ + public: + + enum { DiagIndex = _DiagIndex }; + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Diagonal) + + EIGEN_DEVICE_FUNC + explicit inline Diagonal(MatrixType& matrix, Index a_index = DiagIndex) : m_matrix(matrix), m_index(a_index) + { + eigen_assert( a_index <= m_matrix.cols() && -a_index <= m_matrix.rows() ); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Diagonal) + + EIGEN_DEVICE_FUNC + inline Index rows() const + { + return m_index.value()<0 ? numext::mini(m_matrix.cols(),m_matrix.rows()+m_index.value()) + : numext::mini(m_matrix.rows(),m_matrix.cols()-m_index.value()); + } + + EIGEN_DEVICE_FUNC + inline Index cols() const { return 1; } + + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return m_matrix.outerStride() + 1; + } + + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return 0; + } + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return &(m_matrix.coeffRef(rowOffset(), colOffset())); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return &(m_matrix.coeffRef(rowOffset(), colOffset())); } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index row, Index) + { + EIGEN_STATIC_ASSERT_LVALUE(MatrixType) + return m_matrix.coeffRef(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index row, Index) const + { + return m_matrix.coeffRef(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline CoeffReturnType coeff(Index row, Index) const + { + return m_matrix.coeff(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index idx) + { + EIGEN_STATIC_ASSERT_LVALUE(MatrixType) + return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index idx) const + { + return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline CoeffReturnType coeff(Index idx) const + { + return m_matrix.coeff(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const typename internal::remove_all::type& + nestedExpression() const + { + return m_matrix; + } + + EIGEN_DEVICE_FUNC + inline Index index() const + { + return m_index.value(); + } + + protected: + typename internal::ref_selector::non_const_type m_matrix; + const internal::variable_if_dynamicindex m_index; + + private: + // some compilers may fail to optimize std::max etc in case of compile-time constants... + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index absDiagIndex() const { return m_index.value()>0 ? m_index.value() : -m_index.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rowOffset() const { return m_index.value()>0 ? 0 : -m_index.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index colOffset() const { return m_index.value()>0 ? m_index.value() : 0; } + // trigger a compile-time error if someone try to call packet + template typename MatrixType::PacketReturnType packet(Index) const; + template typename MatrixType::PacketReturnType packet(Index,Index) const; +}; + +/** \returns an expression of the main diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * Example: \include MatrixBase_diagonal.cpp + * Output: \verbinclude MatrixBase_diagonal.out + * + * \sa class Diagonal */ +template +inline typename MatrixBase::DiagonalReturnType +MatrixBase::diagonal() +{ + return DiagonalReturnType(derived()); +} + +/** This is the const version of diagonal(). */ +template +inline typename MatrixBase::ConstDiagonalReturnType +MatrixBase::diagonal() const +{ + return ConstDiagonalReturnType(derived()); +} + +/** \returns an expression of the \a DiagIndex-th sub or super diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * The template parameter \a DiagIndex represent a super diagonal if \a DiagIndex > 0 + * and a sub diagonal otherwise. \a DiagIndex == 0 is equivalent to the main diagonal. + * + * Example: \include MatrixBase_diagonal_int.cpp + * Output: \verbinclude MatrixBase_diagonal_int.out + * + * \sa MatrixBase::diagonal(), class Diagonal */ +template +inline typename MatrixBase::DiagonalDynamicIndexReturnType +MatrixBase::diagonal(Index index) +{ + return DiagonalDynamicIndexReturnType(derived(), index); +} + +/** This is the const version of diagonal(Index). */ +template +inline typename MatrixBase::ConstDiagonalDynamicIndexReturnType +MatrixBase::diagonal(Index index) const +{ + return ConstDiagonalDynamicIndexReturnType(derived(), index); +} + +/** \returns an expression of the \a DiagIndex-th sub or super diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * The template parameter \a DiagIndex represent a super diagonal if \a DiagIndex > 0 + * and a sub diagonal otherwise. \a DiagIndex == 0 is equivalent to the main diagonal. + * + * Example: \include MatrixBase_diagonal_template_int.cpp + * Output: \verbinclude MatrixBase_diagonal_template_int.out + * + * \sa MatrixBase::diagonal(), class Diagonal */ +template +template +inline typename MatrixBase::template DiagonalIndexReturnType::Type +MatrixBase::diagonal() +{ + return typename DiagonalIndexReturnType::Type(derived()); +} + +/** This is the const version of diagonal(). */ +template +template +inline typename MatrixBase::template ConstDiagonalIndexReturnType::Type +MatrixBase::diagonal() const +{ + return typename ConstDiagonalIndexReturnType::Type(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_DIAGONAL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h new file mode 100644 index 00000000000..ecfdce8efa1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h @@ -0,0 +1,343 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// Copyright (C) 2007-2009 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONALMATRIX_H +#define EIGEN_DIAGONALMATRIX_H + +namespace Eigen { + +#ifndef EIGEN_PARSED_BY_DOXYGEN +template +class DiagonalBase : public EigenBase +{ + public: + typedef typename internal::traits::DiagonalVectorType DiagonalVectorType; + typedef typename DiagonalVectorType::Scalar Scalar; + typedef typename DiagonalVectorType::RealScalar RealScalar; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + + enum { + RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + ColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + IsVectorAtCompileTime = 0, + Flags = NoPreferredStorageOrderBit + }; + + typedef Matrix DenseMatrixType; + typedef DenseMatrixType DenseType; + typedef DiagonalMatrix PlainObject; + + EIGEN_DEVICE_FUNC + inline const Derived& derived() const { return *static_cast(this); } + EIGEN_DEVICE_FUNC + inline Derived& derived() { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + DenseMatrixType toDenseMatrix() const { return derived(); } + + EIGEN_DEVICE_FUNC + inline const DiagonalVectorType& diagonal() const { return derived().diagonal(); } + EIGEN_DEVICE_FUNC + inline DiagonalVectorType& diagonal() { return derived().diagonal(); } + + EIGEN_DEVICE_FUNC + inline Index rows() const { return diagonal().size(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return diagonal().size(); } + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const MatrixBase &matrix) const + { + return Product(derived(),matrix.derived()); + } + + typedef DiagonalWrapper, const DiagonalVectorType> > InverseReturnType; + EIGEN_DEVICE_FUNC + inline const InverseReturnType + inverse() const + { + return InverseReturnType(diagonal().cwiseInverse()); + } + + EIGEN_DEVICE_FUNC + inline const DiagonalWrapper + operator*(const Scalar& scalar) const + { + return DiagonalWrapper(diagonal() * scalar); + } + EIGEN_DEVICE_FUNC + friend inline const DiagonalWrapper + operator*(const Scalar& scalar, const DiagonalBase& other) + { + return DiagonalWrapper(scalar * other.diagonal()); + } +}; + +#endif + +/** \class DiagonalMatrix + * \ingroup Core_Module + * + * \brief Represents a diagonal matrix with its storage + * + * \param _Scalar the type of coefficients + * \param SizeAtCompileTime the dimension of the matrix, or Dynamic + * \param MaxSizeAtCompileTime the dimension of the matrix, or Dynamic. This parameter is optional and defaults + * to SizeAtCompileTime. Most of the time, you do not need to specify it. + * + * \sa class DiagonalWrapper + */ + +namespace internal { +template +struct traits > + : traits > +{ + typedef Matrix<_Scalar,SizeAtCompileTime,1,0,MaxSizeAtCompileTime,1> DiagonalVectorType; + typedef DiagonalShape StorageKind; + enum { + Flags = LvalueBit | NoPreferredStorageOrderBit + }; +}; +} +template +class DiagonalMatrix + : public DiagonalBase > +{ + public: + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename internal::traits::DiagonalVectorType DiagonalVectorType; + typedef const DiagonalMatrix& Nested; + typedef _Scalar Scalar; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + #endif + + protected: + + DiagonalVectorType m_diagonal; + + public: + + /** const version of diagonal(). */ + EIGEN_DEVICE_FUNC + inline const DiagonalVectorType& diagonal() const { return m_diagonal; } + /** \returns a reference to the stored vector of diagonal coefficients. */ + EIGEN_DEVICE_FUNC + inline DiagonalVectorType& diagonal() { return m_diagonal; } + + /** Default constructor without initialization */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix() {} + + /** Constructs a diagonal matrix with given dimension */ + EIGEN_DEVICE_FUNC + explicit inline DiagonalMatrix(Index dim) : m_diagonal(dim) {} + + /** 2D constructor. */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const Scalar& x, const Scalar& y) : m_diagonal(x,y) {} + + /** 3D constructor. */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const Scalar& x, const Scalar& y, const Scalar& z) : m_diagonal(x,y,z) {} + + /** Copy constructor. */ + template + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const DiagonalBase& other) : m_diagonal(other.diagonal()) {} + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** copy constructor. prevent a default copy constructor from hiding the other templated constructor */ + inline DiagonalMatrix(const DiagonalMatrix& other) : m_diagonal(other.diagonal()) {} + #endif + + /** generic constructor from expression of the diagonal coefficients */ + template + EIGEN_DEVICE_FUNC + explicit inline DiagonalMatrix(const MatrixBase& other) : m_diagonal(other) + {} + + /** Copy operator. */ + template + EIGEN_DEVICE_FUNC + DiagonalMatrix& operator=(const DiagonalBase& other) + { + m_diagonal = other.diagonal(); + return *this; + } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + DiagonalMatrix& operator=(const DiagonalMatrix& other) + { + m_diagonal = other.diagonal(); + return *this; + } + #endif + + /** Resizes to given size. */ + EIGEN_DEVICE_FUNC + inline void resize(Index size) { m_diagonal.resize(size); } + /** Sets all coefficients to zero. */ + EIGEN_DEVICE_FUNC + inline void setZero() { m_diagonal.setZero(); } + /** Resizes and sets all coefficients to zero. */ + EIGEN_DEVICE_FUNC + inline void setZero(Index size) { m_diagonal.setZero(size); } + /** Sets this matrix to be the identity matrix of the current size. */ + EIGEN_DEVICE_FUNC + inline void setIdentity() { m_diagonal.setOnes(); } + /** Sets this matrix to be the identity matrix of the given size. */ + EIGEN_DEVICE_FUNC + inline void setIdentity(Index size) { m_diagonal.setOnes(size); } +}; + +/** \class DiagonalWrapper + * \ingroup Core_Module + * + * \brief Expression of a diagonal matrix + * + * \param _DiagonalVectorType the type of the vector of diagonal coefficients + * + * This class is an expression of a diagonal matrix, but not storing its own vector of diagonal coefficients, + * instead wrapping an existing vector expression. It is the return type of MatrixBase::asDiagonal() + * and most of the time this is the only way that it is used. + * + * \sa class DiagonalMatrix, class DiagonalBase, MatrixBase::asDiagonal() + */ + +namespace internal { +template +struct traits > +{ + typedef _DiagonalVectorType DiagonalVectorType; + typedef typename DiagonalVectorType::Scalar Scalar; + typedef typename DiagonalVectorType::StorageIndex StorageIndex; + typedef DiagonalShape StorageKind; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + ColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + Flags = (traits::Flags & LvalueBit) | NoPreferredStorageOrderBit + }; +}; +} + +template +class DiagonalWrapper + : public DiagonalBase >, internal::no_assignment_operator +{ + public: + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef _DiagonalVectorType DiagonalVectorType; + typedef DiagonalWrapper Nested; + #endif + + /** Constructor from expression of diagonal coefficients to wrap. */ + EIGEN_DEVICE_FUNC + explicit inline DiagonalWrapper(DiagonalVectorType& a_diagonal) : m_diagonal(a_diagonal) {} + + /** \returns a const reference to the wrapped expression of diagonal coefficients. */ + EIGEN_DEVICE_FUNC + const DiagonalVectorType& diagonal() const { return m_diagonal; } + + protected: + typename DiagonalVectorType::Nested m_diagonal; +}; + +/** \returns a pseudo-expression of a diagonal matrix with *this as vector of diagonal coefficients + * + * \only_for_vectors + * + * Example: \include MatrixBase_asDiagonal.cpp + * Output: \verbinclude MatrixBase_asDiagonal.out + * + * \sa class DiagonalWrapper, class DiagonalMatrix, diagonal(), isDiagonal() + **/ +template +inline const DiagonalWrapper +MatrixBase::asDiagonal() const +{ + return DiagonalWrapper(derived()); +} + +/** \returns true if *this is approximately equal to a diagonal matrix, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isDiagonal.cpp + * Output: \verbinclude MatrixBase_isDiagonal.out + * + * \sa asDiagonal() + */ +template +bool MatrixBase::isDiagonal(const RealScalar& prec) const +{ + if(cols() != rows()) return false; + RealScalar maxAbsOnDiagonal = static_cast(-1); + for(Index j = 0; j < cols(); ++j) + { + RealScalar absOnDiagonal = numext::abs(coeff(j,j)); + if(absOnDiagonal > maxAbsOnDiagonal) maxAbsOnDiagonal = absOnDiagonal; + } + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < j; ++i) + { + if(!internal::isMuchSmallerThan(coeff(i, j), maxAbsOnDiagonal, prec)) return false; + if(!internal::isMuchSmallerThan(coeff(j, i), maxAbsOnDiagonal, prec)) return false; + } + return true; +} + +namespace internal { + +template<> struct storage_kind_to_shape { typedef DiagonalShape Shape; }; + +struct Diagonal2Dense {}; + +template<> struct AssignmentKind { typedef Diagonal2Dense Kind; }; + +// Diagonal matrix to Dense assignment +template< typename DstXprType, typename SrcXprType, typename Functor> +struct Assignment +{ + static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + dst.setZero(); + dst.diagonal() = src.diagonal(); + } + + static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &/*func*/) + { dst.diagonal() += src.diagonal(); } + + static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &/*func*/) + { dst.diagonal() -= src.diagonal(); } +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DIAGONALMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h new file mode 100644 index 00000000000..d372b938f65 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h @@ -0,0 +1,28 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2007-2009 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONALPRODUCT_H +#define EIGEN_DIAGONALPRODUCT_H + +namespace Eigen { + +/** \returns the diagonal matrix product of \c *this by the diagonal matrix \a diagonal. + */ +template +template +inline const Product +MatrixBase::operator*(const DiagonalBase &a_diagonal) const +{ + return Product(derived(),a_diagonal.derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_DIAGONALPRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h new file mode 100644 index 00000000000..1fe7a84a48d --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h @@ -0,0 +1,318 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008, 2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DOT_H +#define EIGEN_DOT_H + +namespace Eigen { + +namespace internal { + +// helper function for dot(). The problem is that if we put that in the body of dot(), then upon calling dot +// with mismatched types, the compiler emits errors about failing to instantiate cwiseProduct BEFORE +// looking at the static assertions. Thus this is a trick to get better compile errors. +template +struct dot_nocheck +{ + typedef scalar_conj_product_op::Scalar,typename traits::Scalar> conj_prod; + typedef typename conj_prod::result_type ResScalar; + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE + static ResScalar run(const MatrixBase& a, const MatrixBase& b) + { + return a.template binaryExpr(b).sum(); + } +}; + +template +struct dot_nocheck +{ + typedef scalar_conj_product_op::Scalar,typename traits::Scalar> conj_prod; + typedef typename conj_prod::result_type ResScalar; + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE + static ResScalar run(const MatrixBase& a, const MatrixBase& b) + { + return a.transpose().template binaryExpr(b).sum(); + } +}; + +} // end namespace internal + +/** \fn MatrixBase::dot + * \returns the dot product of *this with other. + * + * \only_for_vectors + * + * \note If the scalar type is complex numbers, then this function returns the hermitian + * (sesquilinear) dot product, conjugate-linear in the first variable and linear in the + * second variable. + * + * \sa squaredNorm(), norm() + */ +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE +typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType +MatrixBase::dot(const MatrixBase& other) const +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) + EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived) +#if !(defined(EIGEN_NO_STATIC_ASSERT) && defined(EIGEN_NO_DEBUG)) + typedef internal::scalar_conj_product_op func; + EIGEN_CHECK_BINARY_COMPATIBILIY(func,Scalar,typename OtherDerived::Scalar); +#endif + + eigen_assert(size() == other.size()); + + return internal::dot_nocheck::run(*this, other); +} + +//---------- implementation of L2 norm and related functions ---------- + +/** \returns, for vectors, the squared \em l2 norm of \c *this, and for matrices the Frobenius norm. + * In both cases, it consists in the sum of the square of all the matrix entries. + * For vectors, this is also equals to the dot product of \c *this with itself. + * + * \sa dot(), norm(), lpNorm() + */ +template +EIGEN_STRONG_INLINE typename NumTraits::Scalar>::Real MatrixBase::squaredNorm() const +{ + return numext::real((*this).cwiseAbs2().sum()); +} + +/** \returns, for vectors, the \em l2 norm of \c *this, and for matrices the Frobenius norm. + * In both cases, it consists in the square root of the sum of the square of all the matrix entries. + * For vectors, this is also equals to the square root of the dot product of \c *this with itself. + * + * \sa lpNorm(), dot(), squaredNorm() + */ +template +EIGEN_STRONG_INLINE typename NumTraits::Scalar>::Real MatrixBase::norm() const +{ + return numext::sqrt(squaredNorm()); +} + +/** \returns an expression of the quotient of \c *this by its own norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), + * then this function returns a copy of the input. + * + * \only_for_vectors + * + * \sa norm(), normalize() + */ +template +EIGEN_STRONG_INLINE const typename MatrixBase::PlainObject +MatrixBase::normalized() const +{ + typedef typename internal::nested_eval::type _Nested; + _Nested n(derived()); + RealScalar z = n.squaredNorm(); + // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU + if(z>RealScalar(0)) + return n / numext::sqrt(z); + else + return n; +} + +/** Normalizes the vector, i.e. divides it by its own norm. + * + * \only_for_vectors + * + * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged. + * + * \sa norm(), normalized() + */ +template +EIGEN_STRONG_INLINE void MatrixBase::normalize() +{ + RealScalar z = squaredNorm(); + // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU + if(z>RealScalar(0)) + derived() /= numext::sqrt(z); +} + +/** \returns an expression of the quotient of \c *this by its own norm while avoiding underflow and overflow. + * + * \only_for_vectors + * + * This method is analogue to the normalized() method, but it reduces the risk of + * underflow and overflow when computing the norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), + * then this function returns a copy of the input. + * + * \sa stableNorm(), stableNormalize(), normalized() + */ +template +EIGEN_STRONG_INLINE const typename MatrixBase::PlainObject +MatrixBase::stableNormalized() const +{ + typedef typename internal::nested_eval::type _Nested; + _Nested n(derived()); + RealScalar w = n.cwiseAbs().maxCoeff(); + RealScalar z = (n/w).squaredNorm(); + if(z>RealScalar(0)) + return n / (numext::sqrt(z)*w); + else + return n; +} + +/** Normalizes the vector while avoid underflow and overflow + * + * \only_for_vectors + * + * This method is analogue to the normalize() method, but it reduces the risk of + * underflow and overflow when computing the norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged. + * + * \sa stableNorm(), stableNormalized(), normalize() + */ +template +EIGEN_STRONG_INLINE void MatrixBase::stableNormalize() +{ + RealScalar w = cwiseAbs().maxCoeff(); + RealScalar z = (derived()/w).squaredNorm(); + if(z>RealScalar(0)) + derived() /= numext::sqrt(z)*w; +} + +//---------- implementation of other norms ---------- + +namespace internal { + +template +struct lpNorm_selector +{ + typedef typename NumTraits::Scalar>::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const MatrixBase& m) + { + EIGEN_USING_STD_MATH(pow) + return pow(m.cwiseAbs().array().pow(p).sum(), RealScalar(1)/p); + } +}; + +template +struct lpNorm_selector +{ + EIGEN_DEVICE_FUNC + static inline typename NumTraits::Scalar>::Real run(const MatrixBase& m) + { + return m.cwiseAbs().sum(); + } +}; + +template +struct lpNorm_selector +{ + EIGEN_DEVICE_FUNC + static inline typename NumTraits::Scalar>::Real run(const MatrixBase& m) + { + return m.norm(); + } +}; + +template +struct lpNorm_selector +{ + typedef typename NumTraits::Scalar>::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const MatrixBase& m) + { + if(Derived::SizeAtCompileTime==0 || (Derived::SizeAtCompileTime==Dynamic && m.size()==0)) + return RealScalar(0); + return m.cwiseAbs().maxCoeff(); + } +}; + +} // end namespace internal + +/** \returns the \b coefficient-wise \f$ \ell^p \f$ norm of \c *this, that is, returns the p-th root of the sum of the p-th powers of the absolute values + * of the coefficients of \c *this. If \a p is the special value \a Eigen::Infinity, this function returns the \f$ \ell^\infty \f$ + * norm, that is the maximum of the absolute values of the coefficients of \c *this. + * + * In all cases, if \c *this is empty, then the value 0 is returned. + * + * \note For matrices, this function does not compute the operator-norm. That is, if \c *this is a matrix, then its coefficients are interpreted as a 1D vector. Nonetheless, you can easily compute the 1-norm and \f$\infty\f$-norm matrix operator norms using \link TutorialReductionsVisitorsBroadcastingReductionsNorm partial reductions \endlink. + * + * \sa norm() + */ +template +template +#ifndef EIGEN_PARSED_BY_DOXYGEN +inline typename NumTraits::Scalar>::Real +#else +MatrixBase::RealScalar +#endif +MatrixBase::lpNorm() const +{ + return internal::lpNorm_selector::run(*this); +} + +//---------- implementation of isOrthogonal / isUnitary ---------- + +/** \returns true if *this is approximately orthogonal to \a other, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isOrthogonal.cpp + * Output: \verbinclude MatrixBase_isOrthogonal.out + */ +template +template +bool MatrixBase::isOrthogonal +(const MatrixBase& other, const RealScalar& prec) const +{ + typename internal::nested_eval::type nested(derived()); + typename internal::nested_eval::type otherNested(other.derived()); + return numext::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm(); +} + +/** \returns true if *this is approximately an unitary matrix, + * within the precision given by \a prec. In the case where the \a Scalar + * type is real numbers, a unitary matrix is an orthogonal matrix, whence the name. + * + * \note This can be used to check whether a family of vectors forms an orthonormal basis. + * Indeed, \c m.isUnitary() returns true if and only if the columns (equivalently, the rows) of m form an + * orthonormal basis. + * + * Example: \include MatrixBase_isUnitary.cpp + * Output: \verbinclude MatrixBase_isUnitary.out + */ +template +bool MatrixBase::isUnitary(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index i = 0; i < cols(); ++i) + { + if(!internal::isApprox(self.col(i).squaredNorm(), static_cast(1), prec)) + return false; + for(Index j = 0; j < i; ++j) + if(!internal::isMuchSmallerThan(self.col(i).dot(self.col(j)), static_cast(1), prec)) + return false; + } + return true; +} + +} // end namespace Eigen + +#endif // EIGEN_DOT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h new file mode 100644 index 00000000000..b195506a91e --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h @@ -0,0 +1,159 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Benoit Jacob +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_EIGENBASE_H +#define EIGEN_EIGENBASE_H + +namespace Eigen { + +/** \class EigenBase + * \ingroup Core_Module + * + * Common base class for all classes T such that MatrixBase has an operator=(T) and a constructor MatrixBase(T). + * + * In other words, an EigenBase object is an object that can be copied into a MatrixBase. + * + * Besides MatrixBase-derived classes, this also includes special matrix classes such as diagonal matrices, etc. + * + * Notice that this class is trivial, it is only used to disambiguate overloaded functions. + * + * \sa \blank \ref TopicClassHierarchy + */ +template struct EigenBase +{ +// typedef typename internal::plain_matrix_type::type PlainObject; + + /** \brief The interface type of indices + * \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE. + * \deprecated Since Eigen 3.3, its usage is deprecated. Use Eigen::Index instead. + * \sa StorageIndex, \ref TopicPreprocessorDirectives. + */ + typedef Eigen::Index Index; + + // FIXME is it needed? + typedef typename internal::traits::StorageKind StorageKind; + + /** \returns a reference to the derived object */ + EIGEN_DEVICE_FUNC + Derived& derived() { return *static_cast(this); } + /** \returns a const reference to the derived object */ + EIGEN_DEVICE_FUNC + const Derived& derived() const { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + inline Derived& const_cast_derived() const + { return *static_cast(const_cast(this)); } + EIGEN_DEVICE_FUNC + inline const Derived& const_derived() const + { return *static_cast(this); } + + /** \returns the number of rows. \sa cols(), RowsAtCompileTime */ + EIGEN_DEVICE_FUNC + inline Index rows() const { return derived().rows(); } + /** \returns the number of columns. \sa rows(), ColsAtCompileTime*/ + EIGEN_DEVICE_FUNC + inline Index cols() const { return derived().cols(); } + /** \returns the number of coefficients, which is rows()*cols(). + * \sa rows(), cols(), SizeAtCompileTime. */ + EIGEN_DEVICE_FUNC + inline Index size() const { return rows() * cols(); } + + /** \internal Don't use it, but do the equivalent: \code dst = *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const + { derived().evalTo(dst); } + + /** \internal Don't use it, but do the equivalent: \code dst += *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void addTo(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + typename Dest::PlainObject res(rows(),cols()); + evalTo(res); + dst += res; + } + + /** \internal Don't use it, but do the equivalent: \code dst -= *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void subTo(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + typename Dest::PlainObject res(rows(),cols()); + evalTo(res); + dst -= res; + } + + /** \internal Don't use it, but do the equivalent: \code dst.applyOnTheRight(*this); \endcode */ + template + EIGEN_DEVICE_FUNC inline void applyThisOnTheRight(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + dst = dst * this->derived(); + } + + /** \internal Don't use it, but do the equivalent: \code dst.applyOnTheLeft(*this); \endcode */ + template + EIGEN_DEVICE_FUNC inline void applyThisOnTheLeft(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + dst = this->derived() * dst; + } + +}; + +/*************************************************************************** +* Implementation of matrix base methods +***************************************************************************/ + +/** \brief Copies the generic expression \a other into *this. + * + * \details The expression must provide a (templated) evalTo(Derived& dst) const + * function which does the actual job. In practice, this allows any user to write + * its own special matrix without having to modify MatrixBase + * + * \returns a reference to *this. + */ +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator=(const EigenBase &other) +{ + call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator+=(const EigenBase &other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator-=(const EigenBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_EIGENBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ForceAlignedAccess.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ForceAlignedAccess.h new file mode 100644 index 00000000000..7b08b45e67c --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ForceAlignedAccess.h @@ -0,0 +1,146 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_FORCEALIGNEDACCESS_H +#define EIGEN_FORCEALIGNEDACCESS_H + +namespace Eigen { + +/** \class ForceAlignedAccess + * \ingroup Core_Module + * + * \brief Enforce aligned packet loads and stores regardless of what is requested + * + * \param ExpressionType the type of the object of which we are forcing aligned packet access + * + * This class is the return type of MatrixBase::forceAlignedAccess() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::forceAlignedAccess() + */ + +namespace internal { +template +struct traits > : public traits +{}; +} + +template class ForceAlignedAccess + : public internal::dense_xpr_base< ForceAlignedAccess >::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(ForceAlignedAccess) + + EIGEN_DEVICE_FUNC explicit inline ForceAlignedAccess(const ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index row, Index col) const + { + return m_expression.coeff(row, col); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index row, Index col) + { + return m_expression.const_cast_derived().coeffRef(row, col); + } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index index) const + { + return m_expression.coeff(index); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index index) + { + return m_expression.const_cast_derived().coeffRef(index); + } + + template + inline const PacketScalar packet(Index row, Index col) const + { + return m_expression.template packet(row, col); + } + + template + inline void writePacket(Index row, Index col, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(row, col, x); + } + + template + inline const PacketScalar packet(Index index) const + { + return m_expression.template packet(index); + } + + template + inline void writePacket(Index index, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(index, x); + } + + EIGEN_DEVICE_FUNC operator const ExpressionType&() const { return m_expression; } + + protected: + const ExpressionType& m_expression; + + private: + ForceAlignedAccess& operator=(const ForceAlignedAccess&); +}; + +/** \returns an expression of *this with forced aligned access + * \sa forceAlignedAccessIf(),class ForceAlignedAccess + */ +template +inline const ForceAlignedAccess +MatrixBase::forceAlignedAccess() const +{ + return ForceAlignedAccess(derived()); +} + +/** \returns an expression of *this with forced aligned access + * \sa forceAlignedAccessIf(), class ForceAlignedAccess + */ +template +inline ForceAlignedAccess +MatrixBase::forceAlignedAccess() +{ + return ForceAlignedAccess(derived()); +} + +/** \returns an expression of *this with forced aligned access if \a Enable is true. + * \sa forceAlignedAccess(), class ForceAlignedAccess + */ +template +template +inline typename internal::add_const_on_value_type,Derived&>::type>::type +MatrixBase::forceAlignedAccessIf() const +{ + return derived(); // FIXME This should not work but apparently is never used +} + +/** \returns an expression of *this with forced aligned access if \a Enable is true. + * \sa forceAlignedAccess(), class ForceAlignedAccess + */ +template +template +inline typename internal::conditional,Derived&>::type +MatrixBase::forceAlignedAccessIf() +{ + return derived(); // FIXME This should not work but apparently is never used +} + +} // end namespace Eigen + +#endif // EIGEN_FORCEALIGNEDACCESS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h new file mode 100644 index 00000000000..3e403a09d92 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h @@ -0,0 +1,155 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_FUZZY_H +#define EIGEN_FUZZY_H + +namespace Eigen { + +namespace internal +{ + +template::IsInteger> +struct isApprox_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec) + { + typename internal::nested_eval::type nested(x); + typename internal::nested_eval::type otherNested(y); + return (nested - otherNested).cwiseAbs2().sum() <= prec * prec * numext::mini(nested.cwiseAbs2().sum(), otherNested.cwiseAbs2().sum()); + } +}; + +template +struct isApprox_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar&) + { + return x.matrix() == y.matrix(); + } +}; + +template::IsInteger> +struct isMuchSmallerThan_object_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec) + { + return x.cwiseAbs2().sum() <= numext::abs2(prec) * y.cwiseAbs2().sum(); + } +}; + +template +struct isMuchSmallerThan_object_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived&, const typename Derived::RealScalar&) + { + return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix(); + } +}; + +template::IsInteger> +struct isMuchSmallerThan_scalar_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const typename Derived::RealScalar& y, const typename Derived::RealScalar& prec) + { + return x.cwiseAbs2().sum() <= numext::abs2(prec * y); + } +}; + +template +struct isMuchSmallerThan_scalar_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const typename Derived::RealScalar&, const typename Derived::RealScalar&) + { + return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix(); + } +}; + +} // end namespace internal + + +/** \returns \c true if \c *this is approximately equal to \a other, within the precision + * determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. Two vectors \f$ v \f$ and \f$ w \f$ + * are considered to be approximately equal within precision \f$ p \f$ if + * \f[ \Vert v - w \Vert \leqslant p\,\min(\Vert v\Vert, \Vert w\Vert). \f] + * For matrices, the comparison is done using the Hilbert-Schmidt norm (aka Frobenius norm + * L2 norm). + * + * \note Because of the multiplicativeness of this comparison, one can't use this function + * to check whether \c *this is approximately equal to the zero matrix or vector. + * Indeed, \c isApprox(zero) returns false unless \c *this itself is exactly the zero matrix + * or vector. If you want to test whether \c *this is zero, use internal::isMuchSmallerThan(const + * RealScalar&, RealScalar) instead. + * + * \sa internal::isMuchSmallerThan(const RealScalar&, RealScalar) const + */ +template +template +bool DenseBase::isApprox( + const DenseBase& other, + const RealScalar& prec +) const +{ + return internal::isApprox_selector::run(derived(), other.derived(), prec); +} + +/** \returns \c true if the norm of \c *this is much smaller than \a other, + * within the precision determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. A vector \f$ v \f$ is + * considered to be much smaller than \f$ x \f$ within precision \f$ p \f$ if + * \f[ \Vert v \Vert \leqslant p\,\vert x\vert. \f] + * + * For matrices, the comparison is done using the Hilbert-Schmidt norm. For this reason, + * the value of the reference scalar \a other should come from the Hilbert-Schmidt norm + * of a reference matrix of same dimensions. + * + * \sa isApprox(), isMuchSmallerThan(const DenseBase&, RealScalar) const + */ +template +bool DenseBase::isMuchSmallerThan( + const typename NumTraits::Real& other, + const RealScalar& prec +) const +{ + return internal::isMuchSmallerThan_scalar_selector::run(derived(), other, prec); +} + +/** \returns \c true if the norm of \c *this is much smaller than the norm of \a other, + * within the precision determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. A vector \f$ v \f$ is + * considered to be much smaller than a vector \f$ w \f$ within precision \f$ p \f$ if + * \f[ \Vert v \Vert \leqslant p\,\Vert w\Vert. \f] + * For matrices, the comparison is done using the Hilbert-Schmidt norm. + * + * \sa isApprox(), isMuchSmallerThan(const RealScalar&, RealScalar) const + */ +template +template +bool DenseBase::isMuchSmallerThan( + const DenseBase& other, + const RealScalar& prec +) const +{ + return internal::isMuchSmallerThan_object_selector::run(derived(), other.derived(), prec); +} + +} // end namespace Eigen + +#endif // EIGEN_FUZZY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h new file mode 100644 index 00000000000..6f0cc80e940 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h @@ -0,0 +1,455 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008-2011 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GENERAL_PRODUCT_H +#define EIGEN_GENERAL_PRODUCT_H + +namespace Eigen { + +enum { + Large = 2, + Small = 3 +}; + +namespace internal { + +template struct product_type_selector; + +template struct product_size_category +{ + enum { + #ifndef EIGEN_CUDA_ARCH + is_large = MaxSize == Dynamic || + Size >= EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD || + (Size==Dynamic && MaxSize>=EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD), + #else + is_large = 0, + #endif + value = is_large ? Large + : Size == 1 ? 1 + : Small + }; +}; + +template struct product_type +{ + typedef typename remove_all::type _Lhs; + typedef typename remove_all::type _Rhs; + enum { + MaxRows = traits<_Lhs>::MaxRowsAtCompileTime, + Rows = traits<_Lhs>::RowsAtCompileTime, + MaxCols = traits<_Rhs>::MaxColsAtCompileTime, + Cols = traits<_Rhs>::ColsAtCompileTime, + MaxDepth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::MaxColsAtCompileTime, + traits<_Rhs>::MaxRowsAtCompileTime), + Depth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::ColsAtCompileTime, + traits<_Rhs>::RowsAtCompileTime) + }; + + // the splitting into different lines of code here, introducing the _select enums and the typedef below, + // is to work around an internal compiler error with gcc 4.1 and 4.2. +private: + enum { + rows_select = product_size_category::value, + cols_select = product_size_category::value, + depth_select = product_size_category::value + }; + typedef product_type_selector selector; + +public: + enum { + value = selector::ret, + ret = selector::ret + }; +#ifdef EIGEN_DEBUG_PRODUCT + static void debug() + { + EIGEN_DEBUG_VAR(Rows); + EIGEN_DEBUG_VAR(Cols); + EIGEN_DEBUG_VAR(Depth); + EIGEN_DEBUG_VAR(rows_select); + EIGEN_DEBUG_VAR(cols_select); + EIGEN_DEBUG_VAR(depth_select); + EIGEN_DEBUG_VAR(value); + } +#endif +}; + +/* The following allows to select the kind of product at compile time + * based on the three dimensions of the product. + * This is a compile time mapping from {1,Small,Large}^3 -> {product types} */ +// FIXME I'm not sure the current mapping is the ideal one. +template struct product_type_selector { enum { ret = OuterProduct }; }; +template struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template struct product_type_selector<1, N, 1> { enum { ret = LazyCoeffBasedProductMode }; }; +template struct product_type_selector<1, 1, Depth> { enum { ret = InnerProduct }; }; +template<> struct product_type_selector<1, 1, 1> { enum { ret = InnerProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Small,Small> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Large,Small> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Large,Large> { enum { ret = GemvProduct }; }; +template<> struct product_type_selector<1, Small,Large> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemvProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; + +} // end namespace internal + +/*********************************************************************** +* Implementation of Inner Vector Vector Product +***********************************************************************/ + +// FIXME : maybe the "inner product" could return a Scalar +// instead of a 1x1 matrix ?? +// Pro: more natural for the user +// Cons: this could be a problem if in a meta unrolled algorithm a matrix-matrix +// product ends up to a row-vector times col-vector product... To tackle this use +// case, we could have a specialization for Block with: operator=(Scalar x); + +/*********************************************************************** +* Implementation of Outer Vector Vector Product +***********************************************************************/ + +/*********************************************************************** +* Implementation of General Matrix Vector Product +***********************************************************************/ + +/* According to the shape/flags of the matrix we have to distinghish 3 different cases: + * 1 - the matrix is col-major, BLAS compatible and M is large => call fast BLAS-like colmajor routine + * 2 - the matrix is row-major, BLAS compatible and N is large => call fast BLAS-like rowmajor routine + * 3 - all other cases are handled using a simple loop along the outer-storage direction. + * Therefore we need a lower level meta selector. + * Furthermore, if the matrix is the rhs, then the product has to be transposed. + */ +namespace internal { + +template +struct gemv_dense_selector; + +} // end namespace internal + +namespace internal { + +template struct gemv_static_vector_if; + +template +struct gemv_static_vector_if +{ + EIGEN_STRONG_INLINE Scalar* data() { eigen_internal_assert(false && "should never be called"); return 0; } +}; + +template +struct gemv_static_vector_if +{ + EIGEN_STRONG_INLINE Scalar* data() { return 0; } +}; + +template +struct gemv_static_vector_if +{ + enum { + ForceAlignment = internal::packet_traits::Vectorizable, + PacketSize = internal::packet_traits::size + }; + #if EIGEN_MAX_STATIC_ALIGN_BYTES!=0 + internal::plain_array m_data; + EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; } + #else + // Some architectures cannot align on the stack, + // => let's manually enforce alignment by allocating more data and return the address of the first aligned element. + internal::plain_array m_data; + EIGEN_STRONG_INLINE Scalar* data() { + return ForceAlignment + ? reinterpret_cast((internal::UIntPtr(m_data.array) & ~(std::size_t(EIGEN_MAX_ALIGN_BYTES-1))) + EIGEN_MAX_ALIGN_BYTES) + : m_data.array; + } + #endif +}; + +// The vector is on the left => transposition +template +struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + Transpose destT(dest); + enum { OtherStorageOrder = StorageOrder == RowMajor ? ColMajor : RowMajor }; + gemv_dense_selector + ::run(rhs.transpose(), lhs.transpose(), destT, alpha); + } +}; + +template<> struct gemv_dense_selector +{ + template + static inline void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + typedef typename Lhs::Scalar LhsScalar; + typedef typename Rhs::Scalar RhsScalar; + typedef typename Dest::Scalar ResScalar; + typedef typename Dest::RealScalar RealScalar; + + typedef internal::blas_traits LhsBlasTraits; + typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; + typedef internal::blas_traits RhsBlasTraits; + typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; + + typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; + + ActualLhsType actualLhs = LhsBlasTraits::extract(lhs); + ActualRhsType actualRhs = RhsBlasTraits::extract(rhs); + + ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs) + * RhsBlasTraits::extractScalarFactor(rhs); + + // make sure Dest is a compile-time vector type (bug 1166) + typedef typename conditional::type ActualDest; + + enum { + // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 + // on, the other hand it is good for the cache to pack the vector anyways... + EvalToDestAtCompileTime = (ActualDest::InnerStrideAtCompileTime==1), + ComplexByReal = (NumTraits::IsComplex) && (!NumTraits::IsComplex), + MightCannotUseDest = (!EvalToDestAtCompileTime) || ComplexByReal + }; + + typedef const_blas_data_mapper LhsMapper; + typedef const_blas_data_mapper RhsMapper; + RhsScalar compatibleAlpha = get_factor::run(actualAlpha); + + if(!MightCannotUseDest) + { + // shortcut if we are sure to be able to use dest directly, + // this ease the compiler to generate cleaner and more optimzized code for most common cases + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhs.data(), actualRhs.innerStride()), + dest.data(), 1, + compatibleAlpha); + } + else + { + gemv_static_vector_if static_dest; + + const bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0)); + const bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible; + + ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), + evalToDest ? dest.data() : static_dest.data()); + + if(!evalToDest) + { + #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + Index size = dest.size(); + EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #endif + if(!alphaIsCompatible) + { + MappedDest(actualDestPtr, dest.size()).setZero(); + compatibleAlpha = RhsScalar(1); + } + else + MappedDest(actualDestPtr, dest.size()) = dest; + } + + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhs.data(), actualRhs.innerStride()), + actualDestPtr, 1, + compatibleAlpha); + + if (!evalToDest) + { + if(!alphaIsCompatible) + dest.matrix() += actualAlpha * MappedDest(actualDestPtr, dest.size()); + else + dest = MappedDest(actualDestPtr, dest.size()); + } + } + } +}; + +template<> struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + typedef typename Lhs::Scalar LhsScalar; + typedef typename Rhs::Scalar RhsScalar; + typedef typename Dest::Scalar ResScalar; + + typedef internal::blas_traits LhsBlasTraits; + typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; + typedef internal::blas_traits RhsBlasTraits; + typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; + typedef typename internal::remove_all::type ActualRhsTypeCleaned; + + typename add_const::type actualLhs = LhsBlasTraits::extract(lhs); + typename add_const::type actualRhs = RhsBlasTraits::extract(rhs); + + ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs) + * RhsBlasTraits::extractScalarFactor(rhs); + + enum { + // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 + // on, the other hand it is good for the cache to pack the vector anyways... + DirectlyUseRhs = ActualRhsTypeCleaned::InnerStrideAtCompileTime==1 + }; + + gemv_static_vector_if static_rhs; + + ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,actualRhs.size(), + DirectlyUseRhs ? const_cast(actualRhs.data()) : static_rhs.data()); + + if(!DirectlyUseRhs) + { + #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + Index size = actualRhs.size(); + EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #endif + Map(actualRhsPtr, actualRhs.size()) = actualRhs; + } + + typedef const_blas_data_mapper LhsMapper; + typedef const_blas_data_mapper RhsMapper; + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhsPtr, 1), + dest.data(), dest.col(0).innerStride(), //NOTE if dest is not a vector at compile-time, then dest.innerStride() might be wrong. (bug 1166) + actualAlpha); + } +}; + +template<> struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + EIGEN_STATIC_ASSERT((!nested_eval::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE); + // TODO if rhs is large enough it might be beneficial to make sure that dest is sequentially stored in memory, otherwise use a temp + typename nested_eval::type actual_rhs(rhs); + const Index size = rhs.rows(); + for(Index k=0; k struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + EIGEN_STATIC_ASSERT((!nested_eval::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE); + typename nested_eval::type actual_rhs(rhs); + const Index rows = dest.rows(); + for(Index i=0; i +template +inline const Product +MatrixBase::operator*(const MatrixBase &other) const +{ + // A note regarding the function declaration: In MSVC, this function will sometimes + // not be inlined since DenseStorage is an unwindable object for dynamic + // matrices and product types are holding a member to store the result. + // Thus it does not help tagging this function with EIGEN_STRONG_INLINE. + enum { + ProductIsValid = Derived::ColsAtCompileTime==Dynamic + || OtherDerived::RowsAtCompileTime==Dynamic + || int(Derived::ColsAtCompileTime)==int(OtherDerived::RowsAtCompileTime), + AreVectors = Derived::IsVectorAtCompileTime && OtherDerived::IsVectorAtCompileTime, + SameSizes = EIGEN_PREDICATE_SAME_MATRIX_SIZE(Derived,OtherDerived) + }; + // note to the lost user: + // * for a dot product use: v1.dot(v2) + // * for a coeff-wise product use: v1.cwiseProduct(v2) + EIGEN_STATIC_ASSERT(ProductIsValid || !(AreVectors && SameSizes), + INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS) + EIGEN_STATIC_ASSERT(ProductIsValid || !(SameSizes && !AreVectors), + INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION) + EIGEN_STATIC_ASSERT(ProductIsValid || SameSizes, INVALID_MATRIX_PRODUCT) +#ifdef EIGEN_DEBUG_PRODUCT + internal::product_type::debug(); +#endif + + return Product(derived(), other.derived()); +} + +/** \returns an expression of the matrix product of \c *this and \a other without implicit evaluation. + * + * The returned product will behave like any other expressions: the coefficients of the product will be + * computed once at a time as requested. This might be useful in some extremely rare cases when only + * a small and no coherent fraction of the result's coefficients have to be computed. + * + * \warning This version of the matrix product can be much much slower. So use it only if you know + * what you are doing and that you measured a true speed improvement. + * + * \sa operator*(const MatrixBase&) + */ +template +template +const Product +MatrixBase::lazyProduct(const MatrixBase &other) const +{ + enum { + ProductIsValid = Derived::ColsAtCompileTime==Dynamic + || OtherDerived::RowsAtCompileTime==Dynamic + || int(Derived::ColsAtCompileTime)==int(OtherDerived::RowsAtCompileTime), + AreVectors = Derived::IsVectorAtCompileTime && OtherDerived::IsVectorAtCompileTime, + SameSizes = EIGEN_PREDICATE_SAME_MATRIX_SIZE(Derived,OtherDerived) + }; + // note to the lost user: + // * for a dot product use: v1.dot(v2) + // * for a coeff-wise product use: v1.cwiseProduct(v2) + EIGEN_STATIC_ASSERT(ProductIsValid || !(AreVectors && SameSizes), + INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS) + EIGEN_STATIC_ASSERT(ProductIsValid || !(SameSizes && !AreVectors), + INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION) + EIGEN_STATIC_ASSERT(ProductIsValid || SameSizes, INVALID_MATRIX_PRODUCT) + + return Product(derived(), other.derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h new file mode 100644 index 00000000000..e594437791b --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h @@ -0,0 +1,590 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GENERIC_PACKET_MATH_H +#define EIGEN_GENERIC_PACKET_MATH_H + +namespace Eigen { + +namespace internal { + +/** \internal + * \file GenericPacketMath.h + * + * Default implementation for types not supported by the vectorization. + * In practice these functions are provided to make easier the writing + * of generic vectorized code. + */ + +#ifndef EIGEN_DEBUG_ALIGNED_LOAD +#define EIGEN_DEBUG_ALIGNED_LOAD +#endif + +#ifndef EIGEN_DEBUG_UNALIGNED_LOAD +#define EIGEN_DEBUG_UNALIGNED_LOAD +#endif + +#ifndef EIGEN_DEBUG_ALIGNED_STORE +#define EIGEN_DEBUG_ALIGNED_STORE +#endif + +#ifndef EIGEN_DEBUG_UNALIGNED_STORE +#define EIGEN_DEBUG_UNALIGNED_STORE +#endif + +struct default_packet_traits +{ + enum { + HasHalfPacket = 0, + + HasAdd = 1, + HasSub = 1, + HasMul = 1, + HasNegate = 1, + HasAbs = 1, + HasArg = 0, + HasAbs2 = 1, + HasMin = 1, + HasMax = 1, + HasConj = 1, + HasSetLinear = 1, + HasBlend = 0, + + HasDiv = 0, + HasSqrt = 0, + HasRsqrt = 0, + HasExp = 0, + HasLog = 0, + HasLog1p = 0, + HasLog10 = 0, + HasPow = 0, + + HasSin = 0, + HasCos = 0, + HasTan = 0, + HasASin = 0, + HasACos = 0, + HasATan = 0, + HasSinh = 0, + HasCosh = 0, + HasTanh = 0, + HasLGamma = 0, + HasDiGamma = 0, + HasZeta = 0, + HasPolygamma = 0, + HasErf = 0, + HasErfc = 0, + HasIGamma = 0, + HasIGammac = 0, + HasBetaInc = 0, + + HasRound = 0, + HasFloor = 0, + HasCeil = 0, + + HasSign = 0 + }; +}; + +template struct packet_traits : default_packet_traits +{ + typedef T type; + typedef T half; + enum { + Vectorizable = 0, + size = 1, + AlignedOnScalar = 0, + HasHalfPacket = 0 + }; + enum { + HasAdd = 0, + HasSub = 0, + HasMul = 0, + HasNegate = 0, + HasAbs = 0, + HasAbs2 = 0, + HasMin = 0, + HasMax = 0, + HasConj = 0, + HasSetLinear = 0 + }; +}; + +template struct packet_traits : packet_traits { }; + +template struct type_casting_traits { + enum { + VectorizedCast = 0, + SrcCoeffRatio = 1, + TgtCoeffRatio = 1 + }; +}; + + +/** \internal \returns static_cast(a) (coeff-wise) */ +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a) { + return static_cast(a); +} +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a, const SrcPacket& /*b*/) { + return static_cast(a); +} + +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a, const SrcPacket& /*b*/, const SrcPacket& /*c*/, const SrcPacket& /*d*/) { + return static_cast(a); +} + +/** \internal \returns a + b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +padd(const Packet& a, + const Packet& b) { return a+b; } + +/** \internal \returns a - b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +psub(const Packet& a, + const Packet& b) { return a-b; } + +/** \internal \returns -a (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pnegate(const Packet& a) { return -a; } + +/** \internal \returns conj(a) (coeff-wise) */ + +template EIGEN_DEVICE_FUNC inline Packet +pconj(const Packet& a) { return numext::conj(a); } + +/** \internal \returns a * b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmul(const Packet& a, + const Packet& b) { return a*b; } + +/** \internal \returns a / b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pdiv(const Packet& a, + const Packet& b) { return a/b; } + +/** \internal \returns the min of \a a and \a b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmin(const Packet& a, + const Packet& b) { return numext::mini(a, b); } + +/** \internal \returns the max of \a a and \a b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmax(const Packet& a, + const Packet& b) { return numext::maxi(a, b); } + +/** \internal \returns the absolute value of \a a */ +template EIGEN_DEVICE_FUNC inline Packet +pabs(const Packet& a) { using std::abs; return abs(a); } + +/** \internal \returns the phase angle of \a a */ +template EIGEN_DEVICE_FUNC inline Packet +parg(const Packet& a) { using numext::arg; return arg(a); } + +/** \internal \returns the bitwise and of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pand(const Packet& a, const Packet& b) { return a & b; } + +/** \internal \returns the bitwise or of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +por(const Packet& a, const Packet& b) { return a | b; } + +/** \internal \returns the bitwise xor of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pxor(const Packet& a, const Packet& b) { return a ^ b; } + +/** \internal \returns the bitwise andnot of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pandnot(const Packet& a, const Packet& b) { return a & (!b); } + +/** \internal \returns a packet version of \a *from, from must be 16 bytes aligned */ +template EIGEN_DEVICE_FUNC inline Packet +pload(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet version of \a *from, (un-aligned load) */ +template EIGEN_DEVICE_FUNC inline Packet +ploadu(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet with constant coefficients \a a, e.g.: (a,a,a,a) */ +template EIGEN_DEVICE_FUNC inline Packet +pset1(const typename unpacket_traits::type& a) { return a; } + +/** \internal \returns a packet with constant coefficients \a a[0], e.g.: (a[0],a[0],a[0],a[0]) */ +template EIGEN_DEVICE_FUNC inline Packet +pload1(const typename unpacket_traits::type *a) { return pset1(*a); } + +/** \internal \returns a packet with elements of \a *from duplicated. + * For instance, for a packet of 8 elements, 4 scalars will be read from \a *from and + * duplicated to form: {from[0],from[0],from[1],from[1],from[2],from[2],from[3],from[3]} + * Currently, this function is only used for scalar * complex products. + */ +template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet +ploaddup(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet with elements of \a *from quadrupled. + * For instance, for a packet of 8 elements, 2 scalars will be read from \a *from and + * replicated to form: {from[0],from[0],from[0],from[0],from[1],from[1],from[1],from[1]} + * Currently, this function is only used in matrix products. + * For packet-size smaller or equal to 4, this function is equivalent to pload1 + */ +template EIGEN_DEVICE_FUNC inline Packet +ploadquad(const typename unpacket_traits::type* from) +{ return pload1(from); } + +/** \internal equivalent to + * \code + * a0 = pload1(a+0); + * a1 = pload1(a+1); + * a2 = pload1(a+2); + * a3 = pload1(a+3); + * \endcode + * \sa pset1, pload1, ploaddup, pbroadcast2 + */ +template EIGEN_DEVICE_FUNC +inline void pbroadcast4(const typename unpacket_traits::type *a, + Packet& a0, Packet& a1, Packet& a2, Packet& a3) +{ + a0 = pload1(a+0); + a1 = pload1(a+1); + a2 = pload1(a+2); + a3 = pload1(a+3); +} + +/** \internal equivalent to + * \code + * a0 = pload1(a+0); + * a1 = pload1(a+1); + * \endcode + * \sa pset1, pload1, ploaddup, pbroadcast4 + */ +template EIGEN_DEVICE_FUNC +inline void pbroadcast2(const typename unpacket_traits::type *a, + Packet& a0, Packet& a1) +{ + a0 = pload1(a+0); + a1 = pload1(a+1); +} + +/** \internal \brief Returns a packet with coefficients (a,a+1,...,a+packet_size-1). */ +template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet +plset(const typename unpacket_traits::type& a) { return a; } + +/** \internal copy the packet \a from to \a *to, \a to must be 16 bytes aligned */ +template EIGEN_DEVICE_FUNC inline void pstore(Scalar* to, const Packet& from) +{ (*to) = from; } + +/** \internal copy the packet \a from to \a *to, (un-aligned store) */ +template EIGEN_DEVICE_FUNC inline void pstoreu(Scalar* to, const Packet& from) +{ (*to) = from; } + + template EIGEN_DEVICE_FUNC inline Packet pgather(const Scalar* from, Index /*stride*/) + { return ploadu(from); } + + template EIGEN_DEVICE_FUNC inline void pscatter(Scalar* to, const Packet& from, Index /*stride*/) + { pstore(to, from); } + +/** \internal tries to do cache prefetching of \a addr */ +template EIGEN_DEVICE_FUNC inline void prefetch(const Scalar* addr) +{ +#ifdef __CUDA_ARCH__ +#if defined(__LP64__) + // 64-bit pointer operand constraint for inlined asm + asm(" prefetch.L1 [ %1 ];" : "=l"(addr) : "l"(addr)); +#else + // 32-bit pointer operand constraint for inlined asm + asm(" prefetch.L1 [ %1 ];" : "=r"(addr) : "r"(addr)); +#endif +#elif (!EIGEN_COMP_MSVC) && (EIGEN_COMP_GNUC || EIGEN_COMP_CLANG || EIGEN_COMP_ICC) + __builtin_prefetch(addr); +#endif +} + +/** \internal \returns the first element of a packet */ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type pfirst(const Packet& a) +{ return a; } + +/** \internal \returns a packet where the element i contains the sum of the packet of \a vec[i] */ +template EIGEN_DEVICE_FUNC inline Packet +preduxp(const Packet* vecs) { return vecs[0]; } + +/** \internal \returns the sum of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux(const Packet& a) +{ return a; } + +/** \internal \returns the sum of the elements of \a a by block of 4 elements. + * For a packet {a0, a1, a2, a3, a4, a5, a6, a7}, it returns a half packet {a0+a4, a1+a5, a2+a6, a3+a7} + * For packet-size smaller or equal to 4, this boils down to a noop. + */ +template EIGEN_DEVICE_FUNC inline +typename conditional<(unpacket_traits::size%8)==0,typename unpacket_traits::half,Packet>::type +predux_downto4(const Packet& a) +{ return a; } + +/** \internal \returns the product of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_mul(const Packet& a) +{ return a; } + +/** \internal \returns the min of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_min(const Packet& a) +{ return a; } + +/** \internal \returns the max of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_max(const Packet& a) +{ return a; } + +/** \internal \returns the reversed elements of \a a*/ +template EIGEN_DEVICE_FUNC inline Packet preverse(const Packet& a) +{ return a; } + +/** \internal \returns \a a with real and imaginary part flipped (for complex type only) */ +template EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet& a) +{ + return Packet(a.imag(),a.real()); +} + +/************************** +* Special math functions +***************************/ + +/** \internal \returns the sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psin(const Packet& a) { using std::sin; return sin(a); } + +/** \internal \returns the cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pcos(const Packet& a) { using std::cos; return cos(a); } + +/** \internal \returns the tan of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet ptan(const Packet& a) { using std::tan; return tan(a); } + +/** \internal \returns the arc sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pasin(const Packet& a) { using std::asin; return asin(a); } + +/** \internal \returns the arc cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pacos(const Packet& a) { using std::acos; return acos(a); } + +/** \internal \returns the arc tangent of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet patan(const Packet& a) { using std::atan; return atan(a); } + +/** \internal \returns the hyperbolic sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psinh(const Packet& a) { using std::sinh; return sinh(a); } + +/** \internal \returns the hyperbolic cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pcosh(const Packet& a) { using std::cosh; return cosh(a); } + +/** \internal \returns the hyperbolic tan of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet ptanh(const Packet& a) { using std::tanh; return tanh(a); } + +/** \internal \returns the exp of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pexp(const Packet& a) { using std::exp; return exp(a); } + +/** \internal \returns the log of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog(const Packet& a) { using std::log; return log(a); } + +/** \internal \returns the log1p of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog1p(const Packet& a) { return numext::log1p(a); } + +/** \internal \returns the log10 of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog10(const Packet& a) { using std::log10; return log10(a); } + +/** \internal \returns the square-root of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); } + +/** \internal \returns the reciprocal square-root of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet prsqrt(const Packet& a) { + return pdiv(pset1(1), psqrt(a)); +} + +/** \internal \returns the rounded value of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pround(const Packet& a) { using numext::round; return round(a); } + +/** \internal \returns the floor of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pfloor(const Packet& a) { using numext::floor; return floor(a); } + +/** \internal \returns the ceil of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pceil(const Packet& a) { using numext::ceil; return ceil(a); } + +/*************************************************************************** +* The following functions might not have to be overwritten for vectorized types +***************************************************************************/ + +/** \internal copy a packet with constant coeficient \a a (e.g., [a,a,a,a]) to \a *to. \a to must be 16 bytes aligned */ +// NOTE: this function must really be templated on the packet type (think about different packet types for the same scalar type) +template +inline void pstore1(typename unpacket_traits::type* to, const typename unpacket_traits::type& a) +{ + pstore(to, pset1(a)); +} + +/** \internal \returns a * b + c (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmadd(const Packet& a, + const Packet& b, + const Packet& c) +{ return padd(pmul(a, b),c); } + +/** \internal \returns a packet version of \a *from. + * The pointer \a from must be aligned on a \a Alignment bytes boundary. */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt(const typename unpacket_traits::type* from) +{ + if(Alignment >= unpacket_traits::alignment) + return pload(from); + else + return ploadu(from); +} + +/** \internal copy the packet \a from to \a *to. + * The pointer \a from must be aligned on a \a Alignment bytes boundary. */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void pstoret(Scalar* to, const Packet& from) +{ + if(Alignment >= unpacket_traits::alignment) + pstore(to, from); + else + pstoreu(to, from); +} + +/** \internal \returns a packet version of \a *from. + * Unlike ploadt, ploadt_ro takes advantage of the read-only memory path on the + * hardware if available to speedup the loading of data that won't be modified + * by the current computation. + */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt_ro(const typename unpacket_traits::type* from) +{ + return ploadt(from); +} + +/** \internal default implementation of palign() allowing partial specialization */ +template +struct palign_impl +{ + // by default data are aligned, so there is nothing to be done :) + static inline void run(PacketType&, const PacketType&) {} +}; + +/** \internal update \a first using the concatenation of the packet_size minus \a Offset last elements + * of \a first and \a Offset first elements of \a second. + * + * This function is currently only used to optimize matrix-vector products on unligned matrices. + * It takes 2 packets that represent a contiguous memory array, and returns a packet starting + * at the position \a Offset. For instance, for packets of 4 elements, we have: + * Input: + * - first = {f0,f1,f2,f3} + * - second = {s0,s1,s2,s3} + * Output: + * - if Offset==0 then {f0,f1,f2,f3} + * - if Offset==1 then {f1,f2,f3,s0} + * - if Offset==2 then {f2,f3,s0,s1} + * - if Offset==3 then {f3,s0,s1,s3} + */ +template +inline void palign(PacketType& first, const PacketType& second) +{ + palign_impl::run(first,second); +} + +/*************************************************************************** +* Fast complex products (GCC generates a function call which is very slow) +***************************************************************************/ + +// Eigen+CUDA does not support complexes. +#ifndef __CUDACC__ + +template<> inline std::complex pmul(const std::complex& a, const std::complex& b) +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } + +template<> inline std::complex pmul(const std::complex& a, const std::complex& b) +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } + +#endif + + +/*************************************************************************** + * PacketBlock, that is a collection of N packets where the number of words + * in the packet is a multiple of N. +***************************************************************************/ +template ::size> struct PacketBlock { + Packet packet[N]; +}; + +template EIGEN_DEVICE_FUNC inline void +ptranspose(PacketBlock& /*kernel*/) { + // Nothing to do in the scalar case, i.e. a 1x1 matrix. +} + +/*************************************************************************** + * Selector, i.e. vector of N boolean values used to select (i.e. blend) + * words from 2 packets. +***************************************************************************/ +template struct Selector { + bool select[N]; +}; + +template EIGEN_DEVICE_FUNC inline Packet +pblend(const Selector::size>& ifPacket, const Packet& thenPacket, const Packet& elsePacket) { + return ifPacket.select[0] ? thenPacket : elsePacket; +} + +/** \internal \returns \a a with the first coefficient replaced by the scalar b */ +template EIGEN_DEVICE_FUNC inline Packet +pinsertfirst(const Packet& a, typename unpacket_traits::type b) +{ + // Default implementation based on pblend. + // It must be specialized for higher performance. + Selector::size> mask; + mask.select[0] = true; + // This for loop should be optimized away by the compiler. + for(Index i=1; i::size; ++i) + mask.select[i] = false; + return pblend(mask, pset1(b), a); +} + +/** \internal \returns \a a with the last coefficient replaced by the scalar b */ +template EIGEN_DEVICE_FUNC inline Packet +pinsertlast(const Packet& a, typename unpacket_traits::type b) +{ + // Default implementation based on pblend. + // It must be specialized for higher performance. + Selector::size> mask; + // This for loop should be optimized away by the compiler. + for(Index i=0; i::size-1; ++i) + mask.select[i] = false; + mask.select[unpacket_traits::size-1] = true; + return pblend(mask, pset1(b), a); +} + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_GENERIC_PACKET_MATH_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h new file mode 100644 index 00000000000..769dc255c20 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h @@ -0,0 +1,187 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2010-2016 Gael Guennebaud +// Copyright (C) 2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GLOBAL_FUNCTIONS_H +#define EIGEN_GLOBAL_FUNCTIONS_H + +#ifdef EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \ + /** \returns an expression of the coefficient-wise DOC_OP of \a x + + DOC_DETAILS + + \sa Math functions, class CwiseUnaryOp + */ \ + template \ + inline const Eigen::CwiseUnaryOp, const Derived> \ + NAME(const Eigen::ArrayBase& x); + +#else + +#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \ + template \ + inline const Eigen::CwiseUnaryOp, const Derived> \ + (NAME)(const Eigen::ArrayBase& x) { \ + return Eigen::CwiseUnaryOp, const Derived>(x.derived()); \ + } + +#endif // EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(NAME,FUNCTOR) \ + \ + template \ + struct NAME##_retval > \ + { \ + typedef const Eigen::CwiseUnaryOp, const Derived> type; \ + }; \ + template \ + struct NAME##_impl > \ + { \ + static inline typename NAME##_retval >::type run(const Eigen::ArrayBase& x) \ + { \ + return typename NAME##_retval >::type(x.derived()); \ + } \ + }; + +namespace Eigen +{ + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(real,scalar_real_op,real part,\sa ArrayBase::real) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(imag,scalar_imag_op,imaginary part,\sa ArrayBase::imag) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(conj,scalar_conjugate_op,complex conjugate,\sa ArrayBase::conjugate) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(inverse,scalar_inverse_op,inverse,\sa ArrayBase::inverse) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sin,scalar_sin_op,sine,\sa ArrayBase::sin) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cos,scalar_cos_op,cosine,\sa ArrayBase::cos) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tan,scalar_tan_op,tangent,\sa ArrayBase::tan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(atan,scalar_atan_op,arc-tangent,\sa ArrayBase::atan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(asin,scalar_asin_op,arc-sine,\sa ArrayBase::asin) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(acos,scalar_acos_op,arc-consine,\sa ArrayBase::acos) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sinh,scalar_sinh_op,hyperbolic sine,\sa ArrayBase::sinh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cosh,scalar_cosh_op,hyperbolic cosine,\sa ArrayBase::cosh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tanh,scalar_tanh_op,hyperbolic tangent,\sa ArrayBase::tanh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(lgamma,scalar_lgamma_op,natural logarithm of the gamma function,\sa ArrayBase::lgamma) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(digamma,scalar_digamma_op,derivative of lgamma,\sa ArrayBase::digamma) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erf,scalar_erf_op,error function,\sa ArrayBase::erf) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erfc,scalar_erfc_op,complement error function,\sa ArrayBase::erfc) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(exp,scalar_exp_op,exponential,\sa ArrayBase::exp) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log,scalar_log_op,natural logarithm,\sa Eigen::log10 DOXCOMMA ArrayBase::log) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log1p,scalar_log1p_op,natural logarithm of 1 plus the value,\sa ArrayBase::log1p) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log10,scalar_log10_op,base 10 logarithm,\sa Eigen::log DOXCOMMA ArrayBase::log) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs,scalar_abs_op,absolute value,\sa ArrayBase::abs DOXCOMMA MatrixBase::cwiseAbs) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs2,scalar_abs2_op,squared absolute value,\sa ArrayBase::abs2 DOXCOMMA MatrixBase::cwiseAbs2) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(arg,scalar_arg_op,complex argument,\sa ArrayBase::arg) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sqrt,scalar_sqrt_op,square root,\sa ArrayBase::sqrt DOXCOMMA MatrixBase::cwiseSqrt) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(rsqrt,scalar_rsqrt_op,reciprocal square root,\sa ArrayBase::rsqrt) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(square,scalar_square_op,square (power 2),\sa Eigen::abs2 DOXCOMMA Eigen::pow DOXCOMMA ArrayBase::square) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cube,scalar_cube_op,cube (power 3),\sa Eigen::pow DOXCOMMA ArrayBase::cube) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(round,scalar_round_op,nearest integer,\sa Eigen::floor DOXCOMMA Eigen::ceil DOXCOMMA ArrayBase::round) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(floor,scalar_floor_op,nearest integer not greater than the giben value,\sa Eigen::ceil DOXCOMMA ArrayBase::floor) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(ceil,scalar_ceil_op,nearest integer not less than the giben value,\sa Eigen::floor DOXCOMMA ArrayBase::ceil) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isnan,scalar_isnan_op,not-a-number test,\sa Eigen::isinf DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isnan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isinf,scalar_isinf_op,infinite value test,\sa Eigen::isnan DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isinf) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isfinite,scalar_isfinite_op,finite value test,\sa Eigen::isinf DOXCOMMA Eigen::isnan DOXCOMMA ArrayBase::isfinite) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sign,scalar_sign_op,sign (or 0),\sa ArrayBase::sign) + + /** \returns an expression of the coefficient-wise power of \a x to the given constant \a exponent. + * + * \tparam ScalarExponent is the scalar type of \a exponent. It must be compatible with the scalar type of the given expression (\c Derived::Scalar). + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ +#ifdef EIGEN_PARSED_BY_DOXYGEN + template + inline const CwiseBinaryOp,Derived,Constant > + pow(const Eigen::ArrayBase& x, const ScalarExponent& exponent); +#else + template + inline typename internal::enable_if< !(internal::is_same::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,typename Derived::Scalar,ScalarExponent), + const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,ScalarExponent,pow) >::type + pow(const Eigen::ArrayBase& x, const ScalarExponent& exponent) { + return x.derived().pow(exponent); + } + + template + inline const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename Derived::Scalar,pow) + pow(const Eigen::ArrayBase& x, const typename Derived::Scalar& exponent) { + return x.derived().pow(exponent); + } +#endif + + /** \returns an expression of the coefficient-wise power of \a x to the given array of \a exponents. + * + * This function computes the coefficient-wise power. + * + * Example: \include Cwise_array_power_array.cpp + * Output: \verbinclude Cwise_array_power_array.out + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ + template + inline const Eigen::CwiseBinaryOp, const Derived, const ExponentDerived> + pow(const Eigen::ArrayBase& x, const Eigen::ArrayBase& exponents) + { + return Eigen::CwiseBinaryOp, const Derived, const ExponentDerived>( + x.derived(), + exponents.derived() + ); + } + + /** \returns an expression of the coefficient-wise power of the scalar \a x to the given array of \a exponents. + * + * This function computes the coefficient-wise power between a scalar and an array of exponents. + * + * \tparam Scalar is the scalar type of \a x. It must be compatible with the scalar type of the given array expression (\c Derived::Scalar). + * + * Example: \include Cwise_scalar_power_array.cpp + * Output: \verbinclude Cwise_scalar_power_array.out + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ +#ifdef EIGEN_PARSED_BY_DOXYGEN + template + inline const CwiseBinaryOp,Constant,Derived> + pow(const Scalar& x,const Eigen::ArrayBase& x); +#else + template + inline typename internal::enable_if< !(internal::is_same::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,Scalar,typename Derived::Scalar), + const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow) >::type + pow(const Scalar& x, const Eigen::ArrayBase& exponents) + { + return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow)( + typename internal::plain_constant_type::type(exponents.rows(), exponents.cols(), x), exponents.derived() ); + } + + template + inline const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow) + pow(const typename Derived::Scalar& x, const Eigen::ArrayBase& exponents) + { + return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow)( + typename internal::plain_constant_type::type(exponents.rows(), exponents.cols(), x), exponents.derived() ); + } +#endif + + + namespace internal + { + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(real,scalar_real_op) + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(imag,scalar_imag_op) + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(abs2,scalar_abs2_op) + } +} + +// TODO: cleanly disable those functions that are not supported on Array (numext::real_ref, internal::random, internal::isApprox...) + +#endif // EIGEN_GLOBAL_FUNCTIONS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/IO.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/IO.h new file mode 100644 index 00000000000..da7fd6cce2e --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/IO.h @@ -0,0 +1,225 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_IO_H +#define EIGEN_IO_H + +namespace Eigen { + +enum { DontAlignCols = 1 }; +enum { StreamPrecision = -1, + FullPrecision = -2 }; + +namespace internal { +template +std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat& fmt); +} + +/** \class IOFormat + * \ingroup Core_Module + * + * \brief Stores a set of parameters controlling the way matrices are printed + * + * List of available parameters: + * - \b precision number of digits for floating point values, or one of the special constants \c StreamPrecision and \c FullPrecision. + * The default is the special value \c StreamPrecision which means to use the + * stream's own precision setting, as set for instance using \c cout.precision(3). The other special value + * \c FullPrecision means that the number of digits will be computed to match the full precision of each floating-point + * type. + * - \b flags an OR-ed combination of flags, the default value is 0, the only currently available flag is \c DontAlignCols which + * allows to disable the alignment of columns, resulting in faster code. + * - \b coeffSeparator string printed between two coefficients of the same row + * - \b rowSeparator string printed between two rows + * - \b rowPrefix string printed at the beginning of each row + * - \b rowSuffix string printed at the end of each row + * - \b matPrefix string printed at the beginning of the matrix + * - \b matSuffix string printed at the end of the matrix + * + * Example: \include IOFormat.cpp + * Output: \verbinclude IOFormat.out + * + * \sa DenseBase::format(), class WithFormat + */ +struct IOFormat +{ + /** Default constructor, see class IOFormat for the meaning of the parameters */ + IOFormat(int _precision = StreamPrecision, int _flags = 0, + const std::string& _coeffSeparator = " ", + const std::string& _rowSeparator = "\n", const std::string& _rowPrefix="", const std::string& _rowSuffix="", + const std::string& _matPrefix="", const std::string& _matSuffix="") + : matPrefix(_matPrefix), matSuffix(_matSuffix), rowPrefix(_rowPrefix), rowSuffix(_rowSuffix), rowSeparator(_rowSeparator), + rowSpacer(""), coeffSeparator(_coeffSeparator), precision(_precision), flags(_flags) + { + // TODO check if rowPrefix, rowSuffix or rowSeparator contains a newline + // don't add rowSpacer if columns are not to be aligned + if((flags & DontAlignCols)) + return; + int i = int(matSuffix.length())-1; + while (i>=0 && matSuffix[i]!='\n') + { + rowSpacer += ' '; + i--; + } + } + std::string matPrefix, matSuffix; + std::string rowPrefix, rowSuffix, rowSeparator, rowSpacer; + std::string coeffSeparator; + int precision; + int flags; +}; + +/** \class WithFormat + * \ingroup Core_Module + * + * \brief Pseudo expression providing matrix output with given format + * + * \tparam ExpressionType the type of the object on which IO stream operations are performed + * + * This class represents an expression with stream operators controlled by a given IOFormat. + * It is the return type of DenseBase::format() + * and most of the time this is the only way it is used. + * + * See class IOFormat for some examples. + * + * \sa DenseBase::format(), class IOFormat + */ +template +class WithFormat +{ + public: + + WithFormat(const ExpressionType& matrix, const IOFormat& format) + : m_matrix(matrix), m_format(format) + {} + + friend std::ostream & operator << (std::ostream & s, const WithFormat& wf) + { + return internal::print_matrix(s, wf.m_matrix.eval(), wf.m_format); + } + + protected: + typename ExpressionType::Nested m_matrix; + IOFormat m_format; +}; + +namespace internal { + +// NOTE: This helper is kept for backward compatibility with previous code specializing +// this internal::significant_decimals_impl structure. In the future we should directly +// call digits10() which has been introduced in July 2016 in 3.3. +template +struct significant_decimals_impl +{ + static inline int run() + { + return NumTraits::digits10(); + } +}; + +/** \internal + * print the matrix \a _m to the output stream \a s using the output format \a fmt */ +template +std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat& fmt) +{ + if(_m.size() == 0) + { + s << fmt.matPrefix << fmt.matSuffix; + return s; + } + + typename Derived::Nested m = _m; + typedef typename Derived::Scalar Scalar; + + Index width = 0; + + std::streamsize explicit_precision; + if(fmt.precision == StreamPrecision) + { + explicit_precision = 0; + } + else if(fmt.precision == FullPrecision) + { + if (NumTraits::IsInteger) + { + explicit_precision = 0; + } + else + { + explicit_precision = significant_decimals_impl::run(); + } + } + else + { + explicit_precision = fmt.precision; + } + + std::streamsize old_precision = 0; + if(explicit_precision) old_precision = s.precision(explicit_precision); + + bool align_cols = !(fmt.flags & DontAlignCols); + if(align_cols) + { + // compute the largest width + for(Index j = 0; j < m.cols(); ++j) + for(Index i = 0; i < m.rows(); ++i) + { + std::stringstream sstr; + sstr.copyfmt(s); + sstr << m.coeff(i,j); + width = std::max(width, Index(sstr.str().length())); + } + } + s << fmt.matPrefix; + for(Index i = 0; i < m.rows(); ++i) + { + if (i) + s << fmt.rowSpacer; + s << fmt.rowPrefix; + if(width) s.width(width); + s << m.coeff(i, 0); + for(Index j = 1; j < m.cols(); ++j) + { + s << fmt.coeffSeparator; + if (width) s.width(width); + s << m.coeff(i, j); + } + s << fmt.rowSuffix; + if( i < m.rows() - 1) + s << fmt.rowSeparator; + } + s << fmt.matSuffix; + if(explicit_precision) s.precision(old_precision); + return s; +} + +} // end namespace internal + +/** \relates DenseBase + * + * Outputs the matrix, to the given stream. + * + * If you wish to print the matrix with a format different than the default, use DenseBase::format(). + * + * It is also possible to change the default format by defining EIGEN_DEFAULT_IO_FORMAT before including Eigen headers. + * If not defined, this will automatically be defined to Eigen::IOFormat(), that is the Eigen::IOFormat with default parameters. + * + * \sa DenseBase::format() + */ +template +std::ostream & operator << +(std::ostream & s, + const DenseBase & m) +{ + return internal::print_matrix(s, m.eval(), EIGEN_DEFAULT_IO_FORMAT); +} + +} // end namespace Eigen + +#endif // EIGEN_IO_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h new file mode 100644 index 00000000000..b76f0439d80 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h @@ -0,0 +1,118 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2014 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_INVERSE_H +#define EIGEN_INVERSE_H + +namespace Eigen { + +template class InverseImpl; + +namespace internal { + +template +struct traits > + : traits +{ + typedef typename XprType::PlainObject PlainObject; + typedef traits BaseTraits; + enum { + Flags = BaseTraits::Flags & RowMajorBit + }; +}; + +} // end namespace internal + +/** \class Inverse + * + * \brief Expression of the inverse of another expression + * + * \tparam XprType the type of the expression we are taking the inverse + * + * This class represents an abstract expression of A.inverse() + * and most of the time this is the only way it is used. + * + */ +template +class Inverse : public InverseImpl::StorageKind> +{ +public: + typedef typename XprType::StorageIndex StorageIndex; + typedef typename XprType::PlainObject PlainObject; + typedef typename XprType::Scalar Scalar; + typedef typename internal::ref_selector::type XprTypeNested; + typedef typename internal::remove_all::type XprTypeNestedCleaned; + typedef typename internal::ref_selector::type Nested; + typedef typename internal::remove_all::type NestedExpression; + + explicit EIGEN_DEVICE_FUNC Inverse(const XprType &xpr) + : m_xpr(xpr) + {} + + EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } + + EIGEN_DEVICE_FUNC const XprTypeNestedCleaned& nestedExpression() const { return m_xpr; } + +protected: + XprTypeNested m_xpr; +}; + +// Generic API dispatcher +template +class InverseImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; + typedef typename XprType::Scalar Scalar; +private: + + Scalar coeff(Index row, Index col) const; + Scalar coeff(Index i) const; +}; + +namespace internal { + +/** \internal + * \brief Default evaluator for Inverse expression. + * + * This default evaluator for Inverse expression simply evaluate the inverse into a temporary + * by a call to internal::call_assignment_no_alias. + * Therefore, inverse implementers only have to specialize Assignment, ...> for + * there own nested expression. + * + * \sa class Inverse + */ +template +struct unary_evaluator > + : public evaluator::PlainObject> +{ + typedef Inverse InverseType; + typedef typename InverseType::PlainObject PlainObject; + typedef evaluator Base; + + enum { Flags = Base::Flags | EvalBeforeNestingBit }; + + unary_evaluator(const InverseType& inv_xpr) + : m_result(inv_xpr.rows(), inv_xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + internal::call_assignment_no_alias(m_result, inv_xpr); + } + +protected: + PlainObject m_result; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_INVERSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Map.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Map.h new file mode 100644 index 00000000000..548bf9a2d55 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Map.h @@ -0,0 +1,171 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MAP_H +#define EIGEN_MAP_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : public traits +{ + typedef traits TraitsBase; + enum { + PlainObjectTypeInnerSize = ((traits::Flags&RowMajorBit)==RowMajorBit) + ? PlainObjectType::ColsAtCompileTime + : PlainObjectType::RowsAtCompileTime, + + InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0 + ? int(PlainObjectType::InnerStrideAtCompileTime) + : int(StrideType::InnerStrideAtCompileTime), + OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0 + ? (InnerStrideAtCompileTime==Dynamic || PlainObjectTypeInnerSize==Dynamic + ? Dynamic + : int(InnerStrideAtCompileTime) * int(PlainObjectTypeInnerSize)) + : int(StrideType::OuterStrideAtCompileTime), + Alignment = int(MapOptions)&int(AlignedMask), + Flags0 = TraitsBase::Flags & (~NestByRefBit), + Flags = is_lvalue::value ? int(Flags0) : (int(Flags0) & ~LvalueBit) + }; +private: + enum { Options }; // Expressions don't have Options +}; +} + +/** \class Map + * \ingroup Core_Module + * + * \brief A matrix or vector expression mapping an existing array of data. + * + * \tparam PlainObjectType the equivalent matrix type of the mapped data + * \tparam MapOptions specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned. + * The default is \c #Unaligned. + * \tparam StrideType optionally specifies strides. By default, Map assumes the memory layout + * of an ordinary, contiguous array. This can be overridden by specifying strides. + * The type passed here must be a specialization of the Stride template, see examples below. + * + * This class represents a matrix or vector expression mapping an existing array of data. + * It can be used to let Eigen interface without any overhead with non-Eigen data structures, + * such as plain C arrays or structures from other libraries. By default, it assumes that the + * data is laid out contiguously in memory. You can however override this by explicitly specifying + * inner and outer strides. + * + * Here's an example of simply mapping a contiguous array as a \ref TopicStorageOrders "column-major" matrix: + * \include Map_simple.cpp + * Output: \verbinclude Map_simple.out + * + * If you need to map non-contiguous arrays, you can do so by specifying strides: + * + * Here's an example of mapping an array as a vector, specifying an inner stride, that is, the pointer + * increment between two consecutive coefficients. Here, we're specifying the inner stride as a compile-time + * fixed value. + * \include Map_inner_stride.cpp + * Output: \verbinclude Map_inner_stride.out + * + * Here's an example of mapping an array while specifying an outer stride. Here, since we're mapping + * as a column-major matrix, 'outer stride' means the pointer increment between two consecutive columns. + * Here, we're specifying the outer stride as a runtime parameter. Note that here \c OuterStride<> is + * a short version of \c OuterStride because the default template parameter of OuterStride + * is \c Dynamic + * \include Map_outer_stride.cpp + * Output: \verbinclude Map_outer_stride.out + * + * For more details and for an example of specifying both an inner and an outer stride, see class Stride. + * + * \b Tip: to change the array of data mapped by a Map object, you can use the C++ + * placement new syntax: + * + * Example: \include Map_placement_new.cpp + * Output: \verbinclude Map_placement_new.out + * + * This class is the return type of PlainObjectBase::Map() but can also be used directly. + * + * \sa PlainObjectBase::Map(), \ref TopicStorageOrders + */ +template class Map + : public MapBase > +{ + public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Map) + + typedef typename Base::PointerType PointerType; + typedef PointerType PointerArgType; + EIGEN_DEVICE_FUNC + inline PointerType cast_to_pointer_type(PointerArgType ptr) { return ptr; } + + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1; + } + + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return int(StrideType::OuterStrideAtCompileTime) != 0 ? m_stride.outer() + : int(internal::traits::OuterStrideAtCompileTime) != Dynamic ? Index(internal::traits::OuterStrideAtCompileTime) + : IsVectorAtCompileTime ? (this->size() * innerStride()) + : (int(Flags)&RowMajorBit) ? (this->cols() * innerStride()) + : (this->rows() * innerStride()); + } + + /** Constructor in the fixed-size case. + * + * \param dataPtr pointer to the array to map + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + explicit inline Map(PointerArgType dataPtr, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr)), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + /** Constructor in the dynamic-size vector case. + * + * \param dataPtr pointer to the array to map + * \param size the size of the vector expression + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + inline Map(PointerArgType dataPtr, Index size, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr), size), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + /** Constructor in the dynamic-size matrix case. + * + * \param dataPtr pointer to the array to map + * \param rows the number of rows of the matrix expression + * \param cols the number of columns of the matrix expression + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + inline Map(PointerArgType dataPtr, Index rows, Index cols, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr), rows, cols), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Map) + + protected: + StrideType m_stride; +}; + + +} // end namespace Eigen + +#endif // EIGEN_MAP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h new file mode 100644 index 00000000000..92c3b2818a4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h @@ -0,0 +1,308 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MAPBASE_H +#define EIGEN_MAPBASE_H + +#define EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) \ + EIGEN_STATIC_ASSERT((int(internal::evaluator::Flags) & LinearAccessBit) || Derived::IsVectorAtCompileTime, \ + YOU_ARE_TRYING_TO_USE_AN_INDEX_BASED_ACCESSOR_ON_AN_EXPRESSION_THAT_DOES_NOT_SUPPORT_THAT) + +namespace Eigen { + +/** \ingroup Core_Module + * + * \brief Base class for dense Map and Block expression with direct access + * + * This base class provides the const low-level accessors (e.g. coeff, coeffRef) of dense + * Map and Block objects with direct access. + * Typical users do not have to directly deal with this class. + * + * This class can be extended by through the macro plugin \c EIGEN_MAPBASE_PLUGIN. + * See \link TopicCustomizing_Plugins customizing Eigen \endlink for details. + * + * The \c Derived class has to provide the following two methods describing the memory layout: + * \code Index innerStride() const; \endcode + * \code Index outerStride() const; \endcode + * + * \sa class Map, class Block + */ +template class MapBase + : public internal::dense_xpr_base::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + enum { + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + InnerStrideAtCompileTime = internal::traits::InnerStrideAtCompileTime, + SizeAtCompileTime = Base::SizeAtCompileTime + }; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + typedef typename internal::conditional< + bool(internal::is_lvalue::value), + Scalar *, + const Scalar *>::type + PointerType; + + using Base::derived; +// using Base::RowsAtCompileTime; +// using Base::ColsAtCompileTime; +// using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + using Base::IsRowMajor; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::eval; + + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + + // bug 217 - compile error on ICC 11.1 + using Base::operator=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + + /** \copydoc DenseBase::rows() */ + EIGEN_DEVICE_FUNC inline Index rows() const { return m_rows.value(); } + /** \copydoc DenseBase::cols() */ + EIGEN_DEVICE_FUNC inline Index cols() const { return m_cols.value(); } + + /** Returns a pointer to the first coefficient of the matrix or vector. + * + * \note When addressing this data, make sure to honor the strides returned by innerStride() and outerStride(). + * + * \sa innerStride(), outerStride() + */ + EIGEN_DEVICE_FUNC inline const Scalar* data() const { return m_data; } + + /** \copydoc PlainObjectBase::coeff(Index,Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeff(Index rowId, Index colId) const + { + return m_data[colId * colStride() + rowId * rowStride()]; + } + + /** \copydoc PlainObjectBase::coeff(Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeff(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return m_data[index * innerStride()]; + } + + /** \copydoc PlainObjectBase::coeffRef(Index,Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return this->m_data[colId * colStride() + rowId * rowStride()]; + } + + /** \copydoc PlainObjectBase::coeffRef(Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return this->m_data[index * innerStride()]; + } + + /** \internal */ + template + inline PacketScalar packet(Index rowId, Index colId) const + { + return internal::ploadt + (m_data + (colId * colStride() + rowId * rowStride())); + } + + /** \internal */ + template + inline PacketScalar packet(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return internal::ploadt(m_data + index * innerStride()); + } + + /** \internal Constructor for fixed size matrices or vectors */ + EIGEN_DEVICE_FUNC + explicit inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime) + { + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + checkSanity(); + } + + /** \internal Constructor for dynamically sized vectors */ + EIGEN_DEVICE_FUNC + inline MapBase(PointerType dataPtr, Index vecSize) + : m_data(dataPtr), + m_rows(RowsAtCompileTime == Dynamic ? vecSize : Index(RowsAtCompileTime)), + m_cols(ColsAtCompileTime == Dynamic ? vecSize : Index(ColsAtCompileTime)) + { + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + eigen_assert(vecSize >= 0); + eigen_assert(dataPtr == 0 || SizeAtCompileTime == Dynamic || SizeAtCompileTime == vecSize); + checkSanity(); + } + + /** \internal Constructor for dynamically sized matrices */ + EIGEN_DEVICE_FUNC + inline MapBase(PointerType dataPtr, Index rows, Index cols) + : m_data(dataPtr), m_rows(rows), m_cols(cols) + { + eigen_assert( (dataPtr == 0) + || ( rows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows) + && cols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols))); + checkSanity(); + } + + #ifdef EIGEN_MAPBASE_PLUGIN + #include EIGEN_MAPBASE_PLUGIN + #endif + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) + + template + EIGEN_DEVICE_FUNC + void checkSanity(typename internal::enable_if<(internal::traits::Alignment>0),void*>::type = 0) const + { +#if EIGEN_MAX_ALIGN_BYTES>0 + // innerStride() is not set yet when this function is called, so we optimistically assume the lowest plausible value: + const Index minInnerStride = InnerStrideAtCompileTime == Dynamic ? 1 : Index(InnerStrideAtCompileTime); + EIGEN_ONLY_USED_FOR_DEBUG(minInnerStride); + eigen_assert(( ((internal::UIntPtr(m_data) % internal::traits::Alignment) == 0) + || (cols() * rows() * minInnerStride * sizeof(Scalar)) < internal::traits::Alignment ) && "data is not aligned"); +#endif + } + + template + EIGEN_DEVICE_FUNC + void checkSanity(typename internal::enable_if::Alignment==0,void*>::type = 0) const + {} + + PointerType m_data; + const internal::variable_if_dynamic m_rows; + const internal::variable_if_dynamic m_cols; +}; + +/** \ingroup Core_Module + * + * \brief Base class for non-const dense Map and Block expression with direct access + * + * This base class provides the non-const low-level accessors (e.g. coeff and coeffRef) of + * dense Map and Block objects with direct access. + * It inherits MapBase which defines the const variant for reading specific entries. + * + * \sa class Map, class Block + */ +template class MapBase + : public MapBase +{ + typedef MapBase ReadOnlyMapBase; + public: + + typedef MapBase Base; + + typedef typename Base::Scalar Scalar; + typedef typename Base::PacketScalar PacketScalar; + typedef typename Base::StorageIndex StorageIndex; + typedef typename Base::PointerType PointerType; + + using Base::derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return this->m_data; } + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return this->m_data; } // no const-cast here so non-const-correct code will give a compile error + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue& coeffRef(Index row, Index col) + { + return this->m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue& coeffRef(Index index) + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return this->m_data[index * innerStride()]; + } + + template + inline void writePacket(Index row, Index col, const PacketScalar& val) + { + internal::pstoret + (this->m_data + (col * colStride() + row * rowStride()), val); + } + + template + inline void writePacket(Index index, const PacketScalar& val) + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + internal::pstoret + (this->m_data + index * innerStride(), val); + } + + EIGEN_DEVICE_FUNC explicit inline MapBase(PointerType dataPtr) : Base(dataPtr) {} + EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index vecSize) : Base(dataPtr, vecSize) {} + EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index rows, Index cols) : Base(dataPtr, rows, cols) {} + + EIGEN_DEVICE_FUNC + Derived& operator=(const MapBase& other) + { + ReadOnlyMapBase::Base::operator=(other); + return derived(); + } + + // In theory we could simply refer to Base:Base::operator=, but MSVC does not like Base::Base, + // see bugs 821 and 920. + using ReadOnlyMapBase::Base::operator=; + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) +}; + +#undef EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS + +} // end namespace Eigen + +#endif // EIGEN_MAPBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h new file mode 100644 index 00000000000..01736c2a060 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h @@ -0,0 +1,1421 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATHFUNCTIONS_H +#define EIGEN_MATHFUNCTIONS_H + +// source: http://www.geom.uiuc.edu/~huberty/math5337/groupe/digits.html +// TODO this should better be moved to NumTraits +#define EIGEN_PI 3.141592653589793238462643383279502884197169399375105820974944592307816406L + + +namespace Eigen { + +// On WINCE, std::abs is defined for int only, so let's defined our own overloads: +// This issue has been confirmed with MSVC 2008 only, but the issue might exist for more recent versions too. +#if EIGEN_OS_WINCE && EIGEN_COMP_MSVC && EIGEN_COMP_MSVC<=1500 +long abs(long x) { return (labs(x)); } +double abs(double x) { return (fabs(x)); } +float abs(float x) { return (fabsf(x)); } +long double abs(long double x) { return (fabsl(x)); } +#endif + +namespace internal { + +/** \internal \class global_math_functions_filtering_base + * + * What it does: + * Defines a typedef 'type' as follows: + * - if type T has a member typedef Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl, then + * global_math_functions_filtering_base::type is a typedef for it. + * - otherwise, global_math_functions_filtering_base::type is a typedef for T. + * + * How it's used: + * To allow to defined the global math functions (like sin...) in certain cases, like the Array expressions. + * When you do sin(array1+array2), the object array1+array2 has a complicated expression type, all what you want to know + * is that it inherits ArrayBase. So we implement a partial specialization of sin_impl for ArrayBase. + * So we must make sure to use sin_impl > and not sin_impl, otherwise our partial specialization + * won't be used. How does sin know that? That's exactly what global_math_functions_filtering_base tells it. + * + * How it's implemented: + * SFINAE in the style of enable_if. Highly susceptible of breaking compilers. With GCC, it sure does work, but if you replace + * the typename dummy by an integer template parameter, it doesn't work anymore! + */ + +template +struct global_math_functions_filtering_base +{ + typedef T type; +}; + +template struct always_void { typedef void type; }; + +template +struct global_math_functions_filtering_base + ::type + > +{ + typedef typename T::Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl type; +}; + +#define EIGEN_MATHFUNC_IMPL(func, scalar) Eigen::internal::func##_impl::type> +#define EIGEN_MATHFUNC_RETVAL(func, scalar) typename Eigen::internal::func##_retval::type>::type + +/**************************************************************************** +* Implementation of real * +****************************************************************************/ + +template::IsComplex> +struct real_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x; + } +}; + +template +struct real_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + using std::real; + return real(x); + } +}; + +template struct real_impl : real_default_impl {}; + +#ifdef __CUDA_ARCH__ +template +struct real_impl > +{ + typedef T RealScalar; + EIGEN_DEVICE_FUNC + static inline T run(const std::complex& x) + { + return x.real(); + } +}; +#endif + +template +struct real_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of imag * +****************************************************************************/ + +template::IsComplex> +struct imag_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar&) + { + return RealScalar(0); + } +}; + +template +struct imag_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + using std::imag; + return imag(x); + } +}; + +template struct imag_impl : imag_default_impl {}; + +#ifdef __CUDA_ARCH__ +template +struct imag_impl > +{ + typedef T RealScalar; + EIGEN_DEVICE_FUNC + static inline T run(const std::complex& x) + { + return x.imag(); + } +}; +#endif + +template +struct imag_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of real_ref * +****************************************************************************/ + +template +struct real_ref_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar& run(Scalar& x) + { + return reinterpret_cast(&x)[0]; + } + EIGEN_DEVICE_FUNC + static inline const RealScalar& run(const Scalar& x) + { + return reinterpret_cast(&x)[0]; + } +}; + +template +struct real_ref_retval +{ + typedef typename NumTraits::Real & type; +}; + +/**************************************************************************** +* Implementation of imag_ref * +****************************************************************************/ + +template +struct imag_ref_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar& run(Scalar& x) + { + return reinterpret_cast(&x)[1]; + } + EIGEN_DEVICE_FUNC + static inline const RealScalar& run(const Scalar& x) + { + return reinterpret_cast(&x)[1]; + } +}; + +template +struct imag_ref_default_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(Scalar&) + { + return Scalar(0); + } + EIGEN_DEVICE_FUNC + static inline const Scalar run(const Scalar&) + { + return Scalar(0); + } +}; + +template +struct imag_ref_impl : imag_ref_default_impl::IsComplex> {}; + +template +struct imag_ref_retval +{ + typedef typename NumTraits::Real & type; +}; + +/**************************************************************************** +* Implementation of conj * +****************************************************************************/ + +template::IsComplex> +struct conj_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + return x; + } +}; + +template +struct conj_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + using std::conj; + return conj(x); + } +}; + +template +struct conj_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of abs2 * +****************************************************************************/ + +template +struct abs2_impl_default +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x*x; + } +}; + +template +struct abs2_impl_default // IsComplex +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x.real()*x.real() + x.imag()*x.imag(); + } +}; + +template +struct abs2_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return abs2_impl_default::IsComplex>::run(x); + } +}; + +template +struct abs2_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of norm1 * +****************************************************************************/ + +template +struct norm1_default_impl; + +template +struct norm1_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(abs); + return abs(x.real()) + abs(x.imag()); + } +}; + +template +struct norm1_default_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(abs); + return abs(x); + } +}; + +template +struct norm1_impl : norm1_default_impl::IsComplex> {}; + +template +struct norm1_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of hypot * +****************************************************************************/ + +template struct hypot_impl; + +template +struct hypot_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of cast * +****************************************************************************/ + +template +struct cast_impl +{ + EIGEN_DEVICE_FUNC + static inline NewType run(const OldType& x) + { + return static_cast(x); + } +}; + +// here, for once, we're plainly returning NewType: we don't want cast to do weird things. + +template +EIGEN_DEVICE_FUNC +inline NewType cast(const OldType& x) +{ + return cast_impl::run(x); +} + +/**************************************************************************** +* Implementation of round * +****************************************************************************/ + +#if EIGEN_HAS_CXX11_MATH + template + struct round_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT((!NumTraits::IsComplex), NUMERIC_TYPE_MUST_BE_REAL) + using std::round; + return round(x); + } + }; +#else + template + struct round_impl + { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT((!NumTraits::IsComplex), NUMERIC_TYPE_MUST_BE_REAL) + EIGEN_USING_STD_MATH(floor); + EIGEN_USING_STD_MATH(ceil); + return (x > Scalar(0)) ? floor(x + Scalar(0.5)) : ceil(x - Scalar(0.5)); + } + }; +#endif + +template +struct round_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of arg * +****************************************************************************/ + +#if EIGEN_HAS_CXX11_MATH + template + struct arg_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(arg); + return arg(x); + } + }; +#else + template::IsComplex> + struct arg_default_impl + { + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return (x < Scalar(0)) ? Scalar(EIGEN_PI) : Scalar(0); } + }; + + template + struct arg_default_impl + { + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(arg); + return arg(x); + } + }; + + template struct arg_impl : arg_default_impl {}; +#endif + +template +struct arg_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of log1p * +****************************************************************************/ + +namespace std_fallback { + // fallback log1p implementation in case there is no log1p(Scalar) function in namespace of Scalar, + // or that there is no suitable std::log1p function available + template + EIGEN_DEVICE_FUNC inline Scalar log1p(const Scalar& x) { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar) + typedef typename NumTraits::Real RealScalar; + EIGEN_USING_STD_MATH(log); + Scalar x1p = RealScalar(1) + x; + return numext::equal_strict(x1p, Scalar(1)) ? x : x * ( log(x1p) / (x1p - RealScalar(1)) ); + } +} + +template +struct log1p_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar) + #if EIGEN_HAS_CXX11_MATH + using std::log1p; + #endif + using std_fallback::log1p; + return log1p(x); + } +}; + + +template +struct log1p_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of pow * +****************************************************************************/ + +template::IsInteger&&NumTraits::IsInteger> +struct pow_impl +{ + //typedef Scalar retval; + typedef typename ScalarBinaryOpTraits >::ReturnType result_type; + static EIGEN_DEVICE_FUNC inline result_type run(const ScalarX& x, const ScalarY& y) + { + EIGEN_USING_STD_MATH(pow); + return pow(x, y); + } +}; + +template +struct pow_impl +{ + typedef ScalarX result_type; + static EIGEN_DEVICE_FUNC inline ScalarX run(ScalarX x, ScalarY y) + { + ScalarX res(1); + eigen_assert(!NumTraits::IsSigned || y >= 0); + if(y & 1) res *= x; + y >>= 1; + while(y) + { + x *= x; + if(y&1) res *= x; + y >>= 1; + } + return res; + } +}; + +/**************************************************************************** +* Implementation of random * +****************************************************************************/ + +template +struct random_default_impl {}; + +template +struct random_impl : random_default_impl::IsComplex, NumTraits::IsInteger> {}; + +template +struct random_retval +{ + typedef Scalar type; +}; + +template inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(const Scalar& x, const Scalar& y); +template inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(); + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + return x + (y-x) * Scalar(std::rand()) / Scalar(RAND_MAX); + } + static inline Scalar run() + { + return run(Scalar(NumTraits::IsSigned ? -1 : 0), Scalar(1)); + } +}; + +enum { + meta_floor_log2_terminate, + meta_floor_log2_move_up, + meta_floor_log2_move_down, + meta_floor_log2_bogus +}; + +template struct meta_floor_log2_selector +{ + enum { middle = (lower + upper) / 2, + value = (upper <= lower + 1) ? int(meta_floor_log2_terminate) + : (n < (1 << middle)) ? int(meta_floor_log2_move_down) + : (n==0) ? int(meta_floor_log2_bogus) + : int(meta_floor_log2_move_up) + }; +}; + +template::value> +struct meta_floor_log2 {}; + +template +struct meta_floor_log2 +{ + enum { value = meta_floor_log2::middle>::value }; +}; + +template +struct meta_floor_log2 +{ + enum { value = meta_floor_log2::middle, upper>::value }; +}; + +template +struct meta_floor_log2 +{ + enum { value = (n >= ((unsigned int)(1) << (lower+1))) ? lower+1 : lower }; +}; + +template +struct meta_floor_log2 +{ + // no value, error at compile time +}; + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + if (y <= x) + return x; + // ScalarU is the unsigned counterpart of Scalar, possibly Scalar itself. + typedef typename make_unsigned::type ScalarU; + // ScalarX is the widest of ScalarU and unsigned int. + // We'll deal only with ScalarX and unsigned int below thus avoiding signed + // types and arithmetic and signed overflows (which are undefined behavior). + typedef typename conditional<(ScalarU(-1) > unsigned(-1)), ScalarU, unsigned>::type ScalarX; + // The following difference doesn't overflow, provided our integer types are two's + // complement and have the same number of padding bits in signed and unsigned variants. + // This is the case in most modern implementations of C++. + ScalarX range = ScalarX(y) - ScalarX(x); + ScalarX offset = 0; + ScalarX divisor = 1; + ScalarX multiplier = 1; + const unsigned rand_max = RAND_MAX; + if (range <= rand_max) divisor = (rand_max + 1) / (range + 1); + else multiplier = 1 + range / (rand_max + 1); + // Rejection sampling. + do { + offset = (unsigned(std::rand()) * multiplier) / divisor; + } while (offset > range); + return Scalar(ScalarX(x) + offset); + } + + static inline Scalar run() + { +#ifdef EIGEN_MAKING_DOCS + return run(Scalar(NumTraits::IsSigned ? -10 : 0), Scalar(10)); +#else + enum { rand_bits = meta_floor_log2<(unsigned int)(RAND_MAX)+1>::value, + scalar_bits = sizeof(Scalar) * CHAR_BIT, + shift = EIGEN_PLAIN_ENUM_MAX(0, int(rand_bits) - int(scalar_bits)), + offset = NumTraits::IsSigned ? (1 << (EIGEN_PLAIN_ENUM_MIN(rand_bits,scalar_bits)-1)) : 0 + }; + return Scalar((std::rand() >> shift) - offset); +#endif + } +}; + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + return Scalar(random(x.real(), y.real()), + random(x.imag(), y.imag())); + } + static inline Scalar run() + { + typedef typename NumTraits::Real RealScalar; + return Scalar(random(), random()); + } +}; + +template +inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(const Scalar& x, const Scalar& y) +{ + return EIGEN_MATHFUNC_IMPL(random, Scalar)::run(x, y); +} + +template +inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random() +{ + return EIGEN_MATHFUNC_IMPL(random, Scalar)::run(); +} + +// Implementatin of is* functions + +// std::is* do not work with fast-math and gcc, std::is* are available on MSVC 2013 and newer, as well as in clang. +#if (EIGEN_HAS_CXX11_MATH && !(EIGEN_COMP_GNUC_STRICT && __FINITE_MATH_ONLY__)) || (EIGEN_COMP_MSVC>=1800) || (EIGEN_COMP_CLANG) +#define EIGEN_USE_STD_FPCLASSIFY 1 +#else +#define EIGEN_USE_STD_FPCLASSIFY 0 +#endif + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isnan_impl(const T&) { return false; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isinf_impl(const T&) { return false; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isfinite_impl(const T&) { return true; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isfinite_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isfinite)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isfinite; + return isfinite EIGEN_NOT_A_MACRO (x); + #else + return x<=NumTraits::highest() && x>=NumTraits::lowest(); + #endif +} + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isinf_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isinf)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isinf; + return isinf EIGEN_NOT_A_MACRO (x); + #else + return x>NumTraits::highest() || x::lowest(); + #endif +} + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isnan_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isnan)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isnan; + return isnan EIGEN_NOT_A_MACRO (x); + #else + return x != x; + #endif +} + +#if (!EIGEN_USE_STD_FPCLASSIFY) + +#if EIGEN_COMP_MSVC + +template EIGEN_DEVICE_FUNC bool isinf_msvc_helper(T x) +{ + return _fpclass(x)==_FPCLASS_NINF || _fpclass(x)==_FPCLASS_PINF; +} + +//MSVC defines a _isnan builtin function, but for double only +EIGEN_DEVICE_FUNC inline bool isnan_impl(const long double& x) { return _isnan(x)!=0; } +EIGEN_DEVICE_FUNC inline bool isnan_impl(const double& x) { return _isnan(x)!=0; } +EIGEN_DEVICE_FUNC inline bool isnan_impl(const float& x) { return _isnan(x)!=0; } + +EIGEN_DEVICE_FUNC inline bool isinf_impl(const long double& x) { return isinf_msvc_helper(x); } +EIGEN_DEVICE_FUNC inline bool isinf_impl(const double& x) { return isinf_msvc_helper(x); } +EIGEN_DEVICE_FUNC inline bool isinf_impl(const float& x) { return isinf_msvc_helper(x); } + +#elif (defined __FINITE_MATH_ONLY__ && __FINITE_MATH_ONLY__ && EIGEN_COMP_GNUC) + +#if EIGEN_GNUC_AT_LEAST(5,0) + #define EIGEN_TMP_NOOPT_ATTRIB EIGEN_DEVICE_FUNC inline __attribute__((optimize("no-finite-math-only"))) +#else + // NOTE the inline qualifier and noinline attribute are both needed: the former is to avoid linking issue (duplicate symbol), + // while the second prevent too aggressive optimizations in fast-math mode: + #define EIGEN_TMP_NOOPT_ATTRIB EIGEN_DEVICE_FUNC inline __attribute__((noinline,optimize("no-finite-math-only"))) +#endif + +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const long double& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const double& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const float& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const double& x) { return __builtin_isinf(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const float& x) { return __builtin_isinf(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const long double& x) { return __builtin_isinf(x); } + +#undef EIGEN_TMP_NOOPT_ATTRIB + +#endif + +#endif + +// The following overload are defined at the end of this file +template EIGEN_DEVICE_FUNC bool isfinite_impl(const std::complex& x); +template EIGEN_DEVICE_FUNC bool isnan_impl(const std::complex& x); +template EIGEN_DEVICE_FUNC bool isinf_impl(const std::complex& x); + +template T generic_fast_tanh_float(const T& a_x); + +} // end namespace internal + +/**************************************************************************** +* Generic math functions * +****************************************************************************/ + +namespace numext { + +#ifndef __CUDA_ARCH__ +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T mini(const T& x, const T& y) +{ + EIGEN_USING_STD_MATH(min); + return min EIGEN_NOT_A_MACRO (x,y); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T maxi(const T& x, const T& y) +{ + EIGEN_USING_STD_MATH(max); + return max EIGEN_NOT_A_MACRO (x,y); +} +#else +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T mini(const T& x, const T& y) +{ + return y < x ? y : x; +} +template<> +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE float mini(const float& x, const float& y) +{ + return fminf(x, y); +} +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T maxi(const T& x, const T& y) +{ + return x < y ? y : x; +} +template<> +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE float maxi(const float& x, const float& y) +{ + return fmaxf(x, y); +} +#endif + + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(real, Scalar) real(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(real, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) >::type real_ref(const Scalar& x) +{ + return internal::real_ref_impl::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) real_ref(Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(real_ref, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(imag, Scalar) imag(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(imag, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(arg, Scalar) arg(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(arg, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) >::type imag_ref(const Scalar& x) +{ + return internal::imag_ref_impl::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) imag_ref(Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(imag_ref, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(conj, Scalar) conj(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(conj, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x); +} + +EIGEN_DEVICE_FUNC +inline bool abs2(bool x) { return x; } + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(norm1, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(hypot, Scalar) hypot(const Scalar& x, const Scalar& y) +{ + return EIGEN_MATHFUNC_IMPL(hypot, Scalar)::run(x, y); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(log1p, Scalar) log1p(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(log1p, Scalar)::run(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float log1p(const float &x) { return ::log1pf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double log1p(const double &x) { return ::log1p(x); } +#endif + +template +EIGEN_DEVICE_FUNC +inline typename internal::pow_impl::result_type pow(const ScalarX& x, const ScalarY& y) +{ + return internal::pow_impl::run(x, y); +} + +template EIGEN_DEVICE_FUNC bool (isnan) (const T &x) { return internal::isnan_impl(x); } +template EIGEN_DEVICE_FUNC bool (isinf) (const T &x) { return internal::isinf_impl(x); } +template EIGEN_DEVICE_FUNC bool (isfinite)(const T &x) { return internal::isfinite_impl(x); } + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(round, Scalar) round(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(round, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +T (floor)(const T& x) +{ + EIGEN_USING_STD_MATH(floor); + return floor(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float floor(const float &x) { return ::floorf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double floor(const double &x) { return ::floor(x); } +#endif + +template +EIGEN_DEVICE_FUNC +T (ceil)(const T& x) +{ + EIGEN_USING_STD_MATH(ceil); + return ceil(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float ceil(const float &x) { return ::ceilf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double ceil(const double &x) { return ::ceil(x); } +#endif + + +/** Log base 2 for 32 bits positive integers. + * Conveniently returns 0 for x==0. */ +inline int log2(int x) +{ + eigen_assert(x>=0); + unsigned int v(x); + static const int table[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 }; + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + return table[(v * 0x07C4ACDDU) >> 27]; +} + +/** \returns the square root of \a x. + * + * It is essentially equivalent to + * \code using std::sqrt; return sqrt(x); \endcode + * but slightly faster for float/double and some compilers (e.g., gcc), thanks to + * specializations when SSE is enabled. + * + * It's usage is justified in performance critical functions, like norm/normalize. + */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sqrt(const T &x) +{ + EIGEN_USING_STD_MATH(sqrt); + return sqrt(x); +} + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T log(const T &x) { + EIGEN_USING_STD_MATH(log); + return log(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float log(const float &x) { return ::logf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double log(const double &x) { return ::log(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +typename internal::enable_if::IsSigned || NumTraits::IsComplex,typename NumTraits::Real>::type +abs(const T &x) { + EIGEN_USING_STD_MATH(abs); + return abs(x); +} + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +typename internal::enable_if::IsSigned || NumTraits::IsComplex),typename NumTraits::Real>::type +abs(const T &x) { + return x; +} + +#if defined(__SYCL_DEVICE_ONLY__) +EIGEN_ALWAYS_INLINE float abs(float x) { return cl::sycl::fabs(x); } +EIGEN_ALWAYS_INLINE double abs(double x) { return cl::sycl::fabs(x); } +#endif // defined(__SYCL_DEVICE_ONLY__) + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float abs(const float &x) { return ::fabsf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double abs(const double &x) { return ::fabs(x); } + +template <> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float abs(const std::complex& x) { + return ::hypotf(x.real(), x.imag()); +} + +template <> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double abs(const std::complex& x) { + return ::hypot(x.real(), x.imag()); +} +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T exp(const T &x) { + EIGEN_USING_STD_MATH(exp); + return exp(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float exp(const float &x) { return ::expf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double exp(const double &x) { return ::exp(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T cos(const T &x) { + EIGEN_USING_STD_MATH(cos); + return cos(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float cos(const float &x) { return ::cosf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double cos(const double &x) { return ::cos(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sin(const T &x) { + EIGEN_USING_STD_MATH(sin); + return sin(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float sin(const float &x) { return ::sinf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double sin(const double &x) { return ::sin(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T tan(const T &x) { + EIGEN_USING_STD_MATH(tan); + return tan(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tan(const float &x) { return ::tanf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double tan(const double &x) { return ::tan(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T acos(const T &x) { + EIGEN_USING_STD_MATH(acos); + return acos(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float acos(const float &x) { return ::acosf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double acos(const double &x) { return ::acos(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T asin(const T &x) { + EIGEN_USING_STD_MATH(asin); + return asin(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float asin(const float &x) { return ::asinf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double asin(const double &x) { return ::asin(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T atan(const T &x) { + EIGEN_USING_STD_MATH(atan); + return atan(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float atan(const float &x) { return ::atanf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double atan(const double &x) { return ::atan(x); } +#endif + + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T cosh(const T &x) { + EIGEN_USING_STD_MATH(cosh); + return cosh(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float cosh(const float &x) { return ::coshf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double cosh(const double &x) { return ::cosh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sinh(const T &x) { + EIGEN_USING_STD_MATH(sinh); + return sinh(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float sinh(const float &x) { return ::sinhf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double sinh(const double &x) { return ::sinh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T tanh(const T &x) { + EIGEN_USING_STD_MATH(tanh); + return tanh(x); +} + +#if (!defined(__CUDACC__)) && EIGEN_FAST_MATH +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tanh(float x) { return internal::generic_fast_tanh_float(x); } +#endif + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tanh(const float &x) { return ::tanhf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double tanh(const double &x) { return ::tanh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T fmod(const T& a, const T& b) { + EIGEN_USING_STD_MATH(fmod); + return fmod(a, b); +} + +#ifdef __CUDACC__ +template <> +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float fmod(const float& a, const float& b) { + return ::fmodf(a, b); +} + +template <> +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double fmod(const double& a, const double& b) { + return ::fmod(a, b); +} +#endif + +} // end namespace numext + +namespace internal { + +template +EIGEN_DEVICE_FUNC bool isfinite_impl(const std::complex& x) +{ + return (numext::isfinite)(numext::real(x)) && (numext::isfinite)(numext::imag(x)); +} + +template +EIGEN_DEVICE_FUNC bool isnan_impl(const std::complex& x) +{ + return (numext::isnan)(numext::real(x)) || (numext::isnan)(numext::imag(x)); +} + +template +EIGEN_DEVICE_FUNC bool isinf_impl(const std::complex& x) +{ + return ((numext::isinf)(numext::real(x)) || (numext::isinf)(numext::imag(x))) && (!(numext::isnan)(x)); +} + +/**************************************************************************** +* Implementation of fuzzy comparisons * +****************************************************************************/ + +template +struct scalar_fuzzy_default_impl {}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec) + { + return numext::abs(x) <= numext::abs(y) * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return numext::abs(x - y) <= numext::mini(numext::abs(x), numext::abs(y)) * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return x <= y || isApprox(x, y, prec); + } +}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const Scalar&, const RealScalar&) + { + return x == Scalar(0); + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar&) + { + return x == y; + } + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, const RealScalar&) + { + return x <= y; + } +}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec) + { + return numext::abs2(x) <= numext::abs2(y) * prec * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return numext::abs2(x - y) <= numext::mini(numext::abs2(x), numext::abs2(y)) * prec * prec; + } +}; + +template +struct scalar_fuzzy_impl : scalar_fuzzy_default_impl::IsComplex, NumTraits::IsInteger> {}; + +template EIGEN_DEVICE_FUNC +inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::template isMuchSmallerThan(x, y, precision); +} + +template EIGEN_DEVICE_FUNC +inline bool isApprox(const Scalar& x, const Scalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::isApprox(x, y, precision); +} + +template EIGEN_DEVICE_FUNC +inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::isApproxOrLessThan(x, y, precision); +} + +/****************************************** +*** The special case of the bool type *** +******************************************/ + +template<> struct random_impl +{ + static inline bool run() + { + return random(0,1)==0 ? false : true; + } +}; + +template<> struct scalar_fuzzy_impl +{ + typedef bool RealScalar; + + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const bool& x, const bool&, const bool&) + { + return !x; + } + + EIGEN_DEVICE_FUNC + static inline bool isApprox(bool x, bool y, bool) + { + return x == y; + } + + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const bool& x, const bool& y, const bool&) + { + return (!x) || y; + } + +}; + + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_MATHFUNCTIONS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h new file mode 100644 index 00000000000..9c1ceb0eb0f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h @@ -0,0 +1,101 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2014 Pedro Gonnet (pedro.gonnet@gmail.com) +// Copyright (C) 2016 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATHFUNCTIONSIMPL_H +#define EIGEN_MATHFUNCTIONSIMPL_H + +namespace Eigen { + +namespace internal { + +/** \internal \returns the hyperbolic tan of \a a (coeff-wise) + Doesn't do anything fancy, just a 13/6-degree rational interpolant which + is accurate up to a couple of ulp in the range [-9, 9], outside of which + the tanh(x) = +/-1. + + This implementation works on both scalars and packets. +*/ +template +T generic_fast_tanh_float(const T& a_x) +{ + // Clamp the inputs to the range [-9, 9] since anything outside + // this range is +/-1.0f in single-precision. + const T plus_9 = pset1(9.f); + const T minus_9 = pset1(-9.f); + // NOTE GCC prior to 6.3 might improperly optimize this max/min + // step such that if a_x is nan, x will be either 9 or -9, + // and tanh will return 1 or -1 instead of nan. + // This is supposed to be fixed in gcc6.3, + // see: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72867 + const T x = pmax(minus_9,pmin(plus_9,a_x)); + // The monomial coefficients of the numerator polynomial (odd). + const T alpha_1 = pset1(4.89352455891786e-03f); + const T alpha_3 = pset1(6.37261928875436e-04f); + const T alpha_5 = pset1(1.48572235717979e-05f); + const T alpha_7 = pset1(5.12229709037114e-08f); + const T alpha_9 = pset1(-8.60467152213735e-11f); + const T alpha_11 = pset1(2.00018790482477e-13f); + const T alpha_13 = pset1(-2.76076847742355e-16f); + + // The monomial coefficients of the denominator polynomial (even). + const T beta_0 = pset1(4.89352518554385e-03f); + const T beta_2 = pset1(2.26843463243900e-03f); + const T beta_4 = pset1(1.18534705686654e-04f); + const T beta_6 = pset1(1.19825839466702e-06f); + + // Since the polynomials are odd/even, we need x^2. + const T x2 = pmul(x, x); + + // Evaluate the numerator polynomial p. + T p = pmadd(x2, alpha_13, alpha_11); + p = pmadd(x2, p, alpha_9); + p = pmadd(x2, p, alpha_7); + p = pmadd(x2, p, alpha_5); + p = pmadd(x2, p, alpha_3); + p = pmadd(x2, p, alpha_1); + p = pmul(x, p); + + // Evaluate the denominator polynomial p. + T q = pmadd(x2, beta_6, beta_4); + q = pmadd(x2, q, beta_2); + q = pmadd(x2, q, beta_0); + + // Divide the numerator by the denominator. + return pdiv(p, q); +} + +template +EIGEN_STRONG_INLINE +RealScalar positive_real_hypot(const RealScalar& x, const RealScalar& y) +{ + EIGEN_USING_STD_MATH(sqrt); + RealScalar p, qp; + p = numext::maxi(x,y); + if(p==RealScalar(0)) return RealScalar(0); + qp = numext::mini(y,x) / p; + return p * sqrt(RealScalar(1) + qp*qp); +} + +template +struct hypot_impl +{ + typedef typename NumTraits::Real RealScalar; + static inline RealScalar run(const Scalar& x, const Scalar& y) + { + EIGEN_USING_STD_MATH(abs); + return positive_real_hypot(abs(x), abs(y)); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_MATHFUNCTIONSIMPL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h new file mode 100644 index 00000000000..7f4a7af93c6 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h @@ -0,0 +1,459 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// Copyright (C) 2008-2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIX_H +#define EIGEN_MATRIX_H + +namespace Eigen { + +namespace internal { +template +struct traits > +{ +private: + enum { size = internal::size_at_compile_time<_Rows,_Cols>::ret }; + typedef typename find_best_packet<_Scalar,size>::type PacketScalar; + enum { + row_major_bit = _Options&RowMajor ? RowMajorBit : 0, + is_dynamic_size_storage = _MaxRows==Dynamic || _MaxCols==Dynamic, + max_size = is_dynamic_size_storage ? Dynamic : _MaxRows*_MaxCols, + default_alignment = compute_default_alignment<_Scalar,max_size>::value, + actual_alignment = ((_Options&DontAlign)==0) ? default_alignment : 0, + required_alignment = unpacket_traits::alignment, + packet_access_bit = (packet_traits<_Scalar>::Vectorizable && (EIGEN_UNALIGNED_VECTORIZE || (actual_alignment>=required_alignment))) ? PacketAccessBit : 0 + }; + +public: + typedef _Scalar Scalar; + typedef Dense StorageKind; + typedef Eigen::Index StorageIndex; + typedef MatrixXpr XprKind; + enum { + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _MaxRows, + MaxColsAtCompileTime = _MaxCols, + Flags = compute_matrix_flags<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::ret, + Options = _Options, + InnerStrideAtCompileTime = 1, + OuterStrideAtCompileTime = (Options&RowMajor) ? ColsAtCompileTime : RowsAtCompileTime, + + // FIXME, the following flag in only used to define NeedsToAlign in PlainObjectBase + EvaluatorFlags = LinearAccessBit | DirectAccessBit | packet_access_bit | row_major_bit, + Alignment = actual_alignment + }; +}; +} + +/** \class Matrix + * \ingroup Core_Module + * + * \brief The matrix class, also used for vectors and row-vectors + * + * The %Matrix class is the work-horse for all \em dense (\ref dense "note") matrices and vectors within Eigen. + * Vectors are matrices with one column, and row-vectors are matrices with one row. + * + * The %Matrix class encompasses \em both fixed-size and dynamic-size objects (\ref fixedsize "note"). + * + * The first three template parameters are required: + * \tparam _Scalar Numeric type, e.g. float, double, int or std::complex. + * User defined scalar types are supported as well (see \ref user_defined_scalars "here"). + * \tparam _Rows Number of rows, or \b Dynamic + * \tparam _Cols Number of columns, or \b Dynamic + * + * The remaining template parameters are optional -- in most cases you don't have to worry about them. + * \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of either + * \b #AutoAlign or \b #DontAlign. + * The former controls \ref TopicStorageOrders "storage order", and defaults to column-major. The latter controls alignment, which is required + * for vectorization. It defaults to aligning matrices except for fixed sizes that aren't a multiple of the packet size. + * \tparam _MaxRows Maximum number of rows. Defaults to \a _Rows (\ref maxrows "note"). + * \tparam _MaxCols Maximum number of columns. Defaults to \a _Cols (\ref maxrows "note"). + * + * Eigen provides a number of typedefs covering the usual cases. Here are some examples: + * + * \li \c Matrix2d is a 2x2 square matrix of doubles (\c Matrix) + * \li \c Vector4f is a vector of 4 floats (\c Matrix) + * \li \c RowVector3i is a row-vector of 3 ints (\c Matrix) + * + * \li \c MatrixXf is a dynamic-size matrix of floats (\c Matrix) + * \li \c VectorXf is a dynamic-size vector of floats (\c Matrix) + * + * \li \c Matrix2Xf is a partially fixed-size (dynamic-size) matrix of floats (\c Matrix) + * \li \c MatrixX3d is a partially dynamic-size (fixed-size) matrix of double (\c Matrix) + * + * See \link matrixtypedefs this page \endlink for a complete list of predefined \em %Matrix and \em Vector typedefs. + * + * You can access elements of vectors and matrices using normal subscripting: + * + * \code + * Eigen::VectorXd v(10); + * v[0] = 0.1; + * v[1] = 0.2; + * v(0) = 0.3; + * v(1) = 0.4; + * + * Eigen::MatrixXi m(10, 10); + * m(0, 1) = 1; + * m(0, 2) = 2; + * m(0, 3) = 3; + * \endcode + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIX_PLUGIN. + * + * Some notes: + * + *

+ *
\anchor dense Dense versus sparse:
+ *
This %Matrix class handles dense, not sparse matrices and vectors. For sparse matrices and vectors, see the Sparse module. + * + * Dense matrices and vectors are plain usual arrays of coefficients. All the coefficients are stored, in an ordinary contiguous array. + * This is unlike Sparse matrices and vectors where the coefficients are stored as a list of nonzero coefficients.
+ * + *
\anchor fixedsize Fixed-size versus dynamic-size:
+ *
Fixed-size means that the numbers of rows and columns are known are compile-time. In this case, Eigen allocates the array + * of coefficients as a fixed-size array, as a class member. This makes sense for very small matrices, typically up to 4x4, sometimes up + * to 16x16. Larger matrices should be declared as dynamic-size even if one happens to know their size at compile-time. + * + * Dynamic-size means that the numbers of rows or columns are not necessarily known at compile-time. In this case they are runtime + * variables, and the array of coefficients is allocated dynamically on the heap. + * + * Note that \em dense matrices, be they Fixed-size or Dynamic-size, do not expand dynamically in the sense of a std::map. + * If you want this behavior, see the Sparse module.
+ * + *
\anchor maxrows _MaxRows and _MaxCols:
+ *
In most cases, one just leaves these parameters to the default values. + * These parameters mean the maximum size of rows and columns that the matrix may have. They are useful in cases + * when the exact numbers of rows and columns are not known are compile-time, but it is known at compile-time that they cannot + * exceed a certain value. This happens when taking dynamic-size blocks inside fixed-size matrices: in this case _MaxRows and _MaxCols + * are the dimensions of the original matrix, while _Rows and _Cols are Dynamic.
+ *
+ * + * ABI and storage layout + * + * The table below summarizes the ABI of some possible Matrix instances which is fixed thorough the lifetime of Eigen 3. + * + * + * + * + * + * + *
Matrix typeEquivalent C structure
\code Matrix \endcode\code + * struct { + * T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0 + * Eigen::Index rows, cols; + * }; + * \endcode
\code + * Matrix + * Matrix \endcode\code + * struct { + * T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0 + * Eigen::Index size; + * }; + * \endcode
\code Matrix \endcode\code + * struct { + * T data[Rows*Cols]; // with (size_t(data)%A(Rows*Cols*sizeof(T)))==0 + * }; + * \endcode
\code Matrix \endcode\code + * struct { + * T data[MaxRows*MaxCols]; // with (size_t(data)%A(MaxRows*MaxCols*sizeof(T)))==0 + * Eigen::Index rows, cols; + * }; + * \endcode
+ * Note that in this table Rows, Cols, MaxRows and MaxCols are all positive integers. A(S) is defined to the largest possible power-of-two + * smaller to EIGEN_MAX_STATIC_ALIGN_BYTES. + * + * \see MatrixBase for the majority of the API methods for matrices, \ref TopicClassHierarchy, + * \ref TopicStorageOrders + */ + +template +class Matrix + : public PlainObjectBase > +{ + public: + + /** \brief Base class typedef. + * \sa PlainObjectBase + */ + typedef PlainObjectBase Base; + + enum { Options = _Options }; + + EIGEN_DENSE_PUBLIC_INTERFACE(Matrix) + + typedef typename Base::PlainObject PlainObject; + + using Base::base; + using Base::coeffRef; + + /** + * \brief Assigns matrices to each other. + * + * \note This is a special case of the templated operator=. Its purpose is + * to prevent a default operator= from hiding the templated operator=. + * + * \callgraph + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const Matrix& other) + { + return Base::_set(other); + } + + /** \internal + * \brief Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const DenseBase& other) + { + return Base::_set(other); + } + + /* Here, doxygen failed to copy the brief information when using \copydoc */ + + /** + * \brief Copies the generic expression \a other into *this. + * \copydetails DenseBase::operator=(const EigenBase &other) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const EigenBase &other) + { + return Base::operator=(other); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const ReturnByValue& func) + { + return Base::operator=(func); + } + + /** \brief Default constructor. + * + * For fixed-size matrices, does nothing. + * + * For dynamic-size matrices, creates an empty matrix of size 0. Does not allocate any array. Such a matrix + * is called a null matrix. This constructor is the unique way to create null matrices: resizing + * a matrix to 0 is not supported. + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix() : Base() + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + + // FIXME is it still needed + EIGEN_DEVICE_FUNC + explicit Matrix(internal::constructor_without_unaligned_array_assert) + : Base(internal::constructor_without_unaligned_array_assert()) + { Base::_check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED } + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + Matrix(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible::value) + : Base(std::move(other)) + { + Base::_check_template_params(); + } + EIGEN_DEVICE_FUNC + Matrix& operator=(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable::value) + { + other.swap(*this); + return *this; + } +#endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + + // This constructor is for both 1x1 matrices and dynamic vectors + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Matrix(const T& x) + { + Base::_check_template_params(); + Base::template _init1(x); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const T0& x, const T1& y) + { + Base::_check_template_params(); + Base::template _init2(x, y); + } + #else + /** \brief Constructs a fixed-sized matrix initialized with coefficients starting at \a data */ + EIGEN_DEVICE_FUNC + explicit Matrix(const Scalar *data); + + /** \brief Constructs a vector or row-vector with given dimension. \only_for_vectors + * + * This is useful for dynamic-size vectors. For fixed-size vectors, + * it is redundant to pass these parameters, so one should use the default constructor + * Matrix() instead. + * + * \warning This constructor is disabled for fixed-size \c 1x1 matrices. For instance, + * calling Matrix(1) will call the initialization constructor: Matrix(const Scalar&). + * For fixed-size \c 1x1 matrices it is therefore recommended to use the default + * constructor Matrix() instead, especially when using one of the non standard + * \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives). + */ + EIGEN_STRONG_INLINE explicit Matrix(Index dim); + /** \brief Constructs an initialized 1x1 matrix with the given coefficient */ + Matrix(const Scalar& x); + /** \brief Constructs an uninitialized matrix with \a rows rows and \a cols columns. + * + * This is useful for dynamic-size matrices. For fixed-size matrices, + * it is redundant to pass these parameters, so one should use the default constructor + * Matrix() instead. + * + * \warning This constructor is disabled for fixed-size \c 1x2 and \c 2x1 vectors. For instance, + * calling Matrix2f(2,1) will call the initialization constructor: Matrix(const Scalar& x, const Scalar& y). + * For fixed-size \c 1x2 or \c 2x1 vectors it is therefore recommended to use the default + * constructor Matrix() instead, especially when using one of the non standard + * \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives). + */ + EIGEN_DEVICE_FUNC + Matrix(Index rows, Index cols); + + /** \brief Constructs an initialized 2D vector with given coefficients */ + Matrix(const Scalar& x, const Scalar& y); + #endif + + /** \brief Constructs an initialized 3D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Matrix, 3) + m_storage.data()[0] = x; + m_storage.data()[1] = y; + m_storage.data()[2] = z; + } + /** \brief Constructs an initialized 4D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z, const Scalar& w) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Matrix, 4) + m_storage.data()[0] = x; + m_storage.data()[1] = y; + m_storage.data()[2] = z; + m_storage.data()[3] = w; + } + + + /** \brief Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Matrix& other) : Base(other) + { } + + /** \brief Copy constructor for generic expressions. + * \sa MatrixBase::operator=(const EigenBase&) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const EigenBase &other) + : Base(other.derived()) + { } + + EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); } + + /////////// Geometry module /////////// + + template + EIGEN_DEVICE_FUNC + explicit Matrix(const RotationBase& r); + template + EIGEN_DEVICE_FUNC + Matrix& operator=(const RotationBase& r); + + // allow to extend Matrix outside Eigen + #ifdef EIGEN_MATRIX_PLUGIN + #include EIGEN_MATRIX_PLUGIN + #endif + + protected: + template + friend struct internal::conservative_resize_like_impl; + + using Base::m_storage; +}; + +/** \defgroup matrixtypedefs Global matrix typedefs + * + * \ingroup Core_Module + * + * Eigen defines several typedef shortcuts for most common matrix and vector types. + * + * The general patterns are the following: + * + * \c MatrixSizeType where \c Size can be \c 2,\c 3,\c 4 for fixed size square matrices or \c X for dynamic size, + * and where \c Type can be \c i for integer, \c f for float, \c d for double, \c cf for complex float, \c cd + * for complex double. + * + * For example, \c Matrix3d is a fixed-size 3x3 matrix type of doubles, and \c MatrixXf is a dynamic-size matrix of floats. + * + * There are also \c VectorSizeType and \c RowVectorSizeType which are self-explanatory. For example, \c Vector4cf is + * a fixed-size vector of 4 complex floats. + * + * \sa class Matrix + */ + +#define EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Size, SizeSuffix) \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##SizeSuffix##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Vector##SizeSuffix##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix RowVector##SizeSuffix##TypeSuffix; + +#define EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, Size) \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##Size##X##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##X##Size##TypeSuffix; + +#define EIGEN_MAKE_TYPEDEFS_ALL_SIZES(Type, TypeSuffix) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 2, 2) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 3, 3) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 4, 4) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Dynamic, X) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 2) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 3) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 4) + +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(int, i) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(float, f) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(double, d) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(std::complex, cf) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(std::complex, cd) + +#undef EIGEN_MAKE_TYPEDEFS_ALL_SIZES +#undef EIGEN_MAKE_TYPEDEFS +#undef EIGEN_MAKE_FIXED_TYPEDEFS + +} // end namespace Eigen + +#endif // EIGEN_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h new file mode 100644 index 00000000000..f8bcc8c6f56 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h @@ -0,0 +1,530 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2009 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIXBASE_H +#define EIGEN_MATRIXBASE_H + +namespace Eigen { + +/** \class MatrixBase + * \ingroup Core_Module + * + * \brief Base class for all dense matrices, vectors, and expressions + * + * This class is the base that is inherited by all matrix, vector, and related expression + * types. Most of the Eigen API is contained in this class, and its base classes. Other important + * classes for the Eigen API are Matrix, and VectorwiseOp. + * + * Note that some methods are defined in other modules such as the \ref LU_Module LU module + * for all functions related to matrix inversions. + * + * \tparam Derived is the derived type, e.g. a matrix type, or an expression, etc. + * + * When writing a function taking Eigen objects as argument, if you want your function + * to take as argument any matrix, vector, or expression, just let it take a + * MatrixBase argument. As an example, here is a function printFirstRow which, given + * a matrix, vector, or expression \a x, prints the first row of \a x. + * + * \code + template + void printFirstRow(const Eigen::MatrixBase& x) + { + cout << x.row(0) << endl; + } + * \endcode + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIXBASE_PLUGIN. + * + * \sa \blank \ref TopicClassHierarchy + */ +template class MatrixBase + : public DenseBase +{ + public: +#ifndef EIGEN_PARSED_BY_DOXYGEN + typedef MatrixBase StorageBaseType; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + typedef DenseBase Base; + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::eval; + using Base::operator+=; + using Base::operator-=; + using Base::operator*=; + using Base::operator/=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + typedef typename Base::ConstTransposeReturnType ConstTransposeReturnType; + typedef typename Base::RowXpr RowXpr; + typedef typename Base::ColXpr ColXpr; +#endif // not EIGEN_PARSED_BY_DOXYGEN + + + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** type of the equivalent square matrix */ + typedef Matrix SquareMatrixType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + + /** \returns the size of the main diagonal, which is min(rows(),cols()). + * \sa rows(), cols(), SizeAtCompileTime. */ + EIGEN_DEVICE_FUNC + inline Index diagonalSize() const { return (numext::mini)(rows(),cols()); } + + typedef typename Base::PlainObject PlainObject; + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; + /** \internal the return type of MatrixBase::adjoint() */ + typedef typename internal::conditional::IsComplex, + CwiseUnaryOp, ConstTransposeReturnType>, + ConstTransposeReturnType + >::type AdjointReturnType; + /** \internal Return type of eigenvalues() */ + typedef Matrix, internal::traits::ColsAtCompileTime, 1, ColMajor> EigenvaluesReturnType; + /** \internal the return type of identity */ + typedef CwiseNullaryOp,PlainObject> IdentityReturnType; + /** \internal the return type of unit vectors */ + typedef Block, SquareMatrixType>, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime> BasisReturnType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::MatrixBase +#define EIGEN_DOC_UNARY_ADDONS(X,Y) +# include "../plugins/CommonCwiseUnaryOps.h" +# include "../plugins/CommonCwiseBinaryOps.h" +# include "../plugins/MatrixCwiseUnaryOps.h" +# include "../plugins/MatrixCwiseBinaryOps.h" +# ifdef EIGEN_MATRIXBASE_PLUGIN +# include EIGEN_MATRIXBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_UNARY_ADDONS + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const MatrixBase& other); + + // We cannot inherit here via Base::operator= since it is causing + // trouble with MSVC. + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const EigenBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const ReturnByValue& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const MatrixBase& other); + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const MatrixBase& other); + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const MatrixBase &other) const; + + template + EIGEN_DEVICE_FUNC + const Product + lazyProduct(const MatrixBase &other) const; + + template + Derived& operator*=(const EigenBase& other); + + template + void applyOnTheLeft(const EigenBase& other); + + template + void applyOnTheRight(const EigenBase& other); + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const DiagonalBase &diagonal) const; + + template + EIGEN_DEVICE_FUNC + typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType + dot(const MatrixBase& other) const; + + EIGEN_DEVICE_FUNC RealScalar squaredNorm() const; + EIGEN_DEVICE_FUNC RealScalar norm() const; + RealScalar stableNorm() const; + RealScalar blueNorm() const; + RealScalar hypotNorm() const; + EIGEN_DEVICE_FUNC const PlainObject normalized() const; + EIGEN_DEVICE_FUNC const PlainObject stableNormalized() const; + EIGEN_DEVICE_FUNC void normalize(); + EIGEN_DEVICE_FUNC void stableNormalize(); + + EIGEN_DEVICE_FUNC const AdjointReturnType adjoint() const; + EIGEN_DEVICE_FUNC void adjointInPlace(); + + typedef Diagonal DiagonalReturnType; + EIGEN_DEVICE_FUNC + DiagonalReturnType diagonal(); + + typedef typename internal::add_const >::type ConstDiagonalReturnType; + EIGEN_DEVICE_FUNC + ConstDiagonalReturnType diagonal() const; + + template struct DiagonalIndexReturnType { typedef Diagonal Type; }; + template struct ConstDiagonalIndexReturnType { typedef const Diagonal Type; }; + + template + EIGEN_DEVICE_FUNC + typename DiagonalIndexReturnType::Type diagonal(); + + template + EIGEN_DEVICE_FUNC + typename ConstDiagonalIndexReturnType::Type diagonal() const; + + typedef Diagonal DiagonalDynamicIndexReturnType; + typedef typename internal::add_const >::type ConstDiagonalDynamicIndexReturnType; + + EIGEN_DEVICE_FUNC + DiagonalDynamicIndexReturnType diagonal(Index index); + EIGEN_DEVICE_FUNC + ConstDiagonalDynamicIndexReturnType diagonal(Index index) const; + + template struct TriangularViewReturnType { typedef TriangularView Type; }; + template struct ConstTriangularViewReturnType { typedef const TriangularView Type; }; + + template + EIGEN_DEVICE_FUNC + typename TriangularViewReturnType::Type triangularView(); + template + EIGEN_DEVICE_FUNC + typename ConstTriangularViewReturnType::Type triangularView() const; + + template struct SelfAdjointViewReturnType { typedef SelfAdjointView Type; }; + template struct ConstSelfAdjointViewReturnType { typedef const SelfAdjointView Type; }; + + template + EIGEN_DEVICE_FUNC + typename SelfAdjointViewReturnType::Type selfadjointView(); + template + EIGEN_DEVICE_FUNC + typename ConstSelfAdjointViewReturnType::Type selfadjointView() const; + + const SparseView sparseView(const Scalar& m_reference = Scalar(0), + const typename NumTraits::Real& m_epsilon = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC static const IdentityReturnType Identity(); + EIGEN_DEVICE_FUNC static const IdentityReturnType Identity(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index size, Index i); + EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index i); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitX(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitY(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitZ(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitW(); + + EIGEN_DEVICE_FUNC + const DiagonalWrapper asDiagonal() const; + const PermutationWrapper asPermutation() const; + + EIGEN_DEVICE_FUNC + Derived& setIdentity(); + EIGEN_DEVICE_FUNC + Derived& setIdentity(Index rows, Index cols); + + bool isIdentity(const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isDiagonal(const RealScalar& prec = NumTraits::dummy_precision()) const; + + bool isUpperTriangular(const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isLowerTriangular(const RealScalar& prec = NumTraits::dummy_precision()) const; + + template + bool isOrthogonal(const MatrixBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isUnitary(const RealScalar& prec = NumTraits::dummy_precision()) const; + + /** \returns true if each coefficients of \c *this and \a other are all exactly equal. + * \warning When using floating point scalar values you probably should rather use a + * fuzzy comparison such as isApprox() + * \sa isApprox(), operator!= */ + template + EIGEN_DEVICE_FUNC inline bool operator==(const MatrixBase& other) const + { return cwiseEqual(other).all(); } + + /** \returns true if at least one pair of coefficients of \c *this and \a other are not exactly equal to each other. + * \warning When using floating point scalar values you probably should rather use a + * fuzzy comparison such as isApprox() + * \sa isApprox(), operator== */ + template + EIGEN_DEVICE_FUNC inline bool operator!=(const MatrixBase& other) const + { return cwiseNotEqual(other).any(); } + + NoAlias noalias(); + + // TODO forceAlignedAccess is temporarily disabled + // Need to find a nicer workaround. + inline const Derived& forceAlignedAccess() const { return derived(); } + inline Derived& forceAlignedAccess() { return derived(); } + template inline const Derived& forceAlignedAccessIf() const { return derived(); } + template inline Derived& forceAlignedAccessIf() { return derived(); } + + EIGEN_DEVICE_FUNC Scalar trace() const; + + template EIGEN_DEVICE_FUNC RealScalar lpNorm() const; + + EIGEN_DEVICE_FUNC MatrixBase& matrix() { return *this; } + EIGEN_DEVICE_FUNC const MatrixBase& matrix() const { return *this; } + + /** \returns an \link Eigen::ArrayBase Array \endlink expression of this matrix + * \sa ArrayBase::matrix() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE ArrayWrapper array() { return ArrayWrapper(derived()); } + /** \returns a const \link Eigen::ArrayBase Array \endlink expression of this matrix + * \sa ArrayBase::matrix() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const ArrayWrapper array() const { return ArrayWrapper(derived()); } + +/////////// LU module /////////// + + inline const FullPivLU fullPivLu() const; + inline const PartialPivLU partialPivLu() const; + + inline const PartialPivLU lu() const; + + inline const Inverse inverse() const; + + template + inline void computeInverseAndDetWithCheck( + ResultType& inverse, + typename ResultType::Scalar& determinant, + bool& invertible, + const RealScalar& absDeterminantThreshold = NumTraits::dummy_precision() + ) const; + template + inline void computeInverseWithCheck( + ResultType& inverse, + bool& invertible, + const RealScalar& absDeterminantThreshold = NumTraits::dummy_precision() + ) const; + Scalar determinant() const; + +/////////// Cholesky module /////////// + + inline const LLT llt() const; + inline const LDLT ldlt() const; + +/////////// QR module /////////// + + inline const HouseholderQR householderQr() const; + inline const ColPivHouseholderQR colPivHouseholderQr() const; + inline const FullPivHouseholderQR fullPivHouseholderQr() const; + inline const CompleteOrthogonalDecomposition completeOrthogonalDecomposition() const; + +/////////// Eigenvalues module /////////// + + inline EigenvaluesReturnType eigenvalues() const; + inline RealScalar operatorNorm() const; + +/////////// SVD module /////////// + + inline JacobiSVD jacobiSvd(unsigned int computationOptions = 0) const; + inline BDCSVD bdcSvd(unsigned int computationOptions = 0) const; + +/////////// Geometry module /////////// + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /// \internal helper struct to form the return type of the cross product + template struct cross_product_return_type { + typedef typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType Scalar; + typedef Matrix type; + }; + #endif // EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC +#ifndef EIGEN_PARSED_BY_DOXYGEN + inline typename cross_product_return_type::type +#else + inline PlainObject +#endif + cross(const MatrixBase& other) const; + + template + EIGEN_DEVICE_FUNC + inline PlainObject cross3(const MatrixBase& other) const; + + EIGEN_DEVICE_FUNC + inline PlainObject unitOrthogonal(void) const; + + EIGEN_DEVICE_FUNC + inline Matrix eulerAngles(Index a0, Index a1, Index a2) const; + + // put this as separate enum value to work around possible GCC 4.3 bug (?) + enum { HomogeneousReturnTypeDirection = ColsAtCompileTime==1&&RowsAtCompileTime==1 ? ((internal::traits::Flags&RowMajorBit)==RowMajorBit ? Horizontal : Vertical) + : ColsAtCompileTime==1 ? Vertical : Horizontal }; + typedef Homogeneous HomogeneousReturnType; + EIGEN_DEVICE_FUNC + inline HomogeneousReturnType homogeneous() const; + + enum { + SizeMinusOne = SizeAtCompileTime==Dynamic ? Dynamic : SizeAtCompileTime-1 + }; + typedef Block::ColsAtCompileTime==1 ? SizeMinusOne : 1, + internal::traits::ColsAtCompileTime==1 ? 1 : SizeMinusOne> ConstStartMinusOne; + typedef EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(ConstStartMinusOne,Scalar,quotient) HNormalizedReturnType; + EIGEN_DEVICE_FUNC + inline const HNormalizedReturnType hnormalized() const; + +////////// Householder module /////////// + + void makeHouseholderInPlace(Scalar& tau, RealScalar& beta); + template + void makeHouseholder(EssentialPart& essential, + Scalar& tau, RealScalar& beta) const; + template + void applyHouseholderOnTheLeft(const EssentialPart& essential, + const Scalar& tau, + Scalar* workspace); + template + void applyHouseholderOnTheRight(const EssentialPart& essential, + const Scalar& tau, + Scalar* workspace); + +///////// Jacobi module ///////// + + template + void applyOnTheLeft(Index p, Index q, const JacobiRotation& j); + template + void applyOnTheRight(Index p, Index q, const JacobiRotation& j); + +///////// SparseCore module ///////// + + template + EIGEN_STRONG_INLINE const typename SparseMatrixBase::template CwiseProductDenseReturnType::Type + cwiseProduct(const SparseMatrixBase &other) const + { + return other.cwiseProduct(derived()); + } + +///////// MatrixFunctions module ///////// + + typedef typename internal::stem_function::type StemFunction; +#define EIGEN_MATRIX_FUNCTION(ReturnType, Name, Description) \ + /** \returns an expression of the matrix Description of \c *this. \brief This function requires the unsupported MatrixFunctions module. To compute the coefficient-wise Description use ArrayBase::##Name . */ \ + const ReturnType Name() const; +#define EIGEN_MATRIX_FUNCTION_1(ReturnType, Name, Description, Argument) \ + /** \returns an expression of the matrix Description of \c *this. \brief This function requires the unsupported MatrixFunctions module. To compute the coefficient-wise Description use ArrayBase::##Name . */ \ + const ReturnType Name(Argument) const; + + EIGEN_MATRIX_FUNCTION(MatrixExponentialReturnValue, exp, exponential) + /** \brief Helper function for the unsupported MatrixFunctions module.*/ + const MatrixFunctionReturnValue matrixFunction(StemFunction f) const; + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cosh, hyperbolic cosine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sinh, hyperbolic sine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cos, cosine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sin, sine) + EIGEN_MATRIX_FUNCTION(MatrixSquareRootReturnValue, sqrt, square root) + EIGEN_MATRIX_FUNCTION(MatrixLogarithmReturnValue, log, logarithm) + EIGEN_MATRIX_FUNCTION_1(MatrixPowerReturnValue, pow, power to \c p, const RealScalar& p) + EIGEN_MATRIX_FUNCTION_1(MatrixComplexPowerReturnValue, pow, power to \c p, const std::complex& p) + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MatrixBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MatrixBase) + + private: + EIGEN_DEVICE_FUNC explicit MatrixBase(int); + EIGEN_DEVICE_FUNC MatrixBase(int,int); + template EIGEN_DEVICE_FUNC explicit MatrixBase(const MatrixBase&); + protected: + // mixing arrays and matrices is not legal + template Derived& operator+=(const ArrayBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} + // mixing arrays and matrices is not legal + template Derived& operator-=(const ArrayBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} +}; + + +/*************************************************************************** +* Implementation of matrix base methods +***************************************************************************/ + +/** replaces \c *this by \c *this * \a other. + * + * \returns a reference to \c *this + * + * Example: \include MatrixBase_applyOnTheRight.cpp + * Output: \verbinclude MatrixBase_applyOnTheRight.out + */ +template +template +inline Derived& +MatrixBase::operator*=(const EigenBase &other) +{ + other.derived().applyThisOnTheRight(derived()); + return derived(); +} + +/** replaces \c *this by \c *this * \a other. It is equivalent to MatrixBase::operator*=(). + * + * Example: \include MatrixBase_applyOnTheRight.cpp + * Output: \verbinclude MatrixBase_applyOnTheRight.out + */ +template +template +inline void MatrixBase::applyOnTheRight(const EigenBase &other) +{ + other.derived().applyThisOnTheRight(derived()); +} + +/** replaces \c *this by \a other * \c *this. + * + * Example: \include MatrixBase_applyOnTheLeft.cpp + * Output: \verbinclude MatrixBase_applyOnTheLeft.out + */ +template +template +inline void MatrixBase::applyOnTheLeft(const EigenBase &other) +{ + other.derived().applyThisOnTheLeft(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_MATRIXBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h new file mode 100644 index 00000000000..13adf070e89 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h @@ -0,0 +1,110 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NESTBYVALUE_H +#define EIGEN_NESTBYVALUE_H + +namespace Eigen { + +namespace internal { +template +struct traits > : public traits +{}; +} + +/** \class NestByValue + * \ingroup Core_Module + * + * \brief Expression which must be nested by value + * + * \tparam ExpressionType the type of the object of which we are requiring nesting-by-value + * + * This class is the return type of MatrixBase::nestByValue() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::nestByValue() + */ +template class NestByValue + : public internal::dense_xpr_base< NestByValue >::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(NestByValue) + + EIGEN_DEVICE_FUNC explicit inline NestByValue(const ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index row, Index col) const + { + return m_expression.coeff(row, col); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index row, Index col) + { + return m_expression.const_cast_derived().coeffRef(row, col); + } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index index) const + { + return m_expression.coeff(index); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index index) + { + return m_expression.const_cast_derived().coeffRef(index); + } + + template + inline const PacketScalar packet(Index row, Index col) const + { + return m_expression.template packet(row, col); + } + + template + inline void writePacket(Index row, Index col, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(row, col, x); + } + + template + inline const PacketScalar packet(Index index) const + { + return m_expression.template packet(index); + } + + template + inline void writePacket(Index index, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(index, x); + } + + EIGEN_DEVICE_FUNC operator const ExpressionType&() const { return m_expression; } + + protected: + const ExpressionType m_expression; +}; + +/** \returns an expression of the temporary version of *this. + */ +template +inline const NestByValue +DenseBase::nestByValue() const +{ + return NestByValue(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_NESTBYVALUE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h new file mode 100644 index 00000000000..33908010b4f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h @@ -0,0 +1,108 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NOALIAS_H +#define EIGEN_NOALIAS_H + +namespace Eigen { + +/** \class NoAlias + * \ingroup Core_Module + * + * \brief Pseudo expression providing an operator = assuming no aliasing + * + * \tparam ExpressionType the type of the object on which to do the lazy assignment + * + * This class represents an expression with special assignment operators + * assuming no aliasing between the target expression and the source expression. + * More precisely it alloas to bypass the EvalBeforeAssignBit flag of the source expression. + * It is the return type of MatrixBase::noalias() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::noalias() + */ +template class StorageBase> +class NoAlias +{ + public: + typedef typename ExpressionType::Scalar Scalar; + + explicit NoAlias(ExpressionType& expression) : m_expression(expression) {} + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::assign_op()); + return m_expression; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator+=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::add_assign_op()); + return m_expression; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator-=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::sub_assign_op()); + return m_expression; + } + + EIGEN_DEVICE_FUNC + ExpressionType& expression() const + { + return m_expression; + } + + protected: + ExpressionType& m_expression; +}; + +/** \returns a pseudo expression of \c *this with an operator= assuming + * no aliasing between \c *this and the source expression. + * + * More precisely, noalias() allows to bypass the EvalBeforeAssignBit flag. + * Currently, even though several expressions may alias, only product + * expressions have this flag. Therefore, noalias() is only usefull when + * the source expression contains a matrix product. + * + * Here are some examples where noalias is usefull: + * \code + * D.noalias() = A * B; + * D.noalias() += A.transpose() * B; + * D.noalias() -= 2 * A * B.adjoint(); + * \endcode + * + * On the other hand the following example will lead to a \b wrong result: + * \code + * A.noalias() = A * B; + * \endcode + * because the result matrix A is also an operand of the matrix product. Therefore, + * there is no alternative than evaluating A * B in a temporary, that is the default + * behavior when you write: + * \code + * A = A * B; + * \endcode + * + * \sa class NoAlias + */ +template +NoAlias MatrixBase::noalias() +{ + return NoAlias(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_NOALIAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h new file mode 100644 index 00000000000..daf48987898 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h @@ -0,0 +1,248 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NUMTRAITS_H +#define EIGEN_NUMTRAITS_H + +namespace Eigen { + +namespace internal { + +// default implementation of digits10(), based on numeric_limits if specialized, +// 0 for integer types, and log10(epsilon()) otherwise. +template< typename T, + bool use_numeric_limits = std::numeric_limits::is_specialized, + bool is_integer = NumTraits::IsInteger> +struct default_digits10_impl +{ + static int run() { return std::numeric_limits::digits10; } +}; + +template +struct default_digits10_impl // Floating point +{ + static int run() { + using std::log10; + using std::ceil; + typedef typename NumTraits::Real Real; + return int(ceil(-log10(NumTraits::epsilon()))); + } +}; + +template +struct default_digits10_impl // Integer +{ + static int run() { return 0; } +}; + +} // end namespace internal + +/** \class NumTraits + * \ingroup Core_Module + * + * \brief Holds information about the various numeric (i.e. scalar) types allowed by Eigen. + * + * \tparam T the numeric type at hand + * + * This class stores enums, typedefs and static methods giving information about a numeric type. + * + * The provided data consists of: + * \li A typedef \c Real, giving the "real part" type of \a T. If \a T is already real, + * then \c Real is just a typedef to \a T. If \a T is \c std::complex then \c Real + * is a typedef to \a U. + * \li A typedef \c NonInteger, giving the type that should be used for operations producing non-integral values, + * such as quotients, square roots, etc. If \a T is a floating-point type, then this typedef just gives + * \a T again. Note however that many Eigen functions such as internal::sqrt simply refuse to + * take integers. Outside of a few cases, Eigen doesn't do automatic type promotion. Thus, this typedef is + * only intended as a helper for code that needs to explicitly promote types. + * \li A typedef \c Literal giving the type to use for numeric literals such as "2" or "0.5". For instance, for \c std::complex, Literal is defined as \c U. + * Of course, this type must be fully compatible with \a T. In doubt, just use \a T here. + * \li A typedef \a Nested giving the type to use to nest a value inside of the expression tree. If you don't know what + * this means, just use \a T here. + * \li An enum value \a IsComplex. It is equal to 1 if \a T is a \c std::complex + * type, and to 0 otherwise. + * \li An enum value \a IsInteger. It is equal to \c 1 if \a T is an integer type such as \c int, + * and to \c 0 otherwise. + * \li Enum values ReadCost, AddCost and MulCost representing a rough estimate of the number of CPU cycles needed + * to by move / add / mul instructions respectively, assuming the data is already stored in CPU registers. + * Stay vague here. No need to do architecture-specific stuff. + * \li An enum value \a IsSigned. It is equal to \c 1 if \a T is a signed type and to 0 if \a T is unsigned. + * \li An enum value \a RequireInitialization. It is equal to \c 1 if the constructor of the numeric type \a T must + * be called, and to 0 if it is safe not to call it. Default is 0 if \a T is an arithmetic type, and 1 otherwise. + * \li An epsilon() function which, unlike std::numeric_limits::epsilon(), + * it returns a \a Real instead of a \a T. + * \li A dummy_precision() function returning a weak epsilon value. It is mainly used as a default + * value by the fuzzy comparison operators. + * \li highest() and lowest() functions returning the highest and lowest possible values respectively. + * \li digits10() function returning the number of decimal digits that can be represented without change. This is + * the analogue of std::numeric_limits::digits10 + * which is used as the default implementation if specialized. + */ + +template struct GenericNumTraits +{ + enum { + IsInteger = std::numeric_limits::is_integer, + IsSigned = std::numeric_limits::is_signed, + IsComplex = 0, + RequireInitialization = internal::is_arithmetic::value ? 0 : 1, + ReadCost = 1, + AddCost = 1, + MulCost = 1 + }; + + typedef T Real; + typedef typename internal::conditional< + IsInteger, + typename internal::conditional::type, + T + >::type NonInteger; + typedef T Nested; + typedef T Literal; + + EIGEN_DEVICE_FUNC + static inline Real epsilon() + { + return numext::numeric_limits::epsilon(); + } + + EIGEN_DEVICE_FUNC + static inline int digits10() + { + return internal::default_digits10_impl::run(); + } + + EIGEN_DEVICE_FUNC + static inline Real dummy_precision() + { + // make sure to override this for floating-point types + return Real(0); + } + + + EIGEN_DEVICE_FUNC + static inline T highest() { + return (numext::numeric_limits::max)(); + } + + EIGEN_DEVICE_FUNC + static inline T lowest() { + return IsInteger ? (numext::numeric_limits::min)() : (-(numext::numeric_limits::max)()); + } + + EIGEN_DEVICE_FUNC + static inline T infinity() { + return numext::numeric_limits::infinity(); + } + + EIGEN_DEVICE_FUNC + static inline T quiet_NaN() { + return numext::numeric_limits::quiet_NaN(); + } +}; + +template struct NumTraits : GenericNumTraits +{}; + +template<> struct NumTraits + : GenericNumTraits +{ + EIGEN_DEVICE_FUNC + static inline float dummy_precision() { return 1e-5f; } +}; + +template<> struct NumTraits : GenericNumTraits +{ + EIGEN_DEVICE_FUNC + static inline double dummy_precision() { return 1e-12; } +}; + +template<> struct NumTraits + : GenericNumTraits +{ + static inline long double dummy_precision() { return 1e-15l; } +}; + +template struct NumTraits > + : GenericNumTraits > +{ + typedef _Real Real; + typedef typename NumTraits<_Real>::Literal Literal; + enum { + IsComplex = 1, + RequireInitialization = NumTraits<_Real>::RequireInitialization, + ReadCost = 2 * NumTraits<_Real>::ReadCost, + AddCost = 2 * NumTraits::AddCost, + MulCost = 4 * NumTraits::MulCost + 2 * NumTraits::AddCost + }; + + EIGEN_DEVICE_FUNC + static inline Real epsilon() { return NumTraits::epsilon(); } + EIGEN_DEVICE_FUNC + static inline Real dummy_precision() { return NumTraits::dummy_precision(); } + EIGEN_DEVICE_FUNC + static inline int digits10() { return NumTraits::digits10(); } +}; + +template +struct NumTraits > +{ + typedef Array ArrayType; + typedef typename NumTraits::Real RealScalar; + typedef Array Real; + typedef typename NumTraits::NonInteger NonIntegerScalar; + typedef Array NonInteger; + typedef ArrayType & Nested; + typedef typename NumTraits::Literal Literal; + + enum { + IsComplex = NumTraits::IsComplex, + IsInteger = NumTraits::IsInteger, + IsSigned = NumTraits::IsSigned, + RequireInitialization = 1, + ReadCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::ReadCost, + AddCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::AddCost, + MulCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::MulCost + }; + + EIGEN_DEVICE_FUNC + static inline RealScalar epsilon() { return NumTraits::epsilon(); } + EIGEN_DEVICE_FUNC + static inline RealScalar dummy_precision() { return NumTraits::dummy_precision(); } + + static inline int digits10() { return NumTraits::digits10(); } +}; + +template<> struct NumTraits + : GenericNumTraits +{ + enum { + RequireInitialization = 1, + ReadCost = HugeCost, + AddCost = HugeCost, + MulCost = HugeCost + }; + + static inline int digits10() { return 0; } + +private: + static inline std::string epsilon(); + static inline std::string dummy_precision(); + static inline std::string lowest(); + static inline std::string highest(); + static inline std::string infinity(); + static inline std::string quiet_NaN(); +}; + +// Empty specialization for void to allow template specialization based on NumTraits::Real with T==void and SFINAE. +template<> struct NumTraits {}; + +} // end namespace Eigen + +#endif // EIGEN_NUMTRAITS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h new file mode 100644 index 00000000000..47c06ba7706 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h @@ -0,0 +1,605 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Benoit Jacob +// Copyright (C) 2009-2015 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PERMUTATIONMATRIX_H +#define EIGEN_PERMUTATIONMATRIX_H + +namespace Eigen { + +namespace internal { + +enum PermPermProduct_t {PermPermProduct}; + +} // end namespace internal + +/** \class PermutationBase + * \ingroup Core_Module + * + * \brief Base class for permutations + * + * \tparam Derived the derived class + * + * This class is the base class for all expressions representing a permutation matrix, + * internally stored as a vector of integers. + * The convention followed here is that if \f$ \sigma \f$ is a permutation, the corresponding permutation matrix + * \f$ P_\sigma \f$ is such that if \f$ (e_1,\ldots,e_p) \f$ is the canonical basis, we have: + * \f[ P_\sigma(e_i) = e_{\sigma(i)}. \f] + * This convention ensures that for any two permutations \f$ \sigma, \tau \f$, we have: + * \f[ P_{\sigma\circ\tau} = P_\sigma P_\tau. \f] + * + * Permutation matrices are square and invertible. + * + * Notice that in addition to the member functions and operators listed here, there also are non-member + * operator* to multiply any kind of permutation object with any kind of matrix expression (MatrixBase) + * on either side. + * + * \sa class PermutationMatrix, class PermutationWrapper + */ +template +class PermutationBase : public EigenBase +{ + typedef internal::traits Traits; + typedef EigenBase Base; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + enum { + Flags = Traits::Flags, + RowsAtCompileTime = Traits::RowsAtCompileTime, + ColsAtCompileTime = Traits::ColsAtCompileTime, + MaxRowsAtCompileTime = Traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = Traits::MaxColsAtCompileTime + }; + typedef typename Traits::StorageIndex StorageIndex; + typedef Matrix + DenseMatrixType; + typedef PermutationMatrix + PlainPermutationType; + typedef PlainPermutationType PlainObject; + using Base::derived; + typedef Inverse InverseReturnType; + typedef void Scalar; + #endif + + /** Copies the other permutation into *this */ + template + Derived& operator=(const PermutationBase& other) + { + indices() = other.indices(); + return derived(); + } + + /** Assignment from the Transpositions \a tr */ + template + Derived& operator=(const TranspositionsBase& tr) + { + setIdentity(tr.size()); + for(Index k=size()-1; k>=0; --k) + applyTranspositionOnTheRight(k,tr.coeff(k)); + return derived(); + } + + /** \returns the number of rows */ + inline Index rows() const { return Index(indices().size()); } + + /** \returns the number of columns */ + inline Index cols() const { return Index(indices().size()); } + + /** \returns the size of a side of the respective square matrix, i.e., the number of indices */ + inline Index size() const { return Index(indices().size()); } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + void evalTo(MatrixBase& other) const + { + other.setZero(); + for (Index i=0; i=0 && j>=0 && i=0 && j>=0 && i + void assignTranspose(const PermutationBase& other) + { + for (Index i=0; i + void assignProduct(const Lhs& lhs, const Rhs& rhs) + { + eigen_assert(lhs.cols() == rhs.rows()); + for (Index i=0; i + inline PlainPermutationType operator*(const PermutationBase& other) const + { return PlainPermutationType(internal::PermPermProduct, derived(), other.derived()); } + + /** \returns the product of a permutation with another inverse permutation. + * + * \note \blank \note_try_to_help_rvo + */ + template + inline PlainPermutationType operator*(const InverseImpl& other) const + { return PlainPermutationType(internal::PermPermProduct, *this, other.eval()); } + + /** \returns the product of an inverse permutation with another permutation. + * + * \note \blank \note_try_to_help_rvo + */ + template friend + inline PlainPermutationType operator*(const InverseImpl& other, const PermutationBase& perm) + { return PlainPermutationType(internal::PermPermProduct, other.eval(), perm); } + + /** \returns the determinant of the permutation matrix, which is either 1 or -1 depending on the parity of the permutation. + * + * This function is O(\c n) procedure allocating a buffer of \c n booleans. + */ + Index determinant() const + { + Index res = 1; + Index n = size(); + Matrix mask(n); + mask.fill(false); + Index r = 0; + while(r < n) + { + // search for the next seed + while(r=n) + break; + // we got one, let's follow it until we are back to the seed + Index k0 = r++; + mask.coeffRef(k0) = true; + for(Index k=indices().coeff(k0); k!=k0; k=indices().coeff(k)) + { + mask.coeffRef(k) = true; + res = -res; + } + } + return res; + } + + protected: + +}; + +namespace internal { +template +struct traits > + : traits > +{ + typedef PermutationStorage StorageKind; + typedef Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType; + typedef _StorageIndex StorageIndex; + typedef void Scalar; +}; +} + +/** \class PermutationMatrix + * \ingroup Core_Module + * + * \brief Permutation matrix + * + * \tparam SizeAtCompileTime the number of rows/cols, or Dynamic + * \tparam MaxSizeAtCompileTime the maximum number of rows/cols, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it. + * \tparam _StorageIndex the integer type of the indices + * + * This class represents a permutation matrix, internally stored as a vector of integers. + * + * \sa class PermutationBase, class PermutationWrapper, class DiagonalMatrix + */ +template +class PermutationMatrix : public PermutationBase > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + typedef const PermutationMatrix& Nested; + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + typedef typename Traits::StorageIndex StorageIndex; + #endif + + inline PermutationMatrix() + {} + + /** Constructs an uninitialized permutation matrix of given size. + */ + explicit inline PermutationMatrix(Index size) : m_indices(size) + { + eigen_internal_assert(size <= NumTraits::highest()); + } + + /** Copy constructor. */ + template + inline PermutationMatrix(const PermutationBase& other) + : m_indices(other.indices()) {} + + /** Generic constructor from expression of the indices. The indices + * array has the meaning that the permutations sends each integer i to indices[i]. + * + * \warning It is your responsibility to check that the indices array that you passes actually + * describes a permutation, i.e., each value between 0 and n-1 occurs exactly once, where n is the + * array's size. + */ + template + explicit inline PermutationMatrix(const MatrixBase& indices) : m_indices(indices) + {} + + /** Convert the Transpositions \a tr to a permutation matrix */ + template + explicit PermutationMatrix(const TranspositionsBase& tr) + : m_indices(tr.size()) + { + *this = tr; + } + + /** Copies the other permutation into *this */ + template + PermutationMatrix& operator=(const PermutationBase& other) + { + m_indices = other.indices(); + return *this; + } + + /** Assignment from the Transpositions \a tr */ + template + PermutationMatrix& operator=(const TranspositionsBase& tr) + { + return Base::operator=(tr.derived()); + } + + /** const version of indices(). */ + const IndicesType& indices() const { return m_indices; } + /** \returns a reference to the stored array representing the permutation. */ + IndicesType& indices() { return m_indices; } + + + /**** multiplication helpers to hopefully get RVO ****/ + +#ifndef EIGEN_PARSED_BY_DOXYGEN + template + PermutationMatrix(const InverseImpl& other) + : m_indices(other.derived().nestedExpression().size()) + { + eigen_internal_assert(m_indices.size() <= NumTraits::highest()); + StorageIndex end = StorageIndex(m_indices.size()); + for (StorageIndex i=0; i + PermutationMatrix(internal::PermPermProduct_t, const Lhs& lhs, const Rhs& rhs) + : m_indices(lhs.indices().size()) + { + Base::assignProduct(lhs,rhs); + } +#endif + + protected: + + IndicesType m_indices; +}; + + +namespace internal { +template +struct traits,_PacketAccess> > + : traits > +{ + typedef PermutationStorage StorageKind; + typedef Map, _PacketAccess> IndicesType; + typedef _StorageIndex StorageIndex; + typedef void Scalar; +}; +} + +template +class Map,_PacketAccess> + : public PermutationBase,_PacketAccess> > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + typedef typename IndicesType::Scalar StorageIndex; + #endif + + inline Map(const StorageIndex* indicesPtr) + : m_indices(indicesPtr) + {} + + inline Map(const StorageIndex* indicesPtr, Index size) + : m_indices(indicesPtr,size) + {} + + /** Copies the other permutation into *this */ + template + Map& operator=(const PermutationBase& other) + { return Base::operator=(other.derived()); } + + /** Assignment from the Transpositions \a tr */ + template + Map& operator=(const TranspositionsBase& tr) + { return Base::operator=(tr.derived()); } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + Map& operator=(const Map& other) + { + m_indices = other.m_indices; + return *this; + } + #endif + + /** const version of indices(). */ + const IndicesType& indices() const { return m_indices; } + /** \returns a reference to the stored array representing the permutation. */ + IndicesType& indices() { return m_indices; } + + protected: + + IndicesType m_indices; +}; + +template class TranspositionsWrapper; +namespace internal { +template +struct traits > +{ + typedef PermutationStorage StorageKind; + typedef void Scalar; + typedef typename _IndicesType::Scalar StorageIndex; + typedef _IndicesType IndicesType; + enum { + RowsAtCompileTime = _IndicesType::SizeAtCompileTime, + ColsAtCompileTime = _IndicesType::SizeAtCompileTime, + MaxRowsAtCompileTime = IndicesType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = IndicesType::MaxSizeAtCompileTime, + Flags = 0 + }; +}; +} + +/** \class PermutationWrapper + * \ingroup Core_Module + * + * \brief Class to view a vector of integers as a permutation matrix + * + * \tparam _IndicesType the type of the vector of integer (can be any compatible expression) + * + * This class allows to view any vector expression of integers as a permutation matrix. + * + * \sa class PermutationBase, class PermutationMatrix + */ +template +class PermutationWrapper : public PermutationBase > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + #endif + + inline PermutationWrapper(const IndicesType& indices) + : m_indices(indices) + {} + + /** const version of indices(). */ + const typename internal::remove_all::type& + indices() const { return m_indices; } + + protected: + + typename IndicesType::Nested m_indices; +}; + + +/** \returns the matrix with the permutation applied to the columns. + */ +template +EIGEN_DEVICE_FUNC +const Product +operator*(const MatrixBase &matrix, + const PermutationBase& permutation) +{ + return Product + (matrix.derived(), permutation.derived()); +} + +/** \returns the matrix with the permutation applied to the rows. + */ +template +EIGEN_DEVICE_FUNC +const Product +operator*(const PermutationBase &permutation, + const MatrixBase& matrix) +{ + return Product + (permutation.derived(), matrix.derived()); +} + + +template +class InverseImpl + : public EigenBase > +{ + typedef typename PermutationType::PlainPermutationType PlainPermutationType; + typedef internal::traits PermTraits; + protected: + InverseImpl() {} + public: + typedef Inverse InverseType; + using EigenBase >::derived; + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename PermutationType::DenseMatrixType DenseMatrixType; + enum { + RowsAtCompileTime = PermTraits::RowsAtCompileTime, + ColsAtCompileTime = PermTraits::ColsAtCompileTime, + MaxRowsAtCompileTime = PermTraits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = PermTraits::MaxColsAtCompileTime + }; + #endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + void evalTo(MatrixBase& other) const + { + other.setZero(); + for (Index i=0; i friend + const Product + operator*(const MatrixBase& matrix, const InverseType& trPerm) + { + return Product(matrix.derived(), trPerm.derived()); + } + + /** \returns the matrix with the inverse permutation applied to the rows. + */ + template + const Product + operator*(const MatrixBase& matrix) const + { + return Product(derived(), matrix.derived()); + } +}; + +template +const PermutationWrapper MatrixBase::asPermutation() const +{ + return derived(); +} + +namespace internal { + +template<> struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_PERMUTATIONMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h new file mode 100644 index 00000000000..0f3632cfd14 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h @@ -0,0 +1,1037 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2009 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSESTORAGEBASE_H +#define EIGEN_DENSESTORAGEBASE_H + +#if defined(EIGEN_INITIALIZE_MATRICES_BY_ZERO) +# define EIGEN_INITIALIZE_COEFFS +# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED for(int i=0;i::quiet_NaN(); +#else +# undef EIGEN_INITIALIZE_COEFFS +# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED +#endif + +namespace Eigen { + +namespace internal { + +template struct check_rows_cols_for_overflow { + template + EIGEN_DEVICE_FUNC + static EIGEN_ALWAYS_INLINE void run(Index, Index) + { + } +}; + +template<> struct check_rows_cols_for_overflow { + template + EIGEN_DEVICE_FUNC + static EIGEN_ALWAYS_INLINE void run(Index rows, Index cols) + { + // http://hg.mozilla.org/mozilla-central/file/6c8a909977d3/xpcom/ds/CheckedInt.h#l242 + // we assume Index is signed + Index max_index = (std::size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed + bool error = (rows == 0 || cols == 0) ? false + : (rows > max_index / cols); + if (error) + throw_std_bad_alloc(); + } +}; + +template +struct conservative_resize_like_impl; + +template struct matrix_swap_impl; + +} // end namespace internal + +#ifdef EIGEN_PARSED_BY_DOXYGEN +namespace doxygen { + +// This is a workaround to doxygen not being able to understand the inheritance logic +// when it is hidden by the dense_xpr_base helper struct. +// Moreover, doxygen fails to include members that are not documented in the declaration body of +// MatrixBase if we inherits MatrixBase >, +// this is why we simply inherits MatrixBase, though this does not make sense. + +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template struct dense_xpr_base_dispatcher; +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template +struct dense_xpr_base_dispatcher > + : public MatrixBase {}; +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template +struct dense_xpr_base_dispatcher > + : public ArrayBase {}; + +} // namespace doxygen + +/** \class PlainObjectBase + * \ingroup Core_Module + * \brief %Dense storage base class for matrices and arrays. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_PLAINOBJECTBASE_PLUGIN. + * + * \tparam Derived is the derived type, e.g., a Matrix or Array + * + * \sa \ref TopicClassHierarchy + */ +template +class PlainObjectBase : public doxygen::dense_xpr_base_dispatcher +#else +template +class PlainObjectBase : public internal::dense_xpr_base::type +#endif +{ + public: + enum { Options = internal::traits::Options }; + typedef typename internal::dense_xpr_base::type Base; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + typedef Derived DenseType; + + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + template friend class Eigen::Map; + friend class Eigen::Map; + typedef Eigen::Map MapType; + friend class Eigen::Map; + typedef const Eigen::Map ConstMapType; +#if EIGEN_MAX_ALIGN_BYTES>0 + // for EIGEN_MAX_ALIGN_BYTES==0, AlignedMax==Unaligned, and many compilers generate warnings for friend-ing a class twice. + friend class Eigen::Map; + friend class Eigen::Map; +#endif + typedef Eigen::Map AlignedMapType; + typedef const Eigen::Map ConstAlignedMapType; + template struct StridedMapType { typedef Eigen::Map type; }; + template struct StridedConstMapType { typedef Eigen::Map type; }; + template struct StridedAlignedMapType { typedef Eigen::Map type; }; + template struct StridedConstAlignedMapType { typedef Eigen::Map type; }; + + protected: + DenseStorage m_storage; + + public: + enum { NeedsToAlign = (SizeAtCompileTime != Dynamic) && (internal::traits::Alignment>0) }; + EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) + + EIGEN_DEVICE_FUNC + Base& base() { return *static_cast(this); } + EIGEN_DEVICE_FUNC + const Base& base() const { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { return m_storage.rows(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { return m_storage.cols(); } + + /** This is an overloaded version of DenseCoeffsBase::coeff(Index,Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeff(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeff(Index rowId, Index colId) const + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeff(Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeff(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeff(Index index) const + { + return m_storage.data()[index]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeffRef(Index,Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeffRef(Index,Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index rowId, Index colId) + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeffRef(Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeffRef(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index index) + { + return m_storage.data()[index]; + } + + /** This is the const version of coeffRef(Index,Index) which is thus synonym of coeff(Index,Index). + * It is provided for convenience. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeffRef(Index rowId, Index colId) const + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is the const version of coeffRef(Index) which is thus synonym of coeff(Index). + * It is provided for convenience. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeffRef(Index index) const + { + return m_storage.data()[index]; + } + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const + { + return internal::ploadt + (m_storage.data() + (Flags & RowMajorBit + ? colId + rowId * m_storage.cols() + : rowId + colId * m_storage.rows())); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketScalar packet(Index index) const + { + return internal::ploadt(m_storage.data() + index); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE void writePacket(Index rowId, Index colId, const PacketScalar& val) + { + internal::pstoret + (m_storage.data() + (Flags & RowMajorBit + ? colId + rowId * m_storage.cols() + : rowId + colId * m_storage.rows()), val); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE void writePacket(Index index, const PacketScalar& val) + { + internal::pstoret(m_storage.data() + index, val); + } + + /** \returns a const pointer to the data array of this matrix */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar *data() const + { return m_storage.data(); } + + /** \returns a pointer to the data array of this matrix */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar *data() + { return m_storage.data(); } + + /** Resizes \c *this to a \a rows x \a cols matrix. + * + * This method is intended for dynamic-size matrices, although it is legal to call it on any + * matrix as long as fixed dimensions are left unchanged. If you only want to change the number + * of rows and/or of columns, you can use resize(NoChange_t, Index), resize(Index, NoChange_t). + * + * If the current number of coefficients of \c *this exactly matches the + * product \a rows * \a cols, then no memory allocation is performed and + * the current values are left unchanged. In all other cases, including + * shrinking, the data is reallocated and all previous values are lost. + * + * Example: \include Matrix_resize_int_int.cpp + * Output: \verbinclude Matrix_resize_int_int.out + * + * \sa resize(Index) for vectors, resize(NoChange_t, Index), resize(Index, NoChange_t) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void resize(Index rows, Index cols) + { + eigen_assert( EIGEN_IMPLIES(RowsAtCompileTime!=Dynamic,rows==RowsAtCompileTime) + && EIGEN_IMPLIES(ColsAtCompileTime!=Dynamic,cols==ColsAtCompileTime) + && EIGEN_IMPLIES(RowsAtCompileTime==Dynamic && MaxRowsAtCompileTime!=Dynamic,rows<=MaxRowsAtCompileTime) + && EIGEN_IMPLIES(ColsAtCompileTime==Dynamic && MaxColsAtCompileTime!=Dynamic,cols<=MaxColsAtCompileTime) + && rows>=0 && cols>=0 && "Invalid sizes when resizing a matrix or array."); + internal::check_rows_cols_for_overflow::run(rows, cols); + #ifdef EIGEN_INITIALIZE_COEFFS + Index size = rows*cols; + bool size_changed = size != this->size(); + m_storage.resize(size, rows, cols); + if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + #else + m_storage.resize(rows*cols, rows, cols); + #endif + } + + /** Resizes \c *this to a vector of length \a size + * + * \only_for_vectors. This method does not work for + * partially dynamic matrices when the static dimension is anything other + * than 1. For example it will not work with Matrix. + * + * Example: \include Matrix_resize_int.cpp + * Output: \verbinclude Matrix_resize_int.out + * + * \sa resize(Index,Index), resize(NoChange_t, Index), resize(Index, NoChange_t) + */ + EIGEN_DEVICE_FUNC + inline void resize(Index size) + { + EIGEN_STATIC_ASSERT_VECTOR_ONLY(PlainObjectBase) + eigen_assert(((SizeAtCompileTime == Dynamic && (MaxSizeAtCompileTime==Dynamic || size<=MaxSizeAtCompileTime)) || SizeAtCompileTime == size) && size>=0); + #ifdef EIGEN_INITIALIZE_COEFFS + bool size_changed = size != this->size(); + #endif + if(RowsAtCompileTime == 1) + m_storage.resize(size, 1, size); + else + m_storage.resize(size, size, 1); + #ifdef EIGEN_INITIALIZE_COEFFS + if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + #endif + } + + /** Resizes the matrix, changing only the number of columns. For the parameter of type NoChange_t, just pass the special value \c NoChange + * as in the example below. + * + * Example: \include Matrix_resize_NoChange_int.cpp + * Output: \verbinclude Matrix_resize_NoChange_int.out + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + inline void resize(NoChange_t, Index cols) + { + resize(rows(), cols); + } + + /** Resizes the matrix, changing only the number of rows. For the parameter of type NoChange_t, just pass the special value \c NoChange + * as in the example below. + * + * Example: \include Matrix_resize_int_NoChange.cpp + * Output: \verbinclude Matrix_resize_int_NoChange.out + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + inline void resize(Index rows, NoChange_t) + { + resize(rows, cols()); + } + + /** Resizes \c *this to have the same dimensions as \a other. + * Takes care of doing all the checking that's needed. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void resizeLike(const EigenBase& _other) + { + const OtherDerived& other = _other.derived(); + internal::check_rows_cols_for_overflow::run(other.rows(), other.cols()); + const Index othersize = other.rows()*other.cols(); + if(RowsAtCompileTime == 1) + { + eigen_assert(other.rows() == 1 || other.cols() == 1); + resize(1, othersize); + } + else if(ColsAtCompileTime == 1) + { + eigen_assert(other.rows() == 1 || other.cols() == 1); + resize(othersize, 1); + } + else resize(other.rows(), other.cols()); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * The method is intended for matrices of dynamic size. If you only want to change the number + * of rows and/or of columns, you can use conservativeResize(NoChange_t, Index) or + * conservativeResize(Index, NoChange_t). + * + * Matrices are resized relative to the top-left element. In case values need to be + * appended to the matrix they will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index rows, Index cols) + { + internal::conservative_resize_like_impl::run(*this, rows, cols); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * As opposed to conservativeResize(Index rows, Index cols), this version leaves + * the number of columns unchanged. + * + * In case the matrix is growing, new rows will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index rows, NoChange_t) + { + // Note: see the comment in conservativeResize(Index,Index) + conservativeResize(rows, cols()); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * As opposed to conservativeResize(Index rows, Index cols), this version leaves + * the number of rows unchanged. + * + * In case the matrix is growing, new columns will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index cols) + { + // Note: see the comment in conservativeResize(Index,Index) + conservativeResize(rows(), cols); + } + + /** Resizes the vector to \a size while retaining old values. + * + * \only_for_vectors. This method does not work for + * partially dynamic matrices when the static dimension is anything other + * than 1. For example it will not work with Matrix. + * + * When values are appended, they will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index size) + { + internal::conservative_resize_like_impl::run(*this, size); + } + + /** Resizes the matrix to \a rows x \a cols of \c other, while leaving old values untouched. + * + * The method is intended for matrices of dynamic size. If you only want to change the number + * of rows and/or of columns, you can use conservativeResize(NoChange_t, Index) or + * conservativeResize(Index, NoChange_t). + * + * Matrices are resized relative to the top-left element. In case values need to be + * appended to the matrix they will copied from \c other. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResizeLike(const DenseBase& other) + { + internal::conservative_resize_like_impl::run(*this, other); + } + + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const PlainObjectBase& other) + { + return _set(other); + } + + /** \sa MatrixBase::lazyAssign() */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& lazyAssign(const DenseBase& other) + { + _resize_to_match(other); + return Base::lazyAssign(other.derived()); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const ReturnByValue& func) + { + resize(func.rows(), func.cols()); + return Base::operator=(func); + } + + // Prevent user from trying to instantiate PlainObjectBase objects + // by making all its constructor protected. See bug 1074. + protected: + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase() : m_storage() + { +// _check_template_params(); +// EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + // FIXME is it still needed ? + /** \internal */ + EIGEN_DEVICE_FUNC + explicit PlainObjectBase(internal::constructor_without_unaligned_array_assert) + : m_storage(internal::constructor_without_unaligned_array_assert()) + { +// _check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } +#endif + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + PlainObjectBase(PlainObjectBase&& other) EIGEN_NOEXCEPT + : m_storage( std::move(other.m_storage) ) + { + } + + EIGEN_DEVICE_FUNC + PlainObjectBase& operator=(PlainObjectBase&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_storage, other.m_storage); + return *this; + } +#endif + + /** Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const PlainObjectBase& other) + : Base(), m_storage(other.m_storage) { } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(Index size, Index rows, Index cols) + : m_storage(size, rows, cols) + { +// _check_template_params(); +// EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + + /** \sa PlainObjectBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase &other) + : m_storage() + { + _check_template_params(); + resizeLike(other); + _set_noalias(other); + } + + /** \sa PlainObjectBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const EigenBase &other) + : m_storage() + { + _check_template_params(); + resizeLike(other); + *this = other.derived(); + } + /** \brief Copy constructor with in-place evaluation */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const ReturnByValue& other) + { + _check_template_params(); + // FIXME this does not automatically transpose vectors if necessary + resize(other.rows(), other.cols()); + other.evalTo(this->derived()); + } + + public: + + /** \brief Copies the generic expression \a other into *this. + * \copydetails DenseBase::operator=(const EigenBase &other) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const EigenBase &other) + { + _resize_to_match(other); + Base::operator=(other.derived()); + return this->derived(); + } + + /** \name Map + * These are convenience functions returning Map objects. The Map() static functions return unaligned Map objects, + * while the AlignedMap() functions return aligned Map objects and thus should be called only with 16-byte-aligned + * \a data pointers. + * + * Here is an example using strides: + * \include Matrix_Map_stride.cpp + * Output: \verbinclude Matrix_Map_stride.out + * + * \see class Map + */ + //@{ + static inline ConstMapType Map(const Scalar* data) + { return ConstMapType(data); } + static inline MapType Map(Scalar* data) + { return MapType(data); } + static inline ConstMapType Map(const Scalar* data, Index size) + { return ConstMapType(data, size); } + static inline MapType Map(Scalar* data, Index size) + { return MapType(data, size); } + static inline ConstMapType Map(const Scalar* data, Index rows, Index cols) + { return ConstMapType(data, rows, cols); } + static inline MapType Map(Scalar* data, Index rows, Index cols) + { return MapType(data, rows, cols); } + + static inline ConstAlignedMapType MapAligned(const Scalar* data) + { return ConstAlignedMapType(data); } + static inline AlignedMapType MapAligned(Scalar* data) + { return AlignedMapType(data); } + static inline ConstAlignedMapType MapAligned(const Scalar* data, Index size) + { return ConstAlignedMapType(data, size); } + static inline AlignedMapType MapAligned(Scalar* data, Index size) + { return AlignedMapType(data, size); } + static inline ConstAlignedMapType MapAligned(const Scalar* data, Index rows, Index cols) + { return ConstAlignedMapType(data, rows, cols); } + static inline AlignedMapType MapAligned(Scalar* data, Index rows, Index cols) + { return AlignedMapType(data, rows, cols); } + + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, const Stride& stride) + { return typename StridedConstMapType >::type(data, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, const Stride& stride) + { return typename StridedMapType >::type(data, stride); } + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, Index size, const Stride& stride) + { return typename StridedConstMapType >::type(data, size, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, Index size, const Stride& stride) + { return typename StridedMapType >::type(data, size, stride); } + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedConstMapType >::type(data, rows, cols, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedMapType >::type(data, rows, cols, stride); } + + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, stride); } + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, Index size, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, size, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, Index size, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, size, stride); } + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, rows, cols, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, rows, cols, stride); } + //@} + + using Base::setConstant; + EIGEN_DEVICE_FUNC Derived& setConstant(Index size, const Scalar& val); + EIGEN_DEVICE_FUNC Derived& setConstant(Index rows, Index cols, const Scalar& val); + + using Base::setZero; + EIGEN_DEVICE_FUNC Derived& setZero(Index size); + EIGEN_DEVICE_FUNC Derived& setZero(Index rows, Index cols); + + using Base::setOnes; + EIGEN_DEVICE_FUNC Derived& setOnes(Index size); + EIGEN_DEVICE_FUNC Derived& setOnes(Index rows, Index cols); + + using Base::setRandom; + Derived& setRandom(Index size); + Derived& setRandom(Index rows, Index cols); + + #ifdef EIGEN_PLAINOBJECTBASE_PLUGIN + #include EIGEN_PLAINOBJECTBASE_PLUGIN + #endif + + protected: + /** \internal Resizes *this in preparation for assigning \a other to it. + * Takes care of doing all the checking that's needed. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _resize_to_match(const EigenBase& other) + { + #ifdef EIGEN_NO_AUTOMATIC_RESIZING + eigen_assert((this->size()==0 || (IsVectorAtCompileTime ? (this->size() == other.size()) + : (rows() == other.rows() && cols() == other.cols()))) + && "Size mismatch. Automatic resizing is disabled because EIGEN_NO_AUTOMATIC_RESIZING is defined"); + EIGEN_ONLY_USED_FOR_DEBUG(other); + #else + resizeLike(other); + #endif + } + + /** + * \brief Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + * + * \sa operator=(const MatrixBase&), _set_noalias() + * + * \internal + */ + // aliasing is dealt once in internall::call_assignment + // so at this stage we have to assume aliasing... and resising has to be done later. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& _set(const DenseBase& other) + { + internal::call_assignment(this->derived(), other.derived()); + return this->derived(); + } + + /** \internal Like _set() but additionally makes the assumption that no aliasing effect can happen (which + * is the case when creating a new matrix) so one can enforce lazy evaluation. + * + * \sa operator=(const MatrixBase&), _set() + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& _set_noalias(const DenseBase& other) + { + // I don't think we need this resize call since the lazyAssign will anyways resize + // and lazyAssign will be called by the assign selector. + //_resize_to_match(other); + // the 'false' below means to enforce lazy evaluation. We don't use lazyAssign() because + // it wouldn't allow to copy a row-vector into a column-vector. + internal::call_assignment_no_alias(this->derived(), other.derived(), internal::assign_op()); + return this->derived(); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if::type* = 0) + { + const bool t0_is_integer_alike = internal::is_valid_index_type::value; + const bool t1_is_integer_alike = internal::is_valid_index_type::value; + EIGEN_STATIC_ASSERT(t0_is_integer_alike && + t1_is_integer_alike, + FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) + resize(rows,cols); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(const T0& val0, const T1& val1, typename internal::enable_if::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2) + m_storage.data()[0] = Scalar(val0); + m_storage.data()[1] = Scalar(val1); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(const Index& val0, const Index& val1, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime==2,T1>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2) + m_storage.data()[0] = Scalar(val0); + m_storage.data()[1] = Scalar(val1); + } + + // The argument is convertible to the Index type and we either have a non 1x1 Matrix, or a dynamic-sized Array, + // then the argument is meant to be the size of the object. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(Index size, typename internal::enable_if< (Base::SizeAtCompileTime!=1 || !internal::is_convertible::value) + && ((!internal::is_same::XprKind,ArrayXpr>::value || Base::SizeAtCompileTime==Dynamic)),T>::type* = 0) + { + // NOTE MSVC 2008 complains if we directly put bool(NumTraits::IsInteger) as the EIGEN_STATIC_ASSERT argument. + const bool is_integer_alike = internal::is_valid_index_type::value; + EIGEN_UNUSED_VARIABLE(is_integer_alike); + EIGEN_STATIC_ASSERT(is_integer_alike, + FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) + resize(size); + } + + // We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type can be implicitely converted) + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar& val0, typename internal::enable_if::value,T>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1) + m_storage.data()[0] = val0; + } + + // We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type match the index type) + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Index& val0, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime==1 + && internal::is_convertible::value,T*>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1) + m_storage.data()[0] = Scalar(val0); + } + + // Initialize a fixed size matrix from a pointer to raw data + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar* data){ + this->_set_noalias(ConstMapType(data)); + } + + // Initialize an arbitrary matrix from a dense expression + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const DenseBase& other){ + this->_set_noalias(other); + } + + // Initialize an arbitrary matrix from an object convertible to the Derived type. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Derived& other){ + this->_set_noalias(other); + } + + // Initialize an arbitrary matrix from a generic Eigen expression + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const EigenBase& other){ + this->derived() = other; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const ReturnByValue& other) + { + resize(other.rows(), other.cols()); + other.evalTo(this->derived()); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const RotationBase& r) + { + this->derived() = r; + } + + // For fixed-size Array + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar& val0, + typename internal::enable_if< Base::SizeAtCompileTime!=Dynamic + && Base::SizeAtCompileTime!=1 + && internal::is_convertible::value + && internal::is_same::XprKind,ArrayXpr>::value,T>::type* = 0) + { + Base::setConstant(val0); + } + + // For fixed-size Array + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Index& val0, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime!=Dynamic + && Base::SizeAtCompileTime!=1 + && internal::is_convertible::value + && internal::is_same::XprKind,ArrayXpr>::value,T*>::type* = 0) + { + Base::setConstant(val0); + } + + template + friend struct internal::matrix_swap_impl; + + public: + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal + * \brief Override DenseBase::swap() since for dynamic-sized matrices + * of same type it is enough to swap the data pointers. + */ + template + EIGEN_DEVICE_FUNC + void swap(DenseBase & other) + { + enum { SwapPointers = internal::is_same::value && Base::SizeAtCompileTime==Dynamic }; + internal::matrix_swap_impl::run(this->derived(), other.derived()); + } + + /** \internal + * \brief const version forwarded to DenseBase::swap + */ + template + EIGEN_DEVICE_FUNC + void swap(DenseBase const & other) + { Base::swap(other.derived()); } + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void _check_template_params() + { + EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, (Options&RowMajor)==RowMajor) + && EIGEN_IMPLIES(MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1, (Options&RowMajor)==0) + && ((RowsAtCompileTime == Dynamic) || (RowsAtCompileTime >= 0)) + && ((ColsAtCompileTime == Dynamic) || (ColsAtCompileTime >= 0)) + && ((MaxRowsAtCompileTime == Dynamic) || (MaxRowsAtCompileTime >= 0)) + && ((MaxColsAtCompileTime == Dynamic) || (MaxColsAtCompileTime >= 0)) + && (MaxRowsAtCompileTime == RowsAtCompileTime || RowsAtCompileTime==Dynamic) + && (MaxColsAtCompileTime == ColsAtCompileTime || ColsAtCompileTime==Dynamic) + && (Options & (DontAlign|RowMajor)) == Options), + INVALID_MATRIX_TEMPLATE_PARAMETERS) + } + + enum { IsPlainObjectBase = 1 }; +#endif +}; + +namespace internal { + +template +struct conservative_resize_like_impl +{ + static void run(DenseBase& _this, Index rows, Index cols) + { + if (_this.rows() == rows && _this.cols() == cols) return; + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(Derived) + + if ( ( Derived::IsRowMajor && _this.cols() == cols) || // row-major and we change only the number of rows + (!Derived::IsRowMajor && _this.rows() == rows) ) // column-major and we change only the number of columns + { + internal::check_rows_cols_for_overflow::run(rows, cols); + _this.derived().m_storage.conservativeResize(rows*cols,rows,cols); + } + else + { + // The storage order does not allow us to use reallocation. + typename Derived::PlainObject tmp(rows,cols); + const Index common_rows = numext::mini(rows, _this.rows()); + const Index common_cols = numext::mini(cols, _this.cols()); + tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols); + _this.derived().swap(tmp); + } + } + + static void run(DenseBase& _this, const DenseBase& other) + { + if (_this.rows() == other.rows() && _this.cols() == other.cols()) return; + + // Note: Here is space for improvement. Basically, for conservativeResize(Index,Index), + // neither RowsAtCompileTime or ColsAtCompileTime must be Dynamic. If only one of the + // dimensions is dynamic, one could use either conservativeResize(Index rows, NoChange_t) or + // conservativeResize(NoChange_t, Index cols). For these methods new static asserts like + // EIGEN_STATIC_ASSERT_DYNAMIC_ROWS and EIGEN_STATIC_ASSERT_DYNAMIC_COLS would be good. + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(Derived) + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(OtherDerived) + + if ( ( Derived::IsRowMajor && _this.cols() == other.cols()) || // row-major and we change only the number of rows + (!Derived::IsRowMajor && _this.rows() == other.rows()) ) // column-major and we change only the number of columns + { + const Index new_rows = other.rows() - _this.rows(); + const Index new_cols = other.cols() - _this.cols(); + _this.derived().m_storage.conservativeResize(other.size(),other.rows(),other.cols()); + if (new_rows>0) + _this.bottomRightCorner(new_rows, other.cols()) = other.bottomRows(new_rows); + else if (new_cols>0) + _this.bottomRightCorner(other.rows(), new_cols) = other.rightCols(new_cols); + } + else + { + // The storage order does not allow us to use reallocation. + typename Derived::PlainObject tmp(other); + const Index common_rows = numext::mini(tmp.rows(), _this.rows()); + const Index common_cols = numext::mini(tmp.cols(), _this.cols()); + tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols); + _this.derived().swap(tmp); + } + } +}; + +// Here, the specialization for vectors inherits from the general matrix case +// to allow calling .conservativeResize(rows,cols) on vectors. +template +struct conservative_resize_like_impl + : conservative_resize_like_impl +{ + using conservative_resize_like_impl::run; + + static void run(DenseBase& _this, Index size) + { + const Index new_rows = Derived::RowsAtCompileTime==1 ? 1 : size; + const Index new_cols = Derived::RowsAtCompileTime==1 ? size : 1; + _this.derived().m_storage.conservativeResize(size,new_rows,new_cols); + } + + static void run(DenseBase& _this, const DenseBase& other) + { + if (_this.rows() == other.rows() && _this.cols() == other.cols()) return; + + const Index num_new_elements = other.size() - _this.size(); + + const Index new_rows = Derived::RowsAtCompileTime==1 ? 1 : other.rows(); + const Index new_cols = Derived::RowsAtCompileTime==1 ? other.cols() : 1; + _this.derived().m_storage.conservativeResize(other.size(),new_rows,new_cols); + + if (num_new_elements > 0) + _this.tail(num_new_elements) = other.tail(num_new_elements); + } +}; + +template +struct matrix_swap_impl +{ + EIGEN_DEVICE_FUNC + static inline void run(MatrixTypeA& a, MatrixTypeB& b) + { + a.base().swap(b); + } +}; + +template +struct matrix_swap_impl +{ + EIGEN_DEVICE_FUNC + static inline void run(MatrixTypeA& a, MatrixTypeB& b) + { + static_cast(a).m_storage.swap(static_cast(b).m_storage); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DENSESTORAGEBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Product.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Product.h new file mode 100644 index 00000000000..676c4802779 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Product.h @@ -0,0 +1,186 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2011 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PRODUCT_H +#define EIGEN_PRODUCT_H + +namespace Eigen { + +template class ProductImpl; + +namespace internal { + +template +struct traits > +{ + typedef typename remove_all::type LhsCleaned; + typedef typename remove_all::type RhsCleaned; + typedef traits LhsTraits; + typedef traits RhsTraits; + + typedef MatrixXpr XprKind; + + typedef typename ScalarBinaryOpTraits::Scalar, typename traits::Scalar>::ReturnType Scalar; + typedef typename product_promote_storage_type::ret>::ret StorageKind; + typedef typename promote_index_type::type StorageIndex; + + enum { + RowsAtCompileTime = LhsTraits::RowsAtCompileTime, + ColsAtCompileTime = RhsTraits::ColsAtCompileTime, + MaxRowsAtCompileTime = LhsTraits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = RhsTraits::MaxColsAtCompileTime, + + // FIXME: only needed by GeneralMatrixMatrixTriangular + InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsTraits::ColsAtCompileTime, RhsTraits::RowsAtCompileTime), + + // The storage order is somewhat arbitrary here. The correct one will be determined through the evaluator. + Flags = (MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1) ? RowMajorBit + : (MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1) ? 0 + : ( ((LhsTraits::Flags&NoPreferredStorageOrderBit) && (RhsTraits::Flags&RowMajorBit)) + || ((RhsTraits::Flags&NoPreferredStorageOrderBit) && (LhsTraits::Flags&RowMajorBit)) ) ? RowMajorBit + : NoPreferredStorageOrderBit + }; +}; + +} // end namespace internal + +/** \class Product + * \ingroup Core_Module + * + * \brief Expression of the product of two arbitrary matrices or vectors + * + * \tparam _Lhs the type of the left-hand side expression + * \tparam _Rhs the type of the right-hand side expression + * + * This class represents an expression of the product of two arbitrary matrices. + * + * The other template parameters are: + * \tparam Option can be DefaultProduct, AliasFreeProduct, or LazyProduct + * + */ +template +class Product : public ProductImpl<_Lhs,_Rhs,Option, + typename internal::product_promote_storage_type::StorageKind, + typename internal::traits<_Rhs>::StorageKind, + internal::product_type<_Lhs,_Rhs>::ret>::ret> +{ + public: + + typedef _Lhs Lhs; + typedef _Rhs Rhs; + + typedef typename ProductImpl< + Lhs, Rhs, Option, + typename internal::product_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + internal::product_type::ret>::ret>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(Product) + + typedef typename internal::ref_selector::type LhsNested; + typedef typename internal::ref_selector::type RhsNested; + typedef typename internal::remove_all::type LhsNestedCleaned; + typedef typename internal::remove_all::type RhsNestedCleaned; + + EIGEN_DEVICE_FUNC Product(const Lhs& lhs, const Rhs& rhs) : m_lhs(lhs), m_rhs(rhs) + { + eigen_assert(lhs.cols() == rhs.rows() + && "invalid matrix product" + && "if you wanted a coeff-wise or a dot product use the respective explicit functions"); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index rows() const { return m_lhs.rows(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index cols() const { return m_rhs.cols(); } + + EIGEN_DEVICE_FUNC const LhsNestedCleaned& lhs() const { return m_lhs; } + EIGEN_DEVICE_FUNC const RhsNestedCleaned& rhs() const { return m_rhs; } + + protected: + + LhsNested m_lhs; + RhsNested m_rhs; +}; + +namespace internal { + +template::ret> +class dense_product_base + : public internal::dense_xpr_base >::type +{}; + +/** Convertion to scalar for inner-products */ +template +class dense_product_base + : public internal::dense_xpr_base >::type +{ + typedef Product ProductXpr; + typedef typename internal::dense_xpr_base::type Base; +public: + using Base::derived; + typedef typename Base::Scalar Scalar; + + EIGEN_STRONG_INLINE operator const Scalar() const + { + return internal::evaluator(derived()).coeff(0,0); + } +}; + +} // namespace internal + +// Generic API dispatcher +template +class ProductImpl : public internal::generic_xpr_base, MatrixXpr, StorageKind>::type +{ + public: + typedef typename internal::generic_xpr_base, MatrixXpr, StorageKind>::type Base; +}; + +template +class ProductImpl + : public internal::dense_product_base +{ + typedef Product Derived; + + public: + + typedef typename internal::dense_product_base Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Derived) + protected: + enum { + IsOneByOne = (RowsAtCompileTime == 1 || RowsAtCompileTime == Dynamic) && + (ColsAtCompileTime == 1 || ColsAtCompileTime == Dynamic), + EnableCoeff = IsOneByOne || Option==LazyProduct + }; + + public: + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index row, Index col) const + { + EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS); + eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) ); + + return internal::evaluator(derived()).coeff(row,col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index i) const + { + EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS); + eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) ); + + return internal::evaluator(derived()).coeff(i); + } + + +}; + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h new file mode 100644 index 00000000000..bce1310c96f --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h @@ -0,0 +1,1138 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008-2010 Gael Guennebaud +// Copyright (C) 2011 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + + +#ifndef EIGEN_PRODUCTEVALUATORS_H +#define EIGEN_PRODUCTEVALUATORS_H + +namespace Eigen { + +namespace internal { + +/** \internal + * Evaluator of a product expression. + * Since products require special treatments to handle all possible cases, + * we simply deffer the evaluation logic to a product_evaluator class + * which offers more partial specialization possibilities. + * + * \sa class product_evaluator + */ +template +struct evaluator > + : public product_evaluator > +{ + typedef Product XprType; + typedef product_evaluator Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +// Catch "scalar * ( A * B )" and transform it to "(A*scalar) * B" +// TODO we should apply that rule only if that's really helpful +template +struct evaluator_assume_aliasing, + const CwiseNullaryOp, Plain1>, + const Product > > +{ + static const bool value = true; +}; +template +struct evaluator, + const CwiseNullaryOp, Plain1>, + const Product > > + : public evaluator > +{ + typedef CwiseBinaryOp, + const CwiseNullaryOp, Plain1>, + const Product > XprType; + typedef evaluator > Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) + : Base(xpr.lhs().functor().m_other * xpr.rhs().lhs() * xpr.rhs().rhs()) + {} +}; + + +template +struct evaluator, DiagIndex> > + : public evaluator, DiagIndex> > +{ + typedef Diagonal, DiagIndex> XprType; + typedef evaluator, DiagIndex> > Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) + : Base(Diagonal, DiagIndex>( + Product(xpr.nestedExpression().lhs(), xpr.nestedExpression().rhs()), + xpr.index() )) + {} +}; + + +// Helper class to perform a matrix product with the destination at hand. +// Depending on the sizes of the factors, there are different evaluation strategies +// as controlled by internal::product_type. +template< typename Lhs, typename Rhs, + typename LhsShape = typename evaluator_traits::Shape, + typename RhsShape = typename evaluator_traits::Shape, + int ProductType = internal::product_type::value> +struct generic_product_impl; + +template +struct evaluator_assume_aliasing > { + static const bool value = true; +}; + +// This is the default evaluator implementation for products: +// It creates a temporary and call generic_product_impl +template +struct product_evaluator, ProductTag, LhsShape, RhsShape> + : public evaluator::PlainObject> +{ + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + typedef evaluator Base; + enum { + Flags = Base::Flags | EvalBeforeNestingBit + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit product_evaluator(const XprType& xpr) + : m_result(xpr.rows(), xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + +// FIXME shall we handle nested_eval here?, +// if so, then we must take care at removing the call to nested_eval in the specializations (e.g., in permutation_matrix_product, transposition_matrix_product, etc.) +// typedef typename internal::nested_eval::type LhsNested; +// typedef typename internal::nested_eval::type RhsNested; +// typedef typename internal::remove_all::type LhsNestedCleaned; +// typedef typename internal::remove_all::type RhsNestedCleaned; +// +// const LhsNested lhs(xpr.lhs()); +// const RhsNested rhs(xpr.rhs()); +// +// generic_product_impl::evalTo(m_result, lhs, rhs); + + generic_product_impl::evalTo(m_result, xpr.lhs(), xpr.rhs()); + } + +protected: + PlainObject m_result; +}; + +// The following three shortcuts are enabled only if the scalar types match excatly. +// TODO: we could enable them for different scalar types when the product is not vectorized. + +// Dense = Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + // FIXME shall we handle nested_eval here? + generic_product_impl::evalTo(dst, src.lhs(), src.rhs()); + } +}; + +// Dense += Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::add_assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &) + { + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + // FIXME shall we handle nested_eval here? + generic_product_impl::addTo(dst, src.lhs(), src.rhs()); + } +}; + +// Dense -= Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::sub_assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &) + { + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + // FIXME shall we handle nested_eval here? + generic_product_impl::subTo(dst, src.lhs(), src.rhs()); + } +}; + + +// Dense ?= scalar * Product +// TODO we should apply that rule if that's really helpful +// for instance, this is not good for inner products +template< typename DstXprType, typename Lhs, typename Rhs, typename AssignFunc, typename Scalar, typename ScalarBis, typename Plain> +struct Assignment, const CwiseNullaryOp,Plain>, + const Product >, AssignFunc, Dense2Dense> +{ + typedef CwiseBinaryOp, + const CwiseNullaryOp,Plain>, + const Product > SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const AssignFunc& func) + { + call_assignment_no_alias(dst, (src.lhs().functor().m_other * src.rhs().lhs())*src.rhs().rhs(), func); + } +}; + +//---------------------------------------- +// Catch "Dense ?= xpr + Product<>" expression to save one temporary +// FIXME we could probably enable these rules for any product, i.e., not only Dense and DefaultProduct + +template +struct evaluator_assume_aliasing::Scalar>, const OtherXpr, + const Product >, DenseShape > { + static const bool value = true; +}; + +template +struct evaluator_assume_aliasing::Scalar>, const OtherXpr, + const Product >, DenseShape > { + static const bool value = true; +}; + +template +struct assignment_from_xpr_op_product +{ + template + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const InitialFunc& /*func*/) + { + call_assignment_no_alias(dst, src.lhs(), Func1()); + call_assignment_no_alias(dst, src.rhs(), Func2()); + } +}; + +#define EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(ASSIGN_OP,BINOP,ASSIGN_OP2) \ + template< typename DstXprType, typename OtherXpr, typename Lhs, typename Rhs, typename DstScalar, typename SrcScalar, typename OtherScalar,typename ProdScalar> \ + struct Assignment, const OtherXpr, \ + const Product >, internal::ASSIGN_OP, Dense2Dense> \ + : assignment_from_xpr_op_product, internal::ASSIGN_OP, internal::ASSIGN_OP2 > \ + {} + +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(assign_op, scalar_sum_op,add_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(add_assign_op,scalar_sum_op,add_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(sub_assign_op,scalar_sum_op,sub_assign_op); + +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(assign_op, scalar_difference_op,sub_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(add_assign_op,scalar_difference_op,sub_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(sub_assign_op,scalar_difference_op,add_assign_op); + +//---------------------------------------- + +template +struct generic_product_impl +{ + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + dst.coeffRef(0,0) = (lhs.transpose().cwiseProduct(rhs)).sum(); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + dst.coeffRef(0,0) += (lhs.transpose().cwiseProduct(rhs)).sum(); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { dst.coeffRef(0,0) -= (lhs.transpose().cwiseProduct(rhs)).sum(); } +}; + + +/*********************************************************************** +* Implementation of outer dense * dense vector product +***********************************************************************/ + +// Column major result +template +void outer_product_selector_run(Dst& dst, const Lhs &lhs, const Rhs &rhs, const Func& func, const false_type&) +{ + evaluator rhsEval(rhs); + typename nested_eval::type actual_lhs(lhs); + // FIXME if cols is large enough, then it might be useful to make sure that lhs is sequentially stored + // FIXME not very good if rhs is real and lhs complex while alpha is real too + const Index cols = dst.cols(); + for (Index j=0; j +void outer_product_selector_run(Dst& dst, const Lhs &lhs, const Rhs &rhs, const Func& func, const true_type&) +{ + evaluator lhsEval(lhs); + typename nested_eval::type actual_rhs(rhs); + // FIXME if rows is large enough, then it might be useful to make sure that rhs is sequentially stored + // FIXME not very good if lhs is real and rhs complex while alpha is real too + const Index rows = dst.rows(); + for (Index i=0; i +struct generic_product_impl +{ + template struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {}; + typedef typename Product::Scalar Scalar; + + // TODO it would be nice to be able to exploit our *_assign_op functors for that purpose + struct set { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() = src; } }; + struct add { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() += src; } }; + struct sub { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() -= src; } }; + struct adds { + Scalar m_scale; + explicit adds(const Scalar& s) : m_scale(s) {} + template void operator()(const Dst& dst, const Src& src) const { + dst.const_cast_derived() += m_scale * src; + } + }; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, set(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, add(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, sub(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + internal::outer_product_selector_run(dst, lhs, rhs, adds(alpha), is_row_major()); + } + +}; + + +// This base class provides default implementations for evalTo, addTo, subTo, in terms of scaleAndAddTo +template +struct generic_product_impl_base +{ + typedef typename Product::Scalar Scalar; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { dst.setZero(); scaleAndAddTo(dst, lhs, rhs, Scalar(1)); } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { scaleAndAddTo(dst,lhs, rhs, Scalar(1)); } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); } + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { Derived::scaleAndAddTo(dst,lhs,rhs,alpha); } + +}; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename nested_eval::type LhsNested; + typedef typename nested_eval::type RhsNested; + typedef typename Product::Scalar Scalar; + enum { Side = Lhs::IsVectorAtCompileTime ? OnTheLeft : OnTheRight }; + typedef typename internal::remove_all::type>::type MatrixType; + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + LhsNested actual_lhs(lhs); + RhsNested actual_rhs(rhs); + internal::gemv_dense_selector::HasUsableDirectAccess) + >::run(actual_lhs, actual_rhs, dst, alpha); + } +}; + +template +struct generic_product_impl +{ + typedef typename Product::Scalar Scalar; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // Same as: dst.noalias() = lhs.lazyProduct(rhs); + // but easier on the compiler side + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::assign_op()); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // dst.noalias() += lhs.lazyProduct(rhs); + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::add_assign_op()); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // dst.noalias() -= lhs.lazyProduct(rhs); + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::sub_assign_op()); + } + + // Catch "dst {,+,-}= (s*A)*B" and evaluate it lazily by moving out the scalar factor: + // dst {,+,-}= s * (A.lazyProduct(B)) + // This is a huge benefit for heap-allocated matrix types as it save one costly allocation. + // For them, this strategy is also faster than simply by-passing the heap allocation through + // stack allocation. + // For fixed sizes matrices, this is less obvious, it is sometimes x2 faster, but sometimes x3 slower, + // and the behavior depends also a lot on the compiler... so let's be conservative and enable them for dynamic-size only, + // that is when coming from generic_product_impl<...,GemmProduct> in file GeneralMatrixMatrix.h + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const CwiseBinaryOp, + const CwiseNullaryOp, Plain1>, Xpr2>& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lhs().functor().m_other * lhs.rhs().lazyProduct(rhs), func); + } + + // Here, we we always have LhsT==Lhs, but we need to make it a template type to make the above + // overload more specialized. + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const LhsT& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), func); + } + + +// template +// static inline void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) +// { dst.noalias() += alpha * lhs.lazyProduct(rhs); } +}; + +// This specialization enforces the use of a coefficient-based evaluation strategy +template +struct generic_product_impl + : generic_product_impl {}; + +// Case 2: Evaluate coeff by coeff +// +// This is mostly taken from CoeffBasedProduct.h +// The main difference is that we add an extra argument to the etor_product_*_impl::run() function +// for the inner dimension of the product, because evaluator object do not know their size. + +template +struct etor_product_coeff_impl; + +template +struct etor_product_packet_impl; + +template +struct product_evaluator, ProductTag, DenseShape, DenseShape> + : evaluator_base > +{ + typedef Product XprType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit product_evaluator(const XprType& xpr) + : m_lhs(xpr.lhs()), + m_rhs(xpr.rhs()), + m_lhsImpl(m_lhs), // FIXME the creation of the evaluator objects should result in a no-op, but check that! + m_rhsImpl(m_rhs), // Moreover, they are only useful for the packet path, so we could completely disable them when not needed, + // or perhaps declare them on the fly on the packet method... We have experiment to check what's best. + m_innerDim(xpr.lhs().cols()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::MulCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::AddCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); +#if 0 + std::cerr << "LhsOuterStrideBytes= " << LhsOuterStrideBytes << "\n"; + std::cerr << "RhsOuterStrideBytes= " << RhsOuterStrideBytes << "\n"; + std::cerr << "LhsAlignment= " << LhsAlignment << "\n"; + std::cerr << "RhsAlignment= " << RhsAlignment << "\n"; + std::cerr << "CanVectorizeLhs= " << CanVectorizeLhs << "\n"; + std::cerr << "CanVectorizeRhs= " << CanVectorizeRhs << "\n"; + std::cerr << "CanVectorizeInner= " << CanVectorizeInner << "\n"; + std::cerr << "EvalToRowMajor= " << EvalToRowMajor << "\n"; + std::cerr << "Alignment= " << Alignment << "\n"; + std::cerr << "Flags= " << Flags << "\n"; +#endif + } + + // Everything below here is taken from CoeffBasedProduct.h + + typedef typename internal::nested_eval::type LhsNested; + typedef typename internal::nested_eval::type RhsNested; + + typedef typename internal::remove_all::type LhsNestedCleaned; + typedef typename internal::remove_all::type RhsNestedCleaned; + + typedef evaluator LhsEtorType; + typedef evaluator RhsEtorType; + + enum { + RowsAtCompileTime = LhsNestedCleaned::RowsAtCompileTime, + ColsAtCompileTime = RhsNestedCleaned::ColsAtCompileTime, + InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsNestedCleaned::ColsAtCompileTime, RhsNestedCleaned::RowsAtCompileTime), + MaxRowsAtCompileTime = LhsNestedCleaned::MaxRowsAtCompileTime, + MaxColsAtCompileTime = RhsNestedCleaned::MaxColsAtCompileTime + }; + + typedef typename find_best_packet::type LhsVecPacketType; + typedef typename find_best_packet::type RhsVecPacketType; + + enum { + + LhsCoeffReadCost = LhsEtorType::CoeffReadCost, + RhsCoeffReadCost = RhsEtorType::CoeffReadCost, + CoeffReadCost = InnerSize==0 ? NumTraits::ReadCost + : InnerSize == Dynamic ? HugeCost + : InnerSize * (NumTraits::MulCost + LhsCoeffReadCost + RhsCoeffReadCost) + + (InnerSize - 1) * NumTraits::AddCost, + + Unroll = CoeffReadCost <= EIGEN_UNROLLING_LIMIT, + + LhsFlags = LhsEtorType::Flags, + RhsFlags = RhsEtorType::Flags, + + LhsRowMajor = LhsFlags & RowMajorBit, + RhsRowMajor = RhsFlags & RowMajorBit, + + LhsVecPacketSize = unpacket_traits::size, + RhsVecPacketSize = unpacket_traits::size, + + // Here, we don't care about alignment larger than the usable packet size. + LhsAlignment = EIGEN_PLAIN_ENUM_MIN(LhsEtorType::Alignment,LhsVecPacketSize*int(sizeof(typename LhsNestedCleaned::Scalar))), + RhsAlignment = EIGEN_PLAIN_ENUM_MIN(RhsEtorType::Alignment,RhsVecPacketSize*int(sizeof(typename RhsNestedCleaned::Scalar))), + + SameType = is_same::value, + + CanVectorizeRhs = bool(RhsRowMajor) && (RhsFlags & PacketAccessBit) && (ColsAtCompileTime!=1), + CanVectorizeLhs = (!LhsRowMajor) && (LhsFlags & PacketAccessBit) && (RowsAtCompileTime!=1), + + EvalToRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0 + : (bool(RhsRowMajor) && !CanVectorizeLhs), + + Flags = ((unsigned int)(LhsFlags | RhsFlags) & HereditaryBits & ~RowMajorBit) + | (EvalToRowMajor ? RowMajorBit : 0) + // TODO enable vectorization for mixed types + | (SameType && (CanVectorizeLhs || CanVectorizeRhs) ? PacketAccessBit : 0) + | (XprType::IsVectorAtCompileTime ? LinearAccessBit : 0), + + LhsOuterStrideBytes = int(LhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename LhsNestedCleaned::Scalar)), + RhsOuterStrideBytes = int(RhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename RhsNestedCleaned::Scalar)), + + Alignment = bool(CanVectorizeLhs) ? (LhsOuterStrideBytes<=0 || (int(LhsOuterStrideBytes) % EIGEN_PLAIN_ENUM_MAX(1,LhsAlignment))!=0 ? 0 : LhsAlignment) + : bool(CanVectorizeRhs) ? (RhsOuterStrideBytes<=0 || (int(RhsOuterStrideBytes) % EIGEN_PLAIN_ENUM_MAX(1,RhsAlignment))!=0 ? 0 : RhsAlignment) + : 0, + + /* CanVectorizeInner deserves special explanation. It does not affect the product flags. It is not used outside + * of Product. If the Product itself is not a packet-access expression, there is still a chance that the inner + * loop of the product might be vectorized. This is the meaning of CanVectorizeInner. Since it doesn't affect + * the Flags, it is safe to make this value depend on ActualPacketAccessBit, that doesn't affect the ABI. + */ + CanVectorizeInner = SameType + && LhsRowMajor + && (!RhsRowMajor) + && (LhsFlags & RhsFlags & ActualPacketAccessBit) + && (InnerSize % packet_traits::size == 0) + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index row, Index col) const + { + return (m_lhs.row(row).transpose().cwiseProduct( m_rhs.col(col) )).sum(); + } + + /* Allow index-based non-packet access. It is impossible though to allow index-based packed access, + * which is why we don't set the LinearAccessBit. + * TODO: this seems possible when the result is a vector + */ + EIGEN_DEVICE_FUNC const CoeffReturnType coeff(Index index) const + { + const Index row = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? 0 : index; + const Index col = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? index : 0; + return (m_lhs.row(row).transpose().cwiseProduct( m_rhs.col(col) )).sum(); + } + + template + const PacketType packet(Index row, Index col) const + { + PacketType res; + typedef etor_product_packet_impl PacketImpl; + PacketImpl::run(row, col, m_lhsImpl, m_rhsImpl, m_innerDim, res); + return res; + } + + template + const PacketType packet(Index index) const + { + const Index row = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? 0 : index; + const Index col = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? index : 0; + return packet(row,col); + } + +protected: + typename internal::add_const_on_value_type::type m_lhs; + typename internal::add_const_on_value_type::type m_rhs; + + LhsEtorType m_lhsImpl; + RhsEtorType m_rhsImpl; + + // TODO: Get rid of m_innerDim if known at compile time + Index m_innerDim; +}; + +template +struct product_evaluator, LazyCoeffBasedProductMode, DenseShape, DenseShape> + : product_evaluator, CoeffBasedProductMode, DenseShape, DenseShape> +{ + typedef Product XprType; + typedef Product BaseProduct; + typedef product_evaluator Base; + enum { + Flags = Base::Flags | EvalBeforeNestingBit + }; + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(BaseProduct(xpr.lhs(),xpr.rhs())) + {} +}; + +/**************************************** +*** Coeff based product, Packet path *** +****************************************/ + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet &res) + { + etor_product_packet_impl::run(row, col, lhs, rhs, innerDim, res); + res = pmadd(pset1(lhs.coeff(row, Index(UnrollingIndex-1))), rhs.template packet(Index(UnrollingIndex-1), col), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet &res) + { + etor_product_packet_impl::run(row, col, lhs, rhs, innerDim, res); + res = pmadd(lhs.template packet(row, Index(UnrollingIndex-1)), pset1(rhs.coeff(Index(UnrollingIndex-1), col)), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index /*innerDim*/, Packet &res) + { + res = pmul(pset1(lhs.coeff(row, Index(0))),rhs.template packet(Index(0), col)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index /*innerDim*/, Packet &res) + { + res = pmul(lhs.template packet(row, Index(0)), pset1(rhs.coeff(Index(0), col))); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Index /*innerDim*/, Packet &res) + { + res = pset1(typename unpacket_traits::type(0)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Index /*innerDim*/, Packet &res) + { + res = pset1(typename unpacket_traits::type(0)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet& res) + { + res = pset1(typename unpacket_traits::type(0)); + for(Index i = 0; i < innerDim; ++i) + res = pmadd(pset1(lhs.coeff(row, i)), rhs.template packet(i, col), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet& res) + { + res = pset1(typename unpacket_traits::type(0)); + for(Index i = 0; i < innerDim; ++i) + res = pmadd(lhs.template packet(row, i), pset1(rhs.coeff(i, col)), res); + } +}; + + +/*************************************************************************** +* Triangular products +***************************************************************************/ +template +struct triangular_product_impl; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + triangular_product_impl + ::run(dst, lhs.nestedExpression(), rhs, alpha); + } +}; + +template +struct generic_product_impl +: generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + triangular_product_impl::run(dst, lhs, rhs.nestedExpression(), alpha); + } +}; + + +/*************************************************************************** +* SelfAdjoint products +***************************************************************************/ +template +struct selfadjoint_product_impl; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + selfadjoint_product_impl::run(dst, lhs.nestedExpression(), rhs, alpha); + } +}; + +template +struct generic_product_impl +: generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + selfadjoint_product_impl::run(dst, lhs, rhs.nestedExpression(), alpha); + } +}; + + +/*************************************************************************** +* Diagonal products +***************************************************************************/ + +template +struct diagonal_product_evaluator_base + : evaluator_base +{ + typedef typename ScalarBinaryOpTraits::ReturnType Scalar; +public: + enum { + CoeffReadCost = NumTraits::MulCost + evaluator::CoeffReadCost + evaluator::CoeffReadCost, + + MatrixFlags = evaluator::Flags, + DiagFlags = evaluator::Flags, + _StorageOrder = MatrixFlags & RowMajorBit ? RowMajor : ColMajor, + _ScalarAccessOnDiag = !((int(_StorageOrder) == ColMajor && int(ProductOrder) == OnTheLeft) + ||(int(_StorageOrder) == RowMajor && int(ProductOrder) == OnTheRight)), + _SameTypes = is_same::value, + // FIXME currently we need same types, but in the future the next rule should be the one + //_Vectorizable = bool(int(MatrixFlags)&PacketAccessBit) && ((!_PacketOnDiag) || (_SameTypes && bool(int(DiagFlags)&PacketAccessBit))), + _Vectorizable = bool(int(MatrixFlags)&PacketAccessBit) && _SameTypes && (_ScalarAccessOnDiag || (bool(int(DiagFlags)&PacketAccessBit))), + _LinearAccessMask = (MatrixType::RowsAtCompileTime==1 || MatrixType::ColsAtCompileTime==1) ? LinearAccessBit : 0, + Flags = ((HereditaryBits|_LinearAccessMask) & (unsigned int)(MatrixFlags)) | (_Vectorizable ? PacketAccessBit : 0), + Alignment = evaluator::Alignment, + + AsScalarProduct = (DiagonalType::SizeAtCompileTime==1) + || (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::RowsAtCompileTime==1 && ProductOrder==OnTheLeft) + || (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::ColsAtCompileTime==1 && ProductOrder==OnTheRight) + }; + + diagonal_product_evaluator_base(const MatrixType &mat, const DiagonalType &diag) + : m_diagImpl(diag), m_matImpl(mat) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::MulCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index idx) const + { + if(AsScalarProduct) + return m_diagImpl.coeff(0) * m_matImpl.coeff(idx); + else + return m_diagImpl.coeff(idx) * m_matImpl.coeff(idx); + } + +protected: + template + EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::true_type) const + { + return internal::pmul(m_matImpl.template packet(row, col), + internal::pset1(m_diagImpl.coeff(id))); + } + + template + EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::false_type) const + { + enum { + InnerSize = (MatrixType::Flags & RowMajorBit) ? MatrixType::ColsAtCompileTime : MatrixType::RowsAtCompileTime, + DiagonalPacketLoadMode = EIGEN_PLAIN_ENUM_MIN(LoadMode,((InnerSize%16) == 0) ? int(Aligned16) : int(evaluator::Alignment)) // FIXME hardcoded 16!! + }; + return internal::pmul(m_matImpl.template packet(row, col), + m_diagImpl.template packet(id)); + } + + evaluator m_diagImpl; + evaluator m_matImpl; +}; + +// diagonal * dense +template +struct product_evaluator, ProductTag, DiagonalShape, DenseShape> + : diagonal_product_evaluator_base, OnTheLeft> +{ + typedef diagonal_product_evaluator_base, OnTheLeft> Base; + using Base::m_diagImpl; + using Base::m_matImpl; + using Base::coeff; + typedef typename Base::Scalar Scalar; + + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + + enum { + StorageOrder = int(Rhs::Flags) & RowMajorBit ? RowMajor : ColMajor + }; + + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(xpr.rhs(), xpr.lhs().diagonal()) + { + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const + { + return m_diagImpl.coeff(row) * m_matImpl.coeff(row, col); + } + +#ifndef __CUDACC__ + template + EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const + { + // FIXME: NVCC used to complain about the template keyword, but we have to check whether this is still the case. + // See also similar calls below. + return this->template packet_impl(row,col, row, + typename internal::conditional::type()); + } + + template + EIGEN_STRONG_INLINE PacketType packet(Index idx) const + { + return packet(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx); + } +#endif +}; + +// dense * diagonal +template +struct product_evaluator, ProductTag, DenseShape, DiagonalShape> + : diagonal_product_evaluator_base, OnTheRight> +{ + typedef diagonal_product_evaluator_base, OnTheRight> Base; + using Base::m_diagImpl; + using Base::m_matImpl; + using Base::coeff; + typedef typename Base::Scalar Scalar; + + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + + enum { StorageOrder = int(Lhs::Flags) & RowMajorBit ? RowMajor : ColMajor }; + + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(xpr.lhs(), xpr.rhs().diagonal()) + { + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const + { + return m_matImpl.coeff(row, col) * m_diagImpl.coeff(col); + } + +#ifndef __CUDACC__ + template + EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const + { + return this->template packet_impl(row,col, col, + typename internal::conditional::type()); + } + + template + EIGEN_STRONG_INLINE PacketType packet(Index idx) const + { + return packet(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx); + } +#endif +}; + +/*************************************************************************** +* Products with permutation matrices +***************************************************************************/ + +/** \internal + * \class permutation_matrix_product + * Internal helper class implementing the product between a permutation matrix and a matrix. + * This class is specialized for DenseShape below and for SparseShape in SparseCore/SparsePermutation.h + */ +template +struct permutation_matrix_product; + +template +struct permutation_matrix_product +{ + typedef typename nested_eval::type MatrixType; + typedef typename remove_all::type MatrixTypeCleaned; + + template + static inline void run(Dest& dst, const PermutationType& perm, const ExpressionType& xpr) + { + MatrixType mat(xpr); + const Index n = Side==OnTheLeft ? mat.rows() : mat.cols(); + // FIXME we need an is_same for expression that is not sensitive to constness. For instance + // is_same_xpr, Block >::value should be true. + //if(is_same::value && extract_data(dst) == extract_data(mat)) + if(is_same_dense(dst, mat)) + { + // apply the permutation inplace + Matrix mask(perm.size()); + mask.fill(false); + Index r = 0; + while(r < perm.size()) + { + // search for the next seed + while(r=perm.size()) + break; + // we got one, let's follow it until we are back to the seed + Index k0 = r++; + Index kPrev = k0; + mask.coeffRef(k0) = true; + for(Index k=perm.indices().coeff(k0); k!=k0; k=perm.indices().coeff(k)) + { + Block(dst, k) + .swap(Block + (dst,((Side==OnTheLeft) ^ Transposed) ? k0 : kPrev)); + + mask.coeffRef(k) = true; + kPrev = k; + } + } + } + else + { + for(Index i = 0; i < n; ++i) + { + Block + (dst, ((Side==OnTheLeft) ^ Transposed) ? perm.indices().coeff(i) : i) + + = + + Block + (mat, ((Side==OnTheRight) ^ Transposed) ? perm.indices().coeff(i) : i); + } + } + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, lhs, rhs); + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, rhs, lhs); + } +}; + +template +struct generic_product_impl, Rhs, PermutationShape, MatrixShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Inverse& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, lhs.nestedExpression(), rhs); + } +}; + +template +struct generic_product_impl, MatrixShape, PermutationShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Inverse& rhs) + { + permutation_matrix_product::run(dst, rhs.nestedExpression(), lhs); + } +}; + + +/*************************************************************************** +* Products with transpositions matrices +***************************************************************************/ + +// FIXME could we unify Transpositions and Permutation into a single "shape"?? + +/** \internal + * \class transposition_matrix_product + * Internal helper class implementing the product between a permutation matrix and a matrix. + */ +template +struct transposition_matrix_product +{ + typedef typename nested_eval::type MatrixType; + typedef typename remove_all::type MatrixTypeCleaned; + + template + static inline void run(Dest& dst, const TranspositionType& tr, const ExpressionType& xpr) + { + MatrixType mat(xpr); + typedef typename TranspositionType::StorageIndex StorageIndex; + const Index size = tr.size(); + StorageIndex j = 0; + + if(!is_same_dense(dst,mat)) + dst = mat; + + for(Index k=(Transposed?size-1:0) ; Transposed?k>=0:k +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, lhs, rhs); + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, rhs, lhs); + } +}; + + +template +struct generic_product_impl, Rhs, TranspositionsShape, MatrixShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Transpose& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, lhs.nestedExpression(), rhs); + } +}; + +template +struct generic_product_impl, MatrixShape, TranspositionsShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Transpose& rhs) + { + transposition_matrix_product::run(dst, rhs.nestedExpression(), lhs); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_EVALUATORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Random.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Random.h new file mode 100644 index 00000000000..6faf789c761 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Random.h @@ -0,0 +1,182 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_RANDOM_H +#define EIGEN_RANDOM_H + +namespace Eigen { + +namespace internal { + +template struct scalar_random_op { + EIGEN_EMPTY_STRUCT_CTOR(scalar_random_op) + inline const Scalar operator() () const { return random(); } +}; + +template +struct functor_traits > +{ enum { Cost = 5 * NumTraits::MulCost, PacketAccess = false, IsRepeatable = false }; }; + +} // end namespace internal + +/** \returns a random matrix expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * \not_reentrant + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Random() should be used + * instead. + * + * + * Example: \include MatrixBase_random_int_int.cpp + * Output: \verbinclude MatrixBase_random_int_int.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary matrix whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * See DenseBase::NullaryExpr(Index, const CustomNullaryOp&) for an example using C++11 random generators. + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index), DenseBase::Random() + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random(Index rows, Index cols) +{ + return NullaryExpr(rows, cols, internal::scalar_random_op()); +} + +/** \returns a random vector expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * \not_reentrant + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Random() should be used + * instead. + * + * Example: \include MatrixBase_random_int.cpp + * Output: \verbinclude MatrixBase_random_int.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary vector whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random() + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random(Index size) +{ + return NullaryExpr(size, internal::scalar_random_op()); +} + +/** \returns a fixed-size random matrix or vector expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_random.cpp + * Output: \verbinclude MatrixBase_random.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary matrix whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * \not_reentrant + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random(Index) + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random() +{ + return NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_random_op()); +} + +/** Sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \not_reentrant + * + * Example: \include MatrixBase_setRandom.cpp + * Output: \verbinclude MatrixBase_setRandom.out + * + * \sa class CwiseNullaryOp, setRandom(Index), setRandom(Index,Index) + */ +template +inline Derived& DenseBase::setRandom() +{ + return *this = Random(rows(), cols()); +} + +/** Resizes to the given \a newSize, and sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \only_for_vectors + * \not_reentrant + * + * Example: \include Matrix_setRandom_int.cpp + * Output: \verbinclude Matrix_setRandom_int.out + * + * \sa DenseBase::setRandom(), setRandom(Index,Index), class CwiseNullaryOp, DenseBase::Random() + */ +template +EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setRandom(Index newSize) +{ + resize(newSize); + return setRandom(); +} + +/** Resizes to the given size, and sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \not_reentrant + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setRandom_int_int.cpp + * Output: \verbinclude Matrix_setRandom_int_int.out + * + * \sa DenseBase::setRandom(), setRandom(Index), class CwiseNullaryOp, DenseBase::Random() + */ +template +EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setRandom(Index rows, Index cols) +{ + resize(rows, cols); + return setRandom(); +} + +} // end namespace Eigen + +#endif // EIGEN_RANDOM_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h new file mode 100644 index 00000000000..760e9f86154 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h @@ -0,0 +1,505 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REDUX_H +#define EIGEN_REDUX_H + +namespace Eigen { + +namespace internal { + +// TODO +// * implement other kind of vectorization +// * factorize code + +/*************************************************************************** +* Part 1 : the logic deciding a strategy for vectorization and unrolling +***************************************************************************/ + +template +struct redux_traits +{ +public: + typedef typename find_best_packet::type PacketType; + enum { + PacketSize = unpacket_traits::size, + InnerMaxSize = int(Derived::IsRowMajor) + ? Derived::MaxColsAtCompileTime + : Derived::MaxRowsAtCompileTime + }; + + enum { + MightVectorize = (int(Derived::Flags)&ActualPacketAccessBit) + && (functor_traits::PacketAccess), + MayLinearVectorize = bool(MightVectorize) && (int(Derived::Flags)&LinearAccessBit), + MaySliceVectorize = bool(MightVectorize) && int(InnerMaxSize)>=3*PacketSize + }; + +public: + enum { + Traversal = int(MayLinearVectorize) ? int(LinearVectorizedTraversal) + : int(MaySliceVectorize) ? int(SliceVectorizedTraversal) + : int(DefaultTraversal) + }; + +public: + enum { + Cost = Derived::SizeAtCompileTime == Dynamic ? HugeCost + : Derived::SizeAtCompileTime * Derived::CoeffReadCost + (Derived::SizeAtCompileTime-1) * functor_traits::Cost, + UnrollingLimit = EIGEN_UNROLLING_LIMIT * (int(Traversal) == int(DefaultTraversal) ? 1 : int(PacketSize)) + }; + +public: + enum { + Unrolling = Cost <= UnrollingLimit ? CompleteUnrolling : NoUnrolling + }; + +#ifdef EIGEN_DEBUG_ASSIGN + static void debug() + { + std::cerr << "Xpr: " << typeid(typename Derived::XprType).name() << std::endl; + std::cerr.setf(std::ios::hex, std::ios::basefield); + EIGEN_DEBUG_VAR(Derived::Flags) + std::cerr.unsetf(std::ios::hex); + EIGEN_DEBUG_VAR(InnerMaxSize) + EIGEN_DEBUG_VAR(PacketSize) + EIGEN_DEBUG_VAR(MightVectorize) + EIGEN_DEBUG_VAR(MayLinearVectorize) + EIGEN_DEBUG_VAR(MaySliceVectorize) + EIGEN_DEBUG_VAR(Traversal) + EIGEN_DEBUG_VAR(UnrollingLimit) + EIGEN_DEBUG_VAR(Unrolling) + std::cerr << std::endl; + } +#endif +}; + +/*************************************************************************** +* Part 2 : unrollers +***************************************************************************/ + +/*** no vectorization ***/ + +template +struct redux_novec_unroller +{ + enum { + HalfLength = Length/2 + }; + + typedef typename Derived::Scalar Scalar; + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + return func(redux_novec_unroller::run(mat,func), + redux_novec_unroller::run(mat,func)); + } +}; + +template +struct redux_novec_unroller +{ + enum { + outer = Start / Derived::InnerSizeAtCompileTime, + inner = Start % Derived::InnerSizeAtCompileTime + }; + + typedef typename Derived::Scalar Scalar; + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func&) + { + return mat.coeffByOuterInner(outer, inner); + } +}; + +// This is actually dead code and will never be called. It is required +// to prevent false warnings regarding failed inlining though +// for 0 length run() will never be called at all. +template +struct redux_novec_unroller +{ + typedef typename Derived::Scalar Scalar; + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived&, const Func&) { return Scalar(); } +}; + +/*** vectorization ***/ + +template +struct redux_vec_unroller +{ + enum { + PacketSize = redux_traits::PacketSize, + HalfLength = Length/2 + }; + + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func& func) + { + return func.packetOp( + redux_vec_unroller::run(mat,func), + redux_vec_unroller::run(mat,func) ); + } +}; + +template +struct redux_vec_unroller +{ + enum { + index = Start * redux_traits::PacketSize, + outer = index / int(Derived::InnerSizeAtCompileTime), + inner = index % int(Derived::InnerSizeAtCompileTime), + alignment = Derived::Alignment + }; + + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func&) + { + return mat.template packetByOuterInner(outer, inner); + } +}; + +/*************************************************************************** +* Part 3 : implementation of all cases +***************************************************************************/ + +template::Traversal, + int Unrolling = redux_traits::Unrolling +> +struct redux_impl; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + Scalar res; + res = mat.coeffByOuterInner(0, 0); + for(Index i = 1; i < mat.innerSize(); ++i) + res = func(res, mat.coeffByOuterInner(0, i)); + for(Index i = 1; i < mat.outerSize(); ++i) + for(Index j = 0; j < mat.innerSize(); ++j) + res = func(res, mat.coeffByOuterInner(i, j)); + return res; + } +}; + +template +struct redux_impl + : public redux_novec_unroller +{}; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static Scalar run(const Derived &mat, const Func& func) + { + const Index size = mat.size(); + + const Index packetSize = redux_traits::PacketSize; + const int packetAlignment = unpacket_traits::alignment; + enum { + alignment0 = (bool(Derived::Flags & DirectAccessBit) && bool(packet_traits::AlignedOnScalar)) ? int(packetAlignment) : int(Unaligned), + alignment = EIGEN_PLAIN_ENUM_MAX(alignment0, Derived::Alignment) + }; + const Index alignedStart = internal::first_default_aligned(mat.nestedExpression()); + const Index alignedSize2 = ((size-alignedStart)/(2*packetSize))*(2*packetSize); + const Index alignedSize = ((size-alignedStart)/(packetSize))*(packetSize); + const Index alignedEnd2 = alignedStart + alignedSize2; + const Index alignedEnd = alignedStart + alignedSize; + Scalar res; + if(alignedSize) + { + PacketScalar packet_res0 = mat.template packet(alignedStart); + if(alignedSize>packetSize) // we have at least two packets to partly unroll the loop + { + PacketScalar packet_res1 = mat.template packet(alignedStart+packetSize); + for(Index index = alignedStart + 2*packetSize; index < alignedEnd2; index += 2*packetSize) + { + packet_res0 = func.packetOp(packet_res0, mat.template packet(index)); + packet_res1 = func.packetOp(packet_res1, mat.template packet(index+packetSize)); + } + + packet_res0 = func.packetOp(packet_res0,packet_res1); + if(alignedEnd>alignedEnd2) + packet_res0 = func.packetOp(packet_res0, mat.template packet(alignedEnd2)); + } + res = func.predux(packet_res0); + + for(Index index = 0; index < alignedStart; ++index) + res = func(res,mat.coeff(index)); + + for(Index index = alignedEnd; index < size; ++index) + res = func(res,mat.coeff(index)); + } + else // too small to vectorize anything. + // since this is dynamic-size hence inefficient anyway for such small sizes, don't try to optimize. + { + res = mat.coeff(0); + for(Index index = 1; index < size; ++index) + res = func(res,mat.coeff(index)); + } + + return res; + } +}; + +// NOTE: for SliceVectorizedTraversal we simply bypass unrolling +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketType; + + EIGEN_DEVICE_FUNC static Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + const Index innerSize = mat.innerSize(); + const Index outerSize = mat.outerSize(); + enum { + packetSize = redux_traits::PacketSize + }; + const Index packetedInnerSize = ((innerSize)/packetSize)*packetSize; + Scalar res; + if(packetedInnerSize) + { + PacketType packet_res = mat.template packet(0,0); + for(Index j=0; j(j,i)); + + res = func.predux(packet_res); + for(Index j=0; j::run(mat, func); + } + + return res; + } +}; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + + typedef typename redux_traits::PacketType PacketScalar; + enum { + PacketSize = redux_traits::PacketSize, + Size = Derived::SizeAtCompileTime, + VectorizedSize = (Size / PacketSize) * PacketSize + }; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + if (VectorizedSize > 0) { + Scalar res = func.predux(redux_vec_unroller::run(mat,func)); + if (VectorizedSize != Size) + res = func(res,redux_novec_unroller::run(mat,func)); + return res; + } + else { + return redux_novec_unroller::run(mat,func); + } + } +}; + +// evaluator adaptor +template +class redux_evaluator +{ +public: + typedef _XprType XprType; + EIGEN_DEVICE_FUNC explicit redux_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {} + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + typedef typename XprType::PacketScalar PacketScalar; + typedef typename XprType::PacketReturnType PacketReturnType; + + enum { + MaxRowsAtCompileTime = XprType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = XprType::MaxColsAtCompileTime, + // TODO we should not remove DirectAccessBit and rather find an elegant way to query the alignment offset at runtime from the evaluator + Flags = evaluator::Flags & ~DirectAccessBit, + IsRowMajor = XprType::IsRowMajor, + SizeAtCompileTime = XprType::SizeAtCompileTime, + InnerSizeAtCompileTime = XprType::InnerSizeAtCompileTime, + CoeffReadCost = evaluator::CoeffReadCost, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } + EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); } + EIGEN_DEVICE_FUNC Index innerSize() const { return m_xpr.innerSize(); } + EIGEN_DEVICE_FUNC Index outerSize() const { return m_xpr.outerSize(); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeff(Index row, Index col) const + { return m_evaluator.coeff(row, col); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeff(Index index) const + { return m_evaluator.coeff(index); } + + template + PacketType packet(Index row, Index col) const + { return m_evaluator.template packet(row, col); } + + template + PacketType packet(Index index) const + { return m_evaluator.template packet(index); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeffByOuterInner(Index outer, Index inner) const + { return m_evaluator.coeff(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); } + + template + PacketType packetByOuterInner(Index outer, Index inner) const + { return m_evaluator.template packet(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); } + + const XprType & nestedExpression() const { return m_xpr; } + +protected: + internal::evaluator m_evaluator; + const XprType &m_xpr; +}; + +} // end namespace internal + +/*************************************************************************** +* Part 4 : public API +***************************************************************************/ + + +/** \returns the result of a full redux operation on the whole matrix or vector using \a func + * + * The template parameter \a BinaryOp is the type of the functor \a func which must be + * an associative operator. Both current C++98 and C++11 functor styles are handled. + * + * \sa DenseBase::sum(), DenseBase::minCoeff(), DenseBase::maxCoeff(), MatrixBase::colwise(), MatrixBase::rowwise() + */ +template +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::redux(const Func& func) const +{ + eigen_assert(this->rows()>0 && this->cols()>0 && "you are using an empty matrix"); + + typedef typename internal::redux_evaluator ThisEvaluator; + ThisEvaluator thisEval(derived()); + + return internal::redux_impl::run(thisEval, func); +} + +/** \returns the minimum of all coefficients of \c *this. + * \warning the result is undefined if \c *this contains NaN. + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::minCoeff() const +{ + return derived().redux(Eigen::internal::scalar_min_op()); +} + +/** \returns the maximum of all coefficients of \c *this. + * \warning the result is undefined if \c *this contains NaN. + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::maxCoeff() const +{ + return derived().redux(Eigen::internal::scalar_max_op()); +} + +/** \returns the sum of all coefficients of \c *this + * + * If \c *this is empty, then the value 0 is returned. + * + * \sa trace(), prod(), mean() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::sum() const +{ + if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0)) + return Scalar(0); + return derived().redux(Eigen::internal::scalar_sum_op()); +} + +/** \returns the mean of all coefficients of *this +* +* \sa trace(), prod(), sum() +*/ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::mean() const +{ +#ifdef __INTEL_COMPILER + #pragma warning push + #pragma warning ( disable : 2259 ) +#endif + return Scalar(derived().redux(Eigen::internal::scalar_sum_op())) / Scalar(this->size()); +#ifdef __INTEL_COMPILER + #pragma warning pop +#endif +} + +/** \returns the product of all coefficients of *this + * + * Example: \include MatrixBase_prod.cpp + * Output: \verbinclude MatrixBase_prod.out + * + * \sa sum(), mean(), trace() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::prod() const +{ + if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0)) + return Scalar(1); + return derived().redux(Eigen::internal::scalar_product_op()); +} + +/** \returns the trace of \c *this, i.e. the sum of the coefficients on the main diagonal. + * + * \c *this can be any matrix, not necessarily square. + * + * \sa diagonal(), sum() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +MatrixBase::trace() const +{ + return derived().diagonal().sum(); +} + +} // end namespace Eigen + +#endif // EIGEN_REDUX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h new file mode 100644 index 00000000000..17a1496b842 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h @@ -0,0 +1,284 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2012 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REF_H +#define EIGEN_REF_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : public traits > +{ + typedef _PlainObjectType PlainObjectType; + typedef _StrideType StrideType; + enum { + Options = _Options, + Flags = traits >::Flags | NestByRefBit, + Alignment = traits >::Alignment + }; + + template struct match { + enum { + IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime, + HasDirectAccess = internal::has_direct_access::ret, + StorageOrderMatch = IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), + InnerStrideMatch = int(StrideType::InnerStrideAtCompileTime)==int(Dynamic) + || int(StrideType::InnerStrideAtCompileTime)==int(Derived::InnerStrideAtCompileTime) + || (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1), + OuterStrideMatch = IsVectorAtCompileTime + || int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime), + // NOTE, this indirection of evaluator::Alignment is needed + // to workaround a very strange bug in MSVC related to the instantiation + // of has_*ary_operator in evaluator. + // This line is surprisingly very sensitive. For instance, simply adding parenthesis + // as "DerivedAlignment = (int(evaluator::Alignment))," will make MSVC fail... + DerivedAlignment = int(evaluator::Alignment), + AlignmentMatch = (int(traits::Alignment)==int(Unaligned)) || (DerivedAlignment >= int(Alignment)), // FIXME the first condition is not very clear, it should be replaced by the required alignment + ScalarTypeMatch = internal::is_same::value, + MatchAtCompileTime = HasDirectAccess && StorageOrderMatch && InnerStrideMatch && OuterStrideMatch && AlignmentMatch && ScalarTypeMatch + }; + typedef typename internal::conditional::type type; + }; + +}; + +template +struct traits > : public traits {}; + +} + +template class RefBase + : public MapBase +{ + typedef typename internal::traits::PlainObjectType PlainObjectType; + typedef typename internal::traits::StrideType StrideType; + +public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(RefBase) + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1; + } + + EIGEN_DEVICE_FUNC inline Index outerStride() const + { + return StrideType::OuterStrideAtCompileTime != 0 ? m_stride.outer() + : IsVectorAtCompileTime ? this->size() + : int(Flags)&RowMajorBit ? this->cols() + : this->rows(); + } + + EIGEN_DEVICE_FUNC RefBase() + : Base(0,RowsAtCompileTime==Dynamic?0:RowsAtCompileTime,ColsAtCompileTime==Dynamic?0:ColsAtCompileTime), + // Stride<> does not allow default ctor for Dynamic strides, so let' initialize it with dummy values: + m_stride(StrideType::OuterStrideAtCompileTime==Dynamic?0:StrideType::OuterStrideAtCompileTime, + StrideType::InnerStrideAtCompileTime==Dynamic?0:StrideType::InnerStrideAtCompileTime) + {} + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(RefBase) + +protected: + + typedef Stride StrideBase; + + template + EIGEN_DEVICE_FUNC void construct(Expression& expr) + { + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(PlainObjectType,Expression); + + if(PlainObjectType::RowsAtCompileTime==1) + { + eigen_assert(expr.rows()==1 || expr.cols()==1); + ::new (static_cast(this)) Base(expr.data(), 1, expr.size()); + } + else if(PlainObjectType::ColsAtCompileTime==1) + { + eigen_assert(expr.rows()==1 || expr.cols()==1); + ::new (static_cast(this)) Base(expr.data(), expr.size(), 1); + } + else + ::new (static_cast(this)) Base(expr.data(), expr.rows(), expr.cols()); + + if(Expression::IsVectorAtCompileTime && (!PlainObjectType::IsVectorAtCompileTime) && ((Expression::Flags&RowMajorBit)!=(PlainObjectType::Flags&RowMajorBit))) + ::new (&m_stride) StrideBase(expr.innerStride(), StrideType::InnerStrideAtCompileTime==0?0:1); + else + ::new (&m_stride) StrideBase(StrideType::OuterStrideAtCompileTime==0?0:expr.outerStride(), + StrideType::InnerStrideAtCompileTime==0?0:expr.innerStride()); + } + + StrideBase m_stride; +}; + +/** \class Ref + * \ingroup Core_Module + * + * \brief A matrix or vector expression mapping an existing expression + * + * \tparam PlainObjectType the equivalent matrix type of the mapped data + * \tparam Options specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned. + * The default is \c #Unaligned. + * \tparam StrideType optionally specifies strides. By default, Ref implies a contiguous storage along the inner dimension (inner stride==1), + * but accepts a variable outer stride (leading dimension). + * This can be overridden by specifying strides. + * The type passed here must be a specialization of the Stride template, see examples below. + * + * This class provides a way to write non-template functions taking Eigen objects as parameters while limiting the number of copies. + * A Ref<> object can represent either a const expression or a l-value: + * \code + * // in-out argument: + * void foo1(Ref x); + * + * // read-only const argument: + * void foo2(const Ref& x); + * \endcode + * + * In the in-out case, the input argument must satisfy the constraints of the actual Ref<> type, otherwise a compilation issue will be triggered. + * By default, a Ref can reference any dense vector expression of float having a contiguous memory layout. + * Likewise, a Ref can reference any column-major dense matrix expression of float whose column's elements are contiguously stored with + * the possibility to have a constant space in-between each column, i.e. the inner stride must be equal to 1, but the outer stride (or leading dimension) + * can be greater than the number of rows. + * + * In the const case, if the input expression does not match the above requirement, then it is evaluated into a temporary before being passed to the function. + * Here are some examples: + * \code + * MatrixXf A; + * VectorXf a; + * foo1(a.head()); // OK + * foo1(A.col()); // OK + * foo1(A.row()); // Compilation error because here innerstride!=1 + * foo2(A.row()); // Compilation error because A.row() is a 1xN object while foo2 is expecting a Nx1 object + * foo2(A.row().transpose()); // The row is copied into a contiguous temporary + * foo2(2*a); // The expression is evaluated into a temporary + * foo2(A.col().segment(2,4)); // No temporary + * \endcode + * + * The range of inputs that can be referenced without temporary can be enlarged using the last two template parameters. + * Here is an example accepting an innerstride!=1: + * \code + * // in-out argument: + * void foo3(Ref > x); + * foo3(A.row()); // OK + * \endcode + * The downside here is that the function foo3 might be significantly slower than foo1 because it won't be able to exploit vectorization, and will involve more + * expensive address computations even if the input is contiguously stored in memory. To overcome this issue, one might propose to overload internally calling a + * template function, e.g.: + * \code + * // in the .h: + * void foo(const Ref& A); + * void foo(const Ref >& A); + * + * // in the .cpp: + * template void foo_impl(const TypeOfA& A) { + * ... // crazy code goes here + * } + * void foo(const Ref& A) { foo_impl(A); } + * void foo(const Ref >& A) { foo_impl(A); } + * \endcode + * + * + * \sa PlainObjectBase::Map(), \ref TopicStorageOrders + */ +template class Ref + : public RefBase > +{ + private: + typedef internal::traits Traits; + template + EIGEN_DEVICE_FUNC inline Ref(const PlainObjectBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0); + public: + + typedef RefBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Ref) + + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC inline Ref(PlainObjectBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0) + { + EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); + Base::construct(expr.derived()); + } + template + EIGEN_DEVICE_FUNC inline Ref(const DenseBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0) + #else + /** Implicit constructor from any dense expression */ + template + inline Ref(DenseBase& expr) + #endif + { + EIGEN_STATIC_ASSERT(bool(internal::is_lvalue::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); + EIGEN_STATIC_ASSERT(!Derived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + Base::construct(expr.const_cast_derived()); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Ref) + +}; + +// this is the const ref version +template class Ref + : public RefBase > +{ + typedef internal::traits Traits; + public: + + typedef RefBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Ref) + + template + EIGEN_DEVICE_FUNC inline Ref(const DenseBase& expr, + typename internal::enable_if::ScalarTypeMatch),Derived>::type* = 0) + { +// std::cout << match_helper::HasDirectAccess << "," << match_helper::OuterStrideMatch << "," << match_helper::InnerStrideMatch << "\n"; +// std::cout << int(StrideType::OuterStrideAtCompileTime) << " - " << int(Derived::OuterStrideAtCompileTime) << "\n"; +// std::cout << int(StrideType::InnerStrideAtCompileTime) << " - " << int(Derived::InnerStrideAtCompileTime) << "\n"; + construct(expr.derived(), typename Traits::template match::type()); + } + + EIGEN_DEVICE_FUNC inline Ref(const Ref& other) : Base(other) { + // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy + } + + template + EIGEN_DEVICE_FUNC inline Ref(const RefBase& other) { + construct(other.derived(), typename Traits::template match::type()); + } + + protected: + + template + EIGEN_DEVICE_FUNC void construct(const Expression& expr,internal::true_type) + { + Base::construct(expr); + } + + template + EIGEN_DEVICE_FUNC void construct(const Expression& expr, internal::false_type) + { + internal::call_assignment_no_alias(m_object,expr,internal::assign_op()); + Base::construct(m_object); + } + + protected: + TPlainObjectType m_object; +}; + +} // end namespace Eigen + +#endif // EIGEN_REF_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h new file mode 100644 index 00000000000..9960ef884ef --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h @@ -0,0 +1,142 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REPLICATE_H +#define EIGEN_REPLICATE_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename MatrixType::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + enum { + RowsAtCompileTime = RowFactor==Dynamic || int(MatrixType::RowsAtCompileTime)==Dynamic + ? Dynamic + : RowFactor * MatrixType::RowsAtCompileTime, + ColsAtCompileTime = ColFactor==Dynamic || int(MatrixType::ColsAtCompileTime)==Dynamic + ? Dynamic + : ColFactor * MatrixType::ColsAtCompileTime, + //FIXME we don't propagate the max sizes !!! + MaxRowsAtCompileTime = RowsAtCompileTime, + MaxColsAtCompileTime = ColsAtCompileTime, + IsRowMajor = MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1 ? 1 + : MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1 ? 0 + : (MatrixType::Flags & RowMajorBit) ? 1 : 0, + + // FIXME enable DirectAccess with negative strides? + Flags = IsRowMajor ? RowMajorBit : 0 + }; +}; +} + +/** + * \class Replicate + * \ingroup Core_Module + * + * \brief Expression of the multiple replication of a matrix or vector + * + * \tparam MatrixType the type of the object we are replicating + * \tparam RowFactor number of repetitions at compile time along the vertical direction, can be Dynamic. + * \tparam ColFactor number of repetitions at compile time along the horizontal direction, can be Dynamic. + * + * This class represents an expression of the multiple replication of a matrix or vector. + * It is the return type of DenseBase::replicate() and most of the time + * this is the only way it is used. + * + * \sa DenseBase::replicate() + */ +template class Replicate + : public internal::dense_xpr_base< Replicate >::type +{ + typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; + typedef typename internal::traits::_MatrixTypeNested _MatrixTypeNested; + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Replicate) + typedef typename internal::remove_all::type NestedExpression; + + template + EIGEN_DEVICE_FUNC + inline explicit Replicate(const OriginalMatrixType& matrix) + : m_matrix(matrix), m_rowFactor(RowFactor), m_colFactor(ColFactor) + { + EIGEN_STATIC_ASSERT((internal::is_same::type,OriginalMatrixType>::value), + THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE) + eigen_assert(RowFactor!=Dynamic && ColFactor!=Dynamic); + } + + template + EIGEN_DEVICE_FUNC + inline Replicate(const OriginalMatrixType& matrix, Index rowFactor, Index colFactor) + : m_matrix(matrix), m_rowFactor(rowFactor), m_colFactor(colFactor) + { + EIGEN_STATIC_ASSERT((internal::is_same::type,OriginalMatrixType>::value), + THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE) + } + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_matrix.rows() * m_rowFactor.value(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_matrix.cols() * m_colFactor.value(); } + + EIGEN_DEVICE_FUNC + const _MatrixTypeNested& nestedExpression() const + { + return m_matrix; + } + + protected: + MatrixTypeNested m_matrix; + const internal::variable_if_dynamic m_rowFactor; + const internal::variable_if_dynamic m_colFactor; +}; + +/** + * \return an expression of the replication of \c *this + * + * Example: \include MatrixBase_replicate.cpp + * Output: \verbinclude MatrixBase_replicate.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(Index,Index), class Replicate + */ +template +template +const Replicate +DenseBase::replicate() const +{ + return Replicate(derived()); +} + +/** + * \return an expression of the replication of each column (or row) of \c *this + * + * Example: \include DirectionWise_replicate_int.cpp + * Output: \verbinclude DirectionWise_replicate_int.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(), class Replicate + */ +template +const typename VectorwiseOp::ReplicateReturnType +VectorwiseOp::replicate(Index factor) const +{ + return typename VectorwiseOp::ReplicateReturnType + (_expression(),Direction==Vertical?factor:1,Direction==Horizontal?factor:1); +} + +} // end namespace Eigen + +#endif // EIGEN_REPLICATE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h new file mode 100644 index 00000000000..c44b7673bb3 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h @@ -0,0 +1,117 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// Copyright (C) 2009-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_RETURNBYVALUE_H +#define EIGEN_RETURNBYVALUE_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : public traits::ReturnType> +{ + enum { + // We're disabling the DirectAccess because e.g. the constructor of + // the Block-with-DirectAccess expression requires to have a coeffRef method. + // Also, we don't want to have to implement the stride stuff. + Flags = (traits::ReturnType>::Flags + | EvalBeforeNestingBit) & ~DirectAccessBit + }; +}; + +/* The ReturnByValue object doesn't even have a coeff() method. + * So the only way that nesting it in an expression can work, is by evaluating it into a plain matrix. + * So internal::nested always gives the plain return matrix type. + * + * FIXME: I don't understand why we need this specialization: isn't this taken care of by the EvalBeforeNestingBit ?? + * Answer: EvalBeforeNestingBit should be deprecated since we have the evaluators + */ +template +struct nested_eval, n, PlainObject> +{ + typedef typename traits::ReturnType type; +}; + +} // end namespace internal + +/** \class ReturnByValue + * \ingroup Core_Module + * + */ +template class ReturnByValue + : public internal::dense_xpr_base< ReturnByValue >::type, internal::no_assignment_operator +{ + public: + typedef typename internal::traits::ReturnType ReturnType; + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(ReturnByValue) + + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const + { static_cast(this)->evalTo(dst); } + EIGEN_DEVICE_FUNC inline Index rows() const { return static_cast(this)->rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return static_cast(this)->cols(); } + +#ifndef EIGEN_PARSED_BY_DOXYGEN +#define Unusable YOU_ARE_TRYING_TO_ACCESS_A_SINGLE_COEFFICIENT_IN_A_SPECIAL_EXPRESSION_WHERE_THAT_IS_NOT_ALLOWED_BECAUSE_THAT_WOULD_BE_INEFFICIENT + class Unusable{ + Unusable(const Unusable&) {} + Unusable& operator=(const Unusable&) {return *this;} + }; + const Unusable& coeff(Index) const { return *reinterpret_cast(this); } + const Unusable& coeff(Index,Index) const { return *reinterpret_cast(this); } + Unusable& coeffRef(Index) { return *reinterpret_cast(this); } + Unusable& coeffRef(Index,Index) { return *reinterpret_cast(this); } +#undef Unusable +#endif +}; + +template +template +Derived& DenseBase::operator=(const ReturnByValue& other) +{ + other.evalTo(derived()); + return derived(); +} + +namespace internal { + +// Expression is evaluated in a temporary; default implementation of Assignment is bypassed so that +// when a ReturnByValue expression is assigned, the evaluator is not constructed. +// TODO: Finalize port to new regime; ReturnByValue should not exist in the expression world + +template +struct evaluator > + : public evaluator::ReturnType> +{ + typedef ReturnByValue XprType; + typedef typename internal::traits::ReturnType PlainObject; + typedef evaluator Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) + : m_result(xpr.rows(), xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + xpr.evalTo(m_result); + } + +protected: + PlainObject m_result; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_RETURNBYVALUE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h new file mode 100644 index 00000000000..0640cda2a15 --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h @@ -0,0 +1,211 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2009 Ricard Marxer +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REVERSE_H +#define EIGEN_REVERSE_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : traits +{ + typedef typename MatrixType::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + enum { + RowsAtCompileTime = MatrixType::RowsAtCompileTime, + ColsAtCompileTime = MatrixType::ColsAtCompileTime, + MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, + Flags = _MatrixTypeNested::Flags & (RowMajorBit | LvalueBit) + }; +}; + +template struct reverse_packet_cond +{ + static inline PacketType run(const PacketType& x) { return preverse(x); } +}; + +template struct reverse_packet_cond +{ + static inline PacketType run(const PacketType& x) { return x; } +}; + +} // end namespace internal + +/** \class Reverse + * \ingroup Core_Module + * + * \brief Expression of the reverse of a vector or matrix + * + * \tparam MatrixType the type of the object of which we are taking the reverse + * \tparam Direction defines the direction of the reverse operation, can be Vertical, Horizontal, or BothDirections + * + * This class represents an expression of the reverse of a vector. + * It is the return type of MatrixBase::reverse() and VectorwiseOp::reverse() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::reverse(), VectorwiseOp::reverse() + */ +template class Reverse + : public internal::dense_xpr_base< Reverse >::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Reverse) + typedef typename internal::remove_all::type NestedExpression; + using Base::IsRowMajor; + + protected: + enum { + PacketSize = internal::packet_traits::size, + IsColMajor = !IsRowMajor, + ReverseRow = (Direction == Vertical) || (Direction == BothDirections), + ReverseCol = (Direction == Horizontal) || (Direction == BothDirections), + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1, + ReversePacket = (Direction == BothDirections) + || ((Direction == Vertical) && IsColMajor) + || ((Direction == Horizontal) && IsRowMajor) + }; + typedef internal::reverse_packet_cond reverse_packet; + public: + + EIGEN_DEVICE_FUNC explicit inline Reverse(const MatrixType& matrix) : m_matrix(matrix) { } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Reverse) + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_matrix.rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_matrix.cols(); } + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return -m_matrix.innerStride(); + } + + EIGEN_DEVICE_FUNC const typename internal::remove_all::type& + nestedExpression() const + { + return m_matrix; + } + + protected: + typename MatrixType::Nested m_matrix; +}; + +/** \returns an expression of the reverse of *this. + * + * Example: \include MatrixBase_reverse.cpp + * Output: \verbinclude MatrixBase_reverse.out + * + */ +template +inline typename DenseBase::ReverseReturnType +DenseBase::reverse() +{ + return ReverseReturnType(derived()); +} + + +//reverse const overload moved DenseBase.h due to a CUDA compiler bug + +/** This is the "in place" version of reverse: it reverses \c *this. + * + * In most cases it is probably better to simply use the reversed expression + * of a matrix. However, when reversing the matrix data itself is really needed, + * then this "in-place" version is probably the right choice because it provides + * the following additional benefits: + * - less error prone: doing the same operation with .reverse() requires special care: + * \code m = m.reverse().eval(); \endcode + * - this API enables reverse operations without the need for a temporary + * - it allows future optimizations (cache friendliness, etc.) + * + * \sa VectorwiseOp::reverseInPlace(), reverse() */ +template +inline void DenseBase::reverseInPlace() +{ + if(cols()>rows()) + { + Index half = cols()/2; + leftCols(half).swap(rightCols(half).reverse()); + if((cols()%2)==1) + { + Index half2 = rows()/2; + col(half).head(half2).swap(col(half).tail(half2).reverse()); + } + } + else + { + Index half = rows()/2; + topRows(half).swap(bottomRows(half).reverse()); + if((rows()%2)==1) + { + Index half2 = cols()/2; + row(half).head(half2).swap(row(half).tail(half2).reverse()); + } + } +} + +namespace internal { + +template +struct vectorwise_reverse_inplace_impl; + +template<> +struct vectorwise_reverse_inplace_impl +{ + template + static void run(ExpressionType &xpr) + { + Index half = xpr.rows()/2; + xpr.topRows(half).swap(xpr.bottomRows(half).colwise().reverse()); + } +}; + +template<> +struct vectorwise_reverse_inplace_impl +{ + template + static void run(ExpressionType &xpr) + { + Index half = xpr.cols()/2; + xpr.leftCols(half).swap(xpr.rightCols(half).rowwise().reverse()); + } +}; + +} // end namespace internal + +/** This is the "in place" version of VectorwiseOp::reverse: it reverses each column or row of \c *this. + * + * In most cases it is probably better to simply use the reversed expression + * of a matrix. However, when reversing the matrix data itself is really needed, + * then this "in-place" version is probably the right choice because it provides + * the following additional benefits: + * - less error prone: doing the same operation with .reverse() requires special care: + * \code m = m.reverse().eval(); \endcode + * - this API enables reverse operations without the need for a temporary + * + * \sa DenseBase::reverseInPlace(), reverse() */ +template +void VectorwiseOp::reverseInPlace() +{ + internal::vectorwise_reverse_inplace_impl::run(_expression().const_cast_derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_REVERSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Select.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Select.h new file mode 100644 index 00000000000..79eec1b5b0e --- /dev/null +++ b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Select.h @@ -0,0 +1,162 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SELECT_H +#define EIGEN_SELECT_H + +namespace Eigen { + +/** \class Select + * \ingroup Core_Module + * + * \brief Expression of a coefficient wise version of the C++ ternary operator ?: + * + * \param ConditionMatrixType the type of the \em condition expression which must be a boolean matrix + * \param ThenMatrixType the type of the \em then expression + * \param ElseMatrixType the type of the \em else expression + * + * This class represents an expression of a coefficient wise version of the C++ ternary operator ?:. + * It is the return type of DenseBase::select() and most of the time this is the only way it is used. + * + * \sa DenseBase::select(const DenseBase&, const DenseBase&) const + */ + +namespace internal { +template +struct traits > + : traits +{ + typedef typename traits::Scalar Scalar; + typedef Dense StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ConditionMatrixType::Nested ConditionMatrixNested; + typedef typename ThenMatrixType::Nested ThenMatrixNested; + typedef typename ElseMatrixType::Nested ElseMatrixNested; + enum { + RowsAtCompileTime = ConditionMatrixType::RowsAtCompileTime, + ColsAtCompileTime = ConditionMatrixType::ColsAtCompileTime, + MaxRowsAtCompileTime = ConditionMatrixType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = ConditionMatrixType::MaxColsAtCompileTime, + Flags = (unsigned int)ThenMatrixType::Flags & ElseMatrixType::Flags & RowMajorBit + }; +}; +} + +template +class Select : public internal::dense_xpr_base< Select >::type, + internal::no_assignment_operator +{ + public: + + typedef typename internal::dense_xpr_base::type Base; - EIGEN_DENSE_PUBLIC_INTERFACE(Select) - - inline EIGEN_DEVICE_FUNC - Select(const ConditionMatrixType& a_conditionMatrix, - const ThenMatrixType& a_thenMatrix, - const ElseMatrixType& a_elseMatrix) - : m_condition(a_conditionMatrix), m_then(a_thenMatrix), m_else(a_elseMatrix) - { - eigen_assert(m_condition.rows() == m_then.rows() && m_condition.rows() == m_else.rows()); - eigen_assert(m_condition.cols() == m_then.cols() && m_condition.cols() == m_else.cols()); - } - - inline EIGEN_DEVICE_FUNC Index rows() const { return m_condition.rows(); } - inline EIGEN_DEVICE_FUNC Index cols() const { return m_condition.cols(); } - - inline EIGEN_DEVICE_FUNC - const Scalar coeff(Index i, Index j) const - { - if (m_condition.coeff(i,j)) - return m_then.coeff(i,j); - else - return m_else.coeff(i,j); - } - - inline EIGEN_DEVICE_FUNC - const Scalar coeff(Index i) const - { - if (m_condition.coeff(i)) - return m_then.coeff(i); - else - return m_else.coeff(i); - } - - inline EIGEN_DEVICE_FUNC const ConditionMatrixType& conditionMatrix() const - { - return m_condition; - } - - inline EIGEN_DEVICE_FUNC const ThenMatrixType& thenMatrix() const - { - return m_then; - } - - inline EIGEN_DEVICE_FUNC const ElseMatrixType& elseMatrix() const - { - return m_else; - } - - protected: - typename ConditionMatrixType::Nested m_condition; - typename ThenMatrixType::Nested m_then; - typename ElseMatrixType::Nested m_else; -}; - - -/** \returns a matrix where each coefficient (i,j) is equal to \a thenMatrix(i,j) - * if \c *this(i,j), and \a elseMatrix(i,j) otherwise. - * - * Example: \include MatrixBase_select.cpp - * Output: \verbinclude MatrixBase_select.out - * - * \sa class Select - */ -template -template -inline const Select -DenseBase::select(const DenseBase& thenMatrix, - const DenseBase& elseMatrix) const -{ - return Select(derived(), thenMatrix.derived(), elseMatrix.derived()); -} - -/** Version of DenseBase::select(const DenseBase&, const DenseBase&) with - * the \em else expression being a scalar value. - * - * \sa DenseBase::select(const DenseBase&, const DenseBase&) const, class Select - */ -template -template -inline const Select -DenseBase::select(const DenseBase& thenMatrix, - const typename ThenDerived::Scalar& elseScalar) const -{ - return Select( - derived(), thenMatrix.derived(), ThenDerived::Constant(rows(),cols(),elseScalar)); -} - -/** Version of DenseBase::select(const DenseBase&, const DenseBase&) with - * the \em then expression being a scalar value. - * - * \sa DenseBase::select(const DenseBase&, const DenseBase&) const, class Select - */ -template -template -inline const Select -DenseBase::select(const typename ElseDerived::Scalar& thenScalar, - const DenseBase& elseMatrix) const -{ - return Select( - derived(), ElseDerived::Constant(rows(),cols(),thenScalar), elseMatrix.derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SELECT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h deleted file mode 100644 index b2e51f37ac4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h +++ /dev/null @@ -1,352 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINTMATRIX_H -#define EIGEN_SELFADJOINTMATRIX_H - -namespace Eigen { - -/** \class SelfAdjointView - * \ingroup Core_Module - * - * - * \brief Expression of a selfadjoint matrix from a triangular part of a dense matrix - * - * \param MatrixType the type of the dense matrix storing the coefficients - * \param TriangularPart can be either \c #Lower or \c #Upper - * - * This class is an expression of a sefladjoint matrix from a triangular part of a matrix - * with given dense storage of the coefficients. It is the return type of MatrixBase::selfadjointView() - * and most of the time this is the only way that it is used. - * - * \sa class TriangularBase, MatrixBase::selfadjointView() - */ - -namespace internal { -template -struct traits > : traits -{ - typedef typename ref_selector::non_const_type MatrixTypeNested; - typedef typename remove_all::type MatrixTypeNestedCleaned; - typedef MatrixType ExpressionType; - typedef typename MatrixType::PlainObject FullMatrixType; - enum { - Mode = UpLo | SelfAdjoint, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags = MatrixTypeNestedCleaned::Flags & (HereditaryBits|FlagsLvalueBit) - & (~(PacketAccessBit | DirectAccessBit | LinearAccessBit)) // FIXME these flags should be preserved - }; -}; -} - - -template class SelfAdjointView - : public TriangularBase > -{ - public: - - typedef _MatrixType MatrixType; - typedef TriangularBase Base; - typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; - typedef typename internal::traits::MatrixTypeNestedCleaned MatrixTypeNestedCleaned; - typedef MatrixTypeNestedCleaned NestedExpression; - - /** \brief The type of coefficients in this matrix */ - typedef typename internal::traits::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename internal::remove_all::type MatrixConjugateReturnType; - - enum { - Mode = internal::traits::Mode, - Flags = internal::traits::Flags, - TransposeMode = ((Mode & Upper) ? Lower : 0) | ((Mode & Lower) ? Upper : 0) - }; - typedef typename MatrixType::PlainObject PlainObject; - - EIGEN_DEVICE_FUNC - explicit inline SelfAdjointView(MatrixType& matrix) : m_matrix(matrix) - { - EIGEN_STATIC_ASSERT(UpLo==Lower || UpLo==Upper,SELFADJOINTVIEW_ACCEPTS_UPPER_AND_LOWER_MODE_ONLY); - } - - EIGEN_DEVICE_FUNC - inline Index rows() const { return m_matrix.rows(); } - EIGEN_DEVICE_FUNC - inline Index cols() const { return m_matrix.cols(); } - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return m_matrix.outerStride(); } - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return m_matrix.innerStride(); } - - /** \sa MatrixBase::coeff() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const - { - Base::check_coordinates_internal(row, col); - return m_matrix.coeff(row, col); - } - - /** \sa MatrixBase::coeffRef() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) - { - EIGEN_STATIC_ASSERT_LVALUE(SelfAdjointView); - Base::check_coordinates_internal(row, col); - return m_matrix.coeffRef(row, col); - } - - /** \internal */ - EIGEN_DEVICE_FUNC - const MatrixTypeNestedCleaned& _expression() const { return m_matrix; } - - EIGEN_DEVICE_FUNC - const MatrixTypeNestedCleaned& nestedExpression() const { return m_matrix; } - EIGEN_DEVICE_FUNC - MatrixTypeNestedCleaned& nestedExpression() { return m_matrix; } - - /** Efficient triangular matrix times vector/matrix product */ - template - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& rhs) const - { - return Product(*this, rhs.derived()); - } - - /** Efficient vector/matrix times triangular matrix product */ - template friend - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& lhs, const SelfAdjointView& rhs) - { - return Product(lhs.derived(),rhs); - } - - friend EIGEN_DEVICE_FUNC - const SelfAdjointView - operator*(const Scalar& s, const SelfAdjointView& mat) - { - return (s*mat.nestedExpression()).template selfadjointView(); - } - - /** Perform a symmetric rank 2 update of the selfadjoint matrix \c *this: - * \f$ this = this + \alpha u v^* + conj(\alpha) v u^* \f$ - * \returns a reference to \c *this - * - * The vectors \a u and \c v \b must be column vectors, however they can be - * a adjoint expression without any overhead. Only the meaningful triangular - * part of the matrix is updated, the rest is left unchanged. - * - * \sa rankUpdate(const MatrixBase&, Scalar) - */ - template - EIGEN_DEVICE_FUNC - SelfAdjointView& rankUpdate(const MatrixBase& u, const MatrixBase& v, const Scalar& alpha = Scalar(1)); - - /** Perform a symmetric rank K update of the selfadjoint matrix \c *this: - * \f$ this = this + \alpha ( u u^* ) \f$ where \a u is a vector or matrix. - * - * \returns a reference to \c *this - * - * Note that to perform \f$ this = this + \alpha ( u^* u ) \f$ you can simply - * call this function with u.adjoint(). - * - * \sa rankUpdate(const MatrixBase&, const MatrixBase&, Scalar) - */ - template - EIGEN_DEVICE_FUNC - SelfAdjointView& rankUpdate(const MatrixBase& u, const Scalar& alpha = Scalar(1)); - - /** \returns an expression of a triangular view extracted from the current selfadjoint view of a given triangular part - * - * The parameter \a TriMode can have the following values: \c #Upper, \c #StrictlyUpper, \c #UnitUpper, - * \c #Lower, \c #StrictlyLower, \c #UnitLower. - * - * If \c TriMode references the same triangular part than \c *this, then this method simply return a \c TriangularView of the nested expression, - * otherwise, the nested expression is first transposed, thus returning a \c TriangularView> object. - * - * \sa MatrixBase::triangularView(), class TriangularView - */ - template - EIGEN_DEVICE_FUNC - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), - TriangularView, - TriangularView >::type - triangularView() const - { - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), MatrixType&, typename MatrixType::ConstTransposeReturnType>::type tmp1(m_matrix); - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), MatrixType&, typename MatrixType::AdjointReturnType>::type tmp2(tmp1); - return typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), - TriangularView, - TriangularView >::type(tmp2); - } - - typedef SelfAdjointView ConjugateReturnType; - /** \sa MatrixBase::conjugate() const */ - EIGEN_DEVICE_FUNC - inline const ConjugateReturnType conjugate() const - { return ConjugateReturnType(m_matrix.conjugate()); } - - typedef SelfAdjointView AdjointReturnType; - /** \sa MatrixBase::adjoint() const */ - EIGEN_DEVICE_FUNC - inline const AdjointReturnType adjoint() const - { return AdjointReturnType(m_matrix.adjoint()); } - - typedef SelfAdjointView TransposeReturnType; - /** \sa MatrixBase::transpose() */ - EIGEN_DEVICE_FUNC - inline TransposeReturnType transpose() - { - EIGEN_STATIC_ASSERT_LVALUE(MatrixType) - typename MatrixType::TransposeReturnType tmp(m_matrix); - return TransposeReturnType(tmp); - } - - typedef SelfAdjointView ConstTransposeReturnType; - /** \sa MatrixBase::transpose() const */ - EIGEN_DEVICE_FUNC - inline const ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(m_matrix.transpose()); - } - - /** \returns a const expression of the main diagonal of the matrix \c *this - * - * This method simply returns the diagonal of the nested expression, thus by-passing the SelfAdjointView decorator. - * - * \sa MatrixBase::diagonal(), class Diagonal */ - EIGEN_DEVICE_FUNC - typename MatrixType::ConstDiagonalReturnType diagonal() const - { - return typename MatrixType::ConstDiagonalReturnType(m_matrix); - } - -/////////// Cholesky module /////////// - - const LLT llt() const; - const LDLT ldlt() const; - -/////////// Eigenvalue module /////////// - - /** Real part of #Scalar */ - typedef typename NumTraits::Real RealScalar; - /** Return type of eigenvalues() */ - typedef Matrix::ColsAtCompileTime, 1> EigenvaluesReturnType; - - EIGEN_DEVICE_FUNC - EigenvaluesReturnType eigenvalues() const; - EIGEN_DEVICE_FUNC - RealScalar operatorNorm() const; - - protected: - MatrixTypeNested m_matrix; -}; - - -// template -// internal::selfadjoint_matrix_product_returntype > -// operator*(const MatrixBase& lhs, const SelfAdjointView& rhs) -// { -// return internal::matrix_selfadjoint_product_returntype >(lhs.derived(),rhs); -// } - -// selfadjoint to dense matrix - -namespace internal { - -// TODO currently a selfadjoint expression has the form SelfAdjointView<.,.> -// in the future selfadjoint-ness should be defined by the expression traits -// such that Transpose > is valid. (currently TriangularBase::transpose() is overloaded to make it work) -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef SelfAdjointShape Shape; -}; - -template -class triangular_dense_assignment_kernel - : public generic_dense_assignment_kernel -{ -protected: - typedef generic_dense_assignment_kernel Base; - typedef typename Base::DstXprType DstXprType; - typedef typename Base::SrcXprType SrcXprType; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; -public: - - typedef typename Base::DstEvaluatorType DstEvaluatorType; - typedef typename Base::SrcEvaluatorType SrcEvaluatorType; - typedef typename Base::Scalar Scalar; - typedef typename Base::AssignmentTraits AssignmentTraits; - - - EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - - EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - Scalar tmp = m_src.coeff(row,col); - m_functor.assignCoeff(m_dst.coeffRef(row,col), tmp); - m_functor.assignCoeff(m_dst.coeffRef(col,row), numext::conj(tmp)); - } - - EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id) - { - Base::assignCoeff(id,id); - } - - EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index, Index) - { eigen_internal_assert(false && "should never be called"); } -}; - -} // end namespace internal - -/*************************************************************************** -* Implementation of MatrixBase methods -***************************************************************************/ - -/** This is the const version of MatrixBase::selfadjointView() */ -template -template -typename MatrixBase::template ConstSelfAdjointViewReturnType::Type -MatrixBase::selfadjointView() const -{ - return typename ConstSelfAdjointViewReturnType::Type(derived()); -} - -/** \returns an expression of a symmetric/self-adjoint view extracted from the upper or lower triangular part of the current matrix - * - * The parameter \a UpLo can be either \c #Upper or \c #Lower - * - * Example: \include MatrixBase_selfadjointView.cpp - * Output: \verbinclude MatrixBase_selfadjointView.out - * - * \sa class SelfAdjointView - */ -template -template -typename MatrixBase::template SelfAdjointViewReturnType::Type -MatrixBase::selfadjointView() -{ - return typename SelfAdjointViewReturnType::Type(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINTMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h deleted file mode 100644 index 7c89c2e23c6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h +++ /dev/null @@ -1,47 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFCWISEBINARYOP_H -#define EIGEN_SELFCWISEBINARYOP_H - -namespace Eigen { - -// TODO generalize the scalar type of 'other' - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::operator*=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::mul_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& ArrayBase::operator+=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::add_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& ArrayBase::operator-=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::sub_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::operator/=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::div_assign_op()); - return derived(); -} - -} // end namespace Eigen - -#endif // EIGEN_SELFCWISEBINARYOP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h deleted file mode 100644 index a8daea51135..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h +++ /dev/null @@ -1,188 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVE_H -#define EIGEN_SOLVE_H - -namespace Eigen { - -template class SolveImpl; - -/** \class Solve - * \ingroup Core_Module - * - * \brief Pseudo expression representing a solving operation - * - * \tparam Decomposition the type of the matrix or decomposion object - * \tparam Rhstype the type of the right-hand side - * - * This class represents an expression of A.solve(B) - * and most of the time this is the only way it is used. - * - */ -namespace internal { - -// this solve_traits class permits to determine the evaluation type with respect to storage kind (Dense vs Sparse) -template struct solve_traits; - -template -struct solve_traits -{ - typedef typename make_proper_matrix_type::type PlainObject; -}; - -template -struct traits > - : traits::StorageKind>::PlainObject> -{ - typedef typename solve_traits::StorageKind>::PlainObject PlainObject; - typedef typename promote_index_type::type StorageIndex; - typedef traits BaseTraits; - enum { - Flags = BaseTraits::Flags & RowMajorBit, - CoeffReadCost = HugeCost - }; -}; - -} - - -template -class Solve : public SolveImpl::StorageKind> -{ -public: - typedef typename internal::traits::PlainObject PlainObject; - typedef typename internal::traits::StorageIndex StorageIndex; - - Solve(const Decomposition &dec, const RhsType &rhs) - : m_dec(dec), m_rhs(rhs) - {} - - EIGEN_DEVICE_FUNC Index rows() const { return m_dec.cols(); } - EIGEN_DEVICE_FUNC Index cols() const { return m_rhs.cols(); } - - EIGEN_DEVICE_FUNC const Decomposition& dec() const { return m_dec; } - EIGEN_DEVICE_FUNC const RhsType& rhs() const { return m_rhs; } - -protected: - const Decomposition &m_dec; - const RhsType &m_rhs; -}; - - -// Specialization of the Solve expression for dense results -template -class SolveImpl - : public MatrixBase > -{ - typedef Solve Derived; - -public: - - typedef MatrixBase > Base; - EIGEN_DENSE_PUBLIC_INTERFACE(Derived) - -private: - - Scalar coeff(Index row, Index col) const; - Scalar coeff(Index i) const; -}; - -// Generic API dispatcher -template -class SolveImpl : public internal::generic_xpr_base, MatrixXpr, StorageKind>::type -{ - public: - typedef typename internal::generic_xpr_base, MatrixXpr, StorageKind>::type Base; -}; - -namespace internal { - -// Evaluator of Solve -> eval into a temporary -template -struct evaluator > - : public evaluator::PlainObject> -{ - typedef Solve SolveType; - typedef typename SolveType::PlainObject PlainObject; - typedef evaluator Base; - - enum { Flags = Base::Flags | EvalBeforeNestingBit }; - - EIGEN_DEVICE_FUNC explicit evaluator(const SolveType& solve) - : m_result(solve.rows(), solve.cols()) - { - ::new (static_cast(this)) Base(m_result); - solve.dec()._solve_impl(solve.rhs(), m_result); - } - -protected: - PlainObject m_result; -}; - -// Specialization for "dst = dec.solve(rhs)" -// NOTE we need to specialize it for Dense2Dense to avoid ambiguous specialization error and a Sparse2Sparse specialization must exist somewhere -template -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Solve SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec()._solve_impl(src.rhs(), dst); - } -}; - -// Specialization for "dst = dec.transpose().solve(rhs)" -template -struct Assignment,RhsType>, internal::assign_op, Dense2Dense> -{ - typedef Solve,RhsType> SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec().nestedExpression().template _solve_impl_transposed(src.rhs(), dst); - } -}; - -// Specialization for "dst = dec.adjoint().solve(rhs)" -template -struct Assignment, const Transpose >,RhsType>, - internal::assign_op, Dense2Dense> -{ - typedef Solve, const Transpose >,RhsType> SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec().nestedExpression().nestedExpression().template _solve_impl_transposed(src.rhs(), dst); - } -}; - -} // end namepsace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h deleted file mode 100644 index fd0acb1a58e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h +++ /dev/null @@ -1,235 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVETRIANGULAR_H -#define EIGEN_SOLVETRIANGULAR_H - -namespace Eigen { - -namespace internal { - -// Forward declarations: -// The following two routines are implemented in the products/TriangularSolver*.h files -template -struct triangular_solve_vector; - -template -struct triangular_solve_matrix; - -// small helper struct extracting some traits on the underlying solver operation -template -class trsolve_traits -{ - private: - enum { - RhsIsVectorAtCompileTime = (Side==OnTheLeft ? Rhs::ColsAtCompileTime : Rhs::RowsAtCompileTime)==1 - }; - public: - enum { - Unrolling = (RhsIsVectorAtCompileTime && Rhs::SizeAtCompileTime != Dynamic && Rhs::SizeAtCompileTime <= 8) - ? CompleteUnrolling : NoUnrolling, - RhsVectors = RhsIsVectorAtCompileTime ? 1 : Dynamic - }; -}; - -template::Unrolling, - int RhsVectors = trsolve_traits::RhsVectors - > -struct triangular_solver_selector; - -template -struct triangular_solver_selector -{ - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef blas_traits LhsProductTraits; - typedef typename LhsProductTraits::ExtractType ActualLhsType; - typedef Map, Aligned> MappedRhs; - static void run(const Lhs& lhs, Rhs& rhs) - { - ActualLhsType actualLhs = LhsProductTraits::extract(lhs); - - // FIXME find a way to allow an inner stride if packet_traits::size==1 - - bool useRhsDirectly = Rhs::InnerStrideAtCompileTime==1 || rhs.innerStride()==1; - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhs,rhs.size(), - (useRhsDirectly ? rhs.data() : 0)); - - if(!useRhsDirectly) - MappedRhs(actualRhs,rhs.size()) = rhs; - - triangular_solve_vector - ::run(actualLhs.cols(), actualLhs.data(), actualLhs.outerStride(), actualRhs); - - if(!useRhsDirectly) - rhs = MappedRhs(actualRhs, rhs.size()); - } -}; - -// the rhs is a matrix -template -struct triangular_solver_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef blas_traits LhsProductTraits; - typedef typename LhsProductTraits::DirectLinearAccessType ActualLhsType; - - static void run(const Lhs& lhs, Rhs& rhs) - { - typename internal::add_const_on_value_type::type actualLhs = LhsProductTraits::extract(lhs); - - const Index size = lhs.rows(); - const Index othersize = Side==OnTheLeft? rhs.cols() : rhs.rows(); - - typedef internal::gemm_blocking_space<(Rhs::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Rhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxRowsAtCompileTime,4> BlockingType; - - BlockingType blocking(rhs.rows(), rhs.cols(), size, 1, false); - - triangular_solve_matrix - ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.innerStride(), rhs.outerStride(), blocking); - } -}; - -/*************************************************************************** -* meta-unrolling implementation -***************************************************************************/ - -template -struct triangular_solver_unroller; - -template -struct triangular_solver_unroller { - enum { - IsLower = ((Mode&Lower)==Lower), - DiagIndex = IsLower ? LoopIndex : Size - LoopIndex - 1, - StartIndex = IsLower ? 0 : DiagIndex+1 - }; - static void run(const Lhs& lhs, Rhs& rhs) - { - if (LoopIndex>0) - rhs.coeffRef(DiagIndex) -= lhs.row(DiagIndex).template segment(StartIndex).transpose() - .cwiseProduct(rhs.template segment(StartIndex)).sum(); - - if(!(Mode & UnitDiag)) - rhs.coeffRef(DiagIndex) /= lhs.coeff(DiagIndex,DiagIndex); - - triangular_solver_unroller::run(lhs,rhs); - } -}; - -template -struct triangular_solver_unroller { - static void run(const Lhs&, Rhs&) {} -}; - -template -struct triangular_solver_selector { - static void run(const Lhs& lhs, Rhs& rhs) - { triangular_solver_unroller::run(lhs,rhs); } -}; - -template -struct triangular_solver_selector { - static void run(const Lhs& lhs, Rhs& rhs) - { - Transpose trLhs(lhs); - Transpose trRhs(rhs); - - triangular_solver_unroller,Transpose, - ((Mode&Upper)==Upper ? Lower : Upper) | (Mode&UnitDiag), - 0,Rhs::SizeAtCompileTime>::run(trLhs,trRhs); - } -}; - -} // end namespace internal - -/*************************************************************************** -* TriangularView methods -***************************************************************************/ - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void TriangularViewImpl::solveInPlace(const MatrixBase& _other) const -{ - OtherDerived& other = _other.const_cast_derived(); - eigen_assert( derived().cols() == derived().rows() && ((Side==OnTheLeft && derived().cols() == other.rows()) || (Side==OnTheRight && derived().cols() == other.cols())) ); - eigen_assert((!(Mode & ZeroDiag)) && bool(Mode & (Upper|Lower))); - // If solving for a 0x0 matrix, nothing to do, simply return. - if (derived().cols() == 0) - return; - - enum { copy = (internal::traits::Flags & RowMajorBit) && OtherDerived::IsVectorAtCompileTime && OtherDerived::SizeAtCompileTime!=1}; - typedef typename internal::conditional::type, OtherDerived&>::type OtherCopy; - OtherCopy otherCopy(other); - - internal::triangular_solver_selector::type, - Side, Mode>::run(derived().nestedExpression(), otherCopy); - - if (copy) - other = otherCopy; -} - -template -template -const internal::triangular_solve_retval,Other> -TriangularViewImpl::solve(const MatrixBase& other) const -{ - return internal::triangular_solve_retval(derived(), other.derived()); -} -#endif - -namespace internal { - - -template -struct traits > -{ - typedef typename internal::plain_matrix_type_column_major::type ReturnType; -}; - -template struct triangular_solve_retval - : public ReturnByValue > -{ - typedef typename remove_all::type RhsNestedCleaned; - typedef ReturnByValue Base; - - triangular_solve_retval(const TriangularType& tri, const Rhs& rhs) - : m_triangularMatrix(tri), m_rhs(rhs) - {} - - inline Index rows() const { return m_rhs.rows(); } - inline Index cols() const { return m_rhs.cols(); } - - template inline void evalTo(Dest& dst) const - { - if(!is_same_dense(dst,m_rhs)) - dst = m_rhs; - m_triangularMatrix.template solveInPlace(dst); - } - - protected: - const TriangularType& m_triangularMatrix; - typename Rhs::Nested m_rhs; -}; - -} // namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVETRIANGULAR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h deleted file mode 100644 index 8a4adc22973..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h +++ /dev/null @@ -1,130 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVERBASE_H -#define EIGEN_SOLVERBASE_H - -namespace Eigen { - -namespace internal { - - - -} // end namespace internal - -/** \class SolverBase - * \brief A base class for matrix decomposition and solvers - * - * \tparam Derived the actual type of the decomposition/solver. - * - * Any matrix decomposition inheriting this base class provide the following API: - * - * \code - * MatrixType A, b, x; - * DecompositionType dec(A); - * x = dec.solve(b); // solve A * x = b - * x = dec.transpose().solve(b); // solve A^T * x = b - * x = dec.adjoint().solve(b); // solve A' * x = b - * \endcode - * - * \warning Currently, any other usage of transpose() and adjoint() are not supported and will produce compilation errors. - * - * \sa class PartialPivLU, class FullPivLU - */ -template -class SolverBase : public EigenBase -{ - public: - - typedef EigenBase Base; - typedef typename internal::traits::Scalar Scalar; - typedef Scalar CoeffReturnType; - - enum { - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, - internal::traits::ColsAtCompileTime>::ret), - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, - MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, - internal::traits::MaxColsAtCompileTime>::ret), - IsVectorAtCompileTime = internal::traits::MaxRowsAtCompileTime == 1 - || internal::traits::MaxColsAtCompileTime == 1 - }; - - /** Default constructor */ - SolverBase() - {} - - ~SolverBase() - {} - - using Base::derived; - - /** \returns an expression of the solution x of \f$ A x = b \f$ using the current decomposition of A. - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(derived().rows()==b.rows() && "solve(): invalid number of rows of the right hand side matrix b"); - return Solve(derived(), b.derived()); - } - - /** \internal the return type of transpose() */ - typedef typename internal::add_const >::type ConstTransposeReturnType; - /** \returns an expression of the transposed of the factored matrix. - * - * A typical usage is to solve for the transposed problem A^T x = b: - * \code x = dec.transpose().solve(b); \endcode - * - * \sa adjoint(), solve() - */ - inline ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(derived()); - } - - /** \internal the return type of adjoint() */ - typedef typename internal::conditional::IsComplex, - CwiseUnaryOp, ConstTransposeReturnType>, - ConstTransposeReturnType - >::type AdjointReturnType; - /** \returns an expression of the adjoint of the factored matrix - * - * A typical usage is to solve for the adjoint problem A' x = b: - * \code x = dec.adjoint().solve(b); \endcode - * - * For real scalar types, this function is equivalent to transpose(). - * - * \sa transpose(), solve() - */ - inline AdjointReturnType adjoint() const - { - return AdjointReturnType(derived().transpose()); - } - - protected: -}; - -namespace internal { - -template -struct generic_xpr_base -{ - typedef SolverBase type; - -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVERBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h deleted file mode 100644 index 88c8d989024..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h +++ /dev/null @@ -1,221 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STABLENORM_H -#define EIGEN_STABLENORM_H - -namespace Eigen { - -namespace internal { - -template -inline void stable_norm_kernel(const ExpressionType& bl, Scalar& ssq, Scalar& scale, Scalar& invScale) -{ - Scalar maxCoeff = bl.cwiseAbs().maxCoeff(); - - if(maxCoeff>scale) - { - ssq = ssq * numext::abs2(scale/maxCoeff); - Scalar tmp = Scalar(1)/maxCoeff; - if(tmp > NumTraits::highest()) - { - invScale = NumTraits::highest(); - scale = Scalar(1)/invScale; - } - else if(maxCoeff>NumTraits::highest()) // we got a INF - { - invScale = Scalar(1); - scale = maxCoeff; - } - else - { - scale = maxCoeff; - invScale = tmp; - } - } - else if(maxCoeff!=maxCoeff) // we got a NaN - { - scale = maxCoeff; - } - - // TODO if the maxCoeff is much much smaller than the current scale, - // then we can neglect this sub vector - if(scale>Scalar(0)) // if scale==0, then bl is 0 - ssq += (bl*invScale).squaredNorm(); -} - -template -inline typename NumTraits::Scalar>::Real -blueNorm_impl(const EigenBase& _vec) -{ - typedef typename Derived::RealScalar RealScalar; - using std::pow; - using std::sqrt; - using std::abs; - const Derived& vec(_vec.derived()); - static bool initialized = false; - static RealScalar b1, b2, s1m, s2m, rbig, relerr; - if(!initialized) - { - int ibeta, it, iemin, iemax, iexp; - RealScalar eps; - // This program calculates the machine-dependent constants - // bl, b2, slm, s2m, relerr overfl - // from the "basic" machine-dependent numbers - // nbig, ibeta, it, iemin, iemax, rbig. - // The following define the basic machine-dependent constants. - // For portability, the PORT subprograms "ilmaeh" and "rlmach" - // are used. For any specific computer, each of the assignment - // statements can be replaced - ibeta = std::numeric_limits::radix; // base for floating-point numbers - it = std::numeric_limits::digits; // number of base-beta digits in mantissa - iemin = std::numeric_limits::min_exponent; // minimum exponent - iemax = std::numeric_limits::max_exponent; // maximum exponent - rbig = (std::numeric_limits::max)(); // largest floating-point number - - iexp = -((1-iemin)/2); - b1 = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // lower boundary of midrange - iexp = (iemax + 1 - it)/2; - b2 = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // upper boundary of midrange - - iexp = (2-iemin)/2; - s1m = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // scaling factor for lower range - iexp = - ((iemax+it)/2); - s2m = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // scaling factor for upper range - - eps = RealScalar(pow(double(ibeta), 1-it)); - relerr = sqrt(eps); // tolerance for neglecting asml - initialized = true; - } - Index n = vec.size(); - RealScalar ab2 = b2 / RealScalar(n); - RealScalar asml = RealScalar(0); - RealScalar amed = RealScalar(0); - RealScalar abig = RealScalar(0); - for(typename Derived::InnerIterator it(vec, 0); it; ++it) - { - RealScalar ax = abs(it.value()); - if(ax > ab2) abig += numext::abs2(ax*s2m); - else if(ax < b1) asml += numext::abs2(ax*s1m); - else amed += numext::abs2(ax); - } - if(amed!=amed) - return amed; // we got a NaN - if(abig > RealScalar(0)) - { - abig = sqrt(abig); - if(abig > rbig) // overflow, or *this contains INF values - return abig; // return INF - if(amed > RealScalar(0)) - { - abig = abig/s2m; - amed = sqrt(amed); - } - else - return abig/s2m; - } - else if(asml > RealScalar(0)) - { - if (amed > RealScalar(0)) - { - abig = sqrt(amed); - amed = sqrt(asml) / s1m; - } - else - return sqrt(asml)/s1m; - } - else - return sqrt(amed); - asml = numext::mini(abig, amed); - abig = numext::maxi(abig, amed); - if(asml <= abig*relerr) - return abig; - else - return abig * sqrt(RealScalar(1) + numext::abs2(asml/abig)); -} - -} // end namespace internal - -/** \returns the \em l2 norm of \c *this avoiding underflow and overflow. - * This version use a blockwise two passes algorithm: - * 1 - find the absolute largest coefficient \c s - * 2 - compute \f$ s \Vert \frac{*this}{s} \Vert \f$ in a standard way - * - * For architecture/scalar types supporting vectorization, this version - * is faster than blueNorm(). Otherwise the blueNorm() is much faster. - * - * \sa norm(), blueNorm(), hypotNorm() - */ -template -inline typename NumTraits::Scalar>::Real -MatrixBase::stableNorm() const -{ - using std::sqrt; - using std::abs; - const Index blockSize = 4096; - RealScalar scale(0); - RealScalar invScale(1); - RealScalar ssq(0); // sum of square - - typedef typename internal::nested_eval::type DerivedCopy; - typedef typename internal::remove_all::type DerivedCopyClean; - const DerivedCopy copy(derived()); - - enum { - CanAlign = ( (int(DerivedCopyClean::Flags)&DirectAccessBit) - || (int(internal::evaluator::Alignment)>0) // FIXME Alignment)>0 might not be enough - ) && (blockSize*sizeof(Scalar)*20) // if we cannot allocate on the stack, then let's not bother about this optimization - }; - typedef typename internal::conditional, internal::evaluator::Alignment>, - typename DerivedCopyClean::ConstSegmentReturnType>::type SegmentWrapper; - Index n = size(); - - if(n==1) - return abs(this->coeff(0)); - - Index bi = internal::first_default_aligned(copy); - if (bi>0) - internal::stable_norm_kernel(copy.head(bi), ssq, scale, invScale); - for (; bi -inline typename NumTraits::Scalar>::Real -MatrixBase::blueNorm() const -{ - return internal::blueNorm_impl(*this); -} - -/** \returns the \em l2 norm of \c *this avoiding undeflow and overflow. - * This version use a concatenation of hypot() calls, and it is very slow. - * - * \sa norm(), stableNorm() - */ -template -inline typename NumTraits::Scalar>::Real -MatrixBase::hypotNorm() const -{ - return this->cwiseAbs().redux(internal::scalar_hypot_op()); -} - -} // end namespace Eigen - -#endif // EIGEN_STABLENORM_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h deleted file mode 100644 index 513742f34b9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h +++ /dev/null @@ -1,111 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STRIDE_H -#define EIGEN_STRIDE_H - -namespace Eigen { - -/** \class Stride - * \ingroup Core_Module - * - * \brief Holds strides information for Map - * - * This class holds the strides information for mapping arrays with strides with class Map. - * - * It holds two values: the inner stride and the outer stride. - * - * The inner stride is the pointer increment between two consecutive entries within a given row of a - * row-major matrix or within a given column of a column-major matrix. - * - * The outer stride is the pointer increment between two consecutive rows of a row-major matrix or - * between two consecutive columns of a column-major matrix. - * - * These two values can be passed either at compile-time as template parameters, or at runtime as - * arguments to the constructor. - * - * Indeed, this class takes two template parameters: - * \tparam _OuterStrideAtCompileTime the outer stride, or Dynamic if you want to specify it at runtime. - * \tparam _InnerStrideAtCompileTime the inner stride, or Dynamic if you want to specify it at runtime. - * - * Here is an example: - * \include Map_general_stride.cpp - * Output: \verbinclude Map_general_stride.out - * - * \sa class InnerStride, class OuterStride, \ref TopicStorageOrders - */ -template -class Stride -{ - public: - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - enum { - InnerStrideAtCompileTime = _InnerStrideAtCompileTime, - OuterStrideAtCompileTime = _OuterStrideAtCompileTime - }; - - /** Default constructor, for use when strides are fixed at compile time */ - EIGEN_DEVICE_FUNC - Stride() - : m_outer(OuterStrideAtCompileTime), m_inner(InnerStrideAtCompileTime) - { - eigen_assert(InnerStrideAtCompileTime != Dynamic && OuterStrideAtCompileTime != Dynamic); - } - - /** Constructor allowing to pass the strides at runtime */ - EIGEN_DEVICE_FUNC - Stride(Index outerStride, Index innerStride) - : m_outer(outerStride), m_inner(innerStride) - { - eigen_assert(innerStride>=0 && outerStride>=0); - } - - /** Copy constructor */ - EIGEN_DEVICE_FUNC - Stride(const Stride& other) - : m_outer(other.outer()), m_inner(other.inner()) - {} - - /** \returns the outer stride */ - EIGEN_DEVICE_FUNC - inline Index outer() const { return m_outer.value(); } - /** \returns the inner stride */ - EIGEN_DEVICE_FUNC - inline Index inner() const { return m_inner.value(); } - - protected: - internal::variable_if_dynamic m_outer; - internal::variable_if_dynamic m_inner; -}; - -/** \brief Convenience specialization of Stride to specify only an inner stride - * See class Map for some examples */ -template -class InnerStride : public Stride<0, Value> -{ - typedef Stride<0, Value> Base; - public: - EIGEN_DEVICE_FUNC InnerStride() : Base() {} - EIGEN_DEVICE_FUNC InnerStride(Index v) : Base(0, v) {} // FIXME making this explicit could break valid code -}; - -/** \brief Convenience specialization of Stride to specify only an outer stride - * See class Map for some examples */ -template -class OuterStride : public Stride -{ - typedef Stride Base; - public: - EIGEN_DEVICE_FUNC OuterStride() : Base() {} - EIGEN_DEVICE_FUNC OuterStride(Index v) : Base(v,0) {} // FIXME making this explicit could break valid code -}; - -} // end namespace Eigen - -#endif // EIGEN_STRIDE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h deleted file mode 100644 index d702009185e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h +++ /dev/null @@ -1,67 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SWAP_H -#define EIGEN_SWAP_H - -namespace Eigen { - -namespace internal { - -// Overload default assignPacket behavior for swapping them -template -class generic_dense_assignment_kernel, Specialized> - : public generic_dense_assignment_kernel, BuiltIn> -{ -protected: - typedef generic_dense_assignment_kernel, BuiltIn> Base; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; - -public: - typedef typename Base::Scalar Scalar; - typedef typename Base::DstXprType DstXprType; - typedef swap_assign_op Functor; - - EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorTypeT &dst, const SrcEvaluatorTypeT &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - - template - void assignPacket(Index row, Index col) - { - PacketType tmp = m_src.template packet(row,col); - const_cast(m_src).template writePacket(row,col, m_dst.template packet(row,col)); - m_dst.template writePacket(row,col,tmp); - } - - template - void assignPacket(Index index) - { - PacketType tmp = m_src.template packet(index); - const_cast(m_src).template writePacket(index, m_dst.template packet(index)); - m_dst.template writePacket(index,tmp); - } - - // TODO find a simple way not to have to copy/paste this function from generic_dense_assignment_kernel, by simple I mean no CRTP (Gael) - template - void assignPacketByOuterInner(Index outer, Index inner) - { - Index row = Base::rowIndexByOuterInner(outer, inner); - Index col = Base::colIndexByOuterInner(outer, inner); - assignPacket(row, col); - } -}; - -} // namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SWAP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h deleted file mode 100644 index 960dc4510c7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h +++ /dev/null @@ -1,405 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2008 Benoit Jacob -// Copyright (C) 2009-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSPOSE_H -#define EIGEN_TRANSPOSE_H - -namespace Eigen { - -namespace internal { -template -struct traits > : public traits -{ - typedef typename ref_selector::type MatrixTypeNested; - typedef typename remove_reference::type MatrixTypeNestedPlain; - enum { - RowsAtCompileTime = MatrixType::ColsAtCompileTime, - ColsAtCompileTime = MatrixType::RowsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags0 = traits::Flags & ~(LvalueBit | NestByRefBit), - Flags1 = Flags0 | FlagsLvalueBit, - Flags = Flags1 ^ RowMajorBit, - InnerStrideAtCompileTime = inner_stride_at_compile_time::ret, - OuterStrideAtCompileTime = outer_stride_at_compile_time::ret - }; -}; -} - -template class TransposeImpl; - -/** \class Transpose - * \ingroup Core_Module - * - * \brief Expression of the transpose of a matrix - * - * \tparam MatrixType the type of the object of which we are taking the transpose - * - * This class represents an expression of the transpose of a matrix. - * It is the return type of MatrixBase::transpose() and MatrixBase::adjoint() - * and most of the time this is the only way it is used. - * - * \sa MatrixBase::transpose(), MatrixBase::adjoint() - */ -template class Transpose - : public TransposeImpl::StorageKind> -{ - public: - - typedef typename internal::ref_selector::non_const_type MatrixTypeNested; - - typedef typename TransposeImpl::StorageKind>::Base Base; - EIGEN_GENERIC_PUBLIC_INTERFACE(Transpose) - typedef typename internal::remove_all::type NestedExpression; - - EIGEN_DEVICE_FUNC - explicit inline Transpose(MatrixType& matrix) : m_matrix(matrix) {} - - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Transpose) - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_matrix.cols(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_matrix.rows(); } - - /** \returns the nested expression */ - EIGEN_DEVICE_FUNC - const typename internal::remove_all::type& - nestedExpression() const { return m_matrix; } - - /** \returns the nested expression */ - EIGEN_DEVICE_FUNC - typename internal::remove_reference::type& - nestedExpression() { return m_matrix; } - - /** \internal */ - void resize(Index nrows, Index ncols) { - m_matrix.resize(ncols,nrows); - } - - protected: - typename internal::ref_selector::non_const_type m_matrix; -}; - -namespace internal { - -template::ret> -struct TransposeImpl_base -{ - typedef typename dense_xpr_base >::type type; -}; - -template -struct TransposeImpl_base -{ - typedef typename dense_xpr_base >::type type; -}; - -} // end namespace internal - -// Generic API dispatcher -template -class TransposeImpl - : public internal::generic_xpr_base >::type -{ -public: - typedef typename internal::generic_xpr_base >::type Base; -}; - -template class TransposeImpl - : public internal::TransposeImpl_base::type -{ - public: - - typedef typename internal::TransposeImpl_base::type Base; - using Base::coeffRef; - EIGEN_DENSE_PUBLIC_INTERFACE(Transpose) - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TransposeImpl) - - EIGEN_DEVICE_FUNC inline Index innerStride() const { return derived().nestedExpression().innerStride(); } - EIGEN_DEVICE_FUNC inline Index outerStride() const { return derived().nestedExpression().outerStride(); } - - typedef typename internal::conditional< - internal::is_lvalue::value, - Scalar, - const Scalar - >::type ScalarWithConstIfNotLvalue; - - EIGEN_DEVICE_FUNC inline ScalarWithConstIfNotLvalue* data() { return derived().nestedExpression().data(); } - EIGEN_DEVICE_FUNC inline const Scalar* data() const { return derived().nestedExpression().data(); } - - // FIXME: shall we keep the const version of coeffRef? - EIGEN_DEVICE_FUNC - inline const Scalar& coeffRef(Index rowId, Index colId) const - { - return derived().nestedExpression().coeffRef(colId, rowId); - } - - EIGEN_DEVICE_FUNC - inline const Scalar& coeffRef(Index index) const - { - return derived().nestedExpression().coeffRef(index); - } - protected: - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TransposeImpl) -}; - -/** \returns an expression of the transpose of *this. - * - * Example: \include MatrixBase_transpose.cpp - * Output: \verbinclude MatrixBase_transpose.out - * - * \warning If you want to replace a matrix by its own transpose, do \b NOT do this: - * \code - * m = m.transpose(); // bug!!! caused by aliasing effect - * \endcode - * Instead, use the transposeInPlace() method: - * \code - * m.transposeInPlace(); - * \endcode - * which gives Eigen good opportunities for optimization, or alternatively you can also do: - * \code - * m = m.transpose().eval(); - * \endcode - * - * \sa transposeInPlace(), adjoint() */ -template -inline Transpose -DenseBase::transpose() -{ - return TransposeReturnType(derived()); -} - -/** This is the const version of transpose(). - * - * Make sure you read the warning for transpose() ! - * - * \sa transposeInPlace(), adjoint() */ -template -inline typename DenseBase::ConstTransposeReturnType -DenseBase::transpose() const -{ - return ConstTransposeReturnType(derived()); -} - -/** \returns an expression of the adjoint (i.e. conjugate transpose) of *this. - * - * Example: \include MatrixBase_adjoint.cpp - * Output: \verbinclude MatrixBase_adjoint.out - * - * \warning If you want to replace a matrix by its own adjoint, do \b NOT do this: - * \code - * m = m.adjoint(); // bug!!! caused by aliasing effect - * \endcode - * Instead, use the adjointInPlace() method: - * \code - * m.adjointInPlace(); - * \endcode - * which gives Eigen good opportunities for optimization, or alternatively you can also do: - * \code - * m = m.adjoint().eval(); - * \endcode - * - * \sa adjointInPlace(), transpose(), conjugate(), class Transpose, class internal::scalar_conjugate_op */ -template -inline const typename MatrixBase::AdjointReturnType -MatrixBase::adjoint() const -{ - return AdjointReturnType(this->transpose()); -} - -/*************************************************************************** -* "in place" transpose implementation -***************************************************************************/ - -namespace internal { - -template::size)) - && (internal::evaluator::Flags&PacketAccessBit) > -struct inplace_transpose_selector; - -template -struct inplace_transpose_selector { // square matrix - static void run(MatrixType& m) { - m.matrix().template triangularView().swap(m.matrix().transpose()); - } -}; - -// TODO: vectorized path is currently limited to LargestPacketSize x LargestPacketSize cases only. -template -struct inplace_transpose_selector { // PacketSize x PacketSize - static void run(MatrixType& m) { - typedef typename MatrixType::Scalar Scalar; - typedef typename internal::packet_traits::type Packet; - const Index PacketSize = internal::packet_traits::size; - const Index Alignment = internal::evaluator::Alignment; - PacketBlock A; - for (Index i=0; i(i,0); - internal::ptranspose(A); - for (Index i=0; i(m.rowIndexByOuterInner(i,0), m.colIndexByOuterInner(i,0), A.packet[i]); - } -}; - -template -struct inplace_transpose_selector { // non square matrix - static void run(MatrixType& m) { - if (m.rows()==m.cols()) - m.matrix().template triangularView().swap(m.matrix().transpose()); - else - m = m.transpose().eval(); - } -}; - -} // end namespace internal - -/** This is the "in place" version of transpose(): it replaces \c *this by its own transpose. - * Thus, doing - * \code - * m.transposeInPlace(); - * \endcode - * has the same effect on m as doing - * \code - * m = m.transpose().eval(); - * \endcode - * and is faster and also safer because in the latter line of code, forgetting the eval() results - * in a bug caused by \ref TopicAliasing "aliasing". - * - * Notice however that this method is only useful if you want to replace a matrix by its own transpose. - * If you just need the transpose of a matrix, use transpose(). - * - * \note if the matrix is not square, then \c *this must be a resizable matrix. - * This excludes (non-square) fixed-size matrices, block-expressions and maps. - * - * \sa transpose(), adjoint(), adjointInPlace() */ -template -inline void DenseBase::transposeInPlace() -{ - eigen_assert((rows() == cols() || (RowsAtCompileTime == Dynamic && ColsAtCompileTime == Dynamic)) - && "transposeInPlace() called on a non-square non-resizable matrix"); - internal::inplace_transpose_selector::run(derived()); -} - -/*************************************************************************** -* "in place" adjoint implementation -***************************************************************************/ - -/** This is the "in place" version of adjoint(): it replaces \c *this by its own transpose. - * Thus, doing - * \code - * m.adjointInPlace(); - * \endcode - * has the same effect on m as doing - * \code - * m = m.adjoint().eval(); - * \endcode - * and is faster and also safer because in the latter line of code, forgetting the eval() results - * in a bug caused by aliasing. - * - * Notice however that this method is only useful if you want to replace a matrix by its own adjoint. - * If you just need the adjoint of a matrix, use adjoint(). - * - * \note if the matrix is not square, then \c *this must be a resizable matrix. - * This excludes (non-square) fixed-size matrices, block-expressions and maps. - * - * \sa transpose(), adjoint(), transposeInPlace() */ -template -inline void MatrixBase::adjointInPlace() -{ - derived() = adjoint().eval(); -} - -#ifndef EIGEN_NO_DEBUG - -// The following is to detect aliasing problems in most common cases. - -namespace internal { - -template -struct check_transpose_aliasing_compile_time_selector -{ - enum { ret = bool(blas_traits::IsTransposed) != DestIsTransposed }; -}; - -template -struct check_transpose_aliasing_compile_time_selector > -{ - enum { ret = bool(blas_traits::IsTransposed) != DestIsTransposed - || bool(blas_traits::IsTransposed) != DestIsTransposed - }; -}; - -template -struct check_transpose_aliasing_run_time_selector -{ - static bool run(const Scalar* dest, const OtherDerived& src) - { - return (bool(blas_traits::IsTransposed) != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src)); - } -}; - -template -struct check_transpose_aliasing_run_time_selector > -{ - static bool run(const Scalar* dest, const CwiseBinaryOp& src) - { - return ((blas_traits::IsTransposed != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src.lhs()))) - || ((blas_traits::IsTransposed != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src.rhs()))); - } -}; - -// the following selector, checkTransposeAliasing_impl, based on MightHaveTransposeAliasing, -// is because when the condition controlling the assert is known at compile time, ICC emits a warning. -// This is actually a good warning: in expressions that don't have any transposing, the condition is -// known at compile time to be false, and using that, we can avoid generating the code of the assert again -// and again for all these expressions that don't need it. - -template::IsTransposed,OtherDerived>::ret - > -struct checkTransposeAliasing_impl -{ - static void run(const Derived& dst, const OtherDerived& other) - { - eigen_assert((!check_transpose_aliasing_run_time_selector - ::IsTransposed,OtherDerived> - ::run(extract_data(dst), other)) - && "aliasing detected during transposition, use transposeInPlace() " - "or evaluate the rhs into a temporary using .eval()"); - - } -}; - -template -struct checkTransposeAliasing_impl -{ - static void run(const Derived&, const OtherDerived&) - { - } -}; - -template -void check_for_aliasing(const Dst &dst, const Src &src) -{ - internal::checkTransposeAliasing_impl::run(dst, src); -} - -} // end namespace internal - -#endif // EIGEN_NO_DEBUG - -} // end namespace Eigen - -#endif // EIGEN_TRANSPOSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h deleted file mode 100644 index 7718625e80b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h +++ /dev/null @@ -1,368 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010-2011 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSPOSITIONS_H -#define EIGEN_TRANSPOSITIONS_H - -namespace Eigen { - -template -class TranspositionsBase -{ - typedef internal::traits Traits; - - public: - - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - Derived& derived() { return *static_cast(this); } - const Derived& derived() const { return *static_cast(this); } - - /** Copies the \a other transpositions into \c *this */ - template - Derived& operator=(const TranspositionsBase& other) - { - indices() = other.indices(); - return derived(); - } - - /** \returns the number of transpositions */ - Index size() const { return indices().size(); } - /** \returns the number of rows of the equivalent permutation matrix */ - Index rows() const { return indices().size(); } - /** \returns the number of columns of the equivalent permutation matrix */ - Index cols() const { return indices().size(); } - - /** Direct access to the underlying index vector */ - inline const StorageIndex& coeff(Index i) const { return indices().coeff(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& coeffRef(Index i) { return indices().coeffRef(i); } - /** Direct access to the underlying index vector */ - inline const StorageIndex& operator()(Index i) const { return indices()(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& operator()(Index i) { return indices()(i); } - /** Direct access to the underlying index vector */ - inline const StorageIndex& operator[](Index i) const { return indices()(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& operator[](Index i) { return indices()(i); } - - /** const version of indices(). */ - const IndicesType& indices() const { return derived().indices(); } - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return derived().indices(); } - - /** Resizes to given size. */ - inline void resize(Index newSize) - { - indices().resize(newSize); - } - - /** Sets \c *this to represents an identity transformation */ - void setIdentity() - { - for(StorageIndex i = 0; i < indices().size(); ++i) - coeffRef(i) = i; - } - - // FIXME: do we want such methods ? - // might be usefull when the target matrix expression is complex, e.g.: - // object.matrix().block(..,..,..,..) = trans * object.matrix().block(..,..,..,..); - /* - template - void applyForwardToRows(MatrixType& mat) const - { - for(Index k=0 ; k - void applyBackwardToRows(MatrixType& mat) const - { - for(Index k=size()-1 ; k>=0 ; --k) - if(m_indices(k)!=k) - mat.row(k).swap(mat.row(m_indices(k))); - } - */ - - /** \returns the inverse transformation */ - inline Transpose inverse() const - { return Transpose(derived()); } - - /** \returns the tranpose transformation */ - inline Transpose transpose() const - { return Transpose(derived()); } - - protected: -}; - -namespace internal { -template -struct traits > - : traits > -{ - typedef Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType; - typedef TranspositionsStorage StorageKind; -}; -} - -/** \class Transpositions - * \ingroup Core_Module - * - * \brief Represents a sequence of transpositions (row/column interchange) - * - * \tparam SizeAtCompileTime the number of transpositions, or Dynamic - * \tparam MaxSizeAtCompileTime the maximum number of transpositions, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it. - * - * This class represents a permutation transformation as a sequence of \em n transpositions - * \f$[T_{n-1} \ldots T_{i} \ldots T_{0}]\f$. It is internally stored as a vector of integers \c indices. - * Each transposition \f$ T_{i} \f$ applied on the left of a matrix (\f$ T_{i} M\f$) interchanges - * the rows \c i and \c indices[i] of the matrix \c M. - * A transposition applied on the right (e.g., \f$ M T_{i}\f$) yields a column interchange. - * - * Compared to the class PermutationMatrix, such a sequence of transpositions is what is - * computed during a decomposition with pivoting, and it is faster when applying the permutation in-place. - * - * To apply a sequence of transpositions to a matrix, simply use the operator * as in the following example: - * \code - * Transpositions tr; - * MatrixXf mat; - * mat = tr * mat; - * \endcode - * In this example, we detect that the matrix appears on both side, and so the transpositions - * are applied in-place without any temporary or extra copy. - * - * \sa class PermutationMatrix - */ - -template -class Transpositions : public TranspositionsBase > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - inline Transpositions() {} - - /** Copy constructor. */ - template - inline Transpositions(const TranspositionsBase& other) - : m_indices(other.indices()) {} - - /** Generic constructor from expression of the transposition indices. */ - template - explicit inline Transpositions(const MatrixBase& indices) : m_indices(indices) - {} - - /** Copies the \a other transpositions into \c *this */ - template - Transpositions& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - /** Constructs an uninitialized permutation matrix of given size. - */ - inline Transpositions(Index size) : m_indices(size) - {} - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - IndicesType m_indices; -}; - - -namespace internal { -template -struct traits,_PacketAccess> > - : traits > -{ - typedef Map, _PacketAccess> IndicesType; - typedef _StorageIndex StorageIndex; - typedef TranspositionsStorage StorageKind; -}; -} - -template -class Map,PacketAccess> - : public TranspositionsBase,PacketAccess> > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - explicit inline Map(const StorageIndex* indicesPtr) - : m_indices(indicesPtr) - {} - - inline Map(const StorageIndex* indicesPtr, Index size) - : m_indices(indicesPtr,size) - {} - - /** Copies the \a other transpositions into \c *this */ - template - Map& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** This is a special case of the templated operator=. Its purpose is to - * prevent a default operator= from hiding the templated operator=. - */ - Map& operator=(const Map& other) - { - m_indices = other.m_indices; - return *this; - } - #endif - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - IndicesType m_indices; -}; - -namespace internal { -template -struct traits > - : traits > -{ - typedef TranspositionsStorage StorageKind; -}; -} - -template -class TranspositionsWrapper - : public TranspositionsBase > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - explicit inline TranspositionsWrapper(IndicesType& indices) - : m_indices(indices) - {} - - /** Copies the \a other transpositions into \c *this */ - template - TranspositionsWrapper& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - typename IndicesType::Nested m_indices; -}; - - - -/** \returns the \a matrix with the \a transpositions applied to the columns. - */ -template -EIGEN_DEVICE_FUNC -const Product -operator*(const MatrixBase &matrix, - const TranspositionsBase& transpositions) -{ - return Product - (matrix.derived(), transpositions.derived()); -} - -/** \returns the \a matrix with the \a transpositions applied to the rows. - */ -template -EIGEN_DEVICE_FUNC -const Product -operator*(const TranspositionsBase &transpositions, - const MatrixBase& matrix) -{ - return Product - (transpositions.derived(), matrix.derived()); -} - -// Template partial specialization for transposed/inverse transpositions - -namespace internal { - -template -struct traits > > - : traits -{}; - -} // end namespace internal - -template -class Transpose > -{ - typedef TranspositionsDerived TranspositionType; - typedef typename TranspositionType::IndicesType IndicesType; - public: - - explicit Transpose(const TranspositionType& t) : m_transpositions(t) {} - - Index size() const { return m_transpositions.size(); } - Index rows() const { return m_transpositions.size(); } - Index cols() const { return m_transpositions.size(); } - - /** \returns the \a matrix with the inverse transpositions applied to the columns. - */ - template friend - const Product - operator*(const MatrixBase& matrix, const Transpose& trt) - { - return Product(matrix.derived(), trt); - } - - /** \returns the \a matrix with the inverse transpositions applied to the rows. - */ - template - const Product - operator*(const MatrixBase& matrix) const - { - return Product(*this, matrix.derived()); - } - - const TranspositionType& nestedExpression() const { return m_transpositions; } - - protected: - const TranspositionType& m_transpositions; -}; - -} // end namespace Eigen - -#endif // EIGEN_TRANSPOSITIONS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h deleted file mode 100644 index 9abb7e31a8e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h +++ /dev/null @@ -1,985 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Benoit Jacob -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULARMATRIX_H -#define EIGEN_TRIANGULARMATRIX_H - -namespace Eigen { - -namespace internal { - -template struct triangular_solve_retval; - -} - -/** \class TriangularBase - * \ingroup Core_Module - * - * \brief Base class for triangular part in a matrix - */ -template class TriangularBase : public EigenBase -{ - public: - - enum { - Mode = internal::traits::Mode, - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, - - SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, - internal::traits::ColsAtCompileTime>::ret), - /**< This is equal to the number of coefficients, i.e. the number of - * rows times the number of columns, or to \a Dynamic if this is not - * known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */ - - MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, - internal::traits::MaxColsAtCompileTime>::ret) - - }; - typedef typename internal::traits::Scalar Scalar; - typedef typename internal::traits::StorageKind StorageKind; - typedef typename internal::traits::StorageIndex StorageIndex; - typedef typename internal::traits::FullMatrixType DenseMatrixType; - typedef DenseMatrixType DenseType; - typedef Derived const& Nested; - - EIGEN_DEVICE_FUNC - inline TriangularBase() { eigen_assert(!((Mode&UnitDiag) && (Mode&ZeroDiag))); } - - EIGEN_DEVICE_FUNC - inline Index rows() const { return derived().rows(); } - EIGEN_DEVICE_FUNC - inline Index cols() const { return derived().cols(); } - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return derived().outerStride(); } - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return derived().innerStride(); } - - // dummy resize function - void resize(Index rows, Index cols) - { - EIGEN_UNUSED_VARIABLE(rows); - EIGEN_UNUSED_VARIABLE(cols); - eigen_assert(rows==this->rows() && cols==this->cols()); - } - - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const { return derived().coeff(row,col); } - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) { return derived().coeffRef(row,col); } - - /** \see MatrixBase::copyCoeff(row,col) - */ - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE void copyCoeff(Index row, Index col, Other& other) - { - derived().coeffRef(row, col) = other.coeff(row, col); - } - - EIGEN_DEVICE_FUNC - inline Scalar operator()(Index row, Index col) const - { - check_coordinates(row, col); - return coeff(row,col); - } - EIGEN_DEVICE_FUNC - inline Scalar& operator()(Index row, Index col) - { - check_coordinates(row, col); - return coeffRef(row,col); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_DEVICE_FUNC - inline const Derived& derived() const { return *static_cast(this); } - EIGEN_DEVICE_FUNC - inline Derived& derived() { return *static_cast(this); } - #endif // not EIGEN_PARSED_BY_DOXYGEN - - template - EIGEN_DEVICE_FUNC - void evalTo(MatrixBase &other) const; - template - EIGEN_DEVICE_FUNC - void evalToLazy(MatrixBase &other) const; - - EIGEN_DEVICE_FUNC - DenseMatrixType toDenseMatrix() const - { - DenseMatrixType res(rows(), cols()); - evalToLazy(res); - return res; - } - - protected: - - void check_coordinates(Index row, Index col) const - { - EIGEN_ONLY_USED_FOR_DEBUG(row); - EIGEN_ONLY_USED_FOR_DEBUG(col); - eigen_assert(col>=0 && col=0 && row=row) - || (mode==Lower && col<=row) - || ((mode==StrictlyUpper || mode==UnitUpper) && col>row) - || ((mode==StrictlyLower || mode==UnitLower) && col -struct traits > : traits -{ - typedef typename ref_selector::non_const_type MatrixTypeNested; - typedef typename remove_reference::type MatrixTypeNestedNonRef; - typedef typename remove_all::type MatrixTypeNestedCleaned; - typedef typename MatrixType::PlainObject FullMatrixType; - typedef MatrixType ExpressionType; - enum { - Mode = _Mode, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags = (MatrixTypeNestedCleaned::Flags & (HereditaryBits | FlagsLvalueBit) & (~(PacketAccessBit | DirectAccessBit | LinearAccessBit))) - }; -}; -} - -template class TriangularViewImpl; - -template class TriangularView - : public TriangularViewImpl<_MatrixType, _Mode, typename internal::traits<_MatrixType>::StorageKind > -{ - public: - - typedef TriangularViewImpl<_MatrixType, _Mode, typename internal::traits<_MatrixType>::StorageKind > Base; - typedef typename internal::traits::Scalar Scalar; - typedef _MatrixType MatrixType; - - protected: - typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; - typedef typename internal::traits::MatrixTypeNestedNonRef MatrixTypeNestedNonRef; - - typedef typename internal::remove_all::type MatrixConjugateReturnType; - - public: - - typedef typename internal::traits::StorageKind StorageKind; - typedef typename internal::traits::MatrixTypeNestedCleaned NestedExpression; - - enum { - Mode = _Mode, - Flags = internal::traits::Flags, - TransposeMode = (Mode & Upper ? Lower : 0) - | (Mode & Lower ? Upper : 0) - | (Mode & (UnitDiag)) - | (Mode & (ZeroDiag)), - IsVectorAtCompileTime = false - }; - - EIGEN_DEVICE_FUNC - explicit inline TriangularView(MatrixType& matrix) : m_matrix(matrix) - {} - - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TriangularView) - - /** \copydoc EigenBase::rows() */ - EIGEN_DEVICE_FUNC - inline Index rows() const { return m_matrix.rows(); } - /** \copydoc EigenBase::cols() */ - EIGEN_DEVICE_FUNC - inline Index cols() const { return m_matrix.cols(); } - - /** \returns a const reference to the nested expression */ - EIGEN_DEVICE_FUNC - const NestedExpression& nestedExpression() const { return m_matrix; } - - /** \returns a reference to the nested expression */ - EIGEN_DEVICE_FUNC - NestedExpression& nestedExpression() { return m_matrix; } - - typedef TriangularView ConjugateReturnType; - /** \sa MatrixBase::conjugate() const */ - EIGEN_DEVICE_FUNC - inline const ConjugateReturnType conjugate() const - { return ConjugateReturnType(m_matrix.conjugate()); } - - typedef TriangularView AdjointReturnType; - /** \sa MatrixBase::adjoint() const */ - EIGEN_DEVICE_FUNC - inline const AdjointReturnType adjoint() const - { return AdjointReturnType(m_matrix.adjoint()); } - - typedef TriangularView TransposeReturnType; - /** \sa MatrixBase::transpose() */ - EIGEN_DEVICE_FUNC - inline TransposeReturnType transpose() - { - EIGEN_STATIC_ASSERT_LVALUE(MatrixType) - typename MatrixType::TransposeReturnType tmp(m_matrix); - return TransposeReturnType(tmp); - } - - typedef TriangularView ConstTransposeReturnType; - /** \sa MatrixBase::transpose() const */ - EIGEN_DEVICE_FUNC - inline const ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(m_matrix.transpose()); - } - - template - EIGEN_DEVICE_FUNC - inline const Solve - solve(const MatrixBase& other) const - { return Solve(*this, other.derived()); } - - // workaround MSVC ICE - #if EIGEN_COMP_MSVC - template - EIGEN_DEVICE_FUNC - inline const internal::triangular_solve_retval - solve(const MatrixBase& other) const - { return Base::template solve(other); } - #else - using Base::solve; - #endif - - /** \returns a selfadjoint view of the referenced triangular part which must be either \c #Upper or \c #Lower. - * - * This is a shortcut for \code this->nestedExpression().selfadjointView<(*this)::Mode>() \endcode - * \sa MatrixBase::selfadjointView() */ - EIGEN_DEVICE_FUNC - SelfAdjointView selfadjointView() - { - EIGEN_STATIC_ASSERT((Mode&(UnitDiag|ZeroDiag))==0,PROGRAMMING_ERROR); - return SelfAdjointView(m_matrix); - } - - /** This is the const version of selfadjointView() */ - EIGEN_DEVICE_FUNC - const SelfAdjointView selfadjointView() const - { - EIGEN_STATIC_ASSERT((Mode&(UnitDiag|ZeroDiag))==0,PROGRAMMING_ERROR); - return SelfAdjointView(m_matrix); - } - - - /** \returns the determinant of the triangular matrix - * \sa MatrixBase::determinant() */ - EIGEN_DEVICE_FUNC - Scalar determinant() const - { - if (Mode & UnitDiag) - return 1; - else if (Mode & ZeroDiag) - return 0; - else - return m_matrix.diagonal().prod(); - } - - protected: - - MatrixTypeNested m_matrix; -}; - -/** \ingroup Core_Module - * - * \brief Base class for a triangular part in a \b dense matrix - * - * This class is an abstract base class of class TriangularView, and objects of type TriangularViewImpl cannot be instantiated. - * It extends class TriangularView with additional methods which available for dense expressions only. - * - * \sa class TriangularView, MatrixBase::triangularView() - */ -template class TriangularViewImpl<_MatrixType,_Mode,Dense> - : public TriangularBase > -{ - public: - - typedef TriangularView<_MatrixType, _Mode> TriangularViewType; - typedef TriangularBase Base; - typedef typename internal::traits::Scalar Scalar; - - typedef _MatrixType MatrixType; - typedef typename MatrixType::PlainObject DenseMatrixType; - typedef DenseMatrixType PlainObject; - - public: - using Base::evalToLazy; - using Base::derived; - - typedef typename internal::traits::StorageKind StorageKind; - - enum { - Mode = _Mode, - Flags = internal::traits::Flags - }; - - /** \returns the outer-stride of the underlying dense matrix - * \sa DenseCoeffsBase::outerStride() */ - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return derived().nestedExpression().outerStride(); } - /** \returns the inner-stride of the underlying dense matrix - * \sa DenseCoeffsBase::innerStride() */ - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return derived().nestedExpression().innerStride(); } - - /** \sa MatrixBase::operator+=() */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator+=(const DenseBase& other) { - internal::call_assignment_no_alias(derived(), other.derived(), internal::add_assign_op()); - return derived(); - } - /** \sa MatrixBase::operator-=() */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator-=(const DenseBase& other) { - internal::call_assignment_no_alias(derived(), other.derived(), internal::sub_assign_op()); - return derived(); - } - - /** \sa MatrixBase::operator*=() */ - EIGEN_DEVICE_FUNC - TriangularViewType& operator*=(const typename internal::traits::Scalar& other) { return *this = derived().nestedExpression() * other; } - /** \sa DenseBase::operator/=() */ - EIGEN_DEVICE_FUNC - TriangularViewType& operator/=(const typename internal::traits::Scalar& other) { return *this = derived().nestedExpression() / other; } - - /** \sa MatrixBase::fill() */ - EIGEN_DEVICE_FUNC - void fill(const Scalar& value) { setConstant(value); } - /** \sa MatrixBase::setConstant() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setConstant(const Scalar& value) - { return *this = MatrixType::Constant(derived().rows(), derived().cols(), value); } - /** \sa MatrixBase::setZero() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setZero() { return setConstant(Scalar(0)); } - /** \sa MatrixBase::setOnes() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setOnes() { return setConstant(Scalar(1)); } - - /** \sa MatrixBase::coeff() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const - { - Base::check_coordinates_internal(row, col); - return derived().nestedExpression().coeff(row, col); - } - - /** \sa MatrixBase::coeffRef() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) - { - EIGEN_STATIC_ASSERT_LVALUE(TriangularViewType); - Base::check_coordinates_internal(row, col); - return derived().nestedExpression().coeffRef(row, col); - } - - /** Assigns a triangular matrix to a triangular part of a dense matrix */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const TriangularBase& other); - - /** Shortcut for\code *this = other.other.triangularView<(*this)::Mode>() \endcode */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const MatrixBase& other); - -#ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const TriangularViewImpl& other) - { return *this = other.derived().nestedExpression(); } - - /** \deprecated */ - template - EIGEN_DEVICE_FUNC - void lazyAssign(const TriangularBase& other); - - /** \deprecated */ - template - EIGEN_DEVICE_FUNC - void lazyAssign(const MatrixBase& other); -#endif - - /** Efficient triangular matrix times vector/matrix product */ - template - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& rhs) const - { - return Product(derived(), rhs.derived()); - } - - /** Efficient vector/matrix times triangular matrix product */ - template friend - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& lhs, const TriangularViewImpl& rhs) - { - return Product(lhs.derived(),rhs.derived()); - } - - /** \returns the product of the inverse of \c *this with \a other, \a *this being triangular. - * - * This function computes the inverse-matrix matrix product inverse(\c *this) * \a other if - * \a Side==OnTheLeft (the default), or the right-inverse-multiply \a other * inverse(\c *this) if - * \a Side==OnTheRight. - * - * Note that the template parameter \c Side can be ommitted, in which case \c Side==OnTheLeft - * - * The matrix \c *this must be triangular and invertible (i.e., all the coefficients of the - * diagonal must be non zero). It works as a forward (resp. backward) substitution if \c *this - * is an upper (resp. lower) triangular matrix. - * - * Example: \include Triangular_solve.cpp - * Output: \verbinclude Triangular_solve.out - * - * This function returns an expression of the inverse-multiply and can works in-place if it is assigned - * to the same matrix or vector \a other. - * - * For users coming from BLAS, this function (and more specifically solveInPlace()) offer - * all the operations supported by the \c *TRSV and \c *TRSM BLAS routines. - * - * \sa TriangularView::solveInPlace() - */ - template - EIGEN_DEVICE_FUNC - inline const internal::triangular_solve_retval - solve(const MatrixBase& other) const; - - /** "in-place" version of TriangularView::solve() where the result is written in \a other - * - * \warning The parameter is only marked 'const' to make the C++ compiler accept a temporary expression here. - * This function will const_cast it, so constness isn't honored here. - * - * Note that the template parameter \c Side can be ommitted, in which case \c Side==OnTheLeft - * - * See TriangularView:solve() for the details. - */ - template - EIGEN_DEVICE_FUNC - void solveInPlace(const MatrixBase& other) const; - - template - EIGEN_DEVICE_FUNC - void solveInPlace(const MatrixBase& other) const - { return solveInPlace(other); } - - /** Swaps the coefficients of the common triangular parts of two matrices */ - template - EIGEN_DEVICE_FUNC -#ifdef EIGEN_PARSED_BY_DOXYGEN - void swap(TriangularBase &other) -#else - void swap(TriangularBase const & other) -#endif - { - EIGEN_STATIC_ASSERT_LVALUE(OtherDerived); - call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); - } - - /** \deprecated - * Shortcut for \code (*this).swap(other.triangularView<(*this)::Mode>()) \endcode */ - template - EIGEN_DEVICE_FUNC - void swap(MatrixBase const & other) - { - EIGEN_STATIC_ASSERT_LVALUE(OtherDerived); - call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); - } - - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE void _solve_impl(const RhsType &rhs, DstType &dst) const { - if(!internal::is_same_dense(dst,rhs)) - dst = rhs; - this->solveInPlace(dst); - } - - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE TriangularViewType& _assignProduct(const ProductType& prod, const Scalar& alpha, bool beta); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(TriangularViewImpl) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TriangularViewImpl) - -}; - -/*************************************************************************** -* Implementation of triangular evaluation/assignment -***************************************************************************/ - -#ifndef EIGEN_PARSED_BY_DOXYGEN -// FIXME should we keep that possibility -template -template -inline TriangularView& -TriangularViewImpl::operator=(const MatrixBase& other) -{ - internal::call_assignment_no_alias(derived(), other.derived(), internal::assign_op()); - return derived(); -} - -// FIXME should we keep that possibility -template -template -void TriangularViewImpl::lazyAssign(const MatrixBase& other) -{ - internal::call_assignment_no_alias(derived(), other.template triangularView()); -} - - - -template -template -inline TriangularView& -TriangularViewImpl::operator=(const TriangularBase& other) -{ - eigen_assert(Mode == int(OtherDerived::Mode)); - internal::call_assignment(derived(), other.derived()); - return derived(); -} - -template -template -void TriangularViewImpl::lazyAssign(const TriangularBase& other) -{ - eigen_assert(Mode == int(OtherDerived::Mode)); - internal::call_assignment_no_alias(derived(), other.derived()); -} -#endif - -/*************************************************************************** -* Implementation of TriangularBase methods -***************************************************************************/ - -/** Assigns a triangular or selfadjoint matrix to a dense matrix. - * If the matrix is triangular, the opposite part is set to zero. */ -template -template -void TriangularBase::evalTo(MatrixBase &other) const -{ - evalToLazy(other.derived()); -} - -/*************************************************************************** -* Implementation of TriangularView methods -***************************************************************************/ - -/*************************************************************************** -* Implementation of MatrixBase methods -***************************************************************************/ - -/** - * \returns an expression of a triangular view extracted from the current matrix - * - * The parameter \a Mode can have the following values: \c #Upper, \c #StrictlyUpper, \c #UnitUpper, - * \c #Lower, \c #StrictlyLower, \c #UnitLower. - * - * Example: \include MatrixBase_triangularView.cpp - * Output: \verbinclude MatrixBase_triangularView.out - * - * \sa class TriangularView - */ -template -template -typename MatrixBase::template TriangularViewReturnType::Type -MatrixBase::triangularView() -{ - return typename TriangularViewReturnType::Type(derived()); -} - -/** This is the const version of MatrixBase::triangularView() */ -template -template -typename MatrixBase::template ConstTriangularViewReturnType::Type -MatrixBase::triangularView() const -{ - return typename ConstTriangularViewReturnType::Type(derived()); -} - -/** \returns true if *this is approximately equal to an upper triangular matrix, - * within the precision given by \a prec. - * - * \sa isLowerTriangular() - */ -template -bool MatrixBase::isUpperTriangular(const RealScalar& prec) const -{ - RealScalar maxAbsOnUpperPart = static_cast(-1); - for(Index j = 0; j < cols(); ++j) - { - Index maxi = numext::mini(j, rows()-1); - for(Index i = 0; i <= maxi; ++i) - { - RealScalar absValue = numext::abs(coeff(i,j)); - if(absValue > maxAbsOnUpperPart) maxAbsOnUpperPart = absValue; - } - } - RealScalar threshold = maxAbsOnUpperPart * prec; - for(Index j = 0; j < cols(); ++j) - for(Index i = j+1; i < rows(); ++i) - if(numext::abs(coeff(i, j)) > threshold) return false; - return true; -} - -/** \returns true if *this is approximately equal to a lower triangular matrix, - * within the precision given by \a prec. - * - * \sa isUpperTriangular() - */ -template -bool MatrixBase::isLowerTriangular(const RealScalar& prec) const -{ - RealScalar maxAbsOnLowerPart = static_cast(-1); - for(Index j = 0; j < cols(); ++j) - for(Index i = j; i < rows(); ++i) - { - RealScalar absValue = numext::abs(coeff(i,j)); - if(absValue > maxAbsOnLowerPart) maxAbsOnLowerPart = absValue; - } - RealScalar threshold = maxAbsOnLowerPart * prec; - for(Index j = 1; j < cols(); ++j) - { - Index maxi = numext::mini(j, rows()-1); - for(Index i = 0; i < maxi; ++i) - if(numext::abs(coeff(i, j)) > threshold) return false; - } - return true; -} - - -/*************************************************************************** -**************************************************************************** -* Evaluators and Assignment of triangular expressions -*************************************************************************** -***************************************************************************/ - -namespace internal { - - -// TODO currently a triangular expression has the form TriangularView<.,.> -// in the future triangular-ness should be defined by the expression traits -// such that Transpose > is valid. (currently TriangularBase::transpose() is overloaded to make it work) -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef typename glue_shapes::Shape, TriangularShape>::type Shape; -}; - -template -struct unary_evaluator, IndexBased> - : evaluator::type> -{ - typedef TriangularView XprType; - typedef evaluator::type> Base; - unary_evaluator(const XprType &xpr) : Base(xpr.nestedExpression()) {} -}; - -// Additional assignment kinds: -struct Triangular2Triangular {}; -struct Triangular2Dense {}; -struct Dense2Triangular {}; - - -template struct triangular_assignment_loop; - - -/** \internal Specialization of the dense assignment kernel for triangular matrices. - * The main difference is that the triangular, diagonal, and opposite parts are processed through three different functions. - * \tparam UpLo must be either Lower or Upper - * \tparam Mode must be either 0, UnitDiag, ZeroDiag, or SelfAdjoint - */ -template -class triangular_dense_assignment_kernel : public generic_dense_assignment_kernel -{ -protected: - typedef generic_dense_assignment_kernel Base; - typedef typename Base::DstXprType DstXprType; - typedef typename Base::SrcXprType SrcXprType; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; -public: - - typedef typename Base::DstEvaluatorType DstEvaluatorType; - typedef typename Base::SrcEvaluatorType SrcEvaluatorType; - typedef typename Base::Scalar Scalar; - typedef typename Base::AssignmentTraits AssignmentTraits; - - - EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - -#ifdef EIGEN_INTERNAL_DEBUGGING - EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - Base::assignCoeff(row,col); - } -#else - using Base::assignCoeff; -#endif - - EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id) - { - if(Mode==UnitDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(1)); - else if(Mode==ZeroDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(0)); - else if(Mode==0) Base::assignCoeff(id,id); - } - - EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - if(SetOpposite) - m_functor.assignCoeff(m_dst.coeffRef(row,col), Scalar(0)); - } -}; - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -void call_triangular_assignment_loop(DstXprType& dst, const SrcXprType& src, const Functor &func) -{ - typedef evaluator DstEvaluatorType; - typedef evaluator SrcEvaluatorType; - - SrcEvaluatorType srcEvaluator(src); - - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - DstEvaluatorType dstEvaluator(dst); - - typedef triangular_dense_assignment_kernel< Mode&(Lower|Upper),Mode&(UnitDiag|ZeroDiag|SelfAdjoint),SetOpposite, - DstEvaluatorType,SrcEvaluatorType,Functor> Kernel; - Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived()); - - enum { - unroll = DstXprType::SizeAtCompileTime != Dynamic - && SrcEvaluatorType::CoeffReadCost < HugeCost - && DstXprType::SizeAtCompileTime * (DstEvaluatorType::CoeffReadCost+SrcEvaluatorType::CoeffReadCost) / 2 <= EIGEN_UNROLLING_LIMIT - }; - - triangular_assignment_loop::run(kernel); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -void call_triangular_assignment_loop(DstXprType& dst, const SrcXprType& src) -{ - call_triangular_assignment_loop(dst, src, internal::assign_op()); -} - -template<> struct AssignmentKind { typedef Triangular2Triangular Kind; }; -template<> struct AssignmentKind { typedef Triangular2Dense Kind; }; -template<> struct AssignmentKind { typedef Dense2Triangular Kind; }; - - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - eigen_assert(int(DstXprType::Mode) == int(SrcXprType::Mode)); - - call_triangular_assignment_loop(dst, src, func); - } -}; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - call_triangular_assignment_loop(dst, src, func); - } -}; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - call_triangular_assignment_loop(dst, src, func); - } -}; - - -template -struct triangular_assignment_loop -{ - // FIXME: this is not very clean, perhaps this information should be provided by the kernel? - typedef typename Kernel::DstEvaluatorType DstEvaluatorType; - typedef typename DstEvaluatorType::XprType DstXprType; - - enum { - col = (UnrollCount-1) / DstXprType::RowsAtCompileTime, - row = (UnrollCount-1) % DstXprType::RowsAtCompileTime - }; - - typedef typename Kernel::Scalar Scalar; - - EIGEN_DEVICE_FUNC - static inline void run(Kernel &kernel) - { - triangular_assignment_loop::run(kernel); - - if(row==col) - kernel.assignDiagonalCoeff(row); - else if( ((Mode&Lower) && row>col) || ((Mode&Upper) && row -struct triangular_assignment_loop -{ - EIGEN_DEVICE_FUNC - static inline void run(Kernel &) {} -}; - - - -// TODO: experiment with a recursive assignment procedure splitting the current -// triangular part into one rectangular and two triangular parts. - - -template -struct triangular_assignment_loop -{ - typedef typename Kernel::Scalar Scalar; - EIGEN_DEVICE_FUNC - static inline void run(Kernel &kernel) - { - for(Index j = 0; j < kernel.cols(); ++j) - { - Index maxi = numext::mini(j, kernel.rows()); - Index i = 0; - if (((Mode&Lower) && SetOpposite) || (Mode&Upper)) - { - for(; i < maxi; ++i) - if(Mode&Upper) kernel.assignCoeff(i, j); - else kernel.assignOppositeCoeff(i, j); - } - else - i = maxi; - - if(i -template -void TriangularBase::evalToLazy(MatrixBase &other) const -{ - other.derived().resize(this->rows(), this->cols()); - internal::call_triangular_assignment_loop(other.derived(), derived().nestedExpression()); -} - -namespace internal { - -// Triangular = Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - dst._assignProduct(src, 1, 0); - } -}; - -// Triangular += Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::add_assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &) - { - dst._assignProduct(src, 1, 1); - } -}; - -// Triangular -= Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::sub_assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &) - { - dst._assignProduct(src, -1, 1); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULARMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h deleted file mode 100644 index d72fbf7e99d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h +++ /dev/null @@ -1,96 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_VECTORBLOCK_H -#define EIGEN_VECTORBLOCK_H - -namespace Eigen { - -namespace internal { -template -struct traits > - : public traits::Flags & RowMajorBit ? 1 : Size, - traits::Flags & RowMajorBit ? Size : 1> > -{ -}; -} - -/** \class VectorBlock - * \ingroup Core_Module - * - * \brief Expression of a fixed-size or dynamic-size sub-vector - * - * \tparam VectorType the type of the object in which we are taking a sub-vector - * \tparam Size size of the sub-vector we are taking at compile time (optional) - * - * This class represents an expression of either a fixed-size or dynamic-size sub-vector. - * It is the return type of DenseBase::segment(Index,Index) and DenseBase::segment(Index) and - * most of the time this is the only way it is used. - * - * However, if you want to directly maniputate sub-vector expressions, - * for instance if you want to write a function returning such an expression, you - * will need to use this class. - * - * Here is an example illustrating the dynamic case: - * \include class_VectorBlock.cpp - * Output: \verbinclude class_VectorBlock.out - * - * \note Even though this expression has dynamic size, in the case where \a VectorType - * has fixed size, this expression inherits a fixed maximal size which means that evaluating - * it does not cause a dynamic memory allocation. - * - * Here is an example illustrating the fixed-size case: - * \include class_FixedVectorBlock.cpp - * Output: \verbinclude class_FixedVectorBlock.out - * - * \sa class Block, DenseBase::segment(Index,Index,Index,Index), DenseBase::segment(Index,Index) - */ -template class VectorBlock - : public Block::Flags & RowMajorBit ? 1 : Size, - internal::traits::Flags & RowMajorBit ? Size : 1> -{ - typedef Block::Flags & RowMajorBit ? 1 : Size, - internal::traits::Flags & RowMajorBit ? Size : 1> Base; - enum { - IsColVector = !(internal::traits::Flags & RowMajorBit) - }; - public: - EIGEN_DENSE_PUBLIC_INTERFACE(VectorBlock) - - using Base::operator=; - - /** Dynamic-size constructor - */ - EIGEN_DEVICE_FUNC - inline VectorBlock(VectorType& vector, Index start, Index size) - : Base(vector, - IsColVector ? start : 0, IsColVector ? 0 : start, - IsColVector ? size : 1, IsColVector ? 1 : size) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorBlock); - } - - /** Fixed-size constructor - */ - EIGEN_DEVICE_FUNC - inline VectorBlock(VectorType& vector, Index start) - : Base(vector, IsColVector ? start : 0, IsColVector ? 0 : start) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorBlock); - } -}; - - -} // end namespace Eigen - -#endif // EIGEN_VECTORBLOCK_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h deleted file mode 100644 index 4fe267e9f11..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h +++ /dev/null @@ -1,695 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARTIAL_REDUX_H -#define EIGEN_PARTIAL_REDUX_H - -namespace Eigen { - -/** \class PartialReduxExpr - * \ingroup Core_Module - * - * \brief Generic expression of a partially reduxed matrix - * - * \tparam MatrixType the type of the matrix we are applying the redux operation - * \tparam MemberOp type of the member functor - * \tparam Direction indicates the direction of the redux (#Vertical or #Horizontal) - * - * This class represents an expression of a partial redux operator of a matrix. - * It is the return type of some VectorwiseOp functions, - * and most of the time this is the only way it is used. - * - * \sa class VectorwiseOp - */ - -template< typename MatrixType, typename MemberOp, int Direction> -class PartialReduxExpr; - -namespace internal { -template -struct traits > - : traits -{ - typedef typename MemberOp::result_type Scalar; - typedef typename traits::StorageKind StorageKind; - typedef typename traits::XprKind XprKind; - typedef typename MatrixType::Scalar InputScalar; - enum { - RowsAtCompileTime = Direction==Vertical ? 1 : MatrixType::RowsAtCompileTime, - ColsAtCompileTime = Direction==Horizontal ? 1 : MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = Direction==Vertical ? 1 : MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = Direction==Horizontal ? 1 : MatrixType::MaxColsAtCompileTime, - Flags = RowsAtCompileTime == 1 ? RowMajorBit : 0, - TraversalSize = Direction==Vertical ? MatrixType::RowsAtCompileTime : MatrixType::ColsAtCompileTime - }; -}; -} - -template< typename MatrixType, typename MemberOp, int Direction> -class PartialReduxExpr : public internal::dense_xpr_base< PartialReduxExpr >::type, - internal::no_assignment_operator -{ - public: - - typedef typename internal::dense_xpr_base::type Base; - EIGEN_DENSE_PUBLIC_INTERFACE(PartialReduxExpr) - - EIGEN_DEVICE_FUNC - explicit PartialReduxExpr(const MatrixType& mat, const MemberOp& func = MemberOp()) - : m_matrix(mat), m_functor(func) {} - - EIGEN_DEVICE_FUNC - Index rows() const { return (Direction==Vertical ? 1 : m_matrix.rows()); } - EIGEN_DEVICE_FUNC - Index cols() const { return (Direction==Horizontal ? 1 : m_matrix.cols()); } - - EIGEN_DEVICE_FUNC - typename MatrixType::Nested nestedExpression() const { return m_matrix; } - - EIGEN_DEVICE_FUNC - const MemberOp& functor() const { return m_functor; } - - protected: - typename MatrixType::Nested m_matrix; - const MemberOp m_functor; -}; - -#define EIGEN_MEMBER_FUNCTOR(MEMBER,COST) \ - template \ - struct member_##MEMBER { \ - EIGEN_EMPTY_STRUCT_CTOR(member_##MEMBER) \ - typedef ResultType result_type; \ - template struct Cost \ - { enum { value = COST }; }; \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE \ - ResultType operator()(const XprType& mat) const \ - { return mat.MEMBER(); } \ - } - -namespace internal { - -EIGEN_MEMBER_FUNCTOR(squaredNorm, Size * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(norm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(stableNorm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(blueNorm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(hypotNorm, (Size-1) * functor_traits >::Cost ); -EIGEN_MEMBER_FUNCTOR(sum, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(mean, (Size-1)*NumTraits::AddCost + NumTraits::MulCost); -EIGEN_MEMBER_FUNCTOR(minCoeff, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(maxCoeff, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(all, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(any, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(count, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(prod, (Size-1)*NumTraits::MulCost); - -template -struct member_lpnorm { - typedef ResultType result_type; - template struct Cost - { enum { value = (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost }; }; - EIGEN_DEVICE_FUNC member_lpnorm() {} - template - EIGEN_DEVICE_FUNC inline ResultType operator()(const XprType& mat) const - { return mat.template lpNorm

(); } -}; - -template -struct member_redux { - typedef typename result_of< - BinaryOp(const Scalar&,const Scalar&) - >::type result_type; - template struct Cost - { enum { value = (Size-1) * functor_traits::Cost }; }; - EIGEN_DEVICE_FUNC explicit member_redux(const BinaryOp func) : m_functor(func) {} - template - EIGEN_DEVICE_FUNC inline result_type operator()(const DenseBase& mat) const - { return mat.redux(m_functor); } - const BinaryOp m_functor; -}; -} - -/** \class VectorwiseOp - * \ingroup Core_Module - * - * \brief Pseudo expression providing partial reduction operations - * - * \tparam ExpressionType the type of the object on which to do partial reductions - * \tparam Direction indicates the direction of the redux (#Vertical or #Horizontal) - * - * This class represents a pseudo expression with partial reduction features. - * It is the return type of DenseBase::colwise() and DenseBase::rowwise() - * and most of the time this is the only way it is used. - * - * Example: \include MatrixBase_colwise.cpp - * Output: \verbinclude MatrixBase_colwise.out - * - * \sa DenseBase::colwise(), DenseBase::rowwise(), class PartialReduxExpr - */ -template class VectorwiseOp -{ - public: - - typedef typename ExpressionType::Scalar Scalar; - typedef typename ExpressionType::RealScalar RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef typename internal::ref_selector::non_const_type ExpressionTypeNested; - typedef typename internal::remove_all::type ExpressionTypeNestedCleaned; - - template class Functor, - typename Scalar_=Scalar> struct ReturnType - { - typedef PartialReduxExpr, - Direction - > Type; - }; - - template struct ReduxReturnType - { - typedef PartialReduxExpr, - Direction - > Type; - }; - - enum { - isVertical = (Direction==Vertical) ? 1 : 0, - isHorizontal = (Direction==Horizontal) ? 1 : 0 - }; - - protected: - - typedef typename internal::conditional::type SubVector; - /** \internal - * \returns the i-th subvector according to the \c Direction */ - EIGEN_DEVICE_FUNC - SubVector subVector(Index i) - { - return SubVector(m_matrix.derived(),i); - } - - /** \internal - * \returns the number of subvectors in the direction \c Direction */ - EIGEN_DEVICE_FUNC - Index subVectors() const - { return isVertical?m_matrix.cols():m_matrix.rows(); } - - template struct ExtendedType { - typedef Replicate Type; - }; - - /** \internal - * Replicates a vector to match the size of \c *this */ - template - EIGEN_DEVICE_FUNC - typename ExtendedType::Type - extendedTo(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isVertical, OtherDerived::MaxColsAtCompileTime==1), - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED) - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isHorizontal, OtherDerived::MaxRowsAtCompileTime==1), - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED) - return typename ExtendedType::Type - (other.derived(), - isVertical ? 1 : m_matrix.rows(), - isHorizontal ? 1 : m_matrix.cols()); - } - - template struct OppositeExtendedType { - typedef Replicate Type; - }; - - /** \internal - * Replicates a vector in the opposite direction to match the size of \c *this */ - template - EIGEN_DEVICE_FUNC - typename OppositeExtendedType::Type - extendedToOpposite(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isHorizontal, OtherDerived::MaxColsAtCompileTime==1), - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED) - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isVertical, OtherDerived::MaxRowsAtCompileTime==1), - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED) - return typename OppositeExtendedType::Type - (other.derived(), - isHorizontal ? 1 : m_matrix.rows(), - isVertical ? 1 : m_matrix.cols()); - } - - public: - EIGEN_DEVICE_FUNC - explicit inline VectorwiseOp(ExpressionType& matrix) : m_matrix(matrix) {} - - /** \internal */ - EIGEN_DEVICE_FUNC - inline const ExpressionType& _expression() const { return m_matrix; } - - /** \returns a row or column vector expression of \c *this reduxed by \a func - * - * The template parameter \a BinaryOp is the type of the functor - * of the custom redux operator. Note that func must be an associative operator. - * - * \sa class VectorwiseOp, DenseBase::colwise(), DenseBase::rowwise() - */ - template - EIGEN_DEVICE_FUNC - const typename ReduxReturnType::Type - redux(const BinaryOp& func = BinaryOp()) const - { return typename ReduxReturnType::Type(_expression(), internal::member_redux(func)); } - - typedef typename ReturnType::Type MinCoeffReturnType; - typedef typename ReturnType::Type MaxCoeffReturnType; - typedef typename ReturnType::Type SquaredNormReturnType; - typedef typename ReturnType::Type NormReturnType; - typedef typename ReturnType::Type BlueNormReturnType; - typedef typename ReturnType::Type StableNormReturnType; - typedef typename ReturnType::Type HypotNormReturnType; - typedef typename ReturnType::Type SumReturnType; - typedef typename ReturnType::Type MeanReturnType; - typedef typename ReturnType::Type AllReturnType; - typedef typename ReturnType::Type AnyReturnType; - typedef PartialReduxExpr, Direction> CountReturnType; - typedef typename ReturnType::Type ProdReturnType; - typedef Reverse ConstReverseReturnType; - typedef Reverse ReverseReturnType; - - template struct LpNormReturnType { - typedef PartialReduxExpr,Direction> Type; - }; - - /** \returns a row (or column) vector expression of the smallest coefficient - * of each column (or row) of the referenced expression. - * - * \warning the result is undefined if \c *this contains NaN. - * - * Example: \include PartialRedux_minCoeff.cpp - * Output: \verbinclude PartialRedux_minCoeff.out - * - * \sa DenseBase::minCoeff() */ - EIGEN_DEVICE_FUNC - const MinCoeffReturnType minCoeff() const - { return MinCoeffReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the largest coefficient - * of each column (or row) of the referenced expression. - * - * \warning the result is undefined if \c *this contains NaN. - * - * Example: \include PartialRedux_maxCoeff.cpp - * Output: \verbinclude PartialRedux_maxCoeff.out - * - * \sa DenseBase::maxCoeff() */ - EIGEN_DEVICE_FUNC - const MaxCoeffReturnType maxCoeff() const - { return MaxCoeffReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the squared norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_squaredNorm.cpp - * Output: \verbinclude PartialRedux_squaredNorm.out - * - * \sa DenseBase::squaredNorm() */ - EIGEN_DEVICE_FUNC - const SquaredNormReturnType squaredNorm() const - { return SquaredNormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_norm.cpp - * Output: \verbinclude PartialRedux_norm.out - * - * \sa DenseBase::norm() */ - EIGEN_DEVICE_FUNC - const NormReturnType norm() const - { return NormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_norm.cpp - * Output: \verbinclude PartialRedux_norm.out - * - * \sa DenseBase::norm() */ - template - EIGEN_DEVICE_FUNC - const typename LpNormReturnType

::Type lpNorm() const - { return typename LpNormReturnType

::Type(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, using - * Blue's algorithm. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::blueNorm() */ - EIGEN_DEVICE_FUNC - const BlueNormReturnType blueNorm() const - { return BlueNormReturnType(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, avoiding - * underflow and overflow. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::stableNorm() */ - EIGEN_DEVICE_FUNC - const StableNormReturnType stableNorm() const - { return StableNormReturnType(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, avoiding - * underflow and overflow using a concatenation of hypot() calls. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::hypotNorm() */ - EIGEN_DEVICE_FUNC - const HypotNormReturnType hypotNorm() const - { return HypotNormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the sum - * of each column (or row) of the referenced expression. - * - * Example: \include PartialRedux_sum.cpp - * Output: \verbinclude PartialRedux_sum.out - * - * \sa DenseBase::sum() */ - EIGEN_DEVICE_FUNC - const SumReturnType sum() const - { return SumReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the mean - * of each column (or row) of the referenced expression. - * - * \sa DenseBase::mean() */ - EIGEN_DEVICE_FUNC - const MeanReturnType mean() const - { return MeanReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * whether \b all coefficients of each respective column (or row) are \c true. - * This expression can be assigned to a vector with entries of type \c bool. - * - * \sa DenseBase::all() */ - EIGEN_DEVICE_FUNC - const AllReturnType all() const - { return AllReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * whether \b at \b least one coefficient of each respective column (or row) is \c true. - * This expression can be assigned to a vector with entries of type \c bool. - * - * \sa DenseBase::any() */ - EIGEN_DEVICE_FUNC - const AnyReturnType any() const - { return AnyReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * the number of \c true coefficients of each respective column (or row). - * This expression can be assigned to a vector whose entries have the same type as is used to - * index entries of the original matrix; for dense matrices, this is \c std::ptrdiff_t . - * - * Example: \include PartialRedux_count.cpp - * Output: \verbinclude PartialRedux_count.out - * - * \sa DenseBase::count() */ - EIGEN_DEVICE_FUNC - const CountReturnType count() const - { return CountReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the product - * of each column (or row) of the referenced expression. - * - * Example: \include PartialRedux_prod.cpp - * Output: \verbinclude PartialRedux_prod.out - * - * \sa DenseBase::prod() */ - EIGEN_DEVICE_FUNC - const ProdReturnType prod() const - { return ProdReturnType(_expression()); } - - - /** \returns a matrix expression - * where each column (or row) are reversed. - * - * Example: \include Vectorwise_reverse.cpp - * Output: \verbinclude Vectorwise_reverse.out - * - * \sa DenseBase::reverse() */ - EIGEN_DEVICE_FUNC - const ConstReverseReturnType reverse() const - { return ConstReverseReturnType( _expression() ); } - - /** \returns a writable matrix expression - * where each column (or row) are reversed. - * - * \sa reverse() const */ - EIGEN_DEVICE_FUNC - ReverseReturnType reverse() - { return ReverseReturnType( _expression() ); } - - typedef Replicate ReplicateReturnType; - EIGEN_DEVICE_FUNC - const ReplicateReturnType replicate(Index factor) const; - - /** - * \return an expression of the replication of each column (or row) of \c *this - * - * Example: \include DirectionWise_replicate.cpp - * Output: \verbinclude DirectionWise_replicate.out - * - * \sa VectorwiseOp::replicate(Index), DenseBase::replicate(), class Replicate - */ - // NOTE implemented here because of sunstudio's compilation errors - // isVertical*Factor+isHorizontal instead of (isVertical?Factor:1) to handle CUDA bug with ternary operator - template const Replicate - EIGEN_DEVICE_FUNC - replicate(Index factor = Factor) const - { - return Replicate - (_expression(),isVertical?factor:1,isHorizontal?factor:1); - } - -/////////// Artithmetic operators /////////// - - /** Copies the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - //eigen_assert((m_matrix.isNull()) == (other.isNull())); FIXME - return const_cast(m_matrix = extendedTo(other.derived())); - } - - /** Adds the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator+=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return const_cast(m_matrix += extendedTo(other.derived())); - } - - /** Substracts the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator-=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return const_cast(m_matrix -= extendedTo(other.derived())); - } - - /** Multiples each subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator*=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - m_matrix *= extendedTo(other.derived()); - return const_cast(m_matrix); - } - - /** Divides each subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator/=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - m_matrix /= extendedTo(other.derived()); - return const_cast(m_matrix); - } - - /** Returns the expression of the sum of the vector \a other to each subvector of \c *this */ - template EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator+(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix + extendedTo(other.derived()); - } - - /** Returns the expression of the difference between each subvector of \c *this and the vector \a other */ - template - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator-(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix - extendedTo(other.derived()); - } - - /** Returns the expression where each subvector is the product of the vector \a other - * by the corresponding subvector of \c *this */ - template EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - EIGEN_DEVICE_FUNC - operator*(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix * extendedTo(other.derived()); - } - - /** Returns the expression where each subvector is the quotient of the corresponding - * subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator/(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix / extendedTo(other.derived()); - } - - /** \returns an expression where each column (or row) of the referenced matrix are normalized. - * The referenced matrix is \b not modified. - * \sa MatrixBase::normalized(), normalize() - */ - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename OppositeExtendedType::Type>::Type> - normalized() const { return m_matrix.cwiseQuotient(extendedToOpposite(this->norm())); } - - - /** Normalize in-place each row or columns of the referenced matrix. - * \sa MatrixBase::normalize(), normalized() - */ - EIGEN_DEVICE_FUNC void normalize() { - m_matrix = this->normalized(); - } - - EIGEN_DEVICE_FUNC inline void reverseInPlace(); - -/////////// Geometry module /////////// - - typedef Homogeneous HomogeneousReturnType; - EIGEN_DEVICE_FUNC - HomogeneousReturnType homogeneous() const; - - typedef typename ExpressionType::PlainObject CrossReturnType; - template - EIGEN_DEVICE_FUNC - const CrossReturnType cross(const MatrixBase& other) const; - - enum { - HNormalized_Size = Direction==Vertical ? internal::traits::RowsAtCompileTime - : internal::traits::ColsAtCompileTime, - HNormalized_SizeMinusOne = HNormalized_Size==Dynamic ? Dynamic : HNormalized_Size-1 - }; - typedef Block::RowsAtCompileTime), - Direction==Horizontal ? int(HNormalized_SizeMinusOne) - : int(internal::traits::ColsAtCompileTime)> - HNormalized_Block; - typedef Block::RowsAtCompileTime), - Direction==Horizontal ? 1 : int(internal::traits::ColsAtCompileTime)> - HNormalized_Factors; - typedef CwiseBinaryOp::Scalar>, - const HNormalized_Block, - const Replicate > - HNormalizedReturnType; - - EIGEN_DEVICE_FUNC - const HNormalizedReturnType hnormalized() const; - - protected: - ExpressionTypeNested m_matrix; -}; - -//const colwise moved to DenseBase.h due to CUDA compiler bug - - -/** \returns a writable VectorwiseOp wrapper of *this providing additional partial reduction operations - * - * \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting - */ -template -inline typename DenseBase::ColwiseReturnType -DenseBase::colwise() -{ - return ColwiseReturnType(derived()); -} - -//const rowwise moved to DenseBase.h due to CUDA compiler bug - - -/** \returns a writable VectorwiseOp wrapper of *this providing additional partial reduction operations - * - * \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting - */ -template -inline typename DenseBase::RowwiseReturnType -DenseBase::rowwise() -{ - return RowwiseReturnType(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_PARTIAL_REDUX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h deleted file mode 100644 index 54c1883d98f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h +++ /dev/null @@ -1,273 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_VISITOR_H -#define EIGEN_VISITOR_H - -namespace Eigen { - -namespace internal { - -template -struct visitor_impl -{ - enum { - col = (UnrollCount-1) / Derived::RowsAtCompileTime, - row = (UnrollCount-1) % Derived::RowsAtCompileTime - }; - - EIGEN_DEVICE_FUNC - static inline void run(const Derived &mat, Visitor& visitor) - { - visitor_impl::run(mat, visitor); - visitor(mat.coeff(row, col), row, col); - } -}; - -template -struct visitor_impl -{ - EIGEN_DEVICE_FUNC - static inline void run(const Derived &mat, Visitor& visitor) - { - return visitor.init(mat.coeff(0, 0), 0, 0); - } -}; - -template -struct visitor_impl -{ - EIGEN_DEVICE_FUNC - static inline void run(const Derived& mat, Visitor& visitor) - { - visitor.init(mat.coeff(0,0), 0, 0); - for(Index i = 1; i < mat.rows(); ++i) - visitor(mat.coeff(i, 0), i, 0); - for(Index j = 1; j < mat.cols(); ++j) - for(Index i = 0; i < mat.rows(); ++i) - visitor(mat.coeff(i, j), i, j); - } -}; - -// evaluator adaptor -template -class visitor_evaluator -{ -public: - EIGEN_DEVICE_FUNC - explicit visitor_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {} - - typedef typename XprType::Scalar Scalar; - typedef typename XprType::CoeffReturnType CoeffReturnType; - - enum { - RowsAtCompileTime = XprType::RowsAtCompileTime, - CoeffReadCost = internal::evaluator::CoeffReadCost - }; - - EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } - EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } - EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); } - - EIGEN_DEVICE_FUNC CoeffReturnType coeff(Index row, Index col) const - { return m_evaluator.coeff(row, col); } - -protected: - internal::evaluator m_evaluator; - const XprType &m_xpr; -}; -} // end namespace internal - -/** Applies the visitor \a visitor to the whole coefficients of the matrix or vector. - * - * The template parameter \a Visitor is the type of the visitor and provides the following interface: - * \code - * struct MyVisitor { - * // called for the first coefficient - * void init(const Scalar& value, Index i, Index j); - * // called for all other coefficients - * void operator() (const Scalar& value, Index i, Index j); - * }; - * \endcode - * - * \note compared to one or two \em for \em loops, visitors offer automatic - * unrolling for small fixed size matrix. - * - * \sa minCoeff(Index*,Index*), maxCoeff(Index*,Index*), DenseBase::redux() - */ -template -template -EIGEN_DEVICE_FUNC -void DenseBase::visit(Visitor& visitor) const -{ - typedef typename internal::visitor_evaluator ThisEvaluator; - ThisEvaluator thisEval(derived()); - - enum { - unroll = SizeAtCompileTime != Dynamic - && SizeAtCompileTime * ThisEvaluator::CoeffReadCost + (SizeAtCompileTime-1) * internal::functor_traits::Cost <= EIGEN_UNROLLING_LIMIT - }; - return internal::visitor_impl::run(thisEval, visitor); -} - -namespace internal { - -/** \internal - * \brief Base class to implement min and max visitors - */ -template -struct coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - Index row, col; - Scalar res; - EIGEN_DEVICE_FUNC - inline void init(const Scalar& value, Index i, Index j) - { - res = value; - row = i; - col = j; - } -}; - -/** \internal - * \brief Visitor computing the min coefficient with its value and coordinates - * - * \sa DenseBase::minCoeff(Index*, Index*) - */ -template -struct min_coeff_visitor : coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - EIGEN_DEVICE_FUNC - void operator() (const Scalar& value, Index i, Index j) - { - if(value < this->res) - { - this->res = value; - this->row = i; - this->col = j; - } - } -}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost - }; -}; - -/** \internal - * \brief Visitor computing the max coefficient with its value and coordinates - * - * \sa DenseBase::maxCoeff(Index*, Index*) - */ -template -struct max_coeff_visitor : coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - EIGEN_DEVICE_FUNC - void operator() (const Scalar& value, Index i, Index j) - { - if(value > this->res) - { - this->res = value; - this->row = i; - this->col = j; - } - } -}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost - }; -}; - -} // end namespace internal - -/** \fn DenseBase::minCoeff(IndexType* rowId, IndexType* colId) const - * \returns the minimum of all coefficients of *this and puts in *row and *col its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(Index*), DenseBase::maxCoeff(Index*,Index*), DenseBase::visit(), DenseBase::minCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::minCoeff(IndexType* rowId, IndexType* colId) const -{ - internal::min_coeff_visitor minVisitor; - this->visit(minVisitor); - *rowId = minVisitor.row; - if (colId) *colId = minVisitor.col; - return minVisitor.res; -} - -/** \returns the minimum of all coefficients of *this and puts in *index its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::visit(), DenseBase::minCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::minCoeff(IndexType* index) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - internal::min_coeff_visitor minVisitor; - this->visit(minVisitor); - *index = IndexType((RowsAtCompileTime==1) ? minVisitor.col : minVisitor.row); - return minVisitor.res; -} - -/** \fn DenseBase::maxCoeff(IndexType* rowId, IndexType* colId) const - * \returns the maximum of all coefficients of *this and puts in *row and *col its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visit(), DenseBase::maxCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::maxCoeff(IndexType* rowPtr, IndexType* colPtr) const -{ - internal::max_coeff_visitor maxVisitor; - this->visit(maxVisitor); - *rowPtr = maxVisitor.row; - if (colPtr) *colPtr = maxVisitor.col; - return maxVisitor.res; -} - -/** \returns the maximum of all coefficients of *this and puts in *index its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visitor(), DenseBase::maxCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::maxCoeff(IndexType* index) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - internal::max_coeff_visitor maxVisitor; - this->visit(maxVisitor); - *index = (RowsAtCompileTime==1) ? maxVisitor.col : maxVisitor.row; - return maxVisitor.res; -} - -} // end namespace Eigen - -#endif // EIGEN_VISITOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/Complex.h deleted file mode 100644 index 7fa61969dc1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/Complex.h +++ /dev/null @@ -1,451 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner (benoit.steiner.goog@gmail.com) -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_AVX_H -#define EIGEN_COMPLEX_AVX_H - -namespace Eigen { - -namespace internal { - -//---------- float ---------- -struct Packet4cf -{ - EIGEN_STRONG_INLINE Packet4cf() {} - EIGEN_STRONG_INLINE explicit Packet4cf(const __m256& a) : v(a) {} - __m256 v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet4cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket = 1, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=4, alignment=Aligned32}; typedef Packet2cf half; }; - -template<> EIGEN_STRONG_INLINE Packet4cf padd(const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_add_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet4cf psub(const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_sub_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet4cf pnegate(const Packet4cf& a) -{ - return Packet4cf(pnegate(a.v)); -} -template<> EIGEN_STRONG_INLINE Packet4cf pconj(const Packet4cf& a) -{ - const __m256 mask = _mm256_castsi256_ps(_mm256_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000,0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet4cf(_mm256_xor_ps(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet4cf pmul(const Packet4cf& a, const Packet4cf& b) -{ - __m256 tmp1 = _mm256_mul_ps(_mm256_moveldup_ps(a.v), b.v); - __m256 tmp2 = _mm256_mul_ps(_mm256_movehdup_ps(a.v), _mm256_permute_ps(b.v, _MM_SHUFFLE(2,3,0,1))); - __m256 result = _mm256_addsub_ps(tmp1, tmp2); - return Packet4cf(result); -} - -template<> EIGEN_STRONG_INLINE Packet4cf pand (const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_and_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet4cf por (const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_or_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet4cf pxor (const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_xor_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet4cf pandnot(const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_andnot_ps(a.v,b.v)); } - -template<> EIGEN_STRONG_INLINE Packet4cf pload (const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet4cf(pload(&numext::real_ref(*from))); } -template<> EIGEN_STRONG_INLINE Packet4cf ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet4cf(ploadu(&numext::real_ref(*from))); } - - -template<> EIGEN_STRONG_INLINE Packet4cf pset1(const std::complex& from) -{ - return Packet4cf(_mm256_castpd_ps(_mm256_broadcast_sd((const double*)(const void*)&from))); -} - -template<> EIGEN_STRONG_INLINE Packet4cf ploaddup(const std::complex* from) -{ - // FIXME The following might be optimized using _mm256_movedup_pd - Packet2cf a = ploaddup(from); - Packet2cf b = ploaddup(from+1); - return Packet4cf(_mm256_insertf128_ps(_mm256_castps128_ps256(a.v), b.v, 1)); -} - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex* to, const Packet4cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex* to, const Packet4cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), from.v); } - -template<> EIGEN_DEVICE_FUNC inline Packet4cf pgather, Packet4cf>(const std::complex* from, Index stride) -{ - return Packet4cf(_mm256_set_ps(std::imag(from[3*stride]), std::real(from[3*stride]), - std::imag(from[2*stride]), std::real(from[2*stride]), - std::imag(from[1*stride]), std::real(from[1*stride]), - std::imag(from[0*stride]), std::real(from[0*stride]))); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet4cf>(std::complex* to, const Packet4cf& from, Index stride) -{ - __m128 low = _mm256_extractf128_ps(from.v, 0); - to[stride*0] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(low, low, 0)), - _mm_cvtss_f32(_mm_shuffle_ps(low, low, 1))); - to[stride*1] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(low, low, 2)), - _mm_cvtss_f32(_mm_shuffle_ps(low, low, 3))); - - __m128 high = _mm256_extractf128_ps(from.v, 1); - to[stride*2] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(high, high, 0)), - _mm_cvtss_f32(_mm_shuffle_ps(high, high, 1))); - to[stride*3] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(high, high, 2)), - _mm_cvtss_f32(_mm_shuffle_ps(high, high, 3))); - -} - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet4cf& a) -{ - return pfirst(Packet2cf(_mm256_castps256_ps128(a.v))); -} - -template<> EIGEN_STRONG_INLINE Packet4cf preverse(const Packet4cf& a) { - __m128 low = _mm256_extractf128_ps(a.v, 0); - __m128 high = _mm256_extractf128_ps(a.v, 1); - __m128d lowd = _mm_castps_pd(low); - __m128d highd = _mm_castps_pd(high); - low = _mm_castpd_ps(_mm_shuffle_pd(lowd,lowd,0x1)); - high = _mm_castpd_ps(_mm_shuffle_pd(highd,highd,0x1)); - __m256 result = _mm256_setzero_ps(); - result = _mm256_insertf128_ps(result, low, 1); - result = _mm256_insertf128_ps(result, high, 0); - return Packet4cf(result); -} - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet4cf& a) -{ - return predux(padd(Packet2cf(_mm256_extractf128_ps(a.v,0)), - Packet2cf(_mm256_extractf128_ps(a.v,1)))); -} - -template<> EIGEN_STRONG_INLINE Packet4cf preduxp(const Packet4cf* vecs) -{ - Packet8f t0 = _mm256_shuffle_ps(vecs[0].v, vecs[0].v, _MM_SHUFFLE(3, 1, 2 ,0)); - Packet8f t1 = _mm256_shuffle_ps(vecs[1].v, vecs[1].v, _MM_SHUFFLE(3, 1, 2 ,0)); - t0 = _mm256_hadd_ps(t0,t1); - Packet8f t2 = _mm256_shuffle_ps(vecs[2].v, vecs[2].v, _MM_SHUFFLE(3, 1, 2 ,0)); - Packet8f t3 = _mm256_shuffle_ps(vecs[3].v, vecs[3].v, _MM_SHUFFLE(3, 1, 2 ,0)); - t2 = _mm256_hadd_ps(t2,t3); - - t1 = _mm256_permute2f128_ps(t0,t2, 0 + (2<<4)); - t3 = _mm256_permute2f128_ps(t0,t2, 1 + (3<<4)); - - return Packet4cf(_mm256_add_ps(t1,t3)); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet4cf& a) -{ - return predux_mul(pmul(Packet2cf(_mm256_extractf128_ps(a.v, 0)), - Packet2cf(_mm256_extractf128_ps(a.v, 1)))); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4cf& first, const Packet4cf& second) - { - if (Offset==0) return; - palign_impl::run(first.v, second.v); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet4cf pmadd(const Packet4cf& x, const Packet4cf& y, const Packet4cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet4cf pmul(const Packet4cf& a, const Packet4cf& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet4cf pmadd(const Packet4cf& x, const Packet4cf& y, const Packet4cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet4cf pmul(const Packet4cf& a, const Packet4cf& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet4cf pmadd(const Packet4cf& x, const Packet4cf& y, const Packet4cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet4cf pmul(const Packet4cf& a, const Packet4cf& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet4cf,Packet8f) - -template<> EIGEN_STRONG_INLINE Packet4cf pdiv(const Packet4cf& a, const Packet4cf& b) -{ - Packet4cf num = pmul(a, pconj(b)); - __m256 tmp = _mm256_mul_ps(b.v, b.v); - __m256 tmp2 = _mm256_shuffle_ps(tmp,tmp,0xB1); - __m256 denom = _mm256_add_ps(tmp, tmp2); - return Packet4cf(_mm256_div_ps(num.v, denom)); -} - -template<> EIGEN_STRONG_INLINE Packet4cf pcplxflip(const Packet4cf& x) -{ - return Packet4cf(_mm256_shuffle_ps(x.v, x.v, _MM_SHUFFLE(2, 3, 0 ,1))); -} - -//---------- double ---------- -struct Packet2cd -{ - EIGEN_STRONG_INLINE Packet2cd() {} - EIGEN_STRONG_INLINE explicit Packet2cd(const __m256d& a) : v(a) {} - __m256d v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 0, - size = 2, - HasHalfPacket = 1, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned32}; typedef Packet1cd half; }; - -template<> EIGEN_STRONG_INLINE Packet2cd padd(const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_add_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd psub(const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_sub_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd pnegate(const Packet2cd& a) { return Packet2cd(pnegate(a.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd pconj(const Packet2cd& a) -{ - const __m256d mask = _mm256_castsi256_pd(_mm256_set_epi32(0x80000000,0x0,0x0,0x0,0x80000000,0x0,0x0,0x0)); - return Packet2cd(_mm256_xor_pd(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet2cd pmul(const Packet2cd& a, const Packet2cd& b) -{ - __m256d tmp1 = _mm256_shuffle_pd(a.v,a.v,0x0); - __m256d even = _mm256_mul_pd(tmp1, b.v); - __m256d tmp2 = _mm256_shuffle_pd(a.v,a.v,0xF); - __m256d tmp3 = _mm256_shuffle_pd(b.v,b.v,0x5); - __m256d odd = _mm256_mul_pd(tmp2, tmp3); - return Packet2cd(_mm256_addsub_pd(even, odd)); -} - -template<> EIGEN_STRONG_INLINE Packet2cd pand (const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_and_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd por (const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_or_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd pxor (const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_xor_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cd pandnot(const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_andnot_pd(a.v,b.v)); } - -template<> EIGEN_STRONG_INLINE Packet2cd pload (const std::complex* from) -{ EIGEN_DEBUG_ALIGNED_LOAD return Packet2cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet2cd ploadu(const std::complex* from) -{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cd(ploadu((const double*)from)); } - -template<> EIGEN_STRONG_INLINE Packet2cd pset1(const std::complex& from) -{ - // in case casting to a __m128d* is really not safe, then we can still fallback to this version: (much slower though) -// return Packet2cd(_mm256_loadu2_m128d((const double*)&from,(const double*)&from)); - return Packet2cd(_mm256_broadcast_pd((const __m128d*)(const void*)&from)); -} - -template<> EIGEN_STRONG_INLINE Packet2cd ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, from.v); } - -template<> EIGEN_DEVICE_FUNC inline Packet2cd pgather, Packet2cd>(const std::complex* from, Index stride) -{ - return Packet2cd(_mm256_set_pd(std::imag(from[1*stride]), std::real(from[1*stride]), - std::imag(from[0*stride]), std::real(from[0*stride]))); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cd>(std::complex* to, const Packet2cd& from, Index stride) -{ - __m128d low = _mm256_extractf128_pd(from.v, 0); - to[stride*0] = std::complex(_mm_cvtsd_f64(low), _mm_cvtsd_f64(_mm_shuffle_pd(low, low, 1))); - __m128d high = _mm256_extractf128_pd(from.v, 1); - to[stride*1] = std::complex(_mm_cvtsd_f64(high), _mm_cvtsd_f64(_mm_shuffle_pd(high, high, 1))); -} - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cd& a) -{ - __m128d low = _mm256_extractf128_pd(a.v, 0); - EIGEN_ALIGN16 double res[2]; - _mm_store_pd(res, low); - return std::complex(res[0],res[1]); -} - -template<> EIGEN_STRONG_INLINE Packet2cd preverse(const Packet2cd& a) { - __m256d result = _mm256_permute2f128_pd(a.v, a.v, 1); - return Packet2cd(result); -} - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cd& a) -{ - return predux(padd(Packet1cd(_mm256_extractf128_pd(a.v,0)), - Packet1cd(_mm256_extractf128_pd(a.v,1)))); -} - -template<> EIGEN_STRONG_INLINE Packet2cd preduxp(const Packet2cd* vecs) -{ - Packet4d t0 = _mm256_permute2f128_pd(vecs[0].v,vecs[1].v, 0 + (2<<4)); - Packet4d t1 = _mm256_permute2f128_pd(vecs[0].v,vecs[1].v, 1 + (3<<4)); - - return Packet2cd(_mm256_add_pd(t0,t1)); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cd& a) -{ - return predux(pmul(Packet1cd(_mm256_extractf128_pd(a.v,0)), - Packet1cd(_mm256_extractf128_pd(a.v,1)))); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2cd& first, const Packet2cd& second) - { - if (Offset==0) return; - palign_impl::run(first.v, second.v); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cd pmadd(const Packet2cd& x, const Packet2cd& y, const Packet2cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cd pmul(const Packet2cd& a, const Packet2cd& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cd pmadd(const Packet2cd& x, const Packet2cd& y, const Packet2cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cd pmul(const Packet2cd& a, const Packet2cd& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cd pmadd(const Packet2cd& x, const Packet2cd& y, const Packet2cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cd pmul(const Packet2cd& a, const Packet2cd& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cd,Packet4d) - -template<> EIGEN_STRONG_INLINE Packet2cd pdiv(const Packet2cd& a, const Packet2cd& b) -{ - Packet2cd num = pmul(a, pconj(b)); - __m256d tmp = _mm256_mul_pd(b.v, b.v); - __m256d denom = _mm256_hadd_pd(tmp, tmp); - return Packet2cd(_mm256_div_pd(num.v, denom)); -} - -template<> EIGEN_STRONG_INLINE Packet2cd pcplxflip(const Packet2cd& x) -{ - return Packet2cd(_mm256_shuffle_pd(x.v, x.v, 0x5)); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m256d P0 = _mm256_castps_pd(kernel.packet[0].v); - __m256d P1 = _mm256_castps_pd(kernel.packet[1].v); - __m256d P2 = _mm256_castps_pd(kernel.packet[2].v); - __m256d P3 = _mm256_castps_pd(kernel.packet[3].v); - - __m256d T0 = _mm256_shuffle_pd(P0, P1, 15); - __m256d T1 = _mm256_shuffle_pd(P0, P1, 0); - __m256d T2 = _mm256_shuffle_pd(P2, P3, 15); - __m256d T3 = _mm256_shuffle_pd(P2, P3, 0); - - kernel.packet[1].v = _mm256_castpd_ps(_mm256_permute2f128_pd(T0, T2, 32)); - kernel.packet[3].v = _mm256_castpd_ps(_mm256_permute2f128_pd(T0, T2, 49)); - kernel.packet[0].v = _mm256_castpd_ps(_mm256_permute2f128_pd(T1, T3, 32)); - kernel.packet[2].v = _mm256_castpd_ps(_mm256_permute2f128_pd(T1, T3, 49)); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m256d tmp = _mm256_permute2f128_pd(kernel.packet[0].v, kernel.packet[1].v, 0+(2<<4)); - kernel.packet[1].v = _mm256_permute2f128_pd(kernel.packet[0].v, kernel.packet[1].v, 1+(3<<4)); - kernel.packet[0].v = tmp; -} - -template<> EIGEN_STRONG_INLINE Packet4cf pinsertfirst(const Packet4cf& a, std::complex b) -{ - return Packet4cf(_mm256_blend_ps(a.v,pset1(b).v,1|2)); -} - -template<> EIGEN_STRONG_INLINE Packet2cd pinsertfirst(const Packet2cd& a, std::complex b) -{ - return Packet2cd(_mm256_blend_pd(a.v,pset1(b).v,1|2)); -} - -template<> EIGEN_STRONG_INLINE Packet4cf pinsertlast(const Packet4cf& a, std::complex b) -{ - return Packet4cf(_mm256_blend_ps(a.v,pset1(b).v,(1<<7)|(1<<6))); -} - -template<> EIGEN_STRONG_INLINE Packet2cd pinsertlast(const Packet2cd& a, std::complex b) -{ - return Packet2cd(_mm256_blend_pd(a.v,pset1(b).v,(1<<3)|(1<<2))); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_AVX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/MathFunctions.h deleted file mode 100644 index 6af67ce2d65..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/MathFunctions.h +++ /dev/null @@ -1,439 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Pedro Gonnet (pedro.gonnet@gmail.com) -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MATH_FUNCTIONS_AVX_H -#define EIGEN_MATH_FUNCTIONS_AVX_H - -/* The sin, cos, exp, and log functions of this file are loosely derived from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -namespace Eigen { - -namespace internal { - -inline Packet8i pshiftleft(Packet8i v, int n) -{ -#ifdef EIGEN_VECTORIZE_AVX2 - return _mm256_slli_epi32(v, n); -#else - __m128i lo = _mm_slli_epi32(_mm256_extractf128_si256(v, 0), n); - __m128i hi = _mm_slli_epi32(_mm256_extractf128_si256(v, 1), n); - return _mm256_insertf128_si256(_mm256_castsi128_si256(lo), (hi), 1); -#endif -} - -inline Packet8f pshiftright(Packet8f v, int n) -{ -#ifdef EIGEN_VECTORIZE_AVX2 - return _mm256_cvtepi32_ps(_mm256_srli_epi32(_mm256_castps_si256(v), n)); -#else - __m128i lo = _mm_srli_epi32(_mm256_extractf128_si256(_mm256_castps_si256(v), 0), n); - __m128i hi = _mm_srli_epi32(_mm256_extractf128_si256(_mm256_castps_si256(v), 1), n); - return _mm256_cvtepi32_ps(_mm256_insertf128_si256(_mm256_castsi128_si256(lo), (hi), 1)); -#endif -} - -// Sine function -// Computes sin(x) by wrapping x to the interval [-Pi/4,3*Pi/4] and -// evaluating interpolants in [-Pi/4,Pi/4] or [Pi/4,3*Pi/4]. The interpolants -// are (anti-)symmetric and thus have only odd/even coefficients -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8f -psin(const Packet8f& _x) { - Packet8f x = _x; - - // Some useful values. - _EIGEN_DECLARE_CONST_Packet8i(one, 1); - _EIGEN_DECLARE_CONST_Packet8f(one, 1.0f); - _EIGEN_DECLARE_CONST_Packet8f(two, 2.0f); - _EIGEN_DECLARE_CONST_Packet8f(one_over_four, 0.25f); - _EIGEN_DECLARE_CONST_Packet8f(one_over_pi, 3.183098861837907e-01f); - _EIGEN_DECLARE_CONST_Packet8f(neg_pi_first, -3.140625000000000e+00f); - _EIGEN_DECLARE_CONST_Packet8f(neg_pi_second, -9.670257568359375e-04f); - _EIGEN_DECLARE_CONST_Packet8f(neg_pi_third, -6.278329571784980e-07f); - _EIGEN_DECLARE_CONST_Packet8f(four_over_pi, 1.273239544735163e+00f); - - // Map x from [-Pi/4,3*Pi/4] to z in [-1,3] and subtract the shifted period. - Packet8f z = pmul(x, p8f_one_over_pi); - Packet8f shift = _mm256_floor_ps(padd(z, p8f_one_over_four)); - x = pmadd(shift, p8f_neg_pi_first, x); - x = pmadd(shift, p8f_neg_pi_second, x); - x = pmadd(shift, p8f_neg_pi_third, x); - z = pmul(x, p8f_four_over_pi); - - // Make a mask for the entries that need flipping, i.e. wherever the shift - // is odd. - Packet8i shift_ints = _mm256_cvtps_epi32(shift); - Packet8i shift_isodd = _mm256_castps_si256(_mm256_and_ps(_mm256_castsi256_ps(shift_ints), _mm256_castsi256_ps(p8i_one))); - Packet8i sign_flip_mask = pshiftleft(shift_isodd, 31); - - // Create a mask for which interpolant to use, i.e. if z > 1, then the mask - // is set to ones for that entry. - Packet8f ival_mask = _mm256_cmp_ps(z, p8f_one, _CMP_GT_OQ); - - // Evaluate the polynomial for the interval [1,3] in z. - _EIGEN_DECLARE_CONST_Packet8f(coeff_right_0, 9.999999724233232e-01f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_right_2, -3.084242535619928e-01f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_right_4, 1.584991525700324e-02f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_right_6, -3.188805084631342e-04f); - Packet8f z_minus_two = psub(z, p8f_two); - Packet8f z_minus_two2 = pmul(z_minus_two, z_minus_two); - Packet8f right = pmadd(p8f_coeff_right_6, z_minus_two2, p8f_coeff_right_4); - right = pmadd(right, z_minus_two2, p8f_coeff_right_2); - right = pmadd(right, z_minus_two2, p8f_coeff_right_0); - - // Evaluate the polynomial for the interval [-1,1] in z. - _EIGEN_DECLARE_CONST_Packet8f(coeff_left_1, 7.853981525427295e-01f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_left_3, -8.074536727092352e-02f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_left_5, 2.489871967827018e-03f); - _EIGEN_DECLARE_CONST_Packet8f(coeff_left_7, -3.587725841214251e-05f); - Packet8f z2 = pmul(z, z); - Packet8f left = pmadd(p8f_coeff_left_7, z2, p8f_coeff_left_5); - left = pmadd(left, z2, p8f_coeff_left_3); - left = pmadd(left, z2, p8f_coeff_left_1); - left = pmul(left, z); - - // Assemble the results, i.e. select the left and right polynomials. - left = _mm256_andnot_ps(ival_mask, left); - right = _mm256_and_ps(ival_mask, right); - Packet8f res = _mm256_or_ps(left, right); - - // Flip the sign on the odd intervals and return the result. - res = _mm256_xor_ps(res, _mm256_castsi256_ps(sign_flip_mask)); - return res; -} - -// Natural logarithm -// Computes log(x) as log(2^e * m) = C*e + log(m), where the constant C =log(2) -// and m is in the range [sqrt(1/2),sqrt(2)). In this range, the logarithm can -// be easily approximated by a polynomial centered on m=1 for stability. -// TODO(gonnet): Further reduce the interval allowing for lower-degree -// polynomial interpolants -> ... -> profit! -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8f -plog(const Packet8f& _x) { - Packet8f x = _x; - _EIGEN_DECLARE_CONST_Packet8f(1, 1.0f); - _EIGEN_DECLARE_CONST_Packet8f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet8f(126f, 126.0f); - - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(inv_mant_mask, ~0x7f800000); - - // The smallest non denormalized float number. - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(min_norm_pos, 0x00800000); - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(minus_inf, 0xff800000); - - // Polynomial coefficients. - _EIGEN_DECLARE_CONST_Packet8f(cephes_SQRTHF, 0.707106781186547524f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p0, 7.0376836292E-2f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p1, -1.1514610310E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p2, 1.1676998740E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p3, -1.2420140846E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p4, +1.4249322787E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p5, -1.6668057665E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p6, +2.0000714765E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p7, -2.4999993993E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_p8, +3.3333331174E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_q1, -2.12194440e-4f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_log_q2, 0.693359375f); - - Packet8f invalid_mask = _mm256_cmp_ps(x, _mm256_setzero_ps(), _CMP_NGE_UQ); // not greater equal is true if x is NaN - Packet8f iszero_mask = _mm256_cmp_ps(x, _mm256_setzero_ps(), _CMP_EQ_OQ); - - // Truncate input values to the minimum positive normal. - x = pmax(x, p8f_min_norm_pos); - - Packet8f emm0 = pshiftright(x,23); - Packet8f e = _mm256_sub_ps(emm0, p8f_126f); - - // Set the exponents to -1, i.e. x are in the range [0.5,1). - x = _mm256_and_ps(x, p8f_inv_mant_mask); - x = _mm256_or_ps(x, p8f_half); - - // part2: Shift the inputs from the range [0.5,1) to [sqrt(1/2),sqrt(2)) - // and shift by -1. The values are then centered around 0, which improves - // the stability of the polynomial evaluation. - // if( x < SQRTHF ) { - // e -= 1; - // x = x + x - 1.0; - // } else { x = x - 1.0; } - Packet8f mask = _mm256_cmp_ps(x, p8f_cephes_SQRTHF, _CMP_LT_OQ); - Packet8f tmp = _mm256_and_ps(x, mask); - x = psub(x, p8f_1); - e = psub(e, _mm256_and_ps(p8f_1, mask)); - x = padd(x, tmp); - - Packet8f x2 = pmul(x, x); - Packet8f x3 = pmul(x2, x); - - // Evaluate the polynomial approximant of degree 8 in three parts, probably - // to improve instruction-level parallelism. - Packet8f y, y1, y2; - y = pmadd(p8f_cephes_log_p0, x, p8f_cephes_log_p1); - y1 = pmadd(p8f_cephes_log_p3, x, p8f_cephes_log_p4); - y2 = pmadd(p8f_cephes_log_p6, x, p8f_cephes_log_p7); - y = pmadd(y, x, p8f_cephes_log_p2); - y1 = pmadd(y1, x, p8f_cephes_log_p5); - y2 = pmadd(y2, x, p8f_cephes_log_p8); - y = pmadd(y, x3, y1); - y = pmadd(y, x3, y2); - y = pmul(y, x3); - - // Add the logarithm of the exponent back to the result of the interpolation. - y1 = pmul(e, p8f_cephes_log_q1); - tmp = pmul(x2, p8f_half); - y = padd(y, y1); - x = psub(x, tmp); - y2 = pmul(e, p8f_cephes_log_q2); - x = padd(x, y); - x = padd(x, y2); - - // Filter out invalid inputs, i.e. negative arg will be NAN, 0 will be -INF. - return _mm256_or_ps( - _mm256_andnot_ps(iszero_mask, _mm256_or_ps(x, invalid_mask)), - _mm256_and_ps(iszero_mask, p8f_minus_inf)); -} - -// Exponential function. Works by writing "x = m*log(2) + r" where -// "m = floor(x/log(2)+1/2)" and "r" is the remainder. The result is then -// "exp(x) = 2^m*exp(r)" where exp(r) is in the range [-1,1). -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8f -pexp(const Packet8f& _x) { - _EIGEN_DECLARE_CONST_Packet8f(1, 1.0f); - _EIGEN_DECLARE_CONST_Packet8f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet8f(127, 127.0f); - - _EIGEN_DECLARE_CONST_Packet8f(exp_hi, 88.3762626647950f); - _EIGEN_DECLARE_CONST_Packet8f(exp_lo, -88.3762626647949f); - - _EIGEN_DECLARE_CONST_Packet8f(cephes_LOG2EF, 1.44269504088896341f); - - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p0, 1.9875691500E-4f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p1, 1.3981999507E-3f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p2, 8.3334519073E-3f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p3, 4.1665795894E-2f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p4, 1.6666665459E-1f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_p5, 5.0000001201E-1f); - - // Clamp x. - Packet8f x = pmax(pmin(_x, p8f_exp_hi), p8f_exp_lo); - - // Express exp(x) as exp(m*ln(2) + r), start by extracting - // m = floor(x/ln(2) + 0.5). - Packet8f m = _mm256_floor_ps(pmadd(x, p8f_cephes_LOG2EF, p8f_half)); - -// Get r = x - m*ln(2). If no FMA instructions are available, m*ln(2) is -// subtracted out in two parts, m*C1+m*C2 = m*ln(2), to avoid accumulating -// truncation errors. Note that we don't use the "pmadd" function here to -// ensure that a precision-preserving FMA instruction is used. -#ifdef EIGEN_VECTORIZE_FMA - _EIGEN_DECLARE_CONST_Packet8f(nln2, -0.6931471805599453f); - Packet8f r = _mm256_fmadd_ps(m, p8f_nln2, x); -#else - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_C1, 0.693359375f); - _EIGEN_DECLARE_CONST_Packet8f(cephes_exp_C2, -2.12194440e-4f); - Packet8f r = psub(x, pmul(m, p8f_cephes_exp_C1)); - r = psub(r, pmul(m, p8f_cephes_exp_C2)); -#endif - - Packet8f r2 = pmul(r, r); - - // TODO(gonnet): Split into odd/even polynomials and try to exploit - // instruction-level parallelism. - Packet8f y = p8f_cephes_exp_p0; - y = pmadd(y, r, p8f_cephes_exp_p1); - y = pmadd(y, r, p8f_cephes_exp_p2); - y = pmadd(y, r, p8f_cephes_exp_p3); - y = pmadd(y, r, p8f_cephes_exp_p4); - y = pmadd(y, r, p8f_cephes_exp_p5); - y = pmadd(y, r2, r); - y = padd(y, p8f_1); - - // Build emm0 = 2^m. - Packet8i emm0 = _mm256_cvttps_epi32(padd(m, p8f_127)); - emm0 = pshiftleft(emm0, 23); - - // Return 2^m * exp(r). - return pmax(pmul(y, _mm256_castsi256_ps(emm0)), _x); -} - -// Hyperbolic Tangent function. -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8f -ptanh(const Packet8f& x) { - return internal::generic_fast_tanh_float(x); -} - -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet4d -pexp(const Packet4d& _x) { - Packet4d x = _x; - - _EIGEN_DECLARE_CONST_Packet4d(1, 1.0); - _EIGEN_DECLARE_CONST_Packet4d(2, 2.0); - _EIGEN_DECLARE_CONST_Packet4d(half, 0.5); - - _EIGEN_DECLARE_CONST_Packet4d(exp_hi, 709.437); - _EIGEN_DECLARE_CONST_Packet4d(exp_lo, -709.436139303); - - _EIGEN_DECLARE_CONST_Packet4d(cephes_LOG2EF, 1.4426950408889634073599); - - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_p0, 1.26177193074810590878e-4); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_p1, 3.02994407707441961300e-2); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_p2, 9.99999999999999999910e-1); - - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_q0, 3.00198505138664455042e-6); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_q1, 2.52448340349684104192e-3); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_q2, 2.27265548208155028766e-1); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_q3, 2.00000000000000000009e0); - - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_C1, 0.693145751953125); - _EIGEN_DECLARE_CONST_Packet4d(cephes_exp_C2, 1.42860682030941723212e-6); - _EIGEN_DECLARE_CONST_Packet4i(1023, 1023); - - Packet4d tmp, fx; - - // clamp x - x = pmax(pmin(x, p4d_exp_hi), p4d_exp_lo); - // Express exp(x) as exp(g + n*log(2)). - fx = pmadd(p4d_cephes_LOG2EF, x, p4d_half); - - // Get the integer modulus of log(2), i.e. the "n" described above. - fx = _mm256_floor_pd(fx); - - // Get the remainder modulo log(2), i.e. the "g" described above. Subtract - // n*log(2) out in two steps, i.e. n*C1 + n*C2, C1+C2=log2 to get the last - // digits right. - tmp = pmul(fx, p4d_cephes_exp_C1); - Packet4d z = pmul(fx, p4d_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - Packet4d x2 = pmul(x, x); - - // Evaluate the numerator polynomial of the rational interpolant. - Packet4d px = p4d_cephes_exp_p0; - px = pmadd(px, x2, p4d_cephes_exp_p1); - px = pmadd(px, x2, p4d_cephes_exp_p2); - px = pmul(px, x); - - // Evaluate the denominator polynomial of the rational interpolant. - Packet4d qx = p4d_cephes_exp_q0; - qx = pmadd(qx, x2, p4d_cephes_exp_q1); - qx = pmadd(qx, x2, p4d_cephes_exp_q2); - qx = pmadd(qx, x2, p4d_cephes_exp_q3); - - // I don't really get this bit, copied from the SSE2 routines, so... - // TODO(gonnet): Figure out what is going on here, perhaps find a better - // rational interpolant? - x = _mm256_div_pd(px, psub(qx, px)); - x = pmadd(p4d_2, x, p4d_1); - - // Build e=2^n by constructing the exponents in a 128-bit vector and - // shifting them to where they belong in double-precision values. - __m128i emm0 = _mm256_cvtpd_epi32(fx); - emm0 = _mm_add_epi32(emm0, p4i_1023); - emm0 = _mm_shuffle_epi32(emm0, _MM_SHUFFLE(3, 1, 2, 0)); - __m128i lo = _mm_slli_epi64(emm0, 52); - __m128i hi = _mm_slli_epi64(_mm_srli_epi64(emm0, 32), 52); - __m256i e = _mm256_insertf128_si256(_mm256_setzero_si256(), lo, 0); - e = _mm256_insertf128_si256(e, hi, 1); - - // Construct the result 2^n * exp(g) = e * x. The max is used to catch - // non-finite values in the input. - return pmax(pmul(x, _mm256_castsi256_pd(e)), _x); -} - -// Functions for sqrt. -// The EIGEN_FAST_MATH version uses the _mm_rsqrt_ps approximation and one step -// of Newton's method, at a cost of 1-2 bits of precision as opposed to the -// exact solution. It does not handle +inf, or denormalized numbers correctly. -// The main advantage of this approach is not just speed, but also the fact that -// it can be inlined and pipelined with other computations, further reducing its -// effective latency. This is similar to Quake3's fast inverse square root. -// For detail see here: http://www.beyond3d.com/content/articles/8/ -#if EIGEN_FAST_MATH -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8f -psqrt(const Packet8f& _x) { - Packet8f half = pmul(_x, pset1(.5f)); - Packet8f denormal_mask = _mm256_and_ps( - _mm256_cmp_ps(_x, pset1((std::numeric_limits::min)()), - _CMP_LT_OQ), - _mm256_cmp_ps(_x, _mm256_setzero_ps(), _CMP_GE_OQ)); - - // Compute approximate reciprocal sqrt. - Packet8f x = _mm256_rsqrt_ps(_x); - // Do a single step of Newton's iteration. - x = pmul(x, psub(pset1(1.5f), pmul(half, pmul(x,x)))); - // Flush results for denormals to zero. - return _mm256_andnot_ps(denormal_mask, pmul(_x,x)); -} -#else -template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet8f psqrt(const Packet8f& x) { - return _mm256_sqrt_ps(x); -} -#endif -template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4d psqrt(const Packet4d& x) { - return _mm256_sqrt_pd(x); -} -#if EIGEN_FAST_MATH - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet8f prsqrt(const Packet8f& _x) { - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(inf, 0x7f800000); - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(nan, 0x7fc00000); - _EIGEN_DECLARE_CONST_Packet8f(one_point_five, 1.5f); - _EIGEN_DECLARE_CONST_Packet8f(minus_half, -0.5f); - _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(flt_min, 0x00800000); - - Packet8f neg_half = pmul(_x, p8f_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - Packet8f le_zero_mask = _mm256_cmp_ps(_x, p8f_flt_min, _CMP_LT_OQ); - Packet8f x = _mm256_andnot_ps(le_zero_mask, _mm256_rsqrt_ps(_x)); - - // Fill in NaNs and Infs for the negative/zero entries. - Packet8f neg_mask = _mm256_cmp_ps(_x, _mm256_setzero_ps(), _CMP_LT_OQ); - Packet8f zero_mask = _mm256_andnot_ps(neg_mask, le_zero_mask); - Packet8f infs_and_nans = _mm256_or_ps(_mm256_and_ps(neg_mask, p8f_nan), - _mm256_and_ps(zero_mask, p8f_inf)); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p8f_one_point_five)); - - // Insert NaNs and Infs in all the right places. - return _mm256_or_ps(x, infs_and_nans); -} - -#else -template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet8f prsqrt(const Packet8f& x) { - _EIGEN_DECLARE_CONST_Packet8f(one, 1.0f); - return _mm256_div_ps(p8f_one, _mm256_sqrt_ps(x)); -} -#endif - -template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4d prsqrt(const Packet4d& x) { - _EIGEN_DECLARE_CONST_Packet4d(one, 1.0); - return _mm256_div_pd(p4d_one, _mm256_sqrt_pd(x)); -} - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_AVX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/PacketMath.h deleted file mode 100644 index 923a124b206..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/PacketMath.h +++ /dev/null @@ -1,637 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner (benoit.steiner.goog@gmail.com) -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_AVX_H -#define EIGEN_PACKET_MATH_AVX_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS (2*sizeof(void*)) -#endif - -#ifdef __FMA__ -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#endif -#endif - -typedef __m256 Packet8f; -typedef __m256i Packet8i; -typedef __m256d Packet4d; - -template<> struct is_arithmetic<__m256> { enum { value = true }; }; -template<> struct is_arithmetic<__m256i> { enum { value = true }; }; -template<> struct is_arithmetic<__m256d> { enum { value = true }; }; - -#define _EIGEN_DECLARE_CONST_Packet8f(NAME,X) \ - const Packet8f p8f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4d(NAME,X) \ - const Packet4d p4d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet8f_FROM_INT(NAME,X) \ - const Packet8f p8f_##NAME = _mm256_castsi256_ps(pset1(X)) - -#define _EIGEN_DECLARE_CONST_Packet8i(NAME,X) \ - const Packet8i p8i_##NAME = pset1(X) - -// Use the packet_traits defined in AVX512/PacketMath.h instead if we're going -// to leverage AVX512 instructions. -#ifndef EIGEN_VECTORIZE_AVX512 -template<> struct packet_traits : default_packet_traits -{ - typedef Packet8f type; - typedef Packet4f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=8, - HasHalfPacket = 1, - - HasDiv = 1, - HasSin = EIGEN_FAST_MATH, - HasCos = 0, - HasLog = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasTanh = EIGEN_FAST_MATH, - HasBlend = 1, - HasRound = 1, - HasFloor = 1, - HasCeil = 1 - }; -}; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 1, - - HasDiv = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasBlend = 1, - HasRound = 1, - HasFloor = 1, - HasCeil = 1 - }; -}; -#endif - -template<> struct scalar_div_cost { enum { value = 14 }; }; -template<> struct scalar_div_cost { enum { value = 16 }; }; - -/* Proper support for integers is only provided by AVX2. In the meantime, we'll - use SSE instructions and packets to deal with integers. -template<> struct packet_traits : default_packet_traits -{ - typedef Packet8i type; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=8 - }; -}; -*/ - -template<> struct unpacket_traits { typedef float type; typedef Packet4f half; enum {size=8, alignment=Aligned32}; }; -template<> struct unpacket_traits { typedef double type; typedef Packet2d half; enum {size=4, alignment=Aligned32}; }; -template<> struct unpacket_traits { typedef int type; typedef Packet4i half; enum {size=8, alignment=Aligned32}; }; - -template<> EIGEN_STRONG_INLINE Packet8f pset1(const float& from) { return _mm256_set1_ps(from); } -template<> EIGEN_STRONG_INLINE Packet4d pset1(const double& from) { return _mm256_set1_pd(from); } -template<> EIGEN_STRONG_INLINE Packet8i pset1(const int& from) { return _mm256_set1_epi32(from); } - -template<> EIGEN_STRONG_INLINE Packet8f pload1(const float* from) { return _mm256_broadcast_ss(from); } -template<> EIGEN_STRONG_INLINE Packet4d pload1(const double* from) { return _mm256_broadcast_sd(from); } - -template<> EIGEN_STRONG_INLINE Packet8f plset(const float& a) { return _mm256_add_ps(_mm256_set1_ps(a), _mm256_set_ps(7.0,6.0,5.0,4.0,3.0,2.0,1.0,0.0)); } -template<> EIGEN_STRONG_INLINE Packet4d plset(const double& a) { return _mm256_add_pd(_mm256_set1_pd(a), _mm256_set_pd(3.0,2.0,1.0,0.0)); } - -template<> EIGEN_STRONG_INLINE Packet8f padd(const Packet8f& a, const Packet8f& b) { return _mm256_add_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d padd(const Packet4d& a, const Packet4d& b) { return _mm256_add_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f psub(const Packet8f& a, const Packet8f& b) { return _mm256_sub_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d psub(const Packet4d& a, const Packet4d& b) { return _mm256_sub_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pnegate(const Packet8f& a) -{ - return _mm256_sub_ps(_mm256_set1_ps(0.0),a); -} -template<> EIGEN_STRONG_INLINE Packet4d pnegate(const Packet4d& a) -{ - return _mm256_sub_pd(_mm256_set1_pd(0.0),a); -} - -template<> EIGEN_STRONG_INLINE Packet8f pconj(const Packet8f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4d pconj(const Packet4d& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet8i pconj(const Packet8i& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet8f pmul(const Packet8f& a, const Packet8f& b) { return _mm256_mul_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pmul(const Packet4d& a, const Packet4d& b) { return _mm256_mul_pd(a,b); } - - -template<> EIGEN_STRONG_INLINE Packet8f pdiv(const Packet8f& a, const Packet8f& b) { return _mm256_div_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pdiv(const Packet4d& a, const Packet4d& b) { return _mm256_div_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet8i pdiv(const Packet8i& /*a*/, const Packet8i& /*b*/) -{ eigen_assert(false && "packet integer division are not supported by AVX"); - return pset1(0); -} - -#ifdef __FMA__ -template<> EIGEN_STRONG_INLINE Packet8f pmadd(const Packet8f& a, const Packet8f& b, const Packet8f& c) { -#if ( (EIGEN_COMP_GNUC_STRICT && EIGEN_COMP_GNUC<80) || (EIGEN_COMP_CLANG) ) - // Clang stupidly generates a vfmadd213ps instruction plus some vmovaps on registers, - // and even register spilling with clang>=6.0 (bug 1637). - // Gcc stupidly generates a vfmadd132ps instruction. - // So let's enforce it to generate a vfmadd231ps instruction since the most common use - // case is to accumulate the result of the product. - Packet8f res = c; - __asm__("vfmadd231ps %[a], %[b], %[c]" : [c] "+x" (res) : [a] "x" (a), [b] "x" (b)); - return res; -#else - return _mm256_fmadd_ps(a,b,c); -#endif -} -template<> EIGEN_STRONG_INLINE Packet4d pmadd(const Packet4d& a, const Packet4d& b, const Packet4d& c) { -#if ( (EIGEN_COMP_GNUC_STRICT && EIGEN_COMP_GNUC<80) || (EIGEN_COMP_CLANG) ) - // see above - Packet4d res = c; - __asm__("vfmadd231pd %[a], %[b], %[c]" : [c] "+x" (res) : [a] "x" (a), [b] "x" (b)); - return res; -#else - return _mm256_fmadd_pd(a,b,c); -#endif -} -#endif - -template<> EIGEN_STRONG_INLINE Packet8f pmin(const Packet8f& a, const Packet8f& b) { return _mm256_min_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pmin(const Packet4d& a, const Packet4d& b) { return _mm256_min_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pmax(const Packet8f& a, const Packet8f& b) { return _mm256_max_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pmax(const Packet4d& a, const Packet4d& b) { return _mm256_max_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pround(const Packet8f& a) { return _mm256_round_ps(a, _MM_FROUND_CUR_DIRECTION); } -template<> EIGEN_STRONG_INLINE Packet4d pround(const Packet4d& a) { return _mm256_round_pd(a, _MM_FROUND_CUR_DIRECTION); } - -template<> EIGEN_STRONG_INLINE Packet8f pceil(const Packet8f& a) { return _mm256_ceil_ps(a); } -template<> EIGEN_STRONG_INLINE Packet4d pceil(const Packet4d& a) { return _mm256_ceil_pd(a); } - -template<> EIGEN_STRONG_INLINE Packet8f pfloor(const Packet8f& a) { return _mm256_floor_ps(a); } -template<> EIGEN_STRONG_INLINE Packet4d pfloor(const Packet4d& a) { return _mm256_floor_pd(a); } - -template<> EIGEN_STRONG_INLINE Packet8f pand(const Packet8f& a, const Packet8f& b) { return _mm256_and_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pand(const Packet4d& a, const Packet4d& b) { return _mm256_and_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f por(const Packet8f& a, const Packet8f& b) { return _mm256_or_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d por(const Packet4d& a, const Packet4d& b) { return _mm256_or_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pxor(const Packet8f& a, const Packet8f& b) { return _mm256_xor_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pxor(const Packet4d& a, const Packet4d& b) { return _mm256_xor_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pandnot(const Packet8f& a, const Packet8f& b) { return _mm256_andnot_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet4d pandnot(const Packet4d& a, const Packet4d& b) { return _mm256_andnot_pd(a,b); } - -template<> EIGEN_STRONG_INLINE Packet8f pload(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm256_load_ps(from); } -template<> EIGEN_STRONG_INLINE Packet4d pload(const double* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm256_load_pd(from); } -template<> EIGEN_STRONG_INLINE Packet8i pload(const int* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm256_load_si256(reinterpret_cast(from)); } - -template<> EIGEN_STRONG_INLINE Packet8f ploadu(const float* from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm256_loadu_ps(from); } -template<> EIGEN_STRONG_INLINE Packet4d ploadu(const double* from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm256_loadu_pd(from); } -template<> EIGEN_STRONG_INLINE Packet8i ploadu(const int* from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm256_loadu_si256(reinterpret_cast(from)); } - -// Loads 4 floats from memory a returns the packet {a0, a0 a1, a1, a2, a2, a3, a3} -template<> EIGEN_STRONG_INLINE Packet8f ploaddup(const float* from) -{ - // TODO try to find a way to avoid the need of a temporary register -// Packet8f tmp = _mm256_castps128_ps256(_mm_loadu_ps(from)); -// tmp = _mm256_insertf128_ps(tmp, _mm_movehl_ps(_mm256_castps256_ps128(tmp),_mm256_castps256_ps128(tmp)), 1); -// return _mm256_unpacklo_ps(tmp,tmp); - - // _mm256_insertf128_ps is very slow on Haswell, thus: - Packet8f tmp = _mm256_broadcast_ps((const __m128*)(const void*)from); - // mimic an "inplace" permutation of the lower 128bits using a blend - tmp = _mm256_blend_ps(tmp,_mm256_castps128_ps256(_mm_permute_ps( _mm256_castps256_ps128(tmp), _MM_SHUFFLE(1,0,1,0))), 15); - // then we can perform a consistent permutation on the global register to get everything in shape: - return _mm256_permute_ps(tmp, _MM_SHUFFLE(3,3,2,2)); -} -// Loads 2 doubles from memory a returns the packet {a0, a0 a1, a1} -template<> EIGEN_STRONG_INLINE Packet4d ploaddup(const double* from) -{ - Packet4d tmp = _mm256_broadcast_pd((const __m128d*)(const void*)from); - return _mm256_permute_pd(tmp, 3<<2); -} - -// Loads 2 floats from memory a returns the packet {a0, a0 a0, a0, a1, a1, a1, a1} -template<> EIGEN_STRONG_INLINE Packet8f ploadquad(const float* from) -{ - Packet8f tmp = _mm256_castps128_ps256(_mm_broadcast_ss(from)); - return _mm256_insertf128_ps(tmp, _mm_broadcast_ss(from+1), 1); -} - -template<> EIGEN_STRONG_INLINE void pstore(float* to, const Packet8f& from) { EIGEN_DEBUG_ALIGNED_STORE _mm256_store_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet4d& from) { EIGEN_DEBUG_ALIGNED_STORE _mm256_store_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(int* to, const Packet8i& from) { EIGEN_DEBUG_ALIGNED_STORE _mm256_storeu_si256(reinterpret_cast<__m256i*>(to), from); } - -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet8f& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm256_storeu_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet4d& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm256_storeu_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet8i& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm256_storeu_si256(reinterpret_cast<__m256i*>(to), from); } - -// NOTE: leverage _mm256_i32gather_ps and _mm256_i32gather_pd if AVX2 instructions are available -// NOTE: for the record the following seems to be slower: return _mm256_i32gather_ps(from, _mm256_set1_epi32(stride), 4); -template<> EIGEN_DEVICE_FUNC inline Packet8f pgather(const float* from, Index stride) -{ - return _mm256_set_ps(from[7*stride], from[6*stride], from[5*stride], from[4*stride], - from[3*stride], from[2*stride], from[1*stride], from[0*stride]); -} -template<> EIGEN_DEVICE_FUNC inline Packet4d pgather(const double* from, Index stride) -{ - return _mm256_set_pd(from[3*stride], from[2*stride], from[1*stride], from[0*stride]); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet8f& from, Index stride) -{ - __m128 low = _mm256_extractf128_ps(from, 0); - to[stride*0] = _mm_cvtss_f32(low); - to[stride*1] = _mm_cvtss_f32(_mm_shuffle_ps(low, low, 1)); - to[stride*2] = _mm_cvtss_f32(_mm_shuffle_ps(low, low, 2)); - to[stride*3] = _mm_cvtss_f32(_mm_shuffle_ps(low, low, 3)); - - __m128 high = _mm256_extractf128_ps(from, 1); - to[stride*4] = _mm_cvtss_f32(high); - to[stride*5] = _mm_cvtss_f32(_mm_shuffle_ps(high, high, 1)); - to[stride*6] = _mm_cvtss_f32(_mm_shuffle_ps(high, high, 2)); - to[stride*7] = _mm_cvtss_f32(_mm_shuffle_ps(high, high, 3)); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet4d& from, Index stride) -{ - __m128d low = _mm256_extractf128_pd(from, 0); - to[stride*0] = _mm_cvtsd_f64(low); - to[stride*1] = _mm_cvtsd_f64(_mm_shuffle_pd(low, low, 1)); - __m128d high = _mm256_extractf128_pd(from, 1); - to[stride*2] = _mm_cvtsd_f64(high); - to[stride*3] = _mm_cvtsd_f64(_mm_shuffle_pd(high, high, 1)); -} - -template<> EIGEN_STRONG_INLINE void pstore1(float* to, const float& a) -{ - Packet8f pa = pset1(a); - pstore(to, pa); -} -template<> EIGEN_STRONG_INLINE void pstore1(double* to, const double& a) -{ - Packet4d pa = pset1(a); - pstore(to, pa); -} -template<> EIGEN_STRONG_INLINE void pstore1(int* to, const int& a) -{ - Packet8i pa = pset1(a); - pstore(to, pa); -} - -#ifndef EIGEN_VECTORIZE_AVX512 -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -#endif - -template<> EIGEN_STRONG_INLINE float pfirst(const Packet8f& a) { - return _mm_cvtss_f32(_mm256_castps256_ps128(a)); -} -template<> EIGEN_STRONG_INLINE double pfirst(const Packet4d& a) { - return _mm_cvtsd_f64(_mm256_castpd256_pd128(a)); -} -template<> EIGEN_STRONG_INLINE int pfirst(const Packet8i& a) { - return _mm_cvtsi128_si32(_mm256_castsi256_si128(a)); -} - - -template<> EIGEN_STRONG_INLINE Packet8f preverse(const Packet8f& a) -{ - __m256 tmp = _mm256_shuffle_ps(a,a,0x1b); - return _mm256_permute2f128_ps(tmp, tmp, 1); -} -template<> EIGEN_STRONG_INLINE Packet4d preverse(const Packet4d& a) -{ - __m256d tmp = _mm256_shuffle_pd(a,a,5); - return _mm256_permute2f128_pd(tmp, tmp, 1); - #if 0 - // This version is unlikely to be faster as _mm256_shuffle_ps and _mm256_permute_pd - // exhibit the same latency/throughput, but it is here for future reference/benchmarking... - __m256d swap_halves = _mm256_permute2f128_pd(a,a,1); - return _mm256_permute_pd(swap_halves,5); - #endif -} - -// pabs should be ok -template<> EIGEN_STRONG_INLINE Packet8f pabs(const Packet8f& a) -{ - const Packet8f mask = _mm256_castsi256_ps(_mm256_setr_epi32(0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF)); - return _mm256_and_ps(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet4d pabs(const Packet4d& a) -{ - const Packet4d mask = _mm256_castsi256_pd(_mm256_setr_epi32(0xFFFFFFFF,0x7FFFFFFF,0xFFFFFFFF,0x7FFFFFFF,0xFFFFFFFF,0x7FFFFFFF,0xFFFFFFFF,0x7FFFFFFF)); - return _mm256_and_pd(a,mask); -} - -// preduxp should be ok -// FIXME: why is this ok? why isn't the simply implementation working as expected? -template<> EIGEN_STRONG_INLINE Packet8f preduxp(const Packet8f* vecs) -{ - __m256 hsum1 = _mm256_hadd_ps(vecs[0], vecs[1]); - __m256 hsum2 = _mm256_hadd_ps(vecs[2], vecs[3]); - __m256 hsum3 = _mm256_hadd_ps(vecs[4], vecs[5]); - __m256 hsum4 = _mm256_hadd_ps(vecs[6], vecs[7]); - - __m256 hsum5 = _mm256_hadd_ps(hsum1, hsum1); - __m256 hsum6 = _mm256_hadd_ps(hsum2, hsum2); - __m256 hsum7 = _mm256_hadd_ps(hsum3, hsum3); - __m256 hsum8 = _mm256_hadd_ps(hsum4, hsum4); - - __m256 perm1 = _mm256_permute2f128_ps(hsum5, hsum5, 0x23); - __m256 perm2 = _mm256_permute2f128_ps(hsum6, hsum6, 0x23); - __m256 perm3 = _mm256_permute2f128_ps(hsum7, hsum7, 0x23); - __m256 perm4 = _mm256_permute2f128_ps(hsum8, hsum8, 0x23); - - __m256 sum1 = _mm256_add_ps(perm1, hsum5); - __m256 sum2 = _mm256_add_ps(perm2, hsum6); - __m256 sum3 = _mm256_add_ps(perm3, hsum7); - __m256 sum4 = _mm256_add_ps(perm4, hsum8); - - __m256 blend1 = _mm256_blend_ps(sum1, sum2, 0xcc); - __m256 blend2 = _mm256_blend_ps(sum3, sum4, 0xcc); - - __m256 final = _mm256_blend_ps(blend1, blend2, 0xf0); - return final; -} -template<> EIGEN_STRONG_INLINE Packet4d preduxp(const Packet4d* vecs) -{ - Packet4d tmp0, tmp1; - - tmp0 = _mm256_hadd_pd(vecs[0], vecs[1]); - tmp0 = _mm256_add_pd(tmp0, _mm256_permute2f128_pd(tmp0, tmp0, 1)); - - tmp1 = _mm256_hadd_pd(vecs[2], vecs[3]); - tmp1 = _mm256_add_pd(tmp1, _mm256_permute2f128_pd(tmp1, tmp1, 1)); - - return _mm256_blend_pd(tmp0, tmp1, 0xC); -} - -template<> EIGEN_STRONG_INLINE float predux(const Packet8f& a) -{ - return predux(Packet4f(_mm_add_ps(_mm256_castps256_ps128(a),_mm256_extractf128_ps(a,1)))); -} -template<> EIGEN_STRONG_INLINE double predux(const Packet4d& a) -{ - return predux(Packet2d(_mm_add_pd(_mm256_castpd256_pd128(a),_mm256_extractf128_pd(a,1)))); -} - -template<> EIGEN_STRONG_INLINE Packet4f predux_downto4(const Packet8f& a) -{ - return _mm_add_ps(_mm256_castps256_ps128(a),_mm256_extractf128_ps(a,1)); -} - -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet8f& a) -{ - Packet8f tmp; - tmp = _mm256_mul_ps(a, _mm256_permute2f128_ps(a,a,1)); - tmp = _mm256_mul_ps(tmp, _mm256_shuffle_ps(tmp,tmp,_MM_SHUFFLE(1,0,3,2))); - return pfirst(_mm256_mul_ps(tmp, _mm256_shuffle_ps(tmp,tmp,1))); -} -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet4d& a) -{ - Packet4d tmp; - tmp = _mm256_mul_pd(a, _mm256_permute2f128_pd(a,a,1)); - return pfirst(_mm256_mul_pd(tmp, _mm256_shuffle_pd(tmp,tmp,1))); -} - -template<> EIGEN_STRONG_INLINE float predux_min(const Packet8f& a) -{ - Packet8f tmp = _mm256_min_ps(a, _mm256_permute2f128_ps(a,a,1)); - tmp = _mm256_min_ps(tmp, _mm256_shuffle_ps(tmp,tmp,_MM_SHUFFLE(1,0,3,2))); - return pfirst(_mm256_min_ps(tmp, _mm256_shuffle_ps(tmp,tmp,1))); -} -template<> EIGEN_STRONG_INLINE double predux_min(const Packet4d& a) -{ - Packet4d tmp = _mm256_min_pd(a, _mm256_permute2f128_pd(a,a,1)); - return pfirst(_mm256_min_pd(tmp, _mm256_shuffle_pd(tmp, tmp, 1))); -} - -template<> EIGEN_STRONG_INLINE float predux_max(const Packet8f& a) -{ - Packet8f tmp = _mm256_max_ps(a, _mm256_permute2f128_ps(a,a,1)); - tmp = _mm256_max_ps(tmp, _mm256_shuffle_ps(tmp,tmp,_MM_SHUFFLE(1,0,3,2))); - return pfirst(_mm256_max_ps(tmp, _mm256_shuffle_ps(tmp,tmp,1))); -} - -template<> EIGEN_STRONG_INLINE double predux_max(const Packet4d& a) -{ - Packet4d tmp = _mm256_max_pd(a, _mm256_permute2f128_pd(a,a,1)); - return pfirst(_mm256_max_pd(tmp, _mm256_shuffle_pd(tmp, tmp, 1))); -} - - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet8f& first, const Packet8f& second) - { - if (Offset==1) - { - first = _mm256_blend_ps(first, second, 1); - Packet8f tmp1 = _mm256_permute_ps (first, _MM_SHUFFLE(0,3,2,1)); - Packet8f tmp2 = _mm256_permute2f128_ps (tmp1, tmp1, 1); - first = _mm256_blend_ps(tmp1, tmp2, 0x88); - } - else if (Offset==2) - { - first = _mm256_blend_ps(first, second, 3); - Packet8f tmp1 = _mm256_permute_ps (first, _MM_SHUFFLE(1,0,3,2)); - Packet8f tmp2 = _mm256_permute2f128_ps (tmp1, tmp1, 1); - first = _mm256_blend_ps(tmp1, tmp2, 0xcc); - } - else if (Offset==3) - { - first = _mm256_blend_ps(first, second, 7); - Packet8f tmp1 = _mm256_permute_ps (first, _MM_SHUFFLE(2,1,0,3)); - Packet8f tmp2 = _mm256_permute2f128_ps (tmp1, tmp1, 1); - first = _mm256_blend_ps(tmp1, tmp2, 0xee); - } - else if (Offset==4) - { - first = _mm256_blend_ps(first, second, 15); - Packet8f tmp1 = _mm256_permute_ps (first, _MM_SHUFFLE(3,2,1,0)); - Packet8f tmp2 = _mm256_permute2f128_ps (tmp1, tmp1, 1); - first = _mm256_permute_ps(tmp2, _MM_SHUFFLE(3,2,1,0)); - } - else if (Offset==5) - { - first = _mm256_blend_ps(first, second, 31); - first = _mm256_permute2f128_ps(first, first, 1); - Packet8f tmp = _mm256_permute_ps (first, _MM_SHUFFLE(0,3,2,1)); - first = _mm256_permute2f128_ps(tmp, tmp, 1); - first = _mm256_blend_ps(tmp, first, 0x88); - } - else if (Offset==6) - { - first = _mm256_blend_ps(first, second, 63); - first = _mm256_permute2f128_ps(first, first, 1); - Packet8f tmp = _mm256_permute_ps (first, _MM_SHUFFLE(1,0,3,2)); - first = _mm256_permute2f128_ps(tmp, tmp, 1); - first = _mm256_blend_ps(tmp, first, 0xcc); - } - else if (Offset==7) - { - first = _mm256_blend_ps(first, second, 127); - first = _mm256_permute2f128_ps(first, first, 1); - Packet8f tmp = _mm256_permute_ps (first, _MM_SHUFFLE(2,1,0,3)); - first = _mm256_permute2f128_ps(tmp, tmp, 1); - first = _mm256_blend_ps(tmp, first, 0xee); - } - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4d& first, const Packet4d& second) - { - if (Offset==1) - { - first = _mm256_blend_pd(first, second, 1); - __m256d tmp = _mm256_permute_pd(first, 5); - first = _mm256_permute2f128_pd(tmp, tmp, 1); - first = _mm256_blend_pd(tmp, first, 0xA); - } - else if (Offset==2) - { - first = _mm256_blend_pd(first, second, 3); - first = _mm256_permute2f128_pd(first, first, 1); - } - else if (Offset==3) - { - first = _mm256_blend_pd(first, second, 7); - __m256d tmp = _mm256_permute_pd(first, 5); - first = _mm256_permute2f128_pd(tmp, tmp, 1); - first = _mm256_blend_pd(tmp, first, 5); - } - } -}; - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m256 T0 = _mm256_unpacklo_ps(kernel.packet[0], kernel.packet[1]); - __m256 T1 = _mm256_unpackhi_ps(kernel.packet[0], kernel.packet[1]); - __m256 T2 = _mm256_unpacklo_ps(kernel.packet[2], kernel.packet[3]); - __m256 T3 = _mm256_unpackhi_ps(kernel.packet[2], kernel.packet[3]); - __m256 T4 = _mm256_unpacklo_ps(kernel.packet[4], kernel.packet[5]); - __m256 T5 = _mm256_unpackhi_ps(kernel.packet[4], kernel.packet[5]); - __m256 T6 = _mm256_unpacklo_ps(kernel.packet[6], kernel.packet[7]); - __m256 T7 = _mm256_unpackhi_ps(kernel.packet[6], kernel.packet[7]); - __m256 S0 = _mm256_shuffle_ps(T0,T2,_MM_SHUFFLE(1,0,1,0)); - __m256 S1 = _mm256_shuffle_ps(T0,T2,_MM_SHUFFLE(3,2,3,2)); - __m256 S2 = _mm256_shuffle_ps(T1,T3,_MM_SHUFFLE(1,0,1,0)); - __m256 S3 = _mm256_shuffle_ps(T1,T3,_MM_SHUFFLE(3,2,3,2)); - __m256 S4 = _mm256_shuffle_ps(T4,T6,_MM_SHUFFLE(1,0,1,0)); - __m256 S5 = _mm256_shuffle_ps(T4,T6,_MM_SHUFFLE(3,2,3,2)); - __m256 S6 = _mm256_shuffle_ps(T5,T7,_MM_SHUFFLE(1,0,1,0)); - __m256 S7 = _mm256_shuffle_ps(T5,T7,_MM_SHUFFLE(3,2,3,2)); - kernel.packet[0] = _mm256_permute2f128_ps(S0, S4, 0x20); - kernel.packet[1] = _mm256_permute2f128_ps(S1, S5, 0x20); - kernel.packet[2] = _mm256_permute2f128_ps(S2, S6, 0x20); - kernel.packet[3] = _mm256_permute2f128_ps(S3, S7, 0x20); - kernel.packet[4] = _mm256_permute2f128_ps(S0, S4, 0x31); - kernel.packet[5] = _mm256_permute2f128_ps(S1, S5, 0x31); - kernel.packet[6] = _mm256_permute2f128_ps(S2, S6, 0x31); - kernel.packet[7] = _mm256_permute2f128_ps(S3, S7, 0x31); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m256 T0 = _mm256_unpacklo_ps(kernel.packet[0], kernel.packet[1]); - __m256 T1 = _mm256_unpackhi_ps(kernel.packet[0], kernel.packet[1]); - __m256 T2 = _mm256_unpacklo_ps(kernel.packet[2], kernel.packet[3]); - __m256 T3 = _mm256_unpackhi_ps(kernel.packet[2], kernel.packet[3]); - - __m256 S0 = _mm256_shuffle_ps(T0,T2,_MM_SHUFFLE(1,0,1,0)); - __m256 S1 = _mm256_shuffle_ps(T0,T2,_MM_SHUFFLE(3,2,3,2)); - __m256 S2 = _mm256_shuffle_ps(T1,T3,_MM_SHUFFLE(1,0,1,0)); - __m256 S3 = _mm256_shuffle_ps(T1,T3,_MM_SHUFFLE(3,2,3,2)); - - kernel.packet[0] = _mm256_permute2f128_ps(S0, S1, 0x20); - kernel.packet[1] = _mm256_permute2f128_ps(S2, S3, 0x20); - kernel.packet[2] = _mm256_permute2f128_ps(S0, S1, 0x31); - kernel.packet[3] = _mm256_permute2f128_ps(S2, S3, 0x31); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m256d T0 = _mm256_shuffle_pd(kernel.packet[0], kernel.packet[1], 15); - __m256d T1 = _mm256_shuffle_pd(kernel.packet[0], kernel.packet[1], 0); - __m256d T2 = _mm256_shuffle_pd(kernel.packet[2], kernel.packet[3], 15); - __m256d T3 = _mm256_shuffle_pd(kernel.packet[2], kernel.packet[3], 0); - - kernel.packet[1] = _mm256_permute2f128_pd(T0, T2, 32); - kernel.packet[3] = _mm256_permute2f128_pd(T0, T2, 49); - kernel.packet[0] = _mm256_permute2f128_pd(T1, T3, 32); - kernel.packet[2] = _mm256_permute2f128_pd(T1, T3, 49); -} - -template<> EIGEN_STRONG_INLINE Packet8f pblend(const Selector<8>& ifPacket, const Packet8f& thenPacket, const Packet8f& elsePacket) { - const __m256 zero = _mm256_setzero_ps(); - const __m256 select = _mm256_set_ps(ifPacket.select[7], ifPacket.select[6], ifPacket.select[5], ifPacket.select[4], ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m256 false_mask = _mm256_cmp_ps(select, zero, _CMP_EQ_UQ); - return _mm256_blendv_ps(thenPacket, elsePacket, false_mask); -} -template<> EIGEN_STRONG_INLINE Packet4d pblend(const Selector<4>& ifPacket, const Packet4d& thenPacket, const Packet4d& elsePacket) { - const __m256d zero = _mm256_setzero_pd(); - const __m256d select = _mm256_set_pd(ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m256d false_mask = _mm256_cmp_pd(select, zero, _CMP_EQ_UQ); - return _mm256_blendv_pd(thenPacket, elsePacket, false_mask); -} - -template<> EIGEN_STRONG_INLINE Packet8f pinsertfirst(const Packet8f& a, float b) -{ - return _mm256_blend_ps(a,pset1(b),1); -} - -template<> EIGEN_STRONG_INLINE Packet4d pinsertfirst(const Packet4d& a, double b) -{ - return _mm256_blend_pd(a,pset1(b),1); -} - -template<> EIGEN_STRONG_INLINE Packet8f pinsertlast(const Packet8f& a, float b) -{ - return _mm256_blend_ps(a,pset1(b),(1<<7)); -} - -template<> EIGEN_STRONG_INLINE Packet4d pinsertlast(const Packet4d& a, double b) -{ - return _mm256_blend_pd(a,pset1(b),(1<<3)); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PACKET_MATH_AVX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/TypeCasting.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/TypeCasting.h deleted file mode 100644 index 83bfdc604be..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX/TypeCasting.h +++ /dev/null @@ -1,51 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TYPE_CASTING_AVX_H -#define EIGEN_TYPE_CASTING_AVX_H - -namespace Eigen { - -namespace internal { - -// For now we use SSE to handle integers, so we can't use AVX instructions to cast -// from int to float -template <> -struct type_casting_traits { - enum { - VectorizedCast = 0, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 0, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - - - -template<> EIGEN_STRONG_INLINE Packet8i pcast(const Packet8f& a) { - return _mm256_cvtps_epi32(a); -} - -template<> EIGEN_STRONG_INLINE Packet8f pcast(const Packet8i& a) { - return _mm256_cvtepi32_ps(a); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TYPE_CASTING_AVX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/MathFunctions.h deleted file mode 100644 index b259c1e1f92..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/MathFunctions.h +++ /dev/null @@ -1,389 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Pedro Gonnet (pedro.gonnet@gmail.com) -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef THIRD_PARTY_EIGEN3_EIGEN_SRC_CORE_ARCH_AVX512_MATHFUNCTIONS_H_ -#define THIRD_PARTY_EIGEN3_EIGEN_SRC_CORE_ARCH_AVX512_MATHFUNCTIONS_H_ - -namespace Eigen { - -namespace internal { - -// Disable the code for older versions of gcc that don't support many of the required avx512 instrinsics. -#if EIGEN_GNUC_AT_LEAST(5, 3) - -#define _EIGEN_DECLARE_CONST_Packet16f(NAME, X) \ - const Packet16f p16f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(NAME, X) \ - const Packet16f p16f_##NAME = (__m512)pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet8d(NAME, X) \ - const Packet8d p8d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(NAME, X) \ - const Packet8d p8d_##NAME = _mm512_castsi512_pd(_mm512_set1_epi64(X)) - - -// Natural logarithm -// Computes log(x) as log(2^e * m) = C*e + log(m), where the constant C =log(2) -// and m is in the range [sqrt(1/2),sqrt(2)). In this range, the logarithm can -// be easily approximated by a polynomial centered on m=1 for stability. -#if defined(EIGEN_VECTORIZE_AVX512DQ) -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f -plog(const Packet16f& _x) { - Packet16f x = _x; - _EIGEN_DECLARE_CONST_Packet16f(1, 1.0f); - _EIGEN_DECLARE_CONST_Packet16f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet16f(126f, 126.0f); - - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(inv_mant_mask, ~0x7f800000); - - // The smallest non denormalized float number. - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(min_norm_pos, 0x00800000); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(minus_inf, 0xff800000); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(pos_inf, 0x7f800000); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(nan, 0x7fc00000); - - // Polynomial coefficients. - _EIGEN_DECLARE_CONST_Packet16f(cephes_SQRTHF, 0.707106781186547524f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p0, 7.0376836292E-2f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p1, -1.1514610310E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p2, 1.1676998740E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p3, -1.2420140846E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p4, +1.4249322787E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p5, -1.6668057665E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p6, +2.0000714765E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p7, -2.4999993993E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_p8, +3.3333331174E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_q1, -2.12194440e-4f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_log_q2, 0.693359375f); - - // invalid_mask is set to true when x is NaN - __mmask16 invalid_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_NGE_UQ); - __mmask16 iszero_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_EQ_OQ); - - // Truncate input values to the minimum positive normal. - x = pmax(x, p16f_min_norm_pos); - - // Extract the shifted exponents. - Packet16f emm0 = _mm512_cvtepi32_ps(_mm512_srli_epi32((__m512i)x, 23)); - Packet16f e = _mm512_sub_ps(emm0, p16f_126f); - - // Set the exponents to -1, i.e. x are in the range [0.5,1). - x = _mm512_and_ps(x, p16f_inv_mant_mask); - x = _mm512_or_ps(x, p16f_half); - - // part2: Shift the inputs from the range [0.5,1) to [sqrt(1/2),sqrt(2)) - // and shift by -1. The values are then centered around 0, which improves - // the stability of the polynomial evaluation. - // if( x < SQRTHF ) { - // e -= 1; - // x = x + x - 1.0; - // } else { x = x - 1.0; } - __mmask16 mask = _mm512_cmp_ps_mask(x, p16f_cephes_SQRTHF, _CMP_LT_OQ); - Packet16f tmp = _mm512_mask_blend_ps(mask, _mm512_setzero_ps(), x); - x = psub(x, p16f_1); - e = psub(e, _mm512_mask_blend_ps(mask, _mm512_setzero_ps(), p16f_1)); - x = padd(x, tmp); - - Packet16f x2 = pmul(x, x); - Packet16f x3 = pmul(x2, x); - - // Evaluate the polynomial approximant of degree 8 in three parts, probably - // to improve instruction-level parallelism. - Packet16f y, y1, y2; - y = pmadd(p16f_cephes_log_p0, x, p16f_cephes_log_p1); - y1 = pmadd(p16f_cephes_log_p3, x, p16f_cephes_log_p4); - y2 = pmadd(p16f_cephes_log_p6, x, p16f_cephes_log_p7); - y = pmadd(y, x, p16f_cephes_log_p2); - y1 = pmadd(y1, x, p16f_cephes_log_p5); - y2 = pmadd(y2, x, p16f_cephes_log_p8); - y = pmadd(y, x3, y1); - y = pmadd(y, x3, y2); - y = pmul(y, x3); - - // Add the logarithm of the exponent back to the result of the interpolation. - y1 = pmul(e, p16f_cephes_log_q1); - tmp = pmul(x2, p16f_half); - y = padd(y, y1); - x = psub(x, tmp); - y2 = pmul(e, p16f_cephes_log_q2); - x = padd(x, y); - x = padd(x, y2); - - __mmask16 pos_inf_mask = _mm512_cmp_ps_mask(_x,p16f_pos_inf,_CMP_EQ_OQ); - // Filter out invalid inputs, i.e.: - // - negative arg will be NAN, - // - 0 will be -INF. - // - +INF will be +INF - return _mm512_mask_blend_ps(iszero_mask, - _mm512_mask_blend_ps(invalid_mask, - _mm512_mask_blend_ps(pos_inf_mask,x,p16f_pos_inf), - p16f_nan), - p16f_minus_inf); -} - -#endif - -// Exponential function. Works by writing "x = m*log(2) + r" where -// "m = floor(x/log(2)+1/2)" and "r" is the remainder. The result is then -// "exp(x) = 2^m*exp(r)" where exp(r) is in the range [-1,1). -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f -pexp(const Packet16f& _x) { - _EIGEN_DECLARE_CONST_Packet16f(1, 1.0f); - _EIGEN_DECLARE_CONST_Packet16f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet16f(127, 127.0f); - - _EIGEN_DECLARE_CONST_Packet16f(exp_hi, 88.3762626647950f); - _EIGEN_DECLARE_CONST_Packet16f(exp_lo, -88.3762626647949f); - - _EIGEN_DECLARE_CONST_Packet16f(cephes_LOG2EF, 1.44269504088896341f); - - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p0, 1.9875691500E-4f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p1, 1.3981999507E-3f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p2, 8.3334519073E-3f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p3, 4.1665795894E-2f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p4, 1.6666665459E-1f); - _EIGEN_DECLARE_CONST_Packet16f(cephes_exp_p5, 5.0000001201E-1f); - - // Clamp x. - Packet16f x = pmax(pmin(_x, p16f_exp_hi), p16f_exp_lo); - - // Express exp(x) as exp(m*ln(2) + r), start by extracting - // m = floor(x/ln(2) + 0.5). - Packet16f m = _mm512_floor_ps(pmadd(x, p16f_cephes_LOG2EF, p16f_half)); - - // Get r = x - m*ln(2). Note that we can do this without losing more than one - // ulp precision due to the FMA instruction. - _EIGEN_DECLARE_CONST_Packet16f(nln2, -0.6931471805599453f); - Packet16f r = _mm512_fmadd_ps(m, p16f_nln2, x); - Packet16f r2 = pmul(r, r); - - // TODO(gonnet): Split into odd/even polynomials and try to exploit - // instruction-level parallelism. - Packet16f y = p16f_cephes_exp_p0; - y = pmadd(y, r, p16f_cephes_exp_p1); - y = pmadd(y, r, p16f_cephes_exp_p2); - y = pmadd(y, r, p16f_cephes_exp_p3); - y = pmadd(y, r, p16f_cephes_exp_p4); - y = pmadd(y, r, p16f_cephes_exp_p5); - y = pmadd(y, r2, r); - y = padd(y, p16f_1); - - // Build emm0 = 2^m. - Packet16i emm0 = _mm512_cvttps_epi32(padd(m, p16f_127)); - emm0 = _mm512_slli_epi32(emm0, 23); - - // Return 2^m * exp(r). - return pmax(pmul(y, _mm512_castsi512_ps(emm0)), _x); -} - -/*template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8d -pexp(const Packet8d& _x) { - Packet8d x = _x; - - _EIGEN_DECLARE_CONST_Packet8d(1, 1.0); - _EIGEN_DECLARE_CONST_Packet8d(2, 2.0); - - _EIGEN_DECLARE_CONST_Packet8d(exp_hi, 709.437); - _EIGEN_DECLARE_CONST_Packet8d(exp_lo, -709.436139303); - - _EIGEN_DECLARE_CONST_Packet8d(cephes_LOG2EF, 1.4426950408889634073599); - - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_p0, 1.26177193074810590878e-4); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_p1, 3.02994407707441961300e-2); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_p2, 9.99999999999999999910e-1); - - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_q0, 3.00198505138664455042e-6); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_q1, 2.52448340349684104192e-3); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_q2, 2.27265548208155028766e-1); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_q3, 2.00000000000000000009e0); - - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_C1, 0.693145751953125); - _EIGEN_DECLARE_CONST_Packet8d(cephes_exp_C2, 1.42860682030941723212e-6); - - // clamp x - x = pmax(pmin(x, p8d_exp_hi), p8d_exp_lo); - - // Express exp(x) as exp(g + n*log(2)). - const Packet8d n = - _mm512_mul_round_pd(p8d_cephes_LOG2EF, x, _MM_FROUND_TO_NEAREST_INT); - - // Get the remainder modulo log(2), i.e. the "g" described above. Subtract - // n*log(2) out in two steps, i.e. n*C1 + n*C2, C1+C2=log2 to get the last - // digits right. - const Packet8d nC1 = pmul(n, p8d_cephes_exp_C1); - const Packet8d nC2 = pmul(n, p8d_cephes_exp_C2); - x = psub(x, nC1); - x = psub(x, nC2); - - const Packet8d x2 = pmul(x, x); - - // Evaluate the numerator polynomial of the rational interpolant. - Packet8d px = p8d_cephes_exp_p0; - px = pmadd(px, x2, p8d_cephes_exp_p1); - px = pmadd(px, x2, p8d_cephes_exp_p2); - px = pmul(px, x); - - // Evaluate the denominator polynomial of the rational interpolant. - Packet8d qx = p8d_cephes_exp_q0; - qx = pmadd(qx, x2, p8d_cephes_exp_q1); - qx = pmadd(qx, x2, p8d_cephes_exp_q2); - qx = pmadd(qx, x2, p8d_cephes_exp_q3); - - // I don't really get this bit, copied from the SSE2 routines, so... - // TODO(gonnet): Figure out what is going on here, perhaps find a better - // rational interpolant? - x = _mm512_div_pd(px, psub(qx, px)); - x = pmadd(p8d_2, x, p8d_1); - - // Build e=2^n. - const Packet8d e = _mm512_castsi512_pd(_mm512_slli_epi64( - _mm512_add_epi64(_mm512_cvtpd_epi64(n), _mm512_set1_epi64(1023)), 52)); - - // Construct the result 2^n * exp(g) = e * x. The max is used to catch - // non-finite values in the input. - return pmax(pmul(x, e), _x); - }*/ - -// Functions for sqrt. -// The EIGEN_FAST_MATH version uses the _mm_rsqrt_ps approximation and one step -// of Newton's method, at a cost of 1-2 bits of precision as opposed to the -// exact solution. The main advantage of this approach is not just speed, but -// also the fact that it can be inlined and pipelined with other computations, -// further reducing its effective latency. -#if EIGEN_FAST_MATH -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f -psqrt(const Packet16f& _x) { - Packet16f neg_half = pmul(_x, pset1(-.5f)); - __mmask16 denormal_mask = _mm512_kand( - _mm512_cmp_ps_mask(_x, pset1((std::numeric_limits::min)()), - _CMP_LT_OQ), - _mm512_cmp_ps_mask(_x, _mm512_setzero_ps(), _CMP_GE_OQ)); - - Packet16f x = _mm512_rsqrt14_ps(_x); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5f))); - - // Flush results for denormals to zero. - return _mm512_mask_blend_ps(denormal_mask, pmul(_x,x), _mm512_setzero_ps()); -} - -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8d -psqrt(const Packet8d& _x) { - Packet8d neg_half = pmul(_x, pset1(-.5)); - __mmask16 denormal_mask = _mm512_kand( - _mm512_cmp_pd_mask(_x, pset1((std::numeric_limits::min)()), - _CMP_LT_OQ), - _mm512_cmp_pd_mask(_x, _mm512_setzero_pd(), _CMP_GE_OQ)); - - Packet8d x = _mm512_rsqrt14_pd(_x); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); - - // Do a second step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); - - return _mm512_mask_blend_pd(denormal_mask, pmul(_x,x), _mm512_setzero_pd()); -} -#else -template <> -EIGEN_STRONG_INLINE Packet16f psqrt(const Packet16f& x) { - return _mm512_sqrt_ps(x); -} -template <> -EIGEN_STRONG_INLINE Packet8d psqrt(const Packet8d& x) { - return _mm512_sqrt_pd(x); -} -#endif - -// Functions for rsqrt. -// Almost identical to the sqrt routine, just leave out the last multiplication -// and fill in NaN/Inf where needed. Note that this function only exists as an -// iterative version for doubles since there is no instruction for diretly -// computing the reciprocal square root in AVX-512. -#ifdef EIGEN_FAST_MATH -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f -prsqrt(const Packet16f& _x) { - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(inf, 0x7f800000); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(nan, 0x7fc00000); - _EIGEN_DECLARE_CONST_Packet16f(one_point_five, 1.5f); - _EIGEN_DECLARE_CONST_Packet16f(minus_half, -0.5f); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(flt_min, 0x00800000); - - Packet16f neg_half = pmul(_x, p16f_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - __mmask16 le_zero_mask = _mm512_cmp_ps_mask(_x, p16f_flt_min, _CMP_LT_OQ); - Packet16f x = _mm512_mask_blend_ps(le_zero_mask, _mm512_rsqrt14_ps(_x), _mm512_setzero_ps()); - - // Fill in NaNs and Infs for the negative/zero entries. - __mmask16 neg_mask = _mm512_cmp_ps_mask(_x, _mm512_setzero_ps(), _CMP_LT_OQ); - Packet16f infs_and_nans = _mm512_mask_blend_ps( - neg_mask, _mm512_mask_blend_ps(le_zero_mask, _mm512_setzero_ps(), p16f_inf), p16f_nan); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p16f_one_point_five)); - - // Insert NaNs and Infs in all the right places. - return _mm512_mask_blend_ps(le_zero_mask, x, infs_and_nans); -} - -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8d -prsqrt(const Packet8d& _x) { - _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(inf, 0x7ff0000000000000LL); - _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(nan, 0x7ff1000000000000LL); - _EIGEN_DECLARE_CONST_Packet8d(one_point_five, 1.5); - _EIGEN_DECLARE_CONST_Packet8d(minus_half, -0.5); - _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(dbl_min, 0x0010000000000000LL); - - Packet8d neg_half = pmul(_x, p8d_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - __mmask8 le_zero_mask = _mm512_cmp_pd_mask(_x, p8d_dbl_min, _CMP_LT_OQ); - Packet8d x = _mm512_mask_blend_pd(le_zero_mask, _mm512_rsqrt14_pd(_x), _mm512_setzero_pd()); - - // Fill in NaNs and Infs for the negative/zero entries. - __mmask8 neg_mask = _mm512_cmp_pd_mask(_x, _mm512_setzero_pd(), _CMP_LT_OQ); - Packet8d infs_and_nans = _mm512_mask_blend_pd( - neg_mask, _mm512_mask_blend_pd(le_zero_mask, _mm512_setzero_pd(), p8d_inf), p8d_nan); - - // Do a first step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); - - // Do a second step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); - - // Insert NaNs and Infs in all the right places. - return _mm512_mask_blend_pd(le_zero_mask, x, infs_and_nans); -} -#elif defined(EIGEN_VECTORIZE_AVX512ER) -template <> -EIGEN_STRONG_INLINE Packet16f prsqrt(const Packet16f& x) { - return _mm512_rsqrt28_ps(x); -} -#endif -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // THIRD_PARTY_EIGEN3_EIGEN_SRC_CORE_ARCH_AVX512_MATHFUNCTIONS_H_ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/PacketMath.h deleted file mode 100644 index 000b7762ff2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AVX512/PacketMath.h +++ /dev/null @@ -1,1305 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Benoit Steiner (benoit.steiner.goog@gmail.com) -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_AVX512_H -#define EIGEN_PACKET_MATH_AVX512_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32 -#endif - -#ifdef EIGEN_VECTORIZE_FMA -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#endif -#endif - -typedef __m512 Packet16f; -typedef __m512i Packet16i; -typedef __m512d Packet8d; - -template <> -struct is_arithmetic<__m512> { - enum { value = true }; -}; -template <> -struct is_arithmetic<__m512i> { - enum { value = true }; -}; -template <> -struct is_arithmetic<__m512d> { - enum { value = true }; -}; - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet16f type; - typedef Packet8f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 16, - HasHalfPacket = 1, - HasBlend = 0, -#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) -#ifdef EIGEN_VECTORIZE_AVX512DQ - HasLog = 1, -#endif - HasExp = 1, - HasSqrt = EIGEN_FAST_MATH, - HasRsqrt = EIGEN_FAST_MATH, -#endif - HasDiv = 1 - }; - }; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet8d type; - typedef Packet4d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 8, - HasHalfPacket = 1, -#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) - HasSqrt = EIGEN_FAST_MATH, - HasRsqrt = EIGEN_FAST_MATH, -#endif - HasDiv = 1 - }; -}; - -/* TODO Implement AVX512 for integers -template<> struct packet_traits : default_packet_traits -{ - typedef Packet16i type; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=8 - }; -}; -*/ - -template <> -struct unpacket_traits { - typedef float type; - typedef Packet8f half; - typedef Packet16i integer_packet; - enum { size = 16, alignment=Aligned64 }; -}; -template <> -struct unpacket_traits { - typedef double type; - typedef Packet4d half; - enum { size = 8, alignment=Aligned64 }; -}; -template <> -struct unpacket_traits { - typedef int type; - typedef Packet8i half; - enum { size = 16, alignment=Aligned64 }; -}; - -template <> -EIGEN_STRONG_INLINE Packet16f pset1(const float& from) { - return _mm512_set1_ps(from); -} -template <> -EIGEN_STRONG_INLINE Packet8d pset1(const double& from) { - return _mm512_set1_pd(from); -} -template <> -EIGEN_STRONG_INLINE Packet16i pset1(const int& from) { - return _mm512_set1_epi32(from); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pload1(const float* from) { - return _mm512_broadcastss_ps(_mm_load_ps1(from)); -} -template <> -EIGEN_STRONG_INLINE Packet8d pload1(const double* from) { - return _mm512_set1_pd(*from); -} - -template <> -EIGEN_STRONG_INLINE Packet16f plset(const float& a) { - return _mm512_add_ps( - _mm512_set1_ps(a), - _mm512_set_ps(15.0f, 14.0f, 13.0f, 12.0f, 11.0f, 10.0f, 9.0f, 8.0f, 7.0f, 6.0f, 5.0f, - 4.0f, 3.0f, 2.0f, 1.0f, 0.0f)); -} -template <> -EIGEN_STRONG_INLINE Packet8d plset(const double& a) { - return _mm512_add_pd(_mm512_set1_pd(a), - _mm512_set_pd(7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, 0.0)); -} - -template <> -EIGEN_STRONG_INLINE Packet16f padd(const Packet16f& a, - const Packet16f& b) { - return _mm512_add_ps(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet8d padd(const Packet8d& a, - const Packet8d& b) { - return _mm512_add_pd(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet16i padd(const Packet16i& a, - const Packet16i& b) { - return _mm512_add_epi32(a, b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f psub(const Packet16f& a, - const Packet16f& b) { - return _mm512_sub_ps(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet8d psub(const Packet8d& a, - const Packet8d& b) { - return _mm512_sub_pd(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet16i psub(const Packet16i& a, - const Packet16i& b) { - return _mm512_sub_epi32(a, b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pnegate(const Packet16f& a) { - return _mm512_sub_ps(_mm512_set1_ps(0.0), a); -} -template <> -EIGEN_STRONG_INLINE Packet8d pnegate(const Packet8d& a) { - return _mm512_sub_pd(_mm512_set1_pd(0.0), a); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pconj(const Packet16f& a) { - return a; -} -template <> -EIGEN_STRONG_INLINE Packet8d pconj(const Packet8d& a) { - return a; -} -template <> -EIGEN_STRONG_INLINE Packet16i pconj(const Packet16i& a) { - return a; -} - -template <> -EIGEN_STRONG_INLINE Packet16f pmul(const Packet16f& a, - const Packet16f& b) { - return _mm512_mul_ps(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet8d pmul(const Packet8d& a, - const Packet8d& b) { - return _mm512_mul_pd(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet16i pmul(const Packet16i& a, - const Packet16i& b) { - return _mm512_mul_epi32(a, b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pdiv(const Packet16f& a, - const Packet16f& b) { - return _mm512_div_ps(a, b); -} -template <> -EIGEN_STRONG_INLINE Packet8d pdiv(const Packet8d& a, - const Packet8d& b) { - return _mm512_div_pd(a, b); -} - -#ifdef EIGEN_VECTORIZE_FMA -template <> -EIGEN_STRONG_INLINE Packet16f pmadd(const Packet16f& a, const Packet16f& b, - const Packet16f& c) { - return _mm512_fmadd_ps(a, b, c); -} -template <> -EIGEN_STRONG_INLINE Packet8d pmadd(const Packet8d& a, const Packet8d& b, - const Packet8d& c) { - return _mm512_fmadd_pd(a, b, c); -} -#endif - -template <> -EIGEN_STRONG_INLINE Packet16f pmin(const Packet16f& a, - const Packet16f& b) { - // Arguments are reversed to match NaN propagation behavior of std::min. - return _mm512_min_ps(b, a); -} -template <> -EIGEN_STRONG_INLINE Packet8d pmin(const Packet8d& a, - const Packet8d& b) { - // Arguments are reversed to match NaN propagation behavior of std::min. - return _mm512_min_pd(b, a); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pmax(const Packet16f& a, - const Packet16f& b) { - // Arguments are reversed to match NaN propagation behavior of std::max. - return _mm512_max_ps(b, a); -} -template <> -EIGEN_STRONG_INLINE Packet8d pmax(const Packet8d& a, - const Packet8d& b) { - // Arguments are reversed to match NaN propagation behavior of std::max. - return _mm512_max_pd(b, a); -} - -#ifdef EIGEN_VECTORIZE_AVX512DQ -template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { return _mm512_extractf32x8_ps(x,I_); } -template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { return _mm512_extractf64x2_pd(x,I_); } -EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { return _mm512_insertf32x8(_mm512_castps256_ps512(a),b,1); } -#else -// AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 -template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { - return _mm256_castsi256_ps(_mm512_extracti64x4_epi64( _mm512_castps_si512(x),I_)); -} - -// AVX512F does not define _mm512_extractf64x2_pd to extract _m128 from _m512 -template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { - return _mm_castsi128_pd(_mm512_extracti32x4_epi32( _mm512_castpd_si512(x),I_)); -} - -EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { - return _mm512_castsi512_ps(_mm512_inserti64x4(_mm512_castsi256_si512(_mm256_castps_si256(a)), - _mm256_castps_si256(b),1)); -} -#endif - -// Helper function for bit packing snippet of low precision comparison. -// It packs the flags from 32x16 to 16x16. -EIGEN_STRONG_INLINE __m256i Pack32To16(Packet16f rf) { - // Split data into small pieces and handle with AVX instructions - // to guarantee internal order of vector. - // Operation: - // dst[15:0] := Saturate16(rf[31:0]) - // dst[31:16] := Saturate16(rf[63:32]) - // ... - // dst[255:240] := Saturate16(rf[255:224]) - __m256i lo = _mm256_castps_si256(extract256<0>(rf)); - __m256i hi = _mm256_castps_si256(extract256<1>(rf)); - __m128i result_lo = _mm_packs_epi32(_mm256_extractf128_si256(lo, 0), - _mm256_extractf128_si256(lo, 1)); - __m128i result_hi = _mm_packs_epi32(_mm256_extractf128_si256(hi, 0), - _mm256_extractf128_si256(hi, 1)); - return _mm256_insertf128_si256(_mm256_castsi128_si256(result_lo), result_hi, 1); -} - -template <> -EIGEN_STRONG_INLINE Packet16i pand(const Packet16i& a, - const Packet16i& b) { - return _mm512_and_si512(a,b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pand(const Packet16f& a, - const Packet16f& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_and_ps(a, b); -#else - return _mm512_castsi512_ps(pand(_mm512_castps_si512(a),_mm512_castps_si512(b))); -#endif -} -template <> -EIGEN_STRONG_INLINE Packet8d pand(const Packet8d& a, - const Packet8d& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_and_pd(a, b); -#else - Packet8d res = _mm512_undefined_pd(); - Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); - Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); - res = _mm512_insertf64x4(res, _mm256_and_pd(lane0_a, lane0_b), 0); - - Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); - Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - return _mm512_insertf64x4(res, _mm256_and_pd(lane1_a, lane1_b), 1); -#endif -} - -template <> -EIGEN_STRONG_INLINE Packet16i por(const Packet16i& a, const Packet16i& b) { - return _mm512_or_si512(a, b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f por(const Packet16f& a, const Packet16f& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_or_ps(a, b); -#else - return _mm512_castsi512_ps(por(_mm512_castps_si512(a),_mm512_castps_si512(b))); -#endif -} - -template <> -EIGEN_STRONG_INLINE Packet8d por(const Packet8d& a, - const Packet8d& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_or_pd(a, b); -#else - return _mm512_castsi512_pd(por(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} - -template <> -EIGEN_STRONG_INLINE Packet16i pxor(const Packet16i& a, const Packet16i& b) { - return _mm512_xor_si512(a, b); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pxor(const Packet16f& a, const Packet16f& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_xor_ps(a, b); -#else - return _mm512_castsi512_ps(pxor(_mm512_castps_si512(a),_mm512_castps_si512(b))); -#endif -} - -template <> -EIGEN_STRONG_INLINE Packet8d pxor(const Packet8d& a, const Packet8d& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_xor_pd(a, b); -#else - return _mm512_castsi512_pd(pxor(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} - -template <> -EIGEN_STRONG_INLINE Packet16i pandnot(const Packet16i& a, const Packet16i& b) { - return _mm512_andnot_si512(b, a); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pandnot(const Packet16f& a, const Packet16f& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_ps(b, a); -#else - return _mm512_castsi512_ps(pandnot(_mm512_castps_si512(a),_mm512_castps_si512(b))); -#endif -} -template <> -EIGEN_STRONG_INLINE Packet8d pandnot(const Packet8d& a,const Packet8d& b) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_pd(b, a); -#else - return _mm512_castsi512_pd(pandnot(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} - -template EIGEN_STRONG_INLINE Packet16i parithmetic_shift_right(Packet16i a) { - return _mm512_srai_epi32(a, N); -} - -template EIGEN_STRONG_INLINE Packet16i plogical_shift_right(Packet16i a) { - return _mm512_srli_epi32(a, N); -} - -template EIGEN_STRONG_INLINE Packet16i plogical_shift_left(Packet16i a) { - return _mm512_slli_epi32(a, N); -} - -template <> -EIGEN_STRONG_INLINE Packet16f pload(const float* from) { - EIGEN_DEBUG_ALIGNED_LOAD return _mm512_load_ps(from); -} -template <> -EIGEN_STRONG_INLINE Packet8d pload(const double* from) { - EIGEN_DEBUG_ALIGNED_LOAD return _mm512_load_pd(from); -} -template <> -EIGEN_STRONG_INLINE Packet16i pload(const int* from) { - EIGEN_DEBUG_ALIGNED_LOAD return _mm512_load_si512( - reinterpret_cast(from)); -} - -template <> -EIGEN_STRONG_INLINE Packet16f ploadu(const float* from) { - EIGEN_DEBUG_UNALIGNED_LOAD return _mm512_loadu_ps(from); -} -template <> -EIGEN_STRONG_INLINE Packet8d ploadu(const double* from) { - EIGEN_DEBUG_UNALIGNED_LOAD return _mm512_loadu_pd(from); -} -template <> -EIGEN_STRONG_INLINE Packet16i ploadu(const int* from) { - EIGEN_DEBUG_UNALIGNED_LOAD return _mm512_loadu_si512( - reinterpret_cast(from)); -} - -// Loads 8 floats from memory a returns the packet -// {a0, a0 a1, a1, a2, a2, a3, a3, a4, a4, a5, a5, a6, a6, a7, a7} -template <> -EIGEN_STRONG_INLINE Packet16f ploaddup(const float* from) { - // an unaligned load is required here as there is no requirement - // on the alignment of input pointer 'from' - __m256i low_half = _mm256_loadu_si256(reinterpret_cast(from)); - __m512 even_elements = _mm512_castsi512_ps(_mm512_cvtepu32_epi64(low_half)); - __m512 pairs = _mm512_permute_ps(even_elements, _MM_SHUFFLE(2, 2, 0, 0)); - return pairs; -} - -#ifdef EIGEN_VECTORIZE_AVX512DQ -// FIXME: this does not look optimal, better load a Packet4d and shuffle... -// Loads 4 doubles from memory a returns the packet {a0, a0 a1, a1, a2, a2, a3, -// a3} -template <> -EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { - __m512d x = _mm512_setzero_pd(); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[0]), 0); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[1]), 1); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[2]), 2); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[3]), 3); - return x; -} -#else -template <> -EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { - __m512d x = _mm512_setzero_pd(); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<0, _mm_load_sd(from+0)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<2, _mm_load_sd(from+1)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<4, _mm_load_sd(from+2)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<6, _mm_load_sd(from+3)); - return x; -} -#endif - -// Loads 4 floats from memory a returns the packet -// {a0, a0 a0, a0, a1, a1, a1, a1, a2, a2, a2, a2, a3, a3, a3, a3} -template <> -EIGEN_STRONG_INLINE Packet16f ploadquad(const float* from) { - Packet16f tmp = _mm512_castps128_ps512(ploadu(from)); - const Packet16i scatter_mask = _mm512_set_epi32(3,3,3,3, 2,2,2,2, 1,1,1,1, 0,0,0,0); - return _mm512_permutexvar_ps(scatter_mask, tmp); -} - -// Loads 2 doubles from memory a returns the packet -// {a0, a0 a0, a0, a1, a1, a1, a1} -template <> -EIGEN_STRONG_INLINE Packet8d ploadquad(const double* from) { - __m256d lane0 = _mm256_set1_pd(*from); - __m256d lane1 = _mm256_set1_pd(*(from+1)); - __m512d tmp = _mm512_undefined_pd(); - tmp = _mm512_insertf64x4(tmp, lane0, 0); - return _mm512_insertf64x4(tmp, lane1, 1); -} - -template <> -EIGEN_STRONG_INLINE void pstore(float* to, const Packet16f& from) { - EIGEN_DEBUG_ALIGNED_STORE _mm512_store_ps(to, from); -} -template <> -EIGEN_STRONG_INLINE void pstore(double* to, const Packet8d& from) { - EIGEN_DEBUG_ALIGNED_STORE _mm512_store_pd(to, from); -} -template <> -EIGEN_STRONG_INLINE void pstore(int* to, const Packet16i& from) { - EIGEN_DEBUG_ALIGNED_STORE _mm512_storeu_si512(reinterpret_cast<__m512i*>(to), - from); -} - -template <> -EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet16f& from) { - EIGEN_DEBUG_UNALIGNED_STORE _mm512_storeu_ps(to, from); -} -template <> -EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet8d& from) { - EIGEN_DEBUG_UNALIGNED_STORE _mm512_storeu_pd(to, from); -} -template <> -EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet16i& from) { - EIGEN_DEBUG_UNALIGNED_STORE _mm512_storeu_si512( - reinterpret_cast<__m512i*>(to), from); -} - -template <> -EIGEN_DEVICE_FUNC inline Packet16f pgather(const float* from, - Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); - Packet16i stride_multiplier = - _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); - Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); - - return _mm512_i32gather_ps(indices, from, 4); -} -template <> -EIGEN_DEVICE_FUNC inline Packet8d pgather(const double* from, - Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); - Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); - Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); - - return _mm512_i32gather_pd(indices, from, 8); -} - -template <> -EIGEN_DEVICE_FUNC inline void pscatter(float* to, - const Packet16f& from, - Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); - Packet16i stride_multiplier = - _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); - Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); - _mm512_i32scatter_ps(to, indices, from, 4); -} -template <> -EIGEN_DEVICE_FUNC inline void pscatter(double* to, - const Packet8d& from, - Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); - Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); - Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); - _mm512_i32scatter_pd(to, indices, from, 8); -} - -template <> -EIGEN_STRONG_INLINE void pstore1(float* to, const float& a) { - Packet16f pa = pset1(a); - pstore(to, pa); -} -template <> -EIGEN_STRONG_INLINE void pstore1(double* to, const double& a) { - Packet8d pa = pset1(a); - pstore(to, pa); -} -template <> -EIGEN_STRONG_INLINE void pstore1(int* to, const int& a) { - Packet16i pa = pset1(a); - pstore(to, pa); -} - -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } - -template <> -EIGEN_STRONG_INLINE float pfirst(const Packet16f& a) { - return _mm_cvtss_f32(_mm512_extractf32x4_ps(a, 0)); -} -template <> -EIGEN_STRONG_INLINE double pfirst(const Packet8d& a) { - return _mm_cvtsd_f64(_mm256_extractf128_pd(_mm512_extractf64x4_pd(a, 0), 0)); -} -template <> -EIGEN_STRONG_INLINE int pfirst(const Packet16i& a) { - return _mm_extract_epi32(_mm512_extracti32x4_epi32(a, 0), 0); -} - -template<> EIGEN_STRONG_INLINE Packet16f preverse(const Packet16f& a) -{ - return _mm512_permutexvar_ps(_mm512_set_epi32(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15), a); -} - -template<> EIGEN_STRONG_INLINE Packet8d preverse(const Packet8d& a) -{ - return _mm512_permutexvar_pd(_mm512_set_epi32(0, 0, 0, 1, 0, 2, 0, 3, 0, 4, 0, 5, 0, 6, 0, 7), a); -} - -template<> EIGEN_STRONG_INLINE Packet16f pabs(const Packet16f& a) -{ - // _mm512_abs_ps intrinsic not found, so hack around it - return _mm512_castsi512_ps(_mm512_and_si512(_mm512_castps_si512(a), _mm512_set1_epi32(0x7fffffff))); -} -template <> -EIGEN_STRONG_INLINE Packet8d pabs(const Packet8d& a) { - // _mm512_abs_ps intrinsic not found, so hack around it - return _mm512_castsi512_pd(_mm512_and_si512(_mm512_castpd_si512(a), - _mm512_set1_epi64(0x7fffffffffffffff))); -} - -#ifdef EIGEN_VECTORIZE_AVX512DQ -// AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 -#define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ - __m256 OUTPUT##_0 = _mm512_extractf32x8_ps(INPUT, 0); \ - __m256 OUTPUT##_1 = _mm512_extractf32x8_ps(INPUT, 1) -#else -#define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ - __m256 OUTPUT##_0 = _mm256_insertf128_ps( \ - _mm256_castps128_ps256(_mm512_extractf32x4_ps(INPUT, 0)), \ - _mm512_extractf32x4_ps(INPUT, 1), 1); \ - __m256 OUTPUT##_1 = _mm256_insertf128_ps( \ - _mm256_castps128_ps256(_mm512_extractf32x4_ps(INPUT, 2)), \ - _mm512_extractf32x4_ps(INPUT, 3), 1); -#endif - -#ifdef EIGEN_VECTORIZE_AVX512DQ -#define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ - OUTPUT = _mm512_insertf32x8(_mm512_castps256_ps512(INPUTA), INPUTB, 1); -#else -#define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ - OUTPUT = _mm512_undefined_ps(); \ - OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 0), 0); \ - OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 1), 1); \ - OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 0), 2); \ - OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 1), 3); -#endif - -template <> -EIGEN_STRONG_INLINE float predux(const Packet16f& a) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - __m256 lane0 = _mm512_extractf32x8_ps(a, 0); - __m256 lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f x = _mm256_add_ps(lane0, lane1); - return predux(x); -#else - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 sum = _mm_add_ps(_mm_add_ps(lane0, lane1), _mm_add_ps(lane2, lane3)); - sum = _mm_hadd_ps(sum, sum); - sum = _mm_hadd_ps(sum, _mm_permute_ps(sum, 1)); - return _mm_cvtss_f32(sum); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d sum = _mm256_add_pd(lane0, lane1); - __m256d tmp0 = _mm256_hadd_pd(sum, _mm256_permute2f128_pd(sum, sum, 1)); - return _mm_cvtsd_f64(_mm256_castpd256_pd128(_mm256_hadd_pd(tmp0, tmp0))); -} - -template <> -EIGEN_STRONG_INLINE Packet8f predux_downto4(const Packet16f& a) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - return padd(lane0, lane1); -#else - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f sum0 = padd(lane0, lane2); - Packet4f sum1 = padd(lane1, lane3); - return _mm256_insertf128_ps(_mm256_castps128_ps256(sum0), sum1, 1); -#endif -} -template <> -EIGEN_STRONG_INLINE Packet4d predux_downto4(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = padd(lane0, lane1); - return res; -} - -template <> -EIGEN_STRONG_INLINE float predux_mul(const Packet16f& a) { -//#ifdef EIGEN_VECTORIZE_AVX512DQ -#if 0 - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_ps(res, res, 1)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#else - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = pmul(pmul(lane0, lane1), pmul(lane2, lane3)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux_mul(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(pmul(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_min(const Packet16f& a) { - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = _mm_min_ps(_mm_min_ps(lane0, lane1), _mm_min_ps(lane2, lane3)); - res = _mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} -template <> -EIGEN_STRONG_INLINE double predux_min(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = _mm256_min_pd(lane0, lane1); - res = _mm256_min_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_min_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_max(const Packet16f& a) { - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = _mm_max_ps(_mm_max_ps(lane0, lane1), _mm_max_ps(lane2, lane3)); - res = _mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} - -template <> -EIGEN_STRONG_INLINE double predux_max(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = _mm256_max_pd(lane0, lane1); - res = _mm256_max_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_max_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template<> EIGEN_STRONG_INLINE Packet16f preduxp(const Packet16f* vecs) -{ - EIGEN_EXTRACT_8f_FROM_16f(vecs[0], vecs0); - EIGEN_EXTRACT_8f_FROM_16f(vecs[1], vecs1); - EIGEN_EXTRACT_8f_FROM_16f(vecs[2], vecs2); - EIGEN_EXTRACT_8f_FROM_16f(vecs[3], vecs3); - EIGEN_EXTRACT_8f_FROM_16f(vecs[4], vecs4); - EIGEN_EXTRACT_8f_FROM_16f(vecs[5], vecs5); - EIGEN_EXTRACT_8f_FROM_16f(vecs[6], vecs6); - EIGEN_EXTRACT_8f_FROM_16f(vecs[7], vecs7); - EIGEN_EXTRACT_8f_FROM_16f(vecs[8], vecs8); - EIGEN_EXTRACT_8f_FROM_16f(vecs[9], vecs9); - EIGEN_EXTRACT_8f_FROM_16f(vecs[10], vecs10); - EIGEN_EXTRACT_8f_FROM_16f(vecs[11], vecs11); - EIGEN_EXTRACT_8f_FROM_16f(vecs[12], vecs12); - EIGEN_EXTRACT_8f_FROM_16f(vecs[13], vecs13); - EIGEN_EXTRACT_8f_FROM_16f(vecs[14], vecs14); - EIGEN_EXTRACT_8f_FROM_16f(vecs[15], vecs15); - - __m256 hsum1 = _mm256_hadd_ps(vecs0_0, vecs1_0); - __m256 hsum2 = _mm256_hadd_ps(vecs2_0, vecs3_0); - __m256 hsum3 = _mm256_hadd_ps(vecs4_0, vecs5_0); - __m256 hsum4 = _mm256_hadd_ps(vecs6_0, vecs7_0); - - __m256 hsum5 = _mm256_hadd_ps(hsum1, hsum1); - __m256 hsum6 = _mm256_hadd_ps(hsum2, hsum2); - __m256 hsum7 = _mm256_hadd_ps(hsum3, hsum3); - __m256 hsum8 = _mm256_hadd_ps(hsum4, hsum4); - - __m256 perm1 = _mm256_permute2f128_ps(hsum5, hsum5, 0x23); - __m256 perm2 = _mm256_permute2f128_ps(hsum6, hsum6, 0x23); - __m256 perm3 = _mm256_permute2f128_ps(hsum7, hsum7, 0x23); - __m256 perm4 = _mm256_permute2f128_ps(hsum8, hsum8, 0x23); - - __m256 sum1 = _mm256_add_ps(perm1, hsum5); - __m256 sum2 = _mm256_add_ps(perm2, hsum6); - __m256 sum3 = _mm256_add_ps(perm3, hsum7); - __m256 sum4 = _mm256_add_ps(perm4, hsum8); - - __m256 blend1 = _mm256_blend_ps(sum1, sum2, 0xcc); - __m256 blend2 = _mm256_blend_ps(sum3, sum4, 0xcc); - - __m256 final = _mm256_blend_ps(blend1, blend2, 0xf0); - - hsum1 = _mm256_hadd_ps(vecs0_1, vecs1_1); - hsum2 = _mm256_hadd_ps(vecs2_1, vecs3_1); - hsum3 = _mm256_hadd_ps(vecs4_1, vecs5_1); - hsum4 = _mm256_hadd_ps(vecs6_1, vecs7_1); - - hsum5 = _mm256_hadd_ps(hsum1, hsum1); - hsum6 = _mm256_hadd_ps(hsum2, hsum2); - hsum7 = _mm256_hadd_ps(hsum3, hsum3); - hsum8 = _mm256_hadd_ps(hsum4, hsum4); - - perm1 = _mm256_permute2f128_ps(hsum5, hsum5, 0x23); - perm2 = _mm256_permute2f128_ps(hsum6, hsum6, 0x23); - perm3 = _mm256_permute2f128_ps(hsum7, hsum7, 0x23); - perm4 = _mm256_permute2f128_ps(hsum8, hsum8, 0x23); - - sum1 = _mm256_add_ps(perm1, hsum5); - sum2 = _mm256_add_ps(perm2, hsum6); - sum3 = _mm256_add_ps(perm3, hsum7); - sum4 = _mm256_add_ps(perm4, hsum8); - - blend1 = _mm256_blend_ps(sum1, sum2, 0xcc); - blend2 = _mm256_blend_ps(sum3, sum4, 0xcc); - - final = padd(final, _mm256_blend_ps(blend1, blend2, 0xf0)); - - hsum1 = _mm256_hadd_ps(vecs8_0, vecs9_0); - hsum2 = _mm256_hadd_ps(vecs10_0, vecs11_0); - hsum3 = _mm256_hadd_ps(vecs12_0, vecs13_0); - hsum4 = _mm256_hadd_ps(vecs14_0, vecs15_0); - - hsum5 = _mm256_hadd_ps(hsum1, hsum1); - hsum6 = _mm256_hadd_ps(hsum2, hsum2); - hsum7 = _mm256_hadd_ps(hsum3, hsum3); - hsum8 = _mm256_hadd_ps(hsum4, hsum4); - - perm1 = _mm256_permute2f128_ps(hsum5, hsum5, 0x23); - perm2 = _mm256_permute2f128_ps(hsum6, hsum6, 0x23); - perm3 = _mm256_permute2f128_ps(hsum7, hsum7, 0x23); - perm4 = _mm256_permute2f128_ps(hsum8, hsum8, 0x23); - - sum1 = _mm256_add_ps(perm1, hsum5); - sum2 = _mm256_add_ps(perm2, hsum6); - sum3 = _mm256_add_ps(perm3, hsum7); - sum4 = _mm256_add_ps(perm4, hsum8); - - blend1 = _mm256_blend_ps(sum1, sum2, 0xcc); - blend2 = _mm256_blend_ps(sum3, sum4, 0xcc); - - __m256 final_1 = _mm256_blend_ps(blend1, blend2, 0xf0); - - hsum1 = _mm256_hadd_ps(vecs8_1, vecs9_1); - hsum2 = _mm256_hadd_ps(vecs10_1, vecs11_1); - hsum3 = _mm256_hadd_ps(vecs12_1, vecs13_1); - hsum4 = _mm256_hadd_ps(vecs14_1, vecs15_1); - - hsum5 = _mm256_hadd_ps(hsum1, hsum1); - hsum6 = _mm256_hadd_ps(hsum2, hsum2); - hsum7 = _mm256_hadd_ps(hsum3, hsum3); - hsum8 = _mm256_hadd_ps(hsum4, hsum4); - - perm1 = _mm256_permute2f128_ps(hsum5, hsum5, 0x23); - perm2 = _mm256_permute2f128_ps(hsum6, hsum6, 0x23); - perm3 = _mm256_permute2f128_ps(hsum7, hsum7, 0x23); - perm4 = _mm256_permute2f128_ps(hsum8, hsum8, 0x23); - - sum1 = _mm256_add_ps(perm1, hsum5); - sum2 = _mm256_add_ps(perm2, hsum6); - sum3 = _mm256_add_ps(perm3, hsum7); - sum4 = _mm256_add_ps(perm4, hsum8); - - blend1 = _mm256_blend_ps(sum1, sum2, 0xcc); - blend2 = _mm256_blend_ps(sum3, sum4, 0xcc); - - final_1 = padd(final_1, _mm256_blend_ps(blend1, blend2, 0xf0)); - - __m512 final_output; - - EIGEN_INSERT_8f_INTO_16f(final_output, final, final_1); - return final_output; -} - -template<> EIGEN_STRONG_INLINE Packet8d preduxp(const Packet8d* vecs) -{ - Packet4d vecs0_0 = _mm512_extractf64x4_pd(vecs[0], 0); - Packet4d vecs0_1 = _mm512_extractf64x4_pd(vecs[0], 1); - - Packet4d vecs1_0 = _mm512_extractf64x4_pd(vecs[1], 0); - Packet4d vecs1_1 = _mm512_extractf64x4_pd(vecs[1], 1); - - Packet4d vecs2_0 = _mm512_extractf64x4_pd(vecs[2], 0); - Packet4d vecs2_1 = _mm512_extractf64x4_pd(vecs[2], 1); - - Packet4d vecs3_0 = _mm512_extractf64x4_pd(vecs[3], 0); - Packet4d vecs3_1 = _mm512_extractf64x4_pd(vecs[3], 1); - - Packet4d vecs4_0 = _mm512_extractf64x4_pd(vecs[4], 0); - Packet4d vecs4_1 = _mm512_extractf64x4_pd(vecs[4], 1); - - Packet4d vecs5_0 = _mm512_extractf64x4_pd(vecs[5], 0); - Packet4d vecs5_1 = _mm512_extractf64x4_pd(vecs[5], 1); - - Packet4d vecs6_0 = _mm512_extractf64x4_pd(vecs[6], 0); - Packet4d vecs6_1 = _mm512_extractf64x4_pd(vecs[6], 1); - - Packet4d vecs7_0 = _mm512_extractf64x4_pd(vecs[7], 0); - Packet4d vecs7_1 = _mm512_extractf64x4_pd(vecs[7], 1); - - Packet4d tmp0, tmp1; - - tmp0 = _mm256_hadd_pd(vecs0_0, vecs1_0); - tmp0 = _mm256_add_pd(tmp0, _mm256_permute2f128_pd(tmp0, tmp0, 1)); - - tmp1 = _mm256_hadd_pd(vecs2_0, vecs3_0); - tmp1 = _mm256_add_pd(tmp1, _mm256_permute2f128_pd(tmp1, tmp1, 1)); - - __m256d final_0 = _mm256_blend_pd(tmp0, tmp1, 0xC); - - tmp0 = _mm256_hadd_pd(vecs0_1, vecs1_1); - tmp0 = _mm256_add_pd(tmp0, _mm256_permute2f128_pd(tmp0, tmp0, 1)); - - tmp1 = _mm256_hadd_pd(vecs2_1, vecs3_1); - tmp1 = _mm256_add_pd(tmp1, _mm256_permute2f128_pd(tmp1, tmp1, 1)); - - final_0 = padd(final_0, _mm256_blend_pd(tmp0, tmp1, 0xC)); - - tmp0 = _mm256_hadd_pd(vecs4_0, vecs5_0); - tmp0 = _mm256_add_pd(tmp0, _mm256_permute2f128_pd(tmp0, tmp0, 1)); - - tmp1 = _mm256_hadd_pd(vecs6_0, vecs7_0); - tmp1 = _mm256_add_pd(tmp1, _mm256_permute2f128_pd(tmp1, tmp1, 1)); - - __m256d final_1 = _mm256_blend_pd(tmp0, tmp1, 0xC); - - tmp0 = _mm256_hadd_pd(vecs4_1, vecs5_1); - tmp0 = _mm256_add_pd(tmp0, _mm256_permute2f128_pd(tmp0, tmp0, 1)); - - tmp1 = _mm256_hadd_pd(vecs6_1, vecs7_1); - tmp1 = _mm256_add_pd(tmp1, _mm256_permute2f128_pd(tmp1, tmp1, 1)); - - final_1 = padd(final_1, _mm256_blend_pd(tmp0, tmp1, 0xC)); - - __m512d final_output = _mm512_insertf64x4(final_output, final_0, 0); - - return _mm512_insertf64x4(final_output, final_1, 1); -} - - - -#define PACK_OUTPUT(OUTPUT, INPUT, INDEX, STRIDE) \ - EIGEN_INSERT_8f_INTO_16f(OUTPUT[INDEX], INPUT[INDEX], INPUT[INDEX + STRIDE]); - -EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock& kernel) { - __m512 T0 = _mm512_unpacklo_ps(kernel.packet[0], kernel.packet[1]); - __m512 T1 = _mm512_unpackhi_ps(kernel.packet[0], kernel.packet[1]); - __m512 T2 = _mm512_unpacklo_ps(kernel.packet[2], kernel.packet[3]); - __m512 T3 = _mm512_unpackhi_ps(kernel.packet[2], kernel.packet[3]); - __m512 T4 = _mm512_unpacklo_ps(kernel.packet[4], kernel.packet[5]); - __m512 T5 = _mm512_unpackhi_ps(kernel.packet[4], kernel.packet[5]); - __m512 T6 = _mm512_unpacklo_ps(kernel.packet[6], kernel.packet[7]); - __m512 T7 = _mm512_unpackhi_ps(kernel.packet[6], kernel.packet[7]); - __m512 T8 = _mm512_unpacklo_ps(kernel.packet[8], kernel.packet[9]); - __m512 T9 = _mm512_unpackhi_ps(kernel.packet[8], kernel.packet[9]); - __m512 T10 = _mm512_unpacklo_ps(kernel.packet[10], kernel.packet[11]); - __m512 T11 = _mm512_unpackhi_ps(kernel.packet[10], kernel.packet[11]); - __m512 T12 = _mm512_unpacklo_ps(kernel.packet[12], kernel.packet[13]); - __m512 T13 = _mm512_unpackhi_ps(kernel.packet[12], kernel.packet[13]); - __m512 T14 = _mm512_unpacklo_ps(kernel.packet[14], kernel.packet[15]); - __m512 T15 = _mm512_unpackhi_ps(kernel.packet[14], kernel.packet[15]); - __m512 S0 = _mm512_shuffle_ps(T0, T2, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S1 = _mm512_shuffle_ps(T0, T2, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S2 = _mm512_shuffle_ps(T1, T3, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S3 = _mm512_shuffle_ps(T1, T3, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S4 = _mm512_shuffle_ps(T4, T6, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S5 = _mm512_shuffle_ps(T4, T6, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S6 = _mm512_shuffle_ps(T5, T7, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S7 = _mm512_shuffle_ps(T5, T7, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S8 = _mm512_shuffle_ps(T8, T10, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S9 = _mm512_shuffle_ps(T8, T10, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S10 = _mm512_shuffle_ps(T9, T11, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S11 = _mm512_shuffle_ps(T9, T11, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S12 = _mm512_shuffle_ps(T12, T14, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S13 = _mm512_shuffle_ps(T12, T14, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S14 = _mm512_shuffle_ps(T13, T15, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S15 = _mm512_shuffle_ps(T13, T15, _MM_SHUFFLE(3, 2, 3, 2)); - - EIGEN_EXTRACT_8f_FROM_16f(S0, S0); - EIGEN_EXTRACT_8f_FROM_16f(S1, S1); - EIGEN_EXTRACT_8f_FROM_16f(S2, S2); - EIGEN_EXTRACT_8f_FROM_16f(S3, S3); - EIGEN_EXTRACT_8f_FROM_16f(S4, S4); - EIGEN_EXTRACT_8f_FROM_16f(S5, S5); - EIGEN_EXTRACT_8f_FROM_16f(S6, S6); - EIGEN_EXTRACT_8f_FROM_16f(S7, S7); - EIGEN_EXTRACT_8f_FROM_16f(S8, S8); - EIGEN_EXTRACT_8f_FROM_16f(S9, S9); - EIGEN_EXTRACT_8f_FROM_16f(S10, S10); - EIGEN_EXTRACT_8f_FROM_16f(S11, S11); - EIGEN_EXTRACT_8f_FROM_16f(S12, S12); - EIGEN_EXTRACT_8f_FROM_16f(S13, S13); - EIGEN_EXTRACT_8f_FROM_16f(S14, S14); - EIGEN_EXTRACT_8f_FROM_16f(S15, S15); - - PacketBlock tmp; - - tmp.packet[0] = _mm256_permute2f128_ps(S0_0, S4_0, 0x20); - tmp.packet[1] = _mm256_permute2f128_ps(S1_0, S5_0, 0x20); - tmp.packet[2] = _mm256_permute2f128_ps(S2_0, S6_0, 0x20); - tmp.packet[3] = _mm256_permute2f128_ps(S3_0, S7_0, 0x20); - tmp.packet[4] = _mm256_permute2f128_ps(S0_0, S4_0, 0x31); - tmp.packet[5] = _mm256_permute2f128_ps(S1_0, S5_0, 0x31); - tmp.packet[6] = _mm256_permute2f128_ps(S2_0, S6_0, 0x31); - tmp.packet[7] = _mm256_permute2f128_ps(S3_0, S7_0, 0x31); - - tmp.packet[8] = _mm256_permute2f128_ps(S0_1, S4_1, 0x20); - tmp.packet[9] = _mm256_permute2f128_ps(S1_1, S5_1, 0x20); - tmp.packet[10] = _mm256_permute2f128_ps(S2_1, S6_1, 0x20); - tmp.packet[11] = _mm256_permute2f128_ps(S3_1, S7_1, 0x20); - tmp.packet[12] = _mm256_permute2f128_ps(S0_1, S4_1, 0x31); - tmp.packet[13] = _mm256_permute2f128_ps(S1_1, S5_1, 0x31); - tmp.packet[14] = _mm256_permute2f128_ps(S2_1, S6_1, 0x31); - tmp.packet[15] = _mm256_permute2f128_ps(S3_1, S7_1, 0x31); - - // Second set of _m256 outputs - tmp.packet[16] = _mm256_permute2f128_ps(S8_0, S12_0, 0x20); - tmp.packet[17] = _mm256_permute2f128_ps(S9_0, S13_0, 0x20); - tmp.packet[18] = _mm256_permute2f128_ps(S10_0, S14_0, 0x20); - tmp.packet[19] = _mm256_permute2f128_ps(S11_0, S15_0, 0x20); - tmp.packet[20] = _mm256_permute2f128_ps(S8_0, S12_0, 0x31); - tmp.packet[21] = _mm256_permute2f128_ps(S9_0, S13_0, 0x31); - tmp.packet[22] = _mm256_permute2f128_ps(S10_0, S14_0, 0x31); - tmp.packet[23] = _mm256_permute2f128_ps(S11_0, S15_0, 0x31); - - tmp.packet[24] = _mm256_permute2f128_ps(S8_1, S12_1, 0x20); - tmp.packet[25] = _mm256_permute2f128_ps(S9_1, S13_1, 0x20); - tmp.packet[26] = _mm256_permute2f128_ps(S10_1, S14_1, 0x20); - tmp.packet[27] = _mm256_permute2f128_ps(S11_1, S15_1, 0x20); - tmp.packet[28] = _mm256_permute2f128_ps(S8_1, S12_1, 0x31); - tmp.packet[29] = _mm256_permute2f128_ps(S9_1, S13_1, 0x31); - tmp.packet[30] = _mm256_permute2f128_ps(S10_1, S14_1, 0x31); - tmp.packet[31] = _mm256_permute2f128_ps(S11_1, S15_1, 0x31); - - // Pack them into the output - PACK_OUTPUT(kernel.packet, tmp.packet, 0, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 1, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 2, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 3, 16); - - PACK_OUTPUT(kernel.packet, tmp.packet, 4, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 5, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 6, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 7, 16); - - PACK_OUTPUT(kernel.packet, tmp.packet, 8, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 9, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 10, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 11, 16); - - PACK_OUTPUT(kernel.packet, tmp.packet, 12, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 13, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 14, 16); - PACK_OUTPUT(kernel.packet, tmp.packet, 15, 16); -} -#define PACK_OUTPUT_2(OUTPUT, INPUT, INDEX, STRIDE) \ - EIGEN_INSERT_8f_INTO_16f(OUTPUT[INDEX], INPUT[2 * INDEX], \ - INPUT[2 * INDEX + STRIDE]); - -EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock& kernel) { - __m512 T0 = _mm512_unpacklo_ps(kernel.packet[0], kernel.packet[1]); - __m512 T1 = _mm512_unpackhi_ps(kernel.packet[0], kernel.packet[1]); - __m512 T2 = _mm512_unpacklo_ps(kernel.packet[2], kernel.packet[3]); - __m512 T3 = _mm512_unpackhi_ps(kernel.packet[2], kernel.packet[3]); - - __m512 S0 = _mm512_shuffle_ps(T0, T2, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S1 = _mm512_shuffle_ps(T0, T2, _MM_SHUFFLE(3, 2, 3, 2)); - __m512 S2 = _mm512_shuffle_ps(T1, T3, _MM_SHUFFLE(1, 0, 1, 0)); - __m512 S3 = _mm512_shuffle_ps(T1, T3, _MM_SHUFFLE(3, 2, 3, 2)); - - EIGEN_EXTRACT_8f_FROM_16f(S0, S0); - EIGEN_EXTRACT_8f_FROM_16f(S1, S1); - EIGEN_EXTRACT_8f_FROM_16f(S2, S2); - EIGEN_EXTRACT_8f_FROM_16f(S3, S3); - - PacketBlock tmp; - - tmp.packet[0] = _mm256_permute2f128_ps(S0_0, S1_0, 0x20); - tmp.packet[1] = _mm256_permute2f128_ps(S2_0, S3_0, 0x20); - tmp.packet[2] = _mm256_permute2f128_ps(S0_0, S1_0, 0x31); - tmp.packet[3] = _mm256_permute2f128_ps(S2_0, S3_0, 0x31); - - tmp.packet[4] = _mm256_permute2f128_ps(S0_1, S1_1, 0x20); - tmp.packet[5] = _mm256_permute2f128_ps(S2_1, S3_1, 0x20); - tmp.packet[6] = _mm256_permute2f128_ps(S0_1, S1_1, 0x31); - tmp.packet[7] = _mm256_permute2f128_ps(S2_1, S3_1, 0x31); - - PACK_OUTPUT_2(kernel.packet, tmp.packet, 0, 1); - PACK_OUTPUT_2(kernel.packet, tmp.packet, 1, 1); - PACK_OUTPUT_2(kernel.packet, tmp.packet, 2, 1); - PACK_OUTPUT_2(kernel.packet, tmp.packet, 3, 1); -} - -#define PACK_OUTPUT_SQ_D(OUTPUT, INPUT, INDEX, STRIDE) \ - OUTPUT[INDEX] = _mm512_insertf64x4(OUTPUT[INDEX], INPUT[INDEX], 0); \ - OUTPUT[INDEX] = _mm512_insertf64x4(OUTPUT[INDEX], INPUT[INDEX + STRIDE], 1); - -#define PACK_OUTPUT_D(OUTPUT, INPUT, INDEX, STRIDE) \ - OUTPUT[INDEX] = _mm512_insertf64x4(OUTPUT[INDEX], INPUT[(2 * INDEX)], 0); \ - OUTPUT[INDEX] = \ - _mm512_insertf64x4(OUTPUT[INDEX], INPUT[(2 * INDEX) + STRIDE], 1); - -EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock& kernel) { - __m512d T0 = _mm512_shuffle_pd(kernel.packet[0], kernel.packet[1], 0); - __m512d T1 = _mm512_shuffle_pd(kernel.packet[0], kernel.packet[1], 0xff); - __m512d T2 = _mm512_shuffle_pd(kernel.packet[2], kernel.packet[3], 0); - __m512d T3 = _mm512_shuffle_pd(kernel.packet[2], kernel.packet[3], 0xff); - - PacketBlock tmp; - - tmp.packet[0] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 0), - _mm512_extractf64x4_pd(T2, 0), 0x20); - tmp.packet[1] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 0), - _mm512_extractf64x4_pd(T3, 0), 0x20); - tmp.packet[2] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 0), - _mm512_extractf64x4_pd(T2, 0), 0x31); - tmp.packet[3] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 0), - _mm512_extractf64x4_pd(T3, 0), 0x31); - - tmp.packet[4] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 1), - _mm512_extractf64x4_pd(T2, 1), 0x20); - tmp.packet[5] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 1), - _mm512_extractf64x4_pd(T3, 1), 0x20); - tmp.packet[6] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 1), - _mm512_extractf64x4_pd(T2, 1), 0x31); - tmp.packet[7] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 1), - _mm512_extractf64x4_pd(T3, 1), 0x31); - - PACK_OUTPUT_D(kernel.packet, tmp.packet, 0, 1); - PACK_OUTPUT_D(kernel.packet, tmp.packet, 1, 1); - PACK_OUTPUT_D(kernel.packet, tmp.packet, 2, 1); - PACK_OUTPUT_D(kernel.packet, tmp.packet, 3, 1); -} - -EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock& kernel) { - __m512d T0 = _mm512_unpacklo_pd(kernel.packet[0], kernel.packet[1]); - __m512d T1 = _mm512_unpackhi_pd(kernel.packet[0], kernel.packet[1]); - __m512d T2 = _mm512_unpacklo_pd(kernel.packet[2], kernel.packet[3]); - __m512d T3 = _mm512_unpackhi_pd(kernel.packet[2], kernel.packet[3]); - __m512d T4 = _mm512_unpacklo_pd(kernel.packet[4], kernel.packet[5]); - __m512d T5 = _mm512_unpackhi_pd(kernel.packet[4], kernel.packet[5]); - __m512d T6 = _mm512_unpacklo_pd(kernel.packet[6], kernel.packet[7]); - __m512d T7 = _mm512_unpackhi_pd(kernel.packet[6], kernel.packet[7]); - - PacketBlock tmp; - - tmp.packet[0] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 0), - _mm512_extractf64x4_pd(T2, 0), 0x20); - tmp.packet[1] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 0), - _mm512_extractf64x4_pd(T3, 0), 0x20); - tmp.packet[2] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 0), - _mm512_extractf64x4_pd(T2, 0), 0x31); - tmp.packet[3] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 0), - _mm512_extractf64x4_pd(T3, 0), 0x31); - - tmp.packet[4] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 1), - _mm512_extractf64x4_pd(T2, 1), 0x20); - tmp.packet[5] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 1), - _mm512_extractf64x4_pd(T3, 1), 0x20); - tmp.packet[6] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T0, 1), - _mm512_extractf64x4_pd(T2, 1), 0x31); - tmp.packet[7] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T1, 1), - _mm512_extractf64x4_pd(T3, 1), 0x31); - - tmp.packet[8] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T4, 0), - _mm512_extractf64x4_pd(T6, 0), 0x20); - tmp.packet[9] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T5, 0), - _mm512_extractf64x4_pd(T7, 0), 0x20); - tmp.packet[10] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T4, 0), - _mm512_extractf64x4_pd(T6, 0), 0x31); - tmp.packet[11] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T5, 0), - _mm512_extractf64x4_pd(T7, 0), 0x31); - - tmp.packet[12] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T4, 1), - _mm512_extractf64x4_pd(T6, 1), 0x20); - tmp.packet[13] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T5, 1), - _mm512_extractf64x4_pd(T7, 1), 0x20); - tmp.packet[14] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T4, 1), - _mm512_extractf64x4_pd(T6, 1), 0x31); - tmp.packet[15] = _mm256_permute2f128_pd(_mm512_extractf64x4_pd(T5, 1), - _mm512_extractf64x4_pd(T7, 1), 0x31); - - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 0, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 1, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 2, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 3, 8); - - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 4, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 5, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 6, 8); - PACK_OUTPUT_SQ_D(kernel.packet, tmp.packet, 7, 8); -} -template <> -EIGEN_STRONG_INLINE Packet16f pblend(const Selector<16>& /*ifPacket*/, - const Packet16f& /*thenPacket*/, - const Packet16f& /*elsePacket*/) { - assert(false && "To be implemented"); - return Packet16f(); -} -template <> -EIGEN_STRONG_INLINE Packet8d pblend(const Selector<8>& ifPacket, - const Packet8d& thenPacket, - const Packet8d& elsePacket) { - __mmask8 m = (ifPacket.select[0] ) - | (ifPacket.select[1]<<1) - | (ifPacket.select[2]<<2) - | (ifPacket.select[3]<<3) - | (ifPacket.select[4]<<4) - | (ifPacket.select[5]<<5) - | (ifPacket.select[6]<<6) - | (ifPacket.select[7]<<7); - return _mm512_mask_blend_pd(m, elsePacket, thenPacket); -} - -template<> EIGEN_STRONG_INLINE Packet16i pcast(const Packet16f& a) { - return _mm512_cvttps_epi32(a); -} - -template<> EIGEN_STRONG_INLINE Packet16f pcast(const Packet16i& a) { - return _mm512_cvtepi32_ps(a); -} - -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet16f& first, - const Packet16f& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - Offset + 15, Offset + 14, Offset + 13, Offset + 12, Offset + 11, - Offset + 10, Offset + 9, Offset + 8, Offset + 7, Offset + 6, - Offset + 5, Offset + 4, Offset + 3, Offset + 2, Offset + 1, Offset); - - __m512i second_idx = - _mm512_set_epi32(Offset - 1, Offset - 2, Offset - 3, Offset - 4, - Offset - 5, Offset - 6, Offset - 7, Offset - 8, - Offset - 9, Offset - 10, Offset - 11, Offset - 12, - Offset - 13, Offset - 14, Offset - 15, Offset - 16); - - unsigned short mask = 0xFFFF; - mask <<= (16 - Offset); - - first = _mm512_permutexvar_ps(first_idx, first); - Packet16f tmp = _mm512_permutexvar_ps(second_idx, second); - first = _mm512_mask_blend_ps(mask, first, tmp); - } - } -}; -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet8d& first, const Packet8d& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - 0, Offset + 7, 0, Offset + 6, 0, Offset + 5, 0, Offset + 4, 0, - Offset + 3, 0, Offset + 2, 0, Offset + 1, 0, Offset); - - __m512i second_idx = _mm512_set_epi32( - 0, Offset - 1, 0, Offset - 2, 0, Offset - 3, 0, Offset - 4, 0, - Offset - 5, 0, Offset - 6, 0, Offset - 7, 0, Offset - 8); - - unsigned char mask = 0xFF; - mask <<= (8 - Offset); - - first = _mm512_permutexvar_pd(first_idx, first); - Packet8d tmp = _mm512_permutexvar_pd(second_idx, second); - first = _mm512_mask_blend_pd(mask, first, tmp); - } - } -}; - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PACKET_MATH_AVX512_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/Complex.h deleted file mode 100644 index 3e665730cfa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/Complex.h +++ /dev/null @@ -1,430 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// Copyright (C) 2010-2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX32_ALTIVEC_H -#define EIGEN_COMPLEX32_ALTIVEC_H - -namespace Eigen { - -namespace internal { - -static Packet4ui p4ui_CONJ_XOR = vec_mergeh((Packet4ui)p4i_ZERO, (Packet4ui)p4f_MZERO);//{ 0x00000000, 0x80000000, 0x00000000, 0x80000000 }; -#ifdef __VSX__ -#if defined(_BIG_ENDIAN) -static Packet2ul p2ul_CONJ_XOR1 = (Packet2ul) vec_sld((Packet4ui) p2d_MZERO, (Packet4ui) p2l_ZERO, 8);//{ 0x8000000000000000, 0x0000000000000000 }; -static Packet2ul p2ul_CONJ_XOR2 = (Packet2ul) vec_sld((Packet4ui) p2l_ZERO, (Packet4ui) p2d_MZERO, 8);//{ 0x8000000000000000, 0x0000000000000000 }; -#else -static Packet2ul p2ul_CONJ_XOR1 = (Packet2ul) vec_sld((Packet4ui) p2l_ZERO, (Packet4ui) p2d_MZERO, 8);//{ 0x8000000000000000, 0x0000000000000000 }; -static Packet2ul p2ul_CONJ_XOR2 = (Packet2ul) vec_sld((Packet4ui) p2d_MZERO, (Packet4ui) p2l_ZERO, 8);//{ 0x8000000000000000, 0x0000000000000000 }; -#endif -#endif - -//---------- float ---------- -struct Packet2cf -{ - EIGEN_STRONG_INLINE explicit Packet2cf() : v(p4f_ZERO) {} - EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {} - Packet4f v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, -#ifdef __VSX__ - HasBlend = 1, -#endif - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; - -template<> EIGEN_STRONG_INLINE Packet2cf pset1(const std::complex& from) -{ - Packet2cf res; - if((std::ptrdiff_t(&from) % 16) == 0) - res.v = pload((const float *)&from); - else - res.v = ploadu((const float *)&from); - res.v = vec_perm(res.v, res.v, p16uc_PSET64_HI); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet2cf pload(const std::complex* from) { return Packet2cf(pload((const float *) from)); } -template<> EIGEN_STRONG_INLINE Packet2cf ploadu(const std::complex* from) { return Packet2cf(ploadu((const float*) from)); } -template<> EIGEN_STRONG_INLINE Packet2cf ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cf& from) { pstore((float*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cf& from) { pstoreu((float*)to, from.v); } - -template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather, Packet2cf>(const std::complex* from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - return pload(af); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cf>(std::complex* to, const Packet2cf& from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - pstore >((std::complex *) af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; -} - -template<> EIGEN_STRONG_INLINE Packet2cf padd(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(a.v + b.v); } -template<> EIGEN_STRONG_INLINE Packet2cf psub(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(a.v - b.v); } -template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate(a.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) { return Packet2cf(pxor(a.v, reinterpret_cast(p4ui_CONJ_XOR))); } - -template<> EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) -{ - Packet4f v1, v2; - - // Permute and multiply the real parts of a and b - v1 = vec_perm(a.v, a.v, p16uc_PSET32_WODD); - // Get the imaginary parts of a - v2 = vec_perm(a.v, a.v, p16uc_PSET32_WEVEN); - // multiply a_re * b - v1 = vec_madd(v1, b.v, p4f_ZERO); - // multiply a_im * b and get the conjugate result - v2 = vec_madd(v2, b.v, p4f_ZERO); - v2 = reinterpret_cast(pxor(v2, reinterpret_cast(p4ui_CONJ_XOR))); - // permute back to a proper order - v2 = vec_perm(v2, v2, p16uc_COMPLEX32_REV); - - return Packet2cf(padd(v1, v2)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pand (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pand(a.v, b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf por (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(por(a.v, b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pxor (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pxor(a.v, b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pandnot(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pandnot(a.v, b.v)); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_PPC_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cf& a) -{ - std::complex EIGEN_ALIGN16 res[2]; - pstore((float *)&res, a.v); - - return res[0]; -} - -template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) -{ - Packet4f rev_a; - rev_a = vec_perm(a.v, a.v, p16uc_COMPLEX32_REV2); - return Packet2cf(rev_a); -} - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cf& a) -{ - Packet4f b; - b = vec_sld(a.v, a.v, 8); - b = padd(a.v, b); - return pfirst(Packet2cf(b)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf preduxp(const Packet2cf* vecs) -{ - Packet4f b1, b2; -#ifdef _BIG_ENDIAN - b1 = vec_sld(vecs[0].v, vecs[1].v, 8); - b2 = vec_sld(vecs[1].v, vecs[0].v, 8); -#else - b1 = vec_sld(vecs[1].v, vecs[0].v, 8); - b2 = vec_sld(vecs[0].v, vecs[1].v, 8); -#endif - b2 = vec_sld(b2, b2, 8); - b2 = padd(b1, b2); - - return Packet2cf(b2); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cf& a) -{ - Packet4f b; - Packet2cf prod; - b = vec_sld(a.v, a.v, 8); - prod = pmul(a, Packet2cf(b)); - - return pfirst(prod); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) - { - if (Offset==1) - { -#ifdef _BIG_ENDIAN - first.v = vec_sld(first.v, second.v, 8); -#else - first.v = vec_sld(second.v, first.v, 8); -#endif - } - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) - -template<> EIGEN_STRONG_INLINE Packet2cf pdiv(const Packet2cf& a, const Packet2cf& b) -{ - // TODO optimize it for AltiVec - Packet2cf res = conj_helper().pmul(a, b); - Packet4f s = pmul(b.v, b.v); - return Packet2cf(pdiv(res.v, padd(s, vec_perm(s, s, p16uc_COMPLEX32_REV)))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pcplxflip(const Packet2cf& x) -{ - return Packet2cf(vec_perm(x.v, x.v, p16uc_COMPLEX32_REV)); -} - -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel) -{ - Packet4f tmp = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_HI); - kernel.packet[1].v = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_LO); - kernel.packet[0].v = tmp; -} - -#ifdef __VSX__ -template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { - Packet2cf result; - result.v = reinterpret_cast(pblend(ifPacket, reinterpret_cast(thenPacket.v), reinterpret_cast(elsePacket.v))); - return result; -} -#endif - -//---------- double ---------- -#ifdef __VSX__ -struct Packet1cd -{ - EIGEN_STRONG_INLINE Packet1cd() {} - EIGEN_STRONG_INLINE explicit Packet1cd(const Packet2d& a) : v(a) {} - Packet2d v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet1cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 0, - size = 1, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; - -template<> EIGEN_STRONG_INLINE Packet1cd pload (const std::complex* from) { return Packet1cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd ploadu(const std::complex* from) { return Packet1cd(ploadu((const double*)from)); } -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet1cd& from) { pstore((double*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet1cd& from) { pstoreu((double*)to, from.v); } - -template<> EIGEN_STRONG_INLINE Packet1cd pset1(const std::complex& from) -{ /* here we really have to use unaligned loads :( */ return ploadu(&from); } - -template<> EIGEN_DEVICE_FUNC inline Packet1cd pgather, Packet1cd>(const std::complex* from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - return pload(af); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet1cd>(std::complex* to, const Packet1cd& from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - pstore >(af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; -} - -template<> EIGEN_STRONG_INLINE Packet1cd padd(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(a.v + b.v); } -template<> EIGEN_STRONG_INLINE Packet1cd psub(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(a.v - b.v); } -template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } -template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) { return Packet1cd(pxor(a.v, reinterpret_cast(p2ul_CONJ_XOR2))); } - -template<> EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) -{ - Packet2d a_re, a_im, v1, v2; - - // Permute and multiply the real parts of a and b - a_re = vec_perm(a.v, a.v, p16uc_PSET64_HI); - // Get the imaginary parts of a - a_im = vec_perm(a.v, a.v, p16uc_PSET64_LO); - // multiply a_re * b - v1 = vec_madd(a_re, b.v, p2d_ZERO); - // multiply a_im * b and get the conjugate result - v2 = vec_madd(a_im, b.v, p2d_ZERO); - v2 = reinterpret_cast(vec_sld(reinterpret_cast(v2), reinterpret_cast(v2), 8)); - v2 = pxor(v2, reinterpret_cast(p2ul_CONJ_XOR1)); - - return Packet1cd(padd(v1, v2)); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pand (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(pand(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd por (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(por(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pxor (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(pxor(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pandnot(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(pandnot(a.v, b.v)); } - -template<> EIGEN_STRONG_INLINE Packet1cd ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_PPC_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet1cd& a) -{ - std::complex EIGEN_ALIGN16 res[2]; - pstore >(res, a); - - return res[0]; -} - -template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet1cd& a) { return pfirst(a); } -template<> EIGEN_STRONG_INLINE Packet1cd preduxp(const Packet1cd* vecs) { return vecs[0]; } - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet1cd& a) { return pfirst(a); } - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) - { - // FIXME is it sure we never have to align a Packet1cd? - // Even though a std::complex has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) - -template<> EIGEN_STRONG_INLINE Packet1cd pdiv(const Packet1cd& a, const Packet1cd& b) -{ - // TODO optimize it for AltiVec - Packet1cd res = conj_helper().pmul(a,b); - Packet2d s = pmul(b.v, b.v); - return Packet1cd(pdiv(res.v, padd(s, vec_perm(s, s, p16uc_REVERSE64)))); -} - -EIGEN_STRONG_INLINE Packet1cd pcplxflip/**/(const Packet1cd& x) -{ - return Packet1cd(preverse(Packet2d(x.v))); -} - -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel) -{ - Packet2d tmp = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_HI); - kernel.packet[1].v = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_LO); - kernel.packet[0].v = tmp; -} -#endif // __VSX__ -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX32_ALTIVEC_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/MathFunctions.h deleted file mode 100644 index c5e4bede74d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/MathFunctions.h +++ /dev/null @@ -1,322 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007 Julien Pommier -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* The sin, cos, exp, and log functions of this file come from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -#ifndef EIGEN_MATH_FUNCTIONS_ALTIVEC_H -#define EIGEN_MATH_FUNCTIONS_ALTIVEC_H - -namespace Eigen { - -namespace internal { - -static _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); -static _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); -static _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); -static _EIGEN_DECLARE_CONST_Packet4i(23, 23); - -static _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(inv_mant_mask, ~0x7f800000); - -/* the smallest non denormalized float number */ -static _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(min_norm_pos, 0x00800000); -static _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(minus_inf, 0xff800000); // -1.f/0.f -static _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(minus_nan, 0xffffffff); - -/* natural logarithm computed for 4 simultaneous float - return NaN for x <= 0 -*/ -static _EIGEN_DECLARE_CONST_Packet4f(cephes_SQRTHF, 0.707106781186547524f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p0, 7.0376836292E-2f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p1, - 1.1514610310E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p2, 1.1676998740E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p3, - 1.2420140846E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p4, + 1.4249322787E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p5, - 1.6668057665E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p6, + 2.0000714765E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p7, - 2.4999993993E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p8, + 3.3333331174E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q1, -2.12194440e-4f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q2, 0.693359375f); - -static _EIGEN_DECLARE_CONST_Packet4f(exp_hi, 88.3762626647950f); -static _EIGEN_DECLARE_CONST_Packet4f(exp_lo, -88.3762626647949f); - -static _EIGEN_DECLARE_CONST_Packet4f(cephes_LOG2EF, 1.44269504088896341f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C1, 0.693359375f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C2, -2.12194440e-4f); - -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p0, 1.9875691500E-4f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p1, 1.3981999507E-3f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p2, 8.3334519073E-3f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p3, 4.1665795894E-2f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459E-1f); -static _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201E-1f); - -#ifdef __VSX__ -static _EIGEN_DECLARE_CONST_Packet2d(1 , 1.0); -static _EIGEN_DECLARE_CONST_Packet2d(2 , 2.0); -static _EIGEN_DECLARE_CONST_Packet2d(half, 0.5); - -static _EIGEN_DECLARE_CONST_Packet2d(exp_hi, 709.437); -static _EIGEN_DECLARE_CONST_Packet2d(exp_lo, -709.436139303); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_LOG2EF, 1.4426950408889634073599); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p0, 1.26177193074810590878e-4); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p1, 3.02994407707441961300e-2); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p2, 9.99999999999999999910e-1); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q0, 3.00198505138664455042e-6); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q1, 2.52448340349684104192e-3); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q2, 2.27265548208155028766e-1); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q3, 2.00000000000000000009e0); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C1, 0.693145751953125); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C2, 1.42860682030941723212e-6); - -#ifdef __POWER8_VECTOR__ -static Packet2l p2l_1023 = { 1023, 1023 }; -static Packet2ul p2ul_52 = { 52, 52 }; -#endif - -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f plog(const Packet4f& _x) -{ - Packet4f x = _x; - - Packet4i emm0; - - /* isvalid_mask is 0 if x < 0 or x is NaN. */ - Packet4ui isvalid_mask = reinterpret_cast(vec_cmpge(x, p4f_ZERO)); - Packet4ui iszero_mask = reinterpret_cast(vec_cmpeq(x, p4f_ZERO)); - - x = pmax(x, p4f_min_norm_pos); /* cut off denormalized stuff */ - emm0 = vec_sr(reinterpret_cast(x), - reinterpret_cast(p4i_23)); - - /* keep only the fractional part */ - x = pand(x, p4f_inv_mant_mask); - x = por(x, p4f_half); - - emm0 = psub(emm0, p4i_0x7f); - Packet4f e = padd(vec_ctf(emm0, 0), p4f_1); - - /* part2: - if( x < SQRTHF ) { - e -= 1; - x = x + x - 1.0; - } else { x = x - 1.0; } - */ - Packet4f mask = reinterpret_cast(vec_cmplt(x, p4f_cephes_SQRTHF)); - Packet4f tmp = pand(x, mask); - x = psub(x, p4f_1); - e = psub(e, pand(p4f_1, mask)); - x = padd(x, tmp); - - Packet4f x2 = pmul(x,x); - Packet4f x3 = pmul(x2,x); - - Packet4f y, y1, y2; - y = pmadd(p4f_cephes_log_p0, x, p4f_cephes_log_p1); - y1 = pmadd(p4f_cephes_log_p3, x, p4f_cephes_log_p4); - y2 = pmadd(p4f_cephes_log_p6, x, p4f_cephes_log_p7); - y = pmadd(y , x, p4f_cephes_log_p2); - y1 = pmadd(y1, x, p4f_cephes_log_p5); - y2 = pmadd(y2, x, p4f_cephes_log_p8); - y = pmadd(y, x3, y1); - y = pmadd(y, x3, y2); - y = pmul(y, x3); - - y1 = pmul(e, p4f_cephes_log_q1); - tmp = pmul(x2, p4f_half); - y = padd(y, y1); - x = psub(x, tmp); - y2 = pmul(e, p4f_cephes_log_q2); - x = padd(x, y); - x = padd(x, y2); - // negative arg will be NAN, 0 will be -INF - x = vec_sel(x, p4f_minus_inf, iszero_mask); - x = vec_sel(p4f_minus_nan, x, isvalid_mask); - return x; -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pexp(const Packet4f& _x) -{ - Packet4f x = _x; - - Packet4f tmp, fx; - Packet4i emm0; - - // clamp x - x = pmax(pmin(x, p4f_exp_hi), p4f_exp_lo); - - // express exp(x) as exp(g + n*log(2)) - fx = pmadd(x, p4f_cephes_LOG2EF, p4f_half); - - fx = pfloor(fx); - - tmp = pmul(fx, p4f_cephes_exp_C1); - Packet4f z = pmul(fx, p4f_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - z = pmul(x,x); - - Packet4f y = p4f_cephes_exp_p0; - y = pmadd(y, x, p4f_cephes_exp_p1); - y = pmadd(y, x, p4f_cephes_exp_p2); - y = pmadd(y, x, p4f_cephes_exp_p3); - y = pmadd(y, x, p4f_cephes_exp_p4); - y = pmadd(y, x, p4f_cephes_exp_p5); - y = pmadd(y, z, x); - y = padd(y, p4f_1); - - // build 2^n - emm0 = vec_cts(fx, 0); - emm0 = vec_add(emm0, p4i_0x7f); - emm0 = vec_sl(emm0, reinterpret_cast(p4i_23)); - - // Altivec's max & min operators just drop silent NaNs. Check NaNs in - // inputs and return them unmodified. - Packet4ui isnumber_mask = reinterpret_cast(vec_cmpeq(_x, _x)); - return vec_sel(_x, pmax(pmul(y, reinterpret_cast(emm0)), _x), - isnumber_mask); -} - -#ifndef EIGEN_COMP_CLANG -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& x) -{ - return vec_rsqrt(x); -} -#endif - -#ifdef __VSX__ -#ifndef EIGEN_COMP_CLANG -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d prsqrt(const Packet2d& x) -{ - return vec_rsqrt(x); -} -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& x) -{ - return vec_sqrt(x); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d psqrt(const Packet2d& x) -{ - return vec_sqrt(x); -} - -// VSX support varies between different compilers and even different -// versions of the same compiler. For gcc version >= 4.9.3, we can use -// vec_cts to efficiently convert Packet2d to Packet2l. Otherwise, use -// a slow version that works with older compilers. -// Update: apparently vec_cts/vec_ctf intrinsics for 64-bit doubles -// are buggy, https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70963 -static inline Packet2l ConvertToPacket2l(const Packet2d& x) { -#if EIGEN_GNUC_AT_LEAST(5, 4) || \ - (EIGEN_GNUC_AT(6, 1) && __GNUC_PATCHLEVEL__ >= 1) - return vec_cts(x, 0); // TODO: check clang version. -#else - double tmp[2]; - memcpy(tmp, &x, sizeof(tmp)); - Packet2l l = { static_cast(tmp[0]), - static_cast(tmp[1]) }; - return l; -#endif -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d pexp(const Packet2d& _x) -{ - Packet2d x = _x; - - Packet2d tmp, fx; - Packet2l emm0; - - // clamp x - x = pmax(pmin(x, p2d_exp_hi), p2d_exp_lo); - - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(x, p2d_cephes_LOG2EF, p2d_half); - - fx = pfloor(fx); - - tmp = pmul(fx, p2d_cephes_exp_C1); - Packet2d z = pmul(fx, p2d_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - Packet2d x2 = pmul(x,x); - - Packet2d px = p2d_cephes_exp_p0; - px = pmadd(px, x2, p2d_cephes_exp_p1); - px = pmadd(px, x2, p2d_cephes_exp_p2); - px = pmul (px, x); - - Packet2d qx = p2d_cephes_exp_q0; - qx = pmadd(qx, x2, p2d_cephes_exp_q1); - qx = pmadd(qx, x2, p2d_cephes_exp_q2); - qx = pmadd(qx, x2, p2d_cephes_exp_q3); - - x = pdiv(px,psub(qx,px)); - x = pmadd(p2d_2,x,p2d_1); - - // build 2^n - emm0 = ConvertToPacket2l(fx); - -#ifdef __POWER8_VECTOR__ - emm0 = vec_add(emm0, p2l_1023); - emm0 = vec_sl(emm0, p2ul_52); -#else - // Code is a bit complex for POWER7. There is actually a - // vec_xxsldi intrinsic but it is not supported by some gcc versions. - // So we shift (52-32) bits and do a word swap with zeros. - _EIGEN_DECLARE_CONST_Packet4i(1023, 1023); - _EIGEN_DECLARE_CONST_Packet4i(20, 20); // 52 - 32 - - Packet4i emm04i = reinterpret_cast(emm0); - emm04i = vec_add(emm04i, p4i_1023); - emm04i = vec_sl(emm04i, reinterpret_cast(p4i_20)); - static const Packet16uc perm = { - 0x14, 0x15, 0x16, 0x17, 0x00, 0x01, 0x02, 0x03, - 0x1c, 0x1d, 0x1e, 0x1f, 0x08, 0x09, 0x0a, 0x0b }; -#ifdef _BIG_ENDIAN - emm0 = reinterpret_cast(vec_perm(p4i_ZERO, emm04i, perm)); -#else - emm0 = reinterpret_cast(vec_perm(emm04i, p4i_ZERO, perm)); -#endif - -#endif - - // Altivec's max & min operators just drop silent NaNs. Check NaNs in - // inputs and return them unmodified. - Packet2ul isnumber_mask = reinterpret_cast(vec_cmpeq(_x, _x)); - return vec_sel(_x, pmax(pmul(x, reinterpret_cast(emm0)), _x), - isnumber_mask); -} -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_ALTIVEC_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/PacketMath.h deleted file mode 100644 index 08a27d15302..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/AltiVec/PacketMath.h +++ /dev/null @@ -1,1061 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_ALTIVEC_H -#define EIGEN_PACKET_MATH_ALTIVEC_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 4 -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#endif - -// NOTE Altivec has 32 registers, but Eigen only accepts a value of 8 or 16 -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32 -#endif - -typedef __vector float Packet4f; -typedef __vector int Packet4i; -typedef __vector unsigned int Packet4ui; -typedef __vector __bool int Packet4bi; -typedef __vector short int Packet8i; -typedef __vector unsigned char Packet16uc; - -// We don't want to write the same code all the time, but we need to reuse the constants -// and it doesn't really work to declare them global, so we define macros instead - -#define _EIGEN_DECLARE_CONST_FAST_Packet4f(NAME,X) \ - Packet4f p4f_##NAME = reinterpret_cast(vec_splat_s32(X)) - -#define _EIGEN_DECLARE_CONST_FAST_Packet4i(NAME,X) \ - Packet4i p4i_##NAME = vec_splat_s32(X) - -#define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \ - Packet4f p4f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ - Packet4i p4i_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2d(NAME,X) \ - Packet2d p2d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2l(NAME,X) \ - Packet2l p2l_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \ - const Packet4f p4f_##NAME = reinterpret_cast(pset1(X)) - -#define DST_CHAN 1 -#define DST_CTRL(size, count, stride) (((size) << 24) | ((count) << 16) | (stride)) - - -// These constants are endian-agnostic -static _EIGEN_DECLARE_CONST_FAST_Packet4f(ZERO, 0); //{ 0.0, 0.0, 0.0, 0.0} -static _EIGEN_DECLARE_CONST_FAST_Packet4i(ZERO, 0); //{ 0, 0, 0, 0,} -static _EIGEN_DECLARE_CONST_FAST_Packet4i(ONE,1); //{ 1, 1, 1, 1} -static _EIGEN_DECLARE_CONST_FAST_Packet4i(MINUS16,-16); //{ -16, -16, -16, -16} -static _EIGEN_DECLARE_CONST_FAST_Packet4i(MINUS1,-1); //{ -1, -1, -1, -1} -static Packet4f p4f_MZERO = (Packet4f) vec_sl((Packet4ui)p4i_MINUS1, (Packet4ui)p4i_MINUS1); //{ 0x80000000, 0x80000000, 0x80000000, 0x80000000} -#ifndef __VSX__ -static Packet4f p4f_ONE = vec_ctf(p4i_ONE, 0); //{ 1.0, 1.0, 1.0, 1.0} -#endif - -static Packet4f p4f_COUNTDOWN = { 0.0, 1.0, 2.0, 3.0 }; -static Packet4i p4i_COUNTDOWN = { 0, 1, 2, 3 }; - -static Packet16uc p16uc_REVERSE32 = { 12,13,14,15, 8,9,10,11, 4,5,6,7, 0,1,2,3 }; -static Packet16uc p16uc_DUPLICATE32_HI = { 0,1,2,3, 0,1,2,3, 4,5,6,7, 4,5,6,7 }; - -// Mask alignment -#ifdef __PPC64__ -#define _EIGEN_MASK_ALIGNMENT 0xfffffffffffffff0 -#else -#define _EIGEN_MASK_ALIGNMENT 0xfffffff0 -#endif - -#define _EIGEN_ALIGNED_PTR(x) ((std::ptrdiff_t)(x) & _EIGEN_MASK_ALIGNMENT) - -// Handle endianness properly while loading constants -// Define global static constants: -#ifdef _BIG_ENDIAN -static Packet16uc p16uc_FORWARD = vec_lvsl(0, (float*)0); -#ifdef __VSX__ -static Packet16uc p16uc_REVERSE64 = { 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; -#endif -static Packet16uc p16uc_PSET32_WODD = vec_sld((Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 0), (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 2), 8);//{ 0,1,2,3, 0,1,2,3, 8,9,10,11, 8,9,10,11 }; -static Packet16uc p16uc_PSET32_WEVEN = vec_sld(p16uc_DUPLICATE32_HI, (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 3), 8);//{ 4,5,6,7, 4,5,6,7, 12,13,14,15, 12,13,14,15 }; -static Packet16uc p16uc_HALF64_0_16 = vec_sld((Packet16uc)p4i_ZERO, vec_splat((Packet16uc) vec_abs(p4i_MINUS16), 3), 8); //{ 0,0,0,0, 0,0,0,0, 16,16,16,16, 16,16,16,16}; -#else -static Packet16uc p16uc_FORWARD = p16uc_REVERSE32; -static Packet16uc p16uc_REVERSE64 = { 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; -static Packet16uc p16uc_PSET32_WODD = vec_sld((Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 1), (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 3), 8);//{ 0,1,2,3, 0,1,2,3, 8,9,10,11, 8,9,10,11 }; -static Packet16uc p16uc_PSET32_WEVEN = vec_sld((Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 0), (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 2), 8);//{ 4,5,6,7, 4,5,6,7, 12,13,14,15, 12,13,14,15 }; -static Packet16uc p16uc_HALF64_0_16 = vec_sld(vec_splat((Packet16uc) vec_abs(p4i_MINUS16), 0), (Packet16uc)p4i_ZERO, 8); //{ 0,0,0,0, 0,0,0,0, 16,16,16,16, 16,16,16,16}; -#endif // _BIG_ENDIAN - -static Packet16uc p16uc_PSET64_HI = (Packet16uc) vec_mergeh((Packet4ui)p16uc_PSET32_WODD, (Packet4ui)p16uc_PSET32_WEVEN); //{ 0,1,2,3, 4,5,6,7, 0,1,2,3, 4,5,6,7 }; -static Packet16uc p16uc_PSET64_LO = (Packet16uc) vec_mergel((Packet4ui)p16uc_PSET32_WODD, (Packet4ui)p16uc_PSET32_WEVEN); //{ 8,9,10,11, 12,13,14,15, 8,9,10,11, 12,13,14,15 }; -static Packet16uc p16uc_TRANSPOSE64_HI = p16uc_PSET64_HI + p16uc_HALF64_0_16; //{ 0,1,2,3, 4,5,6,7, 16,17,18,19, 20,21,22,23}; -static Packet16uc p16uc_TRANSPOSE64_LO = p16uc_PSET64_LO + p16uc_HALF64_0_16; //{ 8,9,10,11, 12,13,14,15, 24,25,26,27, 28,29,30,31}; - -static Packet16uc p16uc_COMPLEX32_REV = vec_sld(p16uc_REVERSE32, p16uc_REVERSE32, 8); //{ 4,5,6,7, 0,1,2,3, 12,13,14,15, 8,9,10,11 }; - -#ifdef _BIG_ENDIAN -static Packet16uc p16uc_COMPLEX32_REV2 = vec_sld(p16uc_FORWARD, p16uc_FORWARD, 8); //{ 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; -#else -static Packet16uc p16uc_COMPLEX32_REV2 = vec_sld(p16uc_PSET64_HI, p16uc_PSET64_LO, 8); //{ 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; -#endif // _BIG_ENDIAN - -#if EIGEN_HAS_BUILTIN(__builtin_prefetch) || EIGEN_COMP_GNUC - #define EIGEN_PPC_PREFETCH(ADDR) __builtin_prefetch(ADDR); -#else - #define EIGEN_PPC_PREFETCH(ADDR) asm( " dcbt [%[addr]]\n" :: [addr] "r" (ADDR) : "cc" ); -#endif - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4f type; - typedef Packet4f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 1, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasMin = 1, - HasMax = 1, - HasAbs = 1, - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 1, -#ifdef __VSX__ - HasSqrt = 1, -#if !EIGEN_COMP_CLANG - HasRsqrt = 1, -#else - HasRsqrt = 0, -#endif -#else - HasSqrt = 0, - HasRsqrt = 0, -#endif - HasRound = 1, - HasFloor = 1, - HasCeil = 1, - HasNegate = 1, - HasBlend = 1 - }; -}; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4i type; - typedef Packet4i half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 0, - HasBlend = 1 - }; -}; - - -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; -template<> struct unpacket_traits { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; - -inline std::ostream & operator <<(std::ostream & s, const Packet16uc & v) -{ - union { - Packet16uc v; - unsigned char n[16]; - } vt; - vt.v = v; - for (int i=0; i< 16; i++) - s << (int)vt.n[i] << ", "; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet4f & v) -{ - union { - Packet4f v; - float n[4]; - } vt; - vt.v = v; - s << vt.n[0] << ", " << vt.n[1] << ", " << vt.n[2] << ", " << vt.n[3]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet4i & v) -{ - union { - Packet4i v; - int n[4]; - } vt; - vt.v = v; - s << vt.n[0] << ", " << vt.n[1] << ", " << vt.n[2] << ", " << vt.n[3]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet4ui & v) -{ - union { - Packet4ui v; - unsigned int n[4]; - } vt; - vt.v = v; - s << vt.n[0] << ", " << vt.n[1] << ", " << vt.n[2] << ", " << vt.n[3]; - return s; -} - -// Need to define them first or we get specialization after instantiation errors -template<> EIGEN_STRONG_INLINE Packet4f pload(const float* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD -#ifdef __VSX__ - return vec_vsx_ld(0, from); -#else - return vec_ld(0, from); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4i pload(const int* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD -#ifdef __VSX__ - return vec_vsx_ld(0, from); -#else - return vec_ld(0, from); -#endif -} - -template<> EIGEN_STRONG_INLINE void pstore(float* to, const Packet4f& from) -{ - EIGEN_DEBUG_ALIGNED_STORE -#ifdef __VSX__ - vec_vsx_st(from, 0, to); -#else - vec_st(from, 0, to); -#endif -} - -template<> EIGEN_STRONG_INLINE void pstore(int* to, const Packet4i& from) -{ - EIGEN_DEBUG_ALIGNED_STORE -#ifdef __VSX__ - vec_vsx_st(from, 0, to); -#else - vec_st(from, 0, to); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { - Packet4f v = {from, from, from, from}; - return v; -} - -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) { - Packet4i v = {from, from, from, from}; - return v; -} -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const float *a, - Packet4f& a0, Packet4f& a1, Packet4f& a2, Packet4f& a3) -{ - a3 = pload(a); - a0 = vec_splat(a3, 0); - a1 = vec_splat(a3, 1); - a2 = vec_splat(a3, 2); - a3 = vec_splat(a3, 3); -} -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const int *a, - Packet4i& a0, Packet4i& a1, Packet4i& a2, Packet4i& a3) -{ - a3 = pload(a); - a0 = vec_splat(a3, 0); - a1 = vec_splat(a3, 1); - a2 = vec_splat(a3, 2); - a3 = vec_splat(a3, 3); -} - -template<> EIGEN_DEVICE_FUNC inline Packet4f pgather(const float* from, Index stride) -{ - float EIGEN_ALIGN16 af[4]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - af[2] = from[2*stride]; - af[3] = from[3*stride]; - return pload(af); -} -template<> EIGEN_DEVICE_FUNC inline Packet4i pgather(const int* from, Index stride) -{ - int EIGEN_ALIGN16 ai[4]; - ai[0] = from[0*stride]; - ai[1] = from[1*stride]; - ai[2] = from[2*stride]; - ai[3] = from[3*stride]; - return pload(ai); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet4f& from, Index stride) -{ - float EIGEN_ALIGN16 af[4]; - pstore(af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; - to[2*stride] = af[2]; - to[3*stride] = af[3]; -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(int* to, const Packet4i& from, Index stride) -{ - int EIGEN_ALIGN16 ai[4]; - pstore((int *)ai, from); - to[0*stride] = ai[0]; - to[1*stride] = ai[1]; - to[2*stride] = ai[2]; - to[3*stride] = ai[3]; -} - -template<> EIGEN_STRONG_INLINE Packet4f plset(const float& a) { return pset1(a) + p4f_COUNTDOWN; } -template<> EIGEN_STRONG_INLINE Packet4i plset(const int& a) { return pset1(a) + p4i_COUNTDOWN; } - -template<> EIGEN_STRONG_INLINE Packet4f padd(const Packet4f& a, const Packet4f& b) { return a + b; } -template<> EIGEN_STRONG_INLINE Packet4i padd(const Packet4i& a, const Packet4i& b) { return a + b; } - -template<> EIGEN_STRONG_INLINE Packet4f psub(const Packet4f& a, const Packet4f& b) { return a - b; } -template<> EIGEN_STRONG_INLINE Packet4i psub(const Packet4i& a, const Packet4i& b) { return a - b; } - -template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) { return p4f_ZERO - a; } -template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return p4i_ZERO - a; } - -template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4f pmul(const Packet4f& a, const Packet4f& b) { return vec_madd(a,b, p4f_MZERO); } -template<> EIGEN_STRONG_INLINE Packet4i pmul(const Packet4i& a, const Packet4i& b) { return a * b; } - -template<> EIGEN_STRONG_INLINE Packet4f pdiv(const Packet4f& a, const Packet4f& b) -{ -#ifndef __VSX__ // VSX actually provides a div instruction - Packet4f t, y_0, y_1; - - // Altivec does not offer a divide instruction, we have to do a reciprocal approximation - y_0 = vec_re(b); - - // Do one Newton-Raphson iteration to get the needed accuracy - t = vec_nmsub(y_0, b, p4f_ONE); - y_1 = vec_madd(y_0, t, y_0); - - return vec_madd(a, y_1, p4f_MZERO); -#else - return vec_div(a, b); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4i pdiv(const Packet4i& /*a*/, const Packet4i& /*b*/) -{ eigen_assert(false && "packet integer division are not supported by AltiVec"); - return pset1(0); -} - -// for some weird raisons, it has to be overloaded for packet of integers -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return vec_madd(a,b,c); } -template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return a*b + c; } - -template<> EIGEN_STRONG_INLINE Packet4f pmin(const Packet4f& a, const Packet4f& b) -{ - #ifdef __VSX__ - Packet4f ret; - __asm__ ("xvcmpgesp %x0,%x1,%x2\n\txxsel %x0,%x1,%x2,%x0" : "=&wa" (ret) : "wa" (a), "wa" (b)); - return ret; - #else - return vec_min(a, b); - #endif -} -template<> EIGEN_STRONG_INLINE Packet4i pmin(const Packet4i& a, const Packet4i& b) { return vec_min(a, b); } - -template<> EIGEN_STRONG_INLINE Packet4f pmax(const Packet4f& a, const Packet4f& b) -{ - #ifdef __VSX__ - Packet4f ret; - __asm__ ("xvcmpgtsp %x0,%x2,%x1\n\txxsel %x0,%x1,%x2,%x0" : "=&wa" (ret) : "wa" (a), "wa" (b)); - return ret; - #else - return vec_max(a, b); - #endif -} -template<> EIGEN_STRONG_INLINE Packet4i pmax(const Packet4i& a, const Packet4i& b) { return vec_max(a, b); } - -template<> EIGEN_STRONG_INLINE Packet4f pand(const Packet4f& a, const Packet4f& b) { return vec_and(a, b); } -template<> EIGEN_STRONG_INLINE Packet4i pand(const Packet4i& a, const Packet4i& b) { return vec_and(a, b); } - -template<> EIGEN_STRONG_INLINE Packet4f por(const Packet4f& a, const Packet4f& b) { return vec_or(a, b); } -template<> EIGEN_STRONG_INLINE Packet4i por(const Packet4i& a, const Packet4i& b) { return vec_or(a, b); } - -template<> EIGEN_STRONG_INLINE Packet4f pxor(const Packet4f& a, const Packet4f& b) { return vec_xor(a, b); } -template<> EIGEN_STRONG_INLINE Packet4i pxor(const Packet4i& a, const Packet4i& b) { return vec_xor(a, b); } - -template<> EIGEN_STRONG_INLINE Packet4f pandnot(const Packet4f& a, const Packet4f& b) { return vec_and(a, vec_nor(b, b)); } -template<> EIGEN_STRONG_INLINE Packet4i pandnot(const Packet4i& a, const Packet4i& b) { return vec_and(a, vec_nor(b, b)); } - -template<> EIGEN_STRONG_INLINE Packet4f pround(const Packet4f& a) { return vec_round(a); } -template<> EIGEN_STRONG_INLINE Packet4f pceil(const Packet4f& a) { return vec_ceil(a); } -template<> EIGEN_STRONG_INLINE Packet4f pfloor(const Packet4f& a) { return vec_floor(a); } - -#ifdef _BIG_ENDIAN -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD - Packet16uc MSQ, LSQ; - Packet16uc mask; - MSQ = vec_ld(0, (unsigned char *)from); // most significant quadword - LSQ = vec_ld(15, (unsigned char *)from); // least significant quadword - mask = vec_lvsl(0, from); // create the permute mask - return (Packet4f) vec_perm(MSQ, LSQ, mask); // align the data - -} -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD - // Taken from http://developer.apple.com/hardwaredrivers/ve/alignment.html - Packet16uc MSQ, LSQ; - Packet16uc mask; - MSQ = vec_ld(0, (unsigned char *)from); // most significant quadword - LSQ = vec_ld(15, (unsigned char *)from); // least significant quadword - mask = vec_lvsl(0, from); // create the permute mask - return (Packet4i) vec_perm(MSQ, LSQ, mask); // align the data -} -#else -// We also need ot redefine little endian loading of Packet4i/Packet4f using VSX -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return (Packet4i) vec_vsx_ld((long)from & 15, (const int*) _EIGEN_ALIGNED_PTR(from)); -} -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return (Packet4f) vec_vsx_ld((long)from & 15, (const float*) _EIGEN_ALIGNED_PTR(from)); -} -#endif - -template<> EIGEN_STRONG_INLINE Packet4f ploaddup(const float* from) -{ - Packet4f p; - if((std::ptrdiff_t(from) % 16) == 0) p = pload(from); - else p = ploadu(from); - return vec_perm(p, p, p16uc_DUPLICATE32_HI); -} -template<> EIGEN_STRONG_INLINE Packet4i ploaddup(const int* from) -{ - Packet4i p; - if((std::ptrdiff_t(from) % 16) == 0) p = pload(from); - else p = ploadu(from); - return vec_perm(p, p, p16uc_DUPLICATE32_HI); -} - -#ifdef _BIG_ENDIAN -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet4f& from) -{ - EIGEN_DEBUG_UNALIGNED_STORE - // Taken from http://developer.apple.com/hardwaredrivers/ve/alignment.html - // Warning: not thread safe! - Packet16uc MSQ, LSQ, edges; - Packet16uc edgeAlign, align; - - MSQ = vec_ld(0, (unsigned char *)to); // most significant quadword - LSQ = vec_ld(15, (unsigned char *)to); // least significant quadword - edgeAlign = vec_lvsl(0, to); // permute map to extract edges - edges=vec_perm(LSQ,MSQ,edgeAlign); // extract the edges - align = vec_lvsr( 0, to ); // permute map to misalign data - MSQ = vec_perm(edges,(Packet16uc)from,align); // misalign the data (MSQ) - LSQ = vec_perm((Packet16uc)from,edges,align); // misalign the data (LSQ) - vec_st( LSQ, 15, (unsigned char *)to ); // Store the LSQ part first - vec_st( MSQ, 0, (unsigned char *)to ); // Store the MSQ part -} -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet4i& from) -{ - EIGEN_DEBUG_UNALIGNED_STORE - // Taken from http://developer.apple.com/hardwaredrivers/ve/alignment.html - // Warning: not thread safe! - Packet16uc MSQ, LSQ, edges; - Packet16uc edgeAlign, align; - - MSQ = vec_ld(0, (unsigned char *)to); // most significant quadword - LSQ = vec_ld(15, (unsigned char *)to); // least significant quadword - edgeAlign = vec_lvsl(0, to); // permute map to extract edges - edges=vec_perm(LSQ, MSQ, edgeAlign); // extract the edges - align = vec_lvsr( 0, to ); // permute map to misalign data - MSQ = vec_perm(edges, (Packet16uc) from, align); // misalign the data (MSQ) - LSQ = vec_perm((Packet16uc) from, edges, align); // misalign the data (LSQ) - vec_st( LSQ, 15, (unsigned char *)to ); // Store the LSQ part first - vec_st( MSQ, 0, (unsigned char *)to ); // Store the MSQ part -} -#else -// We also need ot redefine little endian loading of Packet4i/Packet4f using VSX -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet4i& from) -{ - EIGEN_DEBUG_ALIGNED_STORE - vec_vsx_st(from, (long)to & 15, (int*) _EIGEN_ALIGNED_PTR(to)); -} -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet4f& from) -{ - EIGEN_DEBUG_ALIGNED_STORE - vec_vsx_st(from, (long)to & 15, (float*) _EIGEN_ALIGNED_PTR(to)); -} -#endif - -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { EIGEN_PPC_PREFETCH(addr); } -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { EIGEN_PPC_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { float EIGEN_ALIGN16 x; vec_ste(a, 0, &x); return x; } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int EIGEN_ALIGN16 x; vec_ste(a, 0, &x); return x; } - -template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) -{ - return reinterpret_cast(vec_perm(reinterpret_cast(a), reinterpret_cast(a), p16uc_REVERSE32)); -} -template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) -{ - return reinterpret_cast(vec_perm(reinterpret_cast(a), reinterpret_cast(a), p16uc_REVERSE32)); } - -template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) { return vec_abs(a); } -template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vec_abs(a); } - -template<> EIGEN_STRONG_INLINE float predux(const Packet4f& a) -{ - Packet4f b, sum; - b = vec_sld(a, a, 8); - sum = a + b; - b = vec_sld(sum, sum, 4); - sum += b; - return pfirst(sum); -} - -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - Packet4f v[4], sum[4]; - - // It's easier and faster to transpose then add as columns - // Check: http://www.freevec.org/function/matrix_4x4_transpose_floats for explanation - // Do the transpose, first set of moves - v[0] = vec_mergeh(vecs[0], vecs[2]); - v[1] = vec_mergel(vecs[0], vecs[2]); - v[2] = vec_mergeh(vecs[1], vecs[3]); - v[3] = vec_mergel(vecs[1], vecs[3]); - // Get the resulting vectors - sum[0] = vec_mergeh(v[0], v[2]); - sum[1] = vec_mergel(v[0], v[2]); - sum[2] = vec_mergeh(v[1], v[3]); - sum[3] = vec_mergel(v[1], v[3]); - - // Now do the summation: - // Lines 0+1 - sum[0] = sum[0] + sum[1]; - // Lines 2+3 - sum[1] = sum[2] + sum[3]; - // Add the results - sum[0] = sum[0] + sum[1]; - - return sum[0]; -} - -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i sum; - sum = vec_sums(a, p4i_ZERO); -#ifdef _BIG_ENDIAN - sum = vec_sld(sum, p4i_ZERO, 12); -#else - sum = vec_sld(p4i_ZERO, sum, 4); -#endif - return pfirst(sum); -} - -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - Packet4i v[4], sum[4]; - - // It's easier and faster to transpose then add as columns - // Check: http://www.freevec.org/function/matrix_4x4_transpose_floats for explanation - // Do the transpose, first set of moves - v[0] = vec_mergeh(vecs[0], vecs[2]); - v[1] = vec_mergel(vecs[0], vecs[2]); - v[2] = vec_mergeh(vecs[1], vecs[3]); - v[3] = vec_mergel(vecs[1], vecs[3]); - // Get the resulting vectors - sum[0] = vec_mergeh(v[0], v[2]); - sum[1] = vec_mergel(v[0], v[2]); - sum[2] = vec_mergeh(v[1], v[3]); - sum[3] = vec_mergel(v[1], v[3]); - - // Now do the summation: - // Lines 0+1 - sum[0] = sum[0] + sum[1]; - // Lines 2+3 - sum[1] = sum[2] + sum[3]; - // Add the results - sum[0] = sum[0] + sum[1]; - - return sum[0]; -} - -// Other reduction functions: -// mul -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet4f& a) -{ - Packet4f prod; - prod = pmul(a, vec_sld(a, a, 8)); - return pfirst(pmul(prod, vec_sld(prod, prod, 4))); -} - -template<> EIGEN_STRONG_INLINE int predux_mul(const Packet4i& a) -{ - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - return aux[0] * aux[1] * aux[2] * aux[3]; -} - -// min -template<> EIGEN_STRONG_INLINE float predux_min(const Packet4f& a) -{ - Packet4f b, res; - b = vec_min(a, vec_sld(a, a, 8)); - res = vec_min(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -template<> EIGEN_STRONG_INLINE int predux_min(const Packet4i& a) -{ - Packet4i b, res; - b = vec_min(a, vec_sld(a, a, 8)); - res = vec_min(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -// max -template<> EIGEN_STRONG_INLINE float predux_max(const Packet4f& a) -{ - Packet4f b, res; - b = vec_max(a, vec_sld(a, a, 8)); - res = vec_max(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -template<> EIGEN_STRONG_INLINE int predux_max(const Packet4i& a) -{ - Packet4i b, res; - b = vec_max(a, vec_sld(a, a, 8)); - res = vec_max(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { -#ifdef _BIG_ENDIAN - switch (Offset % 4) { - case 1: - first = vec_sld(first, second, 4); break; - case 2: - first = vec_sld(first, second, 8); break; - case 3: - first = vec_sld(first, second, 12); break; - } -#else - switch (Offset % 4) { - case 1: - first = vec_sld(second, first, 12); break; - case 2: - first = vec_sld(second, first, 8); break; - case 3: - first = vec_sld(second, first, 4); break; - } -#endif - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { -#ifdef _BIG_ENDIAN - switch (Offset % 4) { - case 1: - first = vec_sld(first, second, 4); break; - case 2: - first = vec_sld(first, second, 8); break; - case 3: - first = vec_sld(first, second, 12); break; - } -#else - switch (Offset % 4) { - case 1: - first = vec_sld(second, first, 12); break; - case 2: - first = vec_sld(second, first, 8); break; - case 3: - first = vec_sld(second, first, 4); break; - } -#endif - } -}; - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet4f t0, t1, t2, t3; - t0 = vec_mergeh(kernel.packet[0], kernel.packet[2]); - t1 = vec_mergel(kernel.packet[0], kernel.packet[2]); - t2 = vec_mergeh(kernel.packet[1], kernel.packet[3]); - t3 = vec_mergel(kernel.packet[1], kernel.packet[3]); - kernel.packet[0] = vec_mergeh(t0, t2); - kernel.packet[1] = vec_mergel(t0, t2); - kernel.packet[2] = vec_mergeh(t1, t3); - kernel.packet[3] = vec_mergel(t1, t3); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet4i t0, t1, t2, t3; - t0 = vec_mergeh(kernel.packet[0], kernel.packet[2]); - t1 = vec_mergel(kernel.packet[0], kernel.packet[2]); - t2 = vec_mergeh(kernel.packet[1], kernel.packet[3]); - t3 = vec_mergel(kernel.packet[1], kernel.packet[3]); - kernel.packet[0] = vec_mergeh(t0, t2); - kernel.packet[1] = vec_mergel(t0, t2); - kernel.packet[2] = vec_mergeh(t1, t3); - kernel.packet[3] = vec_mergel(t1, t3); -} - -template<> EIGEN_STRONG_INLINE Packet4i pblend(const Selector<4>& ifPacket, const Packet4i& thenPacket, const Packet4i& elsePacket) { - Packet4ui select = { ifPacket.select[0], ifPacket.select[1], ifPacket.select[2], ifPacket.select[3] }; - Packet4ui mask = reinterpret_cast(vec_cmpeq(reinterpret_cast(select), reinterpret_cast(p4i_ONE))); - return vec_sel(elsePacket, thenPacket, mask); -} - -template<> EIGEN_STRONG_INLINE Packet4f pblend(const Selector<4>& ifPacket, const Packet4f& thenPacket, const Packet4f& elsePacket) { - Packet4ui select = { ifPacket.select[0], ifPacket.select[1], ifPacket.select[2], ifPacket.select[3] }; - Packet4ui mask = reinterpret_cast(vec_cmpeq(reinterpret_cast(select), reinterpret_cast(p4i_ONE))); - return vec_sel(elsePacket, thenPacket, mask); -} - - -//---------- double ---------- -#ifdef __VSX__ -typedef __vector double Packet2d; -typedef __vector unsigned long long Packet2ul; -typedef __vector long long Packet2l; -#if EIGEN_COMP_CLANG -typedef Packet2ul Packet2bl; -#else -typedef __vector __bool long Packet2bl; -#endif - -static Packet2l p2l_ONE = { 1, 1 }; -static Packet2l p2l_ZERO = reinterpret_cast(p4i_ZERO); -static Packet2d p2d_ONE = { 1.0, 1.0 }; -static Packet2d p2d_ZERO = reinterpret_cast(p4f_ZERO); -static Packet2d p2d_MZERO = { -0.0, -0.0 }; - -#ifdef _BIG_ENDIAN -static Packet2d p2d_COUNTDOWN = reinterpret_cast(vec_sld(reinterpret_cast(p2d_ZERO), reinterpret_cast(p2d_ONE), 8)); -#else -static Packet2d p2d_COUNTDOWN = reinterpret_cast(vec_sld(reinterpret_cast(p2d_ONE), reinterpret_cast(p2d_ZERO), 8)); -#endif - -template Packet2d vec_splat_dbl(Packet2d& a); - -template<> EIGEN_STRONG_INLINE Packet2d vec_splat_dbl<0>(Packet2d& a) -{ - return reinterpret_cast(vec_perm(a, a, p16uc_PSET64_HI)); -} - -template<> EIGEN_STRONG_INLINE Packet2d vec_splat_dbl<1>(Packet2d& a) -{ - return reinterpret_cast(vec_perm(a, a, p16uc_PSET64_LO)); -} - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet2d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 1, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasMin = 1, - HasMax = 1, - HasAbs = 1, - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasRound = 1, - HasFloor = 1, - HasCeil = 1, - HasNegate = 1, - HasBlend = 1 - }; -}; - -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; - -inline std::ostream & operator <<(std::ostream & s, const Packet2l & v) -{ - union { - Packet2l v; - int64_t n[2]; - } vt; - vt.v = v; - s << vt.n[0] << ", " << vt.n[1]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet2d & v) -{ - union { - Packet2d v; - double n[2]; - } vt; - vt.v = v; - s << vt.n[0] << ", " << vt.n[1]; - return s; -} - -// Need to define them first or we get specialization after instantiation errors -template<> EIGEN_STRONG_INLINE Packet2d pload(const double* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD -#ifdef __VSX__ - return vec_vsx_ld(0, from); -#else - return vec_ld(0, from); -#endif -} - -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet2d& from) -{ - EIGEN_DEBUG_ALIGNED_STORE -#ifdef __VSX__ - vec_vsx_st(from, 0, to); -#else - vec_st(from, 0, to); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { - Packet2d v = {from, from}; - return v; -} - -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const double *a, - Packet2d& a0, Packet2d& a1, Packet2d& a2, Packet2d& a3) -{ - a1 = pload(a); - a0 = vec_splat_dbl<0>(a1); - a1 = vec_splat_dbl<1>(a1); - a3 = pload(a+2); - a2 = vec_splat_dbl<0>(a3); - a3 = vec_splat_dbl<1>(a3); -} - -template<> EIGEN_DEVICE_FUNC inline Packet2d pgather(const double* from, Index stride) -{ - double EIGEN_ALIGN16 af[2]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - return pload(af); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet2d& from, Index stride) -{ - double EIGEN_ALIGN16 af[2]; - pstore(af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; -} - -template<> EIGEN_STRONG_INLINE Packet2d plset(const double& a) { return pset1(a) + p2d_COUNTDOWN; } - -template<> EIGEN_STRONG_INLINE Packet2d padd(const Packet2d& a, const Packet2d& b) { return a + b; } - -template<> EIGEN_STRONG_INLINE Packet2d psub(const Packet2d& a, const Packet2d& b) { return a - b; } - -template<> EIGEN_STRONG_INLINE Packet2d pnegate(const Packet2d& a) { return p2d_ZERO - a; } - -template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet2d pmul(const Packet2d& a, const Packet2d& b) { return vec_madd(a,b,p2d_MZERO); } -template<> EIGEN_STRONG_INLINE Packet2d pdiv(const Packet2d& a, const Packet2d& b) { return vec_div(a,b); } - -// for some weird raisons, it has to be overloaded for packet of integers -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return vec_madd(a, b, c); } - -template<> EIGEN_STRONG_INLINE Packet2d pmin(const Packet2d& a, const Packet2d& b) -{ - Packet2d ret; - __asm__ ("xvcmpgedp %x0,%x1,%x2\n\txxsel %x0,%x1,%x2,%x0" : "=&wa" (ret) : "wa" (a), "wa" (b)); - return ret; - } - -template<> EIGEN_STRONG_INLINE Packet2d pmax(const Packet2d& a, const Packet2d& b) -{ - Packet2d ret; - __asm__ ("xvcmpgtdp %x0,%x2,%x1\n\txxsel %x0,%x1,%x2,%x0" : "=&wa" (ret) : "wa" (a), "wa" (b)); - return ret; -} - -template<> EIGEN_STRONG_INLINE Packet2d pand(const Packet2d& a, const Packet2d& b) { return vec_and(a, b); } - -template<> EIGEN_STRONG_INLINE Packet2d por(const Packet2d& a, const Packet2d& b) { return vec_or(a, b); } - -template<> EIGEN_STRONG_INLINE Packet2d pxor(const Packet2d& a, const Packet2d& b) { return vec_xor(a, b); } - -template<> EIGEN_STRONG_INLINE Packet2d pandnot(const Packet2d& a, const Packet2d& b) { return vec_and(a, vec_nor(b, b)); } - -template<> EIGEN_STRONG_INLINE Packet2d pround(const Packet2d& a) { return vec_round(a); } -template<> EIGEN_STRONG_INLINE Packet2d pceil(const Packet2d& a) { return vec_ceil(a); } -template<> EIGEN_STRONG_INLINE Packet2d pfloor(const Packet2d& a) { return vec_floor(a); } - -template<> EIGEN_STRONG_INLINE Packet2d ploadu(const double* from) -{ - EIGEN_DEBUG_ALIGNED_LOAD - return (Packet2d) vec_vsx_ld((long)from & 15, (const double*) _EIGEN_ALIGNED_PTR(from)); -} - -template<> EIGEN_STRONG_INLINE Packet2d ploaddup(const double* from) -{ - Packet2d p; - if((std::ptrdiff_t(from) % 16) == 0) p = pload(from); - else p = ploadu(from); - return vec_splat_dbl<0>(p); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet2d& from) -{ - EIGEN_DEBUG_ALIGNED_STORE - vec_vsx_st((Packet4f)from, (long)to & 15, (float*) _EIGEN_ALIGNED_PTR(to)); -} - -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { EIGEN_PPC_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { double EIGEN_ALIGN16 x[2]; pstore(x, a); return x[0]; } - -template<> EIGEN_STRONG_INLINE Packet2d preverse(const Packet2d& a) -{ - return reinterpret_cast(vec_perm(reinterpret_cast(a), reinterpret_cast(a), p16uc_REVERSE64)); -} -template<> EIGEN_STRONG_INLINE Packet2d pabs(const Packet2d& a) { return vec_abs(a); } - -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) -{ - Packet2d b, sum; - b = reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)); - sum = a + b; - return pfirst(sum); -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - Packet2d v[2], sum; - v[0] = vecs[0] + reinterpret_cast(vec_sld(reinterpret_cast(vecs[0]), reinterpret_cast(vecs[0]), 8)); - v[1] = vecs[1] + reinterpret_cast(vec_sld(reinterpret_cast(vecs[1]), reinterpret_cast(vecs[1]), 8)); - -#ifdef _BIG_ENDIAN - sum = reinterpret_cast(vec_sld(reinterpret_cast(v[0]), reinterpret_cast(v[1]), 8)); -#else - sum = reinterpret_cast(vec_sld(reinterpret_cast(v[1]), reinterpret_cast(v[0]), 8)); -#endif - - return sum; -} -// Other reduction functions: -// mul -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) -{ - return pfirst(pmul(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -// min -template<> EIGEN_STRONG_INLINE double predux_min(const Packet2d& a) -{ - return pfirst(pmin(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -// max -template<> EIGEN_STRONG_INLINE double predux_max(const Packet2d& a) -{ - return pfirst(pmax(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset == 1) -#ifdef _BIG_ENDIAN - first = reinterpret_cast(vec_sld(reinterpret_cast(first), reinterpret_cast(second), 8)); -#else - first = reinterpret_cast(vec_sld(reinterpret_cast(second), reinterpret_cast(first), 8)); -#endif - } -}; - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet2d t0, t1; - t0 = vec_perm(kernel.packet[0], kernel.packet[1], p16uc_TRANSPOSE64_HI); - t1 = vec_perm(kernel.packet[0], kernel.packet[1], p16uc_TRANSPOSE64_LO); - kernel.packet[0] = t0; - kernel.packet[1] = t1; -} - -template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, const Packet2d& thenPacket, const Packet2d& elsePacket) { - Packet2l select = { ifPacket.select[0], ifPacket.select[1] }; - Packet2bl mask = reinterpret_cast( vec_cmpeq(reinterpret_cast(select), reinterpret_cast(p2l_ONE)) ); - return vec_sel(elsePacket, thenPacket, mask); -} -#endif // __VSX__ -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PACKET_MATH_ALTIVEC_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h deleted file mode 100644 index 9c25365090b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h +++ /dev/null @@ -1,103 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_CUDA_H -#define EIGEN_COMPLEX_CUDA_H - -// clang-format off - -namespace Eigen { - -namespace internal { - -#if defined(__CUDACC__) && defined(EIGEN_USE_GPU) - -// Many std::complex methods such as operator+, operator-, operator* and -// operator/ are not constexpr. Due to this, clang does not treat them as device -// functions and thus Eigen functors making use of these operators fail to -// compile. Here, we manually specialize these functors for complex types when -// building for CUDA to avoid non-constexpr methods. - -// Sum -template struct scalar_sum_op, const std::complex > : binary_op_base, const std::complex > { - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_sum_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - return std::complex(numext::real(a) + numext::real(b), - numext::imag(a) + numext::imag(b)); - } -}; - -template struct scalar_sum_op, std::complex > : scalar_sum_op, const std::complex > {}; - - -// Difference -template struct scalar_difference_op, const std::complex > : binary_op_base, const std::complex > { - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_difference_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - return std::complex(numext::real(a) - numext::real(b), - numext::imag(a) - numext::imag(b)); - } -}; - -template struct scalar_difference_op, std::complex > : scalar_difference_op, const std::complex > {}; - - -// Product -template struct scalar_product_op, const std::complex > : binary_op_base, const std::complex > { - enum { - Vectorizable = packet_traits>::HasMul - }; - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_product_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - const T a_real = numext::real(a); - const T a_imag = numext::imag(a); - const T b_real = numext::real(b); - const T b_imag = numext::imag(b); - return std::complex(a_real * b_real - a_imag * b_imag, - a_real * b_imag + a_imag * b_real); - } -}; - -template struct scalar_product_op, std::complex > : scalar_product_op, const std::complex > {}; - - -// Quotient -template struct scalar_quotient_op, const std::complex > : binary_op_base, const std::complex > { - enum { - Vectorizable = packet_traits>::HasDiv - }; - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - const T a_real = numext::real(a); - const T a_imag = numext::imag(a); - const T b_real = numext::real(b); - const T b_imag = numext::imag(b); - const T norm = T(1) / (b_real * b_real + b_imag * b_imag); - return std::complex((a_real * b_real + a_imag * b_imag) * norm, - (a_imag * b_real - a_real * b_imag) * norm); - } -}; - -template struct scalar_quotient_op, std::complex > : scalar_quotient_op, const std::complex > {}; - -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h deleted file mode 100644 index 59717b4fe6c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h +++ /dev/null @@ -1,675 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -// -// The conversion routines are Copyright (c) Fabian Giesen, 2016. -// The original license follows: -// -// Copyright (c) Fabian Giesen, 2016 -// All rights reserved. -// Redistribution and use in source and binary forms, with or without -// modification, are permitted. -// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS -// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT -// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR -// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT -// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, -// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT -// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, -// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY -// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT -// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE -// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - -// Standard 16-bit float type, mostly useful for GPUs. Defines a new -// type Eigen::half (inheriting from CUDA's __half struct) with -// operator overloads such that it behaves basically as an arithmetic -// type. It will be quite slow on CPUs (so it is recommended to stay -// in float32_bits for CPUs, except for simple parameter conversions, I/O -// to disk and the likes), but fast on GPUs. - - -#ifndef EIGEN_HALF_CUDA_H -#define EIGEN_HALF_CUDA_H - -#if __cplusplus > 199711L -#define EIGEN_EXPLICIT_CAST(tgt_type) explicit operator tgt_type() -#else -#define EIGEN_EXPLICIT_CAST(tgt_type) operator tgt_type() -#endif - -#include - -namespace Eigen { - -struct half; - -namespace half_impl { - -#if !defined(EIGEN_HAS_CUDA_FP16) -// Make our own __half_raw definition that is similar to CUDA's. -struct __half_raw { - EIGEN_DEVICE_FUNC __half_raw() : x(0) {} - explicit EIGEN_DEVICE_FUNC __half_raw(unsigned short raw) : x(raw) {} - unsigned short x; -}; -#elif defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER < 90000 -// In CUDA < 9.0, __half is the equivalent of CUDA 9's __half_raw -typedef __half __half_raw; -#endif - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw raw_uint16_to_half(unsigned short x); -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw float_to_half_rtne(float ff); -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC float half_to_float(__half_raw h); - -struct half_base : public __half_raw { - EIGEN_DEVICE_FUNC half_base() {} - EIGEN_DEVICE_FUNC half_base(const half_base& h) : __half_raw(h) {} - EIGEN_DEVICE_FUNC half_base(const __half_raw& h) : __half_raw(h) {} -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER >= 90000 - EIGEN_DEVICE_FUNC half_base(const __half& h) : __half_raw(*(__half_raw*)&h) {} -#endif -}; - -} // namespace half_impl - -// Class definition. -struct half : public half_impl::half_base { - #if !defined(EIGEN_HAS_CUDA_FP16) || (defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER < 90000) - typedef half_impl::__half_raw __half_raw; - #endif - - EIGEN_DEVICE_FUNC half() {} - - EIGEN_DEVICE_FUNC half(const __half_raw& h) : half_impl::half_base(h) {} - EIGEN_DEVICE_FUNC half(const half& h) : half_impl::half_base(h) {} -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER >= 90000 - EIGEN_DEVICE_FUNC half(const __half& h) : half_impl::half_base(h) {} -#endif - - explicit EIGEN_DEVICE_FUNC half(bool b) - : half_impl::half_base(half_impl::raw_uint16_to_half(b ? 0x3c00 : 0)) {} - template - explicit EIGEN_DEVICE_FUNC half(const T& val) - : half_impl::half_base(half_impl::float_to_half_rtne(static_cast(val))) {} - explicit EIGEN_DEVICE_FUNC half(float f) - : half_impl::half_base(half_impl::float_to_half_rtne(f)) {} - - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(bool) const { - // +0.0 and -0.0 become false, everything else becomes true. - return (x & 0x7fff) != 0; - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(signed char) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned char) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(short) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned short) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(int) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned int) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(long long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned long long) const { - return static_cast(half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(float) const { - return half_impl::half_to_float(*this); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(double) const { - return static_cast(half_impl::half_to_float(*this)); - } - - EIGEN_DEVICE_FUNC half& operator=(const half& other) { - x = other.x; - return *this; - } -}; - -} // end namespace Eigen - -namespace std { -template<> -struct numeric_limits { - static const bool is_specialized = true; - static const bool is_signed = true; - static const bool is_integer = false; - static const bool is_exact = false; - static const bool has_infinity = true; - static const bool has_quiet_NaN = true; - static const bool has_signaling_NaN = true; - static const float_denorm_style has_denorm = denorm_present; - static const bool has_denorm_loss = false; - static const std::float_round_style round_style = std::round_to_nearest; - static const bool is_iec559 = false; - static const bool is_bounded = false; - static const bool is_modulo = false; - static const int digits = 11; - static const int digits10 = 3; // according to http://half.sourceforge.net/structstd_1_1numeric__limits_3_01half__float_1_1half_01_4.html - static const int max_digits10 = 5; // according to http://half.sourceforge.net/structstd_1_1numeric__limits_3_01half__float_1_1half_01_4.html - static const int radix = 2; - static const int min_exponent = -13; - static const int min_exponent10 = -4; - static const int max_exponent = 16; - static const int max_exponent10 = 4; - static const bool traps = true; - static const bool tinyness_before = false; - - static Eigen::half (min)() { return Eigen::half_impl::raw_uint16_to_half(0x400); } - static Eigen::half lowest() { return Eigen::half_impl::raw_uint16_to_half(0xfbff); } - static Eigen::half (max)() { return Eigen::half_impl::raw_uint16_to_half(0x7bff); } - static Eigen::half epsilon() { return Eigen::half_impl::raw_uint16_to_half(0x0800); } - static Eigen::half round_error() { return Eigen::half(0.5); } - static Eigen::half infinity() { return Eigen::half_impl::raw_uint16_to_half(0x7c00); } - static Eigen::half quiet_NaN() { return Eigen::half_impl::raw_uint16_to_half(0x7e00); } - static Eigen::half signaling_NaN() { return Eigen::half_impl::raw_uint16_to_half(0x7e00); } - static Eigen::half denorm_min() { return Eigen::half_impl::raw_uint16_to_half(0x1); } -}; - -// If std::numeric_limits is specialized, should also specialize -// std::numeric_limits, std::numeric_limits, and -// std::numeric_limits -// https://stackoverflow.com/a/16519653/ -template<> -struct numeric_limits : numeric_limits {}; -template<> -struct numeric_limits : numeric_limits {}; -template<> -struct numeric_limits : numeric_limits {}; -} // end namespace std - -namespace Eigen { - -namespace half_impl { - -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - -// Intrinsics for native fp16 support. Note that on current hardware, -// these are no faster than float32_bits arithmetic (you need to use the half2 -// versions to get the ALU speed increased), but you do save the -// conversion steps back and forth. - -EIGEN_STRONG_INLINE __device__ half operator + (const half& a, const half& b) { - return __hadd(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator * (const half& a, const half& b) { - return __hmul(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator - (const half& a, const half& b) { - return __hsub(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator / (const half& a, const half& b) { - float num = __half2float(a); - float denom = __half2float(b); - return __float2half(num / denom); -} -EIGEN_STRONG_INLINE __device__ half operator - (const half& a) { - return __hneg(a); -} -EIGEN_STRONG_INLINE __device__ half& operator += (half& a, const half& b) { - a = a + b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator *= (half& a, const half& b) { - a = a * b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator -= (half& a, const half& b) { - a = a - b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator /= (half& a, const half& b) { - a = a / b; - return a; -} -EIGEN_STRONG_INLINE __device__ bool operator == (const half& a, const half& b) { - return __heq(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator != (const half& a, const half& b) { - return __hne(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator < (const half& a, const half& b) { - return __hlt(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator <= (const half& a, const half& b) { - return __hle(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator > (const half& a, const half& b) { - return __hgt(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator >= (const half& a, const half& b) { - return __hge(a, b); -} - -#else // Emulate support for half floats - -// Definitions for CPUs and older CUDA, mostly working through conversion -// to/from float32_bits. - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator + (const half& a, const half& b) { - return half(float(a) + float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator * (const half& a, const half& b) { - return half(float(a) * float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator - (const half& a, const half& b) { - return half(float(a) - float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator / (const half& a, const half& b) { - return half(float(a) / float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator - (const half& a) { - half result; - result.x = a.x ^ 0x8000; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator += (half& a, const half& b) { - a = half(float(a) + float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator *= (half& a, const half& b) { - a = half(float(a) * float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator -= (half& a, const half& b) { - a = half(float(a) - float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator /= (half& a, const half& b) { - a = half(float(a) / float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator == (const half& a, const half& b) { - return numext::equal_strict(float(a),float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator != (const half& a, const half& b) { - return numext::not_equal_strict(float(a), float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator < (const half& a, const half& b) { - return float(a) < float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator <= (const half& a, const half& b) { - return float(a) <= float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator > (const half& a, const half& b) { - return float(a) > float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator >= (const half& a, const half& b) { - return float(a) >= float(b); -} - -#endif // Emulate support for half floats - -// Division by an index. Do it in full float precision to avoid accuracy -// issues in converting the denominator to half. -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator / (const half& a, Index b) { - return half(static_cast(a) / static_cast(b)); -} - -// Conversion routines, including fallbacks for the host or older CUDA. -// Note that newer Intel CPUs (Haswell or newer) have vectorized versions of -// these in hardware. If we need more performance on older/other CPUs, they are -// also possible to vectorize directly. - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw raw_uint16_to_half(unsigned short x) { - __half_raw h; - h.x = x; - return h; -} - -union float32_bits { - unsigned int u; - float f; -}; - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw float_to_half_rtne(float ff) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 - __half tmp_ff = __float2half(ff); - return *(__half_raw*)&tmp_ff; - -#elif defined(EIGEN_HAS_FP16_C) - __half_raw h; - h.x = _cvtss_sh(ff, 0); - return h; - -#else - float32_bits f; f.f = ff; - - const float32_bits f32infty = { 255 << 23 }; - const float32_bits f16max = { (127 + 16) << 23 }; - const float32_bits denorm_magic = { ((127 - 15) + (23 - 10) + 1) << 23 }; - unsigned int sign_mask = 0x80000000u; - __half_raw o; - o.x = static_cast(0x0u); - - unsigned int sign = f.u & sign_mask; - f.u ^= sign; - - // NOTE all the integer compares in this function can be safely - // compiled into signed compares since all operands are below - // 0x80000000. Important if you want fast straight SSE2 code - // (since there's no unsigned PCMPGTD). - - if (f.u >= f16max.u) { // result is Inf or NaN (all exponent bits set) - o.x = (f.u > f32infty.u) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf - } else { // (De)normalized number or zero - if (f.u < (113 << 23)) { // resulting FP16 is subnormal or zero - // use a magic value to align our 10 mantissa bits at the bottom of - // the float. as long as FP addition is round-to-nearest-even this - // just works. - f.f += denorm_magic.f; - - // and one integer subtract of the bias later, we have our final float! - o.x = static_cast(f.u - denorm_magic.u); - } else { - unsigned int mant_odd = (f.u >> 13) & 1; // resulting mantissa is odd - - // update exponent, rounding bias part 1 - f.u += ((unsigned int)(15 - 127) << 23) + 0xfff; - // rounding bias part 2 - f.u += mant_odd; - // take the bits! - o.x = static_cast(f.u >> 13); - } - } - - o.x |= static_cast(sign >> 16); - return o; -#endif -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC float half_to_float(__half_raw h) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 - return __half2float(h); - -#elif defined(EIGEN_HAS_FP16_C) - return _cvtsh_ss(h.x); - -#else - const float32_bits magic = { 113 << 23 }; - const unsigned int shifted_exp = 0x7c00 << 13; // exponent mask after shift - float32_bits o; - - o.u = (h.x & 0x7fff) << 13; // exponent/mantissa bits - unsigned int exp = shifted_exp & o.u; // just the exponent - o.u += (127 - 15) << 23; // exponent adjust - - // handle exponent special cases - if (exp == shifted_exp) { // Inf/NaN? - o.u += (128 - 16) << 23; // extra exp adjust - } else if (exp == 0) { // Zero/Denormal? - o.u += 1 << 23; // extra exp adjust - o.f -= magic.f; // renormalize - } - - o.u |= (h.x & 0x8000) << 16; // sign bit - return o.f; -#endif -} - -// --- standard functions --- - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isinf)(const half& a) { - return (a.x & 0x7fff) == 0x7c00; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isnan)(const half& a) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hisnan(a); -#else - return (a.x & 0x7fff) > 0x7c00; -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isfinite)(const half& a) { - return !(isinf EIGEN_NOT_A_MACRO (a)) && !(isnan EIGEN_NOT_A_MACRO (a)); -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half abs(const half& a) { - half result; - result.x = a.x & 0x7FFF; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half exp(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - return half(hexp(a)); -#else - return half(::expf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log(const half& a) { -#if defined(EIGEN_HAS_CUDA_FP16) && EIGEN_CUDACC_VER >= 80000 && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return half(::hlog(a)); -#else - return half(::logf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log1p(const half& a) { - return half(numext::log1p(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log10(const half& a) { - return half(::log10f(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half sqrt(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - return half(hsqrt(a)); -#else - return half(::sqrtf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half pow(const half& a, const half& b) { - return half(::powf(float(a), float(b))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half sin(const half& a) { - return half(::sinf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half cos(const half& a) { - return half(::cosf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half tan(const half& a) { - return half(::tanf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half tanh(const half& a) { - return half(::tanhf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half floor(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 300 - return half(hfloor(a)); -#else - return half(::floorf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half ceil(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 300 - return half(hceil(a)); -#else - return half(::ceilf(float(a))); -#endif -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half (min)(const half& a, const half& b) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hlt(b, a) ? b : a; -#else - const float f1 = static_cast(a); - const float f2 = static_cast(b); - return f2 < f1 ? b : a; -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half (max)(const half& a, const half& b) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hlt(a, b) ? b : a; -#else - const float f1 = static_cast(a); - const float f2 = static_cast(b); - return f1 < f2 ? b : a; -#endif -} - -EIGEN_ALWAYS_INLINE std::ostream& operator << (std::ostream& os, const half& v) { - os << static_cast(v); - return os; -} - -} // end namespace half_impl - -// import Eigen::half_impl::half into Eigen namespace -// using half_impl::half; - -namespace internal { - -template<> -struct random_default_impl -{ - static inline half run(const half& x, const half& y) - { - return x + (y-x) * half(float(std::rand()) / float(RAND_MAX)); - } - static inline half run() - { - return run(half(-1.f), half(1.f)); - } -}; - -template<> struct is_arithmetic { enum { value = true }; }; - -} // end namespace internal - -template<> struct NumTraits - : GenericNumTraits -{ - enum { - IsSigned = true, - IsInteger = false, - IsComplex = false, - RequireInitialization = false - }; - - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half epsilon() { - return half_impl::raw_uint16_to_half(0x0800); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half dummy_precision() { return Eigen::half(1e-2f); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half highest() { - return half_impl::raw_uint16_to_half(0x7bff); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half lowest() { - return half_impl::raw_uint16_to_half(0xfbff); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half infinity() { - return half_impl::raw_uint16_to_half(0x7c00); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half quiet_NaN() { - return half_impl::raw_uint16_to_half(0x7c01); - } -}; - -} // end namespace Eigen - -// C-like standard mathematical functions and trancendentals. -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half fabsh(const Eigen::half& a) { - Eigen::half result; - result.x = a.x & 0x7FFF; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half exph(const Eigen::half& a) { - return Eigen::half(::expf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half logh(const Eigen::half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return Eigen::half(::hlog(a)); -#else - return Eigen::half(::logf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half sqrth(const Eigen::half& a) { - return Eigen::half(::sqrtf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half powh(const Eigen::half& a, const Eigen::half& b) { - return Eigen::half(::powf(float(a), float(b))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half floorh(const Eigen::half& a) { - return Eigen::half(::floorf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half ceilh(const Eigen::half& a) { - return Eigen::half(::ceilf(float(a))); -} - -namespace std { - -#if __cplusplus > 199711L -template <> -struct hash { - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::size_t operator()(const Eigen::half& a) const { - return static_cast(a.x); - } -}; -#endif - -} // end namespace std - - -// Add the missing shfl_xor intrinsic -#if defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 -__device__ EIGEN_STRONG_INLINE Eigen::half __shfl_xor(Eigen::half var, int laneMask, int width=warpSize) { - #if EIGEN_CUDACC_VER < 90000 - return static_cast(__shfl_xor(static_cast(var), laneMask, width)); - #else - return static_cast(__shfl_xor_sync(0xFFFFFFFF, static_cast(var), laneMask, width)); - #endif -} -#endif - -// ldg() has an overload for __half_raw, but we also need one for Eigen::half. -#if defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 350 -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half __ldg(const Eigen::half* ptr) { - return Eigen::half_impl::raw_uint16_to_half( - __ldg(reinterpret_cast(ptr))); -} -#endif - - -#if defined(EIGEN_CUDA_ARCH) -namespace Eigen { -namespace numext { - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isnan)(const Eigen::half& h) { - return (half_impl::isnan)(h); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isinf)(const Eigen::half& h) { - return (half_impl::isinf)(h); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isfinite)(const Eigen::half& h) { - return (half_impl::isfinite)(h); -} - -} // namespace Eigen -} // namespace numext -#endif - -#endif // EIGEN_HALF_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/MathFunctions.h deleted file mode 100644 index 0348b41db05..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/MathFunctions.h +++ /dev/null @@ -1,91 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MATH_FUNCTIONS_CUDA_H -#define EIGEN_MATH_FUNCTIONS_CUDA_H - -namespace Eigen { - -namespace internal { - -// Make sure this is only available when targeting a GPU: we don't want to -// introduce conflicts between these packet_traits definitions and the ones -// we'll use on the host side (SSE, AVX, ...) -#if defined(__CUDACC__) && defined(EIGEN_USE_GPU) -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -float4 plog(const float4& a) -{ - return make_float4(logf(a.x), logf(a.y), logf(a.z), logf(a.w)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -double2 plog(const double2& a) -{ - using ::log; - return make_double2(log(a.x), log(a.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -float4 plog1p(const float4& a) -{ - return make_float4(log1pf(a.x), log1pf(a.y), log1pf(a.z), log1pf(a.w)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -double2 plog1p(const double2& a) -{ - return make_double2(log1p(a.x), log1p(a.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -float4 pexp(const float4& a) -{ - return make_float4(expf(a.x), expf(a.y), expf(a.z), expf(a.w)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -double2 pexp(const double2& a) -{ - using ::exp; - return make_double2(exp(a.x), exp(a.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -float4 psqrt(const float4& a) -{ - return make_float4(sqrtf(a.x), sqrtf(a.y), sqrtf(a.z), sqrtf(a.w)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -double2 psqrt(const double2& a) -{ - using ::sqrt; - return make_double2(sqrt(a.x), sqrt(a.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -float4 prsqrt(const float4& a) -{ - return make_float4(rsqrtf(a.x), rsqrtf(a.y), rsqrtf(a.z), rsqrtf(a.w)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -double2 prsqrt(const double2& a) -{ - return make_double2(rsqrt(a.x), rsqrt(a.y)); -} - - -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMath.h deleted file mode 100644 index 4dda63188d2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMath.h +++ /dev/null @@ -1,333 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_CUDA_H -#define EIGEN_PACKET_MATH_CUDA_H - -namespace Eigen { - -namespace internal { - -// Make sure this is only available when targeting a GPU: we don't want to -// introduce conflicts between these packet_traits definitions and the ones -// we'll use on the host side (SSE, AVX, ...) -#if defined(__CUDACC__) && defined(EIGEN_USE_GPU) -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; - -template<> struct packet_traits : default_packet_traits -{ - typedef float4 type; - typedef float4 half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 0, - - HasDiv = 1, - HasSin = 0, - HasCos = 0, - HasLog = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasLGamma = 1, - HasDiGamma = 1, - HasZeta = 1, - HasPolygamma = 1, - HasErf = 1, - HasErfc = 1, - HasIGamma = 1, - HasIGammac = 1, - HasBetaInc = 1, - - HasBlend = 0, - }; -}; - -template<> struct packet_traits : default_packet_traits -{ - typedef double2 type; - typedef double2 half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 0, - - HasDiv = 1, - HasLog = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasLGamma = 1, - HasDiGamma = 1, - HasZeta = 1, - HasPolygamma = 1, - HasErf = 1, - HasErfc = 1, - HasIGamma = 1, - HasIGammac = 1, - HasBetaInc = 1, - - HasBlend = 0, - }; -}; - - -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef float4 half; }; -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef double2 half; }; - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pset1(const float& from) { - return make_float4(from, from, from, from); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pset1(const double& from) { - return make_double2(from, from); -} - - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 plset(const float& a) { - return make_float4(a, a+1, a+2, a+3); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 plset(const double& a) { - return make_double2(a, a+1); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 padd(const float4& a, const float4& b) { - return make_float4(a.x+b.x, a.y+b.y, a.z+b.z, a.w+b.w); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 padd(const double2& a, const double2& b) { - return make_double2(a.x+b.x, a.y+b.y); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 psub(const float4& a, const float4& b) { - return make_float4(a.x-b.x, a.y-b.y, a.z-b.z, a.w-b.w); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 psub(const double2& a, const double2& b) { - return make_double2(a.x-b.x, a.y-b.y); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pnegate(const float4& a) { - return make_float4(-a.x, -a.y, -a.z, -a.w); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pnegate(const double2& a) { - return make_double2(-a.x, -a.y); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pconj(const float4& a) { return a; } -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pconj(const double2& a) { return a; } - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pmul(const float4& a, const float4& b) { - return make_float4(a.x*b.x, a.y*b.y, a.z*b.z, a.w*b.w); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pmul(const double2& a, const double2& b) { - return make_double2(a.x*b.x, a.y*b.y); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pdiv(const float4& a, const float4& b) { - return make_float4(a.x/b.x, a.y/b.y, a.z/b.z, a.w/b.w); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pdiv(const double2& a, const double2& b) { - return make_double2(a.x/b.x, a.y/b.y); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pmin(const float4& a, const float4& b) { - return make_float4(fminf(a.x, b.x), fminf(a.y, b.y), fminf(a.z, b.z), fminf(a.w, b.w)); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pmin(const double2& a, const double2& b) { - return make_double2(fmin(a.x, b.x), fmin(a.y, b.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pmax(const float4& a, const float4& b) { - return make_float4(fmaxf(a.x, b.x), fmaxf(a.y, b.y), fmaxf(a.z, b.z), fmaxf(a.w, b.w)); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pmax(const double2& a, const double2& b) { - return make_double2(fmax(a.x, b.x), fmax(a.y, b.y)); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pload(const float* from) { - return *reinterpret_cast(from); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 pload(const double* from) { - return *reinterpret_cast(from); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 ploadu(const float* from) { - return make_float4(from[0], from[1], from[2], from[3]); -} -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 ploadu(const double* from) { - return make_double2(from[0], from[1]); -} - -template<> EIGEN_STRONG_INLINE float4 ploaddup(const float* from) { - return make_float4(from[0], from[0], from[1], from[1]); -} -template<> EIGEN_STRONG_INLINE double2 ploaddup(const double* from) { - return make_double2(from[0], from[0]); -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void pstore(float* to, const float4& from) { - *reinterpret_cast(to) = from; -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void pstore(double* to, const double2& from) { - *reinterpret_cast(to) = from; -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void pstoreu(float* to, const float4& from) { - to[0] = from.x; - to[1] = from.y; - to[2] = from.z; - to[3] = from.w; -} - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void pstoreu(double* to, const double2& from) { - to[0] = from.x; - to[1] = from.y; -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE float4 ploadt_ro(const float* from) { -#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 350 - return __ldg((const float4*)from); -#else - return make_float4(from[0], from[1], from[2], from[3]); -#endif -} -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE double2 ploadt_ro(const double* from) { -#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 350 - return __ldg((const double2*)from); -#else - return make_double2(from[0], from[1]); -#endif -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE float4 ploadt_ro(const float* from) { -#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 350 - return make_float4(__ldg(from+0), __ldg(from+1), __ldg(from+2), __ldg(from+3)); -#else - return make_float4(from[0], from[1], from[2], from[3]); -#endif -} -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE double2 ploadt_ro(const double* from) { -#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 350 - return make_double2(__ldg(from+0), __ldg(from+1)); -#else - return make_double2(from[0], from[1]); -#endif -} - -template<> EIGEN_DEVICE_FUNC inline float4 pgather(const float* from, Index stride) { - return make_float4(from[0*stride], from[1*stride], from[2*stride], from[3*stride]); -} - -template<> EIGEN_DEVICE_FUNC inline double2 pgather(const double* from, Index stride) { - return make_double2(from[0*stride], from[1*stride]); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const float4& from, Index stride) { - to[stride*0] = from.x; - to[stride*1] = from.y; - to[stride*2] = from.z; - to[stride*3] = from.w; -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const double2& from, Index stride) { - to[stride*0] = from.x; - to[stride*1] = from.y; -} - -template<> EIGEN_DEVICE_FUNC inline float pfirst(const float4& a) { - return a.x; -} -template<> EIGEN_DEVICE_FUNC inline double pfirst(const double2& a) { - return a.x; -} - -template<> EIGEN_DEVICE_FUNC inline float predux(const float4& a) { - return a.x + a.y + a.z + a.w; -} -template<> EIGEN_DEVICE_FUNC inline double predux(const double2& a) { - return a.x + a.y; -} - -template<> EIGEN_DEVICE_FUNC inline float predux_max(const float4& a) { - return fmaxf(fmaxf(a.x, a.y), fmaxf(a.z, a.w)); -} -template<> EIGEN_DEVICE_FUNC inline double predux_max(const double2& a) { - return fmax(a.x, a.y); -} - -template<> EIGEN_DEVICE_FUNC inline float predux_min(const float4& a) { - return fminf(fminf(a.x, a.y), fminf(a.z, a.w)); -} -template<> EIGEN_DEVICE_FUNC inline double predux_min(const double2& a) { - return fmin(a.x, a.y); -} - -template<> EIGEN_DEVICE_FUNC inline float predux_mul(const float4& a) { - return a.x * a.y * a.z * a.w; -} -template<> EIGEN_DEVICE_FUNC inline double predux_mul(const double2& a) { - return a.x * a.y; -} - -template<> EIGEN_DEVICE_FUNC inline float4 pabs(const float4& a) { - return make_float4(fabsf(a.x), fabsf(a.y), fabsf(a.z), fabsf(a.w)); -} -template<> EIGEN_DEVICE_FUNC inline double2 pabs(const double2& a) { - return make_double2(fabs(a.x), fabs(a.y)); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - float tmp = kernel.packet[0].y; - kernel.packet[0].y = kernel.packet[1].x; - kernel.packet[1].x = tmp; - - tmp = kernel.packet[0].z; - kernel.packet[0].z = kernel.packet[2].x; - kernel.packet[2].x = tmp; - - tmp = kernel.packet[0].w; - kernel.packet[0].w = kernel.packet[3].x; - kernel.packet[3].x = tmp; - - tmp = kernel.packet[1].z; - kernel.packet[1].z = kernel.packet[2].y; - kernel.packet[2].y = tmp; - - tmp = kernel.packet[1].w; - kernel.packet[1].w = kernel.packet[3].y; - kernel.packet[3].y = tmp; - - tmp = kernel.packet[2].w; - kernel.packet[2].w = kernel.packet[3].z; - kernel.packet[3].z = tmp; -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - double tmp = kernel.packet[0].y; - kernel.packet[0].y = kernel.packet[1].x; - kernel.packet[1].x = tmp; -} - -#endif - -} // end namespace internal - -} // end namespace Eigen - - -#endif // EIGEN_PACKET_MATH_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h deleted file mode 100644 index f749c573ff6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +++ /dev/null @@ -1,1124 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_HALF_CUDA_H -#define EIGEN_PACKET_MATH_HALF_CUDA_H - - -namespace Eigen { -namespace internal { - -// Most of the following operations require arch >= 3.0 -#if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDACC__) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - -template<> struct is_arithmetic { enum { value = true }; }; - -template<> struct packet_traits : default_packet_traits -{ - typedef half2 type; - typedef half2 half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 0, - HasAdd = 1, - HasMul = 1, - HasDiv = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasExp = 1, - HasLog = 1, - HasLog1p = 1 - }; -}; - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=2, alignment=Aligned16}; typedef half2 half; }; - -template<> __device__ EIGEN_STRONG_INLINE half2 pset1(const Eigen::half& from) { - return __half2half2(from); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pload(const Eigen::half* from) { - return *reinterpret_cast(from); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 ploadu(const Eigen::half* from) { - return __halves2half2(from[0], from[1]); -} - -template<> EIGEN_STRONG_INLINE half2 ploaddup(const Eigen::half* from) { - return __halves2half2(from[0], from[0]); -} - -template<> __device__ EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const half2& from) { - *reinterpret_cast(to) = from; -} - -template<> __device__ EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const half2& from) { - to[0] = __low2half(from); - to[1] = __high2half(from); -} - -template<> - __device__ EIGEN_ALWAYS_INLINE half2 ploadt_ro(const Eigen::half* from) { -#if __CUDA_ARCH__ >= 350 - return __ldg((const half2*)from); -#else - return __halves2half2(*(from+0), *(from+1)); -#endif -} - -template<> -__device__ EIGEN_ALWAYS_INLINE half2 ploadt_ro(const Eigen::half* from) { -#if __CUDA_ARCH__ >= 350 - return __halves2half2(__ldg(from+0), __ldg(from+1)); -#else - return __halves2half2(*(from+0), *(from+1)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pgather(const Eigen::half* from, Index stride) { - return __halves2half2(from[0*stride], from[1*stride]); -} - -template<> __device__ EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const half2& from, Index stride) { - to[stride*0] = __low2half(from); - to[stride*1] = __high2half(from); -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half pfirst(const half2& a) { - return __low2half(a); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pabs(const half2& a) { - half2 result; - unsigned temp = *(reinterpret_cast(&(a))); - *(reinterpret_cast(&(result))) = temp & 0x7FFF7FFF; - return result; -} - - -__device__ EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __half a1 = __low2half(kernel.packet[0]); - __half a2 = __high2half(kernel.packet[0]); - __half b1 = __low2half(kernel.packet[1]); - __half b2 = __high2half(kernel.packet[1]); - kernel.packet[0] = __halves2half2(a1, b1); - kernel.packet[1] = __halves2half2(a2, b2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 plset(const Eigen::half& a) { -#if __CUDA_ARCH__ >= 530 - return __halves2half2(a, __hadd(a, __float2half(1.0f))); -#else - float f = __half2float(a) + 1.0f; - return __halves2half2(a, __float2half(f)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 padd(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hadd2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 + b1; - float r2 = a2 + b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 psub(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hsub2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 - b1; - float r2 = a2 - b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pnegate(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hneg2(a); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return __floats2half2_rn(-a1, -a2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pconj(const half2& a) { return a; } - -template<> __device__ EIGEN_STRONG_INLINE half2 pmul(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hmul2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 * b1; - float r2 = a2 * b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmadd(const half2& a, const half2& b, const half2& c) { -#if __CUDA_ARCH__ >= 530 - return __hfma2(a, b, c); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float c1 = __low2float(c); - float c2 = __high2float(c); - float r1 = a1 * b1 + c1; - float r2 = a2 * b2 + c2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pdiv(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 / b1; - float r2 = a2 / b2; - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmin(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - __half r1 = a1 < b1 ? __low2half(a) : __low2half(b); - __half r2 = a2 < b2 ? __high2half(a) : __high2half(b); - return __halves2half2(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmax(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - __half r1 = a1 > b1 ? __low2half(a) : __low2half(b); - __half r2 = a2 > b2 ? __high2half(a) : __high2half(b); - return __halves2half2(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hadd(__low2half(a), __high2half(a)); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 + a2)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_max(const half2& a) { -#if __CUDA_ARCH__ >= 530 - __half first = __low2half(a); - __half second = __high2half(a); - return __hgt(first, second) ? first : second; -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return a1 > a2 ? __low2half(a) : __high2half(a); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_min(const half2& a) { -#if __CUDA_ARCH__ >= 530 - __half first = __low2half(a); - __half second = __high2half(a); - return __hlt(first, second) ? first : second; -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return a1 < a2 ? __low2half(a) : __high2half(a); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_mul(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hmul(__low2half(a), __high2half(a)); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 * a2)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 plog1p(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = log1pf(a1); - float r2 = log1pf(a2); - return __floats2half2_rn(r1, r2); -} - -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - -template<> __device__ EIGEN_STRONG_INLINE -half2 plog(const half2& a) { - return h2log(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 pexp(const half2& a) { - return h2exp(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 psqrt(const half2& a) { - return h2sqrt(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 prsqrt(const half2& a) { - return h2rsqrt(a); -} - -#else - -template<> __device__ EIGEN_STRONG_INLINE half2 plog(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = logf(a1); - float r2 = logf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pexp(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = expf(a1); - float r2 = expf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 psqrt(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = sqrtf(a1); - float r2 = sqrtf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 prsqrt(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = rsqrtf(a1); - float r2 = rsqrtf(a2); - return __floats2half2_rn(r1, r2); -} - -#endif - -#elif defined EIGEN_VECTORIZE_AVX512 - -typedef struct { - __m256i x; -} Packet16h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet16h type; - // There is no half-size packet for Packet16h. - typedef Packet16h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 16, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=16, alignment=Aligned32}; typedef Packet16h half; }; - -template<> EIGEN_STRONG_INLINE Packet16h pset1(const Eigen::half& from) { - Packet16h result; - result.x = _mm256_set1_epi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet16h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm256_extract_epi16(from.x, 0))); -} - -template<> EIGEN_STRONG_INLINE Packet16h pload(const Eigen::half* from) { - Packet16h result; - result.x = _mm256_load_si256(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet16h ploadu(const Eigen::half* from) { - Packet16h result; - result.x = _mm256_loadu_si256(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet16h& from) { - _mm256_store_si256((__m256i*)to, from.x); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet16h& from) { - _mm256_storeu_si256((__m256i*)to, from.x); -} - -template<> EIGEN_STRONG_INLINE Packet16h -ploadquad(const Eigen::half* from) { - Packet16h result; - unsigned short a = from[0].x; - unsigned short b = from[1].x; - unsigned short c = from[2].x; - unsigned short d = from[3].x; - result.x = _mm256_set_epi16(d, d, d, d, c, c, c, c, b, b, b, b, a, a, a, a); - return result; -} - -EIGEN_STRONG_INLINE Packet16f half2float(const Packet16h& a) { -#ifdef EIGEN_HAS_FP16_C - return _mm512_cvtph_ps(a.x); -#else - EIGEN_ALIGN64 half aux[16]; - pstore(aux, a); - float f0(aux[0]); - float f1(aux[1]); - float f2(aux[2]); - float f3(aux[3]); - float f4(aux[4]); - float f5(aux[5]); - float f6(aux[6]); - float f7(aux[7]); - float f8(aux[8]); - float f9(aux[9]); - float fa(aux[10]); - float fb(aux[11]); - float fc(aux[12]); - float fd(aux[13]); - float fe(aux[14]); - float ff(aux[15]); - - return _mm512_set_ps( - ff, fe, fd, fc, fb, fa, f9, f8, f7, f6, f5, f4, f3, f2, f1, f0); -#endif -} - -EIGEN_STRONG_INLINE Packet16h float2half(const Packet16f& a) { -#ifdef EIGEN_HAS_FP16_C - Packet16h result; - result.x = _mm512_cvtps_ph(a, _MM_FROUND_TO_NEAREST_INT|_MM_FROUND_NO_EXC); - return result; -#else - EIGEN_ALIGN64 float aux[16]; - pstore(aux, a); - half h0(aux[0]); - half h1(aux[1]); - half h2(aux[2]); - half h3(aux[3]); - half h4(aux[4]); - half h5(aux[5]); - half h6(aux[6]); - half h7(aux[7]); - half h8(aux[8]); - half h9(aux[9]); - half ha(aux[10]); - half hb(aux[11]); - half hc(aux[12]); - half hd(aux[13]); - half he(aux[14]); - half hf(aux[15]); - - Packet16h result; - result.x = _mm256_set_epi16( - hf.x, he.x, hd.x, hc.x, hb.x, ha.x, h9.x, h8.x, - h7.x, h6.x, h5.x, h4.x, h3.x, h2.x, h1.x, h0.x); - return result; -#endif -} - -template<> EIGEN_STRONG_INLINE Packet16h padd(const Packet16h& a, const Packet16h& b) { - Packet16f af = half2float(a); - Packet16f bf = half2float(b); - Packet16f rf = padd(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet16h pmul(const Packet16h& a, const Packet16h& b) { - Packet16f af = half2float(a); - Packet16f bf = half2float(b); - Packet16f rf = pmul(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE half predux(const Packet16h& from) { - Packet16f from_float = half2float(from); - return half(predux(from_float)); -} - -template<> EIGEN_STRONG_INLINE Packet16h pgather(const Eigen::half* from, Index stride) -{ - Packet16h result; - result.x = _mm256_set_epi16( - from[15*stride].x, from[14*stride].x, from[13*stride].x, from[12*stride].x, - from[11*stride].x, from[10*stride].x, from[9*stride].x, from[8*stride].x, - from[7*stride].x, from[6*stride].x, from[5*stride].x, from[4*stride].x, - from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(half* to, const Packet16h& from, Index stride) -{ - EIGEN_ALIGN64 half aux[16]; - pstore(aux, from); - to[stride*0].x = aux[0].x; - to[stride*1].x = aux[1].x; - to[stride*2].x = aux[2].x; - to[stride*3].x = aux[3].x; - to[stride*4].x = aux[4].x; - to[stride*5].x = aux[5].x; - to[stride*6].x = aux[6].x; - to[stride*7].x = aux[7].x; - to[stride*8].x = aux[8].x; - to[stride*9].x = aux[9].x; - to[stride*10].x = aux[10].x; - to[stride*11].x = aux[11].x; - to[stride*12].x = aux[12].x; - to[stride*13].x = aux[13].x; - to[stride*14].x = aux[14].x; - to[stride*15].x = aux[15].x; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m256i a = kernel.packet[0].x; - __m256i b = kernel.packet[1].x; - __m256i c = kernel.packet[2].x; - __m256i d = kernel.packet[3].x; - __m256i e = kernel.packet[4].x; - __m256i f = kernel.packet[5].x; - __m256i g = kernel.packet[6].x; - __m256i h = kernel.packet[7].x; - __m256i i = kernel.packet[8].x; - __m256i j = kernel.packet[9].x; - __m256i k = kernel.packet[10].x; - __m256i l = kernel.packet[11].x; - __m256i m = kernel.packet[12].x; - __m256i n = kernel.packet[13].x; - __m256i o = kernel.packet[14].x; - __m256i p = kernel.packet[15].x; - - __m256i ab_07 = _mm256_unpacklo_epi16(a, b); - __m256i cd_07 = _mm256_unpacklo_epi16(c, d); - __m256i ef_07 = _mm256_unpacklo_epi16(e, f); - __m256i gh_07 = _mm256_unpacklo_epi16(g, h); - __m256i ij_07 = _mm256_unpacklo_epi16(i, j); - __m256i kl_07 = _mm256_unpacklo_epi16(k, l); - __m256i mn_07 = _mm256_unpacklo_epi16(m, n); - __m256i op_07 = _mm256_unpacklo_epi16(o, p); - - __m256i ab_8f = _mm256_unpackhi_epi16(a, b); - __m256i cd_8f = _mm256_unpackhi_epi16(c, d); - __m256i ef_8f = _mm256_unpackhi_epi16(e, f); - __m256i gh_8f = _mm256_unpackhi_epi16(g, h); - __m256i ij_8f = _mm256_unpackhi_epi16(i, j); - __m256i kl_8f = _mm256_unpackhi_epi16(k, l); - __m256i mn_8f = _mm256_unpackhi_epi16(m, n); - __m256i op_8f = _mm256_unpackhi_epi16(o, p); - - __m256i abcd_03 = _mm256_unpacklo_epi32(ab_07, cd_07); - __m256i abcd_47 = _mm256_unpackhi_epi32(ab_07, cd_07); - __m256i efgh_03 = _mm256_unpacklo_epi32(ef_07, gh_07); - __m256i efgh_47 = _mm256_unpackhi_epi32(ef_07, gh_07); - __m256i ijkl_03 = _mm256_unpacklo_epi32(ij_07, kl_07); - __m256i ijkl_47 = _mm256_unpackhi_epi32(ij_07, kl_07); - __m256i mnop_03 = _mm256_unpacklo_epi32(mn_07, op_07); - __m256i mnop_47 = _mm256_unpackhi_epi32(mn_07, op_07); - - __m256i abcd_8b = _mm256_unpacklo_epi32(ab_8f, cd_8f); - __m256i abcd_cf = _mm256_unpackhi_epi32(ab_8f, cd_8f); - __m256i efgh_8b = _mm256_unpacklo_epi32(ef_8f, gh_8f); - __m256i efgh_cf = _mm256_unpackhi_epi32(ef_8f, gh_8f); - __m256i ijkl_8b = _mm256_unpacklo_epi32(ij_8f, kl_8f); - __m256i ijkl_cf = _mm256_unpackhi_epi32(ij_8f, kl_8f); - __m256i mnop_8b = _mm256_unpacklo_epi32(mn_8f, op_8f); - __m256i mnop_cf = _mm256_unpackhi_epi32(mn_8f, op_8f); - - __m256i abcdefgh_01 = _mm256_unpacklo_epi64(abcd_03, efgh_03); - __m256i abcdefgh_23 = _mm256_unpackhi_epi64(abcd_03, efgh_03); - __m256i ijklmnop_01 = _mm256_unpacklo_epi64(ijkl_03, mnop_03); - __m256i ijklmnop_23 = _mm256_unpackhi_epi64(ijkl_03, mnop_03); - __m256i abcdefgh_45 = _mm256_unpacklo_epi64(abcd_47, efgh_47); - __m256i abcdefgh_67 = _mm256_unpackhi_epi64(abcd_47, efgh_47); - __m256i ijklmnop_45 = _mm256_unpacklo_epi64(ijkl_47, mnop_47); - __m256i ijklmnop_67 = _mm256_unpackhi_epi64(ijkl_47, mnop_47); - __m256i abcdefgh_89 = _mm256_unpacklo_epi64(abcd_8b, efgh_8b); - __m256i abcdefgh_ab = _mm256_unpackhi_epi64(abcd_8b, efgh_8b); - __m256i ijklmnop_89 = _mm256_unpacklo_epi64(ijkl_8b, mnop_8b); - __m256i ijklmnop_ab = _mm256_unpackhi_epi64(ijkl_8b, mnop_8b); - __m256i abcdefgh_cd = _mm256_unpacklo_epi64(abcd_cf, efgh_cf); - __m256i abcdefgh_ef = _mm256_unpackhi_epi64(abcd_cf, efgh_cf); - __m256i ijklmnop_cd = _mm256_unpacklo_epi64(ijkl_cf, mnop_cf); - __m256i ijklmnop_ef = _mm256_unpackhi_epi64(ijkl_cf, mnop_cf); - - // NOTE: no unpacklo/hi instr in this case, so using permute instr. - __m256i a_p_0 = _mm256_permute2x128_si256(abcdefgh_01, ijklmnop_01, 0x20); - __m256i a_p_1 = _mm256_permute2x128_si256(abcdefgh_01, ijklmnop_01, 0x31); - __m256i a_p_2 = _mm256_permute2x128_si256(abcdefgh_23, ijklmnop_23, 0x20); - __m256i a_p_3 = _mm256_permute2x128_si256(abcdefgh_23, ijklmnop_23, 0x31); - __m256i a_p_4 = _mm256_permute2x128_si256(abcdefgh_45, ijklmnop_45, 0x20); - __m256i a_p_5 = _mm256_permute2x128_si256(abcdefgh_45, ijklmnop_45, 0x31); - __m256i a_p_6 = _mm256_permute2x128_si256(abcdefgh_67, ijklmnop_67, 0x20); - __m256i a_p_7 = _mm256_permute2x128_si256(abcdefgh_67, ijklmnop_67, 0x31); - __m256i a_p_8 = _mm256_permute2x128_si256(abcdefgh_89, ijklmnop_89, 0x20); - __m256i a_p_9 = _mm256_permute2x128_si256(abcdefgh_89, ijklmnop_89, 0x31); - __m256i a_p_a = _mm256_permute2x128_si256(abcdefgh_ab, ijklmnop_ab, 0x20); - __m256i a_p_b = _mm256_permute2x128_si256(abcdefgh_ab, ijklmnop_ab, 0x31); - __m256i a_p_c = _mm256_permute2x128_si256(abcdefgh_cd, ijklmnop_cd, 0x20); - __m256i a_p_d = _mm256_permute2x128_si256(abcdefgh_cd, ijklmnop_cd, 0x31); - __m256i a_p_e = _mm256_permute2x128_si256(abcdefgh_ef, ijklmnop_ef, 0x20); - __m256i a_p_f = _mm256_permute2x128_si256(abcdefgh_ef, ijklmnop_ef, 0x31); - - kernel.packet[0].x = a_p_0; - kernel.packet[1].x = a_p_1; - kernel.packet[2].x = a_p_2; - kernel.packet[3].x = a_p_3; - kernel.packet[4].x = a_p_4; - kernel.packet[5].x = a_p_5; - kernel.packet[6].x = a_p_6; - kernel.packet[7].x = a_p_7; - kernel.packet[8].x = a_p_8; - kernel.packet[9].x = a_p_9; - kernel.packet[10].x = a_p_a; - kernel.packet[11].x = a_p_b; - kernel.packet[12].x = a_p_c; - kernel.packet[13].x = a_p_d; - kernel.packet[14].x = a_p_e; - kernel.packet[15].x = a_p_f; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN64 half in[8][16]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - pstore(in[4], kernel.packet[4]); - pstore(in[5], kernel.packet[5]); - pstore(in[6], kernel.packet[6]); - pstore(in[7], kernel.packet[7]); - - EIGEN_ALIGN64 half out[8][16]; - - for (int i = 0; i < 8; ++i) { - for (int j = 0; j < 8; ++j) { - out[i][j] = in[j][2*i]; - } - for (int j = 0; j < 8; ++j) { - out[i][j+8] = in[j][2*i+1]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); - kernel.packet[4] = pload(out[4]); - kernel.packet[5] = pload(out[5]); - kernel.packet[6] = pload(out[6]); - kernel.packet[7] = pload(out[7]); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN64 half in[4][16]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - - EIGEN_ALIGN64 half out[4][16]; - - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 4; ++j) { - out[i][j] = in[j][4*i]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+4] = in[j][4*i+1]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+8] = in[j][4*i+2]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+12] = in[j][4*i+3]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); -} - - -#elif defined EIGEN_VECTORIZE_AVX - -typedef struct { - __m128i x; -} Packet8h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet8h type; - // There is no half-size packet for Packet8h. - typedef Packet8h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 8, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=8, alignment=Aligned16}; typedef Packet8h half; }; - -template<> EIGEN_STRONG_INLINE Packet8h pset1(const Eigen::half& from) { - Packet8h result; - result.x = _mm_set1_epi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet8h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm_extract_epi16(from.x, 0))); -} - -template<> EIGEN_STRONG_INLINE Packet8h pload(const Eigen::half* from) { - Packet8h result; - result.x = _mm_load_si128(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet8h ploadu(const Eigen::half* from) { - Packet8h result; - result.x = _mm_loadu_si128(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet8h& from) { - _mm_store_si128(reinterpret_cast<__m128i*>(to), from.x); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet8h& from) { - _mm_storeu_si128(reinterpret_cast<__m128i*>(to), from.x); -} - -template<> EIGEN_STRONG_INLINE Packet8h -ploadquad(const Eigen::half* from) { - Packet8h result; - unsigned short a = from[0].x; - unsigned short b = from[1].x; - result.x = _mm_set_epi16(b, b, b, b, a, a, a, a); - return result; -} - -EIGEN_STRONG_INLINE Packet8f half2float(const Packet8h& a) { -#ifdef EIGEN_HAS_FP16_C - return _mm256_cvtph_ps(a.x); -#else - EIGEN_ALIGN32 Eigen::half aux[8]; - pstore(aux, a); - float f0(aux[0]); - float f1(aux[1]); - float f2(aux[2]); - float f3(aux[3]); - float f4(aux[4]); - float f5(aux[5]); - float f6(aux[6]); - float f7(aux[7]); - - return _mm256_set_ps(f7, f6, f5, f4, f3, f2, f1, f0); -#endif -} - -EIGEN_STRONG_INLINE Packet8h float2half(const Packet8f& a) { -#ifdef EIGEN_HAS_FP16_C - Packet8h result; - result.x = _mm256_cvtps_ph(a, _MM_FROUND_TO_NEAREST_INT|_MM_FROUND_NO_EXC); - return result; -#else - EIGEN_ALIGN32 float aux[8]; - pstore(aux, a); - Eigen::half h0(aux[0]); - Eigen::half h1(aux[1]); - Eigen::half h2(aux[2]); - Eigen::half h3(aux[3]); - Eigen::half h4(aux[4]); - Eigen::half h5(aux[5]); - Eigen::half h6(aux[6]); - Eigen::half h7(aux[7]); - - Packet8h result; - result.x = _mm_set_epi16(h7.x, h6.x, h5.x, h4.x, h3.x, h2.x, h1.x, h0.x); - return result; -#endif -} - -template<> EIGEN_STRONG_INLINE Packet8h pconj(const Packet8h& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet8h padd(const Packet8h& a, const Packet8h& b) { - Packet8f af = half2float(a); - Packet8f bf = half2float(b); - Packet8f rf = padd(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet8h pmul(const Packet8h& a, const Packet8h& b) { - Packet8f af = half2float(a); - Packet8f bf = half2float(b); - Packet8f rf = pmul(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet8h pgather(const Eigen::half* from, Index stride) -{ - Packet8h result; - result.x = _mm_set_epi16(from[7*stride].x, from[6*stride].x, from[5*stride].x, from[4*stride].x, from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const Packet8h& from, Index stride) -{ - EIGEN_ALIGN32 Eigen::half aux[8]; - pstore(aux, from); - to[stride*0].x = aux[0].x; - to[stride*1].x = aux[1].x; - to[stride*2].x = aux[2].x; - to[stride*3].x = aux[3].x; - to[stride*4].x = aux[4].x; - to[stride*5].x = aux[5].x; - to[stride*6].x = aux[6].x; - to[stride*7].x = aux[7].x; -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_max(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_max(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_min(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_min(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_mul(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_mul(af); - return Eigen::half(reduced); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m128i a = kernel.packet[0].x; - __m128i b = kernel.packet[1].x; - __m128i c = kernel.packet[2].x; - __m128i d = kernel.packet[3].x; - __m128i e = kernel.packet[4].x; - __m128i f = kernel.packet[5].x; - __m128i g = kernel.packet[6].x; - __m128i h = kernel.packet[7].x; - - __m128i a03b03 = _mm_unpacklo_epi16(a, b); - __m128i c03d03 = _mm_unpacklo_epi16(c, d); - __m128i e03f03 = _mm_unpacklo_epi16(e, f); - __m128i g03h03 = _mm_unpacklo_epi16(g, h); - __m128i a47b47 = _mm_unpackhi_epi16(a, b); - __m128i c47d47 = _mm_unpackhi_epi16(c, d); - __m128i e47f47 = _mm_unpackhi_epi16(e, f); - __m128i g47h47 = _mm_unpackhi_epi16(g, h); - - __m128i a01b01c01d01 = _mm_unpacklo_epi32(a03b03, c03d03); - __m128i a23b23c23d23 = _mm_unpackhi_epi32(a03b03, c03d03); - __m128i e01f01g01h01 = _mm_unpacklo_epi32(e03f03, g03h03); - __m128i e23f23g23h23 = _mm_unpackhi_epi32(e03f03, g03h03); - __m128i a45b45c45d45 = _mm_unpacklo_epi32(a47b47, c47d47); - __m128i a67b67c67d67 = _mm_unpackhi_epi32(a47b47, c47d47); - __m128i e45f45g45h45 = _mm_unpacklo_epi32(e47f47, g47h47); - __m128i e67f67g67h67 = _mm_unpackhi_epi32(e47f47, g47h47); - - __m128i a0b0c0d0e0f0g0h0 = _mm_unpacklo_epi64(a01b01c01d01, e01f01g01h01); - __m128i a1b1c1d1e1f1g1h1 = _mm_unpackhi_epi64(a01b01c01d01, e01f01g01h01); - __m128i a2b2c2d2e2f2g2h2 = _mm_unpacklo_epi64(a23b23c23d23, e23f23g23h23); - __m128i a3b3c3d3e3f3g3h3 = _mm_unpackhi_epi64(a23b23c23d23, e23f23g23h23); - __m128i a4b4c4d4e4f4g4h4 = _mm_unpacklo_epi64(a45b45c45d45, e45f45g45h45); - __m128i a5b5c5d5e5f5g5h5 = _mm_unpackhi_epi64(a45b45c45d45, e45f45g45h45); - __m128i a6b6c6d6e6f6g6h6 = _mm_unpacklo_epi64(a67b67c67d67, e67f67g67h67); - __m128i a7b7c7d7e7f7g7h7 = _mm_unpackhi_epi64(a67b67c67d67, e67f67g67h67); - - kernel.packet[0].x = a0b0c0d0e0f0g0h0; - kernel.packet[1].x = a1b1c1d1e1f1g1h1; - kernel.packet[2].x = a2b2c2d2e2f2g2h2; - kernel.packet[3].x = a3b3c3d3e3f3g3h3; - kernel.packet[4].x = a4b4c4d4e4f4g4h4; - kernel.packet[5].x = a5b5c5d5e5f5g5h5; - kernel.packet[6].x = a6b6c6d6e6f6g6h6; - kernel.packet[7].x = a7b7c7d7e7f7g7h7; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN32 Eigen::half in[4][8]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - - EIGEN_ALIGN32 Eigen::half out[4][8]; - - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 4; ++j) { - out[i][j] = in[j][2*i]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+4] = in[j][2*i+1]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); -} - - -// Disable the following code since it's broken on too many platforms / compilers. -//#elif defined(EIGEN_VECTORIZE_SSE) && (!EIGEN_ARCH_x86_64) && (!EIGEN_COMP_MSVC) -#elif 0 - -typedef struct { - __m64 x; -} Packet4h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet4h type; - // There is no half-size packet for Packet4h. - typedef Packet4h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=4, alignment=Aligned16}; typedef Packet4h half; }; - -template<> EIGEN_STRONG_INLINE Packet4h pset1(const Eigen::half& from) { - Packet4h result; - result.x = _mm_set1_pi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet4h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm_cvtsi64_si32(from.x))); -} - -template<> EIGEN_STRONG_INLINE Packet4h pconj(const Packet4h& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4h padd(const Packet4h& a, const Packet4h& b) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - __int64_t b64 = _mm_cvtm64_si64(b.x); - - Eigen::half h[4]; - - Eigen::half ha = half_impl::raw_uint16_to_half(static_cast(a64)); - Eigen::half hb = half_impl::raw_uint16_to_half(static_cast(b64)); - h[0] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 16)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 16)); - h[1] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 32)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 32)); - h[2] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 48)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 48)); - h[3] = ha + hb; - Packet4h result; - result.x = _mm_set_pi16(h[3].x, h[2].x, h[1].x, h[0].x); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h pmul(const Packet4h& a, const Packet4h& b) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - __int64_t b64 = _mm_cvtm64_si64(b.x); - - Eigen::half h[4]; - - Eigen::half ha = half_impl::raw_uint16_to_half(static_cast(a64)); - Eigen::half hb = half_impl::raw_uint16_to_half(static_cast(b64)); - h[0] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 16)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 16)); - h[1] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 32)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 32)); - h[2] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 48)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 48)); - h[3] = ha * hb; - Packet4h result; - result.x = _mm_set_pi16(h[3].x, h[2].x, h[1].x, h[0].x); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h pload(const Eigen::half* from) { - Packet4h result; - result.x = _mm_cvtsi64_m64(*reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h ploadu(const Eigen::half* from) { - Packet4h result; - result.x = _mm_cvtsi64_m64(*reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet4h& from) { - __int64_t r = _mm_cvtm64_si64(from.x); - *(reinterpret_cast<__int64_t*>(to)) = r; -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet4h& from) { - __int64_t r = _mm_cvtm64_si64(from.x); - *(reinterpret_cast<__int64_t*>(to)) = r; -} - -template<> EIGEN_STRONG_INLINE Packet4h -ploadquad(const Eigen::half* from) { - return pset1(*from); -} - -template<> EIGEN_STRONG_INLINE Packet4h pgather(const Eigen::half* from, Index stride) -{ - Packet4h result; - result.x = _mm_set_pi16(from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const Packet4h& from, Index stride) -{ - __int64_t a = _mm_cvtm64_si64(from.x); - to[stride*0].x = static_cast(a); - to[stride*1].x = static_cast(a >> 16); - to[stride*2].x = static_cast(a >> 32); - to[stride*3].x = static_cast(a >> 48); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m64 T0 = _mm_unpacklo_pi16(kernel.packet[0].x, kernel.packet[1].x); - __m64 T1 = _mm_unpacklo_pi16(kernel.packet[2].x, kernel.packet[3].x); - __m64 T2 = _mm_unpackhi_pi16(kernel.packet[0].x, kernel.packet[1].x); - __m64 T3 = _mm_unpackhi_pi16(kernel.packet[2].x, kernel.packet[3].x); - - kernel.packet[0].x = _mm_unpacklo_pi32(T0, T1); - kernel.packet[1].x = _mm_unpackhi_pi32(T0, T1); - kernel.packet[2].x = _mm_unpacklo_pi32(T2, T3); - kernel.packet[3].x = _mm_unpackhi_pi32(T2, T3); -} - -#endif - -} -} - -#endif // EIGEN_PACKET_MATH_HALF_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h deleted file mode 100644 index aa5fbce8eac..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h +++ /dev/null @@ -1,212 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TYPE_CASTING_CUDA_H -#define EIGEN_TYPE_CASTING_CUDA_H - -namespace Eigen { - -namespace internal { - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef Eigen::half result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Eigen::half operator() (const float& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __float2half(a); - #else - return Eigen::half(a); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef Eigen::half result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Eigen::half operator() (const int& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __float2half(static_cast(a)); - #else - return Eigen::half(static_cast(a)); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef float result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float operator() (const Eigen::half& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __half2float(a); - #else - return static_cast(a); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - - -#if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 2, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pcast(const half2& a, const half2& b) { - float2 r1 = __half22float2(a); - float2 r2 = __half22float2(b); - return make_float4(r1.x, r1.y, r2.x, r2.y); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 2 - }; -}; - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE half2 pcast(const float4& a) { - // Simply discard the second half of the input - return __floats2half2_rn(a.x, a.y); -} - -#elif defined EIGEN_VECTORIZE_AVX512 -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet16f pcast(const Packet16h& a) { - return half2float(a); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet16h pcast(const Packet16f& a) { - return float2half(a); -} - -#elif defined EIGEN_VECTORIZE_AVX - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet8f pcast(const Packet8h& a) { - return half2float(a); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet8h pcast(const Packet8f& a) { - return float2half(a); -} - -// Disable the following code since it's broken on too many platforms / compilers. -//#elif defined(EIGEN_VECTORIZE_SSE) && (!EIGEN_ARCH_x86_64) && (!EIGEN_COMP_MSVC) -#elif 0 - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet4h& a) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - Eigen::half h = raw_uint16_to_half(static_cast(a64)); - float f1 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 16)); - float f2 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 32)); - float f3 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 48)); - float f4 = static_cast(h); - return _mm_set_ps(f4, f3, f2, f1); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet4h pcast(const Packet4f& a) { - EIGEN_ALIGN16 float aux[4]; - pstore(aux, a); - Eigen::half h0(aux[0]); - Eigen::half h1(aux[1]); - Eigen::half h2(aux[2]); - Eigen::half h3(aux[3]); - - Packet4h result; - result.x = _mm_set_pi16(h3.x, h2.x, h1.x, h0.x); - return result; -} - -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TYPE_CASTING_CUDA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h deleted file mode 100644 index 4cfe34e0526..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h +++ /dev/null @@ -1,29 +0,0 @@ - -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2017 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ARCH_CONJ_HELPER_H -#define EIGEN_ARCH_CONJ_HELPER_H - -#define EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(PACKET_CPLX, PACKET_REAL) \ - template<> struct conj_helper { \ - EIGEN_STRONG_INLINE PACKET_CPLX pmadd(const PACKET_REAL& x, const PACKET_CPLX& y, const PACKET_CPLX& c) const \ - { return padd(c, pmul(x,y)); } \ - EIGEN_STRONG_INLINE PACKET_CPLX pmul(const PACKET_REAL& x, const PACKET_CPLX& y) const \ - { return PACKET_CPLX(Eigen::internal::pmul(x, y.v)); } \ - }; \ - \ - template<> struct conj_helper { \ - EIGEN_STRONG_INLINE PACKET_CPLX pmadd(const PACKET_CPLX& x, const PACKET_REAL& y, const PACKET_CPLX& c) const \ - { return padd(c, pmul(x,y)); } \ - EIGEN_STRONG_INLINE PACKET_CPLX pmul(const PACKET_CPLX& x, const PACKET_REAL& y) const \ - { return PACKET_CPLX(Eigen::internal::pmul(x.v, y)); } \ - }; - -#endif // EIGEN_ARCH_CONJ_HELPER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h deleted file mode 100644 index 097373c84dc..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h +++ /dev/null @@ -1,49 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -/* All the parameters defined in this file can be specialized in the - * architecture specific files, and/or by the user. - * More to come... */ - -#ifndef EIGEN_DEFAULT_SETTINGS_H -#define EIGEN_DEFAULT_SETTINGS_H - -/** Defines the maximal loop size to enable meta unrolling of loops. - * Note that the value here is expressed in Eigen's own notion of "number of FLOPS", - * it does not correspond to the number of iterations or the number of instructions - */ -#ifndef EIGEN_UNROLLING_LIMIT -#define EIGEN_UNROLLING_LIMIT 100 -#endif - -/** Defines the threshold between a "small" and a "large" matrix. - * This threshold is mainly used to select the proper product implementation. - */ -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -/** Defines the maximal width of the blocks used in the triangular product and solver - * for vectors (level 2 blas xTRMV and xTRSV). The default is 8. - */ -#ifndef EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH -#define EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH 8 -#endif - - -/** Defines the default number of registers available for that architecture. - * Currently it must be 8 or 16. Other values will fail. - */ -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 8 -#endif - -#endif // EIGEN_DEFAULT_SETTINGS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/Complex.h deleted file mode 100644 index 306a309beb2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/Complex.h +++ /dev/null @@ -1,490 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// Copyright (C) 2010 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_NEON_H -#define EIGEN_COMPLEX_NEON_H - -namespace Eigen { - -namespace internal { - -inline uint32x4_t p4ui_CONJ_XOR() { -// See bug 1325, clang fails to call vld1q_u64. -#if EIGEN_COMP_CLANG - uint32x4_t ret = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 }; - return ret; -#else - static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 }; - return vld1q_u32( conj_XOR_DATA ); -#endif -} - -inline uint32x2_t p2ui_CONJ_XOR() { - static const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000 }; - return vld1_u32( conj_XOR_DATA ); -} - -//---------- float ---------- -struct Packet2cf -{ - EIGEN_STRONG_INLINE Packet2cf() {} - EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {} - Packet4f v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; - -template<> EIGEN_STRONG_INLINE Packet2cf pset1(const std::complex& from) -{ - float32x2_t r64; - r64 = vld1_f32((const float *)&from); - - return Packet2cf(vcombine_f32(r64, r64)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf padd(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(padd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf psub(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(psub(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate(a.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) -{ - Packet4ui b = vreinterpretq_u32_f32(a.v); - return Packet2cf(vreinterpretq_f32_u32(veorq_u32(b, p4ui_CONJ_XOR()))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) -{ - Packet4f v1, v2; - - // Get the real values of a | a1_re | a1_re | a2_re | a2_re | - v1 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 0), vdup_lane_f32(vget_high_f32(a.v), 0)); - // Get the imag values of a | a1_im | a1_im | a2_im | a2_im | - v2 = vcombine_f32(vdup_lane_f32(vget_low_f32(a.v), 1), vdup_lane_f32(vget_high_f32(a.v), 1)); - // Multiply the real a with b - v1 = vmulq_f32(v1, b.v); - // Multiply the imag a with b - v2 = vmulq_f32(v2, b.v); - // Conjugate v2 - v2 = vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(v2), p4ui_CONJ_XOR())); - // Swap real/imag elements in v2. - v2 = vrev64q_f32(v2); - // Add and return the result - return Packet2cf(vaddq_f32(v1, v2)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pand (const Packet2cf& a, const Packet2cf& b) -{ - return Packet2cf(vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet2cf por (const Packet2cf& a, const Packet2cf& b) -{ - return Packet2cf(vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet2cf pxor (const Packet2cf& a, const Packet2cf& b) -{ - return Packet2cf(vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet2cf pandnot(const Packet2cf& a, const Packet2cf& b) -{ - return Packet2cf(vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a.v),vreinterpretq_u32_f32(b.v)))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pload(const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload((const float*)from)); } -template<> EIGEN_STRONG_INLINE Packet2cf ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu((const float*)from)); } - -template<> EIGEN_STRONG_INLINE Packet2cf ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); } - -template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather, Packet2cf>(const std::complex* from, Index stride) -{ - Packet4f res = pset1(0.f); - res = vsetq_lane_f32(std::real(from[0*stride]), res, 0); - res = vsetq_lane_f32(std::imag(from[0*stride]), res, 1); - res = vsetq_lane_f32(std::real(from[1*stride]), res, 2); - res = vsetq_lane_f32(std::imag(from[1*stride]), res, 3); - return Packet2cf(res); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cf>(std::complex* to, const Packet2cf& from, Index stride) -{ - to[stride*0] = std::complex(vgetq_lane_f32(from.v, 0), vgetq_lane_f32(from.v, 1)); - to[stride*1] = std::complex(vgetq_lane_f32(from.v, 2), vgetq_lane_f32(from.v, 3)); -} - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_ARM_PREFETCH((const float *)addr); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cf& a) -{ - std::complex EIGEN_ALIGN16 x[2]; - vst1q_f32((float *)x, a.v); - return x[0]; -} - -template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) -{ - float32x2_t a_lo, a_hi; - Packet4f a_r128; - - a_lo = vget_low_f32(a.v); - a_hi = vget_high_f32(a.v); - a_r128 = vcombine_f32(a_hi, a_lo); - - return Packet2cf(a_r128); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pcplxflip(const Packet2cf& a) -{ - return Packet2cf(vrev64q_f32(a.v)); -} - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cf& a) -{ - float32x2_t a1, a2; - std::complex s; - - a1 = vget_low_f32(a.v); - a2 = vget_high_f32(a.v); - a2 = vadd_f32(a1, a2); - vst1_f32((float *)&s, a2); - - return s; -} - -template<> EIGEN_STRONG_INLINE Packet2cf preduxp(const Packet2cf* vecs) -{ - Packet4f sum1, sum2, sum; - - // Add the first two 64-bit float32x2_t of vecs[0] - sum1 = vcombine_f32(vget_low_f32(vecs[0].v), vget_low_f32(vecs[1].v)); - sum2 = vcombine_f32(vget_high_f32(vecs[0].v), vget_high_f32(vecs[1].v)); - sum = vaddq_f32(sum1, sum2); - - return Packet2cf(sum); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cf& a) -{ - float32x2_t a1, a2, v1, v2, prod; - std::complex s; - - a1 = vget_low_f32(a.v); - a2 = vget_high_f32(a.v); - // Get the real values of a | a1_re | a1_re | a2_re | a2_re | - v1 = vdup_lane_f32(a1, 0); - // Get the real values of a | a1_im | a1_im | a2_im | a2_im | - v2 = vdup_lane_f32(a1, 1); - // Multiply the real a with b - v1 = vmul_f32(v1, a2); - // Multiply the imag a with b - v2 = vmul_f32(v2, a2); - // Conjugate v2 - v2 = vreinterpret_f32_u32(veor_u32(vreinterpret_u32_f32(v2), p2ui_CONJ_XOR())); - // Swap real/imag elements in v2. - v2 = vrev64_f32(v2); - // Add v1, v2 - prod = vadd_f32(v1, v2); - - vst1_f32((float *)&s, prod); - - return s; -} - -template -struct palign_impl -{ - EIGEN_STRONG_INLINE static void run(Packet2cf& first, const Packet2cf& second) - { - if (Offset==1) - { - first.v = vextq_f32(first.v, second.v, 2); - } - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) - -template<> EIGEN_STRONG_INLINE Packet2cf pdiv(const Packet2cf& a, const Packet2cf& b) -{ - // TODO optimize it for NEON - Packet2cf res = conj_helper().pmul(a,b); - Packet4f s, rev_s; - - // this computes the norm - s = vmulq_f32(b.v, b.v); - rev_s = vrev64q_f32(s); - - return Packet2cf(pdiv(res.v, vaddq_f32(s,rev_s))); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet4f tmp = vcombine_f32(vget_high_f32(kernel.packet[0].v), vget_high_f32(kernel.packet[1].v)); - kernel.packet[0].v = vcombine_f32(vget_low_f32(kernel.packet[0].v), vget_low_f32(kernel.packet[1].v)); - kernel.packet[1].v = tmp; -} - -//---------- double ---------- -#if EIGEN_ARCH_ARM64 && !EIGEN_APPLE_DOUBLE_NEON_BUG - -// See bug 1325, clang fails to call vld1q_u64. -#if EIGEN_COMP_CLANG - static uint64x2_t p2ul_CONJ_XOR = {0x0, 0x8000000000000000}; -#else - const uint64_t p2ul_conj_XOR_DATA[] = { 0x0, 0x8000000000000000 }; - static uint64x2_t p2ul_CONJ_XOR = vld1q_u64( p2ul_conj_XOR_DATA ); -#endif - -struct Packet1cd -{ - EIGEN_STRONG_INLINE Packet1cd() {} - EIGEN_STRONG_INLINE explicit Packet1cd(const Packet2d& a) : v(a) {} - Packet2d v; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet1cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 0, - size = 1, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; - -template<> EIGEN_STRONG_INLINE Packet1cd pload(const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu((const double*)from)); } - -template<> EIGEN_STRONG_INLINE Packet1cd pset1(const std::complex& from) -{ /* here we really have to use unaligned loads :( */ return ploadu(&from); } - -template<> EIGEN_STRONG_INLINE Packet1cd padd(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(padd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd psub(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(psub(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(a.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) { return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v), p2ul_CONJ_XOR))); } - -template<> EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) -{ - Packet2d v1, v2; - - // Get the real values of a - v1 = vdupq_lane_f64(vget_low_f64(a.v), 0); - // Get the imag values of a - v2 = vdupq_lane_f64(vget_high_f64(a.v), 0); - // Multiply the real a with b - v1 = vmulq_f64(v1, b.v); - // Multiply the imag a with b - v2 = vmulq_f64(v2, b.v); - // Conjugate v2 - v2 = vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(v2), p2ul_CONJ_XOR)); - // Swap real/imag elements in v2. - v2 = preverse(v2); - // Add and return the result - return Packet1cd(vaddq_f64(v1, v2)); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pand (const Packet1cd& a, const Packet1cd& b) -{ - return Packet1cd(vreinterpretq_f64_u64(vandq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet1cd por (const Packet1cd& a, const Packet1cd& b) -{ - return Packet1cd(vreinterpretq_f64_u64(vorrq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet1cd pxor (const Packet1cd& a, const Packet1cd& b) -{ - return Packet1cd(vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); -} -template<> EIGEN_STRONG_INLINE Packet1cd pandnot(const Packet1cd& a, const Packet1cd& b) -{ - return Packet1cd(vreinterpretq_f64_u64(vbicq_u64(vreinterpretq_u64_f64(a.v),vreinterpretq_u64_f64(b.v)))); -} - -template<> EIGEN_STRONG_INLINE Packet1cd ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, from.v); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_ARM_PREFETCH((const double *)addr); } - -template<> EIGEN_DEVICE_FUNC inline Packet1cd pgather, Packet1cd>(const std::complex* from, Index stride) -{ - Packet2d res = pset1(0.0); - res = vsetq_lane_f64(std::real(from[0*stride]), res, 0); - res = vsetq_lane_f64(std::imag(from[0*stride]), res, 1); - return Packet1cd(res); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet1cd>(std::complex* to, const Packet1cd& from, Index stride) -{ - to[stride*0] = std::complex(vgetq_lane_f64(from.v, 0), vgetq_lane_f64(from.v, 1)); -} - - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet1cd& a) -{ - std::complex EIGEN_ALIGN16 res; - pstore >(&res, a); - - return res; -} - -template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet1cd& a) { return pfirst(a); } - -template<> EIGEN_STRONG_INLINE Packet1cd preduxp(const Packet1cd* vecs) { return vecs[0]; } - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet1cd& a) { return pfirst(a); } - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) - { - // FIXME is it sure we never have to align a Packet1cd? - // Even though a std::complex has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) - -template<> EIGEN_STRONG_INLINE Packet1cd pdiv(const Packet1cd& a, const Packet1cd& b) -{ - // TODO optimize it for NEON - Packet1cd res = conj_helper().pmul(a,b); - Packet2d s = pmul(b.v, b.v); - Packet2d rev_s = preverse(s); - - return Packet1cd(pdiv(res.v, padd(s,rev_s))); -} - -EIGEN_STRONG_INLINE Packet1cd pcplxflip/**/(const Packet1cd& x) -{ - return Packet1cd(preverse(Packet2d(x.v))); -} - -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel) -{ - Packet2d tmp = vcombine_f64(vget_high_f64(kernel.packet[0].v), vget_high_f64(kernel.packet[1].v)); - kernel.packet[0].v = vcombine_f64(vget_low_f64(kernel.packet[0].v), vget_low_f64(kernel.packet[1].v)); - kernel.packet[1].v = tmp; -} -#endif // EIGEN_ARCH_ARM64 - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_NEON_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/MathFunctions.h deleted file mode 100644 index 6bb05bb922a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/MathFunctions.h +++ /dev/null @@ -1,91 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* The sin, cos, exp, and log functions of this file come from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -#ifndef EIGEN_MATH_FUNCTIONS_NEON_H -#define EIGEN_MATH_FUNCTIONS_NEON_H - -namespace Eigen { - -namespace internal { - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pexp(const Packet4f& _x) -{ - Packet4f x = _x; - Packet4f tmp, fx; - - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); - _EIGEN_DECLARE_CONST_Packet4f(exp_hi, 88.3762626647950f); - _EIGEN_DECLARE_CONST_Packet4f(exp_lo, -88.3762626647949f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_LOG2EF, 1.44269504088896341f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C1, 0.693359375f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C2, -2.12194440e-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p0, 1.9875691500E-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p1, 1.3981999507E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p2, 8.3334519073E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p3, 4.1665795894E-2f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201E-1f); - - x = vminq_f32(x, p4f_exp_hi); - x = vmaxq_f32(x, p4f_exp_lo); - - /* express exp(x) as exp(g + n*log(2)) */ - fx = vmlaq_f32(p4f_half, x, p4f_cephes_LOG2EF); - - /* perform a floorf */ - tmp = vcvtq_f32_s32(vcvtq_s32_f32(fx)); - - /* if greater, substract 1 */ - Packet4ui mask = vcgtq_f32(tmp, fx); - mask = vandq_u32(mask, vreinterpretq_u32_f32(p4f_1)); - - fx = vsubq_f32(tmp, vreinterpretq_f32_u32(mask)); - - tmp = vmulq_f32(fx, p4f_cephes_exp_C1); - Packet4f z = vmulq_f32(fx, p4f_cephes_exp_C2); - x = vsubq_f32(x, tmp); - x = vsubq_f32(x, z); - - Packet4f y = vmulq_f32(p4f_cephes_exp_p0, x); - z = vmulq_f32(x, x); - y = vaddq_f32(y, p4f_cephes_exp_p1); - y = vmulq_f32(y, x); - y = vaddq_f32(y, p4f_cephes_exp_p2); - y = vmulq_f32(y, x); - y = vaddq_f32(y, p4f_cephes_exp_p3); - y = vmulq_f32(y, x); - y = vaddq_f32(y, p4f_cephes_exp_p4); - y = vmulq_f32(y, x); - y = vaddq_f32(y, p4f_cephes_exp_p5); - - y = vmulq_f32(y, z); - y = vaddq_f32(y, x); - y = vaddq_f32(y, p4f_1); - - /* build 2^n */ - int32x4_t mm; - mm = vcvtq_s32_f32(fx); - mm = vaddq_s32(mm, p4i_0x7f); - mm = vshlq_n_s32(mm, 23); - Packet4f pow2n = vreinterpretq_f32_s32(mm); - - y = vmulq_f32(y, pow2n); - return y; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_NEON_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/PacketMath.h deleted file mode 100644 index 3d5ed0d240c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/NEON/PacketMath.h +++ /dev/null @@ -1,760 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2010 Konstantinos Margaritis -// Heavily based on Gael's SSE version. -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_NEON_H -#define EIGEN_PACKET_MATH_NEON_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#if EIGEN_ARCH_ARM64 -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32 -#else -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 16 -#endif -#endif - -#if EIGEN_COMP_MSVC - -// In MSVC's arm_neon.h header file, all NEON vector types -// are aliases to the same underlying type __n128. -// We thus have to wrap them to make them different C++ types. -// (See also bug 1428) - -template -struct eigen_packet_wrapper -{ - operator T&() { return m_val; } - operator const T&() const { return m_val; } - eigen_packet_wrapper() {} - eigen_packet_wrapper(const T &v) : m_val(v) {} - eigen_packet_wrapper& operator=(const T &v) { - m_val = v; - return *this; - } - - T m_val; -}; -typedef eigen_packet_wrapper Packet2f; -typedef eigen_packet_wrapper Packet4f; -typedef eigen_packet_wrapper Packet4i; -typedef eigen_packet_wrapper Packet2i; -typedef eigen_packet_wrapper Packet4ui; - -#else - -typedef float32x2_t Packet2f; -typedef float32x4_t Packet4f; -typedef int32x4_t Packet4i; -typedef int32x2_t Packet2i; -typedef uint32x4_t Packet4ui; - -#endif // EIGEN_COMP_MSVC - -#define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \ - const Packet4f p4f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \ - const Packet4f p4f_##NAME = vreinterpretq_f32_u32(pset1(X)) - -#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ - const Packet4i p4i_##NAME = pset1(X) - -#if EIGEN_ARCH_ARM64 - // __builtin_prefetch tends to do nothing on ARM64 compilers because the - // prefetch instructions there are too detailed for __builtin_prefetch to map - // meaningfully to them. - #define EIGEN_ARM_PREFETCH(ADDR) __asm__ __volatile__("prfm pldl1keep, [%[addr]]\n" ::[addr] "r"(ADDR) : ); -#elif EIGEN_HAS_BUILTIN(__builtin_prefetch) || EIGEN_COMP_GNUC - #define EIGEN_ARM_PREFETCH(ADDR) __builtin_prefetch(ADDR); -#elif defined __pld - #define EIGEN_ARM_PREFETCH(ADDR) __pld(ADDR) -#elif EIGEN_ARCH_ARM32 - #define EIGEN_ARM_PREFETCH(ADDR) __asm__ __volatile__ ("pld [%[addr]]\n" :: [addr] "r" (ADDR) : ); -#else - // by default no explicit prefetching - #define EIGEN_ARM_PREFETCH(ADDR) -#endif - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4f type; - typedef Packet4f half; // Packet2f intrinsics not implemented yet - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket=0, // Packet2f intrinsics not implemented yet - - HasDiv = 1, - // FIXME check the Has* - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 1, - HasSqrt = 0 - }; -}; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4i type; - typedef Packet4i half; // Packet2i intrinsics not implemented yet - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket=0 // Packet2i intrinsics not implemented yet - // FIXME check the Has* - }; -}; - -#if EIGEN_GNUC_AT_MOST(4,4) && !EIGEN_COMP_LLVM -// workaround gcc 4.2, 4.3 and 4.4 compilatin issue -EIGEN_STRONG_INLINE float32x4_t vld1q_f32(const float* x) { return ::vld1q_f32((const float32_t*)x); } -EIGEN_STRONG_INLINE float32x2_t vld1_f32 (const float* x) { return ::vld1_f32 ((const float32_t*)x); } -EIGEN_STRONG_INLINE float32x2_t vld1_dup_f32 (const float* x) { return ::vld1_dup_f32 ((const float32_t*)x); } -EIGEN_STRONG_INLINE void vst1q_f32(float* to, float32x4_t from) { ::vst1q_f32((float32_t*)to,from); } -EIGEN_STRONG_INLINE void vst1_f32 (float* to, float32x2_t from) { ::vst1_f32 ((float32_t*)to,from); } -#endif - -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; -template<> struct unpacket_traits { typedef int32_t type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; - -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { return vdupq_n_f32(from); } -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int32_t& from) { return vdupq_n_s32(from); } - -template<> EIGEN_STRONG_INLINE Packet4f plset(const float& a) -{ - const float f[] = {0, 1, 2, 3}; - Packet4f countdown = vld1q_f32(f); - return vaddq_f32(pset1(a), countdown); -} -template<> EIGEN_STRONG_INLINE Packet4i plset(const int32_t& a) -{ - const int32_t i[] = {0, 1, 2, 3}; - Packet4i countdown = vld1q_s32(i); - return vaddq_s32(pset1(a), countdown); -} - -template<> EIGEN_STRONG_INLINE Packet4f padd(const Packet4f& a, const Packet4f& b) { return vaddq_f32(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i padd(const Packet4i& a, const Packet4i& b) { return vaddq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f psub(const Packet4f& a, const Packet4f& b) { return vsubq_f32(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i psub(const Packet4i& a, const Packet4i& b) { return vsubq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) { return vnegq_f32(a); } -template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return vnegq_s32(a); } - -template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4f pmul(const Packet4f& a, const Packet4f& b) { return vmulq_f32(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmul(const Packet4i& a, const Packet4i& b) { return vmulq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pdiv(const Packet4f& a, const Packet4f& b) -{ -#if EIGEN_ARCH_ARM64 - return vdivq_f32(a,b); -#else - Packet4f inv, restep, div; - - // NEON does not offer a divide instruction, we have to do a reciprocal approximation - // However NEON in contrast to other SIMD engines (AltiVec/SSE), offers - // a reciprocal estimate AND a reciprocal step -which saves a few instructions - // vrecpeq_f32() returns an estimate to 1/b, which we will finetune with - // Newton-Raphson and vrecpsq_f32() - inv = vrecpeq_f32(b); - - // This returns a differential, by which we will have to multiply inv to get a better - // approximation of 1/b. - restep = vrecpsq_f32(b, inv); - inv = vmulq_f32(restep, inv); - - // Finally, multiply a by 1/b and get the wanted result of the division. - div = vmulq_f32(a, inv); - - return div; -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4i pdiv(const Packet4i& /*a*/, const Packet4i& /*b*/) -{ eigen_assert(false && "packet integer division are not supported by NEON"); - return pset1(0); -} - -// Clang/ARM wrongly advertises __ARM_FEATURE_FMA even when it's not available, -// then implements a slow software scalar fallback calling fmaf()! -// Filed LLVM bug: -// https://llvm.org/bugs/show_bug.cgi?id=27216 -#if (defined __ARM_FEATURE_FMA) && !(EIGEN_COMP_CLANG && EIGEN_ARCH_ARM) -// See bug 936. -// FMA is available on VFPv4 i.e. when compiling with -mfpu=neon-vfpv4. -// FMA is a true fused multiply-add i.e. only 1 rounding at the end, no intermediate rounding. -// MLA is not fused i.e. does 2 roundings. -// In addition to giving better accuracy, FMA also gives better performance here on a Krait (Nexus 4): -// MLA: 10 GFlop/s ; FMA: 12 GFlops/s. -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return vfmaq_f32(c,a,b); } -#else -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { -#if EIGEN_COMP_CLANG && EIGEN_ARCH_ARM - // Clang/ARM will replace VMLA by VMUL+VADD at least for some values of -mcpu, - // at least -mcpu=cortex-a8 and -mcpu=cortex-a7. Since the former is the default on - // -march=armv7-a, that is a very common case. - // See e.g. this thread: - // http://lists.llvm.org/pipermail/llvm-dev/2013-December/068806.html - // Filed LLVM bug: - // https://llvm.org/bugs/show_bug.cgi?id=27219 - Packet4f r = c; - asm volatile( - "vmla.f32 %q[r], %q[a], %q[b]" - : [r] "+w" (r) - : [a] "w" (a), - [b] "w" (b) - : ); - return r; -#else - return vmlaq_f32(c,a,b); -#endif -} -#endif - -// No FMA instruction for int, so use MLA unconditionally. -template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return vmlaq_s32(c,a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pmin(const Packet4f& a, const Packet4f& b) { return vminq_f32(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmin(const Packet4i& a, const Packet4i& b) { return vminq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pmax(const Packet4f& a, const Packet4f& b) { return vmaxq_f32(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmax(const Packet4i& a, const Packet4i& b) { return vmaxq_s32(a,b); } - -// Logical Operations are not supported for float, so we have to reinterpret casts using NEON intrinsics -template<> EIGEN_STRONG_INLINE Packet4f pand(const Packet4f& a, const Packet4f& b) -{ - return vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b))); -} -template<> EIGEN_STRONG_INLINE Packet4i pand(const Packet4i& a, const Packet4i& b) { return vandq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f por(const Packet4f& a, const Packet4f& b) -{ - return vreinterpretq_f32_u32(vorrq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b))); -} -template<> EIGEN_STRONG_INLINE Packet4i por(const Packet4i& a, const Packet4i& b) { return vorrq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pxor(const Packet4f& a, const Packet4f& b) -{ - return vreinterpretq_f32_u32(veorq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b))); -} -template<> EIGEN_STRONG_INLINE Packet4i pxor(const Packet4i& a, const Packet4i& b) { return veorq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pandnot(const Packet4f& a, const Packet4f& b) -{ - return vreinterpretq_f32_u32(vbicq_u32(vreinterpretq_u32_f32(a),vreinterpretq_u32_f32(b))); -} -template<> EIGEN_STRONG_INLINE Packet4i pandnot(const Packet4i& a, const Packet4i& b) { return vbicq_s32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pload(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f32(from); } -template<> EIGEN_STRONG_INLINE Packet4i pload(const int32_t* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_s32(from); } - -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f32(from); } -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int32_t* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_s32(from); } - -template<> EIGEN_STRONG_INLINE Packet4f ploaddup(const float* from) -{ - float32x2_t lo, hi; - lo = vld1_dup_f32(from); - hi = vld1_dup_f32(from+1); - return vcombine_f32(lo, hi); -} -template<> EIGEN_STRONG_INLINE Packet4i ploaddup(const int32_t* from) -{ - int32x2_t lo, hi; - lo = vld1_dup_s32(from); - hi = vld1_dup_s32(from+1); - return vcombine_s32(lo, hi); -} - -template<> EIGEN_STRONG_INLINE void pstore (float* to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f32(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(int32_t* to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_s32(to, from); } - -template<> EIGEN_STRONG_INLINE void pstoreu (float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(int32_t* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); } - -template<> EIGEN_DEVICE_FUNC inline Packet4f pgather(const float* from, Index stride) -{ - Packet4f res = pset1(0.f); - res = vsetq_lane_f32(from[0*stride], res, 0); - res = vsetq_lane_f32(from[1*stride], res, 1); - res = vsetq_lane_f32(from[2*stride], res, 2); - res = vsetq_lane_f32(from[3*stride], res, 3); - return res; -} -template<> EIGEN_DEVICE_FUNC inline Packet4i pgather(const int32_t* from, Index stride) -{ - Packet4i res = pset1(0); - res = vsetq_lane_s32(from[0*stride], res, 0); - res = vsetq_lane_s32(from[1*stride], res, 1); - res = vsetq_lane_s32(from[2*stride], res, 2); - res = vsetq_lane_s32(from[3*stride], res, 3); - return res; -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet4f& from, Index stride) -{ - to[stride*0] = vgetq_lane_f32(from, 0); - to[stride*1] = vgetq_lane_f32(from, 1); - to[stride*2] = vgetq_lane_f32(from, 2); - to[stride*3] = vgetq_lane_f32(from, 3); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(int32_t* to, const Packet4i& from, Index stride) -{ - to[stride*0] = vgetq_lane_s32(from, 0); - to[stride*1] = vgetq_lane_s32(from, 1); - to[stride*2] = vgetq_lane_s32(from, 2); - to[stride*3] = vgetq_lane_s32(from, 3); -} - -template<> EIGEN_STRONG_INLINE void prefetch (const float* addr) { EIGEN_ARM_PREFETCH(addr); } -template<> EIGEN_STRONG_INLINE void prefetch(const int32_t* addr) { EIGEN_ARM_PREFETCH(addr); } - -// FIXME only store the 2 first elements ? -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; } -template<> EIGEN_STRONG_INLINE int32_t pfirst(const Packet4i& a) { int32_t EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; } - -template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) { - float32x2_t a_lo, a_hi; - Packet4f a_r64; - - a_r64 = vrev64q_f32(a); - a_lo = vget_low_f32(a_r64); - a_hi = vget_high_f32(a_r64); - return vcombine_f32(a_hi, a_lo); -} -template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) { - int32x2_t a_lo, a_hi; - Packet4i a_r64; - - a_r64 = vrev64q_s32(a); - a_lo = vget_low_s32(a_r64); - a_hi = vget_high_s32(a_r64); - return vcombine_s32(a_hi, a_lo); -} - -template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) { return vabsq_f32(a); } -template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vabsq_s32(a); } - -template<> EIGEN_STRONG_INLINE float predux(const Packet4f& a) -{ - float32x2_t a_lo, a_hi, sum; - - a_lo = vget_low_f32(a); - a_hi = vget_high_f32(a); - sum = vpadd_f32(a_lo, a_hi); - sum = vpadd_f32(sum, sum); - return vget_lane_f32(sum, 0); -} - -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - float32x4x2_t vtrn1, vtrn2, res1, res2; - Packet4f sum1, sum2, sum; - - // NEON zip performs interleaving of the supplied vectors. - // We perform two interleaves in a row to acquire the transposed vector - vtrn1 = vzipq_f32(vecs[0], vecs[2]); - vtrn2 = vzipq_f32(vecs[1], vecs[3]); - res1 = vzipq_f32(vtrn1.val[0], vtrn2.val[0]); - res2 = vzipq_f32(vtrn1.val[1], vtrn2.val[1]); - - // Do the addition of the resulting vectors - sum1 = vaddq_f32(res1.val[0], res1.val[1]); - sum2 = vaddq_f32(res2.val[0], res2.val[1]); - sum = vaddq_f32(sum1, sum2); - - return sum; -} - -template<> EIGEN_STRONG_INLINE int32_t predux(const Packet4i& a) -{ - int32x2_t a_lo, a_hi, sum; - - a_lo = vget_low_s32(a); - a_hi = vget_high_s32(a); - sum = vpadd_s32(a_lo, a_hi); - sum = vpadd_s32(sum, sum); - return vget_lane_s32(sum, 0); -} - -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - int32x4x2_t vtrn1, vtrn2, res1, res2; - Packet4i sum1, sum2, sum; - - // NEON zip performs interleaving of the supplied vectors. - // We perform two interleaves in a row to acquire the transposed vector - vtrn1 = vzipq_s32(vecs[0], vecs[2]); - vtrn2 = vzipq_s32(vecs[1], vecs[3]); - res1 = vzipq_s32(vtrn1.val[0], vtrn2.val[0]); - res2 = vzipq_s32(vtrn1.val[1], vtrn2.val[1]); - - // Do the addition of the resulting vectors - sum1 = vaddq_s32(res1.val[0], res1.val[1]); - sum2 = vaddq_s32(res2.val[0], res2.val[1]); - sum = vaddq_s32(sum1, sum2); - - return sum; -} - -// Other reduction functions: -// mul -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet4f& a) -{ - float32x2_t a_lo, a_hi, prod; - - // Get a_lo = |a1|a2| and a_hi = |a3|a4| - a_lo = vget_low_f32(a); - a_hi = vget_high_f32(a); - // Get the product of a_lo * a_hi -> |a1*a3|a2*a4| - prod = vmul_f32(a_lo, a_hi); - // Multiply prod with its swapped value |a2*a4|a1*a3| - prod = vmul_f32(prod, vrev64_f32(prod)); - - return vget_lane_f32(prod, 0); -} -template<> EIGEN_STRONG_INLINE int32_t predux_mul(const Packet4i& a) -{ - int32x2_t a_lo, a_hi, prod; - - // Get a_lo = |a1|a2| and a_hi = |a3|a4| - a_lo = vget_low_s32(a); - a_hi = vget_high_s32(a); - // Get the product of a_lo * a_hi -> |a1*a3|a2*a4| - prod = vmul_s32(a_lo, a_hi); - // Multiply prod with its swapped value |a2*a4|a1*a3| - prod = vmul_s32(prod, vrev64_s32(prod)); - - return vget_lane_s32(prod, 0); -} - -// min -template<> EIGEN_STRONG_INLINE float predux_min(const Packet4f& a) -{ - float32x2_t a_lo, a_hi, min; - - a_lo = vget_low_f32(a); - a_hi = vget_high_f32(a); - min = vpmin_f32(a_lo, a_hi); - min = vpmin_f32(min, min); - - return vget_lane_f32(min, 0); -} - -template<> EIGEN_STRONG_INLINE int32_t predux_min(const Packet4i& a) -{ - int32x2_t a_lo, a_hi, min; - - a_lo = vget_low_s32(a); - a_hi = vget_high_s32(a); - min = vpmin_s32(a_lo, a_hi); - min = vpmin_s32(min, min); - - return vget_lane_s32(min, 0); -} - -// max -template<> EIGEN_STRONG_INLINE float predux_max(const Packet4f& a) -{ - float32x2_t a_lo, a_hi, max; - - a_lo = vget_low_f32(a); - a_hi = vget_high_f32(a); - max = vpmax_f32(a_lo, a_hi); - max = vpmax_f32(max, max); - - return vget_lane_f32(max, 0); -} - -template<> EIGEN_STRONG_INLINE int32_t predux_max(const Packet4i& a) -{ - int32x2_t a_lo, a_hi, max; - - a_lo = vget_low_s32(a); - a_hi = vget_high_s32(a); - max = vpmax_s32(a_lo, a_hi); - max = vpmax_s32(max, max); - - return vget_lane_s32(max, 0); -} - -// this PALIGN_NEON business is to work around a bug in LLVM Clang 3.0 causing incorrect compilation errors, -// see bug 347 and this LLVM bug: http://llvm.org/bugs/show_bug.cgi?id=11074 -#define PALIGN_NEON(Offset,Type,Command) \ -template<>\ -struct palign_impl\ -{\ - EIGEN_STRONG_INLINE static void run(Type& first, const Type& second)\ - {\ - if (Offset!=0)\ - first = Command(first, second, Offset);\ - }\ -};\ - -PALIGN_NEON(0,Packet4f,vextq_f32) -PALIGN_NEON(1,Packet4f,vextq_f32) -PALIGN_NEON(2,Packet4f,vextq_f32) -PALIGN_NEON(3,Packet4f,vextq_f32) -PALIGN_NEON(0,Packet4i,vextq_s32) -PALIGN_NEON(1,Packet4i,vextq_s32) -PALIGN_NEON(2,Packet4i,vextq_s32) -PALIGN_NEON(3,Packet4i,vextq_s32) - -#undef PALIGN_NEON - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - float32x4x2_t tmp1 = vzipq_f32(kernel.packet[0], kernel.packet[1]); - float32x4x2_t tmp2 = vzipq_f32(kernel.packet[2], kernel.packet[3]); - - kernel.packet[0] = vcombine_f32(vget_low_f32(tmp1.val[0]), vget_low_f32(tmp2.val[0])); - kernel.packet[1] = vcombine_f32(vget_high_f32(tmp1.val[0]), vget_high_f32(tmp2.val[0])); - kernel.packet[2] = vcombine_f32(vget_low_f32(tmp1.val[1]), vget_low_f32(tmp2.val[1])); - kernel.packet[3] = vcombine_f32(vget_high_f32(tmp1.val[1]), vget_high_f32(tmp2.val[1])); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - int32x4x2_t tmp1 = vzipq_s32(kernel.packet[0], kernel.packet[1]); - int32x4x2_t tmp2 = vzipq_s32(kernel.packet[2], kernel.packet[3]); - kernel.packet[0] = vcombine_s32(vget_low_s32(tmp1.val[0]), vget_low_s32(tmp2.val[0])); - kernel.packet[1] = vcombine_s32(vget_high_s32(tmp1.val[0]), vget_high_s32(tmp2.val[0])); - kernel.packet[2] = vcombine_s32(vget_low_s32(tmp1.val[1]), vget_low_s32(tmp2.val[1])); - kernel.packet[3] = vcombine_s32(vget_high_s32(tmp1.val[1]), vget_high_s32(tmp2.val[1])); -} - -//---------- double ---------- - -// Clang 3.5 in the iOS toolchain has an ICE triggered by NEON intrisics for double. -// Confirmed at least with __apple_build_version__ = 6000054. -#ifdef __apple_build_version__ -// Let's hope that by the time __apple_build_version__ hits the 601* range, the bug will be fixed. -// https://gist.github.com/yamaya/2924292 suggests that the 3 first digits are only updated with -// major toolchain updates. -#define EIGEN_APPLE_DOUBLE_NEON_BUG (__apple_build_version__ < 6010000) -#else -#define EIGEN_APPLE_DOUBLE_NEON_BUG 0 -#endif - -#if EIGEN_ARCH_ARM64 && !EIGEN_APPLE_DOUBLE_NEON_BUG - -// Bug 907: workaround missing declarations of the following two functions in the ADK -// Defining these functions as templates ensures that if these intrinsics are -// already defined in arm_neon.h, then our workaround doesn't cause a conflict -// and has lower priority in overload resolution. -template -uint64x2_t vreinterpretq_u64_f64(T a) -{ - return (uint64x2_t) a; -} - -template -float64x2_t vreinterpretq_f64_u64(T a) -{ - return (float64x2_t) a; -} - -typedef float64x2_t Packet2d; -typedef float64x1_t Packet1d; - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet2d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket=0, - - HasDiv = 1, - // FIXME check the Has* - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 0, - HasSqrt = 0 - }; -}; - -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; - -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { return vdupq_n_f64(from); } - -template<> EIGEN_STRONG_INLINE Packet2d plset(const double& a) -{ - const double countdown_raw[] = {0.0,1.0}; - const Packet2d countdown = vld1q_f64(countdown_raw); - return vaddq_f64(pset1(a), countdown); -} -template<> EIGEN_STRONG_INLINE Packet2d padd(const Packet2d& a, const Packet2d& b) { return vaddq_f64(a,b); } - -template<> EIGEN_STRONG_INLINE Packet2d psub(const Packet2d& a, const Packet2d& b) { return vsubq_f64(a,b); } - -template<> EIGEN_STRONG_INLINE Packet2d pnegate(const Packet2d& a) { return vnegq_f64(a); } - -template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet2d pmul(const Packet2d& a, const Packet2d& b) { return vmulq_f64(a,b); } - -template<> EIGEN_STRONG_INLINE Packet2d pdiv(const Packet2d& a, const Packet2d& b) { return vdivq_f64(a,b); } - -#ifdef __ARM_FEATURE_FMA -// See bug 936. See above comment about FMA for float. -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return vfmaq_f64(c,a,b); } -#else -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return vmlaq_f64(c,a,b); } -#endif - -template<> EIGEN_STRONG_INLINE Packet2d pmin(const Packet2d& a, const Packet2d& b) { return vminq_f64(a,b); } - -template<> EIGEN_STRONG_INLINE Packet2d pmax(const Packet2d& a, const Packet2d& b) { return vmaxq_f64(a,b); } - -// Logical Operations are not supported for float, so we have to reinterpret casts using NEON intrinsics -template<> EIGEN_STRONG_INLINE Packet2d pand(const Packet2d& a, const Packet2d& b) -{ - return vreinterpretq_f64_u64(vandq_u64(vreinterpretq_u64_f64(a),vreinterpretq_u64_f64(b))); -} - -template<> EIGEN_STRONG_INLINE Packet2d por(const Packet2d& a, const Packet2d& b) -{ - return vreinterpretq_f64_u64(vorrq_u64(vreinterpretq_u64_f64(a),vreinterpretq_u64_f64(b))); -} - -template<> EIGEN_STRONG_INLINE Packet2d pxor(const Packet2d& a, const Packet2d& b) -{ - return vreinterpretq_f64_u64(veorq_u64(vreinterpretq_u64_f64(a),vreinterpretq_u64_f64(b))); -} - -template<> EIGEN_STRONG_INLINE Packet2d pandnot(const Packet2d& a, const Packet2d& b) -{ - return vreinterpretq_f64_u64(vbicq_u64(vreinterpretq_u64_f64(a),vreinterpretq_u64_f64(b))); -} - -template<> EIGEN_STRONG_INLINE Packet2d pload(const double* from) { EIGEN_DEBUG_ALIGNED_LOAD return vld1q_f64(from); } - -template<> EIGEN_STRONG_INLINE Packet2d ploadu(const double* from) { EIGEN_DEBUG_UNALIGNED_LOAD return vld1q_f64(from); } - -template<> EIGEN_STRONG_INLINE Packet2d ploaddup(const double* from) -{ - return vld1q_dup_f64(from); -} -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet2d& from) { EIGEN_DEBUG_ALIGNED_STORE vst1q_f64(to, from); } - -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet2d& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f64(to, from); } - -template<> EIGEN_DEVICE_FUNC inline Packet2d pgather(const double* from, Index stride) -{ - Packet2d res = pset1(0.0); - res = vsetq_lane_f64(from[0*stride], res, 0); - res = vsetq_lane_f64(from[1*stride], res, 1); - return res; -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet2d& from, Index stride) -{ - to[stride*0] = vgetq_lane_f64(from, 0); - to[stride*1] = vgetq_lane_f64(from, 1); -} -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { EIGEN_ARM_PREFETCH(addr); } - -// FIXME only store the 2 first elements ? -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { return vgetq_lane_f64(a, 0); } - -template<> EIGEN_STRONG_INLINE Packet2d preverse(const Packet2d& a) { return vcombine_f64(vget_high_f64(a), vget_low_f64(a)); } - -template<> EIGEN_STRONG_INLINE Packet2d pabs(const Packet2d& a) { return vabsq_f64(a); } - -#if EIGEN_COMP_CLANG && defined(__apple_build_version__) -// workaround ICE, see bug 907 -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) { return (vget_low_f64(a) + vget_high_f64(a))[0]; } -#else -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) { return vget_lane_f64(vget_low_f64(a) + vget_high_f64(a), 0); } -#endif - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - float64x2_t trn1, trn2; - - // NEON zip performs interleaving of the supplied vectors. - // We perform two interleaves in a row to acquire the transposed vector - trn1 = vzip1q_f64(vecs[0], vecs[1]); - trn2 = vzip2q_f64(vecs[0], vecs[1]); - - // Do the addition of the resulting vectors - return vaddq_f64(trn1, trn2); -} -// Other reduction functions: -// mul -#if EIGEN_COMP_CLANG && defined(__apple_build_version__) -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) { return (vget_low_f64(a) * vget_high_f64(a))[0]; } -#else -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) { return vget_lane_f64(vget_low_f64(a) * vget_high_f64(a), 0); } -#endif - -// min -template<> EIGEN_STRONG_INLINE double predux_min(const Packet2d& a) { return vgetq_lane_f64(vpminq_f64(a, a), 0); } - -// max -template<> EIGEN_STRONG_INLINE double predux_max(const Packet2d& a) { return vgetq_lane_f64(vpmaxq_f64(a, a), 0); } - -// this PALIGN_NEON business is to work around a bug in LLVM Clang 3.0 causing incorrect compilation errors, -// see bug 347 and this LLVM bug: http://llvm.org/bugs/show_bug.cgi?id=11074 -#define PALIGN_NEON(Offset,Type,Command) \ -template<>\ -struct palign_impl\ -{\ - EIGEN_STRONG_INLINE static void run(Type& first, const Type& second)\ - {\ - if (Offset!=0)\ - first = Command(first, second, Offset);\ - }\ -};\ - -PALIGN_NEON(0,Packet2d,vextq_f64) -PALIGN_NEON(1,Packet2d,vextq_f64) -#undef PALIGN_NEON - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - float64x2_t trn1 = vzip1q_f64(kernel.packet[0], kernel.packet[1]); - float64x2_t trn2 = vzip2q_f64(kernel.packet[0], kernel.packet[1]); - - kernel.packet[0] = trn1; - kernel.packet[1] = trn2; -} -#endif // EIGEN_ARCH_ARM64 - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PACKET_MATH_NEON_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h deleted file mode 100644 index d075043ce1b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h +++ /dev/null @@ -1,471 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_SSE_H -#define EIGEN_COMPLEX_SSE_H - -namespace Eigen { - -namespace internal { - -//---------- float ---------- -struct Packet2cf -{ - EIGEN_STRONG_INLINE Packet2cf() {} - EIGEN_STRONG_INLINE explicit Packet2cf(const __m128& a) : v(a) {} - __m128 v; -}; - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0, - HasBlend = 1 - }; -}; -#endif - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; - -template<> EIGEN_STRONG_INLINE Packet2cf padd(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf psub(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) -{ - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); - return Packet2cf(_mm_xor_ps(a.v,mask)); -} -template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) -{ - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_xor_ps(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) -{ - #ifdef EIGEN_VECTORIZE_SSE3 - return Packet2cf(_mm_addsub_ps(_mm_mul_ps(_mm_moveldup_ps(a.v), b.v), - _mm_mul_ps(_mm_movehdup_ps(a.v), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); -// return Packet2cf(_mm_addsub_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), -// _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), -// vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x00000000,0x80000000,0x00000000)); - return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), - _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); - #endif -} - -template<> EIGEN_STRONG_INLINE Packet2cf pand (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_and_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf por (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_or_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pxor (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pandnot(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(a.v,b.v)); } - -template<> EIGEN_STRONG_INLINE Packet2cf pload (const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload(&numext::real_ref(*from))); } -template<> EIGEN_STRONG_INLINE Packet2cf ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu(&numext::real_ref(*from))); } - -template<> EIGEN_STRONG_INLINE Packet2cf pset1(const std::complex& from) -{ - Packet2cf res; -#if EIGEN_GNUC_AT_MOST(4,2) - // Workaround annoying "may be used uninitialized in this function" warning with gcc 4.2 - res.v = _mm_loadl_pi(_mm_set1_ps(0.0f), reinterpret_cast(&from)); -#elif EIGEN_GNUC_AT_LEAST(4,6) - // Suppress annoying "may be used uninitialized in this function" warning with gcc >= 4.6 - #pragma GCC diagnostic push - #pragma GCC diagnostic ignored "-Wuninitialized" - res.v = _mm_loadl_pi(res.v, (const __m64*)&from); - #pragma GCC diagnostic pop -#else - res.v = _mm_loadl_pi(res.v, (const __m64*)&from); -#endif - return Packet2cf(_mm_movelh_ps(res.v,res.v)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), Packet4f(from.v)); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), Packet4f(from.v)); } - - -template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather, Packet2cf>(const std::complex* from, Index stride) -{ - return Packet2cf(_mm_set_ps(std::imag(from[1*stride]), std::real(from[1*stride]), - std::imag(from[0*stride]), std::real(from[0*stride]))); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cf>(std::complex* to, const Packet2cf& from, Index stride) -{ - to[stride*0] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 0)), - _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 1))); - to[stride*1] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 2)), - _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 3))); -} - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cf& a) -{ - #if EIGEN_GNUC_AT_MOST(4,3) - // Workaround gcc 4.2 ICE - this is not performance wise ideal, but who cares... - // This workaround also fix invalid code generation with gcc 4.3 - EIGEN_ALIGN16 std::complex res[2]; - _mm_store_ps((float*)res, a.v); - return res[0]; - #else - std::complex res; - _mm_storel_pi((__m64*)&res, a.v); - return res; - #endif -} - -template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) { return Packet2cf(_mm_castpd_ps(preverse(Packet2d(_mm_castps_pd(a.v))))); } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cf& a) -{ - return pfirst(Packet2cf(_mm_add_ps(a.v, _mm_movehl_ps(a.v,a.v)))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf preduxp(const Packet2cf* vecs) -{ - return Packet2cf(_mm_add_ps(_mm_movelh_ps(vecs[0].v,vecs[1].v), _mm_movehl_ps(vecs[1].v,vecs[0].v))); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cf& a) -{ - return pfirst(pmul(a, Packet2cf(_mm_movehl_ps(a.v,a.v)))); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) - { - if (Offset==1) - { - first.v = _mm_movehl_ps(first.v, first.v); - first.v = _mm_movelh_ps(first.v, second.v); - } - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(a, pconj(b)); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_add_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), - _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(pconj(a), b); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), - _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return pconj(internal::pmul(a, b)); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_sub_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), - _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #endif - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) - -template<> EIGEN_STRONG_INLINE Packet2cf pdiv(const Packet2cf& a, const Packet2cf& b) -{ - // TODO optimize it for SSE3 and 4 - Packet2cf res = conj_helper().pmul(a,b); - __m128 s = _mm_mul_ps(b.v,b.v); - return Packet2cf(_mm_div_ps(res.v,_mm_add_ps(s,_mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(s), 0xb1))))); -} - -EIGEN_STRONG_INLINE Packet2cf pcplxflip/* */(const Packet2cf& x) -{ - return Packet2cf(vec4f_swizzle1(x.v, 1, 0, 3, 2)); -} - - -//---------- double ---------- -struct Packet1cd -{ - EIGEN_STRONG_INLINE Packet1cd() {} - EIGEN_STRONG_INLINE explicit Packet1cd(const __m128d& a) : v(a) {} - __m128d v; -}; - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet1cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 0, - size = 1, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; -#endif - -template<> struct unpacket_traits { typedef std::complex type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; - -template<> EIGEN_STRONG_INLINE Packet1cd padd(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd psub(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } -template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) -{ - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_xor_pd(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) -{ - #ifdef EIGEN_VECTORIZE_SSE3 - return Packet1cd(_mm_addsub_pd(_mm_mul_pd(_mm_movedup_pd(a.v), b.v), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); - return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), - _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)), mask))); - #endif -} - -template<> EIGEN_STRONG_INLINE Packet1cd pand (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_and_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd por (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_or_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pxor (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_xor_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pandnot(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_andnot_pd(a.v,b.v)); } - -// FIXME force unaligned load, this is a temporary fix -template<> EIGEN_STRONG_INLINE Packet1cd pload (const std::complex* from) -{ EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd ploadu(const std::complex* from) -{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd pset1(const std::complex& from) -{ /* here we really have to use unaligned loads :( */ return ploadu(&from); } - -template<> EIGEN_STRONG_INLINE Packet1cd ploaddup(const std::complex* from) { return pset1(*from); } - -// FIXME force unaligned store, this is a temporary fix -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, Packet2d(from.v)); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, Packet2d(from.v)); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet1cd& a) -{ - EIGEN_ALIGN16 double res[2]; - _mm_store_pd(res, a.v); - return std::complex(res[0],res[1]); -} - -template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet1cd& a) -{ - return pfirst(a); -} - -template<> EIGEN_STRONG_INLINE Packet1cd preduxp(const Packet1cd* vecs) -{ - return vecs[0]; -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet1cd& a) -{ - return pfirst(a); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) - { - // FIXME is it sure we never have to align a Packet1cd? - // Even though a std::complex has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(a, pconj(b)); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_add_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(pconj(a), b); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), - _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)), mask))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return pconj(internal::pmul(a, b)); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_sub_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #endif - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) - -template<> EIGEN_STRONG_INLINE Packet1cd pdiv(const Packet1cd& a, const Packet1cd& b) -{ - // TODO optimize it for SSE3 and 4 - Packet1cd res = conj_helper().pmul(a,b); - __m128d s = _mm_mul_pd(b.v,b.v); - return Packet1cd(_mm_div_pd(res.v, _mm_add_pd(s,_mm_shuffle_pd(s, s, 0x1)))); -} - -EIGEN_STRONG_INLINE Packet1cd pcplxflip/* */(const Packet1cd& x) -{ - return Packet1cd(preverse(Packet2d(x.v))); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128d w1 = _mm_castps_pd(kernel.packet[0].v); - __m128d w2 = _mm_castps_pd(kernel.packet[1].v); - - __m128 tmp = _mm_castpd_ps(_mm_unpackhi_pd(w1, w2)); - kernel.packet[0].v = _mm_castpd_ps(_mm_unpacklo_pd(w1, w2)); - kernel.packet[1].v = tmp; -} - -template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { - __m128d result = pblend(ifPacket, _mm_castps_pd(thenPacket.v), _mm_castps_pd(elsePacket.v)); - return Packet2cf(_mm_castpd_ps(result)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pinsertfirst(const Packet2cf& a, std::complex b) -{ - return Packet2cf(_mm_loadl_pi(a.v, reinterpret_cast(&b))); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pinsertfirst(const Packet1cd&, std::complex b) -{ - return pset1(b); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pinsertlast(const Packet2cf& a, std::complex b) -{ - return Packet2cf(_mm_loadh_pi(a.v, reinterpret_cast(&b))); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pinsertlast(const Packet1cd&, std::complex b) -{ - return pset1(b); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h deleted file mode 100644 index 7b5f948e119..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h +++ /dev/null @@ -1,562 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007 Julien Pommier -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* The sin, cos, exp, and log functions of this file come from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -#ifndef EIGEN_MATH_FUNCTIONS_SSE_H -#define EIGEN_MATH_FUNCTIONS_SSE_H - -namespace Eigen { - -namespace internal { - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f plog(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); - - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(inv_mant_mask, ~0x7f800000); - - /* the smallest non denormalized float number */ - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(min_norm_pos, 0x00800000); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(minus_inf, 0xff800000);//-1.f/0.f); - - /* natural logarithm computed for 4 simultaneous float - return NaN for x <= 0 - */ - _EIGEN_DECLARE_CONST_Packet4f(cephes_SQRTHF, 0.707106781186547524f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p0, 7.0376836292E-2f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p1, - 1.1514610310E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p2, 1.1676998740E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p3, - 1.2420140846E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p4, + 1.4249322787E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p5, - 1.6668057665E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p6, + 2.0000714765E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p7, - 2.4999993993E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p8, + 3.3333331174E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q1, -2.12194440e-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q2, 0.693359375f); - - - Packet4i emm0; - - Packet4f invalid_mask = _mm_cmpnge_ps(x, _mm_setzero_ps()); // not greater equal is true if x is NaN - Packet4f iszero_mask = _mm_cmpeq_ps(x, _mm_setzero_ps()); - - x = pmax(x, p4f_min_norm_pos); /* cut off denormalized stuff */ - emm0 = _mm_srli_epi32(_mm_castps_si128(x), 23); - - /* keep only the fractional part */ - x = _mm_and_ps(x, p4f_inv_mant_mask); - x = _mm_or_ps(x, p4f_half); - - emm0 = _mm_sub_epi32(emm0, p4i_0x7f); - Packet4f e = padd(Packet4f(_mm_cvtepi32_ps(emm0)), p4f_1); - - /* part2: - if( x < SQRTHF ) { - e -= 1; - x = x + x - 1.0; - } else { x = x - 1.0; } - */ - Packet4f mask = _mm_cmplt_ps(x, p4f_cephes_SQRTHF); - Packet4f tmp = pand(x, mask); - x = psub(x, p4f_1); - e = psub(e, pand(p4f_1, mask)); - x = padd(x, tmp); - - Packet4f x2 = pmul(x,x); - Packet4f x3 = pmul(x2,x); - - Packet4f y, y1, y2; - y = pmadd(p4f_cephes_log_p0, x, p4f_cephes_log_p1); - y1 = pmadd(p4f_cephes_log_p3, x, p4f_cephes_log_p4); - y2 = pmadd(p4f_cephes_log_p6, x, p4f_cephes_log_p7); - y = pmadd(y , x, p4f_cephes_log_p2); - y1 = pmadd(y1, x, p4f_cephes_log_p5); - y2 = pmadd(y2, x, p4f_cephes_log_p8); - y = pmadd(y, x3, y1); - y = pmadd(y, x3, y2); - y = pmul(y, x3); - - y1 = pmul(e, p4f_cephes_log_q1); - tmp = pmul(x2, p4f_half); - y = padd(y, y1); - x = psub(x, tmp); - y2 = pmul(e, p4f_cephes_log_q2); - x = padd(x, y); - x = padd(x, y2); - // negative arg will be NAN, 0 will be -INF - return _mm_or_ps(_mm_andnot_ps(iszero_mask, _mm_or_ps(x, invalid_mask)), - _mm_and_ps(iszero_mask, p4f_minus_inf)); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pexp(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); - - - _EIGEN_DECLARE_CONST_Packet4f(exp_hi, 88.3762626647950f); - _EIGEN_DECLARE_CONST_Packet4f(exp_lo, -88.3762626647949f); - - _EIGEN_DECLARE_CONST_Packet4f(cephes_LOG2EF, 1.44269504088896341f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C1, 0.693359375f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C2, -2.12194440e-4f); - - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p0, 1.9875691500E-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p1, 1.3981999507E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p2, 8.3334519073E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p3, 4.1665795894E-2f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201E-1f); - - Packet4f tmp, fx; - Packet4i emm0; - - // clamp x - x = pmax(pmin(x, p4f_exp_hi), p4f_exp_lo); - - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(x, p4f_cephes_LOG2EF, p4f_half); - -#ifdef EIGEN_VECTORIZE_SSE4_1 - fx = _mm_floor_ps(fx); -#else - emm0 = _mm_cvttps_epi32(fx); - tmp = _mm_cvtepi32_ps(emm0); - /* if greater, substract 1 */ - Packet4f mask = _mm_cmpgt_ps(tmp, fx); - mask = _mm_and_ps(mask, p4f_1); - fx = psub(tmp, mask); -#endif - - tmp = pmul(fx, p4f_cephes_exp_C1); - Packet4f z = pmul(fx, p4f_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - z = pmul(x,x); - - Packet4f y = p4f_cephes_exp_p0; - y = pmadd(y, x, p4f_cephes_exp_p1); - y = pmadd(y, x, p4f_cephes_exp_p2); - y = pmadd(y, x, p4f_cephes_exp_p3); - y = pmadd(y, x, p4f_cephes_exp_p4); - y = pmadd(y, x, p4f_cephes_exp_p5); - y = pmadd(y, z, x); - y = padd(y, p4f_1); - - // build 2^n - emm0 = _mm_cvttps_epi32(fx); - emm0 = _mm_add_epi32(emm0, p4i_0x7f); - emm0 = _mm_slli_epi32(emm0, 23); - return pmax(pmul(y, Packet4f(_mm_castsi128_ps(emm0))), _x); -} -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d pexp(const Packet2d& _x) -{ - Packet2d x = _x; - - _EIGEN_DECLARE_CONST_Packet2d(1 , 1.0); - _EIGEN_DECLARE_CONST_Packet2d(2 , 2.0); - _EIGEN_DECLARE_CONST_Packet2d(half, 0.5); - - _EIGEN_DECLARE_CONST_Packet2d(exp_hi, 709.437); - _EIGEN_DECLARE_CONST_Packet2d(exp_lo, -709.436139303); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_LOG2EF, 1.4426950408889634073599); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p0, 1.26177193074810590878e-4); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p1, 3.02994407707441961300e-2); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p2, 9.99999999999999999910e-1); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q0, 3.00198505138664455042e-6); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q1, 2.52448340349684104192e-3); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q2, 2.27265548208155028766e-1); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q3, 2.00000000000000000009e0); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C1, 0.693145751953125); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C2, 1.42860682030941723212e-6); - static const __m128i p4i_1023_0 = _mm_setr_epi32(1023, 1023, 0, 0); - - Packet2d tmp, fx; - Packet4i emm0; - - // clamp x - x = pmax(pmin(x, p2d_exp_hi), p2d_exp_lo); - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(p2d_cephes_LOG2EF, x, p2d_half); - -#ifdef EIGEN_VECTORIZE_SSE4_1 - fx = _mm_floor_pd(fx); -#else - emm0 = _mm_cvttpd_epi32(fx); - tmp = _mm_cvtepi32_pd(emm0); - /* if greater, substract 1 */ - Packet2d mask = _mm_cmpgt_pd(tmp, fx); - mask = _mm_and_pd(mask, p2d_1); - fx = psub(tmp, mask); -#endif - - tmp = pmul(fx, p2d_cephes_exp_C1); - Packet2d z = pmul(fx, p2d_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - Packet2d x2 = pmul(x,x); - - Packet2d px = p2d_cephes_exp_p0; - px = pmadd(px, x2, p2d_cephes_exp_p1); - px = pmadd(px, x2, p2d_cephes_exp_p2); - px = pmul (px, x); - - Packet2d qx = p2d_cephes_exp_q0; - qx = pmadd(qx, x2, p2d_cephes_exp_q1); - qx = pmadd(qx, x2, p2d_cephes_exp_q2); - qx = pmadd(qx, x2, p2d_cephes_exp_q3); - - x = pdiv(px,psub(qx,px)); - x = pmadd(p2d_2,x,p2d_1); - - // build 2^n - emm0 = _mm_cvttpd_epi32(fx); - emm0 = _mm_add_epi32(emm0, p4i_1023_0); - emm0 = _mm_slli_epi32(emm0, 20); - emm0 = _mm_shuffle_epi32(emm0, _MM_SHUFFLE(1,2,0,3)); - return pmax(pmul(x, Packet2d(_mm_castsi128_pd(emm0))), _x); -} - -/* evaluation of 4 sines at onces, using SSE2 intrinsics. - - The code is the exact rewriting of the cephes sinf function. - Precision is excellent as long as x < 8192 (I did not bother to - take into account the special handling they have for greater values - -- it does not return garbage for arguments over 8192, though, but - the extra precision is missing). - - Note that it is such that sinf((float)M_PI) = 8.74e-8, which is the - surprising but correct result. -*/ - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psin(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - - _EIGEN_DECLARE_CONST_Packet4i(1, 1); - _EIGEN_DECLARE_CONST_Packet4i(not1, ~1); - _EIGEN_DECLARE_CONST_Packet4i(2, 2); - _EIGEN_DECLARE_CONST_Packet4i(4, 4); - - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(sign_mask, 0x80000000); - - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP1,-0.78515625f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP2, -2.4187564849853515625e-4f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP3, -3.77489497744594108e-8f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p0, -1.9515295891E-4f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p1, 8.3321608736E-3f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p2, -1.6666654611E-1f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p0, 2.443315711809948E-005f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p1, -1.388731625493765E-003f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p2, 4.166664568298827E-002f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI - - Packet4f xmm1, xmm2, xmm3, sign_bit, y; - - Packet4i emm0, emm2; - sign_bit = x; - /* take the absolute value */ - x = pabs(x); - - /* take the modulo */ - - /* extract the sign bit (upper one) */ - sign_bit = _mm_and_ps(sign_bit, p4f_sign_mask); - - /* scale by 4/Pi */ - y = pmul(x, p4f_cephes_FOPI); - - /* store the integer part of y in mm0 */ - emm2 = _mm_cvttps_epi32(y); - /* j=(j+1) & (~1) (see the cephes sources) */ - emm2 = _mm_add_epi32(emm2, p4i_1); - emm2 = _mm_and_si128(emm2, p4i_not1); - y = _mm_cvtepi32_ps(emm2); - /* get the swap sign flag */ - emm0 = _mm_and_si128(emm2, p4i_4); - emm0 = _mm_slli_epi32(emm0, 29); - /* get the polynom selection mask - there is one polynom for 0 <= x <= Pi/4 - and another one for Pi/4 EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pcos(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - - _EIGEN_DECLARE_CONST_Packet4i(1, 1); - _EIGEN_DECLARE_CONST_Packet4i(not1, ~1); - _EIGEN_DECLARE_CONST_Packet4i(2, 2); - _EIGEN_DECLARE_CONST_Packet4i(4, 4); - - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP1,-0.78515625f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP2, -2.4187564849853515625e-4f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP3, -3.77489497744594108e-8f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p0, -1.9515295891E-4f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p1, 8.3321608736E-3f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p2, -1.6666654611E-1f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p0, 2.443315711809948E-005f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p1, -1.388731625493765E-003f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p2, 4.166664568298827E-002f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI - - Packet4f xmm1, xmm2, xmm3, y; - Packet4i emm0, emm2; - - x = pabs(x); - - /* scale by 4/Pi */ - y = pmul(x, p4f_cephes_FOPI); - - /* get the integer part of y */ - emm2 = _mm_cvttps_epi32(y); - /* j=(j+1) & (~1) (see the cephes sources) */ - emm2 = _mm_add_epi32(emm2, p4i_1); - emm2 = _mm_and_si128(emm2, p4i_not1); - y = _mm_cvtepi32_ps(emm2); - - emm2 = _mm_sub_epi32(emm2, p4i_2); - - /* get the swap sign flag */ - emm0 = _mm_andnot_si128(emm2, p4i_4); - emm0 = _mm_slli_epi32(emm0, 29); - /* get the polynom selection mask */ - emm2 = _mm_and_si128(emm2, p4i_2); - emm2 = _mm_cmpeq_epi32(emm2, _mm_setzero_si128()); - - Packet4f sign_bit = _mm_castsi128_ps(emm0); - Packet4f poly_mask = _mm_castsi128_ps(emm2); - - /* The magic pass: "Extended precision modular arithmetic" - x = ((x - y * DP1) - y * DP2) - y * DP3; */ - xmm1 = pmul(y, p4f_minus_cephes_DP1); - xmm2 = pmul(y, p4f_minus_cephes_DP2); - xmm3 = pmul(y, p4f_minus_cephes_DP3); - x = padd(x, xmm1); - x = padd(x, xmm2); - x = padd(x, xmm3); - - /* Evaluate the first polynom (0 <= x <= Pi/4) */ - y = p4f_coscof_p0; - Packet4f z = pmul(x,x); - - y = pmadd(y,z,p4f_coscof_p1); - y = pmadd(y,z,p4f_coscof_p2); - y = pmul(y, z); - y = pmul(y, z); - Packet4f tmp = _mm_mul_ps(z, p4f_half); - y = psub(y, tmp); - y = padd(y, p4f_1); - - /* Evaluate the second polynom (Pi/4 <= x <= 0) */ - Packet4f y2 = p4f_sincof_p0; - y2 = pmadd(y2, z, p4f_sincof_p1); - y2 = pmadd(y2, z, p4f_sincof_p2); - y2 = pmul(y2, z); - y2 = pmadd(y2, x, x); - - /* select the correct result from the two polynoms */ - y2 = _mm_and_ps(poly_mask, y2); - y = _mm_andnot_ps(poly_mask, y); - y = _mm_or_ps(y,y2); - - /* update the sign */ - return _mm_xor_ps(y, sign_bit); -} - -#if EIGEN_FAST_MATH - -// Functions for sqrt. -// The EIGEN_FAST_MATH version uses the _mm_rsqrt_ps approximation and one step -// of Newton's method, at a cost of 1-2 bits of precision as opposed to the -// exact solution. It does not handle +inf, or denormalized numbers correctly. -// The main advantage of this approach is not just speed, but also the fact that -// it can be inlined and pipelined with other computations, further reducing its -// effective latency. This is similar to Quake3's fast inverse square root. -// For detail see here: http://www.beyond3d.com/content/articles/8/ -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& _x) -{ - Packet4f half = pmul(_x, pset1(.5f)); - Packet4f denormal_mask = _mm_and_ps( - _mm_cmpge_ps(_x, _mm_setzero_ps()), - _mm_cmplt_ps(_x, pset1((std::numeric_limits::min)()))); - - // Compute approximate reciprocal sqrt. - Packet4f x = _mm_rsqrt_ps(_x); - // Do a single step of Newton's iteration. - x = pmul(x, psub(pset1(1.5f), pmul(half, pmul(x,x)))); - // Flush results for denormals to zero. - return _mm_andnot_ps(denormal_mask, pmul(_x,x)); -} - -#else - -template<>EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& x) { return _mm_sqrt_ps(x); } - -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d psqrt(const Packet2d& x) { return _mm_sqrt_pd(x); } - -#if EIGEN_FAST_MATH - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& _x) { - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(inf, 0x7f800000); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(nan, 0x7fc00000); - _EIGEN_DECLARE_CONST_Packet4f(one_point_five, 1.5f); - _EIGEN_DECLARE_CONST_Packet4f(minus_half, -0.5f); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(flt_min, 0x00800000); - - Packet4f neg_half = pmul(_x, p4f_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - Packet4f le_zero_mask = _mm_cmple_ps(_x, p4f_flt_min); - Packet4f x = _mm_andnot_ps(le_zero_mask, _mm_rsqrt_ps(_x)); - - // Fill in NaNs and Infs for the negative/zero entries. - Packet4f neg_mask = _mm_cmplt_ps(_x, _mm_setzero_ps()); - Packet4f zero_mask = _mm_andnot_ps(neg_mask, le_zero_mask); - Packet4f infs_and_nans = _mm_or_ps(_mm_and_ps(neg_mask, p4f_nan), - _mm_and_ps(zero_mask, p4f_inf)); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p4f_one_point_five)); - - // Insert NaNs and Infs in all the right places. - return _mm_or_ps(x, infs_and_nans); -} - -#else - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& x) { - // Unfortunately we can't use the much faster mm_rqsrt_ps since it only provides an approximation. - return _mm_div_ps(pset1(1.0f), _mm_sqrt_ps(x)); -} - -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d prsqrt(const Packet2d& x) { - // Unfortunately we can't use the much faster mm_rqsrt_pd since it only provides an approximation. - return _mm_div_pd(pset1(1.0), _mm_sqrt_pd(x)); -} - -// Hyperbolic Tangent function. -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet4f -ptanh(const Packet4f& x) { - return internal::generic_fast_tanh_float(x); -} - -} // end namespace internal - -namespace numext { - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -float sqrt(const float &x) -{ - return internal::pfirst(internal::Packet4f(_mm_sqrt_ss(_mm_set_ss(x)))); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -double sqrt(const double &x) -{ -#if EIGEN_COMP_GNUC_STRICT - // This works around a GCC bug generating poor code for _mm_sqrt_pd - // See https://bitbucket.org/eigen/eigen/commits/14f468dba4d350d7c19c9b93072e19f7b3df563b - return internal::pfirst(internal::Packet2d(__builtin_ia32_sqrtsd(_mm_set_sd(x)))); -#else - return internal::pfirst(internal::Packet2d(_mm_sqrt_pd(_mm_set_sd(x)))); -#endif -} - -} // end namespace numex - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h deleted file mode 100644 index 60e2517e4bd..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h +++ /dev/null @@ -1,895 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_SSE_H -#define EIGEN_PACKET_MATH_SSE_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS (2*sizeof(void*)) -#endif - -#ifdef __FMA__ -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD 1 -#endif -#endif - -#if ((defined EIGEN_VECTORIZE_AVX) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_MINGW) && (__GXX_ABI_VERSION < 1004)) || EIGEN_OS_QNX -// With GCC's default ABI version, a __m128 or __m256 are the same types and therefore we cannot -// have overloads for both types without linking error. -// One solution is to increase ABI version using -fabi-version=4 (or greater). -// Otherwise, we workaround this inconvenience by wrapping 128bit types into the following helper -// structure: -template -struct eigen_packet_wrapper -{ - EIGEN_ALWAYS_INLINE operator T&() { return m_val; } - EIGEN_ALWAYS_INLINE operator const T&() const { return m_val; } - EIGEN_ALWAYS_INLINE eigen_packet_wrapper() {} - EIGEN_ALWAYS_INLINE eigen_packet_wrapper(const T &v) : m_val(v) {} - EIGEN_ALWAYS_INLINE eigen_packet_wrapper& operator=(const T &v) { - m_val = v; - return *this; - } - - T m_val; -}; -typedef eigen_packet_wrapper<__m128> Packet4f; -typedef eigen_packet_wrapper<__m128i> Packet4i; -typedef eigen_packet_wrapper<__m128d> Packet2d; -#else -typedef __m128 Packet4f; -typedef __m128i Packet4i; -typedef __m128d Packet2d; -#endif - -template<> struct is_arithmetic<__m128> { enum { value = true }; }; -template<> struct is_arithmetic<__m128i> { enum { value = true }; }; -template<> struct is_arithmetic<__m128d> { enum { value = true }; }; - -#define vec4f_swizzle1(v,p,q,r,s) \ - (_mm_castsi128_ps(_mm_shuffle_epi32( _mm_castps_si128(v), ((s)<<6|(r)<<4|(q)<<2|(p))))) - -#define vec4i_swizzle1(v,p,q,r,s) \ - (_mm_shuffle_epi32( v, ((s)<<6|(r)<<4|(q)<<2|(p)))) - -#define vec2d_swizzle1(v,p,q) \ - (_mm_castsi128_pd(_mm_shuffle_epi32( _mm_castpd_si128(v), ((q*2+1)<<6|(q*2)<<4|(p*2+1)<<2|(p*2))))) - -#define vec4f_swizzle2(a,b,p,q,r,s) \ - (_mm_shuffle_ps( (a), (b), ((s)<<6|(r)<<4|(q)<<2|(p)))) - -#define vec4i_swizzle2(a,b,p,q,r,s) \ - (_mm_castps_si128( (_mm_shuffle_ps( _mm_castsi128_ps(a), _mm_castsi128_ps(b), ((s)<<6|(r)<<4|(q)<<2|(p)))))) - -#define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \ - const Packet4f p4f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2d(NAME,X) \ - const Packet2d p2d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \ - const Packet4f p4f_##NAME = _mm_castsi128_ps(pset1(X)) - -#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ - const Packet4i p4i_##NAME = pset1(X) - - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4f type; - typedef Packet4f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 0, - - HasDiv = 1, - HasSin = EIGEN_FAST_MATH, - HasCos = EIGEN_FAST_MATH, - HasLog = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasTanh = EIGEN_FAST_MATH, - HasBlend = 1 - -#ifdef EIGEN_VECTORIZE_SSE4_1 - , - HasRound = 1, - HasFloor = 1, - HasCeil = 1 -#endif - }; -}; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet2d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 0, - - HasDiv = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasBlend = 1 - -#ifdef EIGEN_VECTORIZE_SSE4_1 - , - HasRound = 1, - HasFloor = 1, - HasCeil = 1 -#endif - }; -}; -#endif -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4i type; - typedef Packet4i half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - - HasBlend = 1 - }; -}; - -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; -template<> struct unpacket_traits { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; - -#ifndef EIGEN_VECTORIZE_AVX -template<> struct scalar_div_cost { enum { value = 7 }; }; -template<> struct scalar_div_cost { enum { value = 8 }; }; -#endif - -#if EIGEN_COMP_MSVC==1500 -// Workaround MSVC 9 internal compiler error. -// TODO: It has been detected with win64 builds (amd64), so let's check whether it also happens in 32bits+SSE mode -// TODO: let's check whether there does not exist a better fix, like adding a pset0() function. (it crashed on pset1(0)). -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { return _mm_set_ps(from,from,from,from); } -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { return _mm_set_pd(from,from); } -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) { return _mm_set_epi32(from,from,from,from); } -#else -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { return _mm_set_ps1(from); } -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { return _mm_set1_pd(from); } -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) { return _mm_set1_epi32(from); } -#endif - -// GCC generates a shufps instruction for _mm_set1_ps/_mm_load1_ps instead of the more efficient pshufd instruction. -// However, using inrinsics for pset1 makes gcc to generate crappy code in some cases (see bug 203) -// Using inline assembly is also not an option because then gcc fails to reorder properly the instructions. -// Therefore, we introduced the pload1 functions to be used in product kernels for which bug 203 does not apply. -// Also note that with AVX, we want it to generate a vbroadcastss. -#if EIGEN_COMP_GNUC_STRICT && (!defined __AVX__) -template<> EIGEN_STRONG_INLINE Packet4f pload1(const float *from) { - return vec4f_swizzle1(_mm_load_ss(from),0,0,0,0); -} -#endif - -template<> EIGEN_STRONG_INLINE Packet4f plset(const float& a) { return _mm_add_ps(pset1(a), _mm_set_ps(3,2,1,0)); } -template<> EIGEN_STRONG_INLINE Packet2d plset(const double& a) { return _mm_add_pd(pset1(a),_mm_set_pd(1,0)); } -template<> EIGEN_STRONG_INLINE Packet4i plset(const int& a) { return _mm_add_epi32(pset1(a),_mm_set_epi32(3,2,1,0)); } - -template<> EIGEN_STRONG_INLINE Packet4f padd(const Packet4f& a, const Packet4f& b) { return _mm_add_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d padd(const Packet2d& a, const Packet2d& b) { return _mm_add_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i padd(const Packet4i& a, const Packet4i& b) { return _mm_add_epi32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f psub(const Packet4f& a, const Packet4f& b) { return _mm_sub_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d psub(const Packet2d& a, const Packet2d& b) { return _mm_sub_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i psub(const Packet4i& a, const Packet4i& b) { return _mm_sub_epi32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) -{ - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); - return _mm_xor_ps(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet2d pnegate(const Packet2d& a) -{ - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x80000000,0x0,0x80000000)); - return _mm_xor_pd(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) -{ - return psub(Packet4i(_mm_setr_epi32(0,0,0,0)), a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4f pmul(const Packet4f& a, const Packet4f& b) { return _mm_mul_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmul(const Packet2d& a, const Packet2d& b) { return _mm_mul_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmul(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_mullo_epi32(a,b); -#else - // this version is slightly faster than 4 scalar products - return vec4i_swizzle1( - vec4i_swizzle2( - _mm_mul_epu32(a,b), - _mm_mul_epu32(vec4i_swizzle1(a,1,0,3,2), - vec4i_swizzle1(b,1,0,3,2)), - 0,2,0,2), - 0,2,1,3); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pdiv(const Packet4f& a, const Packet4f& b) { return _mm_div_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pdiv(const Packet2d& a, const Packet2d& b) { return _mm_div_pd(a,b); } - -// for some weird raisons, it has to be overloaded for packet of integers -template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return padd(pmul(a,b), c); } -#ifdef __FMA__ -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fmadd_ps(a,b,c); } -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fmadd_pd(a,b,c); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f pmin(const Packet4f& a, const Packet4f& b) { return _mm_min_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmin(const Packet2d& a, const Packet2d& b) { return _mm_min_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmin(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_min_epi32(a,b); -#else - // after some bench, this version *is* faster than a scalar implementation - Packet4i mask = _mm_cmplt_epi32(a,b); - return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pmax(const Packet4f& a, const Packet4f& b) { return _mm_max_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmax(const Packet2d& a, const Packet2d& b) { return _mm_max_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmax(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_max_epi32(a,b); -#else - // after some bench, this version *is* faster than a scalar implementation - Packet4i mask = _mm_cmpgt_epi32(a,b); - return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b)); -#endif -} - -#ifdef EIGEN_VECTORIZE_SSE4_1 -template<> EIGEN_STRONG_INLINE Packet4f pround(const Packet4f& a) { return _mm_round_ps(a, 0); } -template<> EIGEN_STRONG_INLINE Packet2d pround(const Packet2d& a) { return _mm_round_pd(a, 0); } - -template<> EIGEN_STRONG_INLINE Packet4f pceil(const Packet4f& a) { return _mm_ceil_ps(a); } -template<> EIGEN_STRONG_INLINE Packet2d pceil(const Packet2d& a) { return _mm_ceil_pd(a); } - -template<> EIGEN_STRONG_INLINE Packet4f pfloor(const Packet4f& a) { return _mm_floor_ps(a); } -template<> EIGEN_STRONG_INLINE Packet2d pfloor(const Packet2d& a) { return _mm_floor_pd(a); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f pand(const Packet4f& a, const Packet4f& b) { return _mm_and_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pand(const Packet2d& a, const Packet2d& b) { return _mm_and_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pand(const Packet4i& a, const Packet4i& b) { return _mm_and_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f por(const Packet4f& a, const Packet4f& b) { return _mm_or_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d por(const Packet2d& a, const Packet2d& b) { return _mm_or_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i por(const Packet4i& a, const Packet4i& b) { return _mm_or_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pxor(const Packet4f& a, const Packet4f& b) { return _mm_xor_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pxor(const Packet2d& a, const Packet2d& b) { return _mm_xor_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pxor(const Packet4i& a, const Packet4i& b) { return _mm_xor_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pandnot(const Packet4f& a, const Packet4f& b) { return _mm_andnot_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pandnot(const Packet2d& a, const Packet2d& b) { return _mm_andnot_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pandnot(const Packet4i& a, const Packet4i& b) { return _mm_andnot_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pload(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_ps(from); } -template<> EIGEN_STRONG_INLINE Packet2d pload(const double* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_pd(from); } -template<> EIGEN_STRONG_INLINE Packet4i pload(const int* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_si128(reinterpret_cast(from)); } - -#if EIGEN_COMP_MSVC - template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) { - EIGEN_DEBUG_UNALIGNED_LOAD - #if (EIGEN_COMP_MSVC==1600) - // NOTE Some version of MSVC10 generates bad code when using _mm_loadu_ps - // (i.e., it does not generate an unaligned load!! - __m128 res = _mm_loadl_pi(_mm_set1_ps(0.0f), (const __m64*)(from)); - res = _mm_loadh_pi(res, (const __m64*)(from+2)); - return res; - #else - return _mm_loadu_ps(from); - #endif - } -#else -// NOTE: with the code below, MSVC's compiler crashes! - -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_ps(from); -} -#endif - -template<> EIGEN_STRONG_INLINE Packet2d ploadu(const double* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_pd(from); -} -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_si128(reinterpret_cast(from)); -} - - -template<> EIGEN_STRONG_INLINE Packet4f ploaddup(const float* from) -{ - return vec4f_swizzle1(_mm_castpd_ps(_mm_load_sd(reinterpret_cast(from))), 0, 0, 1, 1); -} -template<> EIGEN_STRONG_INLINE Packet2d ploaddup(const double* from) -{ return pset1(from[0]); } -template<> EIGEN_STRONG_INLINE Packet4i ploaddup(const int* from) -{ - Packet4i tmp; - tmp = _mm_loadl_epi64(reinterpret_cast(from)); - return vec4i_swizzle1(tmp, 0, 0, 1, 1); -} - -template<> EIGEN_STRONG_INLINE void pstore(float* to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet2d& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(int* to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_si128(reinterpret_cast<__m128i*>(to), from); } - -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet2d& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_si128(reinterpret_cast<__m128i*>(to), from); } - -template<> EIGEN_DEVICE_FUNC inline Packet4f pgather(const float* from, Index stride) -{ - return _mm_set_ps(from[3*stride], from[2*stride], from[1*stride], from[0*stride]); -} -template<> EIGEN_DEVICE_FUNC inline Packet2d pgather(const double* from, Index stride) -{ - return _mm_set_pd(from[1*stride], from[0*stride]); -} -template<> EIGEN_DEVICE_FUNC inline Packet4i pgather(const int* from, Index stride) -{ - return _mm_set_epi32(from[3*stride], from[2*stride], from[1*stride], from[0*stride]); - } - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet4f& from, Index stride) -{ - to[stride*0] = _mm_cvtss_f32(from); - to[stride*1] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 1)); - to[stride*2] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 2)); - to[stride*3] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 3)); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet2d& from, Index stride) -{ - to[stride*0] = _mm_cvtsd_f64(from); - to[stride*1] = _mm_cvtsd_f64(_mm_shuffle_pd(from, from, 1)); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(int* to, const Packet4i& from, Index stride) -{ - to[stride*0] = _mm_cvtsi128_si32(from); - to[stride*1] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 1)); - to[stride*2] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 2)); - to[stride*3] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 3)); -} - -// some compilers might be tempted to perform multiple moves instead of using a vector path. -template<> EIGEN_STRONG_INLINE void pstore1(float* to, const float& a) -{ - Packet4f pa = _mm_set_ss(a); - pstore(to, Packet4f(vec4f_swizzle1(pa,0,0,0,0))); -} -// some compilers might be tempted to perform multiple moves instead of using a vector path. -template<> EIGEN_STRONG_INLINE void pstore1(double* to, const double& a) -{ - Packet2d pa = _mm_set_sd(a); - pstore(to, Packet2d(vec2d_swizzle1(pa,0,0))); -} - -#if EIGEN_COMP_PGI -typedef const void * SsePrefetchPtrType; -#else -typedef const char * SsePrefetchPtrType; -#endif - -#ifndef EIGEN_VECTORIZE_AVX -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -#endif - -#if EIGEN_COMP_MSVC_STRICT && EIGEN_OS_WIN64 -// The temporary variable fixes an internal compilation error in vs <= 2008 and a wrong-result bug in vs 2010 -// Direct of the struct members fixed bug #62. -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { return a.m128_f32[0]; } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { return a.m128d_f64[0]; } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int x = _mm_cvtsi128_si32(a); return x; } -#elif EIGEN_COMP_MSVC_STRICT -// The temporary variable fixes an internal compilation error in vs <= 2008 and a wrong-result bug in vs 2010 -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { float x = _mm_cvtss_f32(a); return x; } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { double x = _mm_cvtsd_f64(a); return x; } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int x = _mm_cvtsi128_si32(a); return x; } -#else -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { return _mm_cvtss_f32(a); } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { return _mm_cvtsd_f64(a); } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { return _mm_cvtsi128_si32(a); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) -{ return _mm_shuffle_ps(a,a,0x1B); } -template<> EIGEN_STRONG_INLINE Packet2d preverse(const Packet2d& a) -{ return _mm_shuffle_pd(a,a,0x1); } -template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) -{ return _mm_shuffle_epi32(a,0x1B); } - -template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) -{ - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF)); - return _mm_and_ps(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet2d pabs(const Packet2d& a) -{ - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0xFFFFFFFF,0x7FFFFFFF,0xFFFFFFFF,0x7FFFFFFF)); - return _mm_and_pd(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) -{ - #ifdef EIGEN_VECTORIZE_SSSE3 - return _mm_abs_epi32(a); - #else - Packet4i aux = _mm_srai_epi32(a,31); - return _mm_sub_epi32(_mm_xor_si128(a,aux),aux); - #endif -} - -// with AVX, the default implementations based on pload1 are faster -#ifndef __AVX__ -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const float *a, - Packet4f& a0, Packet4f& a1, Packet4f& a2, Packet4f& a3) -{ - a3 = pload(a); - a0 = vec4f_swizzle1(a3, 0,0,0,0); - a1 = vec4f_swizzle1(a3, 1,1,1,1); - a2 = vec4f_swizzle1(a3, 2,2,2,2); - a3 = vec4f_swizzle1(a3, 3,3,3,3); -} -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const double *a, - Packet2d& a0, Packet2d& a1, Packet2d& a2, Packet2d& a3) -{ -#ifdef EIGEN_VECTORIZE_SSE3 - a0 = _mm_loaddup_pd(a+0); - a1 = _mm_loaddup_pd(a+1); - a2 = _mm_loaddup_pd(a+2); - a3 = _mm_loaddup_pd(a+3); -#else - a1 = pload(a); - a0 = vec2d_swizzle1(a1, 0,0); - a1 = vec2d_swizzle1(a1, 1,1); - a3 = pload(a+2); - a2 = vec2d_swizzle1(a3, 0,0); - a3 = vec2d_swizzle1(a3, 1,1); -#endif -} -#endif - -EIGEN_STRONG_INLINE void punpackp(Packet4f* vecs) -{ - vecs[1] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0x55)); - vecs[2] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0xAA)); - vecs[3] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0xFF)); - vecs[0] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0x00)); -} - -#ifdef EIGEN_VECTORIZE_SSE3 -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - return _mm_hadd_ps(_mm_hadd_ps(vecs[0], vecs[1]),_mm_hadd_ps(vecs[2], vecs[3])); -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - return _mm_hadd_pd(vecs[0], vecs[1]); -} - -#else -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - Packet4f tmp0, tmp1, tmp2; - tmp0 = _mm_unpacklo_ps(vecs[0], vecs[1]); - tmp1 = _mm_unpackhi_ps(vecs[0], vecs[1]); - tmp2 = _mm_unpackhi_ps(vecs[2], vecs[3]); - tmp0 = _mm_add_ps(tmp0, tmp1); - tmp1 = _mm_unpacklo_ps(vecs[2], vecs[3]); - tmp1 = _mm_add_ps(tmp1, tmp2); - tmp2 = _mm_movehl_ps(tmp1, tmp0); - tmp0 = _mm_movelh_ps(tmp0, tmp1); - return _mm_add_ps(tmp0, tmp2); -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - return _mm_add_pd(_mm_unpacklo_pd(vecs[0], vecs[1]), _mm_unpackhi_pd(vecs[0], vecs[1])); -} -#endif // SSE3 - -template<> EIGEN_STRONG_INLINE float predux(const Packet4f& a) -{ - // Disable SSE3 _mm_hadd_pd that is extremely slow on all existing Intel's architectures - // (from Nehalem to Haswell) -// #ifdef EIGEN_VECTORIZE_SSE3 -// Packet4f tmp = _mm_add_ps(a, vec4f_swizzle1(a,2,3,2,3)); -// return pfirst(_mm_hadd_ps(tmp, tmp)); -// #else - Packet4f tmp = _mm_add_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_add_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -// #endif -} - -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) -{ - // Disable SSE3 _mm_hadd_pd that is extremely slow on all existing Intel's architectures - // (from Nehalem to Haswell) -// #ifdef EIGEN_VECTORIZE_SSE3 -// return pfirst(_mm_hadd_pd(a, a)); -// #else - return pfirst(_mm_add_sd(a, _mm_unpackhi_pd(a,a))); -// #endif -} - -#ifdef EIGEN_VECTORIZE_SSSE3 -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - return _mm_hadd_epi32(_mm_hadd_epi32(vecs[0], vecs[1]),_mm_hadd_epi32(vecs[2], vecs[3])); -} -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i tmp0 = _mm_hadd_epi32(a,a); - return pfirst(_mm_hadd_epi32(tmp0,tmp0)); -} -#else -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i tmp = _mm_add_epi32(a, _mm_unpackhi_epi64(a,a)); - return pfirst(tmp) + pfirst(_mm_shuffle_epi32(tmp, 1)); -} - -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - Packet4i tmp0, tmp1, tmp2; - tmp0 = _mm_unpacklo_epi32(vecs[0], vecs[1]); - tmp1 = _mm_unpackhi_epi32(vecs[0], vecs[1]); - tmp2 = _mm_unpackhi_epi32(vecs[2], vecs[3]); - tmp0 = _mm_add_epi32(tmp0, tmp1); - tmp1 = _mm_unpacklo_epi32(vecs[2], vecs[3]); - tmp1 = _mm_add_epi32(tmp1, tmp2); - tmp2 = _mm_unpacklo_epi64(tmp0, tmp1); - tmp0 = _mm_unpackhi_epi64(tmp0, tmp1); - return _mm_add_epi32(tmp0, tmp2); -} -#endif -// Other reduction functions: - -// mul -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet4f& a) -{ - Packet4f tmp = _mm_mul_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_mul_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) -{ - return pfirst(_mm_mul_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_mul(const Packet4i& a) -{ - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., reusing pmul is very slow !) - // TODO try to call _mm_mul_epu32 directly - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - return (aux[0] * aux[1]) * (aux[2] * aux[3]);; -} - -// min -template<> EIGEN_STRONG_INLINE float predux_min(const Packet4f& a) -{ - Packet4f tmp = _mm_min_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_min_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_min(const Packet2d& a) -{ - return pfirst(_mm_min_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_min(const Packet4i& a) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - Packet4i tmp = _mm_min_epi32(a, _mm_shuffle_epi32(a, _MM_SHUFFLE(0,0,3,2))); - return pfirst(_mm_min_epi32(tmp,_mm_shuffle_epi32(tmp, 1))); -#else - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., it does not like using std::min after the pstore !!) - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - int aux0 = aux[0] EIGEN_STRONG_INLINE float predux_max(const Packet4f& a) -{ - Packet4f tmp = _mm_max_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_max_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_max(const Packet2d& a) -{ - return pfirst(_mm_max_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_max(const Packet4i& a) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - Packet4i tmp = _mm_max_epi32(a, _mm_shuffle_epi32(a, _MM_SHUFFLE(0,0,3,2))); - return pfirst(_mm_max_epi32(tmp,_mm_shuffle_epi32(tmp, 1))); -#else - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., it does not like using std::min after the pstore !!) - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - int aux0 = aux[0]>aux[1] ? aux[0] : aux[1]; - int aux2 = aux[2]>aux[3] ? aux[2] : aux[3]; - return aux0>aux2 ? aux0 : aux2; -#endif // EIGEN_VECTORIZE_SSE4_1 -} - -#if EIGEN_COMP_GNUC -// template <> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) -// { -// Packet4f res = b; -// asm("mulps %[a], %[b] \n\taddps %[c], %[b]" : [b] "+x" (res) : [a] "x" (a), [c] "x" (c)); -// return res; -// } -// EIGEN_STRONG_INLINE Packet4i _mm_alignr_epi8(const Packet4i& a, const Packet4i& b, const int i) -// { -// Packet4i res = a; -// asm("palignr %[i], %[a], %[b] " : [b] "+x" (res) : [a] "x" (a), [i] "i" (i)); -// return res; -// } -#endif - -#ifdef EIGEN_VECTORIZE_SSSE3 -// SSSE3 versions -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { - if (Offset!=0) - first = _mm_castsi128_ps(_mm_alignr_epi8(_mm_castps_si128(second), _mm_castps_si128(first), Offset*4)); - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { - if (Offset!=0) - first = _mm_alignr_epi8(second,first, Offset*4); - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset==1) - first = _mm_castsi128_pd(_mm_alignr_epi8(_mm_castpd_si128(second), _mm_castpd_si128(first), 8)); - } -}; -#else -// SSE2 versions -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { - if (Offset==1) - { - first = _mm_move_ss(first,second); - first = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(first),0x39)); - } - else if (Offset==2) - { - first = _mm_movehl_ps(first,first); - first = _mm_movelh_ps(first,second); - } - else if (Offset==3) - { - first = _mm_move_ss(first,second); - first = _mm_shuffle_ps(first,second,0x93); - } - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { - if (Offset==1) - { - first = _mm_castps_si128(_mm_move_ss(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - first = _mm_shuffle_epi32(first,0x39); - } - else if (Offset==2) - { - first = _mm_castps_si128(_mm_movehl_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(first))); - first = _mm_castps_si128(_mm_movelh_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - } - else if (Offset==3) - { - first = _mm_castps_si128(_mm_move_ss(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - first = _mm_castps_si128(_mm_shuffle_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(second),0x93)); - } - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset==1) - { - first = _mm_castps_pd(_mm_movehl_ps(_mm_castpd_ps(first),_mm_castpd_ps(first))); - first = _mm_castps_pd(_mm_movelh_ps(_mm_castpd_ps(first),_mm_castpd_ps(second))); - } - } -}; -#endif - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - _MM_TRANSPOSE4_PS(kernel.packet[0], kernel.packet[1], kernel.packet[2], kernel.packet[3]); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128d tmp = _mm_unpackhi_pd(kernel.packet[0], kernel.packet[1]); - kernel.packet[0] = _mm_unpacklo_pd(kernel.packet[0], kernel.packet[1]); - kernel.packet[1] = tmp; -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128i T0 = _mm_unpacklo_epi32(kernel.packet[0], kernel.packet[1]); - __m128i T1 = _mm_unpacklo_epi32(kernel.packet[2], kernel.packet[3]); - __m128i T2 = _mm_unpackhi_epi32(kernel.packet[0], kernel.packet[1]); - __m128i T3 = _mm_unpackhi_epi32(kernel.packet[2], kernel.packet[3]); - - kernel.packet[0] = _mm_unpacklo_epi64(T0, T1); - kernel.packet[1] = _mm_unpackhi_epi64(T0, T1); - kernel.packet[2] = _mm_unpacklo_epi64(T2, T3); - kernel.packet[3] = _mm_unpackhi_epi64(T2, T3); -} - -template<> EIGEN_STRONG_INLINE Packet4i pblend(const Selector<4>& ifPacket, const Packet4i& thenPacket, const Packet4i& elsePacket) { - const __m128i zero = _mm_setzero_si128(); - const __m128i select = _mm_set_epi32(ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m128i false_mask = _mm_cmpeq_epi32(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_epi8(thenPacket, elsePacket, false_mask); -#else - return _mm_or_si128(_mm_andnot_si128(false_mask, thenPacket), _mm_and_si128(false_mask, elsePacket)); -#endif -} -template<> EIGEN_STRONG_INLINE Packet4f pblend(const Selector<4>& ifPacket, const Packet4f& thenPacket, const Packet4f& elsePacket) { - const __m128 zero = _mm_setzero_ps(); - const __m128 select = _mm_set_ps(ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m128 false_mask = _mm_cmpeq_ps(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_ps(thenPacket, elsePacket, false_mask); -#else - return _mm_or_ps(_mm_andnot_ps(false_mask, thenPacket), _mm_and_ps(false_mask, elsePacket)); -#endif -} -template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, const Packet2d& thenPacket, const Packet2d& elsePacket) { - const __m128d zero = _mm_setzero_pd(); - const __m128d select = _mm_set_pd(ifPacket.select[1], ifPacket.select[0]); - __m128d false_mask = _mm_cmpeq_pd(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_pd(thenPacket, elsePacket, false_mask); -#else - return _mm_or_pd(_mm_andnot_pd(false_mask, thenPacket), _mm_and_pd(false_mask, elsePacket)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pinsertfirst(const Packet4f& a, float b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_ps(a,pset1(b),1); -#else - return _mm_move_ss(a, _mm_load_ss(&b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet2d pinsertfirst(const Packet2d& a, double b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_pd(a,pset1(b),1); -#else - return _mm_move_sd(a, _mm_load_sd(&b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pinsertlast(const Packet4f& a, float b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_ps(a,pset1(b),(1<<3)); -#else - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x0,0x0,0x0,0xFFFFFFFF)); - return _mm_or_ps(_mm_andnot_ps(mask, a), _mm_and_ps(mask, pset1(b))); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet2d pinsertlast(const Packet2d& a, double b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_pd(a,pset1(b),(1<<1)); -#else - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x0,0xFFFFFFFF,0xFFFFFFFF)); - return _mm_or_pd(_mm_andnot_pd(mask, a), _mm_and_pd(mask, pset1(b))); -#endif -} - -// Scalar path for pmadd with FMA to ensure consistency with vectorized path. -#ifdef __FMA__ -template<> EIGEN_STRONG_INLINE float pmadd(const float& a, const float& b, const float& c) { - return ::fmaf(a,b,c); -} -template<> EIGEN_STRONG_INLINE double pmadd(const double& a, const double& b, const double& c) { - return ::fma(a,b,c); -} -#endif - -} // end namespace internal - -} // end namespace Eigen - -#if EIGEN_COMP_PGI -// PGI++ does not define the following intrinsics in C++ mode. -static inline __m128 _mm_castpd_ps (__m128d x) { return reinterpret_cast<__m128&>(x); } -static inline __m128i _mm_castpd_si128(__m128d x) { return reinterpret_cast<__m128i&>(x); } -static inline __m128d _mm_castps_pd (__m128 x) { return reinterpret_cast<__m128d&>(x); } -static inline __m128i _mm_castps_si128(__m128 x) { return reinterpret_cast<__m128i&>(x); } -static inline __m128 _mm_castsi128_ps(__m128i x) { return reinterpret_cast<__m128&>(x); } -static inline __m128d _mm_castsi128_pd(__m128i x) { return reinterpret_cast<__m128d&>(x); } -#endif - -#endif // EIGEN_PACKET_MATH_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h deleted file mode 100644 index c6ca8c716c0..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h +++ /dev/null @@ -1,77 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TYPE_CASTING_SSE_H -#define EIGEN_TYPE_CASTING_SSE_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_VECTORIZE_AVX -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 2, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 2 - }; -}; -#endif - -template<> EIGEN_STRONG_INLINE Packet4i pcast(const Packet4f& a) { - return _mm_cvttps_epi32(a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet4i& a) { - return _mm_cvtepi32_ps(a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet2d& a, const Packet2d& b) { - return _mm_shuffle_ps(_mm_cvtpd_ps(a), _mm_cvtpd_ps(b), (1 << 2) | (1 << 6)); -} - -template<> EIGEN_STRONG_INLINE Packet2d pcast(const Packet4f& a) { - // Simply discard the second half of the input - return _mm_cvtps_pd(a); -} - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TYPE_CASTING_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/Complex.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/Complex.h deleted file mode 100644 index 1bfb73397d6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/Complex.h +++ /dev/null @@ -1,397 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// Copyright (C) 2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX32_ALTIVEC_H -#define EIGEN_COMPLEX32_ALTIVEC_H - -namespace Eigen { - -namespace internal { - -static Packet2ul p2ul_CONJ_XOR1 = (Packet2ul) vec_sld((Packet4ui) p2d_ZERO_, (Packet4ui) p2l_ZERO, 8);//{ 0x8000000000000000, 0x0000000000000000 }; -static Packet2ul p2ul_CONJ_XOR2 = (Packet2ul) vec_sld((Packet4ui) p2l_ZERO, (Packet4ui) p2d_ZERO_, 8);//{ 0x8000000000000000, 0x0000000000000000 }; - -struct Packet1cd -{ - EIGEN_STRONG_INLINE Packet1cd() {} - EIGEN_STRONG_INLINE explicit Packet1cd(const Packet2d& a) : v(a) {} - Packet2d v; -}; - -struct Packet2cf -{ - EIGEN_STRONG_INLINE Packet2cf() {} - EIGEN_STRONG_INLINE explicit Packet2cf(const Packet4f& a) : v(a) {} - union { - Packet4f v; - Packet1cd cd[2]; - }; -}; - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasBlend = 1, - HasSetLinear = 0 - }; -}; - - -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet1cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 1, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; -template<> struct unpacket_traits { typedef std::complex type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; - -/* Forward declaration */ -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel); - -template<> EIGEN_STRONG_INLINE Packet2cf pload (const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload((const float*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd pload (const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet2cf ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu((const float*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu((const double*)from)); } -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((float*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((float*)to, from.v); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, from.v); } - -template<> EIGEN_STRONG_INLINE Packet1cd pset1(const std::complex& from) -{ /* here we really have to use unaligned loads :( */ return ploadu(&from); } - -template<> EIGEN_STRONG_INLINE Packet2cf pset1(const std::complex& from) -{ - Packet2cf res; - res.cd[0] = Packet1cd(vec_ld2f((const float *)&from)); - res.cd[1] = res.cd[0]; - return res; -} -template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather, Packet2cf>(const std::complex* from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - return pload(af); -} -template<> EIGEN_DEVICE_FUNC inline Packet1cd pgather, Packet1cd>(const std::complex* from, Index stride EIGEN_UNUSED) -{ - return pload(from); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cf>(std::complex* to, const Packet2cf& from, Index stride) -{ - std::complex EIGEN_ALIGN16 af[2]; - pstore >((std::complex *) af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; -} -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet1cd>(std::complex* to, const Packet1cd& from, Index stride EIGEN_UNUSED) -{ - pstore >(to, from); -} - -template<> EIGEN_STRONG_INLINE Packet2cf padd(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(padd(a.v, b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd padd(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(a.v + b.v); } -template<> EIGEN_STRONG_INLINE Packet2cf psub(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(psub(a.v, b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd psub(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(a.v - b.v); } -template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } -template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) { return Packet2cf(pnegate(Packet4f(a.v))); } -template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) { return Packet1cd((Packet2d)vec_xor((Packet2d)a.v, (Packet2d)p2ul_CONJ_XOR2)); } -template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) -{ - Packet2cf res; - res.v.v4f[0] = pconj(Packet1cd(reinterpret_cast(a.v.v4f[0]))).v; - res.v.v4f[1] = pconj(Packet1cd(reinterpret_cast(a.v.v4f[1]))).v; - return res; -} - -template<> EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) -{ - Packet2d a_re, a_im, v1, v2; - - // Permute and multiply the real parts of a and b - a_re = vec_perm(a.v, a.v, p16uc_PSET64_HI); - // Get the imaginary parts of a - a_im = vec_perm(a.v, a.v, p16uc_PSET64_LO); - // multiply a_re * b - v1 = vec_madd(a_re, b.v, p2d_ZERO); - // multiply a_im * b and get the conjugate result - v2 = vec_madd(a_im, b.v, p2d_ZERO); - v2 = (Packet2d) vec_sld((Packet4ui)v2, (Packet4ui)v2, 8); - v2 = (Packet2d) vec_xor((Packet2d)v2, (Packet2d) p2ul_CONJ_XOR1); - - return Packet1cd(v1 + v2); -} -template<> EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) -{ - Packet2cf res; - res.v.v4f[0] = pmul(Packet1cd(reinterpret_cast(a.v.v4f[0])), Packet1cd(reinterpret_cast(b.v.v4f[0]))).v; - res.v.v4f[1] = pmul(Packet1cd(reinterpret_cast(a.v.v4f[1])), Packet1cd(reinterpret_cast(b.v.v4f[1]))).v; - return res; -} - -template<> EIGEN_STRONG_INLINE Packet1cd pand (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(vec_and(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pand (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pand(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd por (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(vec_or(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf por (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(por(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pxor (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(vec_xor(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pxor (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pxor(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pandnot(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(vec_and(a.v, vec_nor(b.v,b.v))); } -template<> EIGEN_STRONG_INLINE Packet2cf pandnot(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(pandnot(a.v,b.v)); } - -template<> EIGEN_STRONG_INLINE Packet1cd ploaddup(const std::complex* from) { return pset1(*from); } -template<> EIGEN_STRONG_INLINE Packet2cf ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_ZVECTOR_PREFETCH(addr); } -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { EIGEN_ZVECTOR_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet1cd& a) -{ - std::complex EIGEN_ALIGN16 res; - pstore >(&res, a); - - return res; -} -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cf& a) -{ - std::complex EIGEN_ALIGN16 res[2]; - pstore >(res, a); - - return res[0]; -} - -template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) -{ - Packet2cf res; - res.cd[0] = a.cd[1]; - res.cd[1] = a.cd[0]; - return res; -} - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet1cd& a) -{ - return pfirst(a); -} -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cf& a) -{ - std::complex res; - Packet1cd b = padd(a.cd[0], a.cd[1]); - vec_st2f(b.v, (float*)&res); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet1cd preduxp(const Packet1cd* vecs) -{ - return vecs[0]; -} -template<> EIGEN_STRONG_INLINE Packet2cf preduxp(const Packet2cf* vecs) -{ - PacketBlock transpose; - transpose.packet[0] = vecs[0]; - transpose.packet[1] = vecs[1]; - ptranspose(transpose); - - return padd(transpose.packet[0], transpose.packet[1]); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet1cd& a) -{ - return pfirst(a); -} -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cf& a) -{ - std::complex res; - Packet1cd b = pmul(a.cd[0], a.cd[1]); - vec_st2f(b.v, (float*)&res); - return res; -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) - { - // FIXME is it sure we never have to align a Packet1cd? - // Even though a std::complex has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) - { - if (Offset == 1) { - first.cd[0] = first.cd[1]; - first.cd[1] = second.cd[0]; - } - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(a, pconj(b)); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return internal::pmul(pconj(a), b); - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - return pconj(internal::pmul(a, b)); - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) - -template<> EIGEN_STRONG_INLINE Packet1cd pdiv(const Packet1cd& a, const Packet1cd& b) -{ - // TODO optimize it for AltiVec - Packet1cd res = conj_helper().pmul(a,b); - Packet2d s = vec_madd(b.v, b.v, p2d_ZERO_); - return Packet1cd(pdiv(res.v, s + vec_perm(s, s, p16uc_REVERSE64))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pdiv(const Packet2cf& a, const Packet2cf& b) -{ - // TODO optimize it for AltiVec - Packet2cf res; - res.cd[0] = pdiv(a.cd[0], b.cd[0]); - res.cd[1] = pdiv(a.cd[1], b.cd[1]); - return res; -} - -EIGEN_STRONG_INLINE Packet1cd pcplxflip/**/(const Packet1cd& x) -{ - return Packet1cd(preverse(Packet2d(x.v))); -} - -EIGEN_STRONG_INLINE Packet2cf pcplxflip/**/(const Packet2cf& x) -{ - Packet2cf res; - res.cd[0] = pcplxflip(x.cd[0]); - res.cd[1] = pcplxflip(x.cd[1]); - return res; -} - -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel) -{ - Packet2d tmp = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_HI); - kernel.packet[1].v = vec_perm(kernel.packet[0].v, kernel.packet[1].v, p16uc_TRANSPOSE64_LO); - kernel.packet[0].v = tmp; -} - -EIGEN_STRONG_INLINE void ptranspose(PacketBlock& kernel) -{ - Packet1cd tmp = kernel.packet[0].cd[1]; - kernel.packet[0].cd[1] = kernel.packet[1].cd[0]; - kernel.packet[1].cd[0] = tmp; -} - -template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { - Packet2cf result; - const Selector<4> ifPacket4 = { ifPacket.select[0], ifPacket.select[0], ifPacket.select[1], ifPacket.select[1] }; - result.v = pblend(ifPacket4, thenPacket.v, elsePacket.v); - return result; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX32_ALTIVEC_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/MathFunctions.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/MathFunctions.h deleted file mode 100644 index 5c7aa725678..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/MathFunctions.h +++ /dev/null @@ -1,137 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007 Julien Pommier -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* The sin, cos, exp, and log functions of this file come from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -#ifndef EIGEN_MATH_FUNCTIONS_ALTIVEC_H -#define EIGEN_MATH_FUNCTIONS_ALTIVEC_H - -namespace Eigen { - -namespace internal { - -static _EIGEN_DECLARE_CONST_Packet2d(1 , 1.0); -static _EIGEN_DECLARE_CONST_Packet2d(2 , 2.0); -static _EIGEN_DECLARE_CONST_Packet2d(half, 0.5); - -static _EIGEN_DECLARE_CONST_Packet2d(exp_hi, 709.437); -static _EIGEN_DECLARE_CONST_Packet2d(exp_lo, -709.436139303); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_LOG2EF, 1.4426950408889634073599); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p0, 1.26177193074810590878e-4); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p1, 3.02994407707441961300e-2); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p2, 9.99999999999999999910e-1); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q0, 3.00198505138664455042e-6); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q1, 2.52448340349684104192e-3); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q2, 2.27265548208155028766e-1); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q3, 2.00000000000000000009e0); - -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C1, 0.693145751953125); -static _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C2, 1.42860682030941723212e-6); - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d pexp(const Packet2d& _x) -{ - Packet2d x = _x; - - Packet2d tmp, fx; - Packet2l emm0; - - // clamp x - x = pmax(pmin(x, p2d_exp_hi), p2d_exp_lo); - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(p2d_cephes_LOG2EF, x, p2d_half); - - fx = vec_floor(fx); - - tmp = pmul(fx, p2d_cephes_exp_C1); - Packet2d z = pmul(fx, p2d_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - Packet2d x2 = pmul(x,x); - - Packet2d px = p2d_cephes_exp_p0; - px = pmadd(px, x2, p2d_cephes_exp_p1); - px = pmadd(px, x2, p2d_cephes_exp_p2); - px = pmul (px, x); - - Packet2d qx = p2d_cephes_exp_q0; - qx = pmadd(qx, x2, p2d_cephes_exp_q1); - qx = pmadd(qx, x2, p2d_cephes_exp_q2); - qx = pmadd(qx, x2, p2d_cephes_exp_q3); - - x = pdiv(px,psub(qx,px)); - x = pmadd(p2d_2,x,p2d_1); - - // build 2^n - emm0 = vec_ctsl(fx, 0); - - static const Packet2l p2l_1023 = { 1023, 1023 }; - static const Packet2ul p2ul_52 = { 52, 52 }; - - emm0 = emm0 + p2l_1023; - emm0 = emm0 << reinterpret_cast(p2ul_52); - - // Altivec's max & min operators just drop silent NaNs. Check NaNs in - // inputs and return them unmodified. - Packet2ul isnumber_mask = reinterpret_cast(vec_cmpeq(_x, _x)); - return vec_sel(_x, pmax(pmul(x, reinterpret_cast(emm0)), _x), - isnumber_mask); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pexp(const Packet4f& x) -{ - Packet4f res; - res.v4f[0] = pexp(x.v4f[0]); - res.v4f[1] = pexp(x.v4f[1]); - return res; -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d psqrt(const Packet2d& x) -{ - return __builtin_s390_vfsqdb(x); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& x) -{ - Packet4f res; - res.v4f[0] = psqrt(x.v4f[0]); - res.v4f[1] = psqrt(x.v4f[1]); - return res; -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d prsqrt(const Packet2d& x) { - // Unfortunately we can't use the much faster mm_rqsrt_pd since it only provides an approximation. - return pset1(1.0) / psqrt(x); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& x) { - Packet4f res; - res.v4f[0] = prsqrt(x.v4f[0]); - res.v4f[1] = prsqrt(x.v4f[1]); - return res; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_ALTIVEC_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/PacketMath.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/PacketMath.h deleted file mode 100644 index 57b01fc634d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/arch/ZVector/PacketMath.h +++ /dev/null @@ -1,945 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Konstantinos Margaritis -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_ZVECTOR_H -#define EIGEN_PACKET_MATH_ZVECTOR_H - -#include - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 4 -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#endif - -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 16 -#endif - -typedef __vector int Packet4i; -typedef __vector unsigned int Packet4ui; -typedef __vector __bool int Packet4bi; -typedef __vector short int Packet8i; -typedef __vector unsigned char Packet16uc; -typedef __vector double Packet2d; -typedef __vector unsigned long long Packet2ul; -typedef __vector long long Packet2l; - -typedef struct { - Packet2d v4f[2]; -} Packet4f; - -typedef union { - int32_t i[4]; - uint32_t ui[4]; - int64_t l[2]; - uint64_t ul[2]; - double d[2]; - Packet4i v4i; - Packet4ui v4ui; - Packet2l v2l; - Packet2ul v2ul; - Packet2d v2d; -} Packet; - -// We don't want to write the same code all the time, but we need to reuse the constants -// and it doesn't really work to declare them global, so we define macros instead - -#define _EIGEN_DECLARE_CONST_FAST_Packet4i(NAME,X) \ - Packet4i p4i_##NAME = reinterpret_cast(vec_splat_s32(X)) - -#define _EIGEN_DECLARE_CONST_FAST_Packet2d(NAME,X) \ - Packet2d p2d_##NAME = reinterpret_cast(vec_splat_s64(X)) - -#define _EIGEN_DECLARE_CONST_FAST_Packet2l(NAME,X) \ - Packet2l p2l_##NAME = reinterpret_cast(vec_splat_s64(X)) - -#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ - Packet4i p4i_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2d(NAME,X) \ - Packet2d p2d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2l(NAME,X) \ - Packet2l p2l_##NAME = pset1(X) - -// These constants are endian-agnostic -//static _EIGEN_DECLARE_CONST_FAST_Packet4i(ZERO, 0); //{ 0, 0, 0, 0,} -static _EIGEN_DECLARE_CONST_FAST_Packet4i(ONE, 1); //{ 1, 1, 1, 1} - -static _EIGEN_DECLARE_CONST_FAST_Packet2d(ZERO, 0); -static _EIGEN_DECLARE_CONST_FAST_Packet2l(ZERO, 0); -static _EIGEN_DECLARE_CONST_FAST_Packet2l(ONE, 1); - -static Packet2d p2d_ONE = { 1.0, 1.0 }; -static Packet2d p2d_ZERO_ = { -0.0, -0.0 }; - -static Packet4i p4i_COUNTDOWN = { 0, 1, 2, 3 }; -static Packet4f p4f_COUNTDOWN = { 0.0, 1.0, 2.0, 3.0 }; -static Packet2d p2d_COUNTDOWN = reinterpret_cast(vec_sld(reinterpret_cast(p2d_ZERO), reinterpret_cast(p2d_ONE), 8)); - -static Packet16uc p16uc_PSET64_HI = { 0,1,2,3, 4,5,6,7, 0,1,2,3, 4,5,6,7 }; -static Packet16uc p16uc_DUPLICATE32_HI = { 0,1,2,3, 0,1,2,3, 4,5,6,7, 4,5,6,7 }; - -// Mask alignment -#define _EIGEN_MASK_ALIGNMENT 0xfffffffffffffff0 - -#define _EIGEN_ALIGNED_PTR(x) ((std::ptrdiff_t)(x) & _EIGEN_MASK_ALIGNMENT) - -// Handle endianness properly while loading constants -// Define global static constants: - -static Packet16uc p16uc_FORWARD = { 0,1,2,3, 4,5,6,7, 8,9,10,11, 12,13,14,15 }; -static Packet16uc p16uc_REVERSE32 = { 12,13,14,15, 8,9,10,11, 4,5,6,7, 0,1,2,3 }; -static Packet16uc p16uc_REVERSE64 = { 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; - -static Packet16uc p16uc_PSET32_WODD = vec_sld((Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 0), (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 2), 8);//{ 0,1,2,3, 0,1,2,3, 8,9,10,11, 8,9,10,11 }; -static Packet16uc p16uc_PSET32_WEVEN = vec_sld(p16uc_DUPLICATE32_HI, (Packet16uc) vec_splat((Packet4ui)p16uc_FORWARD, 3), 8);//{ 4,5,6,7, 4,5,6,7, 12,13,14,15, 12,13,14,15 }; -/*static Packet16uc p16uc_HALF64_0_16 = vec_sld((Packet16uc)p4i_ZERO, vec_splat((Packet16uc) vec_abs(p4i_MINUS16), 3), 8); //{ 0,0,0,0, 0,0,0,0, 16,16,16,16, 16,16,16,16}; - -static Packet16uc p16uc_PSET64_HI = (Packet16uc) vec_mergeh((Packet4ui)p16uc_PSET32_WODD, (Packet4ui)p16uc_PSET32_WEVEN); //{ 0,1,2,3, 4,5,6,7, 0,1,2,3, 4,5,6,7 };*/ -static Packet16uc p16uc_PSET64_LO = (Packet16uc) vec_mergel((Packet4ui)p16uc_PSET32_WODD, (Packet4ui)p16uc_PSET32_WEVEN); //{ 8,9,10,11, 12,13,14,15, 8,9,10,11, 12,13,14,15 }; -/*static Packet16uc p16uc_TRANSPOSE64_HI = vec_add(p16uc_PSET64_HI, p16uc_HALF64_0_16); //{ 0,1,2,3, 4,5,6,7, 16,17,18,19, 20,21,22,23}; -static Packet16uc p16uc_TRANSPOSE64_LO = vec_add(p16uc_PSET64_LO, p16uc_HALF64_0_16); //{ 8,9,10,11, 12,13,14,15, 24,25,26,27, 28,29,30,31};*/ -static Packet16uc p16uc_TRANSPOSE64_HI = { 0,1,2,3, 4,5,6,7, 16,17,18,19, 20,21,22,23}; -static Packet16uc p16uc_TRANSPOSE64_LO = { 8,9,10,11, 12,13,14,15, 24,25,26,27, 28,29,30,31}; - -//static Packet16uc p16uc_COMPLEX32_REV = vec_sld(p16uc_REVERSE32, p16uc_REVERSE32, 8); //{ 4,5,6,7, 0,1,2,3, 12,13,14,15, 8,9,10,11 }; - -//static Packet16uc p16uc_COMPLEX32_REV2 = vec_sld(p16uc_FORWARD, p16uc_FORWARD, 8); //{ 8,9,10,11, 12,13,14,15, 0,1,2,3, 4,5,6,7 }; - - -#if EIGEN_HAS_BUILTIN(__builtin_prefetch) || EIGEN_COMP_GNUC - #define EIGEN_ZVECTOR_PREFETCH(ADDR) __builtin_prefetch(ADDR); -#else - #define EIGEN_ZVECTOR_PREFETCH(ADDR) asm( " pfd [%[addr]]\n" :: [addr] "r" (ADDR) : "cc" ); -#endif - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4i type; - typedef Packet4i half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasBlend = 1 - }; -}; - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4f type; - typedef Packet4f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasMin = 1, - HasMax = 1, - HasAbs = 1, - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasRound = 1, - HasFloor = 1, - HasCeil = 1, - HasNegate = 1, - HasBlend = 1 - }; -}; - -template<> struct packet_traits : default_packet_traits -{ - typedef Packet2d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 1, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasMin = 1, - HasMax = 1, - HasAbs = 1, - HasSin = 0, - HasCos = 0, - HasLog = 0, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasRound = 1, - HasFloor = 1, - HasCeil = 1, - HasNegate = 1, - HasBlend = 1 - }; -}; - -template<> struct unpacket_traits { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; - -/* Forward declaration */ -EIGEN_DEVICE_FUNC inline void ptranspose(PacketBlock& kernel); - -inline std::ostream & operator <<(std::ostream & s, const Packet4i & v) -{ - Packet vt; - vt.v4i = v; - s << vt.i[0] << ", " << vt.i[1] << ", " << vt.i[2] << ", " << vt.i[3]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet4ui & v) -{ - Packet vt; - vt.v4ui = v; - s << vt.ui[0] << ", " << vt.ui[1] << ", " << vt.ui[2] << ", " << vt.ui[3]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet2l & v) -{ - Packet vt; - vt.v2l = v; - s << vt.l[0] << ", " << vt.l[1]; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet2ul & v) -{ - Packet vt; - vt.v2ul = v; - s << vt.ul[0] << ", " << vt.ul[1] ; - return s; -} - -inline std::ostream & operator <<(std::ostream & s, const Packet2d & v) -{ - Packet vt; - vt.v2d = v; - s << vt.d[0] << ", " << vt.d[1]; - return s; -} - -/* Helper function to simulate a vec_splat_packet4f - */ -template EIGEN_STRONG_INLINE Packet4f vec_splat_packet4f(const Packet4f& from) -{ - Packet4f splat; - switch (element) { - case 0: - splat.v4f[0] = vec_splat(from.v4f[0], 0); - splat.v4f[1] = splat.v4f[0]; - break; - case 1: - splat.v4f[0] = vec_splat(from.v4f[0], 1); - splat.v4f[1] = splat.v4f[0]; - break; - case 2: - splat.v4f[0] = vec_splat(from.v4f[1], 0); - splat.v4f[1] = splat.v4f[0]; - break; - case 3: - splat.v4f[0] = vec_splat(from.v4f[1], 1); - splat.v4f[1] = splat.v4f[0]; - break; - } - return splat; -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { - switch (Offset % 4) { - case 1: - first = vec_sld(first, second, 4); break; - case 2: - first = vec_sld(first, second, 8); break; - case 3: - first = vec_sld(first, second, 12); break; - } - } -}; - -/* This is a tricky one, we have to translate float alignment to vector elements of sizeof double - */ -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { - switch (Offset % 4) { - case 1: - first.v4f[0] = vec_sld(first.v4f[0], first.v4f[1], 8); - first.v4f[1] = vec_sld(first.v4f[1], second.v4f[0], 8); - break; - case 2: - first.v4f[0] = first.v4f[1]; - first.v4f[1] = second.v4f[0]; - break; - case 3: - first.v4f[0] = vec_sld(first.v4f[1], second.v4f[0], 8); - first.v4f[1] = vec_sld(second.v4f[0], second.v4f[1], 8); - break; - } - } -}; - - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset == 1) - first = reinterpret_cast(vec_sld(reinterpret_cast(first), reinterpret_cast(second), 8)); - } -}; - -template<> EIGEN_STRONG_INLINE Packet4i pload(const int* from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_LOAD - Packet *vfrom; - vfrom = (Packet *) from; - return vfrom->v4i; -} - -template<> EIGEN_STRONG_INLINE Packet4f pload(const float* from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_LOAD - Packet4f vfrom; - vfrom.v4f[0] = vec_ld2f(&from[0]); - vfrom.v4f[1] = vec_ld2f(&from[2]); - return vfrom; -} - -template<> EIGEN_STRONG_INLINE Packet2d pload(const double* from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_LOAD - Packet *vfrom; - vfrom = (Packet *) from; - return vfrom->v2d; -} - -template<> EIGEN_STRONG_INLINE void pstore(int* to, const Packet4i& from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_STORE - Packet *vto; - vto = (Packet *) to; - vto->v4i = from; -} - -template<> EIGEN_STRONG_INLINE void pstore(float* to, const Packet4f& from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_STORE - vec_st2f(from.v4f[0], &to[0]); - vec_st2f(from.v4f[1], &to[2]); -} - - -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet2d& from) -{ - // FIXME: No intrinsic yet - EIGEN_DEBUG_ALIGNED_STORE - Packet *vto; - vto = (Packet *) to; - vto->v2d = from; -} - -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) -{ - return vec_splats(from); -} -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { - return vec_splats(from); -} -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) -{ - Packet4f to; - to.v4f[0] = pset1(static_cast(from)); - to.v4f[1] = to.v4f[0]; - return to; -} - -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const int *a, - Packet4i& a0, Packet4i& a1, Packet4i& a2, Packet4i& a3) -{ - a3 = pload(a); - a0 = vec_splat(a3, 0); - a1 = vec_splat(a3, 1); - a2 = vec_splat(a3, 2); - a3 = vec_splat(a3, 3); -} - -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const float *a, - Packet4f& a0, Packet4f& a1, Packet4f& a2, Packet4f& a3) -{ - a3 = pload(a); - a0 = vec_splat_packet4f<0>(a3); - a1 = vec_splat_packet4f<1>(a3); - a2 = vec_splat_packet4f<2>(a3); - a3 = vec_splat_packet4f<3>(a3); -} - -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const double *a, - Packet2d& a0, Packet2d& a1, Packet2d& a2, Packet2d& a3) -{ - a1 = pload(a); - a0 = vec_splat(a1, 0); - a1 = vec_splat(a1, 1); - a3 = pload(a+2); - a2 = vec_splat(a3, 0); - a3 = vec_splat(a3, 1); -} - -template<> EIGEN_DEVICE_FUNC inline Packet4i pgather(const int* from, Index stride) -{ - int EIGEN_ALIGN16 ai[4]; - ai[0] = from[0*stride]; - ai[1] = from[1*stride]; - ai[2] = from[2*stride]; - ai[3] = from[3*stride]; - return pload(ai); -} - -template<> EIGEN_DEVICE_FUNC inline Packet4f pgather(const float* from, Index stride) -{ - float EIGEN_ALIGN16 ai[4]; - ai[0] = from[0*stride]; - ai[1] = from[1*stride]; - ai[2] = from[2*stride]; - ai[3] = from[3*stride]; - return pload(ai); -} - -template<> EIGEN_DEVICE_FUNC inline Packet2d pgather(const double* from, Index stride) -{ - double EIGEN_ALIGN16 af[2]; - af[0] = from[0*stride]; - af[1] = from[1*stride]; - return pload(af); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(int* to, const Packet4i& from, Index stride) -{ - int EIGEN_ALIGN16 ai[4]; - pstore((int *)ai, from); - to[0*stride] = ai[0]; - to[1*stride] = ai[1]; - to[2*stride] = ai[2]; - to[3*stride] = ai[3]; -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet4f& from, Index stride) -{ - float EIGEN_ALIGN16 ai[4]; - pstore((float *)ai, from); - to[0*stride] = ai[0]; - to[1*stride] = ai[1]; - to[2*stride] = ai[2]; - to[3*stride] = ai[3]; -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet2d& from, Index stride) -{ - double EIGEN_ALIGN16 af[2]; - pstore(af, from); - to[0*stride] = af[0]; - to[1*stride] = af[1]; -} - -template<> EIGEN_STRONG_INLINE Packet4i padd(const Packet4i& a, const Packet4i& b) { return (a + b); } -template<> EIGEN_STRONG_INLINE Packet4f padd(const Packet4f& a, const Packet4f& b) -{ - Packet4f c; - c.v4f[0] = a.v4f[0] + b.v4f[0]; - c.v4f[1] = a.v4f[1] + b.v4f[1]; - return c; -} -template<> EIGEN_STRONG_INLINE Packet2d padd(const Packet2d& a, const Packet2d& b) { return (a + b); } - -template<> EIGEN_STRONG_INLINE Packet4i psub(const Packet4i& a, const Packet4i& b) { return (a - b); } -template<> EIGEN_STRONG_INLINE Packet4f psub(const Packet4f& a, const Packet4f& b) -{ - Packet4f c; - c.v4f[0] = a.v4f[0] - b.v4f[0]; - c.v4f[1] = a.v4f[1] - b.v4f[1]; - return c; -} -template<> EIGEN_STRONG_INLINE Packet2d psub(const Packet2d& a, const Packet2d& b) { return (a - b); } - -template<> EIGEN_STRONG_INLINE Packet4i pmul(const Packet4i& a, const Packet4i& b) { return (a * b); } -template<> EIGEN_STRONG_INLINE Packet4f pmul(const Packet4f& a, const Packet4f& b) -{ - Packet4f c; - c.v4f[0] = a.v4f[0] * b.v4f[0]; - c.v4f[1] = a.v4f[1] * b.v4f[1]; - return c; -} -template<> EIGEN_STRONG_INLINE Packet2d pmul(const Packet2d& a, const Packet2d& b) { return (a * b); } - -template<> EIGEN_STRONG_INLINE Packet4i pdiv(const Packet4i& a, const Packet4i& b) { return (a / b); } -template<> EIGEN_STRONG_INLINE Packet4f pdiv(const Packet4f& a, const Packet4f& b) -{ - Packet4f c; - c.v4f[0] = a.v4f[0] / b.v4f[0]; - c.v4f[1] = a.v4f[1] / b.v4f[1]; - return c; -} -template<> EIGEN_STRONG_INLINE Packet2d pdiv(const Packet2d& a, const Packet2d& b) { return (a / b); } - -template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) { return (-a); } -template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) -{ - Packet4f c; - c.v4f[0] = -a.v4f[0]; - c.v4f[1] = -a.v4f[1]; - return c; -} -template<> EIGEN_STRONG_INLINE Packet2d pnegate(const Packet2d& a) { return (-a); } - -template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return padd(pmul(a, b), c); } -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) -{ - Packet4f res; - res.v4f[0] = vec_madd(a.v4f[0], b.v4f[0], c.v4f[0]); - res.v4f[1] = vec_madd(a.v4f[1], b.v4f[1], c.v4f[1]); - return res; -} -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return vec_madd(a, b, c); } - -template<> EIGEN_STRONG_INLINE Packet4i plset(const int& a) { return padd(pset1(a), p4i_COUNTDOWN); } -template<> EIGEN_STRONG_INLINE Packet4f plset(const float& a) { return padd(pset1(a), p4f_COUNTDOWN); } -template<> EIGEN_STRONG_INLINE Packet2d plset(const double& a) { return padd(pset1(a), p2d_COUNTDOWN); } - -template<> EIGEN_STRONG_INLINE Packet4i pmin(const Packet4i& a, const Packet4i& b) { return vec_min(a, b); } -template<> EIGEN_STRONG_INLINE Packet2d pmin(const Packet2d& a, const Packet2d& b) { return vec_min(a, b); } -template<> EIGEN_STRONG_INLINE Packet4f pmin(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pmin(a.v4f[0], b.v4f[0]); - res.v4f[1] = pmin(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4i pmax(const Packet4i& a, const Packet4i& b) { return vec_max(a, b); } -template<> EIGEN_STRONG_INLINE Packet2d pmax(const Packet2d& a, const Packet2d& b) { return vec_max(a, b); } -template<> EIGEN_STRONG_INLINE Packet4f pmax(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pmax(a.v4f[0], b.v4f[0]); - res.v4f[1] = pmax(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4i pand(const Packet4i& a, const Packet4i& b) { return vec_and(a, b); } -template<> EIGEN_STRONG_INLINE Packet2d pand(const Packet2d& a, const Packet2d& b) { return vec_and(a, b); } -template<> EIGEN_STRONG_INLINE Packet4f pand(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pand(a.v4f[0], b.v4f[0]); - res.v4f[1] = pand(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4i por(const Packet4i& a, const Packet4i& b) { return vec_or(a, b); } -template<> EIGEN_STRONG_INLINE Packet2d por(const Packet2d& a, const Packet2d& b) { return vec_or(a, b); } -template<> EIGEN_STRONG_INLINE Packet4f por(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pand(a.v4f[0], b.v4f[0]); - res.v4f[1] = pand(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4i pxor(const Packet4i& a, const Packet4i& b) { return vec_xor(a, b); } -template<> EIGEN_STRONG_INLINE Packet2d pxor(const Packet2d& a, const Packet2d& b) { return vec_xor(a, b); } -template<> EIGEN_STRONG_INLINE Packet4f pxor(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pand(a.v4f[0], b.v4f[0]); - res.v4f[1] = pand(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4i pandnot(const Packet4i& a, const Packet4i& b) { return pand(a, vec_nor(b, b)); } -template<> EIGEN_STRONG_INLINE Packet2d pandnot(const Packet2d& a, const Packet2d& b) { return vec_and(a, vec_nor(b, b)); } -template<> EIGEN_STRONG_INLINE Packet4f pandnot(const Packet4f& a, const Packet4f& b) -{ - Packet4f res; - res.v4f[0] = pandnot(a.v4f[0], b.v4f[0]); - res.v4f[1] = pandnot(a.v4f[1], b.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE Packet4f pround(const Packet4f& a) -{ - Packet4f res; - res.v4f[0] = vec_round(a.v4f[0]); - res.v4f[1] = vec_round(a.v4f[1]); - return res; -} -template<> EIGEN_STRONG_INLINE Packet2d pround(const Packet2d& a) { return vec_round(a); } -template<> EIGEN_STRONG_INLINE Packet4f pceil(const Packet4f& a) -{ - Packet4f res; - res.v4f[0] = vec_ceil(a.v4f[0]); - res.v4f[1] = vec_ceil(a.v4f[1]); - return res; -} -template<> EIGEN_STRONG_INLINE Packet2d pceil(const Packet2d& a) { return vec_ceil(a); } -template<> EIGEN_STRONG_INLINE Packet4f pfloor(const Packet4f& a) -{ - Packet4f res; - res.v4f[0] = vec_floor(a.v4f[0]); - res.v4f[1] = vec_floor(a.v4f[1]); - return res; -} -template<> EIGEN_STRONG_INLINE Packet2d pfloor(const Packet2d& a) { return vec_floor(a); } - -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int* from) { return pload(from); } -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) { return pload(from); } -template<> EIGEN_STRONG_INLINE Packet2d ploadu(const double* from) { return pload(from); } - - -template<> EIGEN_STRONG_INLINE Packet4i ploaddup(const int* from) -{ - Packet4i p = pload(from); - return vec_perm(p, p, p16uc_DUPLICATE32_HI); -} - -template<> EIGEN_STRONG_INLINE Packet4f ploaddup(const float* from) -{ - Packet4f p = pload(from); - p.v4f[1] = vec_splat(p.v4f[0], 1); - p.v4f[0] = vec_splat(p.v4f[0], 0); - return p; -} - -template<> EIGEN_STRONG_INLINE Packet2d ploaddup(const double* from) -{ - Packet2d p = pload(from); - return vec_perm(p, p, p16uc_PSET64_HI); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet4i& from) { pstore(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet4f& from) { pstore(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet2d& from) { pstore(to, from); } - -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { EIGEN_ZVECTOR_PREFETCH(addr); } -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { EIGEN_ZVECTOR_PREFETCH(addr); } -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { EIGEN_ZVECTOR_PREFETCH(addr); } - -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int EIGEN_ALIGN16 x[4]; pstore(x, a); return x[0]; } -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { float EIGEN_ALIGN16 x[2]; vec_st2f(a.v4f[0], &x[0]); return x[0]; } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { double EIGEN_ALIGN16 x[2]; pstore(x, a); return x[0]; } - -template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) -{ - return reinterpret_cast(vec_perm(reinterpret_cast(a), reinterpret_cast(a), p16uc_REVERSE32)); -} - -template<> EIGEN_STRONG_INLINE Packet2d preverse(const Packet2d& a) -{ - return reinterpret_cast(vec_perm(reinterpret_cast(a), reinterpret_cast(a), p16uc_REVERSE64)); -} - -template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) -{ - Packet4f rev; - rev.v4f[0] = preverse(a.v4f[1]); - rev.v4f[1] = preverse(a.v4f[0]); - return rev; -} - -template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) { return vec_abs(a); } -template<> EIGEN_STRONG_INLINE Packet2d pabs(const Packet2d& a) { return vec_abs(a); } -template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) -{ - Packet4f res; - res.v4f[0] = pabs(a.v4f[0]); - res.v4f[1] = pabs(a.v4f[1]); - return res; -} - -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i b, sum; - b = vec_sld(a, a, 8); - sum = padd(a, b); - b = vec_sld(sum, sum, 4); - sum = padd(sum, b); - return pfirst(sum); -} - -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) -{ - Packet2d b, sum; - b = reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)); - sum = padd(a, b); - return pfirst(sum); -} -template<> EIGEN_STRONG_INLINE float predux(const Packet4f& a) -{ - Packet2d sum; - sum = padd(a.v4f[0], a.v4f[1]); - double first = predux(sum); - return static_cast(first); -} - -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - Packet4i v[4], sum[4]; - - // It's easier and faster to transpose then add as columns - // Check: http://www.freevec.org/function/matrix_4x4_transpose_floats for explanation - // Do the transpose, first set of moves - v[0] = vec_mergeh(vecs[0], vecs[2]); - v[1] = vec_mergel(vecs[0], vecs[2]); - v[2] = vec_mergeh(vecs[1], vecs[3]); - v[3] = vec_mergel(vecs[1], vecs[3]); - // Get the resulting vectors - sum[0] = vec_mergeh(v[0], v[2]); - sum[1] = vec_mergel(v[0], v[2]); - sum[2] = vec_mergeh(v[1], v[3]); - sum[3] = vec_mergel(v[1], v[3]); - - // Now do the summation: - // Lines 0+1 - sum[0] = padd(sum[0], sum[1]); - // Lines 2+3 - sum[1] = padd(sum[2], sum[3]); - // Add the results - sum[0] = padd(sum[0], sum[1]); - - return sum[0]; -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - Packet2d v[2], sum; - v[0] = padd(vecs[0], reinterpret_cast(vec_sld(reinterpret_cast(vecs[0]), reinterpret_cast(vecs[0]), 8))); - v[1] = padd(vecs[1], reinterpret_cast(vec_sld(reinterpret_cast(vecs[1]), reinterpret_cast(vecs[1]), 8))); - - sum = reinterpret_cast(vec_sld(reinterpret_cast(v[0]), reinterpret_cast(v[1]), 8)); - - return sum; -} - -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - PacketBlock transpose; - transpose.packet[0] = vecs[0]; - transpose.packet[1] = vecs[1]; - transpose.packet[2] = vecs[2]; - transpose.packet[3] = vecs[3]; - ptranspose(transpose); - - Packet4f sum = padd(transpose.packet[0], transpose.packet[1]); - sum = padd(sum, transpose.packet[2]); - sum = padd(sum, transpose.packet[3]); - return sum; -} - -// Other reduction functions: -// mul -template<> EIGEN_STRONG_INLINE int predux_mul(const Packet4i& a) -{ - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - return aux[0] * aux[1] * aux[2] * aux[3]; -} - -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) -{ - return pfirst(pmul(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet4f& a) -{ - // Return predux_mul of the subvectors product - return static_cast(pfirst(predux_mul(pmul(a.v4f[0], a.v4f[1])))); -} - -// min -template<> EIGEN_STRONG_INLINE int predux_min(const Packet4i& a) -{ - Packet4i b, res; - b = pmin(a, vec_sld(a, a, 8)); - res = pmin(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -template<> EIGEN_STRONG_INLINE double predux_min(const Packet2d& a) -{ - return pfirst(pmin(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -template<> EIGEN_STRONG_INLINE float predux_min(const Packet4f& a) -{ - Packet2d b, res; - b = pmin(a.v4f[0], a.v4f[1]); - res = pmin(b, reinterpret_cast(vec_sld(reinterpret_cast(b), reinterpret_cast(b), 8))); - return static_cast(pfirst(res)); -} - -// max -template<> EIGEN_STRONG_INLINE int predux_max(const Packet4i& a) -{ - Packet4i b, res; - b = pmax(a, vec_sld(a, a, 8)); - res = pmax(b, vec_sld(b, b, 4)); - return pfirst(res); -} - -// max -template<> EIGEN_STRONG_INLINE double predux_max(const Packet2d& a) -{ - return pfirst(pmax(a, reinterpret_cast(vec_sld(reinterpret_cast(a), reinterpret_cast(a), 8)))); -} - -template<> EIGEN_STRONG_INLINE float predux_max(const Packet4f& a) -{ - Packet2d b, res; - b = pmax(a.v4f[0], a.v4f[1]); - res = pmax(b, reinterpret_cast(vec_sld(reinterpret_cast(b), reinterpret_cast(b), 8))); - return static_cast(pfirst(res)); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet4i t0 = vec_mergeh(kernel.packet[0], kernel.packet[2]); - Packet4i t1 = vec_mergel(kernel.packet[0], kernel.packet[2]); - Packet4i t2 = vec_mergeh(kernel.packet[1], kernel.packet[3]); - Packet4i t3 = vec_mergel(kernel.packet[1], kernel.packet[3]); - kernel.packet[0] = vec_mergeh(t0, t2); - kernel.packet[1] = vec_mergel(t0, t2); - kernel.packet[2] = vec_mergeh(t1, t3); - kernel.packet[3] = vec_mergel(t1, t3); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - Packet2d t0 = vec_perm(kernel.packet[0], kernel.packet[1], p16uc_TRANSPOSE64_HI); - Packet2d t1 = vec_perm(kernel.packet[0], kernel.packet[1], p16uc_TRANSPOSE64_LO); - kernel.packet[0] = t0; - kernel.packet[1] = t1; -} - -/* Split the Packet4f PacketBlock into 4 Packet2d PacketBlocks and transpose each one - */ -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - PacketBlock t0,t1,t2,t3; - // copy top-left 2x2 Packet2d block - t0.packet[0] = kernel.packet[0].v4f[0]; - t0.packet[1] = kernel.packet[1].v4f[0]; - - // copy top-right 2x2 Packet2d block - t1.packet[0] = kernel.packet[0].v4f[1]; - t1.packet[1] = kernel.packet[1].v4f[1]; - - // copy bottom-left 2x2 Packet2d block - t2.packet[0] = kernel.packet[2].v4f[0]; - t2.packet[1] = kernel.packet[3].v4f[0]; - - // copy bottom-right 2x2 Packet2d block - t3.packet[0] = kernel.packet[2].v4f[1]; - t3.packet[1] = kernel.packet[3].v4f[1]; - - // Transpose all 2x2 blocks - ptranspose(t0); - ptranspose(t1); - ptranspose(t2); - ptranspose(t3); - - // Copy back transposed blocks, but exchange t1 and t2 due to transposition - kernel.packet[0].v4f[0] = t0.packet[0]; - kernel.packet[0].v4f[1] = t2.packet[0]; - kernel.packet[1].v4f[0] = t0.packet[1]; - kernel.packet[1].v4f[1] = t2.packet[1]; - kernel.packet[2].v4f[0] = t1.packet[0]; - kernel.packet[2].v4f[1] = t3.packet[0]; - kernel.packet[3].v4f[0] = t1.packet[1]; - kernel.packet[3].v4f[1] = t3.packet[1]; -} - -template<> EIGEN_STRONG_INLINE Packet4i pblend(const Selector<4>& ifPacket, const Packet4i& thenPacket, const Packet4i& elsePacket) { - Packet4ui select = { ifPacket.select[0], ifPacket.select[1], ifPacket.select[2], ifPacket.select[3] }; - Packet4ui mask = vec_cmpeq(select, reinterpret_cast(p4i_ONE)); - return vec_sel(elsePacket, thenPacket, mask); -} - -template<> EIGEN_STRONG_INLINE Packet4f pblend(const Selector<4>& ifPacket, const Packet4f& thenPacket, const Packet4f& elsePacket) { - Packet2ul select_hi = { ifPacket.select[0], ifPacket.select[1] }; - Packet2ul select_lo = { ifPacket.select[2], ifPacket.select[3] }; - Packet2ul mask_hi = vec_cmpeq(select_hi, reinterpret_cast(p2l_ONE)); - Packet2ul mask_lo = vec_cmpeq(select_lo, reinterpret_cast(p2l_ONE)); - Packet4f result; - result.v4f[0] = vec_sel(elsePacket.v4f[0], thenPacket.v4f[0], mask_hi); - result.v4f[1] = vec_sel(elsePacket.v4f[1], thenPacket.v4f[1], mask_lo); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, const Packet2d& thenPacket, const Packet2d& elsePacket) { - Packet2ul select = { ifPacket.select[0], ifPacket.select[1] }; - Packet2ul mask = vec_cmpeq(select, reinterpret_cast(p2l_ONE)); - return vec_sel(elsePacket, thenPacket, mask); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PACKET_MATH_ZVECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h deleted file mode 100644 index 4153b877cff..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h +++ /dev/null @@ -1,168 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ASSIGNMENT_FUNCTORS_H -#define EIGEN_ASSIGNMENT_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -/** \internal - * \brief Template functor for scalar/packet assignment - * - */ -template struct assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a = b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,b); } -}; - -// Empty overload for void type (used by PermutationMatrix) -template struct assign_op {}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost, - PacketAccess = is_same::value && packet_traits::Vectorizable && packet_traits::Vectorizable - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with addition - * - */ -template struct add_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(add_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a += b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::padd(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::AddCost, - PacketAccess = is_same::value && packet_traits::HasAdd - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with subtraction - * - */ -template struct sub_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(sub_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a -= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::psub(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::AddCost, - PacketAccess = is_same::value && packet_traits::HasSub - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with multiplication - * - */ -template -struct mul_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(mul_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a *= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::pmul(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::MulCost, - PacketAccess = is_same::value && packet_traits::HasMul - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with diviving - * - */ -template struct div_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(div_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a /= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::pdiv(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::MulCost, - PacketAccess = is_same::value && packet_traits::HasDiv - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with swapping - * - * It works as follow. For a non-vectorized evaluation loop, we have: - * for(i) func(A.coeffRef(i), B.coeff(i)); - * where B is a SwapWrapper expression. The trick is to make SwapWrapper::coeff behaves like a non-const coeffRef. - * Actually, SwapWrapper might not even be needed since even if B is a plain expression, since it has to be writable - * B.coeff already returns a const reference to the underlying scalar value. - * - * The case of a vectorized loop is more tricky: - * for(i,j) func.assignPacket(&A.coeffRef(i,j), B.packet(i,j)); - * Here, B must be a SwapWrapper whose packet function actually returns a proxy object holding a Scalar*, - * the actual alignment and Packet type. - * - */ -template struct swap_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(swap_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Scalar& a, const Scalar& b) const - { -#ifdef __CUDACC__ - // FIXME is there some kind of cuda::swap? - Scalar t=b; const_cast(b)=a; a=t; -#else - using std::swap; - swap(a,const_cast(b)); -#endif - } -}; -template -struct functor_traits > { - enum { - Cost = 3 * NumTraits::ReadCost, - PacketAccess = packet_traits::Vectorizable - }; -}; - -} // namespace internal - -} // namespace Eigen - -#endif // EIGEN_ASSIGNMENT_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h deleted file mode 100644 index 3eae6b8cada..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h +++ /dev/null @@ -1,475 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BINARY_FUNCTORS_H -#define EIGEN_BINARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -//---------- associative binary functors ---------- - -template -struct binary_op_base -{ - typedef Arg1 first_argument_type; - typedef Arg2 second_argument_type; -}; - -/** \internal - * \brief Template functor to compute the sum of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::operator+, class VectorwiseOp, DenseBase::sum() - */ -template -struct scalar_sum_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_sum_op) -#else - scalar_sum_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a + b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::padd(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, // rough estimate! - PacketAccess = is_same::value && packet_traits::HasAdd && packet_traits::HasAdd - // TODO vectorize mixed sum - }; -}; - -/** \internal - * \brief Template specialization to deprecate the summation of boolean expressions. - * This is required to solve Bug 426. - * \sa DenseBase::count(), DenseBase::any(), ArrayBase::cast(), MatrixBase::cast() - */ -template<> struct scalar_sum_op : scalar_sum_op { - EIGEN_DEPRECATED - scalar_sum_op() {} -}; - - -/** \internal - * \brief Template functor to compute the product of two scalars - * - * \sa class CwiseBinaryOp, Cwise::operator*(), class VectorwiseOp, MatrixBase::redux() - */ -template -struct scalar_product_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_product_op) -#else - scalar_product_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a * b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmul(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_mul(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::MulCost + NumTraits::MulCost)/2, // rough estimate! - PacketAccess = is_same::value && packet_traits::HasMul && packet_traits::HasMul - // TODO vectorize mixed product - }; -}; - -/** \internal - * \brief Template functor to compute the conjugate product of two scalars - * - * This is a short cut for conj(x) * y which is needed for optimization purpose; in Eigen2 support mode, this becomes x * conj(y) - */ -template -struct scalar_conj_product_op : binary_op_base -{ - - enum { - Conj = NumTraits::IsComplex - }; - - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_conj_product_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const - { return conj_helper().pmul(a,b); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return conj_helper().pmul(a,b); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::MulCost, - PacketAccess = internal::is_same::value && packet_traits::HasMul - }; -}; - -/** \internal - * \brief Template functor to compute the min of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::cwiseMin, class VectorwiseOp, MatrixBase::minCoeff() - */ -template -struct scalar_min_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_min_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return numext::mini(a, b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmin(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_min(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = internal::is_same::value && packet_traits::HasMin - }; -}; - -/** \internal - * \brief Template functor to compute the max of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::cwiseMax, class VectorwiseOp, MatrixBase::maxCoeff() - */ -template -struct scalar_max_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_max_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return numext::maxi(a, b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmax(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_max(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = internal::is_same::value && packet_traits::HasMax - }; -}; - -/** \internal - * \brief Template functors for comparison of two scalars - * \todo Implement packet-comparisons - */ -template struct scalar_cmp_op; - -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = false - }; -}; - -template -struct result_of(LhsScalar,RhsScalar)> { - typedef bool type; -}; - - -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a==b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a<=b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a>b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a>=b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return !(a<=b || b<=a);} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a!=b;} -}; - - -/** \internal - * \brief Template functor to compute the hypot of two \b positive \b and \b real scalars - * - * \sa MatrixBase::stableNorm(), class Redux - */ -template -struct scalar_hypot_op : binary_op_base -{ - EIGEN_EMPTY_STRUCT_CTOR(scalar_hypot_op) - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar &x, const Scalar &y) const - { - // This functor is used by hypotNorm only for which it is faster to first apply abs - // on all coefficients prior to reduction through hypot. - // This way we avoid calling abs on positive and real entries, and this also permits - // to seamlessly handle complexes. Otherwise we would have to handle both real and complexes - // through the same functor... - return internal::positive_real_hypot(x,y); - } -}; -template -struct functor_traits > { - enum - { - Cost = 3 * NumTraits::AddCost + - 2 * NumTraits::MulCost + - 2 * scalar_div_cost::value, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the pow of two scalars - */ -template -struct scalar_pow_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_pow_op) -#else - scalar_pow_op() { - typedef Scalar LhsScalar; - typedef Exponent RhsScalar; - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC - inline result_type operator() (const Scalar& a, const Exponent& b) const { return numext::pow(a, b); } -}; -template -struct functor_traits > { - enum { Cost = 5 * NumTraits::MulCost, PacketAccess = false }; -}; - - - -//---------- non associative binary functors ---------- - -/** \internal - * \brief Template functor to compute the difference of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::operator- - */ -template -struct scalar_difference_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_difference_op) -#else - scalar_difference_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a - b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::psub(a,b); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = is_same::value && packet_traits::HasSub && packet_traits::HasSub - }; -}; - -/** \internal - * \brief Template functor to compute the quotient of two scalars - * - * \sa class CwiseBinaryOp, Cwise::operator/() - */ -template -struct scalar_quotient_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op) -#else - scalar_quotient_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a / b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pdiv(a,b); } -}; -template -struct functor_traits > { - typedef typename scalar_quotient_op::result_type result_type; - enum { - PacketAccess = is_same::value && packet_traits::HasDiv && packet_traits::HasDiv, - Cost = scalar_div_cost::value - }; -}; - - - -/** \internal - * \brief Template functor to compute the and of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator&& - */ -struct scalar_boolean_and_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_and_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a && b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the or of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator|| - */ -struct scalar_boolean_or_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_or_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a || b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the xor of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator^ - */ -struct scalar_boolean_xor_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_xor_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a ^ b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - - - -//---------- binary functors bound to a constant, thus appearing as a unary functor ---------- - -// The following two classes permits to turn any binary functor into a unary one with one argument bound to a constant value. -// They are analogues to std::binder1st/binder2nd but with the following differences: -// - they are compatible with packetOp -// - they are portable across C++ versions (the std::binder* are deprecated in C++11) -template struct bind1st_op : BinaryOp { - - typedef typename BinaryOp::first_argument_type first_argument_type; - typedef typename BinaryOp::second_argument_type second_argument_type; - typedef typename BinaryOp::result_type result_type; - - bind1st_op(const first_argument_type &val) : m_value(val) {} - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const second_argument_type& b) const { return BinaryOp::operator()(m_value,b); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& b) const - { return BinaryOp::packetOp(internal::pset1(m_value), b); } - - first_argument_type m_value; -}; -template struct functor_traits > : functor_traits {}; - - -template struct bind2nd_op : BinaryOp { - - typedef typename BinaryOp::first_argument_type first_argument_type; - typedef typename BinaryOp::second_argument_type second_argument_type; - typedef typename BinaryOp::result_type result_type; - - bind2nd_op(const second_argument_type &val) : m_value(val) {} - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const first_argument_type& a) const { return BinaryOp::operator()(a,m_value); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return BinaryOp::packetOp(a,internal::pset1(m_value)); } - - second_argument_type m_value; -}; -template struct functor_traits > : functor_traits {}; - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BINARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h deleted file mode 100644 index b03be0269c9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h +++ /dev/null @@ -1,188 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_NULLARY_FUNCTORS_H -#define EIGEN_NULLARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -template -struct scalar_constant_op { - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE scalar_constant_op(const scalar_constant_op& other) : m_other(other.m_other) { } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE scalar_constant_op(const Scalar& other) : m_other(other) { } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() () const { return m_other; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const PacketType packetOp() const { return internal::pset1(m_other); } - const Scalar m_other; -}; -template -struct functor_traits > -{ enum { Cost = 0 /* as the constant value should be loaded in register only once for the whole expression */, - PacketAccess = packet_traits::Vectorizable, IsRepeatable = true }; }; - -template struct scalar_identity_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_identity_op) - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType row, IndexType col) const { return row==col ? Scalar(1) : Scalar(0); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false, IsRepeatable = true }; }; - -template struct linspaced_op_impl; - -template -struct linspaced_op_impl -{ - linspaced_op_impl(const Scalar& low, const Scalar& high, Index num_steps) : - m_low(low), m_high(high), m_size1(num_steps==1 ? 1 : num_steps-1), m_step(num_steps==1 ? Scalar() : (high-low)/Scalar(num_steps-1)), - m_flip(numext::abs(high) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType i) const { - typedef typename NumTraits::Real RealScalar; - if(m_flip) - return (i==0)? m_low : (m_high - RealScalar(m_size1-i)*m_step); - else - return (i==m_size1)? m_high : (m_low + RealScalar(i)*m_step); - } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(IndexType i) const - { - // Principle: - // [low, ..., low] + ( [step, ..., step] * ( [i, ..., i] + [0, ..., size] ) ) - if(m_flip) - { - Packet pi = plset(Scalar(i-m_size1)); - Packet res = padd(pset1(m_high), pmul(pset1(m_step), pi)); - if(i==0) - res = pinsertfirst(res, m_low); - return res; - } - else - { - Packet pi = plset(Scalar(i)); - Packet res = padd(pset1(m_low), pmul(pset1(m_step), pi)); - if(i==m_size1-unpacket_traits::size+1) - res = pinsertlast(res, m_high); - return res; - } - } - - const Scalar m_low; - const Scalar m_high; - const Index m_size1; - const Scalar m_step; - const bool m_flip; -}; - -template -struct linspaced_op_impl -{ - linspaced_op_impl(const Scalar& low, const Scalar& high, Index num_steps) : - m_low(low), - m_multiplier((high-low)/convert_index(num_steps<=1 ? 1 : num_steps-1)), - m_divisor(convert_index((high>=low?num_steps:-num_steps)+(high-low))/((numext::abs(high-low)+1)==0?1:(numext::abs(high-low)+1))), - m_use_divisor(num_steps>1 && (numext::abs(high-low)+1) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE - const Scalar operator() (IndexType i) const - { - if(m_use_divisor) return m_low + convert_index(i)/m_divisor; - else return m_low + convert_index(i)*m_multiplier; - } - - const Scalar m_low; - const Scalar m_multiplier; - const Scalar m_divisor; - const bool m_use_divisor; -}; - -// ----- Linspace functor ---------------------------------------------------------------- - -// Forward declaration (we default to random access which does not really give -// us a speed gain when using packet access but it allows to use the functor in -// nested expressions). -template struct linspaced_op; -template struct functor_traits< linspaced_op > -{ - enum - { - Cost = 1, - PacketAccess = (!NumTraits::IsInteger) && packet_traits::HasSetLinear && packet_traits::HasBlend, - /*&& ((!NumTraits::IsInteger) || packet_traits::HasDiv),*/ // <- vectorization for integer is currently disabled - IsRepeatable = true - }; -}; -template struct linspaced_op -{ - linspaced_op(const Scalar& low, const Scalar& high, Index num_steps) - : impl((num_steps==1 ? high : low),high,num_steps) - {} - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType i) const { return impl(i); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(IndexType i) const { return impl.packetOp(i); } - - // This proxy object handles the actual required temporaries and the different - // implementations (integer vs. floating point). - const linspaced_op_impl::IsInteger> impl; -}; - -// Linear access is automatically determined from the operator() prototypes available for the given functor. -// If it exposes an operator()(i,j), then we assume the i and j coefficients are required independently -// and linear access is not possible. In all other cases, linear access is enabled. -// Users should not have to deal with this structure. -template struct functor_has_linear_access { enum { ret = !has_binary_operator::value }; }; - -// For unreliable compilers, let's specialize the has_*ary_operator -// helpers so that at least built-in nullary functors work fine. -#if !( (EIGEN_COMP_MSVC>1600) || (EIGEN_GNUC_AT_LEAST(4,8)) || (EIGEN_COMP_ICC>=1600)) -template -struct has_nullary_operator,IndexType> { enum { value = 1}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 0}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 1}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 0}; }; -template -struct has_unary_operator,IndexType> { enum { value = 1}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 1}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_NULLARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h deleted file mode 100644 index 9c1d75850be..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h +++ /dev/null @@ -1,136 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STL_FUNCTORS_H -#define EIGEN_STL_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -// default functor traits for STL functors: - -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -#if (__cplusplus < 201103L) && (EIGEN_COMP_MSVC <= 1900) -// std::binder* are deprecated since c++11 and will be removed in c++17 -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost, PacketAccess = false }; }; -#endif - -#if (__cplusplus < 201703L) && (EIGEN_COMP_MSVC < 1910) -// std::unary_negate is deprecated since c++17 and will be removed in c++20 -template -struct functor_traits > -{ enum { Cost = 1 + functor_traits::Cost, PacketAccess = false }; }; - -// std::binary_negate is deprecated since c++17 and will be removed in c++20 -template -struct functor_traits > -{ enum { Cost = 1 + functor_traits::Cost, PacketAccess = false }; }; -#endif - -#ifdef EIGEN_STDEXT_SUPPORT - -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost + functor_traits::Cost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost + functor_traits::Cost + functor_traits::Cost, PacketAccess = false }; }; - -#endif // EIGEN_STDEXT_SUPPORT - -// allow to add new functors and specializations of functor_traits from outside Eigen. -// this macro is really needed because functor_traits must be specialized after it is declared but before it is used... -#ifdef EIGEN_FUNCTORS_PLUGIN -#include EIGEN_FUNCTORS_PLUGIN -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_STL_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h deleted file mode 100644 index b254e96c6ad..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h +++ /dev/null @@ -1,25 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Eugene Brevdo -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TERNARY_FUNCTORS_H -#define EIGEN_TERNARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -//---------- associative ternary functors ---------- - - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TERNARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h deleted file mode 100644 index b56e7afd2c5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h +++ /dev/null @@ -1,792 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_UNARY_FUNCTORS_H -#define EIGEN_UNARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -/** \internal - * \brief Template functor to compute the opposite of a scalar - * - * \sa class CwiseUnaryOp, MatrixBase::operator- - */ -template struct scalar_opposite_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_opposite_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return -a; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pnegate(a); } -}; -template -struct functor_traits > -{ enum { - Cost = NumTraits::AddCost, - PacketAccess = packet_traits::HasNegate }; -}; - -/** \internal - * \brief Template functor to compute the absolute value of a scalar - * - * \sa class CwiseUnaryOp, Cwise::abs - */ -template struct scalar_abs_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_abs_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pabs(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::AddCost, - PacketAccess = packet_traits::HasAbs - }; -}; - -/** \internal - * \brief Template functor to compute the score of a scalar, to chose a pivot - * - * \sa class CwiseUnaryOp - */ -template struct scalar_score_coeff_op : scalar_abs_op -{ - typedef void Score_is_abs; -}; -template -struct functor_traits > : functor_traits > {}; - -/* Avoid recomputing abs when we know the score and they are the same. Not a true Eigen functor. */ -template struct abs_knowing_score -{ - EIGEN_EMPTY_STRUCT_CTOR(abs_knowing_score) - typedef typename NumTraits::Real result_type; - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a, const Score&) const { return numext::abs(a); } -}; -template struct abs_knowing_score::Score_is_abs> -{ - EIGEN_EMPTY_STRUCT_CTOR(abs_knowing_score) - typedef typename NumTraits::Real result_type; - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scal&, const result_type& a) const { return a; } -}; - -/** \internal - * \brief Template functor to compute the squared absolute value of a scalar - * - * \sa class CwiseUnaryOp, Cwise::abs2 - */ -template struct scalar_abs2_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_abs2_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs2(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pmul(a,a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasAbs2 }; }; - -/** \internal - * \brief Template functor to compute the conjugate of a complex value - * - * \sa class CwiseUnaryOp, MatrixBase::conjugate() - */ -template struct scalar_conjugate_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_conjugate_op) - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { using numext::conj; return conj(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const { return internal::pconj(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::IsComplex ? NumTraits::AddCost : 0, - PacketAccess = packet_traits::HasConj - }; -}; - -/** \internal - * \brief Template functor to compute the phase angle of a complex - * - * \sa class CwiseUnaryOp, Cwise::arg - */ -template struct scalar_arg_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_arg_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { using numext::arg; return arg(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::parg(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::IsComplex ? 5 * NumTraits::MulCost : NumTraits::AddCost, - PacketAccess = packet_traits::HasArg - }; -}; -/** \internal - * \brief Template functor to cast a scalar to another type - * - * \sa class CwiseUnaryOp, MatrixBase::cast() - */ -template -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef NewType result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const NewType operator() (const Scalar& a) const { return cast(a); } -}; -template -struct functor_traits > -{ enum { Cost = is_same::value ? 0 : NumTraits::AddCost, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the real part of a complex - * - * \sa class CwiseUnaryOp, MatrixBase::real() - */ -template -struct scalar_real_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_real_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::real(a); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the imaginary part of a complex - * - * \sa class CwiseUnaryOp, MatrixBase::imag() - */ -template -struct scalar_imag_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::imag(a); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the real part of a complex as a reference - * - * \sa class CwiseUnaryOp, MatrixBase::real() - */ -template -struct scalar_real_ref_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_real_ref_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::real_ref(*const_cast(&a)); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the imaginary part of a complex as a reference - * - * \sa class CwiseUnaryOp, MatrixBase::imag() - */ -template -struct scalar_imag_ref_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_ref_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::imag_ref(*const_cast(&a)); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * - * \brief Template functor to compute the exponential of a scalar - * - * \sa class CwiseUnaryOp, Cwise::exp() - */ -template struct scalar_exp_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_exp_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::exp(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pexp(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasExp, - // The following numbers are based on the AVX implementation. -#ifdef EIGEN_VECTORIZE_FMA - // Haswell can issue 2 add/mul/madd per cycle. - Cost = - (sizeof(Scalar) == 4 - // float: 8 pmadd, 4 pmul, 2 padd/psub, 6 other - ? (8 * NumTraits::AddCost + 6 * NumTraits::MulCost) - // double: 7 pmadd, 5 pmul, 3 padd/psub, 1 div, 13 other - : (14 * NumTraits::AddCost + - 6 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value)) -#else - Cost = - (sizeof(Scalar) == 4 - // float: 7 pmadd, 6 pmul, 4 padd/psub, 10 other - ? (21 * NumTraits::AddCost + 13 * NumTraits::MulCost) - // double: 7 pmadd, 5 pmul, 3 padd/psub, 1 div, 13 other - : (23 * NumTraits::AddCost + - 12 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value)) -#endif - }; -}; - -/** \internal - * - * \brief Template functor to compute the logarithm of a scalar - * - * \sa class CwiseUnaryOp, ArrayBase::log() - */ -template struct scalar_log_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::log(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasLog, - Cost = - (PacketAccess - // The following numbers are based on the AVX implementation. -#ifdef EIGEN_VECTORIZE_FMA - // 8 pmadd, 6 pmul, 8 padd/psub, 16 other, can issue 2 add/mul/madd per cycle. - ? (20 * NumTraits::AddCost + 7 * NumTraits::MulCost) -#else - // 8 pmadd, 6 pmul, 8 padd/psub, 20 other - ? (36 * NumTraits::AddCost + 14 * NumTraits::MulCost) -#endif - // Measured cost of std::log. - : sizeof(Scalar)==4 ? 40 : 85) - }; -}; - -/** \internal - * - * \brief Template functor to compute the logarithm of 1 plus a scalar value - * - * \sa class CwiseUnaryOp, ArrayBase::log1p() - */ -template struct scalar_log1p_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log1p_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::log1p(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog1p(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasLog1p, - Cost = functor_traits >::Cost // TODO measure cost of log1p - }; -}; - -/** \internal - * - * \brief Template functor to compute the base-10 logarithm of a scalar - * - * \sa class CwiseUnaryOp, Cwise::log10() - */ -template struct scalar_log10_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log10_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { EIGEN_USING_STD_MATH(log10) return log10(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog10(a); } -}; -template -struct functor_traits > -{ enum { Cost = 5 * NumTraits::MulCost, PacketAccess = packet_traits::HasLog10 }; }; - -/** \internal - * \brief Template functor to compute the square root of a scalar - * \sa class CwiseUnaryOp, Cwise::sqrt() - */ -template struct scalar_sqrt_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sqrt_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sqrt(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psqrt(a); } -}; -template -struct functor_traits > { - enum { -#if EIGEN_FAST_MATH - // The following numbers are based on the AVX implementation. - Cost = (sizeof(Scalar) == 8 ? 28 - // 4 pmul, 1 pmadd, 3 other - : (3 * NumTraits::AddCost + - 5 * NumTraits::MulCost)), -#else - // The following numbers are based on min VSQRT throughput on Haswell. - Cost = (sizeof(Scalar) == 8 ? 28 : 14), -#endif - PacketAccess = packet_traits::HasSqrt - }; -}; - -/** \internal - * \brief Template functor to compute the reciprocal square root of a scalar - * \sa class CwiseUnaryOp, Cwise::rsqrt() - */ -template struct scalar_rsqrt_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_rsqrt_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return Scalar(1)/numext::sqrt(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::prsqrt(a); } -}; - -template -struct functor_traits > -{ enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasRsqrt - }; -}; - -/** \internal - * \brief Template functor to compute the cosine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::cos() - */ -template struct scalar_cos_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cos_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return numext::cos(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pcos(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasCos - }; -}; - -/** \internal - * \brief Template functor to compute the sine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::sin() - */ -template struct scalar_sin_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sin_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sin(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psin(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasSin - }; -}; - - -/** \internal - * \brief Template functor to compute the tan of a scalar - * \sa class CwiseUnaryOp, ArrayBase::tan() - */ -template struct scalar_tan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_tan_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::tan(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::ptan(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasTan - }; -}; - -/** \internal - * \brief Template functor to compute the arc cosine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::acos() - */ -template struct scalar_acos_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_acos_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::acos(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pacos(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasACos - }; -}; - -/** \internal - * \brief Template functor to compute the arc sine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::asin() - */ -template struct scalar_asin_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_asin_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::asin(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pasin(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasASin - }; -}; - - -/** \internal - * \brief Template functor to compute the atan of a scalar - * \sa class CwiseUnaryOp, ArrayBase::atan() - */ -template struct scalar_atan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_atan_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::atan(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::patan(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasATan - }; -}; - -/** \internal - * \brief Template functor to compute the tanh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::tanh() - */ -template -struct scalar_tanh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_tanh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator()(const Scalar& a) const { return numext::tanh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& x) const { return ptanh(x); } -}; - -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasTanh, - Cost = ( (EIGEN_FAST_MATH && is_same::value) -// The following numbers are based on the AVX implementation, -#ifdef EIGEN_VECTORIZE_FMA - // Haswell can issue 2 add/mul/madd per cycle. - // 9 pmadd, 2 pmul, 1 div, 2 other - ? (2 * NumTraits::AddCost + - 6 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value) -#else - ? (11 * NumTraits::AddCost + - 11 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value) -#endif - // This number assumes a naive implementation of tanh - : (6 * NumTraits::AddCost + - 3 * NumTraits::MulCost + - 2 * scalar_div_cost::HasDiv>::value + - functor_traits >::Cost)) - }; -}; - -/** \internal - * \brief Template functor to compute the sinh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::sinh() - */ -template struct scalar_sinh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sinh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sinh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psinh(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasSinh - }; -}; - -/** \internal - * \brief Template functor to compute the cosh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::cosh() - */ -template struct scalar_cosh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cosh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::cosh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pcosh(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasCosh - }; -}; - -/** \internal - * \brief Template functor to compute the inverse of a scalar - * \sa class CwiseUnaryOp, Cwise::inverse() - */ -template -struct scalar_inverse_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_inverse_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return Scalar(1)/a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pdiv(pset1(Scalar(1)),a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasDiv }; }; - -/** \internal - * \brief Template functor to compute the square of a scalar - * \sa class CwiseUnaryOp, Cwise::square() - */ -template -struct scalar_square_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_square_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return a*a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pmul(a,a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasMul }; }; - -/** \internal - * \brief Template functor to compute the cube of a scalar - * \sa class CwiseUnaryOp, Cwise::cube() - */ -template -struct scalar_cube_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cube_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return a*a*a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pmul(a,pmul(a,a)); } -}; -template -struct functor_traits > -{ enum { Cost = 2*NumTraits::MulCost, PacketAccess = packet_traits::HasMul }; }; - -/** \internal - * \brief Template functor to compute the rounded value of a scalar - * \sa class CwiseUnaryOp, ArrayBase::round() - */ -template struct scalar_round_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_round_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::round(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pround(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasRound - }; -}; - -/** \internal - * \brief Template functor to compute the floor of a scalar - * \sa class CwiseUnaryOp, ArrayBase::floor() - */ -template struct scalar_floor_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_floor_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::floor(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pfloor(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasFloor - }; -}; - -/** \internal - * \brief Template functor to compute the ceil of a scalar - * \sa class CwiseUnaryOp, ArrayBase::ceil() - */ -template struct scalar_ceil_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_ceil_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::ceil(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pceil(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasCeil - }; -}; - -/** \internal - * \brief Template functor to compute whether a scalar is NaN - * \sa class CwiseUnaryOp, ArrayBase::isnan() - */ -template struct scalar_isnan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isnan_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isnan)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to check whether a scalar is +/-inf - * \sa class CwiseUnaryOp, ArrayBase::isinf() - */ -template struct scalar_isinf_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isinf_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isinf)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to check whether a scalar has a finite value - * \sa class CwiseUnaryOp, ArrayBase::isfinite() - */ -template struct scalar_isfinite_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isfinite_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isfinite)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the logical not of a boolean - * - * \sa class CwiseUnaryOp, ArrayBase::operator! - */ -template struct scalar_boolean_not_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_not_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a) const { return !a; } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the signum of a scalar - * \sa class CwiseUnaryOp, Cwise::sign() - */ -template::IsComplex!=0) > struct scalar_sign_op; -template -struct scalar_sign_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sign_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const - { - return Scalar( (a>Scalar(0)) - (a - //EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psign(a); } -}; -template -struct scalar_sign_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sign_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const - { - typedef typename NumTraits::Real real_type; - real_type aa = numext::abs(a); - if (aa==real_type(0)) - return Scalar(0); - aa = real_type(1)/aa; - return Scalar(a.real()*aa, a.imag()*aa ); - } - //TODO - //template - //EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psign(a); } -}; -template -struct functor_traits > -{ enum { - Cost = - NumTraits::IsComplex - ? ( 8*NumTraits::MulCost ) // roughly - : ( 3*NumTraits::AddCost), - PacketAccess = packet_traits::HasSign - }; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h deleted file mode 100644 index 681451cc309..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ /dev/null @@ -1,2157 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_BLOCK_PANEL_H -#define EIGEN_GENERAL_BLOCK_PANEL_H - - -namespace Eigen { - -namespace internal { - -template -class gebp_traits; - - -/** \internal \returns b if a<=0, and returns a otherwise. */ -inline std::ptrdiff_t manage_caching_sizes_helper(std::ptrdiff_t a, std::ptrdiff_t b) -{ - return a<=0 ? b : a; -} - -#if EIGEN_ARCH_i386_OR_x86_64 -const std::ptrdiff_t defaultL1CacheSize = 32*1024; -const std::ptrdiff_t defaultL2CacheSize = 256*1024; -const std::ptrdiff_t defaultL3CacheSize = 2*1024*1024; -#else -const std::ptrdiff_t defaultL1CacheSize = 16*1024; -const std::ptrdiff_t defaultL2CacheSize = 512*1024; -const std::ptrdiff_t defaultL3CacheSize = 512*1024; -#endif - -/** \internal */ -struct CacheSizes { - CacheSizes(): m_l1(-1),m_l2(-1),m_l3(-1) { - int l1CacheSize, l2CacheSize, l3CacheSize; - queryCacheSizes(l1CacheSize, l2CacheSize, l3CacheSize); - m_l1 = manage_caching_sizes_helper(l1CacheSize, defaultL1CacheSize); - m_l2 = manage_caching_sizes_helper(l2CacheSize, defaultL2CacheSize); - m_l3 = manage_caching_sizes_helper(l3CacheSize, defaultL3CacheSize); - } - - std::ptrdiff_t m_l1; - std::ptrdiff_t m_l2; - std::ptrdiff_t m_l3; -}; - - -/** \internal */ -inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff_t* l2, std::ptrdiff_t* l3) -{ - static CacheSizes m_cacheSizes; - - if(action==SetAction) - { - // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1!=0 && l2!=0); - m_cacheSizes.m_l1 = *l1; - m_cacheSizes.m_l2 = *l2; - m_cacheSizes.m_l3 = *l3; - } - else if(action==GetAction) - { - eigen_internal_assert(l1!=0 && l2!=0); - *l1 = m_cacheSizes.m_l1; - *l2 = m_cacheSizes.m_l2; - *l3 = m_cacheSizes.m_l3; - } - else - { - eigen_internal_assert(false); - } -} - -/* Helper for computeProductBlockingSizes. - * - * Given a m x k times k x n matrix product of scalar types \c LhsScalar and \c RhsScalar, - * this function computes the blocking size parameters along the respective dimensions - * for matrix products and related algorithms. The blocking sizes depends on various - * parameters: - * - the L1 and L2 cache sizes, - * - the register level blocking sizes defined by gebp_traits, - * - the number of scalars that fit into a packet (when vectorization is enabled). - * - * \sa setCpuCacheSizes */ - -template -void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - typedef gebp_traits Traits; - - // Explanations: - // Let's recall that the product algorithms form mc x kc vertical panels A' on the lhs and - // kc x nc blocks B' on the rhs. B' has to fit into L2/L3 cache. Moreover, A' is processed - // per mr x kc horizontal small panels where mr is the blocking size along the m dimension - // at the register level. This small horizontal panel has to stay within L1 cache. - std::ptrdiff_t l1, l2, l3; - manage_caching_sizes(GetAction, &l1, &l2, &l3); - - if (num_threads > 1) { - typedef typename Traits::ResScalar ResScalar; - enum { - kdiv = KcFactor * (Traits::mr * sizeof(LhsScalar) + Traits::nr * sizeof(RhsScalar)), - ksub = Traits::mr * Traits::nr * sizeof(ResScalar), - kr = 8, - mr = Traits::mr, - nr = Traits::nr - }; - // Increasing k gives us more time to prefetch the content of the "C" - // registers. However once the latency is hidden there is no point in - // increasing the value of k, so we'll cap it at 320 (value determined - // experimentally). - // To avoid that k vanishes, we make k_cache at least as big as kr - const Index k_cache = numext::maxi(kr, (numext::mini)((l1-ksub)/kdiv, 320)); - if (k_cache < k) { - k = k_cache - (k_cache % kr); - eigen_internal_assert(k > 0); - } - - const Index n_cache = (l2-l1) / (nr * sizeof(RhsScalar) * k); - const Index n_per_thread = numext::div_ceil(n, num_threads); - if (n_cache <= n_per_thread) { - // Don't exceed the capacity of the l2 cache. - eigen_internal_assert(n_cache >= static_cast(nr)); - n = n_cache - (n_cache % nr); - eigen_internal_assert(n > 0); - } else { - n = (numext::mini)(n, (n_per_thread + nr - 1) - ((n_per_thread + nr - 1) % nr)); - } - - if (l3 > l2) { - // l3 is shared between all cores, so we'll give each thread its own chunk of l3. - const Index m_cache = (l3-l2) / (sizeof(LhsScalar) * k * num_threads); - const Index m_per_thread = numext::div_ceil(m, num_threads); - if(m_cache < m_per_thread && m_cache >= static_cast(mr)) { - m = m_cache - (m_cache % mr); - eigen_internal_assert(m > 0); - } else { - m = (numext::mini)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); - } - } - } - else { - // In unit tests we do not want to use extra large matrices, - // so we reduce the cache size to check the blocking strategy is not flawed -#ifdef EIGEN_DEBUG_SMALL_PRODUCT_BLOCKS - l1 = 9*1024; - l2 = 32*1024; - l3 = 512*1024; -#endif - - // Early return for small problems because the computation below are time consuming for small problems. - // Perhaps it would make more sense to consider k*n*m?? - // Note that for very tiny problem, this function should be bypassed anyway - // because we use the coefficient-based implementation for them. - if((numext::maxi)(k,(numext::maxi)(m,n))<48) - return; - - typedef typename Traits::ResScalar ResScalar; - enum { - k_peeling = 8, - k_div = KcFactor * (Traits::mr * sizeof(LhsScalar) + Traits::nr * sizeof(RhsScalar)), - k_sub = Traits::mr * Traits::nr * sizeof(ResScalar) - }; - - // ---- 1st level of blocking on L1, yields kc ---- - - // Blocking on the third dimension (i.e., k) is chosen so that an horizontal panel - // of size mr x kc of the lhs plus a vertical panel of kc x nr of the rhs both fits within L1 cache. - // We also include a register-level block of the result (mx x nr). - // (In an ideal world only the lhs panel would stay in L1) - // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi(((l1-k_sub)/k_div) & (~(k_peeling-1)),1); - const Index old_k = k; - if(k>max_kc) - { - // We are really blocking on the third dimension: - // -> reduce blocking size to make sure the last block is as large as possible - // while keeping the same number of sweeps over the result. - k = (k%max_kc)==0 ? max_kc - : max_kc - k_peeling * ((max_kc-1-(k%max_kc))/(k_peeling*(k/max_kc+1))); - - eigen_internal_assert(((old_k/k) == (old_k/max_kc)) && "the number of sweeps has to remain the same"); - } - - // ---- 2nd level of blocking on max(L2,L3), yields nc ---- - - // TODO find a reliable way to get the actual amount of cache per core to use for 2nd level blocking, that is: - // actual_l2 = max(l2, l3/nb_core_sharing_l3) - // The number below is quite conservative: it is better to underestimate the cache size rather than overestimating it) - // For instance, it corresponds to 6MB of L3 shared among 4 cores. - #ifdef EIGEN_DEBUG_SMALL_PRODUCT_BLOCKS - const Index actual_l2 = l3; - #else - const Index actual_l2 = 1572864; // == 1.5 MB - #endif - - // Here, nc is chosen such that a block of kc x nc of the rhs fit within half of L2. - // The second half is implicitly reserved to access the result and lhs coefficients. - // When k= Index(Traits::nr*sizeof(RhsScalar))*k) - { - // L1 blocking - max_nc = remaining_l1 / (k*sizeof(RhsScalar)); - } - else - { - // L2 blocking - max_nc = (3*actual_l2)/(2*2*max_kc*sizeof(RhsScalar)); - } - // WARNING Below, we assume that Traits::nr is a power of two. - Index nc = numext::mini(actual_l2/(2*k*sizeof(RhsScalar)), max_nc) & (~(Traits::nr-1)); - if(n>nc) - { - // We are really blocking over the columns: - // -> reduce blocking size to make sure the last block is as large as possible - // while keeping the same number of sweeps over the packed lhs. - // Here we allow one more sweep if this gives us a perfect match, thus the commented "-1" - n = (n%nc)==0 ? nc - : (nc - Traits::nr * ((nc/*-1*/-(n%nc))/(Traits::nr*(n/nc+1)))); - } - else if(old_k==k) - { - // So far, no blocking at all, i.e., kc==k, and nc==n. - // In this case, let's perform a blocking over the rows such that the packed lhs data is kept in cache L1/L2 - // TODO: part of this blocking strategy is now implemented within the kernel itself, so the L1-based heuristic here should be obsolete. - Index problem_size = k*n*sizeof(LhsScalar); - Index actual_lm = actual_l2; - Index max_mc = m; - if(problem_size<=1024) - { - // problem is small enough to keep in L1 - // Let's choose m such that lhs's block fit in 1/3 of L1 - actual_lm = l1; - } - else if(l3!=0 && problem_size<=32768) - { - // we have both L2 and L3, and problem is small enough to be kept in L2 - // Let's choose m such that lhs's block fit in 1/3 of L2 - actual_lm = l2; - max_mc = (numext::mini)(576,max_mc); - } - Index mc = (numext::mini)(actual_lm/(3*k*sizeof(LhsScalar)), max_mc); - if (mc > Traits::mr) mc -= mc % Traits::mr; - else if (mc==0) return; - m = (m%mc)==0 ? mc - : (mc - Traits::mr * ((mc/*-1*/-(m%mc))/(Traits::mr*(m/mc+1)))); - } - } -} - -template -inline bool useSpecificBlockingSizes(Index& k, Index& m, Index& n) -{ -#ifdef EIGEN_TEST_SPECIFIC_BLOCKING_SIZES - if (EIGEN_TEST_SPECIFIC_BLOCKING_SIZES) { - k = numext::mini(k, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_K); - m = numext::mini(m, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_M); - n = numext::mini(n, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_N); - return true; - } -#else - EIGEN_UNUSED_VARIABLE(k) - EIGEN_UNUSED_VARIABLE(m) - EIGEN_UNUSED_VARIABLE(n) -#endif - return false; -} - -/** \brief Computes the blocking parameters for a m x k times k x n matrix product - * - * \param[in,out] k Input: the third dimension of the product. Output: the blocking size along the same dimension. - * \param[in,out] m Input: the number of rows of the left hand side. Output: the blocking size along the same dimension. - * \param[in,out] n Input: the number of columns of the right hand side. Output: the blocking size along the same dimension. - * - * Given a m x k times k x n matrix product of scalar types \c LhsScalar and \c RhsScalar, - * this function computes the blocking size parameters along the respective dimensions - * for matrix products and related algorithms. - * - * The blocking size parameters may be evaluated: - * - either by a heuristic based on cache sizes; - * - or using fixed prescribed values (for testing purposes). - * - * \sa setCpuCacheSizes */ - -template -void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - if (!useSpecificBlockingSizes(k, m, n)) { - evaluateProductBlockingSizesHeuristic(k, m, n, num_threads); - } -} - -template -inline void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - computeProductBlockingSizes(k, m, n, num_threads); -} - -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD - #define CJMADD(CJ,A,B,C,T) C = CJ.pmadd(A,B,C); -#else - - // FIXME (a bit overkill maybe ?) - - template struct gebp_madd_selector { - EIGEN_ALWAYS_INLINE static void run(const CJ& cj, A& a, B& b, C& c, T& /*t*/) - { - c = cj.pmadd(a,b,c); - } - }; - - template struct gebp_madd_selector { - EIGEN_ALWAYS_INLINE static void run(const CJ& cj, T& a, T& b, T& c, T& t) - { - t = b; t = cj.pmul(a,t); c = padd(c,t); - } - }; - - template - EIGEN_STRONG_INLINE void gebp_madd(const CJ& cj, A& a, B& b, C& c, T& t) - { - gebp_madd_selector::run(cj,a,b,c,t); - } - - #define CJMADD(CJ,A,B,C,T) gebp_madd(CJ,A,B,C,T); -// #define CJMADD(CJ,A,B,C,T) T = B; T = CJ.pmul(A,T); C = padd(C,T); -#endif - -/* Vectorization logic - * real*real: unpack rhs to constant packets, ... - * - * cd*cd : unpack rhs to (b_r,b_r), (b_i,b_i), mul to get (a_r b_r,a_i b_r) (a_r b_i,a_i b_i), - * storing each res packet into two packets (2x2), - * at the end combine them: swap the second and addsub them - * cf*cf : same but with 2x4 blocks - * cplx*real : unpack rhs to constant packets, ... - * real*cplx : load lhs as (a0,a0,a1,a1), and mul as usual - */ -template -class gebp_traits -{ -public: - typedef _LhsScalar LhsScalar; - typedef _RhsScalar RhsScalar; - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - - // register block size along the N direction must be 1 or 4 - nr = 4, - - // register block size along the M direction (currently, this one cannot be modified) - default_mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*LhsPacketSize, -#if defined(EIGEN_HAS_SINGLE_INSTRUCTION_MADD) && !defined(EIGEN_VECTORIZE_ALTIVEC) && !defined(EIGEN_VECTORIZE_VSX) - // we assume 16 registers - // See bug 992, if the scalar type is not vectorizable but that EIGEN_HAS_SINGLE_INSTRUCTION_MADD is defined, - // then using 3*LhsPacketSize triggers non-implemented paths in syrk. - mr = Vectorizable ? 3*LhsPacketSize : default_mr, -#else - mr = default_mr, -#endif - - LhsProgress = LhsPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// pbroadcast2(b, b0, b1); -// } - - template - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacketType& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - dest = ploadquad(b); - } - - template - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacketType& dest) const - { - dest = pload(a); - } - - template - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacketType& dest) const - { - dest = ploadu(a); - } - - template - EIGEN_STRONG_INLINE void madd(const LhsPacketType& a, const RhsPacketType& b, AccPacketType& c, AccPacketType& tmp) const - { - conj_helper cj; - // It would be a lot cleaner to call pmadd all the time. Unfortunately if we - // let gcc allocate the register in which to store the result of the pmul - // (in the case where there is no FMA) gcc fails to figure out how to avoid - // spilling register. -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c = cj.pmadd(a,b,c); -#else - tmp = b; tmp = cj.pmul(a,tmp); c = padd(c,tmp); -#endif - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = pmadd(c,alpha,r); - } - - template - EIGEN_STRONG_INLINE void acc(const ResPacketHalf& c, const ResPacketHalf& alpha, ResPacketHalf& r) const - { - r = pmadd(c,alpha,r); - } - -}; - -template -class gebp_traits, RealScalar, _ConjLhs, false> -{ -public: - typedef std::complex LhsScalar; - typedef RealScalar RhsScalar; - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = false, - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - nr = 4, -#if defined(EIGEN_HAS_SINGLE_INSTRUCTION_MADD) && !defined(EIGEN_VECTORIZE_ALTIVEC) && !defined(EIGEN_VECTORIZE_VSX) - // we assume 16 registers - mr = 3*LhsPacketSize, -#else - mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*LhsPacketSize, -#endif - - LhsProgress = LhsPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = pload(a); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploadu(a); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// pbroadcast2(b, b0, b1); -// } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp) const - { - madd_impl(a, b, c, tmp, typename conditional::type()); - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp, const true_type&) const - { -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c.v = pmadd(a.v,b,c.v); -#else - tmp = b; tmp = pmul(a.v,tmp); c.v = padd(c.v,tmp); -#endif - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsScalar& a, const RhsScalar& b, ResScalar& c, RhsScalar& /*tmp*/, const false_type&) const - { - c += a * b; - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = cj.pmadd(c,alpha,r); - } - -protected: - conj_helper cj; -}; - -template -struct DoublePacket -{ - Packet first; - Packet second; -}; - -template -DoublePacket padd(const DoublePacket &a, const DoublePacket &b) -{ - DoublePacket res; - res.first = padd(a.first, b.first); - res.second = padd(a.second,b.second); - return res; -} - -template -const DoublePacket& predux_downto4(const DoublePacket &a) -{ - return a; -} - -template struct unpacket_traits > { typedef DoublePacket half; }; -// template -// DoublePacket pmadd(const DoublePacket &a, const DoublePacket &b) -// { -// DoublePacket res; -// res.first = padd(a.first, b.first); -// res.second = padd(a.second,b.second); -// return res; -// } - -template -class gebp_traits, std::complex, _ConjLhs, _ConjRhs > -{ -public: - typedef std::complex Scalar; - typedef std::complex LhsScalar; - typedef std::complex RhsScalar; - typedef std::complex ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable - && packet_traits::Vectorizable, - RealPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - - // FIXME: should depend on NumberOfRegisters - nr = 4, - mr = ResPacketSize, - - LhsProgress = ResPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type RealPacket; - typedef typename packet_traits::type ScalarPacket; - typedef DoublePacket DoublePacketType; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - typedef typename conditional::type AccPacket; - - EIGEN_STRONG_INLINE void initAcc(Scalar& p) { p = Scalar(0); } - - EIGEN_STRONG_INLINE void initAcc(DoublePacketType& p) - { - p.first = pset1(RealScalar(0)); - p.second = pset1(RealScalar(0)); - } - - // Scalar path - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, ResPacket& dest) const - { - dest = pset1(*b); - } - - // Vectorized path - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, DoublePacketType& dest) const - { - dest.first = pset1(numext::real(*b)); - dest.second = pset1(numext::imag(*b)); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, ResPacket& dest) const - { - loadRhs(b,dest); - } - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, DoublePacketType& dest) const - { - eigen_internal_assert(unpacket_traits::size<=4); - loadRhs(b,dest); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - loadRhs(b+2, b2); - loadRhs(b+3, b3); - } - - // Vectorized path - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, DoublePacketType& b0, DoublePacketType& b1) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - } - - // Scalar path - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsScalar& b0, RhsScalar& b1) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - } - - // nothing special here - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = pload((const typename unpacket_traits::type*)(a)); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploadu((const typename unpacket_traits::type*)(a)); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, DoublePacketType& c, RhsPacket& /*tmp*/) const - { - c.first = padd(pmul(a,b.first), c.first); - c.second = padd(pmul(a,b.second),c.second); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, ResPacket& c, RhsPacket& /*tmp*/) const - { - c = cj.pmadd(a,b,c); - } - - EIGEN_STRONG_INLINE void acc(const Scalar& c, const Scalar& alpha, Scalar& r) const { r += alpha * c; } - - EIGEN_STRONG_INLINE void acc(const DoublePacketType& c, const ResPacket& alpha, ResPacket& r) const - { - // assemble c - ResPacket tmp; - if((!ConjLhs)&&(!ConjRhs)) - { - tmp = pcplxflip(pconj(ResPacket(c.second))); - tmp = padd(ResPacket(c.first),tmp); - } - else if((!ConjLhs)&&(ConjRhs)) - { - tmp = pconj(pcplxflip(ResPacket(c.second))); - tmp = padd(ResPacket(c.first),tmp); - } - else if((ConjLhs)&&(!ConjRhs)) - { - tmp = pcplxflip(ResPacket(c.second)); - tmp = padd(pconj(ResPacket(c.first)),tmp); - } - else if((ConjLhs)&&(ConjRhs)) - { - tmp = pcplxflip(ResPacket(c.second)); - tmp = psub(pconj(ResPacket(c.first)),tmp); - } - - r = pmadd(tmp,alpha,r); - } - -protected: - conj_helper cj; -}; - -template -class gebp_traits, false, _ConjRhs > -{ -public: - typedef std::complex Scalar; - typedef RealScalar LhsScalar; - typedef Scalar RhsScalar; - typedef Scalar ResScalar; - - enum { - ConjLhs = false, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable - && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - // FIXME: should depend on NumberOfRegisters - nr = 4, - mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*ResPacketSize, - - LhsProgress = ResPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// // FIXME not sure that's the best way to implement it! -// b0 = pload1(b+0); -// b1 = pload1(b+1); -// } - - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploaddup(a); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - eigen_internal_assert(unpacket_traits::size<=4); - loadRhs(b,dest); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploaddup(a); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp) const - { - madd_impl(a, b, c, tmp, typename conditional::type()); - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp, const true_type&) const - { -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c.v = pmadd(a,b.v,c.v); -#else - tmp = b; tmp.v = pmul(a,tmp.v); c = padd(c,tmp); -#endif - - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsScalar& a, const RhsScalar& b, ResScalar& c, RhsScalar& /*tmp*/, const false_type&) const - { - c += a * b; - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = cj.pmadd(alpha,c,r); - } - -protected: - conj_helper cj; -}; - -/* optimized GEneral packed Block * packed Panel product kernel - * - * Mixing type logic: C += A * B - * | A | B | comments - * |real |cplx | no vectorization yet, would require to pack A with duplication - * |cplx |real | easy vectorization - */ -template -struct gebp_kernel -{ - typedef gebp_traits Traits; - typedef typename Traits::ResScalar ResScalar; - typedef typename Traits::LhsPacket LhsPacket; - typedef typename Traits::RhsPacket RhsPacket; - typedef typename Traits::ResPacket ResPacket; - typedef typename Traits::AccPacket AccPacket; - - typedef gebp_traits SwappedTraits; - typedef typename SwappedTraits::ResScalar SResScalar; - typedef typename SwappedTraits::LhsPacket SLhsPacket; - typedef typename SwappedTraits::RhsPacket SRhsPacket; - typedef typename SwappedTraits::ResPacket SResPacket; - typedef typename SwappedTraits::AccPacket SAccPacket; - - typedef typename DataMapper::LinearMapper LinearMapper; - - enum { - Vectorizable = Traits::Vectorizable, - LhsProgress = Traits::LhsProgress, - RhsProgress = Traits::RhsProgress, - ResPacketSize = Traits::ResPacketSize - }; - - EIGEN_DONT_INLINE - void operator()(const DataMapper& res, const LhsScalar* blockA, const RhsScalar* blockB, - Index rows, Index depth, Index cols, ResScalar alpha, - Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0); -}; - -template -EIGEN_DONT_INLINE -void gebp_kernel - ::operator()(const DataMapper& res, const LhsScalar* blockA, const RhsScalar* blockB, - Index rows, Index depth, Index cols, ResScalar alpha, - Index strideA, Index strideB, Index offsetA, Index offsetB) - { - Traits traits; - SwappedTraits straits; - - if(strideA==-1) strideA = depth; - if(strideB==-1) strideB = depth; - conj_helper cj; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - const Index peeled_mc3 = mr>=3*Traits::LhsProgress ? (rows/(3*LhsProgress))*(3*LhsProgress) : 0; - const Index peeled_mc2 = mr>=2*Traits::LhsProgress ? peeled_mc3+((rows-peeled_mc3)/(2*LhsProgress))*(2*LhsProgress) : 0; - const Index peeled_mc1 = mr>=1*Traits::LhsProgress ? (rows/(1*LhsProgress))*(1*LhsProgress) : 0; - enum { pk = 8 }; // NOTE Such a large peeling factor is important for large matrices (~ +5% when >1000 on Haswell) - const Index peeled_kc = depth & ~(pk-1); - const Index prefetch_res_offset = 32/sizeof(ResScalar); -// const Index depth2 = depth & ~1; - - //---------- Process 3 * LhsProgress rows at once ---------- - // This corresponds to 3*LhsProgress x nr register blocks. - // Usually, make sense only with FMA - if(mr>=3*Traits::LhsProgress) - { - // Here, the general idea is to loop on each largest micro horizontal panel of the lhs (3*Traits::LhsProgress x depth) - // and on each largest micro vertical panel of the rhs (depth * nr). - // Blocking sizes, i.e., 'depth' has been computed so that the micro horizontal panel of the lhs fit in L1. - // However, if depth is too small, we can extend the number of rows of these horizontal panels. - // This actual number of rows is computed as follow: - const Index l1 = defaultL1CacheSize; // in Bytes, TODO, l1 should be passed to this function. - // The max(1, ...) here is needed because we may be using blocking params larger than what our known l1 cache size - // suggests we should be using: either because our known l1 cache size is inaccurate (e.g. on Android, we can only guess), - // or because we are testing specific blocking sizes. - const Index actual_panel_rows = (3*LhsProgress) * std::max(1,( (l1 - sizeof(ResScalar)*mr*nr - depth*nr*sizeof(RhsScalar)) / (depth * sizeof(LhsScalar) * 3*LhsProgress) )); - for(Index i1=0; i1(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r0.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C8, alphav, R2); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r0.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r1.loadPacket(0 * Traits::ResPacketSize); - R1 = r1.loadPacket(1 * Traits::ResPacketSize); - R2 = r1.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C1, alphav, R0); - traits.acc(C5, alphav, R1); - traits.acc(C9, alphav, R2); - r1.storePacket(0 * Traits::ResPacketSize, R0); - r1.storePacket(1 * Traits::ResPacketSize, R1); - r1.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r2.loadPacket(1 * Traits::ResPacketSize); - R2 = r2.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C6, alphav, R1); - traits.acc(C10, alphav, R2); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r2.storePacket(1 * Traits::ResPacketSize, R1); - r2.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r3.loadPacket(0 * Traits::ResPacketSize); - R1 = r3.loadPacket(1 * Traits::ResPacketSize); - R2 = r3.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C3, alphav, R0); - traits.acc(C7, alphav, R1); - traits.acc(C11, alphav, R2); - r3.storePacket(0 * Traits::ResPacketSize, R0); - r3.storePacket(1 * Traits::ResPacketSize, R1); - r3.storePacket(2 * Traits::ResPacketSize, R2); - } - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r0.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C8, alphav, R2); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r0.storePacket(2 * Traits::ResPacketSize, R2); - } - } - } - } - - //---------- Process 2 * LhsProgress rows at once ---------- - if(mr>=2*Traits::LhsProgress) - { - const Index l1 = defaultL1CacheSize; // in Bytes, TODO, l1 should be passed to this function. - // The max(1, ...) here is needed because we may be using blocking params larger than what our known l1 cache size - // suggests we should be using: either because our known l1 cache size is inaccurate (e.g. on Android, we can only guess), - // or because we are testing specific blocking sizes. - Index actual_panel_rows = (2*LhsProgress) * std::max(1,( (l1 - sizeof(ResScalar)*mr*nr - depth*nr*sizeof(RhsScalar)) / (depth * sizeof(LhsScalar) * 2*LhsProgress) )); - - for(Index i1=peeled_mc3; i1=6 without FMA (bug 1637) - #if EIGEN_GNUC_AT_LEAST(6,0) && defined(EIGEN_VECTORIZE_SSE) - #define EIGEN_GEBP_2PX4_SPILLING_WORKAROUND __asm__ ("" : [a0] "+x,m" (A0),[a1] "+x,m" (A1)); - #else - #define EIGEN_GEBP_2PX4_SPILLING_WORKAROUND - #endif - #define EIGEN_GEBGP_ONESTEP(K) \ - do { \ - EIGEN_ASM_COMMENT("begin step of gebp micro kernel 2pX4"); \ - traits.loadLhs(&blA[(0+2*K)*LhsProgress], A0); \ - traits.loadLhs(&blA[(1+2*K)*LhsProgress], A1); \ - traits.broadcastRhs(&blB[(0+4*K)*RhsProgress], B_0, B1, B2, B3); \ - traits.madd(A0, B_0, C0, T0); \ - traits.madd(A1, B_0, C4, B_0); \ - traits.madd(A0, B1, C1, T0); \ - traits.madd(A1, B1, C5, B1); \ - traits.madd(A0, B2, C2, T0); \ - traits.madd(A1, B2, C6, B2); \ - traits.madd(A0, B3, C3, T0); \ - traits.madd(A1, B3, C7, B3); \ - EIGEN_GEBP_2PX4_SPILLING_WORKAROUND \ - EIGEN_ASM_COMMENT("end step of gebp micro kernel 2pX4"); \ - } while(false) - - internal::prefetch(blB+(48+0)); - EIGEN_GEBGP_ONESTEP(0); - EIGEN_GEBGP_ONESTEP(1); - EIGEN_GEBGP_ONESTEP(2); - EIGEN_GEBGP_ONESTEP(3); - internal::prefetch(blB+(48+16)); - EIGEN_GEBGP_ONESTEP(4); - EIGEN_GEBGP_ONESTEP(5); - EIGEN_GEBGP_ONESTEP(6); - EIGEN_GEBGP_ONESTEP(7); - - blB += pk*4*RhsProgress; - blA += pk*(2*Traits::LhsProgress); - - EIGEN_ASM_COMMENT("end gebp micro kernel 2pX4"); - } - // process remaining peeled loop - for(Index k=peeled_kc; k(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r1.loadPacket(0 * Traits::ResPacketSize); - R3 = r1.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C1, alphav, R2); - traits.acc(C5, alphav, R3); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r1.storePacket(0 * Traits::ResPacketSize, R2); - r1.storePacket(1 * Traits::ResPacketSize, R3); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r2.loadPacket(1 * Traits::ResPacketSize); - R2 = r3.loadPacket(0 * Traits::ResPacketSize); - R3 = r3.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C6, alphav, R1); - traits.acc(C3, alphav, R2); - traits.acc(C7, alphav, R3); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r2.storePacket(1 * Traits::ResPacketSize, R1); - r3.storePacket(0 * Traits::ResPacketSize, R2); - r3.storePacket(1 * Traits::ResPacketSize, R3); - } - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - } - } - } - } - //---------- Process 1 * LhsProgress rows at once ---------- - if(mr>=1*Traits::LhsProgress) - { - // loops on each largest micro horizontal panel of lhs (1*LhsProgress x depth) - for(Index i=peeled_mc2; i(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r1.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C1, alphav, R1); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r1.storePacket(0 * Traits::ResPacketSize, R1); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r3.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C3, alphav, R1); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r3.storePacket(0 * Traits::ResPacketSize, R1); - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - r0.storePacket(0 * Traits::ResPacketSize, R0); - } - } - } - //---------- Process remaining rows, 1 at once ---------- - if(peeled_mc1::half>::size; - if ((SwappedTraits::LhsProgress % 4) == 0 && - (SwappedTraits::LhsProgress <= 8) && - (SwappedTraits::LhsProgress!=8 || SResPacketHalfSize==nr)) - { - SAccPacket C0, C1, C2, C3; - straits.initAcc(C0); - straits.initAcc(C1); - straits.initAcc(C2); - straits.initAcc(C3); - - const Index spk = (std::max)(1,SwappedTraits::LhsProgress/4); - const Index endk = (depth/spk)*spk; - const Index endk4 = (depth/(spk*4))*(spk*4); - - Index k=0; - for(; k=8,typename unpacket_traits::half,SResPacket>::type SResPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SLhsPacket>::type SLhsPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SRhsPacket>::type SRhsPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SAccPacket>::type SAccPacketHalf; - - SResPacketHalf R = res.template gatherPacket(i, j2); - SResPacketHalf alphav = pset1(alpha); - - if(depth-endk>0) - { - // We have to handle the last row of the rhs which corresponds to a half-packet - SLhsPacketHalf a0; - SRhsPacketHalf b0; - straits.loadLhsUnaligned(blB, a0); - straits.loadRhs(blA, b0); - SAccPacketHalf c0 = predux_downto4(C0); - straits.madd(a0,b0,c0,b0); - straits.acc(c0, alphav, R); - } - else - { - straits.acc(predux_downto4(C0), alphav, R); - } - res.scatterPacket(i, j2, R); - } - else - { - SResPacket R = res.template gatherPacket(i, j2); - SResPacket alphav = pset1(alpha); - straits.acc(C0, alphav, R); - res.scatterPacket(i, j2, R); - } - } - else // scalar path - { - // get a 1 x 4 res block as registers - ResScalar C0(0), C1(0), C2(0), C3(0); - - for(Index k=0; k -struct gemm_pack_lhs -{ - typedef typename DataMapper::LinearMapper LinearMapper; - EIGEN_DONT_INLINE void operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_lhs - ::operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride, Index offset) -{ - typedef typename packet_traits::type Packet; - enum { PacketSize = packet_traits::size }; - - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - eigen_assert( ((Pack1%PacketSize)==0 && Pack1<=4*PacketSize) || (Pack1<=4) ); - conj_if::IsComplex && Conjugate> cj; - Index count = 0; - - const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; - const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; - const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - const Index peeled_mc0 = Pack2>=1*PacketSize ? peeled_mc1 - : Pack2>1 ? (rows/Pack2)*Pack2 : 0; - - Index i=0; - - // Pack 3 packets - if(Pack1>=3*PacketSize) - { - for(; i=2*PacketSize) - { - for(; i=1*PacketSize) - { - for(; i1) - { - for(; i -struct gemm_pack_lhs -{ - typedef typename DataMapper::LinearMapper LinearMapper; - EIGEN_DONT_INLINE void operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_lhs - ::operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride, Index offset) -{ - typedef typename packet_traits::type Packet; - enum { PacketSize = packet_traits::size }; - - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index count = 0; - -// const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; -// const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; -// const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - - int pack = Pack1; - Index i = 0; - while(pack>0) - { - Index remaining_rows = rows-i; - Index peeled_mc = i+(remaining_rows/pack)*pack; - for(; i=PacketSize) - { - for(; k kernel; - for (int p = 0; p < PacketSize; ++p) kernel.packet[p] = lhs.loadPacket(i+p+m, k); - ptranspose(kernel); - for (int p = 0; p < PacketSize; ++p) pstore(blockA+count+m+(pack)*p, cj.pconj(kernel.packet[p])); - } - count += PacketSize*pack; - } - } - for(; k -struct gemm_pack_rhs -{ - typedef typename packet_traits::type Packet; - typedef typename DataMapper::LinearMapper LinearMapper; - enum { PacketSize = packet_traits::size }; - EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_rhs - ::operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride, Index offset) -{ - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS COLMAJOR"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - Index count = 0; - const Index peeled_k = (depth/PacketSize)*PacketSize; -// if(nr>=8) -// { -// for(Index j2=0; j2 kernel; -// for (int p = 0; p < PacketSize; ++p) { -// kernel.packet[p] = ploadu(&rhs[(j2+p)*rhsStride+k]); -// } -// ptranspose(kernel); -// for (int p = 0; p < PacketSize; ++p) { -// pstoreu(blockB+count, cj.pconj(kernel.packet[p])); -// count+=PacketSize; -// } -// } -// } -// for(; k=4) - { - for(Index j2=packet_cols8; j2 kernel; - kernel.packet[0] = dm0.loadPacket(k); - kernel.packet[1%PacketSize] = dm1.loadPacket(k); - kernel.packet[2%PacketSize] = dm2.loadPacket(k); - kernel.packet[3%PacketSize] = dm3.loadPacket(k); - ptranspose(kernel); - pstoreu(blockB+count+0*PacketSize, cj.pconj(kernel.packet[0])); - pstoreu(blockB+count+1*PacketSize, cj.pconj(kernel.packet[1%PacketSize])); - pstoreu(blockB+count+2*PacketSize, cj.pconj(kernel.packet[2%PacketSize])); - pstoreu(blockB+count+3*PacketSize, cj.pconj(kernel.packet[3%PacketSize])); - count+=4*PacketSize; - } - } - for(; k -struct gemm_pack_rhs -{ - typedef typename packet_traits::type Packet; - typedef typename DataMapper::LinearMapper LinearMapper; - enum { PacketSize = packet_traits::size }; - EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_rhs - ::operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride, Index offset) -{ - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - Index count = 0; - -// if(nr>=8) -// { -// for(Index j2=0; j2(&rhs[k*rhsStride + j2]); -// pstoreu(blockB+count, cj.pconj(A)); -// } else if (PacketSize==4) { -// Packet A = ploadu(&rhs[k*rhsStride + j2]); -// Packet B = ploadu(&rhs[k*rhsStride + j2 + PacketSize]); -// pstoreu(blockB+count, cj.pconj(A)); -// pstoreu(blockB+count+PacketSize, cj.pconj(B)); -// } else { -// const Scalar* b0 = &rhs[k*rhsStride + j2]; -// blockB[count+0] = cj(b0[0]); -// blockB[count+1] = cj(b0[1]); -// blockB[count+2] = cj(b0[2]); -// blockB[count+3] = cj(b0[3]); -// blockB[count+4] = cj(b0[4]); -// blockB[count+5] = cj(b0[5]); -// blockB[count+6] = cj(b0[6]); -// blockB[count+7] = cj(b0[7]); -// } -// count += 8; -// } -// // skip what we have after -// if(PanelMode) count += 8 * (stride-offset-depth); -// } -// } - if(nr>=4) - { - for(Index j2=packet_cols8; j2 -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_H -#define EIGEN_GENERAL_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -template class level3_blocking; - -/* Specialization for a row-major destination matrix => simple transposition of the product */ -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResInnerStride> -struct general_matrix_matrix_product -{ - typedef gebp_traits Traits; - - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, Index depth, - const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, - ResScalar* res, Index resIncr, Index resStride, - ResScalar alpha, - level3_blocking& blocking, - GemmParallelInfo* info = 0) - { - // transpose the product such that the result is column major - general_matrix_matrix_product - ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking,info); - } -}; - -/* Specialization for a col-major destination matrix - * => Blocking algorithm following Goto's paper */ -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResInnerStride> -struct general_matrix_matrix_product -{ - -typedef gebp_traits Traits; - -typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; -static void run(Index rows, Index cols, Index depth, - const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, - ResScalar alpha, - level3_blocking& blocking, - GemmParallelInfo* info = 0) -{ - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs, lhsStride); - RhsMapper rhs(_rhs, rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - Index nc = (std::min)(cols,blocking.nc()); // cache block size along the N direction - - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gebp_kernel gebp; - -#ifdef EIGEN_HAS_OPENMP - if(info) - { - // this is the parallel version! - int tid = omp_get_thread_num(); - int threads = omp_get_num_threads(); - - LhsScalar* blockA = blocking.blockA(); - eigen_internal_assert(blockA!=0); - - std::size_t sizeB = kc*nc; - ei_declare_aligned_stack_constructed_variable(RhsScalar, blockB, sizeB, 0); - - // For each horizontal panel of the rhs, and corresponding vertical panel of the lhs... - for(Index k=0; k rows of B', and cols of the A' - - // In order to reduce the chance that a thread has to wait for the other, - // let's start by packing B'. - pack_rhs(blockB, rhs.getSubMapper(k,0), actual_kc, nc); - - // Pack A_k to A' in a parallel fashion: - // each thread packs the sub block A_k,i to A'_i where i is the thread id. - - // However, before copying to A'_i, we have to make sure that no other thread is still using it, - // i.e., we test that info[tid].users equals 0. - // Then, we set info[tid].users to the number of threads to mark that all other threads are going to use it. - while(info[tid].users!=0) {} - info[tid].users += threads; - - pack_lhs(blockA+info[tid].lhs_start*actual_kc, lhs.getSubMapper(info[tid].lhs_start,k), actual_kc, info[tid].lhs_length); - - // Notify the other threads that the part A'_i is ready to go. - info[tid].sync = k; - - // Computes C_i += A' * B' per A'_i - for(int shift=0; shift0) { - while(info[i].sync!=k) { - } - } - - gebp(res.getSubMapper(info[i].lhs_start, 0), blockA+info[i].lhs_start*actual_kc, blockB, info[i].lhs_length, actual_kc, nc, alpha); - } - - // Then keep going as usual with the remaining B' - for(Index j=nc; j Pack lhs's panel into a sequential chunk of memory (L2/L3 caching) - // Note that this panel will be read as many times as the number of blocks in the rhs's - // horizontal panel which is, in practice, a very low number. - pack_lhs(blockA, lhs.getSubMapper(i2,k2), actual_kc, actual_mc); - - // For each kc x nc block of the rhs's horizontal panel... - for(Index j2=0; j2 -struct gemm_functor -{ - gemm_functor(const Lhs& lhs, const Rhs& rhs, Dest& dest, const Scalar& actualAlpha, BlockingType& blocking) - : m_lhs(lhs), m_rhs(rhs), m_dest(dest), m_actualAlpha(actualAlpha), m_blocking(blocking) - {} - - void initParallelSession(Index num_threads) const - { - m_blocking.initParallel(m_lhs.rows(), m_rhs.cols(), m_lhs.cols(), num_threads); - m_blocking.allocateA(); - } - - void operator() (Index row, Index rows, Index col=0, Index cols=-1, GemmParallelInfo* info=0) const - { - if(cols==-1) - cols = m_rhs.cols(); - - Gemm::run(rows, cols, m_lhs.cols(), - &m_lhs.coeffRef(row,0), m_lhs.outerStride(), - &m_rhs.coeffRef(0,col), m_rhs.outerStride(), - (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.innerStride(), m_dest.outerStride(), - m_actualAlpha, m_blocking, info); - } - - typedef typename Gemm::Traits Traits; - - protected: - const Lhs& m_lhs; - const Rhs& m_rhs; - Dest& m_dest; - Scalar m_actualAlpha; - BlockingType& m_blocking; -}; - -template class gemm_blocking_space; - -template -class level3_blocking -{ - typedef _LhsScalar LhsScalar; - typedef _RhsScalar RhsScalar; - - protected: - LhsScalar* m_blockA; - RhsScalar* m_blockB; - - Index m_mc; - Index m_nc; - Index m_kc; - - public: - - level3_blocking() - : m_blockA(0), m_blockB(0), m_mc(0), m_nc(0), m_kc(0) - {} - - inline Index mc() const { return m_mc; } - inline Index nc() const { return m_nc; } - inline Index kc() const { return m_kc; } - - inline LhsScalar* blockA() { return m_blockA; } - inline RhsScalar* blockB() { return m_blockB; } -}; - -template -class gemm_blocking_space - : public level3_blocking< - typename conditional::type, - typename conditional::type> -{ - enum { - Transpose = StorageOrder==RowMajor, - ActualRows = Transpose ? MaxCols : MaxRows, - ActualCols = Transpose ? MaxRows : MaxCols - }; - typedef typename conditional::type LhsScalar; - typedef typename conditional::type RhsScalar; - typedef gebp_traits Traits; - enum { - SizeA = ActualRows * MaxDepth, - SizeB = ActualCols * MaxDepth - }; - -#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES - EIGEN_ALIGN_MAX LhsScalar m_staticA[SizeA]; - EIGEN_ALIGN_MAX RhsScalar m_staticB[SizeB]; -#else - EIGEN_ALIGN_MAX char m_staticA[SizeA * sizeof(LhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1]; - EIGEN_ALIGN_MAX char m_staticB[SizeB * sizeof(RhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1]; -#endif - - public: - - gemm_blocking_space(Index /*rows*/, Index /*cols*/, Index /*depth*/, Index /*num_threads*/, bool /*full_rows = false*/) - { - this->m_mc = ActualRows; - this->m_nc = ActualCols; - this->m_kc = MaxDepth; -#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES - this->m_blockA = m_staticA; - this->m_blockB = m_staticB; -#else - this->m_blockA = reinterpret_cast((internal::UIntPtr(m_staticA) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1)); - this->m_blockB = reinterpret_cast((internal::UIntPtr(m_staticB) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1)); -#endif - } - - void initParallel(Index, Index, Index, Index) - {} - - inline void allocateA() {} - inline void allocateB() {} - inline void allocateAll() {} -}; - -template -class gemm_blocking_space - : public level3_blocking< - typename conditional::type, - typename conditional::type> -{ - enum { - Transpose = StorageOrder==RowMajor - }; - typedef typename conditional::type LhsScalar; - typedef typename conditional::type RhsScalar; - typedef gebp_traits Traits; - - Index m_sizeA; - Index m_sizeB; - - public: - - gemm_blocking_space(Index rows, Index cols, Index depth, Index num_threads, bool l3_blocking) - { - this->m_mc = Transpose ? cols : rows; - this->m_nc = Transpose ? rows : cols; - this->m_kc = depth; - - if(l3_blocking) - { - computeProductBlockingSizes(this->m_kc, this->m_mc, this->m_nc, num_threads); - } - else // no l3 blocking - { - Index n = this->m_nc; - computeProductBlockingSizes(this->m_kc, this->m_mc, n, num_threads); - } - - m_sizeA = this->m_mc * this->m_kc; - m_sizeB = this->m_kc * this->m_nc; - } - - void initParallel(Index rows, Index cols, Index depth, Index num_threads) - { - this->m_mc = Transpose ? cols : rows; - this->m_nc = Transpose ? rows : cols; - this->m_kc = depth; - - eigen_internal_assert(this->m_blockA==0 && this->m_blockB==0); - Index m = this->m_mc; - computeProductBlockingSizes(this->m_kc, m, this->m_nc, num_threads); - m_sizeA = this->m_mc * this->m_kc; - m_sizeB = this->m_kc * this->m_nc; - } - - void allocateA() - { - if(this->m_blockA==0) - this->m_blockA = aligned_new(m_sizeA); - } - - void allocateB() - { - if(this->m_blockB==0) - this->m_blockB = aligned_new(m_sizeB); - } - - void allocateAll() - { - allocateA(); - allocateB(); - } - - ~gemm_blocking_space() - { - aligned_delete(this->m_blockA, m_sizeA); - aligned_delete(this->m_blockB, m_sizeB); - } -}; - -} // end namespace internal - -namespace internal { - -template -struct generic_product_impl - : generic_product_impl_base > -{ - typedef typename Product::Scalar Scalar; - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - enum { - MaxDepthAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(Lhs::MaxColsAtCompileTime,Rhs::MaxRowsAtCompileTime) - }; - - typedef generic_product_impl lazyproduct; - - template - static void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::assign_op()); - else - { - dst.setZero(); - scaleAndAddTo(dst, lhs, rhs, Scalar(1)); - } - } - - template - static void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::add_assign_op()); - else - scaleAndAddTo(dst,lhs, rhs, Scalar(1)); - } - - template - static void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::sub_assign_op()); - else - scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); - } - - template - static void scaleAndAddTo(Dest& dst, const Lhs& a_lhs, const Rhs& a_rhs, const Scalar& alpha) - { - eigen_assert(dst.rows()==a_lhs.rows() && dst.cols()==a_rhs.cols()); - if(a_lhs.cols()==0 || a_lhs.rows()==0 || a_rhs.cols()==0) - return; - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,LhsScalar,RhsScalar, - Dest::MaxRowsAtCompileTime,Dest::MaxColsAtCompileTime,MaxDepthAtCompileTime> BlockingType; - - typedef internal::gemm_functor< - Scalar, Index, - internal::general_matrix_matrix_product< - Index, - LhsScalar, (ActualLhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(LhsBlasTraits::NeedToConjugate), - RhsScalar, (ActualRhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(RhsBlasTraits::NeedToConjugate), - (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime>, - ActualLhsTypeCleaned, ActualRhsTypeCleaned, Dest, BlockingType> GemmFunctor; - - BlockingType blocking(dst.rows(), dst.cols(), lhs.cols(), 1, true); - internal::parallelize_gemm<(Dest::MaxRowsAtCompileTime>32 || Dest::MaxRowsAtCompileTime==Dynamic)> - (GemmFunctor(lhs, rhs, dst, actualAlpha, blocking), a_lhs.rows(), a_rhs.cols(), a_lhs.cols(), Dest::Flags&RowMajorBit); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h deleted file mode 100644 index d68d2f96572..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h +++ /dev/null @@ -1,317 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H -#define EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H - -namespace Eigen { - -template -struct selfadjoint_rank1_update; - -namespace internal { - -/********************************************************************** -* This file implements a general A * B product while -* evaluating only one triangular part of the product. -* This is a more general version of self adjoint product (C += A A^T) -* as the level 3 SYRK Blas routine. -**********************************************************************/ - -// forward declarations (defined at the end of this file) -template -struct tribb_kernel; - -/* Optimized matrix-matrix product evaluating only one triangular half */ -template -struct general_matrix_matrix_triangular_product; - -// as usual if the result is row major => we transpose the product -template -struct general_matrix_matrix_triangular_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resIncr, Index resStride, - const ResScalar& alpha, level3_blocking& blocking) - { - general_matrix_matrix_triangular_product - ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking); - } -}; - -template -struct general_matrix_matrix_triangular_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, - const ResScalar& alpha, level3_blocking& blocking) - { - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); - Index mc = (std::min)(size,blocking.mc()); - - // !!! mc must be a multiple of nr: - if(mc > Traits::nr) - mc = (mc/Traits::nr)*Traits::nr; - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*size; - - ei_declare_aligned_stack_constructed_variable(LhsScalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(RhsScalar, blockB, sizeB, blocking.blockB()); - - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gebp_kernel gebp; - tribb_kernel sybb; - - for(Index k2=0; k2 processed with gebp or skipped - // 2 - the actual_mc x actual_mc symmetric block => processed with a special kernel - // 3 - after the diagonal => processed with gebp or skipped - if (UpLo==Lower) - gebp(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, - (std::min)(size,i2), alpha, -1, -1, 0, 0); - - sybb(_res+resStride*i2 + resIncr*i2, resIncr, resStride, blockA, blockB + actual_kc*i2, actual_mc, actual_kc, alpha); - - if (UpLo==Upper) - { - Index j2 = i2+actual_mc; - gebp(res.getSubMapper(i2, j2), blockA, blockB+actual_kc*j2, actual_mc, - actual_kc, (std::max)(Index(0), size-j2), alpha, -1, -1, 0, 0); - } - } - } - } -}; - -// Optimized packed Block * packed Block product kernel evaluating only one given triangular part -// This kernel is built on top of the gebp kernel: -// - the current destination block is processed per panel of actual_mc x BlockSize -// where BlockSize is set to the minimal value allowing gebp to be as fast as possible -// - then, as usual, each panel is split into three parts along the diagonal, -// the sub blocks above and below the diagonal are processed as usual, -// while the triangular block overlapping the diagonal is evaluated into a -// small temporary buffer which is then accumulated into the result using a -// triangular traversal. -template -struct tribb_kernel -{ - typedef gebp_traits Traits; - typedef typename Traits::ResScalar ResScalar; - - enum { - BlockSize = meta_least_common_multiple::ret - }; - void operator()(ResScalar* _res, Index resIncr, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) - { - typedef blas_data_mapper ResMapper; - typedef blas_data_mapper BufferMapper; - ResMapper res(_res, resStride, resIncr); - gebp_kernel gebp_kernel1; - gebp_kernel gebp_kernel2; - - Matrix buffer((internal::constructor_without_unaligned_array_assert())); - - // let's process the block per panel of actual_mc x BlockSize, - // again, each is split into three parts, etc. - for (Index j=0; j(BlockSize,size - j); - const RhsScalar* actual_b = blockB+j*depth; - - if(UpLo==Upper) - gebp_kernel1(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - - // selfadjoint micro block - { - Index i = j; - buffer.setZero(); - // 1 - apply the kernel on the temporary buffer - gebp_kernel2(BufferMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - - // 2 - triangular accumulation - for(Index j1=0; j1 -struct general_product_to_triangular_selector; - - -template -struct general_product_to_triangular_selector -{ - static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha, bool beta) - { - typedef typename MatrixType::Scalar Scalar; - - typedef typename internal::remove_all::type Lhs; - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs; - typedef typename internal::remove_all::type _ActualLhs; - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(prod.lhs()); - - typedef typename internal::remove_all::type Rhs; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs; - typedef typename internal::remove_all::type _ActualRhs; - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(prod.rhs()); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived()); - - if(!beta) - mat.template triangularView().setZero(); - - enum { - StorageOrder = (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, - UseLhsDirectly = _ActualLhs::InnerStrideAtCompileTime==1, - UseRhsDirectly = _ActualRhs::InnerStrideAtCompileTime==1 - }; - - internal::gemv_static_vector_if static_lhs; - ei_declare_aligned_stack_constructed_variable(Scalar, actualLhsPtr, actualLhs.size(), - (UseLhsDirectly ? const_cast(actualLhs.data()) : static_lhs.data())); - if(!UseLhsDirectly) Map(actualLhsPtr, actualLhs.size()) = actualLhs; - - internal::gemv_static_vector_if static_rhs; - ei_declare_aligned_stack_constructed_variable(Scalar, actualRhsPtr, actualRhs.size(), - (UseRhsDirectly ? const_cast(actualRhs.data()) : static_rhs.data())); - if(!UseRhsDirectly) Map(actualRhsPtr, actualRhs.size()) = actualRhs; - - - selfadjoint_rank1_update::IsComplex, - RhsBlasTraits::NeedToConjugate && NumTraits::IsComplex> - ::run(actualLhs.size(), mat.data(), mat.outerStride(), actualLhsPtr, actualRhsPtr, actualAlpha); - } -}; - -template -struct general_product_to_triangular_selector -{ - static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha, bool beta) - { - typedef typename internal::remove_all::type Lhs; - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs; - typedef typename internal::remove_all::type _ActualLhs; - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(prod.lhs()); - - typedef typename internal::remove_all::type Rhs; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs; - typedef typename internal::remove_all::type _ActualRhs; - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(prod.rhs()); - - typename ProductType::Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived()); - - if(!beta) - mat.template triangularView().setZero(); - - enum { - IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0, - LhsIsRowMajor = _ActualLhs::Flags&RowMajorBit ? 1 : 0, - RhsIsRowMajor = _ActualRhs::Flags&RowMajorBit ? 1 : 0, - SkipDiag = (UpLo&(UnitDiag|ZeroDiag))!=0 - }; - - Index size = mat.cols(); - if(SkipDiag) - size--; - Index depth = actualLhs.cols(); - - typedef internal::gemm_blocking_space BlockingType; - - BlockingType blocking(size, size, depth, 1, false); - - internal::general_matrix_matrix_triangular_product - ::run(size, depth, - &actualLhs.coeffRef(SkipDiag&&(UpLo&Lower)==Lower ? 1 : 0,0), actualLhs.outerStride(), - &actualRhs.coeffRef(0,SkipDiag&&(UpLo&Upper)==Upper ? 1 : 0), actualRhs.outerStride(), - mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? mat.innerStride() : mat.outerStride() ) : 0), - mat.innerStride(), mat.outerStride(), actualAlpha, blocking); - } -}; - -template -template -TriangularView& TriangularViewImpl::_assignProduct(const ProductType& prod, const Scalar& alpha, bool beta) -{ - EIGEN_STATIC_ASSERT((UpLo&UnitDiag)==0, WRITING_TO_TRIANGULAR_PART_WITH_UNIT_DIAGONAL_IS_NOT_SUPPORTED); - eigen_assert(derived().nestedExpression().rows() == prod.rows() && derived().cols() == prod.cols()); - - general_product_to_triangular_selector::InnerSize==1>::run(derived().nestedExpression().const_cast_derived(), prod, alpha, beta); - - return derived(); -} - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h deleted file mode 100644 index 691f95d6979..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h +++ /dev/null @@ -1,145 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Level 3 BLAS SYRK/HERK implementation. - ******************************************************************************** -*/ - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H -#define EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H - -namespace Eigen { - -namespace internal { - -template -struct general_matrix_matrix_rankupdate : - general_matrix_matrix_triangular_product< - Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,1,UpLo,BuiltIn> {}; - - -// try to go to BLAS specialization -#define EIGEN_BLAS_RANKUPDATE_SPECIALIZE(Scalar) \ -template \ -struct general_matrix_matrix_triangular_product { \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const Scalar* lhs, Index lhsStride, \ - const Scalar* rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) \ - { \ - if ( lhs==rhs && ((UpLo&(Lower|Upper))==UpLo) ) { \ - general_matrix_matrix_rankupdate \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resStride,alpha,blocking); \ - } else { \ - general_matrix_matrix_triangular_product \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resIncr,resStride,alpha,blocking); \ - } \ - } \ -}; - -EIGEN_BLAS_RANKUPDATE_SPECIALIZE(double) -EIGEN_BLAS_RANKUPDATE_SPECIALIZE(float) -// TODO handle complex cases -// EIGEN_BLAS_RANKUPDATE_SPECIALIZE(dcomplex) -// EIGEN_BLAS_RANKUPDATE_SPECIALIZE(scomplex) - -// SYRK for float/double -#define EIGEN_BLAS_RANKUPDATE_R(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct general_matrix_matrix_rankupdate { \ - enum { \ - IsLower = (UpLo&Lower) == Lower, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((AStorageOrder==ColMajor) && ConjugateA) ? 1 : 0 \ - }; \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* /*rhs*/, Index /*rhsStride*/, EIGTYPE* res, Index resStride, EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - /* typedef Matrix MatrixRhs;*/ \ -\ - BlasIndex lda=convert_index(lhsStride), ldc=convert_index(resStride), n=convert_index(size), k=convert_index(depth); \ - char uplo=((IsLower) ? 'L' : 'U'), trans=((AStorageOrder==RowMajor) ? 'T':'N'); \ - EIGTYPE beta(1); \ - BLASFUNC(&uplo, &trans, &n, &k, (const BLASTYPE*)&numext::real_ref(alpha), lhs, &lda, (const BLASTYPE*)&numext::real_ref(beta), res, &ldc); \ - } \ -}; - -// HERK for complex data -#define EIGEN_BLAS_RANKUPDATE_C(EIGTYPE, BLASTYPE, RTYPE, BLASFUNC) \ -template \ -struct general_matrix_matrix_rankupdate { \ - enum { \ - IsLower = (UpLo&Lower) == Lower, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = (((AStorageOrder==ColMajor) && ConjugateA) || ((AStorageOrder==RowMajor) && !ConjugateA)) ? 1 : 0 \ - }; \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* /*rhs*/, Index /*rhsStride*/, EIGTYPE* res, Index resStride, EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - typedef Matrix MatrixType; \ -\ - BlasIndex lda=convert_index(lhsStride), ldc=convert_index(resStride), n=convert_index(size), k=convert_index(depth); \ - char uplo=((IsLower) ? 'L' : 'U'), trans=((AStorageOrder==RowMajor) ? 'C':'N'); \ - RTYPE alpha_, beta_; \ - const EIGTYPE* a_ptr; \ -\ - alpha_ = alpha.real(); \ - beta_ = 1.0; \ -/* Copy with conjugation in some cases*/ \ - MatrixType a; \ - if (conjA) { \ - Map > mapA(lhs,n,k,OuterStride<>(lhsStride)); \ - a = mapA.conjugate(); \ - lda = a.outerStride(); \ - a_ptr = a.data(); \ - } else a_ptr=lhs; \ - BLASFUNC(&uplo, &trans, &n, &k, &alpha_, (BLASTYPE*)a_ptr, &lda, &beta_, (BLASTYPE*)res, &ldc); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_RANKUPDATE_R(double, double, dsyrk) -EIGEN_BLAS_RANKUPDATE_R(float, float, ssyrk) -#else -EIGEN_BLAS_RANKUPDATE_R(double, double, dsyrk_) -EIGEN_BLAS_RANKUPDATE_R(float, float, ssyrk_) -#endif - -// TODO hanlde complex cases -// EIGEN_BLAS_RANKUPDATE_C(dcomplex, double, double, zherk_) -// EIGEN_BLAS_RANKUPDATE_C(scomplex, float, float, cherk_) - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h deleted file mode 100644 index 71abf4013d1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h +++ /dev/null @@ -1,124 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * General matrix-matrix product functionality based on ?GEMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H -#define EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements general matrix-matrix multiplication using BLAS -* gemm function via partial specialization of -* general_matrix_matrix_product::run(..) method for float, double, -* std::complex and std::complex types -**********************************************************************/ - -// gemm specialization - -#define GEMM_SPECIALIZATION(EIGTYPE, EIGPREFIX, BLASTYPE, BLASFUNC) \ -template< \ - typename Index, \ - int LhsStorageOrder, bool ConjugateLhs, \ - int RhsStorageOrder, bool ConjugateRhs> \ -struct general_matrix_matrix_product \ -{ \ -typedef gebp_traits Traits; \ -\ -static void run(Index rows, Index cols, Index depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, \ - level3_blocking& /*blocking*/, \ - GemmParallelInfo* /*info = 0*/) \ -{ \ - using std::conj; \ -\ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char transa, transb; \ - BlasIndex m, n, k, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX a_tmp, b_tmp; \ -\ -/* Set transpose options */ \ - transa = (LhsStorageOrder==RowMajor) ? ((ConjugateLhs) ? 'C' : 'T') : 'N'; \ - transb = (RhsStorageOrder==RowMajor) ? ((ConjugateRhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ - k = convert_index(depth); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if ((LhsStorageOrder==ColMajor) && (ConjugateLhs)) { \ - Map > lhs(_lhs,m,k,OuterStride<>(lhsStride)); \ - a_tmp = lhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _lhs; \ -\ - if ((RhsStorageOrder==ColMajor) && (ConjugateRhs)) { \ - Map > rhs(_rhs,k,n,OuterStride<>(rhsStride)); \ - b_tmp = rhs.conjugate(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _rhs; \ -\ - BLASFUNC(&transa, &transb, &m, &n, &k, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -}}; - -#ifdef EIGEN_USE_MKL -GEMM_SPECIALIZATION(double, d, double, dgemm) -GEMM_SPECIALIZATION(float, f, float, sgemm) -GEMM_SPECIALIZATION(dcomplex, cd, MKL_Complex16, zgemm) -GEMM_SPECIALIZATION(scomplex, cf, MKL_Complex8, cgemm) -#else -GEMM_SPECIALIZATION(double, d, double, dgemm_) -GEMM_SPECIALIZATION(float, f, float, sgemm_) -GEMM_SPECIALIZATION(dcomplex, cd, double, zgemm_) -GEMM_SPECIALIZATION(scomplex, cf, float, cgemm_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h deleted file mode 100644 index a597c1f4ee6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h +++ /dev/null @@ -1,619 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_VECTOR_H -#define EIGEN_GENERAL_MATRIX_VECTOR_H - -namespace Eigen { - -namespace internal { - -/* Optimized col-major matrix * vector product: - * This algorithm processes 4 columns at onces that allows to both reduce - * the number of load/stores of the result by a factor 4 and to reduce - * the instruction dependency. Moreover, we know that all bands have the - * same alignment pattern. - * - * Mixing type logic: C += alpha * A * B - * | A | B |alpha| comments - * |real |cplx |cplx | no vectorization - * |real |cplx |real | alpha is converted to a cplx when calling the run function, no vectorization - * |cplx |real |cplx | invalid, the caller has to do tmp: = A * B; C += alpha*tmp - * |cplx |real |real | optimal case, vectorization possible via real-cplx mul - * - * Accesses to the matrix coefficients follow the following logic: - * - * - if all columns have the same alignment then - * - if the columns have the same alignment as the result vector, then easy! (-> AllAligned case) - * - otherwise perform unaligned loads only (-> NoneAligned case) - * - otherwise - * - if even columns have the same alignment then - * // odd columns are guaranteed to have the same alignment too - * - if even or odd columns have the same alignment as the result, then - * // for a register size of 2 scalars, this is guarantee to be the case (e.g., SSE with double) - * - perform half aligned and half unaligned loads (-> EvenAligned case) - * - otherwise perform unaligned loads only (-> NoneAligned case) - * - otherwise, if the register size is 4 scalars (e.g., SSE with float) then - * - one over 4 consecutive columns is guaranteed to be aligned with the result vector, - * perform simple aligned loads for this column and aligned loads plus re-alignment for the other. (-> FirstAligned case) - * // this re-alignment is done by the palign function implemented for SSE in Eigen/src/Core/arch/SSE/PacketMath.h - * - otherwise, - * // if we get here, this means the register size is greater than 4 (e.g., AVX with floats), - * // we currently fall back to the NoneAligned case - * - * The same reasoning apply for the transposed case. - * - * The last case (PacketSize>4) could probably be improved by generalizing the FirstAligned case, but since we do not support AVX yet... - * One might also wonder why in the EvenAligned case we perform unaligned loads instead of using the aligned-loads plus re-alignment - * strategy as in the FirstAligned case. The reason is that we observed that unaligned loads on a 8 byte boundary are not too slow - * compared to unaligned loads on a 4 byte boundary. - * - */ -template -struct general_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - -enum { - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable - && int(packet_traits::size)==int(packet_traits::size), - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1 -}; - -typedef typename packet_traits::type _LhsPacket; -typedef typename packet_traits::type _RhsPacket; -typedef typename packet_traits::type _ResPacket; - -typedef typename conditional::type LhsPacket; -typedef typename conditional::type RhsPacket; -typedef typename conditional::type ResPacket; - -EIGEN_DONT_INLINE static void run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - RhsScalar alpha); -}; - -template -EIGEN_DONT_INLINE void general_matrix_vector_product::run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - RhsScalar alpha) -{ - EIGEN_UNUSED_VARIABLE(resIncr); - eigen_internal_assert(resIncr==1); - #ifdef _EIGEN_ACCUMULATE_PACKETS - #error _EIGEN_ACCUMULATE_PACKETS has already been defined - #endif - #define _EIGEN_ACCUMULATE_PACKETS(Alignment0,Alignment13,Alignment2) \ - pstore(&res[j], \ - padd(pload(&res[j]), \ - padd( \ - padd(pcj.pmul(lhs0.template load(j), ptmp0), \ - pcj.pmul(lhs1.template load(j), ptmp1)), \ - padd(pcj.pmul(lhs2.template load(j), ptmp2), \ - pcj.pmul(lhs3.template load(j), ptmp3)) ))) - - typedef typename LhsMapper::VectorMapper LhsScalars; - - conj_helper cj; - conj_helper pcj; - if(ConjugateRhs) - alpha = numext::conj(alpha); - - enum { AllAligned = 0, EvenAligned, FirstAligned, NoneAligned }; - const Index columnsAtOnce = 4; - const Index peels = 2; - const Index LhsPacketAlignedMask = LhsPacketSize-1; - const Index ResPacketAlignedMask = ResPacketSize-1; -// const Index PeelAlignedMask = ResPacketSize*peels-1; - const Index size = rows; - - const Index lhsStride = lhs.stride(); - - // How many coeffs of the result do we have to skip to be aligned. - // Here we assume data are at least aligned on the base scalar type. - Index alignedStart = internal::first_default_aligned(res,size); - Index alignedSize = ResPacketSize>1 ? alignedStart + ((size-alignedStart) & ~ResPacketAlignedMask) : 0; - const Index peeledSize = alignedSize - RhsPacketSize*peels - RhsPacketSize + 1; - - const Index alignmentStep = LhsPacketSize>1 ? (LhsPacketSize - lhsStride % LhsPacketSize) & LhsPacketAlignedMask : 0; - Index alignmentPattern = alignmentStep==0 ? AllAligned - : alignmentStep==(LhsPacketSize/2) ? EvenAligned - : FirstAligned; - - // we cannot assume the first element is aligned because of sub-matrices - const Index lhsAlignmentOffset = lhs.firstAligned(size); - - // find how many columns do we have to skip to be aligned with the result (if possible) - Index skipColumns = 0; - // if the data cannot be aligned (TODO add some compile time tests when possible, e.g. for floats) - if( (lhsAlignmentOffset < 0) || (lhsAlignmentOffset == size) || (UIntPtr(res)%sizeof(ResScalar)) ) - { - alignedSize = 0; - alignedStart = 0; - alignmentPattern = NoneAligned; - } - else if(LhsPacketSize > 4) - { - // TODO: extend the code to support aligned loads whenever possible when LhsPacketSize > 4. - // Currently, it seems to be better to perform unaligned loads anyway - alignmentPattern = NoneAligned; - } - else if (LhsPacketSize>1) - { - // eigen_internal_assert(size_t(firstLhs+lhsAlignmentOffset)%sizeof(LhsPacket)==0 || size= cols) - || LhsPacketSize > size - || (size_t(firstLhs+alignedStart+lhsStride*skipColumns)%sizeof(LhsPacket))==0);*/ - } - else if(Vectorizable) - { - alignedStart = 0; - alignedSize = size; - alignmentPattern = AllAligned; - } - - const Index offset1 = (alignmentPattern==FirstAligned && alignmentStep==1)?3:1; - const Index offset3 = (alignmentPattern==FirstAligned && alignmentStep==1)?1:3; - - Index columnBound = ((cols-skipColumns)/columnsAtOnce)*columnsAtOnce + skipColumns; - for (Index i=skipColumns; i(alpha*rhs(i, 0)), - ptmp1 = pset1(alpha*rhs(i+offset1, 0)), - ptmp2 = pset1(alpha*rhs(i+2, 0)), - ptmp3 = pset1(alpha*rhs(i+offset3, 0)); - - // this helps a lot generating better binary code - const LhsScalars lhs0 = lhs.getVectorMapper(0, i+0), lhs1 = lhs.getVectorMapper(0, i+offset1), - lhs2 = lhs.getVectorMapper(0, i+2), lhs3 = lhs.getVectorMapper(0, i+offset3); - - if (Vectorizable) - { - /* explicit vectorization */ - // process initial unaligned coeffs - for (Index j=0; jalignedStart) - { - switch(alignmentPattern) - { - case AllAligned: - for (Index j = alignedStart; j1) - { - LhsPacket A00, A01, A02, A03, A10, A11, A12, A13; - ResPacket T0, T1; - - A01 = lhs1.template load(alignedStart-1); - A02 = lhs2.template load(alignedStart-2); - A03 = lhs3.template load(alignedStart-3); - - for (; j(j-1+LhsPacketSize); palign<1>(A01,A11); - A12 = lhs2.template load(j-2+LhsPacketSize); palign<2>(A02,A12); - A13 = lhs3.template load(j-3+LhsPacketSize); palign<3>(A03,A13); - - A00 = lhs0.template load(j); - A10 = lhs0.template load(j+LhsPacketSize); - T0 = pcj.pmadd(A00, ptmp0, pload(&res[j])); - T1 = pcj.pmadd(A10, ptmp0, pload(&res[j+ResPacketSize])); - - T0 = pcj.pmadd(A01, ptmp1, T0); - A01 = lhs1.template load(j-1+2*LhsPacketSize); palign<1>(A11,A01); - T0 = pcj.pmadd(A02, ptmp2, T0); - A02 = lhs2.template load(j-2+2*LhsPacketSize); palign<2>(A12,A02); - T0 = pcj.pmadd(A03, ptmp3, T0); - pstore(&res[j],T0); - A03 = lhs3.template load(j-3+2*LhsPacketSize); palign<3>(A13,A03); - T1 = pcj.pmadd(A11, ptmp1, T1); - T1 = pcj.pmadd(A12, ptmp2, T1); - T1 = pcj.pmadd(A13, ptmp3, T1); - pstore(&res[j+ResPacketSize],T1); - } - } - for (; j(alpha*rhs(k, 0)); - const LhsScalars lhs0 = lhs.getVectorMapper(0, k); - - if (Vectorizable) - { - /* explicit vectorization */ - // process first unaligned result's coeffs - for (Index j=0; j(alignedStart)) - for (Index i = alignedStart;i(i), ptmp0, pload(&res[i]))); - else - for (Index i = alignedStart;i(i), ptmp0, pload(&res[i]))); - } - - // process remaining scalars (or all if no explicit vectorization) - for (Index i=alignedSize; i -struct general_matrix_vector_product -{ -typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - -enum { - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable - && int(packet_traits::size)==int(packet_traits::size), - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1 -}; - -typedef typename packet_traits::type _LhsPacket; -typedef typename packet_traits::type _RhsPacket; -typedef typename packet_traits::type _ResPacket; - -typedef typename conditional::type LhsPacket; -typedef typename conditional::type RhsPacket; -typedef typename conditional::type ResPacket; - -EIGEN_DONT_INLINE static void run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - ResScalar alpha); -}; - -template -EIGEN_DONT_INLINE void general_matrix_vector_product::run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - ResScalar alpha) -{ - eigen_internal_assert(rhs.stride()==1); - - #ifdef _EIGEN_ACCUMULATE_PACKETS - #error _EIGEN_ACCUMULATE_PACKETS has already been defined - #endif - - #define _EIGEN_ACCUMULATE_PACKETS(Alignment0,Alignment13,Alignment2) {\ - RhsPacket b = rhs.getVectorMapper(j, 0).template load(0); \ - ptmp0 = pcj.pmadd(lhs0.template load(j), b, ptmp0); \ - ptmp1 = pcj.pmadd(lhs1.template load(j), b, ptmp1); \ - ptmp2 = pcj.pmadd(lhs2.template load(j), b, ptmp2); \ - ptmp3 = pcj.pmadd(lhs3.template load(j), b, ptmp3); } - - conj_helper cj; - conj_helper pcj; - - typedef typename LhsMapper::VectorMapper LhsScalars; - - enum { AllAligned=0, EvenAligned=1, FirstAligned=2, NoneAligned=3 }; - const Index rowsAtOnce = 4; - const Index peels = 2; - const Index RhsPacketAlignedMask = RhsPacketSize-1; - const Index LhsPacketAlignedMask = LhsPacketSize-1; - const Index depth = cols; - const Index lhsStride = lhs.stride(); - - // How many coeffs of the result do we have to skip to be aligned. - // Here we assume data are at least aligned on the base scalar type - // if that's not the case then vectorization is discarded, see below. - Index alignedStart = rhs.firstAligned(depth); - Index alignedSize = RhsPacketSize>1 ? alignedStart + ((depth-alignedStart) & ~RhsPacketAlignedMask) : 0; - const Index peeledSize = alignedSize - RhsPacketSize*peels - RhsPacketSize + 1; - - const Index alignmentStep = LhsPacketSize>1 ? (LhsPacketSize - lhsStride % LhsPacketSize) & LhsPacketAlignedMask : 0; - Index alignmentPattern = alignmentStep==0 ? AllAligned - : alignmentStep==(LhsPacketSize/2) ? EvenAligned - : FirstAligned; - - // we cannot assume the first element is aligned because of sub-matrices - const Index lhsAlignmentOffset = lhs.firstAligned(depth); - const Index rhsAlignmentOffset = rhs.firstAligned(rows); - - // find how many rows do we have to skip to be aligned with rhs (if possible) - Index skipRows = 0; - // if the data cannot be aligned (TODO add some compile time tests when possible, e.g. for floats) - if( (sizeof(LhsScalar)!=sizeof(RhsScalar)) || - (lhsAlignmentOffset < 0) || (lhsAlignmentOffset == depth) || - (rhsAlignmentOffset < 0) || (rhsAlignmentOffset == rows) ) - { - alignedSize = 0; - alignedStart = 0; - alignmentPattern = NoneAligned; - } - else if(LhsPacketSize > 4) - { - // TODO: extend the code to support aligned loads whenever possible when LhsPacketSize > 4. - alignmentPattern = NoneAligned; - } - else if (LhsPacketSize>1) - { - // eigen_internal_assert(size_t(firstLhs+lhsAlignmentOffset)%sizeof(LhsPacket)==0 || depth= rows) - || LhsPacketSize > depth - || (size_t(firstLhs+alignedStart+lhsStride*skipRows)%sizeof(LhsPacket))==0);*/ - } - else if(Vectorizable) - { - alignedStart = 0; - alignedSize = depth; - alignmentPattern = AllAligned; - } - - const Index offset1 = (alignmentPattern==FirstAligned && alignmentStep==1)?3:1; - const Index offset3 = (alignmentPattern==FirstAligned && alignmentStep==1)?1:3; - - Index rowBound = ((rows-skipRows)/rowsAtOnce)*rowsAtOnce + skipRows; - for (Index i=skipRows; i(ResScalar(0)), ptmp1 = pset1(ResScalar(0)), - ptmp2 = pset1(ResScalar(0)), ptmp3 = pset1(ResScalar(0)); - - // process initial unaligned coeffs - // FIXME this loop get vectorized by the compiler ! - for (Index j=0; jalignedStart) - { - switch(alignmentPattern) - { - case AllAligned: - for (Index j = alignedStart; j1) - { - /* Here we proccess 4 rows with with two peeled iterations to hide - * the overhead of unaligned loads. Moreover unaligned loads are handled - * using special shift/move operations between the two aligned packets - * overlaping the desired unaligned packet. This is *much* more efficient - * than basic unaligned loads. - */ - LhsPacket A01, A02, A03, A11, A12, A13; - A01 = lhs1.template load(alignedStart-1); - A02 = lhs2.template load(alignedStart-2); - A03 = lhs3.template load(alignedStart-3); - - for (; j(0); - A11 = lhs1.template load(j-1+LhsPacketSize); palign<1>(A01,A11); - A12 = lhs2.template load(j-2+LhsPacketSize); palign<2>(A02,A12); - A13 = lhs3.template load(j-3+LhsPacketSize); palign<3>(A03,A13); - - ptmp0 = pcj.pmadd(lhs0.template load(j), b, ptmp0); - ptmp1 = pcj.pmadd(A01, b, ptmp1); - A01 = lhs1.template load(j-1+2*LhsPacketSize); palign<1>(A11,A01); - ptmp2 = pcj.pmadd(A02, b, ptmp2); - A02 = lhs2.template load(j-2+2*LhsPacketSize); palign<2>(A12,A02); - ptmp3 = pcj.pmadd(A03, b, ptmp3); - A03 = lhs3.template load(j-3+2*LhsPacketSize); palign<3>(A13,A03); - - b = rhs.getVectorMapper(j+RhsPacketSize, 0).template load(0); - ptmp0 = pcj.pmadd(lhs0.template load(j+LhsPacketSize), b, ptmp0); - ptmp1 = pcj.pmadd(A11, b, ptmp1); - ptmp2 = pcj.pmadd(A12, b, ptmp2); - ptmp3 = pcj.pmadd(A13, b, ptmp3); - } - } - for (; j(tmp0); - const LhsScalars lhs0 = lhs.getVectorMapper(i, 0); - // process first unaligned result's coeffs - // FIXME this loop get vectorized by the compiler ! - for (Index j=0; jalignedStart) - { - // process aligned rhs coeffs - if (lhs0.template aligned(alignedStart)) - for (Index j = alignedStart;j(j), rhs.getVectorMapper(j, 0).template load(0), ptmp0); - else - for (Index j = alignedStart;j(j), rhs.getVectorMapper(j, 0).template load(0), ptmp0); - tmp0 += predux(ptmp0); - } - - // process remaining scalars - // FIXME this loop get vectorized by the compiler ! - for (Index j=alignedSize; j and std::complex types -**********************************************************************/ - -// gemv specialization - -template -struct general_matrix_vector_product_gemv; - -#define EIGEN_BLAS_GEMV_SPECIALIZE(Scalar) \ -template \ -struct general_matrix_vector_product,ColMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,Specialized> { \ -static void run( \ - Index rows, Index cols, \ - const const_blas_data_mapper &lhs, \ - const const_blas_data_mapper &rhs, \ - Scalar* res, Index resIncr, Scalar alpha) \ -{ \ - if (ConjugateLhs) { \ - general_matrix_vector_product,ColMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,BuiltIn>::run( \ - rows, cols, lhs, rhs, res, resIncr, alpha); \ - } else { \ - general_matrix_vector_product_gemv::run( \ - rows, cols, lhs.data(), lhs.stride(), rhs.data(), rhs.stride(), res, resIncr, alpha); \ - } \ -} \ -}; \ -template \ -struct general_matrix_vector_product,RowMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,Specialized> { \ -static void run( \ - Index rows, Index cols, \ - const const_blas_data_mapper &lhs, \ - const const_blas_data_mapper &rhs, \ - Scalar* res, Index resIncr, Scalar alpha) \ -{ \ - general_matrix_vector_product_gemv::run( \ - rows, cols, lhs.data(), lhs.stride(), rhs.data(), rhs.stride(), res, resIncr, alpha); \ -} \ -}; \ - -EIGEN_BLAS_GEMV_SPECIALIZE(double) -EIGEN_BLAS_GEMV_SPECIALIZE(float) -EIGEN_BLAS_GEMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_GEMV_SPECIALIZE(scomplex) - -#define EIGEN_BLAS_GEMV_SPECIALIZATION(EIGTYPE,BLASTYPE,BLASFUNC) \ -template \ -struct general_matrix_vector_product_gemv \ -{ \ -typedef Matrix GEMVVector;\ -\ -static void run( \ - Index rows, Index cols, \ - const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* rhs, Index rhsIncr, \ - EIGTYPE* res, Index resIncr, EIGTYPE alpha) \ -{ \ - BlasIndex m=convert_index(rows), n=convert_index(cols), \ - lda=convert_index(lhsStride), incx=convert_index(rhsIncr), incy=convert_index(resIncr); \ - const EIGTYPE beta(1); \ - const EIGTYPE *x_ptr; \ - char trans=(LhsStorageOrder==ColMajor) ? 'N' : (ConjugateLhs) ? 'C' : 'T'; \ - if (LhsStorageOrder==RowMajor) { \ - m = convert_index(cols); \ - n = convert_index(rows); \ - }\ - GEMVVector x_tmp; \ - if (ConjugateRhs) { \ - Map > map_x(rhs,cols,1,InnerStride<>(incx)); \ - x_tmp=map_x.conjugate(); \ - x_ptr=x_tmp.data(); \ - incx=1; \ - } else x_ptr=rhs; \ - BLASFUNC(&trans, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)lhs, &lda, (const BLASTYPE*)x_ptr, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &incy); \ -}\ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_GEMV_SPECIALIZATION(double, double, dgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(float, float, sgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(dcomplex, MKL_Complex16, zgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(scomplex, MKL_Complex8 , cgemv) -#else -EIGEN_BLAS_GEMV_SPECIALIZATION(double, double, dgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(float, float, sgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(dcomplex, double, zgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(scomplex, float, cgemv_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h deleted file mode 100644 index a3cc05b77bb..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h +++ /dev/null @@ -1,166 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARALLELIZER_H -#define EIGEN_PARALLELIZER_H - -namespace Eigen { - -namespace internal { - -/** \internal */ -inline void manage_multi_threading(Action action, int* v) -{ - static int m_maxThreads = -1; - EIGEN_UNUSED_VARIABLE(m_maxThreads); - - if(action==SetAction) - { - eigen_internal_assert(v!=0); - m_maxThreads = *v; - } - else if(action==GetAction) - { - eigen_internal_assert(v!=0); - #ifdef EIGEN_HAS_OPENMP - if(m_maxThreads>0) - *v = m_maxThreads; - else - *v = omp_get_max_threads(); - #else - *v = 1; - #endif - } - else - { - eigen_internal_assert(false); - } -} - -} - -/** Must be call first when calling Eigen from multiple threads */ -inline void initParallel() -{ - int nbt; - internal::manage_multi_threading(GetAction, &nbt); - std::ptrdiff_t l1, l2, l3; - internal::manage_caching_sizes(GetAction, &l1, &l2, &l3); -} - -/** \returns the max number of threads reserved for Eigen - * \sa setNbThreads */ -inline int nbThreads() -{ - int ret; - internal::manage_multi_threading(GetAction, &ret); - return ret; -} - -/** Sets the max number of threads reserved for Eigen - * \sa nbThreads */ -inline void setNbThreads(int v) -{ - internal::manage_multi_threading(SetAction, &v); -} - -namespace internal { - -template struct GemmParallelInfo -{ - GemmParallelInfo() : sync(-1), users(0), lhs_start(0), lhs_length(0) {} - - Index volatile sync; - int volatile users; - - Index lhs_start; - Index lhs_length; -}; - -template -void parallelize_gemm(const Functor& func, Index rows, Index cols, Index depth, bool transpose) -{ - // TODO when EIGEN_USE_BLAS is defined, - // we should still enable OMP for other scalar types -#if !(defined (EIGEN_HAS_OPENMP)) || defined (EIGEN_USE_BLAS) - // FIXME the transpose variable is only needed to properly split - // the matrix product when multithreading is enabled. This is a temporary - // fix to support row-major destination matrices. This whole - // parallelizer mechanism has to be redisigned anyway. - EIGEN_UNUSED_VARIABLE(depth); - EIGEN_UNUSED_VARIABLE(transpose); - func(0,rows, 0,cols); -#else - - // Dynamically check whether we should enable or disable OpenMP. - // The conditions are: - // - the max number of threads we can create is greater than 1 - // - we are not already in a parallel code - // - the sizes are large enough - - // compute the maximal number of threads from the size of the product: - // This first heuristic takes into account that the product kernel is fully optimized when working with nr columns at once. - Index size = transpose ? rows : cols; - Index pb_max_threads = std::max(1,size / Functor::Traits::nr); - - // compute the maximal number of threads from the total amount of work: - double work = static_cast(rows) * static_cast(cols) * - static_cast(depth); - double kMinTaskSize = 50000; // FIXME improve this heuristic. - pb_max_threads = std::max(1, std::min(pb_max_threads, work / kMinTaskSize)); - - // compute the number of threads we are going to use - Index threads = std::min(nbThreads(), pb_max_threads); - - // if multi-threading is explicitely disabled, not useful, or if we already are in a parallel session, - // then abort multi-threading - // FIXME omp_get_num_threads()>1 only works for openmp, what if the user does not use openmp? - if((!Condition) || (threads==1) || (omp_get_num_threads()>1)) - return func(0,rows, 0,cols); - - Eigen::initParallel(); - func.initParallelSession(threads); - - if(transpose) - std::swap(rows,cols); - - ei_declare_aligned_stack_constructed_variable(GemmParallelInfo,info,threads,0); - - #pragma omp parallel num_threads(threads) - { - Index i = omp_get_thread_num(); - // Note that the actual number of threads might be lower than the number of request ones. - Index actual_threads = omp_get_num_threads(); - - Index blockCols = (cols / actual_threads) & ~Index(0x3); - Index blockRows = (rows / actual_threads); - blockRows = (blockRows/Functor::Traits::mr)*Functor::Traits::mr; - - Index r0 = i*blockRows; - Index actualBlockRows = (i+1==actual_threads) ? rows-r0 : blockRows; - - Index c0 = i*blockCols; - Index actualBlockCols = (i+1==actual_threads) ? cols-c0 : blockCols; - - info[i].lhs_start = r0; - info[i].lhs_length = actualBlockRows; - - if(transpose) - func(c0, actualBlockCols, 0, rows, info); - else - func(0, rows, c0, actualBlockCols, info); - } -#endif -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PARALLELIZER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h deleted file mode 100644 index 04c933480bd..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h +++ /dev/null @@ -1,527 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_MATRIX_MATRIX_H -#define EIGEN_SELFADJOINT_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -// pack a selfadjoint block diagonal for use with the gebp_kernel -template -struct symm_pack_lhs -{ - template inline - void pack(Scalar* blockA, const const_blas_data_mapper& lhs, Index cols, Index i, Index& count) - { - // normal copy - for(Index k=0; k::size }; - const_blas_data_mapper lhs(_lhs,lhsStride); - Index count = 0; - //Index peeled_mc3 = (rows/Pack1)*Pack1; - - const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; - const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; - const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - - if(Pack1>=3*PacketSize) - for(Index i=0; i(blockA, lhs, cols, i, count); - - if(Pack1>=2*PacketSize) - for(Index i=peeled_mc3; i(blockA, lhs, cols, i, count); - - if(Pack1>=1*PacketSize) - for(Index i=peeled_mc2; i(blockA, lhs, cols, i, count); - - // do the same with mr==1 - for(Index i=peeled_mc1; i -struct symm_pack_rhs -{ - enum { PacketSize = packet_traits::size }; - void operator()(Scalar* blockB, const Scalar* _rhs, Index rhsStride, Index rows, Index cols, Index k2) - { - Index end_k = k2 + rows; - Index count = 0; - const_blas_data_mapper rhs(_rhs,rhsStride); - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - - // first part: normal case - for(Index j2=0; j2=4) - { - blockB[count+2] = rhs(k,j2+2); - blockB[count+3] = rhs(k,j2+3); - } - if (nr>=8) - { - blockB[count+4] = rhs(k,j2+4); - blockB[count+5] = rhs(k,j2+5); - blockB[count+6] = rhs(k,j2+6); - blockB[count+7] = rhs(k,j2+7); - } - count += nr; - } - } - - // second part: diagonal block - Index end8 = nr>=8 ? (std::min)(k2+rows,packet_cols8) : k2; - if(nr>=8) - { - for(Index j2=k2; j2=4) - { - for(Index j2=end8; j2<(std::min)(k2+rows,packet_cols4); j2+=4) - { - // again we can split vertically in three different parts (transpose, symmetric, normal) - // transpose - for(Index k=k2; k=8) - { - for(Index j2=k2+rows; j2=4) - { - for(Index j2=(std::max)(packet_cols8,k2+rows); j2 the same with nr==1) - for(Index j2=packet_cols4; j2 -struct product_selfadjoint_matrix; - -template -struct product_selfadjoint_matrix -{ - - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - product_selfadjoint_matrix::IsComplex && EIGEN_LOGICAL_XOR(RhsSelfAdjoint,ConjugateRhs), - EIGEN_LOGICAL_XOR(LhsSelfAdjoint,LhsStorageOrder==RowMajor) ? ColMajor : RowMajor, - LhsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsSelfAdjoint,ConjugateLhs), - ColMajor,ResInnerStride> - ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); - } -}; - -template -struct product_selfadjoint_matrix -{ - - static EIGEN_DONT_INLINE void run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - Index size = rows; - - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper LhsTransposeMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - LhsTransposeMapper lhs_transpose(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - // kc must be smaller than mc - kc = (std::min)(kc,mc); - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - gebp_kernel gebp_kernel; - symm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gemm_pack_lhs pack_lhs_transposed; - - for(Index k2=0; k2 transposed packed copy - // 2 - the diagonal block => special packed copy - // 3 - the panel below the diagonal block => generic packed copy - for(Index i2=0; i2() - (blockA, lhs.getSubMapper(i2, k2), actual_kc, actual_mc); - - gebp_kernel(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, cols, alpha); - } - } - } - -// matrix * selfadjoint product -template -struct product_selfadjoint_matrix -{ - - static EIGEN_DONT_INLINE void run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - Index size = cols; - - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - ResMapper res(_res,resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - symm_pack_rhs pack_rhs; - - for(Index k2=0; k2 GEPP - for(Index i2=0; i2 -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - - enum { - LhsIsUpper = (LhsMode&(Upper|Lower))==Upper, - LhsIsSelfAdjoint = (LhsMode&SelfAdjoint)==SelfAdjoint, - RhsIsUpper = (RhsMode&(Upper|Lower))==Upper, - RhsIsSelfAdjoint = (RhsMode&SelfAdjoint)==SelfAdjoint - }; - - template - static void run(Dest &dst, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - eigen_assert(dst.rows()==a_lhs.rows() && dst.cols()==a_rhs.cols()); - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Lhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxColsAtCompileTime,1> BlockingType; - - BlockingType blocking(lhs.rows(), rhs.cols(), lhs.cols(), 1, false); - - internal::product_selfadjoint_matrix::Flags &RowMajorBit) ? RowMajor : ColMajor, LhsIsSelfAdjoint, - NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsIsUpper,bool(LhsBlasTraits::NeedToConjugate)), - EIGEN_LOGICAL_XOR(RhsIsUpper,internal::traits::Flags &RowMajorBit) ? RowMajor : ColMajor, RhsIsSelfAdjoint, - NumTraits::IsComplex && EIGEN_LOGICAL_XOR(RhsIsUpper,bool(RhsBlasTraits::NeedToConjugate)), - internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime> - ::run( - lhs.rows(), rhs.cols(), // sizes - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info - actualAlpha, blocking // alpha - ); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h deleted file mode 100644 index 61396dbdf6f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h +++ /dev/null @@ -1,295 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -// - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Self adjoint matrix * matrix product functionality based on ?SYMM/?HEMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H -#define EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - - -/* Optimized selfadjoint matrix * matrix (?SYMM/?HEMM) product */ - -#define EIGEN_BLAS_SYMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ -\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='L', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ -\ -/* Set transpose options */ \ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (LhsStorageOrder==RowMajor) uplo='U'; \ - a = _lhs; \ -\ - if (RhsStorageOrder==RowMajor) { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.adjoint(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _rhs; \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - - -#define EIGEN_BLAS_HEMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='L', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ - Matrix a_tmp; \ -\ -/* Set transpose options */ \ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (((LhsStorageOrder==ColMajor) && ConjugateLhs) || ((LhsStorageOrder==RowMajor) && (!ConjugateLhs))) { \ - Map, 0, OuterStride<> > lhs(_lhs,m,m,OuterStride<>(lhsStride)); \ - a_tmp = lhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _lhs; \ - if (LhsStorageOrder==RowMajor) uplo='U'; \ -\ - if (RhsStorageOrder==ColMajor && (!ConjugateRhs)) { \ - b = _rhs; } \ - else { \ - if (RhsStorageOrder==ColMajor && ConjugateRhs) { \ - Map > rhs(_rhs,m,n,OuterStride<>(rhsStride)); \ - b_tmp = rhs.conjugate(); \ - } else \ - if (ConjugateRhs) { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.adjoint(); \ - } else { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.transpose(); \ - } \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMM_L(double, double, d, dsymm) -EIGEN_BLAS_SYMM_L(float, float, f, ssymm) -EIGEN_BLAS_HEMM_L(dcomplex, MKL_Complex16, cd, zhemm) -EIGEN_BLAS_HEMM_L(scomplex, MKL_Complex8, cf, chemm) -#else -EIGEN_BLAS_SYMM_L(double, double, d, dsymm_) -EIGEN_BLAS_SYMM_L(float, float, f, ssymm_) -EIGEN_BLAS_HEMM_L(dcomplex, double, cd, zhemm_) -EIGEN_BLAS_HEMM_L(scomplex, float, cf, chemm_) -#endif - -/* Optimized matrix * selfadjoint matrix (?SYMM/?HEMM) product */ - -#define EIGEN_BLAS_SYMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ -\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='R', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(rhsStride); \ - ldb = convert_index(lhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (RhsStorageOrder==RowMajor) uplo='U'; \ - a = _rhs; \ -\ - if (LhsStorageOrder==RowMajor) { \ - Map > lhs(_lhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = lhs.adjoint(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _lhs; \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - - -#define EIGEN_BLAS_HEMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='R', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ - Matrix a_tmp; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(rhsStride); \ - ldb = convert_index(lhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (((RhsStorageOrder==ColMajor) && ConjugateRhs) || ((RhsStorageOrder==RowMajor) && (!ConjugateRhs))) { \ - Map, 0, OuterStride<> > rhs(_rhs,n,n,OuterStride<>(rhsStride)); \ - a_tmp = rhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _rhs; \ - if (RhsStorageOrder==RowMajor) uplo='U'; \ -\ - if (LhsStorageOrder==ColMajor && (!ConjugateLhs)) { \ - b = _lhs; } \ - else { \ - if (LhsStorageOrder==ColMajor && ConjugateLhs) { \ - Map > lhs(_lhs,m,n,OuterStride<>(lhsStride)); \ - b_tmp = lhs.conjugate(); \ - } else \ - if (ConjugateLhs) { \ - Map > lhs(_lhs,n,m,OuterStride<>(lhsStride)); \ - b_tmp = lhs.adjoint(); \ - } else { \ - Map > lhs(_lhs,n,m,OuterStride<>(lhsStride)); \ - b_tmp = lhs.transpose(); \ - } \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMM_R(double, double, d, dsymm) -EIGEN_BLAS_SYMM_R(float, float, f, ssymm) -EIGEN_BLAS_HEMM_R(dcomplex, MKL_Complex16, cd, zhemm) -EIGEN_BLAS_HEMM_R(scomplex, MKL_Complex8, cf, chemm) -#else -EIGEN_BLAS_SYMM_R(double, double, d, dsymm_) -EIGEN_BLAS_SYMM_R(float, float, f, ssymm_) -EIGEN_BLAS_HEMM_R(dcomplex, double, cd, zhemm_) -EIGEN_BLAS_HEMM_R(scomplex, float, cf, chemm_) -#endif -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h deleted file mode 100644 index 3fd180e6c0b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h +++ /dev/null @@ -1,260 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_MATRIX_VECTOR_H -#define EIGEN_SELFADJOINT_MATRIX_VECTOR_H - -namespace Eigen { - -namespace internal { - -/* Optimized selfadjoint matrix * vector product: - * This algorithm processes 2 columns at onces that allows to both reduce - * the number of load/stores of the result by a factor 2 and to reduce - * the instruction dependency. - */ - -template -struct selfadjoint_matrix_vector_product; - -template -struct selfadjoint_matrix_vector_product - -{ -static EIGEN_DONT_INLINE void run( - Index size, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, - Scalar* res, - Scalar alpha); -}; - -template -EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product::run( - Index size, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, - Scalar* res, - Scalar alpha) -{ - typedef typename packet_traits::type Packet; - typedef typename NumTraits::Real RealScalar; - const Index PacketSize = sizeof(Packet)/sizeof(Scalar); - - enum { - IsRowMajor = StorageOrder==RowMajor ? 1 : 0, - IsLower = UpLo == Lower ? 1 : 0, - FirstTriangular = IsRowMajor == IsLower - }; - - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, IsRowMajor), ConjugateRhs> cj0; - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, !IsRowMajor), ConjugateRhs> cj1; - conj_helper cjd; - - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, IsRowMajor), ConjugateRhs> pcj0; - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, !IsRowMajor), ConjugateRhs> pcj1; - - Scalar cjAlpha = ConjugateRhs ? numext::conj(alpha) : alpha; - - - Index bound = (std::max)(Index(0),size-8) & 0xfffffffe; - if (FirstTriangular) - bound = size - bound; - - for (Index j=FirstTriangular ? bound : 0; - j<(FirstTriangular ? size : bound);j+=2) - { - const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride; - const Scalar* EIGEN_RESTRICT A1 = lhs + (j+1)*lhsStride; - - Scalar t0 = cjAlpha * rhs[j]; - Packet ptmp0 = pset1(t0); - Scalar t1 = cjAlpha * rhs[j+1]; - Packet ptmp1 = pset1(t1); - - Scalar t2(0); - Packet ptmp2 = pset1(t2); - Scalar t3(0); - Packet ptmp3 = pset1(t3); - - Index starti = FirstTriangular ? 0 : j+2; - Index endi = FirstTriangular ? j : size; - Index alignedStart = (starti) + internal::first_default_aligned(&res[starti], endi-starti); - Index alignedEnd = alignedStart + ((endi-alignedStart)/(PacketSize))*(PacketSize); - - res[j] += cjd.pmul(numext::real(A0[j]), t0); - res[j+1] += cjd.pmul(numext::real(A1[j+1]), t1); - if(FirstTriangular) - { - res[j] += cj0.pmul(A1[j], t1); - t3 += cj1.pmul(A1[j], rhs[j]); - } - else - { - res[j+1] += cj0.pmul(A0[j+1],t0); - t2 += cj1.pmul(A0[j+1], rhs[j+1]); - } - - for (Index i=starti; i huge speed up) - // gcc 4.2 does this optimization automatically. - const Scalar* EIGEN_RESTRICT a0It = A0 + alignedStart; - const Scalar* EIGEN_RESTRICT a1It = A1 + alignedStart; - const Scalar* EIGEN_RESTRICT rhsIt = rhs + alignedStart; - Scalar* EIGEN_RESTRICT resIt = res + alignedStart; - for (Index i=alignedStart; i(a0It); a0It += PacketSize; - Packet A1i = ploadu(a1It); a1It += PacketSize; - Packet Bi = ploadu(rhsIt); rhsIt += PacketSize; // FIXME should be aligned in most cases - Packet Xi = pload (resIt); - - Xi = pcj0.pmadd(A0i,ptmp0, pcj0.pmadd(A1i,ptmp1,Xi)); - ptmp2 = pcj1.pmadd(A0i, Bi, ptmp2); - ptmp3 = pcj1.pmadd(A1i, Bi, ptmp3); - pstore(resIt,Xi); resIt += PacketSize; - } - for (Index i=alignedEnd; i -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - enum { LhsUpLo = LhsMode&(Upper|Lower) }; - - template - static void run(Dest& dest, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - typedef typename Dest::Scalar ResScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; - - eigen_assert(dest.rows()==a_lhs.rows() && dest.cols()==a_rhs.cols()); - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - enum { - EvalToDest = (Dest::InnerStrideAtCompileTime==1), - UseRhs = (ActualRhsTypeCleaned::InnerStrideAtCompileTime==1) - }; - - internal::gemv_static_vector_if static_dest; - internal::gemv_static_vector_if static_rhs; - - ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), - EvalToDest ? dest.data() : static_dest.data()); - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,rhs.size(), - UseRhs ? const_cast(rhs.data()) : static_rhs.data()); - - if(!EvalToDest) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = dest.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - MappedDest(actualDestPtr, dest.size()) = dest; - } - - if(!UseRhs) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = rhs.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - Map(actualRhsPtr, rhs.size()) = rhs; - } - - - internal::selfadjoint_matrix_vector_product::Flags&RowMajorBit) ? RowMajor : ColMajor, - int(LhsUpLo), bool(LhsBlasTraits::NeedToConjugate), bool(RhsBlasTraits::NeedToConjugate)>::run - ( - lhs.rows(), // size - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - actualRhsPtr, // rhs info - actualDestPtr, // result info - actualAlpha // scale factor - ); - - if(!EvalToDest) - dest = MappedDest(actualDestPtr, dest.size()); - } -}; - -template -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - enum { RhsUpLo = RhsMode&(Upper|Lower) }; - - template - static void run(Dest& dest, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - // let's simply transpose the product - Transpose destT(dest); - selfadjoint_product_impl, int(RhsUpLo)==Upper ? Lower : Upper, false, - Transpose, 0, true>::run(destT, a_rhs.transpose(), a_lhs.transpose(), alpha); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_VECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h deleted file mode 100644 index 1238345e3fa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h +++ /dev/null @@ -1,118 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Selfadjoint matrix-vector product functionality based on ?SYMV/HEMV. - ******************************************************************************** -*/ - -#ifndef EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H -#define EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements selfadjoint matrix-vector multiplication using BLAS -**********************************************************************/ - -// symv/hemv specialization - -template -struct selfadjoint_matrix_vector_product_symv : - selfadjoint_matrix_vector_product {}; - -#define EIGEN_BLAS_SYMV_SPECIALIZE(Scalar) \ -template \ -struct selfadjoint_matrix_vector_product { \ -static void run( \ - Index size, const Scalar* lhs, Index lhsStride, \ - const Scalar* _rhs, Scalar* res, Scalar alpha) { \ - enum {\ - IsColMajor = StorageOrder==ColMajor \ - }; \ - if (IsColMajor == ConjugateLhs) {\ - selfadjoint_matrix_vector_product::run( \ - size, lhs, lhsStride, _rhs, res, alpha); \ - } else {\ - selfadjoint_matrix_vector_product_symv::run( \ - size, lhs, lhsStride, _rhs, res, alpha); \ - }\ - } \ -}; \ - -EIGEN_BLAS_SYMV_SPECIALIZE(double) -EIGEN_BLAS_SYMV_SPECIALIZE(float) -EIGEN_BLAS_SYMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_SYMV_SPECIALIZE(scomplex) - -#define EIGEN_BLAS_SYMV_SPECIALIZATION(EIGTYPE,BLASTYPE,BLASFUNC) \ -template \ -struct selfadjoint_matrix_vector_product_symv \ -{ \ -typedef Matrix SYMVVector;\ -\ -static void run( \ -Index size, const EIGTYPE* lhs, Index lhsStride, \ -const EIGTYPE* _rhs, EIGTYPE* res, EIGTYPE alpha) \ -{ \ - enum {\ - IsRowMajor = StorageOrder==RowMajor ? 1 : 0, \ - IsLower = UpLo == Lower ? 1 : 0 \ - }; \ - BlasIndex n=convert_index(size), lda=convert_index(lhsStride), incx=1, incy=1; \ - EIGTYPE beta(1); \ - const EIGTYPE *x_ptr; \ - char uplo=(IsRowMajor) ? (IsLower ? 'U' : 'L') : (IsLower ? 'L' : 'U'); \ - SYMVVector x_tmp; \ - if (ConjugateRhs) { \ - Map map_x(_rhs,size,1); \ - x_tmp=map_x.conjugate(); \ - x_ptr=x_tmp.data(); \ - } else x_ptr=_rhs; \ - BLASFUNC(&uplo, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)lhs, &lda, (const BLASTYPE*)x_ptr, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &incy); \ -}\ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMV_SPECIALIZATION(double, double, dsymv) -EIGEN_BLAS_SYMV_SPECIALIZATION(float, float, ssymv) -EIGEN_BLAS_SYMV_SPECIALIZATION(dcomplex, MKL_Complex16, zhemv) -EIGEN_BLAS_SYMV_SPECIALIZATION(scomplex, MKL_Complex8, chemv) -#else -EIGEN_BLAS_SYMV_SPECIALIZATION(double, double, dsymv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(float, float, ssymv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(dcomplex, double, zhemv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(scomplex, float, chemv_) -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h deleted file mode 100644 index ef12c98f6cb..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h +++ /dev/null @@ -1,133 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_PRODUCT_H -#define EIGEN_SELFADJOINT_PRODUCT_H - -/********************************************************************** -* This file implements a self adjoint product: C += A A^T updating only -* half of the selfadjoint matrix C. -* It corresponds to the level 3 SYRK and level 2 SYR Blas routines. -**********************************************************************/ - -namespace Eigen { - - -template -struct selfadjoint_rank1_update -{ - static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha) - { - internal::conj_if cj; - typedef Map > OtherMap; - typedef typename internal::conditional::type ConjLhsType; - for (Index i=0; i >(mat+stride*i+(UpLo==Lower ? i : 0), (UpLo==Lower ? size-i : (i+1))) - += (alpha * cj(vecY[i])) * ConjLhsType(OtherMap(vecX+(UpLo==Lower ? i : 0),UpLo==Lower ? size-i : (i+1))); - } - } -}; - -template -struct selfadjoint_rank1_update -{ - static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha) - { - selfadjoint_rank1_update::run(size,mat,stride,vecY,vecX,alpha); - } -}; - -template -struct selfadjoint_product_selector; - -template -struct selfadjoint_product_selector -{ - static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha) - { - typedef typename MatrixType::Scalar Scalar; - typedef internal::blas_traits OtherBlasTraits; - typedef typename OtherBlasTraits::DirectLinearAccessType ActualOtherType; - typedef typename internal::remove_all::type _ActualOtherType; - typename internal::add_const_on_value_type::type actualOther = OtherBlasTraits::extract(other.derived()); - - Scalar actualAlpha = alpha * OtherBlasTraits::extractScalarFactor(other.derived()); - - enum { - StorageOrder = (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, - UseOtherDirectly = _ActualOtherType::InnerStrideAtCompileTime==1 - }; - internal::gemv_static_vector_if static_other; - - ei_declare_aligned_stack_constructed_variable(Scalar, actualOtherPtr, other.size(), - (UseOtherDirectly ? const_cast(actualOther.data()) : static_other.data())); - - if(!UseOtherDirectly) - Map(actualOtherPtr, actualOther.size()) = actualOther; - - selfadjoint_rank1_update::IsComplex, - (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex> - ::run(other.size(), mat.data(), mat.outerStride(), actualOtherPtr, actualOtherPtr, actualAlpha); - } -}; - -template -struct selfadjoint_product_selector -{ - static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha) - { - typedef typename MatrixType::Scalar Scalar; - typedef internal::blas_traits OtherBlasTraits; - typedef typename OtherBlasTraits::DirectLinearAccessType ActualOtherType; - typedef typename internal::remove_all::type _ActualOtherType; - typename internal::add_const_on_value_type::type actualOther = OtherBlasTraits::extract(other.derived()); - - Scalar actualAlpha = alpha * OtherBlasTraits::extractScalarFactor(other.derived()); - - enum { - IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0, - OtherIsRowMajor = _ActualOtherType::Flags&RowMajorBit ? 1 : 0 - }; - - Index size = mat.cols(); - Index depth = actualOther.cols(); - - typedef internal::gemm_blocking_space BlockingType; - - BlockingType blocking(size, size, depth, 1, false); - - - internal::general_matrix_matrix_triangular_product::IsComplex, - Scalar, OtherIsRowMajor ? ColMajor : RowMajor, (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex, - IsRowMajor ? RowMajor : ColMajor, MatrixType::InnerStrideAtCompileTime, UpLo> - ::run(size, depth, - &actualOther.coeffRef(0,0), actualOther.outerStride(), &actualOther.coeffRef(0,0), actualOther.outerStride(), - mat.data(), mat.innerStride(), mat.outerStride(), actualAlpha, blocking); - } -}; - -// high level API - -template -template -SelfAdjointView& SelfAdjointView -::rankUpdate(const MatrixBase& u, const Scalar& alpha) -{ - selfadjoint_product_selector::run(_expression().const_cast_derived(), u.derived(), alpha); - - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h deleted file mode 100644 index 2ae3641111c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h +++ /dev/null @@ -1,93 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINTRANK2UPTADE_H -#define EIGEN_SELFADJOINTRANK2UPTADE_H - -namespace Eigen { - -namespace internal { - -/* Optimized selfadjoint matrix += alpha * uv' + conj(alpha)*vu' - * It corresponds to the Level2 syr2 BLAS routine - */ - -template -struct selfadjoint_rank2_update_selector; - -template -struct selfadjoint_rank2_update_selector -{ - static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha) - { - const Index size = u.size(); - for (Index i=0; i >(mat+stride*i+i, size-i) += - (numext::conj(alpha) * numext::conj(u.coeff(i))) * v.tail(size-i) - + (alpha * numext::conj(v.coeff(i))) * u.tail(size-i); - } - } -}; - -template -struct selfadjoint_rank2_update_selector -{ - static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha) - { - const Index size = u.size(); - for (Index i=0; i >(mat+stride*i, i+1) += - (numext::conj(alpha) * numext::conj(u.coeff(i))) * v.head(i+1) - + (alpha * numext::conj(v.coeff(i))) * u.head(i+1); - } -}; - -template struct conj_expr_if - : conditional::Scalar>,T> > {}; - -} // end namespace internal - -template -template -SelfAdjointView& SelfAdjointView -::rankUpdate(const MatrixBase& u, const MatrixBase& v, const Scalar& alpha) -{ - typedef internal::blas_traits UBlasTraits; - typedef typename UBlasTraits::DirectLinearAccessType ActualUType; - typedef typename internal::remove_all::type _ActualUType; - typename internal::add_const_on_value_type::type actualU = UBlasTraits::extract(u.derived()); - - typedef internal::blas_traits VBlasTraits; - typedef typename VBlasTraits::DirectLinearAccessType ActualVType; - typedef typename internal::remove_all::type _ActualVType; - typename internal::add_const_on_value_type::type actualV = VBlasTraits::extract(v.derived()); - - // If MatrixType is row major, then we use the routine for lower triangular in the upper triangular case and - // vice versa, and take the complex conjugate of all coefficients and vector entries. - - enum { IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0 }; - Scalar actualAlpha = alpha * UBlasTraits::extractScalarFactor(u.derived()) - * numext::conj(VBlasTraits::extractScalarFactor(v.derived())); - if (IsRowMajor) - actualAlpha = numext::conj(actualAlpha); - - typedef typename internal::remove_all::type>::type UType; - typedef typename internal::remove_all::type>::type VType; - internal::selfadjoint_rank2_update_selector - ::run(_expression().const_cast_derived().data(),_expression().outerStride(),UType(actualU),VType(actualV),actualAlpha); - - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINTRANK2UPTADE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h deleted file mode 100644 index 2fb408d1d73..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h +++ /dev/null @@ -1,472 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_MATRIX_MATRIX_H -#define EIGEN_TRIANGULAR_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -// template -// struct gemm_pack_lhs_triangular -// { -// Matrix::IsComplex && Conjugate> cj; -// const_blas_data_mapper lhs(_lhs,lhsStride); -// int count = 0; -// const int peeled_mc = (rows/mr)*mr; -// for(int i=0; i -struct product_triangular_matrix_matrix; - -template -struct product_triangular_matrix_matrix -{ - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, Index depth, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - product_triangular_matrix_matrix - ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); - } -}; - -// implements col-major += alpha * op(triangular) * op(general) -template -struct product_triangular_matrix_matrix -{ - - typedef gebp_traits Traits; - enum { - SmallPanelWidth = 2 * EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower, - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1 - }; - - static EIGEN_DONT_INLINE void run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - // strip zeros - Index diagSize = (std::min)(_rows,_depth); - Index rows = IsLower ? _rows : diagSize; - Index depth = IsLower ? diagSize : _depth; - Index cols = _cols; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - // The small panel size must not be larger than blocking size. - // Usually this should never be the case because SmallPanelWidth^2 is very small - // compared to L2 cache size, but let's be safe: - Index panelWidth = (std::min)(Index(SmallPanelWidth),(std::min)(kc,mc)); - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - // To work around an "error: member reference base type 'Matrix<...> - // (Eigen::internal::constructor_without_unaligned_array_assert (*)())' is - // not a structure or union" compilation error in nvcc (tested V8.0.61), - // create a dummy internal::constructor_without_unaligned_array_assert - // object to pass to the Matrix constructor. - internal::constructor_without_unaligned_array_assert a; - Matrix triangularBuffer(a); - triangularBuffer.setZero(); - if((Mode&ZeroDiag)==ZeroDiag) - triangularBuffer.diagonal().setZero(); - else - triangularBuffer.diagonal().setOnes(); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - - for(Index k2=IsLower ? depth : 0; - IsLower ? k2>0 : k2rows)) - { - actual_kc = rows-k2; - k2 = k2+actual_kc-kc; - } - - pack_rhs(blockB, rhs.getSubMapper(actual_k2,0), actual_kc, cols); - - // the selected lhs's panel has to be split in three different parts: - // 1 - the part which is zero => skip it - // 2 - the diagonal block => special kernel - // 3 - the dense panel below (lower case) or above (upper case) the diagonal block => GEPP - - // the block diagonal, if any: - if(IsLower || actual_k2(actual_kc-k1, panelWidth); - Index lengthTarget = IsLower ? actual_kc-k1-actualPanelWidth : k1; - Index startBlock = actual_k2+k1; - Index blockBOffset = k1; - - // => GEBP with the micro triangular block - // The trick is to pack this micro block while filling the opposite triangular part with zeros. - // To this end we do an extra triangular copy to a small temporary buffer - for (Index k=0;k0) - { - Index startTarget = IsLower ? actual_k2+k1+actualPanelWidth : actual_k2; - - pack_lhs(blockA, lhs.getSubMapper(startTarget,startBlock), actualPanelWidth, lengthTarget); - - gebp_kernel(res.getSubMapper(startTarget, 0), blockA, blockB, - lengthTarget, actualPanelWidth, cols, alpha, - actualPanelWidth, actual_kc, 0, blockBOffset); - } - } - } - // the part below (lower case) or above (upper case) the diagonal => GEPP - { - Index start = IsLower ? k2 : 0; - Index end = IsLower ? rows : (std::min)(actual_k2,rows); - for(Index i2=start; i2() - (blockA, lhs.getSubMapper(i2, actual_k2), actual_kc, actual_mc); - - gebp_kernel(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, - actual_kc, cols, alpha, -1, -1, 0, 0); - } - } - } - } - -// implements col-major += alpha * op(general) * op(triangular) -template -struct product_triangular_matrix_matrix -{ - typedef gebp_traits Traits; - enum { - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower, - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1 - }; - - static EIGEN_DONT_INLINE void run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - const Index PacketBytes = packet_traits::size*sizeof(Scalar); - // strip zeros - Index diagSize = (std::min)(_cols,_depth); - Index rows = _rows; - Index depth = IsLower ? _depth : diagSize; - Index cols = IsLower ? diagSize : _cols; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols+EIGEN_MAX_ALIGN_BYTES/sizeof(Scalar); - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - internal::constructor_without_unaligned_array_assert a; - Matrix triangularBuffer(a); - triangularBuffer.setZero(); - if((Mode&ZeroDiag)==ZeroDiag) - triangularBuffer.diagonal().setZero(); - else - triangularBuffer.diagonal().setOnes(); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gemm_pack_rhs pack_rhs_panel; - - for(Index k2=IsLower ? 0 : depth; - IsLower ? k20; - IsLower ? k2+=kc : k2-=kc) - { - Index actual_kc = (std::min)(IsLower ? depth-k2 : k2, kc); - Index actual_k2 = IsLower ? k2 : k2-actual_kc; - - // align blocks with the end of the triangular part for trapezoidal rhs - if(IsLower && (k2cols)) - { - actual_kc = cols-k2; - k2 = actual_k2 + actual_kc - kc; - } - - // remaining size - Index rs = IsLower ? (std::min)(cols,actual_k2) : cols - k2; - // size of the triangular part - Index ts = (IsLower && actual_k2>=cols) ? 0 : actual_kc; - - Scalar* geb = blockB+ts*ts; - geb = geb + internal::first_aligned(geb,PacketBytes/sizeof(Scalar)); - - pack_rhs(geb, rhs.getSubMapper(actual_k2,IsLower ? 0 : k2), actual_kc, rs); - - // pack the triangular part of the rhs padding the unrolled blocks with zeros - if(ts>0) - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index actual_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc-j2-actualPanelWidth : j2; - // general part - pack_rhs_panel(blockB+j2*actual_kc, - rhs.getSubMapper(actual_k2+panelOffset, actual_j2), - panelLength, actualPanelWidth, - actual_kc, panelOffset); - - // append the triangular part via a temporary buffer - for (Index j=0;j0) - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index panelLength = IsLower ? actual_kc-j2 : j2+actualPanelWidth; - Index blockOffset = IsLower ? j2 : 0; - - gebp_kernel(res.getSubMapper(i2, actual_k2 + j2), - blockA, blockB+j2*actual_kc, - actual_mc, panelLength, actualPanelWidth, - alpha, - actual_kc, actual_kc, // strides - blockOffset, blockOffset);// offsets - } - } - gebp_kernel(res.getSubMapper(i2, IsLower ? 0 : k2), - blockA, geb, actual_mc, actual_kc, rs, - alpha, - -1, -1, 0, 0); - } - } - } - -/*************************************************************************** -* Wrapper to product_triangular_matrix_matrix -***************************************************************************/ - -} // end namespace internal - -namespace internal { -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &a_lhs, const Rhs &a_rhs, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(a_lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(a_rhs); - Scalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Lhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxColsAtCompileTime,4> BlockingType; - - enum { IsLower = (Mode&Lower) == Lower }; - Index stripedRows = ((!LhsIsTriangular) || (IsLower)) ? lhs.rows() : (std::min)(lhs.rows(),lhs.cols()); - Index stripedCols = ((LhsIsTriangular) || (!IsLower)) ? rhs.cols() : (std::min)(rhs.cols(),rhs.rows()); - Index stripedDepth = LhsIsTriangular ? ((!IsLower) ? lhs.cols() : (std::min)(lhs.cols(),lhs.rows())) - : ((IsLower) ? rhs.rows() : (std::min)(rhs.rows(),rhs.cols())); - - BlockingType blocking(stripedRows, stripedCols, stripedDepth, 1, false); - - internal::product_triangular_matrix_matrix::Flags&RowMajorBit) ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, Dest::InnerStrideAtCompileTime> - ::run( - stripedRows, stripedCols, stripedDepth, // sizes - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info - actualAlpha, blocking - ); - - // Apply correction if the diagonal is unit and a scalar factor was nested: - if ((Mode&UnitDiag)==UnitDiag) - { - if (LhsIsTriangular && lhs_alpha!=LhsScalar(1)) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dst.topRows(diagSize) -= ((lhs_alpha-LhsScalar(1))*a_rhs).topRows(diagSize); - } - else if ((!LhsIsTriangular) && rhs_alpha!=RhsScalar(1)) - { - Index diagSize = (std::min)(rhs.rows(),rhs.cols()); - dst.leftCols(diagSize) -= (rhs_alpha-RhsScalar(1))*a_lhs.leftCols(diagSize); - } - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h deleted file mode 100644 index a98d12e4ae1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h +++ /dev/null @@ -1,317 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix * matrix product functionality based on ?TRMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H -#define EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - - -template -struct product_triangular_matrix_matrix_trmm : - product_triangular_matrix_matrix {}; - - -// try to go to BLAS specialization -#define EIGEN_BLAS_TRMM_SPECIALIZE(Scalar, LhsIsTriangular) \ -template \ -struct product_triangular_matrix_matrix { \ - static inline void run(Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride,\ - const Scalar* _rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ - } \ -}; - -EIGEN_BLAS_TRMM_SPECIALIZE(double, true) -EIGEN_BLAS_TRMM_SPECIALIZE(double, false) -EIGEN_BLAS_TRMM_SPECIALIZE(dcomplex, true) -EIGEN_BLAS_TRMM_SPECIALIZE(dcomplex, false) -EIGEN_BLAS_TRMM_SPECIALIZE(float, true) -EIGEN_BLAS_TRMM_SPECIALIZE(float, false) -EIGEN_BLAS_TRMM_SPECIALIZE(scomplex, true) -EIGEN_BLAS_TRMM_SPECIALIZE(scomplex, false) - -// implements col-major += alpha * op(triangular) * op(general) -#define EIGEN_BLAS_TRMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_triangular_matrix_matrix_trmm \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((LhsStorageOrder==ColMajor) && ConjugateLhs) ? 1 : 0 \ - }; \ -\ - static void run( \ - Index _rows, Index _cols, Index _depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ - EIGTYPE alpha, level3_blocking& blocking) \ - { \ - Index diagSize = (std::min)(_rows,_depth); \ - Index rows = IsLower ? _rows : diagSize; \ - Index depth = IsLower ? diagSize : _depth; \ - Index cols = _cols; \ -\ - typedef Matrix MatrixLhs; \ - typedef Matrix MatrixRhs; \ -\ -/* Non-square case - doesn't fit to BLAS ?TRMM. Fall to default triangular product or call BLAS ?GEMM*/ \ - if (rows != depth) { \ -\ - /* FIXME handle mkl_domain_get_max_threads */ \ - /*int nthr = mkl_domain_get_max_threads(EIGEN_BLAS_DOMAIN_BLAS);*/ int nthr = 1;\ -\ - if (((nthr==1) && (((std::max)(rows,depth)-diagSize)/(double)diagSize < 0.5))) { \ - /* Most likely no benefit to call TRMM or GEMM from BLAS */ \ - product_triangular_matrix_matrix::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ - /*std::cout << "TRMM_L: A is not square! Go to Eigen TRMM implementation!\n";*/ \ - } else { \ - /* Make sense to call GEMM */ \ - Map > lhsMap(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixLhs aa_tmp=lhsMap.template triangularView(); \ - BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ - gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, 1, resStride, alpha, gemm_blocking, 0); \ -\ - /*std::cout << "TRMM_L: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ - } \ - return; \ - } \ - char side = 'L', transa, uplo, diag = 'N'; \ - EIGTYPE *b; \ - const EIGTYPE *a; \ - BlasIndex m, n, lda, ldb; \ -\ -/* Set m, n */ \ - m = convert_index(diagSize); \ - n = convert_index(cols); \ -\ -/* Set trans */ \ - transa = (LhsStorageOrder==RowMajor) ? ((ConjugateLhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set b, ldb */ \ - Map > rhs(_rhs,depth,cols,OuterStride<>(rhsStride)); \ - MatrixX##EIGPREFIX b_tmp; \ -\ - if (ConjugateRhs) b_tmp = rhs.conjugate(); else b_tmp = rhs; \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ -\ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (LhsStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - Map > lhs(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixLhs a_tmp; \ -\ - if ((conjA!=0) || (SetDiag==0)) { \ - if (conjA) a_tmp = lhs.conjugate(); else a_tmp = lhs; \ - if (IsZeroDiag) \ - a_tmp.diagonal().setZero(); \ - else if (IsUnitDiag) \ - a_tmp.diagonal().setOnes();\ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _lhs; \ - lda = convert_index(lhsStride); \ - } \ - /*std::cout << "TRMM_L: A is square! Go to BLAS TRMM implementation! \n";*/ \ -/* call ?trmm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)b, &ldb); \ -\ -/* Add op(a_triangular)*b into res*/ \ - Map > res_tmp(res,rows,cols,OuterStride<>(resStride)); \ - res_tmp=res_tmp+b_tmp; \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMM_L(double, double, d, dtrmm) -EIGEN_BLAS_TRMM_L(dcomplex, MKL_Complex16, cd, ztrmm) -EIGEN_BLAS_TRMM_L(float, float, f, strmm) -EIGEN_BLAS_TRMM_L(scomplex, MKL_Complex8, cf, ctrmm) -#else -EIGEN_BLAS_TRMM_L(double, double, d, dtrmm_) -EIGEN_BLAS_TRMM_L(dcomplex, double, cd, ztrmm_) -EIGEN_BLAS_TRMM_L(float, float, f, strmm_) -EIGEN_BLAS_TRMM_L(scomplex, float, cf, ctrmm_) -#endif - -// implements col-major += alpha * op(general) * op(triangular) -#define EIGEN_BLAS_TRMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_triangular_matrix_matrix_trmm \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((RhsStorageOrder==ColMajor) && ConjugateRhs) ? 1 : 0 \ - }; \ -\ - static void run( \ - Index _rows, Index _cols, Index _depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ - EIGTYPE alpha, level3_blocking& blocking) \ - { \ - Index diagSize = (std::min)(_cols,_depth); \ - Index rows = _rows; \ - Index depth = IsLower ? _depth : diagSize; \ - Index cols = IsLower ? diagSize : _cols; \ -\ - typedef Matrix MatrixLhs; \ - typedef Matrix MatrixRhs; \ -\ -/* Non-square case - doesn't fit to BLAS ?TRMM. Fall to default triangular product or call BLAS ?GEMM*/ \ - if (cols != depth) { \ -\ - int nthr = 1 /*mkl_domain_get_max_threads(EIGEN_BLAS_DOMAIN_BLAS)*/; \ -\ - if ((nthr==1) && (((std::max)(cols,depth)-diagSize)/(double)diagSize < 0.5)) { \ - /* Most likely no benefit to call TRMM or GEMM from BLAS*/ \ - product_triangular_matrix_matrix::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ - /*std::cout << "TRMM_R: A is not square! Go to Eigen TRMM implementation!\n";*/ \ - } else { \ - /* Make sense to call GEMM */ \ - Map > rhsMap(_rhs,depth,cols, OuterStride<>(rhsStride)); \ - MatrixRhs aa_tmp=rhsMap.template triangularView(); \ - BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ - gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, 1, resStride, alpha, gemm_blocking, 0); \ -\ - /*std::cout << "TRMM_R: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ - } \ - return; \ - } \ - char side = 'R', transa, uplo, diag = 'N'; \ - EIGTYPE *b; \ - const EIGTYPE *a; \ - BlasIndex m, n, lda, ldb; \ -\ -/* Set m, n */ \ - m = convert_index(rows); \ - n = convert_index(diagSize); \ -\ -/* Set trans */ \ - transa = (RhsStorageOrder==RowMajor) ? ((ConjugateRhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set b, ldb */ \ - Map > lhs(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixX##EIGPREFIX b_tmp; \ -\ - if (ConjugateLhs) b_tmp = lhs.conjugate(); else b_tmp = lhs; \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ -\ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (RhsStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - Map > rhs(_rhs,depth,cols, OuterStride<>(rhsStride)); \ - MatrixRhs a_tmp; \ -\ - if ((conjA!=0) || (SetDiag==0)) { \ - if (conjA) a_tmp = rhs.conjugate(); else a_tmp = rhs; \ - if (IsZeroDiag) \ - a_tmp.diagonal().setZero(); \ - else if (IsUnitDiag) \ - a_tmp.diagonal().setOnes();\ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _rhs; \ - lda = convert_index(rhsStride); \ - } \ - /*std::cout << "TRMM_R: A is square! Go to BLAS TRMM implementation! \n";*/ \ -/* call ?trmm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)b, &ldb); \ -\ -/* Add op(a_triangular)*b into res*/ \ - Map > res_tmp(res,rows,cols,OuterStride<>(resStride)); \ - res_tmp=res_tmp+b_tmp; \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMM_R(double, double, d, dtrmm) -EIGEN_BLAS_TRMM_R(dcomplex, MKL_Complex16, cd, ztrmm) -EIGEN_BLAS_TRMM_R(float, float, f, strmm) -EIGEN_BLAS_TRMM_R(scomplex, MKL_Complex8, cf, ctrmm) -#else -EIGEN_BLAS_TRMM_R(double, double, d, dtrmm_) -EIGEN_BLAS_TRMM_R(dcomplex, double, cd, ztrmm_) -EIGEN_BLAS_TRMM_R(float, float, f, strmm_) -EIGEN_BLAS_TRMM_R(scomplex, float, cf, ctrmm_) -#endif -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h deleted file mode 100644 index 76bfa159ced..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h +++ /dev/null @@ -1,350 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULARMATRIXVECTOR_H -#define EIGEN_TRIANGULARMATRIXVECTOR_H - -namespace Eigen { - -namespace internal { - -template -struct triangular_matrix_vector_product; - -template -struct triangular_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - enum { - IsLower = ((Mode&Lower)==Lower), - HasUnitDiag = (Mode & UnitDiag)==UnitDiag, - HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag - }; - static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const RhsScalar& alpha); -}; - -template -EIGEN_DONT_INLINE void triangular_matrix_vector_product - ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const RhsScalar& alpha) - { - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - Index size = (std::min)(_rows,_cols); - Index rows = IsLower ? _rows : (std::min)(_rows,_cols); - Index cols = IsLower ? (std::min)(_rows,_cols) : _cols; - - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,rows,cols,OuterStride<>(lhsStride)); - typename conj_expr_if::type cjLhs(lhs); - - typedef Map, 0, InnerStride<> > RhsMap; - const RhsMap rhs(_rhs,cols,InnerStride<>(rhsIncr)); - typename conj_expr_if::type cjRhs(rhs); - - typedef Map > ResMap; - ResMap res(_res,rows); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - for (Index pi=0; pi0) - res.segment(s,r) += (alpha * cjRhs.coeff(i)) * cjLhs.col(i).segment(s,r); - if (HasUnitDiag) - res.coeffRef(i) += alpha * cjRhs.coeff(i); - } - Index r = IsLower ? rows - pi - actualPanelWidth : pi; - if (r>0) - { - Index s = IsLower ? pi+actualPanelWidth : 0; - general_matrix_vector_product::run( - r, actualPanelWidth, - LhsMapper(&lhs.coeffRef(s,pi), lhsStride), - RhsMapper(&rhs.coeffRef(pi), rhsIncr), - &res.coeffRef(s), resIncr, alpha); - } - } - if((!IsLower) && cols>size) - { - general_matrix_vector_product::run( - rows, cols-size, - LhsMapper(&lhs.coeffRef(0,size), lhsStride), - RhsMapper(&rhs.coeffRef(size), rhsIncr), - _res, resIncr, alpha); - } - } - -template -struct triangular_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - enum { - IsLower = ((Mode&Lower)==Lower), - HasUnitDiag = (Mode & UnitDiag)==UnitDiag, - HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag - }; - static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha); -}; - -template -EIGEN_DONT_INLINE void triangular_matrix_vector_product - ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha) - { - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - Index diagSize = (std::min)(_rows,_cols); - Index rows = IsLower ? _rows : diagSize; - Index cols = IsLower ? diagSize : _cols; - - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,rows,cols,OuterStride<>(lhsStride)); - typename conj_expr_if::type cjLhs(lhs); - - typedef Map > RhsMap; - const RhsMap rhs(_rhs,cols); - typename conj_expr_if::type cjRhs(rhs); - - typedef Map, 0, InnerStride<> > ResMap; - ResMap res(_res,rows,InnerStride<>(resIncr)); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - for (Index pi=0; pi0) - res.coeffRef(i) += alpha * (cjLhs.row(i).segment(s,r).cwiseProduct(cjRhs.segment(s,r).transpose())).sum(); - if (HasUnitDiag) - res.coeffRef(i) += alpha * cjRhs.coeff(i); - } - Index r = IsLower ? pi : cols - pi - actualPanelWidth; - if (r>0) - { - Index s = IsLower ? 0 : pi + actualPanelWidth; - general_matrix_vector_product::run( - actualPanelWidth, r, - LhsMapper(&lhs.coeffRef(pi,s), lhsStride), - RhsMapper(&rhs.coeffRef(s), rhsIncr), - &res.coeffRef(pi), resIncr, alpha); - } - } - if(IsLower && rows>diagSize) - { - general_matrix_vector_product::run( - rows-diagSize, cols, - LhsMapper(&lhs.coeffRef(diagSize,0), lhsStride), - RhsMapper(&rhs.coeffRef(0), rhsIncr), - &res.coeffRef(diagSize), resIncr, alpha); - } - } - -/*************************************************************************** -* Wrapper to product_triangular_vector -***************************************************************************/ - -template -struct trmv_selector; - -} // end namespace internal - -namespace internal { - -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &lhs, const Rhs &rhs, const typename Dest::Scalar& alpha) - { - eigen_assert(dst.rows()==lhs.rows() && dst.cols()==rhs.cols()); - - internal::trmv_selector::Flags)&RowMajorBit) ? RowMajor : ColMajor>::run(lhs, rhs, dst, alpha); - } -}; - -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &lhs, const Rhs &rhs, const typename Dest::Scalar& alpha) - { - eigen_assert(dst.rows()==lhs.rows() && dst.cols()==rhs.cols()); - - Transpose dstT(dst); - internal::trmv_selector<(Mode & (UnitDiag|ZeroDiag)) | ((Mode & Lower) ? Upper : Lower), - (int(internal::traits::Flags)&RowMajorBit) ? ColMajor : RowMajor> - ::run(rhs.transpose(),lhs.transpose(), dstT, alpha); - } -}; - -} // end namespace internal - -namespace internal { - -// TODO: find a way to factorize this piece of code with gemv_selector since the logic is exactly the same. - -template struct trmv_selector -{ - template - static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar ResScalar; - typedef typename Dest::RealScalar RealScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - - typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; - - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(lhs); - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(rhs); - ResScalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - enum { - // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 - // on, the other hand it is good for the cache to pack the vector anyways... - EvalToDestAtCompileTime = Dest::InnerStrideAtCompileTime==1, - ComplexByReal = (NumTraits::IsComplex) && (!NumTraits::IsComplex), - MightCannotUseDest = (Dest::InnerStrideAtCompileTime!=1) || ComplexByReal - }; - - gemv_static_vector_if static_dest; - - bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0)); - bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible; - - RhsScalar compatibleAlpha = get_factor::run(actualAlpha); - - ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), - evalToDest ? dest.data() : static_dest.data()); - - if(!evalToDest) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = dest.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - if(!alphaIsCompatible) - { - MappedDest(actualDestPtr, dest.size()).setZero(); - compatibleAlpha = RhsScalar(1); - } - else - MappedDest(actualDestPtr, dest.size()) = dest; - } - - internal::triangular_matrix_vector_product - - ::run(actualLhs.rows(),actualLhs.cols(), - actualLhs.data(),actualLhs.outerStride(), - actualRhs.data(),actualRhs.innerStride(), - actualDestPtr,1,compatibleAlpha); - - if (!evalToDest) - { - if(!alphaIsCompatible) - dest += actualAlpha * MappedDest(actualDestPtr, dest.size()); - else - dest = MappedDest(actualDestPtr, dest.size()); - } - - if ( ((Mode&UnitDiag)==UnitDiag) && (lhs_alpha!=LhsScalar(1)) ) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dest.head(diagSize) -= (lhs_alpha-LhsScalar(1))*rhs.head(diagSize); - } - } -}; - -template struct trmv_selector -{ - template - static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar ResScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - typename add_const::type actualLhs = LhsBlasTraits::extract(lhs); - typename add_const::type actualRhs = RhsBlasTraits::extract(rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(rhs); - ResScalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - enum { - DirectlyUseRhs = ActualRhsTypeCleaned::InnerStrideAtCompileTime==1 - }; - - gemv_static_vector_if static_rhs; - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,actualRhs.size(), - DirectlyUseRhs ? const_cast(actualRhs.data()) : static_rhs.data()); - - if(!DirectlyUseRhs) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = actualRhs.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - Map(actualRhsPtr, actualRhs.size()) = actualRhs; - } - - internal::triangular_matrix_vector_product - - ::run(actualLhs.rows(),actualLhs.cols(), - actualLhs.data(),actualLhs.outerStride(), - actualRhsPtr,1, - dest.data(),dest.innerStride(), - actualAlpha); - - if ( ((Mode&UnitDiag)==UnitDiag) && (lhs_alpha!=LhsScalar(1)) ) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dest.head(diagSize) -= (lhs_alpha-LhsScalar(1))*rhs.head(diagSize); - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULARMATRIXVECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h deleted file mode 100644 index 3d47a2b94cb..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h +++ /dev/null @@ -1,255 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix-vector product functionality based on ?TRMV. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H -#define EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements triangular matrix-vector multiplication using BLAS -**********************************************************************/ - -// trmv/hemv specialization - -template -struct triangular_matrix_vector_product_trmv : - triangular_matrix_vector_product {}; - -#define EIGEN_BLAS_TRMV_SPECIALIZE(Scalar) \ -template \ -struct triangular_matrix_vector_product { \ - static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \ - const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \ - triangular_matrix_vector_product_trmv::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - } \ -}; \ -template \ -struct triangular_matrix_vector_product { \ - static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \ - const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \ - triangular_matrix_vector_product_trmv::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - } \ -}; - -EIGEN_BLAS_TRMV_SPECIALIZE(double) -EIGEN_BLAS_TRMV_SPECIALIZE(float) -EIGEN_BLAS_TRMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_TRMV_SPECIALIZE(scomplex) - -// implements col-major: res += alpha * op(triangular) * vector -#define EIGEN_BLAS_TRMV_CM(EIGTYPE, BLASTYPE, EIGPREFIX, BLASPREFIX, BLASPOSTFIX) \ -template \ -struct triangular_matrix_vector_product_trmv { \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper \ - }; \ - static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \ - { \ - if (ConjLhs || IsZeroDiag) { \ - triangular_matrix_vector_product::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - return; \ - }\ - Index size = (std::min)(_rows,_cols); \ - Index rows = IsLower ? _rows : size; \ - Index cols = IsLower ? size : _cols; \ -\ - typedef VectorX##EIGPREFIX VectorRhs; \ - EIGTYPE *x, *y;\ -\ -/* Set x*/ \ - Map > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \ - VectorRhs x_tmp; \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ -\ -/* Square part handling */\ -\ - char trans, uplo, diag; \ - BlasIndex m, n, lda, incx, incy; \ - EIGTYPE const *a; \ - EIGTYPE beta(1); \ -\ -/* Set m, n */ \ - n = convert_index(size); \ - lda = convert_index(lhsStride); \ - incx = 1; \ - incy = convert_index(resIncr); \ -\ -/* Set uplo, trans and diag*/ \ - trans = 'N'; \ - uplo = IsLower ? 'L' : 'U'; \ - diag = IsUnitDiag ? 'U' : 'N'; \ -\ -/* call ?TRMV*/ \ - BLASPREFIX##trmv##BLASPOSTFIX(&uplo, &trans, &diag, &n, (const BLASTYPE*)_lhs, &lda, (BLASTYPE*)x, &incx); \ -\ -/* Add op(a_tr)rhs into res*/ \ - BLASPREFIX##axpy##BLASPOSTFIX(&n, (const BLASTYPE*)&numext::real_ref(alpha),(const BLASTYPE*)x, &incx, (BLASTYPE*)_res, &incy); \ -/* Non-square case - doesn't fit to BLAS ?TRMV. Fall to default triangular product*/ \ - if (size<(std::max)(rows,cols)) { \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ - if (size(rows-size); \ - n = convert_index(size); \ - } \ - else { \ - x += size; \ - y = _res; \ - a = _lhs + size*lda; \ - m = convert_index(size); \ - n = convert_index(cols-size); \ - } \ - BLASPREFIX##gemv##BLASPOSTFIX(&trans, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)x, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)y, &incy); \ - } \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMV_CM(double, double, d, d,) -EIGEN_BLAS_TRMV_CM(dcomplex, MKL_Complex16, cd, z,) -EIGEN_BLAS_TRMV_CM(float, float, f, s,) -EIGEN_BLAS_TRMV_CM(scomplex, MKL_Complex8, cf, c,) -#else -EIGEN_BLAS_TRMV_CM(double, double, d, d, _) -EIGEN_BLAS_TRMV_CM(dcomplex, double, cd, z, _) -EIGEN_BLAS_TRMV_CM(float, float, f, s, _) -EIGEN_BLAS_TRMV_CM(scomplex, float, cf, c, _) -#endif - -// implements row-major: res += alpha * op(triangular) * vector -#define EIGEN_BLAS_TRMV_RM(EIGTYPE, BLASTYPE, EIGPREFIX, BLASPREFIX, BLASPOSTFIX) \ -template \ -struct triangular_matrix_vector_product_trmv { \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper \ - }; \ - static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \ - { \ - if (IsZeroDiag) { \ - triangular_matrix_vector_product::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - return; \ - }\ - Index size = (std::min)(_rows,_cols); \ - Index rows = IsLower ? _rows : size; \ - Index cols = IsLower ? size : _cols; \ -\ - typedef VectorX##EIGPREFIX VectorRhs; \ - EIGTYPE *x, *y;\ -\ -/* Set x*/ \ - Map > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \ - VectorRhs x_tmp; \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ -\ -/* Square part handling */\ -\ - char trans, uplo, diag; \ - BlasIndex m, n, lda, incx, incy; \ - EIGTYPE const *a; \ - EIGTYPE beta(1); \ -\ -/* Set m, n */ \ - n = convert_index(size); \ - lda = convert_index(lhsStride); \ - incx = 1; \ - incy = convert_index(resIncr); \ -\ -/* Set uplo, trans and diag*/ \ - trans = ConjLhs ? 'C' : 'T'; \ - uplo = IsLower ? 'U' : 'L'; \ - diag = IsUnitDiag ? 'U' : 'N'; \ -\ -/* call ?TRMV*/ \ - BLASPREFIX##trmv##BLASPOSTFIX(&uplo, &trans, &diag, &n, (const BLASTYPE*)_lhs, &lda, (BLASTYPE*)x, &incx); \ -\ -/* Add op(a_tr)rhs into res*/ \ - BLASPREFIX##axpy##BLASPOSTFIX(&n, (const BLASTYPE*)&numext::real_ref(alpha),(const BLASTYPE*)x, &incx, (BLASTYPE*)_res, &incy); \ -/* Non-square case - doesn't fit to BLAS ?TRMV. Fall to default triangular product*/ \ - if (size<(std::max)(rows,cols)) { \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ - if (size(rows-size); \ - n = convert_index(size); \ - } \ - else { \ - x += size; \ - y = _res; \ - a = _lhs + size; \ - m = convert_index(size); \ - n = convert_index(cols-size); \ - } \ - BLASPREFIX##gemv##BLASPOSTFIX(&trans, &n, &m, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)x, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)y, &incy); \ - } \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMV_RM(double, double, d, d,) -EIGEN_BLAS_TRMV_RM(dcomplex, MKL_Complex16, cd, z,) -EIGEN_BLAS_TRMV_RM(float, float, f, s,) -EIGEN_BLAS_TRMV_RM(scomplex, MKL_Complex8, cf, c,) -#else -EIGEN_BLAS_TRMV_RM(double, double, d, d,_) -EIGEN_BLAS_TRMV_RM(dcomplex, double, cd, z,_) -EIGEN_BLAS_TRMV_RM(float, float, f, s,_) -EIGEN_BLAS_TRMV_RM(scomplex, float, cf, c,_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h deleted file mode 100644 index e3ed2cd19ea..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h +++ /dev/null @@ -1,335 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_SOLVER_MATRIX_H -#define EIGEN_TRIANGULAR_SOLVER_MATRIX_H - -namespace Eigen { - -namespace internal { - -// if the rhs is row major, let's transpose the product -template -struct triangular_solve_matrix -{ - static void run( - Index size, Index cols, - const Scalar* tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - triangular_solve_matrix< - Scalar, Index, Side==OnTheLeft?OnTheRight:OnTheLeft, - (Mode&UnitDiag) | ((Mode&Upper) ? Lower : Upper), - NumTraits::IsComplex && Conjugate, - TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor, OtherInnerStride> - ::run(size, cols, tri, triStride, _other, otherIncr, otherStride, blocking); - } -}; - -/* Optimized triangular solver with multiple right hand side and the triangular matrix on the left - */ -template -struct triangular_solve_matrix -{ - static EIGEN_DONT_INLINE void run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking); -}; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - Index cols = otherSize; - - typedef const_blas_data_mapper TriMapper; - typedef blas_data_mapper OtherMapper; - TriMapper tri(_tri, triStride); - OtherMapper other(_other, otherStride, otherIncr); - - typedef gebp_traits Traits; - - enum { - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower - }; - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(size,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - conj_if conj; - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - - // the goal here is to subdivise the Rhs panels such that we keep some cache - // coherence when accessing the rhs elements - std::ptrdiff_t l1, l2, l3; - manage_caching_sizes(GetAction, &l1, &l2, &l3); - Index subcols = cols>0 ? l2/(4 * sizeof(Scalar) * std::max(otherStride,size)) : 0; - subcols = std::max((subcols/Traits::nr)*Traits::nr, Traits::nr); - - for(Index k2=IsLower ? 0 : size; - IsLower ? k20; - IsLower ? k2+=kc : k2-=kc) - { - const Index actual_kc = (std::min)(IsLower ? size-k2 : k2, kc); - - // We have selected and packed a big horizontal panel R1 of rhs. Let B be the packed copy of this panel, - // and R2 the remaining part of rhs. The corresponding vertical panel of lhs is split into - // A11 (the triangular part) and A21 the remaining rectangular part. - // Then the high level algorithm is: - // - B = R1 => general block copy (done during the next step) - // - R1 = A11^-1 B => tricky part - // - update B from the new R1 => actually this has to be performed continuously during the above step - // - R2 -= A21 * B => GEPP - - // The tricky part: compute R1 = A11^-1 B while updating B from R1 - // The idea is to split A11 into multiple small vertical panels. - // Each panel can be split into a small triangular part T1k which is processed without optimization, - // and the remaining small part T2k which is processed using gebp with appropriate block strides - for(Index j2=0; j2(actual_kc-k1, SmallPanelWidth); - // tr solve - for (Index k=0; k0) - { - Index startTarget = IsLower ? k2+k1+actualPanelWidth : k2-actual_kc; - - pack_lhs(blockA, tri.getSubMapper(startTarget,startBlock), actualPanelWidth, lengthTarget); - - gebp_kernel(other.getSubMapper(startTarget,j2), blockA, blockB+actual_kc*j2, lengthTarget, actualPanelWidth, actual_cols, Scalar(-1), - actualPanelWidth, actual_kc, 0, blockBOffset); - } - } - } - - // R2 -= A21 * B => GEPP - { - Index start = IsLower ? k2+kc : 0; - Index end = IsLower ? size : k2-kc; - for(Index i2=start; i20) - { - pack_lhs(blockA, tri.getSubMapper(i2, IsLower ? k2 : k2-kc), actual_kc, actual_mc); - - gebp_kernel(other.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, cols, Scalar(-1), -1, -1, 0, 0); - } - } - } - } - } - -/* Optimized triangular solver with multiple left hand sides and the triangular matrix on the right - */ -template -struct triangular_solve_matrix -{ - static EIGEN_DONT_INLINE void run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking); -}; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - Index rows = otherSize; - typedef typename NumTraits::Real RealScalar; - - typedef blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - LhsMapper lhs(_other, otherStride, otherIncr); - RhsMapper rhs(_tri, triStride); - - typedef gebp_traits Traits; - enum { - RhsStorageOrder = TriStorageOrder, - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower - }; - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*size; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - conj_if conj; - gebp_kernel gebp_kernel; - gemm_pack_rhs pack_rhs; - gemm_pack_rhs pack_rhs_panel; - gemm_pack_lhs pack_lhs_panel; - - for(Index k2=IsLower ? size : 0; - IsLower ? k2>0 : k20) pack_rhs(geb, rhs.getSubMapper(actual_k2,startPanel), actual_kc, rs); - - // triangular packing (we only pack the panels off the diagonal, - // neglecting the blocks overlapping the diagonal - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index actual_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc-j2-actualPanelWidth : j2; - - if (panelLength>0) - pack_rhs_panel(blockB+j2*actual_kc, - rhs.getSubMapper(actual_k2+panelOffset, actual_j2), - panelLength, actualPanelWidth, - actual_kc, panelOffset); - } - } - - for(Index i2=0; i2 vertical panels of rhs) - for (Index j2 = IsLower - ? (actual_kc - ((actual_kc%SmallPanelWidth) ? Index(actual_kc%SmallPanelWidth) - : Index(SmallPanelWidth))) - : 0; - IsLower ? j2>=0 : j2(actual_kc-j2, SmallPanelWidth); - Index absolute_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc - j2 - actualPanelWidth : j2; - - // GEBP - if(panelLength>0) - { - gebp_kernel(lhs.getSubMapper(i2,absolute_j2), - blockA, blockB+j2*actual_kc, - actual_mc, panelLength, actualPanelWidth, - Scalar(-1), - actual_kc, actual_kc, // strides - panelOffset, panelOffset); // offsets - } - - // unblocked triangular solve - for (Index k=0; k0) - gebp_kernel(lhs.getSubMapper(i2, startPanel), blockA, geb, - actual_mc, actual_kc, rs, Scalar(-1), - -1, -1, 0, 0); - } - } - } - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h deleted file mode 100644 index 621194ce6ca..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h +++ /dev/null @@ -1,167 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix * matrix product functionality based on ?TRMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H -#define EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - -// implements LeftSide op(triangular)^-1 * general -#define EIGEN_BLAS_TRSM_L(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct triangular_solve_matrix \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - conjA = ((TriStorageOrder==ColMajor) && Conjugate) ? 1 : 0 \ - }; \ - static void run( \ - Index size, Index otherSize, \ - const EIGTYPE* _tri, Index triStride, \ - EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ - BlasIndex m = convert_index(size), n = convert_index(otherSize), lda, ldb; \ - char side = 'L', uplo, diag='N', transa; \ - /* Set alpha_ */ \ - EIGTYPE alpha(1); \ - ldb = convert_index(otherStride);\ -\ - const EIGTYPE *a; \ -/* Set trans */ \ - transa = (TriStorageOrder==RowMajor) ? ((Conjugate) ? 'C' : 'T') : 'N'; \ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (TriStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - typedef Matrix MatrixTri; \ - Map > tri(_tri,size,size,OuterStride<>(triStride)); \ - MatrixTri a_tmp; \ -\ - if (conjA) { \ - a_tmp = tri.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _tri; \ - lda = convert_index(triStride); \ - } \ - if (IsUnitDiag) diag='U'; \ -/* call ?trsm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)_other, &ldb); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRSM_L(double, double, dtrsm) -EIGEN_BLAS_TRSM_L(dcomplex, MKL_Complex16, ztrsm) -EIGEN_BLAS_TRSM_L(float, float, strsm) -EIGEN_BLAS_TRSM_L(scomplex, MKL_Complex8, ctrsm) -#else -EIGEN_BLAS_TRSM_L(double, double, dtrsm_) -EIGEN_BLAS_TRSM_L(dcomplex, double, ztrsm_) -EIGEN_BLAS_TRSM_L(float, float, strsm_) -EIGEN_BLAS_TRSM_L(scomplex, float, ctrsm_) -#endif - -// implements RightSide general * op(triangular)^-1 -#define EIGEN_BLAS_TRSM_R(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct triangular_solve_matrix \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - conjA = ((TriStorageOrder==ColMajor) && Conjugate) ? 1 : 0 \ - }; \ - static void run( \ - Index size, Index otherSize, \ - const EIGTYPE* _tri, Index triStride, \ - EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ - BlasIndex m = convert_index(otherSize), n = convert_index(size), lda, ldb; \ - char side = 'R', uplo, diag='N', transa; \ - /* Set alpha_ */ \ - EIGTYPE alpha(1); \ - ldb = convert_index(otherStride);\ -\ - const EIGTYPE *a; \ -/* Set trans */ \ - transa = (TriStorageOrder==RowMajor) ? ((Conjugate) ? 'C' : 'T') : 'N'; \ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (TriStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - typedef Matrix MatrixTri; \ - Map > tri(_tri,size,size,OuterStride<>(triStride)); \ - MatrixTri a_tmp; \ -\ - if (conjA) { \ - a_tmp = tri.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _tri; \ - lda = convert_index(triStride); \ - } \ - if (IsUnitDiag) diag='U'; \ -/* call ?trsm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)_other, &ldb); \ - /*std::cout << "TRMS_L specialization!\n";*/ \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRSM_R(double, double, dtrsm) -EIGEN_BLAS_TRSM_R(dcomplex, MKL_Complex16, ztrsm) -EIGEN_BLAS_TRSM_R(float, float, strsm) -EIGEN_BLAS_TRSM_R(scomplex, MKL_Complex8, ctrsm) -#else -EIGEN_BLAS_TRSM_R(double, double, dtrsm_) -EIGEN_BLAS_TRSM_R(dcomplex, double, ztrsm_) -EIGEN_BLAS_TRSM_R(float, float, strsm_) -EIGEN_BLAS_TRSM_R(scomplex, float, ctrsm_) -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h deleted file mode 100644 index b994759b268..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h +++ /dev/null @@ -1,145 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_SOLVER_VECTOR_H -#define EIGEN_TRIANGULAR_SOLVER_VECTOR_H - -namespace Eigen { - -namespace internal { - -template -struct triangular_solve_vector -{ - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - triangular_solve_vector::run(size, _lhs, lhsStride, rhs); - } -}; - -// forward and backward substitution, row-major, rhs is a vector -template -struct triangular_solve_vector -{ - enum { - IsLower = ((Mode&Lower)==Lower) - }; - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,size,size,OuterStride<>(lhsStride)); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - typename internal::conditional< - Conjugate, - const CwiseUnaryOp,LhsMap>, - const LhsMap&> - ::type cjLhs(lhs); - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - for(Index pi=IsLower ? 0 : size; - IsLower ? pi0; - IsLower ? pi+=PanelWidth : pi-=PanelWidth) - { - Index actualPanelWidth = (std::min)(IsLower ? size - pi : pi, PanelWidth); - - Index r = IsLower ? pi : size - pi; // remaining size - if (r > 0) - { - // let's directly call the low level product function because: - // 1 - it is faster to compile - // 2 - it is slighlty faster at runtime - Index startRow = IsLower ? pi : pi-actualPanelWidth; - Index startCol = IsLower ? 0 : pi; - - general_matrix_vector_product::run( - actualPanelWidth, r, - LhsMapper(&lhs.coeffRef(startRow,startCol), lhsStride), - RhsMapper(rhs + startCol, 1), - rhs + startRow, 1, - RhsScalar(-1)); - } - - for(Index k=0; k0) - rhs[i] -= (cjLhs.row(i).segment(s,k).transpose().cwiseProduct(Map >(rhs+s,k))).sum(); - - if(!(Mode & UnitDiag)) - rhs[i] /= cjLhs(i,i); - } - } - } -}; - -// forward and backward substitution, column-major, rhs is a vector -template -struct triangular_solve_vector -{ - enum { - IsLower = ((Mode&Lower)==Lower) - }; - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,size,size,OuterStride<>(lhsStride)); - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typename internal::conditional,LhsMap>, - const LhsMap& - >::type cjLhs(lhs); - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - - for(Index pi=IsLower ? 0 : size; - IsLower ? pi0; - IsLower ? pi+=PanelWidth : pi-=PanelWidth) - { - Index actualPanelWidth = (std::min)(IsLower ? size - pi : pi, PanelWidth); - Index startBlock = IsLower ? pi : pi-actualPanelWidth; - Index endBlock = IsLower ? pi + actualPanelWidth : 0; - - for(Index k=0; k0) - Map >(rhs+s,r) -= rhs[i] * cjLhs.col(i).segment(s,r); - } - Index r = IsLower ? size - endBlock : startBlock; // remaining size - if (r > 0) - { - // let's directly call the low level product function because: - // 1 - it is faster to compile - // 2 - it is slighlty faster at runtime - general_matrix_vector_product::run( - r, actualPanelWidth, - LhsMapper(&lhs.coeffRef(endBlock,startBlock), lhsStride), - RhsMapper(rhs+startBlock, 1), - rhs+endBlock, 1, RhsScalar(-1)); - } - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_VECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h deleted file mode 100644 index 3dff9bc9b33..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h +++ /dev/null @@ -1,499 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BLASUTIL_H -#define EIGEN_BLASUTIL_H - -// This file contains many lightweight helper classes used to -// implement and control fast level 2 and level 3 BLAS-like routines. - -namespace Eigen { - -namespace internal { - -// forward declarations -template -struct gebp_kernel; - -template -struct gemm_pack_rhs; - -template -struct gemm_pack_lhs; - -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResStorageOrder, int ResInnerStride> -struct general_matrix_matrix_product; - -template -struct general_matrix_vector_product; - - -template struct conj_if; - -template<> struct conj_if { - template - inline T operator()(const T& x) const { return numext::conj(x); } - template - inline T pconj(const T& x) const { return internal::pconj(x); } -}; - -template<> struct conj_if { - template - inline const T& operator()(const T& x) const { return x; } - template - inline const T& pconj(const T& x) const { return x; } -}; - -// Generic implementation for custom complex types. -template -struct conj_helper -{ - typedef typename ScalarBinaryOpTraits::ReturnType Scalar; - - EIGEN_STRONG_INLINE Scalar pmadd(const LhsScalar& x, const RhsScalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - - EIGEN_STRONG_INLINE Scalar pmul(const LhsScalar& x, const RhsScalar& y) const - { return conj_if()(x) * conj_if()(y); } -}; - -template struct conj_helper -{ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return internal::pmadd(x,y,c); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const { return internal::pmul(x,y); } -}; - -template struct conj_helper, std::complex, false,true> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) + numext::imag(x)*numext::imag(y), numext::imag(x)*numext::real(y) - numext::real(x)*numext::imag(y)); } -}; - -template struct conj_helper, std::complex, true,false> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) + numext::imag(x)*numext::imag(y), numext::real(x)*numext::imag(y) - numext::imag(x)*numext::real(y)); } -}; - -template struct conj_helper, std::complex, true,true> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) - numext::imag(x)*numext::imag(y), - numext::real(x)*numext::imag(y) - numext::imag(x)*numext::real(y)); } -}; - -template struct conj_helper, RealScalar, Conj,false> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const RealScalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const RealScalar& y) const - { return conj_if()(x)*y; } -}; - -template struct conj_helper, false,Conj> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const RealScalar& x, const Scalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - EIGEN_STRONG_INLINE Scalar pmul(const RealScalar& x, const Scalar& y) const - { return x*conj_if()(y); } -}; - -template struct get_factor { - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE To run(const From& x) { return To(x); } -}; - -template struct get_factor::Real> { - EIGEN_DEVICE_FUNC - static EIGEN_STRONG_INLINE typename NumTraits::Real run(const Scalar& x) { return numext::real(x); } -}; - - -template -class BlasVectorMapper { - public: - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasVectorMapper(Scalar *data) : m_data(data) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar operator()(Index i) const { - return m_data[i]; - } - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet load(Index i) const { - return ploadt(m_data + i); - } - - template - EIGEN_DEVICE_FUNC bool aligned(Index i) const { - return (UIntPtr(m_data+i)%sizeof(Packet))==0; - } - - protected: - Scalar* m_data; -}; - -template -class BlasLinearMapper; - -template -class BlasLinearMapper { - public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data, Index incr=1) - : m_data(data) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { - internal::prefetch(&operator()(i)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { - return m_data[i]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { - return ploadt(m_data + i); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE HalfPacket loadHalfPacket(Index i) const { - return ploadt(m_data + i); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const Packet &p) const { - pstoret(m_data + i, p); - } - - protected: - Scalar *m_data; -}; - -// Lightweight helper class to access matrix coefficients. -template -class blas_data_mapper; - -template -class blas_data_mapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - typedef BlasLinearMapper LinearMapper; - typedef BlasVectorMapper VectorMapper; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr=1) - : m_data(data), m_stride(stride) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper - getSubMapper(Index i, Index j) const { - return blas_data_mapper(&operator()(i, j), m_stride); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { - return LinearMapper(&operator()(i, j)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE VectorMapper getVectorMapper(Index i, Index j) const { - return VectorMapper(&operator()(i, j)); - } - - - EIGEN_DEVICE_FUNC - EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { - return m_data[StorageOrder==RowMajor ? j + i*m_stride : i + j*m_stride]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { - return ploadt(&operator()(i, j)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE HalfPacket loadHalfPacket(Index i, Index j) const { - return ploadt(&operator()(i, j)); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { - pscatter(&operator()(i, j), p, m_stride); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { - return pgather(&operator()(i, j), m_stride); - } - - EIGEN_DEVICE_FUNC const Index stride() const { return m_stride; } - EIGEN_DEVICE_FUNC const Scalar* data() const { return m_data; } - - EIGEN_DEVICE_FUNC Index firstAligned(Index size) const { - if (UIntPtr(m_data)%sizeof(Scalar)) { - return -1; - } - return internal::first_default_aligned(m_data, size); - } - - protected: - Scalar* EIGEN_RESTRICT m_data; - const Index m_stride; -}; - -// Implementation of non-natural increment (i.e. inner-stride != 1) -// The exposed API is not complete yet compared to the Incr==1 case -// because some features makes less sense in this case. -template -class BlasLinearMapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data,Index incr) : m_data(data), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { - internal::prefetch(&operator()(i)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { - return m_data[i*m_incr.value()]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { - return pgather(m_data + i*m_incr.value(), m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const PacketType &p) const { - pscatter(m_data + i*m_incr.value(), p, m_incr.value()); - } - -protected: - Scalar *m_data; - const internal::variable_if_dynamic m_incr; -}; - -template -class blas_data_mapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - typedef BlasLinearMapper LinearMapper; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr) : m_data(data), m_stride(stride), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper - getSubMapper(Index i, Index j) const { - return blas_data_mapper(&operator()(i, j), m_stride, m_incr.value()); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { - return LinearMapper(&operator()(i, j), m_incr.value()); - } - - EIGEN_DEVICE_FUNC - EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { - return m_data[StorageOrder==RowMajor ? j*m_incr.value() + i*m_stride : i*m_incr.value() + j*m_stride]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE PacketT load(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { - pscatter(&operator()(i, j), p, m_stride); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { - return pgather(&operator()(i, j), m_stride); - } - -protected: - Scalar* EIGEN_RESTRICT m_data; - const Index m_stride; - const internal::variable_if_dynamic m_incr; -}; - -// lightweight helper class to access matrix coefficients (const version) -template -class const_blas_data_mapper : public blas_data_mapper { - public: - EIGEN_ALWAYS_INLINE const_blas_data_mapper(const Scalar *data, Index stride) : blas_data_mapper(data, stride) {} - - EIGEN_ALWAYS_INLINE const_blas_data_mapper getSubMapper(Index i, Index j) const { - return const_blas_data_mapper(&(this->operator()(i, j)), this->m_stride); - } -}; - - -/* Helper class to analyze the factors of a Product expression. - * In particular it allows to pop out operator-, scalar multiples, - * and conjugate */ -template struct blas_traits -{ - typedef typename traits::Scalar Scalar; - typedef const XprType& ExtractType; - typedef XprType _ExtractType; - enum { - IsComplex = NumTraits::IsComplex, - IsTransposed = false, - NeedToConjugate = false, - HasUsableDirectAccess = ( (int(XprType::Flags)&DirectAccessBit) - && ( bool(XprType::IsVectorAtCompileTime) - || int(inner_stride_at_compile_time::ret) == 1) - ) ? 1 : 0 - }; - typedef typename conditional::type DirectLinearAccessType; - static inline ExtractType extract(const XprType& x) { return x; } - static inline const Scalar extractScalarFactor(const XprType&) { return Scalar(1); } -}; - -// pop conjugate -template -struct blas_traits, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseUnaryOp, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - - enum { - IsComplex = NumTraits::IsComplex, - NeedToConjugate = Base::NeedToConjugate ? 0 : IsComplex - }; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } - static inline Scalar extractScalarFactor(const XprType& x) { return conj(Base::extractScalarFactor(x.nestedExpression())); } -}; - -// pop scalar multiple -template -struct blas_traits, const CwiseNullaryOp,Plain>, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseBinaryOp, const CwiseNullaryOp,Plain>, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.rhs()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return x.lhs().functor().m_other * Base::extractScalarFactor(x.rhs()); } -}; -template -struct blas_traits, NestedXpr, const CwiseNullaryOp,Plain> > > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseBinaryOp, NestedXpr, const CwiseNullaryOp,Plain> > XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.lhs()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return Base::extractScalarFactor(x.lhs()) * x.rhs().functor().m_other; } -}; -template -struct blas_traits, const CwiseNullaryOp,Plain1>, - const CwiseNullaryOp,Plain2> > > - : blas_traits,Plain1> > -{}; - -// pop opposite -template -struct blas_traits, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseUnaryOp, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return - Base::extractScalarFactor(x.nestedExpression()); } -}; - -// pop/push transpose -template -struct blas_traits > - : blas_traits -{ - typedef typename NestedXpr::Scalar Scalar; - typedef blas_traits Base; - typedef Transpose XprType; - typedef Transpose ExtractType; // const to get rid of a compile error; anyway blas traits are only used on the RHS - typedef Transpose _ExtractType; - typedef typename conditional::type DirectLinearAccessType; - enum { - IsTransposed = Base::IsTransposed ? 0 : 1 - }; - static inline ExtractType extract(const XprType& x) { return ExtractType(Base::extract(x.nestedExpression())); } - static inline Scalar extractScalarFactor(const XprType& x) { return Base::extractScalarFactor(x.nestedExpression()); } -}; - -template -struct blas_traits - : blas_traits -{}; - -template::HasUsableDirectAccess> -struct extract_data_selector { - static const typename T::Scalar* run(const T& m) - { - return blas_traits::extract(m).data(); - } -}; - -template -struct extract_data_selector { - static typename T::Scalar* run(const T&) { return 0; } -}; - -template const typename T::Scalar* extract_data(const T& m) -{ - return extract_data_selector::run(m); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BLASUTIL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h deleted file mode 100644 index 7587d684243..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h +++ /dev/null @@ -1,547 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2007-2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_CONSTANTS_H -#define EIGEN_CONSTANTS_H - -namespace Eigen { - -/** This value means that a positive quantity (e.g., a size) is not known at compile-time, and that instead the value is - * stored in some runtime variable. - * - * Changing the value of Dynamic breaks the ABI, as Dynamic is often used as a template parameter for Matrix. - */ -const int Dynamic = -1; - -/** This value means that a signed quantity (e.g., a signed index) is not known at compile-time, and that instead its value - * has to be specified at runtime. - */ -const int DynamicIndex = 0xffffff; - -/** This value means +Infinity; it is currently used only as the p parameter to MatrixBase::lpNorm(). - * The value Infinity there means the L-infinity norm. - */ -const int Infinity = -1; - -/** This value means that the cost to evaluate an expression coefficient is either very expensive or - * cannot be known at compile time. - * - * This value has to be positive to (1) simplify cost computation, and (2) allow to distinguish between a very expensive and very very expensive expressions. - * It thus must also be large enough to make sure unrolling won't happen and that sub expressions will be evaluated, but not too large to avoid overflow. - */ -const int HugeCost = 10000; - -/** \defgroup flags Flags - * \ingroup Core_Module - * - * These are the possible bits which can be OR'ed to constitute the flags of a matrix or - * expression. - * - * It is important to note that these flags are a purely compile-time notion. They are a compile-time property of - * an expression type, implemented as enum's. They are not stored in memory at runtime, and they do not incur any - * runtime overhead. - * - * \sa MatrixBase::Flags - */ - -/** \ingroup flags - * - * for a matrix, this means that the storage order is row-major. - * If this bit is not set, the storage order is column-major. - * For an expression, this determines the storage order of - * the matrix created by evaluation of that expression. - * \sa \blank \ref TopicStorageOrders */ -const unsigned int RowMajorBit = 0x1; - -/** \ingroup flags - * means the expression should be evaluated by the calling expression */ -const unsigned int EvalBeforeNestingBit = 0x2; - -/** \ingroup flags - * \deprecated - * means the expression should be evaluated before any assignment */ -EIGEN_DEPRECATED -const unsigned int EvalBeforeAssigningBit = 0x4; // FIXME deprecated - -/** \ingroup flags - * - * Short version: means the expression might be vectorized - * - * Long version: means that the coefficients can be handled by packets - * and start at a memory location whose alignment meets the requirements - * of the present CPU architecture for optimized packet access. In the fixed-size - * case, there is the additional condition that it be possible to access all the - * coefficients by packets (this implies the requirement that the size be a multiple of 16 bytes, - * and that any nontrivial strides don't break the alignment). In the dynamic-size case, - * there is no such condition on the total size and strides, so it might not be possible to access - * all coeffs by packets. - * - * \note This bit can be set regardless of whether vectorization is actually enabled. - * To check for actual vectorizability, see \a ActualPacketAccessBit. - */ -const unsigned int PacketAccessBit = 0x8; - -#ifdef EIGEN_VECTORIZE -/** \ingroup flags - * - * If vectorization is enabled (EIGEN_VECTORIZE is defined) this constant - * is set to the value \a PacketAccessBit. - * - * If vectorization is not enabled (EIGEN_VECTORIZE is not defined) this constant - * is set to the value 0. - */ -const unsigned int ActualPacketAccessBit = PacketAccessBit; -#else -const unsigned int ActualPacketAccessBit = 0x0; -#endif - -/** \ingroup flags - * - * Short version: means the expression can be seen as 1D vector. - * - * Long version: means that one can access the coefficients - * of this expression by coeff(int), and coeffRef(int) in the case of a lvalue expression. These - * index-based access methods are guaranteed - * to not have to do any runtime computation of a (row, col)-pair from the index, so that it - * is guaranteed that whenever it is available, index-based access is at least as fast as - * (row,col)-based access. Expressions for which that isn't possible don't have the LinearAccessBit. - * - * If both PacketAccessBit and LinearAccessBit are set, then the - * packets of this expression can be accessed by packet(int), and writePacket(int) in the case of a - * lvalue expression. - * - * Typically, all vector expressions have the LinearAccessBit, but there is one exception: - * Product expressions don't have it, because it would be troublesome for vectorization, even when the - * Product is a vector expression. Thus, vector Product expressions allow index-based coefficient access but - * not index-based packet access, so they don't have the LinearAccessBit. - */ -const unsigned int LinearAccessBit = 0x10; - -/** \ingroup flags - * - * Means the expression has a coeffRef() method, i.e. is writable as its individual coefficients are directly addressable. - * This rules out read-only expressions. - * - * Note that DirectAccessBit and LvalueBit are mutually orthogonal, as there are examples of expression having one but note - * the other: - * \li writable expressions that don't have a very simple memory layout as a strided array, have LvalueBit but not DirectAccessBit - * \li Map-to-const expressions, for example Map, have DirectAccessBit but not LvalueBit - * - * Expressions having LvalueBit also have their coeff() method returning a const reference instead of returning a new value. - */ -const unsigned int LvalueBit = 0x20; - -/** \ingroup flags - * - * Means that the underlying array of coefficients can be directly accessed as a plain strided array. The memory layout - * of the array of coefficients must be exactly the natural one suggested by rows(), cols(), - * outerStride(), innerStride(), and the RowMajorBit. This rules out expressions such as Diagonal, whose coefficients, - * though referencable, do not have such a regular memory layout. - * - * See the comment on LvalueBit for an explanation of how LvalueBit and DirectAccessBit are mutually orthogonal. - */ -const unsigned int DirectAccessBit = 0x40; - -/** \deprecated \ingroup flags - * - * means the first coefficient packet is guaranteed to be aligned. - * An expression cannot has the AlignedBit without the PacketAccessBit flag. - * In other words, this means we are allow to perform an aligned packet access to the first element regardless - * of the expression kind: - * \code - * expression.packet(0); - * \endcode - */ -EIGEN_DEPRECATED const unsigned int AlignedBit = 0x80; - -const unsigned int NestByRefBit = 0x100; - -/** \ingroup flags - * - * for an expression, this means that the storage order - * can be either row-major or column-major. - * The precise choice will be decided at evaluation time or when - * combined with other expressions. - * \sa \blank \ref RowMajorBit, \ref TopicStorageOrders */ -const unsigned int NoPreferredStorageOrderBit = 0x200; - -/** \ingroup flags - * - * Means that the underlying coefficients can be accessed through pointers to the sparse (un)compressed storage format, - * that is, the expression provides: - * \code - inline const Scalar* valuePtr() const; - inline const Index* innerIndexPtr() const; - inline const Index* outerIndexPtr() const; - inline const Index* innerNonZeroPtr() const; - \endcode - */ -const unsigned int CompressedAccessBit = 0x400; - - -// list of flags that are inherited by default -const unsigned int HereditaryBits = RowMajorBit - | EvalBeforeNestingBit; - -/** \defgroup enums Enumerations - * \ingroup Core_Module - * - * Various enumerations used in %Eigen. Many of these are used as template parameters. - */ - -/** \ingroup enums - * Enum containing possible values for the \c Mode or \c UpLo parameter of - * MatrixBase::selfadjointView() and MatrixBase::triangularView(), and selfadjoint solvers. */ -enum UpLoType { - /** View matrix as a lower triangular matrix. */ - Lower=0x1, - /** View matrix as an upper triangular matrix. */ - Upper=0x2, - /** %Matrix has ones on the diagonal; to be used in combination with #Lower or #Upper. */ - UnitDiag=0x4, - /** %Matrix has zeros on the diagonal; to be used in combination with #Lower or #Upper. */ - ZeroDiag=0x8, - /** View matrix as a lower triangular matrix with ones on the diagonal. */ - UnitLower=UnitDiag|Lower, - /** View matrix as an upper triangular matrix with ones on the diagonal. */ - UnitUpper=UnitDiag|Upper, - /** View matrix as a lower triangular matrix with zeros on the diagonal. */ - StrictlyLower=ZeroDiag|Lower, - /** View matrix as an upper triangular matrix with zeros on the diagonal. */ - StrictlyUpper=ZeroDiag|Upper, - /** Used in BandMatrix and SelfAdjointView to indicate that the matrix is self-adjoint. */ - SelfAdjoint=0x10, - /** Used to support symmetric, non-selfadjoint, complex matrices. */ - Symmetric=0x20 -}; - -/** \ingroup enums - * Enum for indicating whether a buffer is aligned or not. */ -enum AlignmentType { - Unaligned=0, /**< Data pointer has no specific alignment. */ - Aligned8=8, /**< Data pointer is aligned on a 8 bytes boundary. */ - Aligned16=16, /**< Data pointer is aligned on a 16 bytes boundary. */ - Aligned32=32, /**< Data pointer is aligned on a 32 bytes boundary. */ - Aligned64=64, /**< Data pointer is aligned on a 64 bytes boundary. */ - Aligned128=128, /**< Data pointer is aligned on a 128 bytes boundary. */ - AlignedMask=255, - Aligned=16, /**< \deprecated Synonym for Aligned16. */ -#if EIGEN_MAX_ALIGN_BYTES==128 - AlignedMax = Aligned128 -#elif EIGEN_MAX_ALIGN_BYTES==64 - AlignedMax = Aligned64 -#elif EIGEN_MAX_ALIGN_BYTES==32 - AlignedMax = Aligned32 -#elif EIGEN_MAX_ALIGN_BYTES==16 - AlignedMax = Aligned16 -#elif EIGEN_MAX_ALIGN_BYTES==8 - AlignedMax = Aligned8 -#elif EIGEN_MAX_ALIGN_BYTES==0 - AlignedMax = Unaligned -#else -#error Invalid value for EIGEN_MAX_ALIGN_BYTES -#endif -}; - -/** \ingroup enums - * Enum used by DenseBase::corner() in Eigen2 compatibility mode. */ -// FIXME after the corner() API change, this was not needed anymore, except by AlignedBox -// TODO: find out what to do with that. Adapt the AlignedBox API ? -enum CornerType { TopLeft, TopRight, BottomLeft, BottomRight }; - -/** \ingroup enums - * Enum containing possible values for the \p Direction parameter of - * Reverse, PartialReduxExpr and VectorwiseOp. */ -enum DirectionType { - /** For Reverse, all columns are reversed; - * for PartialReduxExpr and VectorwiseOp, act on columns. */ - Vertical, - /** For Reverse, all rows are reversed; - * for PartialReduxExpr and VectorwiseOp, act on rows. */ - Horizontal, - /** For Reverse, both rows and columns are reversed; - * not used for PartialReduxExpr and VectorwiseOp. */ - BothDirections -}; - -/** \internal \ingroup enums - * Enum to specify how to traverse the entries of a matrix. */ -enum TraversalType { - /** \internal Default traversal, no vectorization, no index-based access */ - DefaultTraversal, - /** \internal No vectorization, use index-based access to have only one for loop instead of 2 nested loops */ - LinearTraversal, - /** \internal Equivalent to a slice vectorization for fixed-size matrices having good alignment - * and good size */ - InnerVectorizedTraversal, - /** \internal Vectorization path using a single loop plus scalar loops for the - * unaligned boundaries */ - LinearVectorizedTraversal, - /** \internal Generic vectorization path using one vectorized loop per row/column with some - * scalar loops to handle the unaligned boundaries */ - SliceVectorizedTraversal, - /** \internal Special case to properly handle incompatible scalar types or other defecting cases*/ - InvalidTraversal, - /** \internal Evaluate all entries at once */ - AllAtOnceTraversal -}; - -/** \internal \ingroup enums - * Enum to specify whether to unroll loops when traversing over the entries of a matrix. */ -enum UnrollingType { - /** \internal Do not unroll loops. */ - NoUnrolling, - /** \internal Unroll only the inner loop, but not the outer loop. */ - InnerUnrolling, - /** \internal Unroll both the inner and the outer loop. If there is only one loop, - * because linear traversal is used, then unroll that loop. */ - CompleteUnrolling -}; - -/** \internal \ingroup enums - * Enum to specify whether to use the default (built-in) implementation or the specialization. */ -enum SpecializedType { - Specialized, - BuiltIn -}; - -/** \ingroup enums - * Enum containing possible values for the \p _Options template parameter of - * Matrix, Array and BandMatrix. */ -enum StorageOptions { - /** Storage order is column major (see \ref TopicStorageOrders). */ - ColMajor = 0, - /** Storage order is row major (see \ref TopicStorageOrders). */ - RowMajor = 0x1, // it is only a coincidence that this is equal to RowMajorBit -- don't rely on that - /** Align the matrix itself if it is vectorizable fixed-size */ - AutoAlign = 0, - /** Don't require alignment for the matrix itself (the array of coefficients, if dynamically allocated, may still be requested to be aligned) */ // FIXME --- clarify the situation - DontAlign = 0x2 -}; - -/** \ingroup enums - * Enum for specifying whether to apply or solve on the left or right. */ -enum SideType { - /** Apply transformation on the left. */ - OnTheLeft = 1, - /** Apply transformation on the right. */ - OnTheRight = 2 -}; - -/* the following used to be written as: - * - * struct NoChange_t {}; - * namespace { - * EIGEN_UNUSED NoChange_t NoChange; - * } - * - * on the ground that it feels dangerous to disambiguate overloaded functions on enum/integer types. - * However, this leads to "variable declared but never referenced" warnings on Intel Composer XE, - * and we do not know how to get rid of them (bug 450). - */ - -enum NoChange_t { NoChange }; -enum Sequential_t { Sequential }; -enum Default_t { Default }; - -/** \internal \ingroup enums - * Used in AmbiVector. */ -enum AmbiVectorMode { - IsDense = 0, - IsSparse -}; - -/** \ingroup enums - * Used as template parameter in DenseCoeffBase and MapBase to indicate - * which accessors should be provided. */ -enum AccessorLevels { - /** Read-only access via a member function. */ - ReadOnlyAccessors, - /** Read/write access via member functions. */ - WriteAccessors, - /** Direct read-only access to the coefficients. */ - DirectAccessors, - /** Direct read/write access to the coefficients. */ - DirectWriteAccessors -}; - -/** \ingroup enums - * Enum with options to give to various decompositions. */ -enum DecompositionOptions { - /** \internal Not used (meant for LDLT?). */ - Pivoting = 0x01, - /** \internal Not used (meant for LDLT?). */ - NoPivoting = 0x02, - /** Used in JacobiSVD to indicate that the square matrix U is to be computed. */ - ComputeFullU = 0x04, - /** Used in JacobiSVD to indicate that the thin matrix U is to be computed. */ - ComputeThinU = 0x08, - /** Used in JacobiSVD to indicate that the square matrix V is to be computed. */ - ComputeFullV = 0x10, - /** Used in JacobiSVD to indicate that the thin matrix V is to be computed. */ - ComputeThinV = 0x20, - /** Used in SelfAdjointEigenSolver and GeneralizedSelfAdjointEigenSolver to specify - * that only the eigenvalues are to be computed and not the eigenvectors. */ - EigenvaluesOnly = 0x40, - /** Used in SelfAdjointEigenSolver and GeneralizedSelfAdjointEigenSolver to specify - * that both the eigenvalues and the eigenvectors are to be computed. */ - ComputeEigenvectors = 0x80, - /** \internal */ - EigVecMask = EigenvaluesOnly | ComputeEigenvectors, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ Ax = \lambda B x \f$. */ - Ax_lBx = 0x100, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ ABx = \lambda x \f$. */ - ABx_lx = 0x200, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ BAx = \lambda x \f$. */ - BAx_lx = 0x400, - /** \internal */ - GenEigMask = Ax_lBx | ABx_lx | BAx_lx -}; - -/** \ingroup enums - * Possible values for the \p QRPreconditioner template parameter of JacobiSVD. */ -enum QRPreconditioners { - /** Do not specify what is to be done if the SVD of a non-square matrix is asked for. */ - NoQRPreconditioner, - /** Use a QR decomposition without pivoting as the first step. */ - HouseholderQRPreconditioner, - /** Use a QR decomposition with column pivoting as the first step. */ - ColPivHouseholderQRPreconditioner, - /** Use a QR decomposition with full pivoting as the first step. */ - FullPivHouseholderQRPreconditioner -}; - -#ifdef Success -#error The preprocessor symbol 'Success' is defined, possibly by the X11 header file X.h -#endif - -/** \ingroup enums - * Enum for reporting the status of a computation. */ -enum ComputationInfo { - /** Computation was successful. */ - Success = 0, - /** The provided data did not satisfy the prerequisites. */ - NumericalIssue = 1, - /** Iterative procedure did not converge. */ - NoConvergence = 2, - /** The inputs are invalid, or the algorithm has been improperly called. - * When assertions are enabled, such errors trigger an assert. */ - InvalidInput = 3 -}; - -/** \ingroup enums - * Enum used to specify how a particular transformation is stored in a matrix. - * \sa Transform, Hyperplane::transform(). */ -enum TransformTraits { - /** Transformation is an isometry. */ - Isometry = 0x1, - /** Transformation is an affine transformation stored as a (Dim+1)^2 matrix whose last row is - * assumed to be [0 ... 0 1]. */ - Affine = 0x2, - /** Transformation is an affine transformation stored as a (Dim) x (Dim+1) matrix. */ - AffineCompact = 0x10 | Affine, - /** Transformation is a general projective transformation stored as a (Dim+1)^2 matrix. */ - Projective = 0x20 -}; - -/** \internal \ingroup enums - * Enum used to choose between implementation depending on the computer architecture. */ -namespace Architecture -{ - enum Type { - Generic = 0x0, - SSE = 0x1, - AltiVec = 0x2, - VSX = 0x3, - NEON = 0x4, -#if defined EIGEN_VECTORIZE_SSE - Target = SSE -#elif defined EIGEN_VECTORIZE_ALTIVEC - Target = AltiVec -#elif defined EIGEN_VECTORIZE_VSX - Target = VSX -#elif defined EIGEN_VECTORIZE_NEON - Target = NEON -#else - Target = Generic -#endif - }; -} - -/** \internal \ingroup enums - * Enum used as template parameter in Product and product evaluators. */ -enum ProductImplType -{ DefaultProduct=0, LazyProduct, AliasFreeProduct, CoeffBasedProductMode, LazyCoeffBasedProductMode, OuterProduct, InnerProduct, GemvProduct, GemmProduct }; - -/** \internal \ingroup enums - * Enum used in experimental parallel implementation. */ -enum Action {GetAction, SetAction}; - -/** The type used to identify a dense storage. */ -struct Dense {}; - -/** The type used to identify a general sparse storage. */ -struct Sparse {}; - -/** The type used to identify a general solver (factored) storage. */ -struct SolverStorage {}; - -/** The type used to identify a permutation storage. */ -struct PermutationStorage {}; - -/** The type used to identify a permutation storage. */ -struct TranspositionsStorage {}; - -/** The type used to identify a matrix expression */ -struct MatrixXpr {}; - -/** The type used to identify an array expression */ -struct ArrayXpr {}; - -// An evaluator must define its shape. By default, it can be one of the following: -struct DenseShape { static std::string debugName() { return "DenseShape"; } }; -struct SolverShape { static std::string debugName() { return "SolverShape"; } }; -struct HomogeneousShape { static std::string debugName() { return "HomogeneousShape"; } }; -struct DiagonalShape { static std::string debugName() { return "DiagonalShape"; } }; -struct BandShape { static std::string debugName() { return "BandShape"; } }; -struct TriangularShape { static std::string debugName() { return "TriangularShape"; } }; -struct SelfAdjointShape { static std::string debugName() { return "SelfAdjointShape"; } }; -struct PermutationShape { static std::string debugName() { return "PermutationShape"; } }; -struct TranspositionsShape { static std::string debugName() { return "TranspositionsShape"; } }; -struct SparseShape { static std::string debugName() { return "SparseShape"; } }; - -namespace internal { - - // random access iterators based on coeff*() accessors. -struct IndexBased {}; - -// evaluator based on iterators to access coefficients. -struct IteratorBased {}; - -/** \internal - * Constants for comparison functors - */ -enum ComparisonName { - cmp_EQ = 0, - cmp_LT = 1, - cmp_LE = 2, - cmp_UNORD = 3, - cmp_NEQ = 4, - cmp_GT = 5, - cmp_GE = 6 -}; -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_CONSTANTS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h deleted file mode 100644 index 74f74cc42b8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h +++ /dev/null @@ -1,94 +0,0 @@ -#ifndef EIGEN_WARNINGS_DISABLED -#define EIGEN_WARNINGS_DISABLED - -#ifdef _MSC_VER - // 4100 - unreferenced formal parameter (occurred e.g. in aligned_allocator::destroy(pointer p)) - // 4101 - unreferenced local variable - // 4127 - conditional expression is constant - // 4181 - qualifier applied to reference type ignored - // 4211 - nonstandard extension used : redefined extern to static - // 4244 - 'argument' : conversion from 'type1' to 'type2', possible loss of data - // 4273 - QtAlignedMalloc, inconsistent DLL linkage - // 4324 - structure was padded due to declspec(align()) - // 4503 - decorated name length exceeded, name was truncated - // 4512 - assignment operator could not be generated - // 4522 - 'class' : multiple assignment operators specified - // 4700 - uninitialized local variable 'xyz' used - // 4714 - function marked as __forceinline not inlined - // 4717 - 'function' : recursive on all control paths, function will cause runtime stack overflow - // 4800 - 'type' : forcing value to bool 'true' or 'false' (performance warning) - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma warning( push ) - #endif - #pragma warning( disable : 4100 4101 4127 4181 4211 4244 4273 4324 4503 4512 4522 4700 4714 4717 4800) - -#elif defined __INTEL_COMPILER - // 2196 - routine is both "inline" and "noinline" ("noinline" assumed) - // ICC 12 generates this warning even without any inline keyword, when defining class methods 'inline' i.e. inside of class body - // typedef that may be a reference type. - // 279 - controlling expression is constant - // ICC 12 generates this warning on assert(constant_expression_depending_on_template_params) and frankly this is a legitimate use case. - // 1684 - conversion from pointer to same-sized integral type (potential portability problem) - // 2259 - non-pointer conversion from "Eigen::Index={ptrdiff_t={long}}" to "int" may lose significant bits - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma warning push - #endif - #pragma warning disable 2196 279 1684 2259 - -#elif defined __clang__ - // -Wconstant-logical-operand - warning: use of logical && with constant operand; switch to bitwise & or remove constant - // this is really a stupid warning as it warns on compile-time expressions involving enums - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma clang diagnostic push - #endif - #pragma clang diagnostic ignored "-Wconstant-logical-operand" - -#elif defined __GNUC__ - - #if (!defined(EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS)) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) - #pragma GCC diagnostic push - #endif - // g++ warns about local variables shadowing member functions, which is too strict - #pragma GCC diagnostic ignored "-Wshadow" - #if __GNUC__ == 4 && __GNUC_MINOR__ < 8 - // Until g++-4.7 there are warnings when comparing unsigned int vs 0, even in templated functions: - #pragma GCC diagnostic ignored "-Wtype-limits" - #endif - #if __GNUC__>=6 - #pragma GCC diagnostic ignored "-Wignored-attributes" - #endif - #if __GNUC__==7 - // See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89325 - #pragma GCC diagnostic ignored "-Wattributes" - #endif -#endif - -#if defined __NVCC__ - // Disable the "statement is unreachable" message - #pragma diag_suppress code_is_unreachable - // Disable the "dynamic initialization in unreachable code" message - #pragma diag_suppress initialization_not_reachable - // Disable the "invalid error number" message that we get with older versions of nvcc - #pragma diag_suppress 1222 - // Disable the "calling a __host__ function from a __host__ __device__ function is not allowed" messages (yes, there are many of them and they seem to change with every version of the compiler) - #pragma diag_suppress 2527 - #pragma diag_suppress 2529 - #pragma diag_suppress 2651 - #pragma diag_suppress 2653 - #pragma diag_suppress 2668 - #pragma diag_suppress 2669 - #pragma diag_suppress 2670 - #pragma diag_suppress 2671 - #pragma diag_suppress 2735 - #pragma diag_suppress 2737 -#endif - -#else -// warnings already disabled: -# ifndef EIGEN_WARNINGS_DISABLED_2 -# define EIGEN_WARNINGS_DISABLED_2 -# elif defined(EIGEN_INTERNAL_DEBUGGING) -# error "Do not include \"DisableStupidWarnings.h\" recursively more than twice!" -# endif - -#endif // not EIGEN_WARNINGS_DISABLED diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h deleted file mode 100644 index 134544f9643..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h +++ /dev/null @@ -1,298 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007-2010 Benoit Jacob -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_FORWARDDECLARATIONS_H -#define EIGEN_FORWARDDECLARATIONS_H - -namespace Eigen { -namespace internal { - -template struct traits; - -// here we say once and for all that traits == traits -// When constness must affect traits, it has to be constness on template parameters on which T itself depends. -// For example, traits > != traits >, but -// traits > == traits > -template struct traits : traits {}; - -template struct has_direct_access -{ - enum { ret = (traits::Flags & DirectAccessBit) ? 1 : 0 }; -}; - -template struct accessors_level -{ - enum { has_direct_access = (traits::Flags & DirectAccessBit) ? 1 : 0, - has_write_access = (traits::Flags & LvalueBit) ? 1 : 0, - value = has_direct_access ? (has_write_access ? DirectWriteAccessors : DirectAccessors) - : (has_write_access ? WriteAccessors : ReadOnlyAccessors) - }; -}; - -template struct evaluator_traits; - -template< typename T> struct evaluator; - -} // end namespace internal - -template struct NumTraits; - -template struct EigenBase; -template class DenseBase; -template class PlainObjectBase; -template class DenseCoeffsBase; - -template class Matrix; - -template class MatrixBase; -template class ArrayBase; - -template class Flagged; -template class StorageBase > class NoAlias; -template class NestByValue; -template class ForceAlignedAccess; -template class SwapWrapper; - -template class Block; - -template class VectorBlock; -template class Transpose; -template class Conjugate; -template class CwiseNullaryOp; -template class CwiseUnaryOp; -template class CwiseUnaryView; -template class CwiseBinaryOp; -template class CwiseTernaryOp; -template class Solve; -template class Inverse; - -template class Product; - -template class DiagonalBase; -template class DiagonalWrapper; -template class DiagonalMatrix; -template class DiagonalProduct; -template class Diagonal; -template class PermutationMatrix; -template class Transpositions; -template class PermutationBase; -template class TranspositionsBase; -template class PermutationWrapper; -template class TranspositionsWrapper; - -template::has_write_access ? WriteAccessors : ReadOnlyAccessors -> class MapBase; -template class Stride; -template class InnerStride; -template class OuterStride; -template > class Map; -template class RefBase; -template,OuterStride<> >::type > class Ref; - -template class TriangularBase; -template class TriangularView; -template class SelfAdjointView; -template class SparseView; -template class WithFormat; -template struct CommaInitializer; -template class ReturnByValue; -template class ArrayWrapper; -template class MatrixWrapper; -template class SolverBase; -template class InnerIterator; - -namespace internal { -template struct kernel_retval_base; -template struct kernel_retval; -template struct image_retval_base; -template struct image_retval; -} // end namespace internal - -namespace internal { -template class BandMatrix; -} - -namespace internal { -template struct product_type; - -template struct EnableIf; - -/** \internal - * \class product_evaluator - * Products need their own evaluator with more template arguments allowing for - * easier partial template specializations. - */ -template< typename T, - int ProductTag = internal::product_type::ret, - typename LhsShape = typename evaluator_traits::Shape, - typename RhsShape = typename evaluator_traits::Shape, - typename LhsScalar = typename traits::Scalar, - typename RhsScalar = typename traits::Scalar - > struct product_evaluator; -} - -template::value> -struct ProductReturnType; - -// this is a workaround for sun CC -template struct LazyProductReturnType; - -namespace internal { - -// Provides scalar/packet-wise product and product with accumulation -// with optional conjugation of the arguments. -template struct conj_helper; - -template struct scalar_sum_op; -template struct scalar_difference_op; -template struct scalar_conj_product_op; -template struct scalar_min_op; -template struct scalar_max_op; -template struct scalar_opposite_op; -template struct scalar_conjugate_op; -template struct scalar_real_op; -template struct scalar_imag_op; -template struct scalar_abs_op; -template struct scalar_abs2_op; -template struct scalar_sqrt_op; -template struct scalar_rsqrt_op; -template struct scalar_exp_op; -template struct scalar_log_op; -template struct scalar_cos_op; -template struct scalar_sin_op; -template struct scalar_acos_op; -template struct scalar_asin_op; -template struct scalar_tan_op; -template struct scalar_inverse_op; -template struct scalar_square_op; -template struct scalar_cube_op; -template struct scalar_cast_op; -template struct scalar_random_op; -template struct scalar_constant_op; -template struct scalar_identity_op; -template struct scalar_sign_op; -template struct scalar_pow_op; -template struct scalar_hypot_op; -template struct scalar_product_op; -template struct scalar_quotient_op; - -// SpecialFunctions module -template struct scalar_lgamma_op; -template struct scalar_digamma_op; -template struct scalar_erf_op; -template struct scalar_erfc_op; -template struct scalar_igamma_op; -template struct scalar_igammac_op; -template struct scalar_zeta_op; -template struct scalar_betainc_op; - -} // end namespace internal - -struct IOFormat; - -// Array module -template class Array; -template class Select; -template class PartialReduxExpr; -template class VectorwiseOp; -template class Replicate; -template class Reverse; - -template class FullPivLU; -template class PartialPivLU; -namespace internal { -template struct inverse_impl; -} -template class HouseholderQR; -template class ColPivHouseholderQR; -template class FullPivHouseholderQR; -template class CompleteOrthogonalDecomposition; -template class JacobiSVD; -template class BDCSVD; -template class LLT; -template class LDLT; -template class HouseholderSequence; -template class JacobiRotation; - -// Geometry module: -template class RotationBase; -template class Cross; -template class QuaternionBase; -template class Rotation2D; -template class AngleAxis; -template class Translation; -template class AlignedBox; -template class Quaternion; -template class Transform; -template class ParametrizedLine; -template class Hyperplane; -template class UniformScaling; -template class Homogeneous; - -// Sparse module: -template class SparseMatrixBase; - -// MatrixFunctions module -template struct MatrixExponentialReturnValue; -template class MatrixFunctionReturnValue; -template class MatrixSquareRootReturnValue; -template class MatrixLogarithmReturnValue; -template class MatrixPowerReturnValue; -template class MatrixComplexPowerReturnValue; - -namespace internal { -template -struct stem_function -{ - typedef std::complex::Real> ComplexScalar; - typedef ComplexScalar type(ComplexScalar, int); -}; -} - -} // end namespace Eigen - -#endif // EIGEN_FORWARDDECLARATIONS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h deleted file mode 100644 index b7d6ecc76e8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h +++ /dev/null @@ -1,130 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to Intel(R) MKL - * Include file with common MKL declarations - ******************************************************************************** -*/ - -#ifndef EIGEN_MKL_SUPPORT_H -#define EIGEN_MKL_SUPPORT_H - -#ifdef EIGEN_USE_MKL_ALL - #ifndef EIGEN_USE_BLAS - #define EIGEN_USE_BLAS - #endif - #ifndef EIGEN_USE_LAPACKE - #define EIGEN_USE_LAPACKE - #endif - #ifndef EIGEN_USE_MKL_VML - #define EIGEN_USE_MKL_VML - #endif -#endif - -#ifdef EIGEN_USE_LAPACKE_STRICT - #define EIGEN_USE_LAPACKE -#endif - -#if defined(EIGEN_USE_MKL_VML) && !defined(EIGEN_USE_MKL) - #define EIGEN_USE_MKL -#endif - - -#if defined EIGEN_USE_MKL -# include -/*Check IMKL version for compatibility: < 10.3 is not usable with Eigen*/ -# ifndef INTEL_MKL_VERSION -# undef EIGEN_USE_MKL /* INTEL_MKL_VERSION is not even defined on older versions */ -# elif INTEL_MKL_VERSION < 100305 /* the intel-mkl-103-release-notes say this was when the lapacke.h interface was added*/ -# undef EIGEN_USE_MKL -# endif -# ifndef EIGEN_USE_MKL - /*If the MKL version is too old, undef everything*/ -# undef EIGEN_USE_MKL_ALL -# undef EIGEN_USE_LAPACKE -# undef EIGEN_USE_MKL_VML -# undef EIGEN_USE_LAPACKE_STRICT -# undef EIGEN_USE_LAPACKE -# endif -#endif - -#if defined EIGEN_USE_MKL - -#define EIGEN_MKL_VML_THRESHOLD 128 - -/* MKL_DOMAIN_BLAS, etc are defined only in 10.3 update 7 */ -/* MKL_BLAS, etc are not defined in 11.2 */ -#ifdef MKL_DOMAIN_ALL -#define EIGEN_MKL_DOMAIN_ALL MKL_DOMAIN_ALL -#else -#define EIGEN_MKL_DOMAIN_ALL MKL_ALL -#endif - -#ifdef MKL_DOMAIN_BLAS -#define EIGEN_MKL_DOMAIN_BLAS MKL_DOMAIN_BLAS -#else -#define EIGEN_MKL_DOMAIN_BLAS MKL_BLAS -#endif - -#ifdef MKL_DOMAIN_FFT -#define EIGEN_MKL_DOMAIN_FFT MKL_DOMAIN_FFT -#else -#define EIGEN_MKL_DOMAIN_FFT MKL_FFT -#endif - -#ifdef MKL_DOMAIN_VML -#define EIGEN_MKL_DOMAIN_VML MKL_DOMAIN_VML -#else -#define EIGEN_MKL_DOMAIN_VML MKL_VML -#endif - -#ifdef MKL_DOMAIN_PARDISO -#define EIGEN_MKL_DOMAIN_PARDISO MKL_DOMAIN_PARDISO -#else -#define EIGEN_MKL_DOMAIN_PARDISO MKL_PARDISO -#endif -#endif - -#if defined(EIGEN_USE_BLAS) && !defined(EIGEN_USE_MKL) -#include "../../misc/blas.h" -#endif - -namespace Eigen { - -typedef std::complex dcomplex; -typedef std::complex scomplex; - -#if defined(EIGEN_USE_MKL) -typedef MKL_INT BlasIndex; -#else -typedef int BlasIndex; -#endif - -} // end namespace Eigen - - -#endif // EIGEN_MKL_SUPPORT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h deleted file mode 100644 index 6b0399ebcc5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h +++ /dev/null @@ -1,1053 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MACROS_H -#define EIGEN_MACROS_H - -#define EIGEN_WORLD_VERSION 3 -#define EIGEN_MAJOR_VERSION 3 -#define EIGEN_MINOR_VERSION 9 - -#define EIGEN_VERSION_AT_LEAST(x,y,z) (EIGEN_WORLD_VERSION>x || (EIGEN_WORLD_VERSION>=x && \ - (EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \ - EIGEN_MINOR_VERSION>=z)))) - -// Compiler identification, EIGEN_COMP_* - -/// \internal EIGEN_COMP_GNUC set to 1 for all compilers compatible with GCC -#ifdef __GNUC__ - #define EIGEN_COMP_GNUC 1 -#else - #define EIGEN_COMP_GNUC 0 -#endif - -/// \internal EIGEN_COMP_CLANG set to major+minor version (e.g., 307 for clang 3.7) if the compiler is clang -#if defined(__clang__) - #define EIGEN_COMP_CLANG (__clang_major__*100+__clang_minor__) -#else - #define EIGEN_COMP_CLANG 0 -#endif - - -/// \internal EIGEN_COMP_LLVM set to 1 if the compiler backend is llvm -#if defined(__llvm__) - #define EIGEN_COMP_LLVM 1 -#else - #define EIGEN_COMP_LLVM 0 -#endif - -/// \internal EIGEN_COMP_ICC set to __INTEL_COMPILER if the compiler is Intel compiler, 0 otherwise -#if defined(__INTEL_COMPILER) - #define EIGEN_COMP_ICC __INTEL_COMPILER -#else - #define EIGEN_COMP_ICC 0 -#endif - -/// \internal EIGEN_COMP_MINGW set to 1 if the compiler is mingw -#if defined(__MINGW32__) - #define EIGEN_COMP_MINGW 1 -#else - #define EIGEN_COMP_MINGW 0 -#endif - -/// \internal EIGEN_COMP_SUNCC set to 1 if the compiler is Solaris Studio -#if defined(__SUNPRO_CC) - #define EIGEN_COMP_SUNCC 1 -#else - #define EIGEN_COMP_SUNCC 0 -#endif - -/// \internal EIGEN_COMP_MSVC set to _MSC_VER if the compiler is Microsoft Visual C++, 0 otherwise. -#if defined(_MSC_VER) - #define EIGEN_COMP_MSVC _MSC_VER -#else - #define EIGEN_COMP_MSVC 0 -#endif - -// For the record, here is a table summarizing the possible values for EIGEN_COMP_MSVC: -// name ver MSC_VER -// 2008 9 1500 -// 2010 10 1600 -// 2012 11 1700 -// 2013 12 1800 -// 2015 14 1900 -// "15" 15 1900 - -/// \internal EIGEN_COMP_MSVC_STRICT set to 1 if the compiler is really Microsoft Visual C++ and not ,e.g., ICC or clang-cl -#if EIGEN_COMP_MSVC && !(EIGEN_COMP_ICC || EIGEN_COMP_LLVM || EIGEN_COMP_CLANG) - #define EIGEN_COMP_MSVC_STRICT _MSC_VER -#else - #define EIGEN_COMP_MSVC_STRICT 0 -#endif - -/// \internal EIGEN_COMP_IBM set to 1 if the compiler is IBM XL C++ -#if defined(__IBMCPP__) || defined(__xlc__) - #define EIGEN_COMP_IBM 1 -#else - #define EIGEN_COMP_IBM 0 -#endif - -/// \internal EIGEN_COMP_PGI set to 1 if the compiler is Portland Group Compiler -#if defined(__PGI) - #define EIGEN_COMP_PGI 1 -#else - #define EIGEN_COMP_PGI 0 -#endif - -/// \internal EIGEN_COMP_ARM set to 1 if the compiler is ARM Compiler -#if defined(__CC_ARM) || defined(__ARMCC_VERSION) - #define EIGEN_COMP_ARM 1 -#else - #define EIGEN_COMP_ARM 0 -#endif - -/// \internal EIGEN_COMP_ARM set to 1 if the compiler is ARM Compiler -#if defined(__EMSCRIPTEN__) - #define EIGEN_COMP_EMSCRIPTEN 1 -#else - #define EIGEN_COMP_EMSCRIPTEN 0 -#endif - - -/// \internal EIGEN_GNUC_STRICT set to 1 if the compiler is really GCC and not a compatible compiler (e.g., ICC, clang, mingw, etc.) -#if EIGEN_COMP_GNUC && !(EIGEN_COMP_CLANG || EIGEN_COMP_ICC || EIGEN_COMP_MINGW || EIGEN_COMP_PGI || EIGEN_COMP_IBM || EIGEN_COMP_ARM || EIGEN_COMP_EMSCRIPTEN) - #define EIGEN_COMP_GNUC_STRICT 1 -#else - #define EIGEN_COMP_GNUC_STRICT 0 -#endif - - -#if EIGEN_COMP_GNUC - #define EIGEN_GNUC_AT_LEAST(x,y) ((__GNUC__==x && __GNUC_MINOR__>=y) || __GNUC__>x) - #define EIGEN_GNUC_AT_MOST(x,y) ((__GNUC__==x && __GNUC_MINOR__<=y) || __GNUC__=11 && (defined(__cplusplus) && (__cplusplus >= 201103L) || EIGEN_COMP_MSVC >= 1900) -#define EIGEN_HAS_CXX11 1 -#else -#define EIGEN_HAS_CXX11 0 -#endif - - -// Do we support r-value references? -#ifndef EIGEN_HAS_RVALUE_REFERENCES -#if EIGEN_MAX_CPP_VER>=11 && \ - (__has_feature(cxx_rvalue_references) || \ - (defined(__cplusplus) && __cplusplus >= 201103L) || \ - (EIGEN_COMP_MSVC >= 1600)) - #define EIGEN_HAS_RVALUE_REFERENCES 1 -#else - #define EIGEN_HAS_RVALUE_REFERENCES 0 -#endif -#endif - -// Does the compiler support C99? -#ifndef EIGEN_HAS_C99_MATH -#if EIGEN_MAX_CPP_VER>=11 && \ - ((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)) \ - || (defined(__GNUC__) && defined(_GLIBCXX_USE_C99)) \ - || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER)) \ - || (EIGEN_COMP_MSVC >= 1900) ) - #define EIGEN_HAS_C99_MATH 1 -#else - #define EIGEN_HAS_C99_MATH 0 -#endif -#endif - -// Does the compiler support result_of? -#ifndef EIGEN_HAS_STD_RESULT_OF -#if EIGEN_MAX_CPP_VER>=11 && ((__has_feature(cxx_lambdas) || (defined(__cplusplus) && __cplusplus >= 201103L))) -#define EIGEN_HAS_STD_RESULT_OF 1 -#else -#define EIGEN_HAS_STD_RESULT_OF 0 -#endif -#endif - -// Does the compiler support type_traits? -// - full support of type traits was added only to GCC 5.1.0. -// - 20150626 corresponds to the last release of 4.x libstdc++ -#ifndef EIGEN_HAS_TYPE_TRAITS -#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_HAS_CXX11 || EIGEN_COMP_MSVC >= 1700) \ - && ((!EIGEN_COMP_GNUC_STRICT) || EIGEN_GNUC_AT_LEAST(5, 1)) \ - && ((!defined(__GLIBCXX__)) || __GLIBCXX__ > 20150626) -#define EIGEN_HAS_TYPE_TRAITS 1 -#define EIGEN_INCLUDE_TYPE_TRAITS -#else -#define EIGEN_HAS_TYPE_TRAITS 0 -#endif -#endif - -// Does the compiler support variadic templates? -#ifndef EIGEN_HAS_VARIADIC_TEMPLATES -#if EIGEN_MAX_CPP_VER>=11 && (__cplusplus > 199711L || EIGEN_COMP_MSVC >= 1900) \ - && (!defined(__NVCC__) || !EIGEN_ARCH_ARM_OR_ARM64 || (EIGEN_CUDACC_VER >= 80000) ) - // ^^ Disable the use of variadic templates when compiling with versions of nvcc older than 8.0 on ARM devices: - // this prevents nvcc from crashing when compiling Eigen on Tegra X1 -#define EIGEN_HAS_VARIADIC_TEMPLATES 1 -#else -#define EIGEN_HAS_VARIADIC_TEMPLATES 0 -#endif -#endif - -// Does the compiler fully support const expressions? (as in c++14) -#ifndef EIGEN_HAS_CONSTEXPR - -#ifdef __CUDACC__ -// Const expressions are supported provided that c++11 is enabled and we're using either clang or nvcc 7.5 or above -#if EIGEN_MAX_CPP_VER>=14 && (__cplusplus > 199711L && (EIGEN_COMP_CLANG || EIGEN_CUDACC_VER >= 70500)) - #define EIGEN_HAS_CONSTEXPR 1 -#endif -#elif EIGEN_MAX_CPP_VER>=14 && (__has_feature(cxx_relaxed_constexpr) || (defined(__cplusplus) && __cplusplus >= 201402L) || \ - (EIGEN_GNUC_AT_LEAST(4,8) && (__cplusplus > 199711L))) -#define EIGEN_HAS_CONSTEXPR 1 -#endif - -#ifndef EIGEN_HAS_CONSTEXPR -#define EIGEN_HAS_CONSTEXPR 0 -#endif - -#endif - -// Does the compiler support C++11 math? -// Let's be conservative and enable the default C++11 implementation only if we are sure it exists -#ifndef EIGEN_HAS_CXX11_MATH - #if EIGEN_MAX_CPP_VER>=11 && ((__cplusplus > 201103L) || (__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC) \ - && (EIGEN_ARCH_i386_OR_x86_64) && (EIGEN_OS_GNULINUX || EIGEN_OS_WIN_STRICT || EIGEN_OS_MAC)) - #define EIGEN_HAS_CXX11_MATH 1 - #else - #define EIGEN_HAS_CXX11_MATH 0 - #endif -#endif - -// Does the compiler support proper C++11 containers? -#ifndef EIGEN_HAS_CXX11_CONTAINERS - #if EIGEN_MAX_CPP_VER>=11 && \ - ((__cplusplus > 201103L) \ - || ((__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_ICC>=1400)) \ - || EIGEN_COMP_MSVC >= 1900) - #define EIGEN_HAS_CXX11_CONTAINERS 1 - #else - #define EIGEN_HAS_CXX11_CONTAINERS 0 - #endif -#endif - -// Does the compiler support C++11 noexcept? -#ifndef EIGEN_HAS_CXX11_NOEXCEPT - #if EIGEN_MAX_CPP_VER>=11 && \ - (__has_feature(cxx_noexcept) \ - || (__cplusplus > 201103L) \ - || ((__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_ICC>=1400)) \ - || EIGEN_COMP_MSVC >= 1900) - #define EIGEN_HAS_CXX11_NOEXCEPT 1 - #else - #define EIGEN_HAS_CXX11_NOEXCEPT 0 - #endif -#endif - -/** Allows to disable some optimizations which might affect the accuracy of the result. - * Such optimization are enabled by default, and set EIGEN_FAST_MATH to 0 to disable them. - * They currently include: - * - single precision ArrayBase::sin() and ArrayBase::cos() for SSE and AVX vectorization. - */ -#ifndef EIGEN_FAST_MATH -#define EIGEN_FAST_MATH 1 -#endif - -#define EIGEN_DEBUG_VAR(x) std::cerr << #x << " = " << x << std::endl; - -// concatenate two tokens -#define EIGEN_CAT2(a,b) a ## b -#define EIGEN_CAT(a,b) EIGEN_CAT2(a,b) - -#define EIGEN_COMMA , - -// convert a token to a string -#define EIGEN_MAKESTRING2(a) #a -#define EIGEN_MAKESTRING(a) EIGEN_MAKESTRING2(a) - -// EIGEN_STRONG_INLINE is a stronger version of the inline, using __forceinline on MSVC, -// but it still doesn't use GCC's always_inline. This is useful in (common) situations where MSVC needs forceinline -// but GCC is still doing fine with just inline. -#ifndef EIGEN_STRONG_INLINE -#if EIGEN_COMP_MSVC || EIGEN_COMP_ICC -#define EIGEN_STRONG_INLINE __forceinline -#else -#define EIGEN_STRONG_INLINE inline -#endif -#endif - -// EIGEN_ALWAYS_INLINE is the stronget, it has the effect of making the function inline and adding every possible -// attribute to maximize inlining. This should only be used when really necessary: in particular, -// it uses __attribute__((always_inline)) on GCC, which most of the time is useless and can severely harm compile times. -// FIXME with the always_inline attribute, -// gcc 3.4.x and 4.1 reports the following compilation error: -// Eval.h:91: sorry, unimplemented: inlining failed in call to 'const Eigen::Eval Eigen::MatrixBase::eval() const' -// : function body not available -// See also bug 1367 -#if EIGEN_GNUC_AT_LEAST(4,2) -#define EIGEN_ALWAYS_INLINE __attribute__((always_inline)) inline -#else -#define EIGEN_ALWAYS_INLINE EIGEN_STRONG_INLINE -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_DONT_INLINE __attribute__((noinline)) -#elif EIGEN_COMP_MSVC -#define EIGEN_DONT_INLINE __declspec(noinline) -#else -#define EIGEN_DONT_INLINE -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_PERMISSIVE_EXPR __extension__ -#else -#define EIGEN_PERMISSIVE_EXPR -#endif - -// this macro allows to get rid of linking errors about multiply defined functions. -// - static is not very good because it prevents definitions from different object files to be merged. -// So static causes the resulting linked executable to be bloated with multiple copies of the same function. -// - inline is not perfect either as it unwantedly hints the compiler toward inlining the function. -#define EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS -#define EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS inline - -#ifdef NDEBUG -# ifndef EIGEN_NO_DEBUG -# define EIGEN_NO_DEBUG -# endif -#endif - -// eigen_plain_assert is where we implement the workaround for the assert() bug in GCC <= 4.3, see bug 89 -#ifdef EIGEN_NO_DEBUG - #define eigen_plain_assert(x) -#else - #if EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO - namespace Eigen { - namespace internal { - inline bool copy_bool(bool b) { return b; } - } - } - #define eigen_plain_assert(x) assert(x) - #else - // work around bug 89 - #include // for abort - #include // for std::cerr - - namespace Eigen { - namespace internal { - // trivial function copying a bool. Must be EIGEN_DONT_INLINE, so we implement it after including Eigen headers. - // see bug 89. - namespace { - EIGEN_DONT_INLINE bool copy_bool(bool b) { return b; } - } - inline void assert_fail(const char *condition, const char *function, const char *file, int line) - { - std::cerr << "assertion failed: " << condition << " in function " << function << " at " << file << ":" << line << std::endl; - abort(); - } - } - } - #define eigen_plain_assert(x) \ - do { \ - if(!Eigen::internal::copy_bool(x)) \ - Eigen::internal::assert_fail(EIGEN_MAKESTRING(x), __PRETTY_FUNCTION__, __FILE__, __LINE__); \ - } while(false) - #endif -#endif - -// eigen_assert can be overridden -#ifndef eigen_assert -#define eigen_assert(x) eigen_plain_assert(x) -#endif - -#ifdef EIGEN_INTERNAL_DEBUGGING -#define eigen_internal_assert(x) eigen_assert(x) -#else -#define eigen_internal_assert(x) -#endif - -#ifdef EIGEN_NO_DEBUG -#define EIGEN_ONLY_USED_FOR_DEBUG(x) EIGEN_UNUSED_VARIABLE(x) -#else -#define EIGEN_ONLY_USED_FOR_DEBUG(x) -#endif - -#ifndef EIGEN_NO_DEPRECATED_WARNING - #if EIGEN_COMP_GNUC - #define EIGEN_DEPRECATED __attribute__((deprecated)) - #elif EIGEN_COMP_MSVC - #define EIGEN_DEPRECATED __declspec(deprecated) - #else - #define EIGEN_DEPRECATED - #endif -#else - #define EIGEN_DEPRECATED -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_UNUSED __attribute__((unused)) -#else -#define EIGEN_UNUSED -#endif - -// Suppresses 'unused variable' warnings. -namespace Eigen { - namespace internal { - template EIGEN_DEVICE_FUNC void ignore_unused_variable(const T&) {} - } -} -#define EIGEN_UNUSED_VARIABLE(var) Eigen::internal::ignore_unused_variable(var); - -#if !defined(EIGEN_ASM_COMMENT) - #if EIGEN_COMP_GNUC && (EIGEN_ARCH_i386_OR_x86_64 || EIGEN_ARCH_ARM_OR_ARM64) - #define EIGEN_ASM_COMMENT(X) __asm__("#" X) - #else - #define EIGEN_ASM_COMMENT(X) - #endif -#endif - - -//------------------------------------------------------------------------------------------ -// Static and dynamic alignment control -// -// The main purpose of this section is to define EIGEN_MAX_ALIGN_BYTES and EIGEN_MAX_STATIC_ALIGN_BYTES -// as the maximal boundary in bytes on which dynamically and statically allocated data may be alignment respectively. -// The values of EIGEN_MAX_ALIGN_BYTES and EIGEN_MAX_STATIC_ALIGN_BYTES can be specified by the user. If not, -// a default value is automatically computed based on architecture, compiler, and OS. -// -// This section also defines macros EIGEN_ALIGN_TO_BOUNDARY(N) and the shortcuts EIGEN_ALIGN{8,16,32,_MAX} -// to be used to declare statically aligned buffers. -//------------------------------------------------------------------------------------------ - - -/* EIGEN_ALIGN_TO_BOUNDARY(n) forces data to be n-byte aligned. This is used to satisfy SIMD requirements. - * However, we do that EVEN if vectorization (EIGEN_VECTORIZE) is disabled, - * so that vectorization doesn't affect binary compatibility. - * - * If we made alignment depend on whether or not EIGEN_VECTORIZE is defined, it would be impossible to link - * vectorized and non-vectorized code. - */ -#if (defined __CUDACC__) - #define EIGEN_ALIGN_TO_BOUNDARY(n) __align__(n) -#elif EIGEN_COMP_GNUC || EIGEN_COMP_PGI || EIGEN_COMP_IBM || EIGEN_COMP_ARM - #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n))) -#elif EIGEN_COMP_MSVC - #define EIGEN_ALIGN_TO_BOUNDARY(n) __declspec(align(n)) -#elif EIGEN_COMP_SUNCC - // FIXME not sure about this one: - #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n))) -#else - #error Please tell me what is the equivalent of __attribute__((aligned(n))) for your compiler -#endif - -// If the user explicitly disable vectorization, then we also disable alignment -#if defined(EIGEN_DONT_VECTORIZE) - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 0 -#elif defined(EIGEN_VECTORIZE_AVX512) - // 64 bytes static alignmeent is preferred only if really required - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 64 -#elif defined(__AVX__) - // 32 bytes static alignmeent is preferred only if really required - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 32 -#else - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 16 -#endif - - -// EIGEN_MIN_ALIGN_BYTES defines the minimal value for which the notion of explicit alignment makes sense -#define EIGEN_MIN_ALIGN_BYTES 16 - -// Defined the boundary (in bytes) on which the data needs to be aligned. Note -// that unless EIGEN_ALIGN is defined and not equal to 0, the data may not be -// aligned at all regardless of the value of this #define. - -#if (defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN)) && defined(EIGEN_MAX_STATIC_ALIGN_BYTES) && EIGEN_MAX_STATIC_ALIGN_BYTES>0 -#error EIGEN_MAX_STATIC_ALIGN_BYTES and EIGEN_DONT_ALIGN[_STATICALLY] are both defined with EIGEN_MAX_STATIC_ALIGN_BYTES!=0. Use EIGEN_MAX_STATIC_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN_STATICALLY. -#endif - -// EIGEN_DONT_ALIGN_STATICALLY and EIGEN_DONT_ALIGN are deprectated -// They imply EIGEN_MAX_STATIC_ALIGN_BYTES=0 -#if defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN) - #ifdef EIGEN_MAX_STATIC_ALIGN_BYTES - #undef EIGEN_MAX_STATIC_ALIGN_BYTES - #endif - #define EIGEN_MAX_STATIC_ALIGN_BYTES 0 -#endif - -#ifndef EIGEN_MAX_STATIC_ALIGN_BYTES - - // Try to automatically guess what is the best default value for EIGEN_MAX_STATIC_ALIGN_BYTES - - // 16 byte alignment is only useful for vectorization. Since it affects the ABI, we need to enable - // 16 byte alignment on all platforms where vectorization might be enabled. In theory we could always - // enable alignment, but it can be a cause of problems on some platforms, so we just disable it in - // certain common platform (compiler+architecture combinations) to avoid these problems. - // Only static alignment is really problematic (relies on nonstandard compiler extensions), - // try to keep heap alignment even when we have to disable static alignment. - #if EIGEN_COMP_GNUC && !(EIGEN_ARCH_i386_OR_x86_64 || EIGEN_ARCH_ARM_OR_ARM64 || EIGEN_ARCH_PPC || EIGEN_ARCH_IA64) - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1 - #elif EIGEN_ARCH_ARM_OR_ARM64 && EIGEN_COMP_GNUC_STRICT && EIGEN_GNUC_AT_MOST(4, 6) - // Old versions of GCC on ARM, at least 4.4, were once seen to have buggy static alignment support. - // Not sure which version fixed it, hopefully it doesn't affect 4.7, which is still somewhat in use. - // 4.8 and newer seem definitely unaffected. - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1 - #else - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 0 - #endif - - // static alignment is completely disabled with GCC 3, Sun Studio, and QCC/QNX - #if !EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT \ - && !EIGEN_GCC3_OR_OLDER \ - && !EIGEN_COMP_SUNCC \ - && !EIGEN_OS_QNX - #define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 1 - #else - #define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 0 - #endif - - #if EIGEN_ARCH_WANTS_STACK_ALIGNMENT - #define EIGEN_MAX_STATIC_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES - #else - #define EIGEN_MAX_STATIC_ALIGN_BYTES 0 - #endif - -#endif - -// If EIGEN_MAX_ALIGN_BYTES is defined, then it is considered as an upper bound for EIGEN_MAX_ALIGN_BYTES -#if defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES0 is the true test whether we want to align arrays on the stack or not. -// It takes into account both the user choice to explicitly enable/disable alignment (by settting EIGEN_MAX_STATIC_ALIGN_BYTES) -// and the architecture config (EIGEN_ARCH_WANTS_STACK_ALIGNMENT). -// Henceforth, only EIGEN_MAX_STATIC_ALIGN_BYTES should be used. - - -// Shortcuts to EIGEN_ALIGN_TO_BOUNDARY -#define EIGEN_ALIGN8 EIGEN_ALIGN_TO_BOUNDARY(8) -#define EIGEN_ALIGN16 EIGEN_ALIGN_TO_BOUNDARY(16) -#define EIGEN_ALIGN32 EIGEN_ALIGN_TO_BOUNDARY(32) -#define EIGEN_ALIGN64 EIGEN_ALIGN_TO_BOUNDARY(64) -#if EIGEN_MAX_STATIC_ALIGN_BYTES>0 -#define EIGEN_ALIGN_MAX EIGEN_ALIGN_TO_BOUNDARY(EIGEN_MAX_STATIC_ALIGN_BYTES) -#else -#define EIGEN_ALIGN_MAX -#endif - - -// Dynamic alignment control - -#if defined(EIGEN_DONT_ALIGN) && defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES>0 -#error EIGEN_MAX_ALIGN_BYTES and EIGEN_DONT_ALIGN are both defined with EIGEN_MAX_ALIGN_BYTES!=0. Use EIGEN_MAX_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN. -#endif - -#ifdef EIGEN_DONT_ALIGN - #ifdef EIGEN_MAX_ALIGN_BYTES - #undef EIGEN_MAX_ALIGN_BYTES - #endif - #define EIGEN_MAX_ALIGN_BYTES 0 -#elif !defined(EIGEN_MAX_ALIGN_BYTES) - #define EIGEN_MAX_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES -#endif - -#if EIGEN_IDEAL_MAX_ALIGN_BYTES > EIGEN_MAX_ALIGN_BYTES -#define EIGEN_DEFAULT_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES -#else -#define EIGEN_DEFAULT_ALIGN_BYTES EIGEN_MAX_ALIGN_BYTES -#endif - - -#ifndef EIGEN_UNALIGNED_VECTORIZE -#define EIGEN_UNALIGNED_VECTORIZE 1 -#endif - -//---------------------------------------------------------------------- - - -#ifdef EIGEN_DONT_USE_RESTRICT_KEYWORD - #define EIGEN_RESTRICT -#endif -#ifndef EIGEN_RESTRICT - #define EIGEN_RESTRICT __restrict -#endif - -#ifndef EIGEN_STACK_ALLOCATION_LIMIT -// 131072 == 128 KB -#define EIGEN_STACK_ALLOCATION_LIMIT 131072 -#endif - -#ifndef EIGEN_DEFAULT_IO_FORMAT -#ifdef EIGEN_MAKING_DOCS -// format used in Eigen's documentation -// needed to define it here as escaping characters in CMake add_definition's argument seems very problematic. -#define EIGEN_DEFAULT_IO_FORMAT Eigen::IOFormat(3, 0, " ", "\n", "", "") -#else -#define EIGEN_DEFAULT_IO_FORMAT Eigen::IOFormat() -#endif -#endif - -// just an empty macro ! -#define EIGEN_EMPTY - -#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC < 1900 || EIGEN_CUDACC_VER>0) - // for older MSVC versions, as well as 1900 && CUDA 8, using the base operator is sufficient (cf Bugs 1000, 1324) - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; -#elif EIGEN_COMP_CLANG // workaround clang bug (see http://forum.kde.org/viewtopic.php?f=74&t=102653) - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const Derived& other) { Base::operator=(other); return *this; } \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const DenseBase& other) { Base::operator=(other.derived()); return *this; } -#else - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const Derived& other) \ - { \ - Base::operator=(other); \ - return *this; \ - } -#endif - - -/** - * \internal - * \brief Macro to explicitly define the default copy constructor. - * This is necessary, because the implicit definition is deprecated if the copy-assignment is overridden. - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) EIGEN_DEVICE_FUNC CLASS(const CLASS&) = default; -#else -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) -#endif - - - -/** \internal - * \brief Macro to manually inherit assignment operators. - * This is necessary, because the implicitly defined assignment operator gets deleted when a custom operator= is defined. - * With C++11 or later this also default-implements the copy-constructor - */ -#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) \ - EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - EIGEN_DEFAULT_COPY_CONSTRUCTOR(Derived) - -/** \internal - * \brief Macro to manually define default constructors and destructors. - * This is necessary when the copy constructor is re-defined. - * For empty helper classes this should usually be protected, to avoid accidentally creating empty objects. - * - * Hiding the default destructor lead to problems in C++03 mode together with boost::multiprecision - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() = default; \ - EIGEN_DEVICE_FUNC ~Derived() = default; -#else -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() {}; \ - /* EIGEN_DEVICE_FUNC ~Derived() {}; */ -#endif - - - - - -/** -* Just a side note. Commenting within defines works only by documenting -* behind the object (via '!<'). Comments cannot be multi-line and thus -* we have these extra long lines. What is confusing doxygen over here is -* that we use '\' and basically have a bunch of typedefs with their -* documentation in a single line. -**/ - -#define EIGEN_GENERIC_PUBLIC_INTERFACE(Derived) \ - typedef typename Eigen::internal::traits::Scalar Scalar; /*!< \brief Numeric type, e.g. float, double, int or std::complex. */ \ - typedef typename Eigen::NumTraits::Real RealScalar; /*!< \brief The underlying numeric type for composed scalar types. \details In cases where Scalar is e.g. std::complex, T were corresponding to RealScalar. */ \ - typedef typename Base::CoeffReturnType CoeffReturnType; /*!< \brief The return type for coefficient access. \details Depending on whether the object allows direct coefficient access (e.g. for a MatrixXd), this type is either 'const Scalar&' or simply 'Scalar' for objects that do not allow direct coefficient access. */ \ - typedef typename Eigen::internal::ref_selector::type Nested; \ - typedef typename Eigen::internal::traits::StorageKind StorageKind; \ - typedef typename Eigen::internal::traits::StorageIndex StorageIndex; \ - enum { RowsAtCompileTime = Eigen::internal::traits::RowsAtCompileTime, \ - ColsAtCompileTime = Eigen::internal::traits::ColsAtCompileTime, \ - Flags = Eigen::internal::traits::Flags, \ - SizeAtCompileTime = Base::SizeAtCompileTime, \ - MaxSizeAtCompileTime = Base::MaxSizeAtCompileTime, \ - IsVectorAtCompileTime = Base::IsVectorAtCompileTime }; \ - using Base::derived; \ - using Base::const_cast_derived; - - -// FIXME Maybe the EIGEN_DENSE_PUBLIC_INTERFACE could be removed as importing PacketScalar is rarely needed -#define EIGEN_DENSE_PUBLIC_INTERFACE(Derived) \ - EIGEN_GENERIC_PUBLIC_INTERFACE(Derived) \ - typedef typename Base::PacketScalar PacketScalar; - - -#define EIGEN_PLAIN_ENUM_MIN(a,b) (((int)a <= (int)b) ? (int)a : (int)b) -#define EIGEN_PLAIN_ENUM_MAX(a,b) (((int)a >= (int)b) ? (int)a : (int)b) - -// EIGEN_SIZE_MIN_PREFER_DYNAMIC gives the min between compile-time sizes. 0 has absolute priority, followed by 1, -// followed by Dynamic, followed by other finite values. The reason for giving Dynamic the priority over -// finite values is that min(3, Dynamic) should be Dynamic, since that could be anything between 0 and 3. -#define EIGEN_SIZE_MIN_PREFER_DYNAMIC(a,b) (((int)a == 0 || (int)b == 0) ? 0 \ - : ((int)a == 1 || (int)b == 1) ? 1 \ - : ((int)a == Dynamic || (int)b == Dynamic) ? Dynamic \ - : ((int)a <= (int)b) ? (int)a : (int)b) - -// EIGEN_SIZE_MIN_PREFER_FIXED is a variant of EIGEN_SIZE_MIN_PREFER_DYNAMIC comparing MaxSizes. The difference is that finite values -// now have priority over Dynamic, so that min(3, Dynamic) gives 3. Indeed, whatever the actual value is -// (between 0 and 3), it is not more than 3. -#define EIGEN_SIZE_MIN_PREFER_FIXED(a,b) (((int)a == 0 || (int)b == 0) ? 0 \ - : ((int)a == 1 || (int)b == 1) ? 1 \ - : ((int)a == Dynamic && (int)b == Dynamic) ? Dynamic \ - : ((int)a == Dynamic) ? (int)b \ - : ((int)b == Dynamic) ? (int)a \ - : ((int)a <= (int)b) ? (int)a : (int)b) - -// see EIGEN_SIZE_MIN_PREFER_DYNAMIC. No need for a separate variant for MaxSizes here. -#define EIGEN_SIZE_MAX(a,b) (((int)a == Dynamic || (int)b == Dynamic) ? Dynamic \ - : ((int)a >= (int)b) ? (int)a : (int)b) - -#define EIGEN_LOGICAL_XOR(a,b) (((a) || (b)) && !((a) && (b))) - -#define EIGEN_IMPLIES(a,b) (!(a) || (b)) - -// the expression type of a standard coefficient wise binary operation -#define EIGEN_CWISE_BINARY_RETURN_TYPE(LHS,RHS,OPNAME) \ - CwiseBinaryOp< \ - EIGEN_CAT(EIGEN_CAT(internal::scalar_,OPNAME),_op)< \ - typename internal::traits::Scalar, \ - typename internal::traits::Scalar \ - >, \ - const LHS, \ - const RHS \ - > - -#define EIGEN_MAKE_CWISE_BINARY_OP(METHOD,OPNAME) \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,OPNAME) \ - (METHOD)(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ - { \ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,OPNAME)(derived(), other.derived()); \ - } - -#define EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,TYPEA,TYPEB) \ - (Eigen::internal::has_ReturnType > >::value) - -#define EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(EXPR,SCALAR,OPNAME) \ - CwiseBinaryOp::Scalar,SCALAR>, const EXPR, \ - const typename internal::plain_constant_type::type> - -#define EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(SCALAR,EXPR,OPNAME) \ - CwiseBinaryOp::Scalar>, \ - const typename internal::plain_constant_type::type, const EXPR> - -// Workaround for MSVC 2010 (see ML thread "patch with compile for for MSVC 2010") -#if EIGEN_COMP_MSVC_STRICT<=1600 -#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) typename internal::enable_if::type -#else -#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) X -#endif - -#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME) \ - template EIGEN_DEVICE_FUNC inline \ - EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename internal::promote_scalar_arg::type,OPNAME))\ - (METHOD)(const T& scalar) const { \ - typedef typename internal::promote_scalar_arg::type PromotedT; \ - return EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,PromotedT,OPNAME)(derived(), \ - typename internal::plain_constant_type::type(derived().rows(), derived().cols(), internal::scalar_constant_op(scalar))); \ - } - -#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(METHOD,OPNAME) \ - template EIGEN_DEVICE_FUNC inline friend \ - EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename internal::promote_scalar_arg::type,Derived,OPNAME)) \ - (METHOD)(const T& scalar, const StorageBaseType& matrix) { \ - typedef typename internal::promote_scalar_arg::type PromotedT; \ - return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(PromotedT,Derived,OPNAME)( \ - typename internal::plain_constant_type::type(matrix.derived().rows(), matrix.derived().cols(), internal::scalar_constant_op(scalar)), matrix.derived()); \ - } - -#define EIGEN_MAKE_SCALAR_BINARY_OP(METHOD,OPNAME) \ - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(METHOD,OPNAME) \ - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME) - - -#ifdef EIGEN_EXCEPTIONS -# define EIGEN_THROW_X(X) throw X -# define EIGEN_THROW throw -# define EIGEN_TRY try -# define EIGEN_CATCH(X) catch (X) -#else -# ifdef __CUDA_ARCH__ -# define EIGEN_THROW_X(X) asm("trap;") -# define EIGEN_THROW asm("trap;") -# else -# define EIGEN_THROW_X(X) std::abort() -# define EIGEN_THROW std::abort() -# endif -# define EIGEN_TRY if (true) -# define EIGEN_CATCH(X) else -#endif - - -#if EIGEN_HAS_CXX11_NOEXCEPT -# define EIGEN_INCLUDE_TYPE_TRAITS -# define EIGEN_NOEXCEPT noexcept -# define EIGEN_NOEXCEPT_IF(x) noexcept(x) -# define EIGEN_NO_THROW noexcept(true) -# define EIGEN_EXCEPTION_SPEC(X) noexcept(false) -#else -# define EIGEN_NOEXCEPT -# define EIGEN_NOEXCEPT_IF(x) -# define EIGEN_NO_THROW throw() -# if EIGEN_COMP_MSVC - // MSVC does not support exception specifications (warning C4290), - // and they are deprecated in c++11 anyway. -# define EIGEN_EXCEPTION_SPEC(X) throw() -# else -# define EIGEN_EXCEPTION_SPEC(X) throw(X) -# endif -#endif - -#endif // EIGEN_MACROS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h deleted file mode 100644 index 291383c581a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h +++ /dev/null @@ -1,993 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2008-2009 Benoit Jacob -// Copyright (C) 2009 Kenneth Riddile -// Copyright (C) 2010 Hauke Heibel -// Copyright (C) 2010 Thomas Capricelli -// Copyright (C) 2013 Pavel Holoborodko -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -/***************************************************************************** -*** Platform checks for aligned malloc functions *** -*****************************************************************************/ - -#ifndef EIGEN_MEMORY_H -#define EIGEN_MEMORY_H - -#ifndef EIGEN_MALLOC_ALREADY_ALIGNED - -// Try to determine automatically if malloc is already aligned. - -// On 64-bit systems, glibc's malloc returns 16-byte-aligned pointers, see: -// http://www.gnu.org/s/libc/manual/html_node/Aligned-Memory-Blocks.html -// This is true at least since glibc 2.8. -// This leaves the question how to detect 64-bit. According to this document, -// http://gcc.fyxm.net/summit/2003/Porting%20to%2064%20bit.pdf -// page 114, "[The] LP64 model [...] is used by all 64-bit UNIX ports" so it's indeed -// quite safe, at least within the context of glibc, to equate 64-bit with LP64. -#if defined(__GLIBC__) && ((__GLIBC__>=2 && __GLIBC_MINOR__ >= 8) || __GLIBC__>2) \ - && defined(__LP64__) && ! defined( __SANITIZE_ADDRESS__ ) && (EIGEN_DEFAULT_ALIGN_BYTES == 16) - #define EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED 0 -#endif - -// FreeBSD 6 seems to have 16-byte aligned malloc -// See http://svn.freebsd.org/viewvc/base/stable/6/lib/libc/stdlib/malloc.c?view=markup -// FreeBSD 7 seems to have 16-byte aligned malloc except on ARM and MIPS architectures -// See http://svn.freebsd.org/viewvc/base/stable/7/lib/libc/stdlib/malloc.c?view=markup -#if defined(__FreeBSD__) && !(EIGEN_ARCH_ARM || EIGEN_ARCH_MIPS) && (EIGEN_DEFAULT_ALIGN_BYTES == 16) - #define EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED 0 -#endif - -#if (EIGEN_OS_MAC && (EIGEN_DEFAULT_ALIGN_BYTES == 16)) \ - || (EIGEN_OS_WIN64 && (EIGEN_DEFAULT_ALIGN_BYTES == 16)) \ - || EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED \ - || EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED - #define EIGEN_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_MALLOC_ALREADY_ALIGNED 0 -#endif - -#endif - -namespace Eigen { - -namespace internal { - -EIGEN_DEVICE_FUNC -inline void throw_std_bad_alloc() -{ - #ifdef EIGEN_EXCEPTIONS - throw std::bad_alloc(); - #else - std::size_t huge = static_cast(-1); - ::operator new(huge); - #endif -} - -/***************************************************************************** -*** Implementation of handmade aligned functions *** -*****************************************************************************/ - -/* ----- Hand made implementations of aligned malloc/free and realloc ----- */ - -/** \internal Like malloc, but the returned pointer is guaranteed to be 16-byte aligned. - * Fast, but wastes 16 additional bytes of memory. Does not throw any exception. - */ -inline void* handmade_aligned_malloc(std::size_t size) -{ - void *original = std::malloc(size+EIGEN_DEFAULT_ALIGN_BYTES); - if (original == 0) return 0; - void *aligned = reinterpret_cast((reinterpret_cast(original) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) + EIGEN_DEFAULT_ALIGN_BYTES); - *(reinterpret_cast(aligned) - 1) = original; - return aligned; -} - -/** \internal Frees memory allocated with handmade_aligned_malloc */ -inline void handmade_aligned_free(void *ptr) -{ - if (ptr) std::free(*(reinterpret_cast(ptr) - 1)); -} - -/** \internal - * \brief Reallocates aligned memory. - * Since we know that our handmade version is based on std::malloc - * we can use std::realloc to implement efficient reallocation. - */ -inline void* handmade_aligned_realloc(void* ptr, std::size_t size, std::size_t = 0) -{ - if (ptr == 0) return handmade_aligned_malloc(size); - void *original = *(reinterpret_cast(ptr) - 1); - std::ptrdiff_t previous_offset = static_cast(ptr)-static_cast(original); - original = std::realloc(original,size+EIGEN_DEFAULT_ALIGN_BYTES); - if (original == 0) return 0; - void *aligned = reinterpret_cast((reinterpret_cast(original) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) + EIGEN_DEFAULT_ALIGN_BYTES); - void *previous_aligned = static_cast(original)+previous_offset; - if(aligned!=previous_aligned) - std::memmove(aligned, previous_aligned, size); - - *(reinterpret_cast(aligned) - 1) = original; - return aligned; -} - -/***************************************************************************** -*** Implementation of portable aligned versions of malloc/free/realloc *** -*****************************************************************************/ - -#ifdef EIGEN_NO_MALLOC -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{ - eigen_assert(false && "heap allocation is forbidden (EIGEN_NO_MALLOC is defined)"); -} -#elif defined EIGEN_RUNTIME_NO_MALLOC -EIGEN_DEVICE_FUNC inline bool is_malloc_allowed_impl(bool update, bool new_value = false) -{ - static bool value = true; - if (update == 1) - value = new_value; - return value; -} -EIGEN_DEVICE_FUNC inline bool is_malloc_allowed() { return is_malloc_allowed_impl(false); } -EIGEN_DEVICE_FUNC inline bool set_is_malloc_allowed(bool new_value) { return is_malloc_allowed_impl(true, new_value); } -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{ - eigen_assert(is_malloc_allowed() && "heap allocation is forbidden (EIGEN_RUNTIME_NO_MALLOC is defined and g_is_malloc_allowed is false)"); -} -#else -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{} -#endif - -/** \internal Allocates \a size bytes. The returned pointer is guaranteed to have 16 or 32 bytes alignment depending on the requirements. - * On allocation error, the returned pointer is null, and std::bad_alloc is thrown. - */ -EIGEN_DEVICE_FUNC inline void* aligned_malloc(std::size_t size) -{ - check_that_malloc_is_allowed(); - - void *result; - #if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - result = std::malloc(size); - #if EIGEN_DEFAULT_ALIGN_BYTES==16 - eigen_assert((size<16 || (std::size_t(result)%16)==0) && "System's malloc returned an unaligned pointer. Compile with EIGEN_MALLOC_ALREADY_ALIGNED=0 to fallback to handmade alignd memory allocator."); - #endif - #else - result = handmade_aligned_malloc(size); - #endif - - if(!result && size) - throw_std_bad_alloc(); - - return result; -} - -/** \internal Frees memory allocated with aligned_malloc. */ -EIGEN_DEVICE_FUNC inline void aligned_free(void *ptr) -{ - #if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - std::free(ptr); - #else - handmade_aligned_free(ptr); - #endif -} - -/** - * \internal - * \brief Reallocates an aligned block of memory. - * \throws std::bad_alloc on allocation failure - */ -inline void* aligned_realloc(void *ptr, std::size_t new_size, std::size_t old_size) -{ - EIGEN_UNUSED_VARIABLE(old_size); - - void *result; -#if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - result = std::realloc(ptr,new_size); -#else - result = handmade_aligned_realloc(ptr,new_size,old_size); -#endif - - if (!result && new_size) - throw_std_bad_alloc(); - - return result; -} - -/***************************************************************************** -*** Implementation of conditionally aligned functions *** -*****************************************************************************/ - -/** \internal Allocates \a size bytes. If Align is true, then the returned ptr is 16-byte-aligned. - * On allocation error, the returned pointer is null, and a std::bad_alloc is thrown. - */ -template EIGEN_DEVICE_FUNC inline void* conditional_aligned_malloc(std::size_t size) -{ - return aligned_malloc(size); -} - -template<> EIGEN_DEVICE_FUNC inline void* conditional_aligned_malloc(std::size_t size) -{ - check_that_malloc_is_allowed(); - - void *result = std::malloc(size); - if(!result && size) - throw_std_bad_alloc(); - return result; -} - -/** \internal Frees memory allocated with conditional_aligned_malloc */ -template EIGEN_DEVICE_FUNC inline void conditional_aligned_free(void *ptr) -{ - aligned_free(ptr); -} - -template<> EIGEN_DEVICE_FUNC inline void conditional_aligned_free(void *ptr) -{ - std::free(ptr); -} - -template inline void* conditional_aligned_realloc(void* ptr, std::size_t new_size, std::size_t old_size) -{ - return aligned_realloc(ptr, new_size, old_size); -} - -template<> inline void* conditional_aligned_realloc(void* ptr, std::size_t new_size, std::size_t) -{ - return std::realloc(ptr, new_size); -} - -/***************************************************************************** -*** Construction/destruction of array elements *** -*****************************************************************************/ - -/** \internal Destructs the elements of an array. - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void destruct_elements_of_array(T *ptr, std::size_t size) -{ - // always destruct an array starting from the end. - if(ptr) - while(size) ptr[--size].~T(); -} - -/** \internal Constructs the elements of an array. - * The \a size parameter tells on how many objects to call the constructor of T. - */ -template EIGEN_DEVICE_FUNC inline T* construct_elements_of_array(T *ptr, std::size_t size) -{ - std::size_t i; - EIGEN_TRY - { - for (i = 0; i < size; ++i) ::new (ptr + i) T; - return ptr; - } - EIGEN_CATCH(...) - { - destruct_elements_of_array(ptr, i); - EIGEN_THROW; - } - return NULL; -} - -/***************************************************************************** -*** Implementation of aligned new/delete-like functions *** -*****************************************************************************/ - -template -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void check_size_for_overflow(std::size_t size) -{ - if(size > std::size_t(-1) / sizeof(T)) - throw_std_bad_alloc(); -} - -/** \internal Allocates \a size objects of type T. The returned pointer is guaranteed to have 16 bytes alignment. - * On allocation error, the returned pointer is undefined, but a std::bad_alloc is thrown. - * The default constructor of T is called. - */ -template EIGEN_DEVICE_FUNC inline T* aligned_new(std::size_t size) -{ - check_size_for_overflow(size); - T *result = reinterpret_cast(aligned_malloc(sizeof(T)*size)); - EIGEN_TRY - { - return construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - aligned_free(result); - EIGEN_THROW; - } - return result; -} - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_new(std::size_t size) -{ - check_size_for_overflow(size); - T *result = reinterpret_cast(conditional_aligned_malloc(sizeof(T)*size)); - EIGEN_TRY - { - return construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - return result; -} - -/** \internal Deletes objects constructed with aligned_new - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void aligned_delete(T *ptr, std::size_t size) -{ - destruct_elements_of_array(ptr, size); - aligned_free(ptr); -} - -/** \internal Deletes objects constructed with conditional_aligned_new - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void conditional_aligned_delete(T *ptr, std::size_t size) -{ - destruct_elements_of_array(ptr, size); - conditional_aligned_free(ptr); -} - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_realloc_new(T* pts, std::size_t new_size, std::size_t old_size) -{ - check_size_for_overflow(new_size); - check_size_for_overflow(old_size); - if(new_size < old_size) - destruct_elements_of_array(pts+new_size, old_size-new_size); - T *result = reinterpret_cast(conditional_aligned_realloc(reinterpret_cast(pts), sizeof(T)*new_size, sizeof(T)*old_size)); - if(new_size > old_size) - { - EIGEN_TRY - { - construct_elements_of_array(result+old_size, new_size-old_size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_new_auto(std::size_t size) -{ - if(size==0) - return 0; // short-cut. Also fixes Bug 884 - check_size_for_overflow(size); - T *result = reinterpret_cast(conditional_aligned_malloc(sizeof(T)*size)); - if(NumTraits::RequireInitialization) - { - EIGEN_TRY - { - construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - -template inline T* conditional_aligned_realloc_new_auto(T* pts, std::size_t new_size, std::size_t old_size) -{ - check_size_for_overflow(new_size); - check_size_for_overflow(old_size); - if(NumTraits::RequireInitialization && (new_size < old_size)) - destruct_elements_of_array(pts+new_size, old_size-new_size); - T *result = reinterpret_cast(conditional_aligned_realloc(reinterpret_cast(pts), sizeof(T)*new_size, sizeof(T)*old_size)); - if(NumTraits::RequireInitialization && (new_size > old_size)) - { - EIGEN_TRY - { - construct_elements_of_array(result+old_size, new_size-old_size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - -template EIGEN_DEVICE_FUNC inline void conditional_aligned_delete_auto(T *ptr, std::size_t size) -{ - if(NumTraits::RequireInitialization) - destruct_elements_of_array(ptr, size); - conditional_aligned_free(ptr); -} - -/****************************************************************************/ - -/** \internal Returns the index of the first element of the array that is well aligned with respect to the requested \a Alignment. - * - * \tparam Alignment requested alignment in Bytes. - * \param array the address of the start of the array - * \param size the size of the array - * - * \note If no element of the array is well aligned or the requested alignment is not a multiple of a scalar, - * the size of the array is returned. For example with SSE, the requested alignment is typically 16-bytes. If - * packet size for the given scalar type is 1, then everything is considered well-aligned. - * - * \note Otherwise, if the Alignment is larger that the scalar size, we rely on the assumptions that sizeof(Scalar) is a - * power of 2. On the other hand, we do not assume that the array address is a multiple of sizeof(Scalar), as that fails for - * example with Scalar=double on certain 32-bit platforms, see bug #79. - * - * There is also the variant first_aligned(const MatrixBase&) defined in DenseCoeffsBase.h. - * \sa first_default_aligned() - */ -template -EIGEN_DEVICE_FUNC inline Index first_aligned(const Scalar* array, Index size) -{ - const Index ScalarSize = sizeof(Scalar); - const Index AlignmentSize = Alignment / ScalarSize; - const Index AlignmentMask = AlignmentSize-1; - - if(AlignmentSize<=1) - { - // Either the requested alignment if smaller than a scalar, or it exactly match a 1 scalar - // so that all elements of the array have the same alignment. - return 0; - } - else if( (UIntPtr(array) & (sizeof(Scalar)-1)) || (Alignment%ScalarSize)!=0) - { - // The array is not aligned to the size of a single scalar, or the requested alignment is not a multiple of the scalar size. - // Consequently, no element of the array is well aligned. - return size; - } - else - { - Index first = (AlignmentSize - (Index((UIntPtr(array)/sizeof(Scalar))) & AlignmentMask)) & AlignmentMask; - return (first < size) ? first : size; - } -} - -/** \internal Returns the index of the first element of the array that is well aligned with respect the largest packet requirement. - * \sa first_aligned(Scalar*,Index) and first_default_aligned(DenseBase) */ -template -EIGEN_DEVICE_FUNC inline Index first_default_aligned(const Scalar* array, Index size) -{ - typedef typename packet_traits::type DefaultPacketType; - return first_aligned::alignment>(array, size); -} - -/** \internal Returns the smallest integer multiple of \a base and greater or equal to \a size - */ -template -inline Index first_multiple(Index size, Index base) -{ - return ((size+base-1)/base)*base; -} - -// std::copy is much slower than memcpy, so let's introduce a smart_copy which -// use memcpy on trivial types, i.e., on types that does not require an initialization ctor. -template struct smart_copy_helper; - -template EIGEN_DEVICE_FUNC void smart_copy(const T* start, const T* end, T* target) -{ - smart_copy_helper::RequireInitialization>::run(start, end, target); -} - -template struct smart_copy_helper { - EIGEN_DEVICE_FUNC static inline void run(const T* start, const T* end, T* target) - { - IntPtr size = IntPtr(end)-IntPtr(start); - if(size==0) return; - eigen_internal_assert(start!=0 && end!=0 && target!=0); - std::memcpy(target, start, size); - } -}; - -template struct smart_copy_helper { - EIGEN_DEVICE_FUNC static inline void run(const T* start, const T* end, T* target) - { std::copy(start, end, target); } -}; - -// intelligent memmove. falls back to std::memmove for POD types, uses std::copy otherwise. -template struct smart_memmove_helper; - -template void smart_memmove(const T* start, const T* end, T* target) -{ - smart_memmove_helper::RequireInitialization>::run(start, end, target); -} - -template struct smart_memmove_helper { - static inline void run(const T* start, const T* end, T* target) - { - IntPtr size = IntPtr(end)-IntPtr(start); - if(size==0) return; - eigen_internal_assert(start!=0 && end!=0 && target!=0); - std::memmove(target, start, size); - } -}; - -template struct smart_memmove_helper { - static inline void run(const T* start, const T* end, T* target) - { - if (UIntPtr(target) < UIntPtr(start)) - { - std::copy(start, end, target); - } - else - { - std::ptrdiff_t count = (std::ptrdiff_t(end)-std::ptrdiff_t(start)) / sizeof(T); - std::copy_backward(start, end, target + count); - } - } -}; - - -/***************************************************************************** -*** Implementation of runtime stack allocation (falling back to malloc) *** -*****************************************************************************/ - -// you can overwrite Eigen's default behavior regarding alloca by defining EIGEN_ALLOCA -// to the appropriate stack allocation function -#ifndef EIGEN_ALLOCA - #if EIGEN_OS_LINUX || EIGEN_OS_MAC || (defined alloca) - #define EIGEN_ALLOCA alloca - #elif EIGEN_COMP_MSVC - #define EIGEN_ALLOCA _alloca - #endif -#endif - -// This helper class construct the allocated memory, and takes care of destructing and freeing the handled data -// at destruction time. In practice this helper class is mainly useful to avoid memory leak in case of exceptions. -template class aligned_stack_memory_handler : noncopyable -{ - public: - /* Creates a stack_memory_handler responsible for the buffer \a ptr of size \a size. - * Note that \a ptr can be 0 regardless of the other parameters. - * This constructor takes care of constructing/initializing the elements of the buffer if required by the scalar type T (see NumTraits::RequireInitialization). - * In this case, the buffer elements will also be destructed when this handler will be destructed. - * Finally, if \a dealloc is true, then the pointer \a ptr is freed. - **/ - aligned_stack_memory_handler(T* ptr, std::size_t size, bool dealloc) - : m_ptr(ptr), m_size(size), m_deallocate(dealloc) - { - if(NumTraits::RequireInitialization && m_ptr) - Eigen::internal::construct_elements_of_array(m_ptr, size); - } - ~aligned_stack_memory_handler() - { - if(NumTraits::RequireInitialization && m_ptr) - Eigen::internal::destruct_elements_of_array(m_ptr, m_size); - if(m_deallocate) - Eigen::internal::aligned_free(m_ptr); - } - protected: - T* m_ptr; - std::size_t m_size; - bool m_deallocate; -}; - -template class scoped_array : noncopyable -{ - T* m_ptr; -public: - explicit scoped_array(std::ptrdiff_t size) - { - m_ptr = new T[size]; - } - ~scoped_array() - { - delete[] m_ptr; - } - T& operator[](std::ptrdiff_t i) { return m_ptr[i]; } - const T& operator[](std::ptrdiff_t i) const { return m_ptr[i]; } - T* &ptr() { return m_ptr; } - const T* ptr() const { return m_ptr; } - operator const T*() const { return m_ptr; } -}; - -template void swap(scoped_array &a,scoped_array &b) -{ - std::swap(a.ptr(),b.ptr()); -} - -} // end namespace internal - -/** \internal - * Declares, allocates and construct an aligned buffer named NAME of SIZE elements of type TYPE on the stack - * if SIZE is smaller than EIGEN_STACK_ALLOCATION_LIMIT, and if stack allocation is supported by the platform - * (currently, this is Linux and Visual Studio only). Otherwise the memory is allocated on the heap. - * The allocated buffer is automatically deleted when exiting the scope of this declaration. - * If BUFFER is non null, then the declared variable is simply an alias for BUFFER, and no allocation/deletion occurs. - * Here is an example: - * \code - * { - * ei_declare_aligned_stack_constructed_variable(float,data,size,0); - * // use data[0] to data[size-1] - * } - * \endcode - * The underlying stack allocation function can controlled with the EIGEN_ALLOCA preprocessor token. - */ -#ifdef EIGEN_ALLOCA - - #if EIGEN_DEFAULT_ALIGN_BYTES>0 - // We always manually re-align the result of EIGEN_ALLOCA. - // If alloca is already aligned, the compiler should be smart enough to optimize away the re-alignment. - #define EIGEN_ALIGNED_ALLOCA(SIZE) reinterpret_cast((internal::UIntPtr(EIGEN_ALLOCA(SIZE+EIGEN_DEFAULT_ALIGN_BYTES-1)) + EIGEN_DEFAULT_ALIGN_BYTES-1) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) - #else - #define EIGEN_ALIGNED_ALLOCA(SIZE) EIGEN_ALLOCA(SIZE) - #endif - - #define ei_declare_aligned_stack_constructed_variable(TYPE,NAME,SIZE,BUFFER) \ - Eigen::internal::check_size_for_overflow(SIZE); \ - TYPE* NAME = (BUFFER)!=0 ? (BUFFER) \ - : reinterpret_cast( \ - (sizeof(TYPE)*SIZE<=EIGEN_STACK_ALLOCATION_LIMIT) ? EIGEN_ALIGNED_ALLOCA(sizeof(TYPE)*SIZE) \ - : Eigen::internal::aligned_malloc(sizeof(TYPE)*SIZE) ); \ - Eigen::internal::aligned_stack_memory_handler EIGEN_CAT(NAME,_stack_memory_destructor)((BUFFER)==0 ? NAME : 0,SIZE,sizeof(TYPE)*SIZE>EIGEN_STACK_ALLOCATION_LIMIT) - -#else - - #define ei_declare_aligned_stack_constructed_variable(TYPE,NAME,SIZE,BUFFER) \ - Eigen::internal::check_size_for_overflow(SIZE); \ - TYPE* NAME = (BUFFER)!=0 ? BUFFER : reinterpret_cast(Eigen::internal::aligned_malloc(sizeof(TYPE)*SIZE)); \ - Eigen::internal::aligned_stack_memory_handler EIGEN_CAT(NAME,_stack_memory_destructor)((BUFFER)==0 ? NAME : 0,SIZE,true) - -#endif - - -/***************************************************************************** -*** Implementation of EIGEN_MAKE_ALIGNED_OPERATOR_NEW [_IF] *** -*****************************************************************************/ - -#if EIGEN_MAX_ALIGN_BYTES!=0 - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_NOTHROW(NeedsToAlign) \ - void* operator new(std::size_t size, const std::nothrow_t&) EIGEN_NO_THROW { \ - EIGEN_TRY { return Eigen::internal::conditional_aligned_malloc(size); } \ - EIGEN_CATCH (...) { return 0; } \ - } - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) \ - void *operator new(std::size_t size) { \ - return Eigen::internal::conditional_aligned_malloc(size); \ - } \ - void *operator new[](std::size_t size) { \ - return Eigen::internal::conditional_aligned_malloc(size); \ - } \ - void operator delete(void * ptr) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete[](void * ptr) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete(void * ptr, std::size_t /* sz */) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete[](void * ptr, std::size_t /* sz */) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - /* in-place new and delete. since (at least afaik) there is no actual */ \ - /* memory allocated we can safely let the default implementation handle */ \ - /* this particular case. */ \ - static void *operator new(std::size_t size, void *ptr) { return ::operator new(size,ptr); } \ - static void *operator new[](std::size_t size, void* ptr) { return ::operator new[](size,ptr); } \ - void operator delete(void * memory, void *ptr) EIGEN_NO_THROW { return ::operator delete(memory,ptr); } \ - void operator delete[](void * memory, void *ptr) EIGEN_NO_THROW { return ::operator delete[](memory,ptr); } \ - /* nothrow-new (returns zero instead of std::bad_alloc) */ \ - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_NOTHROW(NeedsToAlign) \ - void operator delete(void *ptr, const std::nothrow_t&) EIGEN_NO_THROW { \ - Eigen::internal::conditional_aligned_free(ptr); \ - } \ - typedef void eigen_aligned_operator_new_marker_type; -#else - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) -#endif - -#define EIGEN_MAKE_ALIGNED_OPERATOR_NEW EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(true) -#define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(Scalar,Size) \ - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(bool(((Size)!=Eigen::Dynamic) && ((sizeof(Scalar)*(Size))%EIGEN_MAX_ALIGN_BYTES==0))) - -/****************************************************************************/ - -/** \class aligned_allocator -* \ingroup Core_Module -* -* \brief STL compatible allocator to use with types requiring a non standrad alignment. -* -* The memory is aligned as for dynamically aligned matrix/array types such as MatrixXd. -* By default, it will thus provide at least 16 bytes alignment and more in following cases: -* - 32 bytes alignment if AVX is enabled. -* - 64 bytes alignment if AVX512 is enabled. -* -* This can be controled using the \c EIGEN_MAX_ALIGN_BYTES macro as documented -* \link TopicPreprocessorDirectivesPerformance there \endlink. -* -* Example: -* \code -* // Matrix4f requires 16 bytes alignment: -* std::map< int, Matrix4f, std::less, -* aligned_allocator > > my_map_mat4; -* // Vector3f does not require 16 bytes alignment, no need to use Eigen's allocator: -* std::map< int, Vector3f > my_map_vec3; -* \endcode -* -* \sa \blank \ref TopicStlContainers. -*/ -template -class aligned_allocator : public std::allocator -{ -public: - typedef std::size_t size_type; - typedef std::ptrdiff_t difference_type; - typedef T* pointer; - typedef const T* const_pointer; - typedef T& reference; - typedef const T& const_reference; - typedef T value_type; - - template - struct rebind - { - typedef aligned_allocator other; - }; - - aligned_allocator() : std::allocator() {} - - aligned_allocator(const aligned_allocator& other) : std::allocator(other) {} - - template - aligned_allocator(const aligned_allocator& other) : std::allocator(other) {} - - ~aligned_allocator() {} - - pointer allocate(size_type num, const void* /*hint*/ = 0) - { - internal::check_size_for_overflow(num); - size_type size = num * sizeof(T); -#if EIGEN_COMP_GNUC_STRICT && EIGEN_GNUC_AT_LEAST(7,0) - // workaround gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87544 - // It triggered eigen/Eigen/src/Core/util/Memory.h:189:12: warning: argument 1 value '18446744073709551612' exceeds maximum object size 9223372036854775807 - if(size>=std::size_t((std::numeric_limits::max)())) - return 0; - else -#endif - return static_cast( internal::aligned_malloc(size) ); - } - - void deallocate(pointer p, size_type /*num*/) - { - internal::aligned_free(p); - } -}; - -//---------- Cache sizes ---------- - -#if !defined(EIGEN_NO_CPUID) -# if EIGEN_COMP_GNUC && EIGEN_ARCH_i386_OR_x86_64 -# if defined(__PIC__) && EIGEN_ARCH_i386 - // Case for x86 with PIC -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("xchgl %%ebx, %k1;cpuid; xchgl %%ebx,%k1": "=a" (abcd[0]), "=&r" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "a" (func), "c" (id)); -# elif defined(__PIC__) && EIGEN_ARCH_x86_64 - // Case for x64 with PIC. In theory this is only a problem with recent gcc and with medium or large code model, not with the default small code model. - // However, we cannot detect which code model is used, and the xchg overhead is negligible anyway. -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("xchg{q}\t{%%}rbx, %q1; cpuid; xchg{q}\t{%%}rbx, %q1": "=a" (abcd[0]), "=&r" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "0" (func), "2" (id)); -# else - // Case for x86_64 or x86 w/o PIC -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("cpuid": "=a" (abcd[0]), "=b" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "0" (func), "2" (id) ); -# endif -# elif EIGEN_COMP_MSVC -# if (EIGEN_COMP_MSVC > 1500) && EIGEN_ARCH_i386_OR_x86_64 -# define EIGEN_CPUID(abcd,func,id) __cpuidex((int*)abcd,func,id) -# endif -# endif -#endif - -namespace internal { - -#ifdef EIGEN_CPUID - -inline bool cpuid_is_vendor(int abcd[4], const int vendor[3]) -{ - return abcd[1]==vendor[0] && abcd[3]==vendor[1] && abcd[2]==vendor[2]; -} - -inline void queryCacheSizes_intel_direct(int& l1, int& l2, int& l3) -{ - int abcd[4]; - l1 = l2 = l3 = 0; - int cache_id = 0; - int cache_type = 0; - do { - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x4,cache_id); - cache_type = (abcd[0] & 0x0F) >> 0; - if(cache_type==1||cache_type==3) // data or unified cache - { - int cache_level = (abcd[0] & 0xE0) >> 5; // A[7:5] - int ways = (abcd[1] & 0xFFC00000) >> 22; // B[31:22] - int partitions = (abcd[1] & 0x003FF000) >> 12; // B[21:12] - int line_size = (abcd[1] & 0x00000FFF) >> 0; // B[11:0] - int sets = (abcd[2]); // C[31:0] - - int cache_size = (ways+1) * (partitions+1) * (line_size+1) * (sets+1); - - switch(cache_level) - { - case 1: l1 = cache_size; break; - case 2: l2 = cache_size; break; - case 3: l3 = cache_size; break; - default: break; - } - } - cache_id++; - } while(cache_type>0 && cache_id<16); -} - -inline void queryCacheSizes_intel_codes(int& l1, int& l2, int& l3) -{ - int abcd[4]; - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - l1 = l2 = l3 = 0; - EIGEN_CPUID(abcd,0x00000002,0); - unsigned char * bytes = reinterpret_cast(abcd)+2; - bool check_for_p2_core2 = false; - for(int i=0; i<14; ++i) - { - switch(bytes[i]) - { - case 0x0A: l1 = 8; break; // 0Ah data L1 cache, 8 KB, 2 ways, 32 byte lines - case 0x0C: l1 = 16; break; // 0Ch data L1 cache, 16 KB, 4 ways, 32 byte lines - case 0x0E: l1 = 24; break; // 0Eh data L1 cache, 24 KB, 6 ways, 64 byte lines - case 0x10: l1 = 16; break; // 10h data L1 cache, 16 KB, 4 ways, 32 byte lines (IA-64) - case 0x15: l1 = 16; break; // 15h code L1 cache, 16 KB, 4 ways, 32 byte lines (IA-64) - case 0x2C: l1 = 32; break; // 2Ch data L1 cache, 32 KB, 8 ways, 64 byte lines - case 0x30: l1 = 32; break; // 30h code L1 cache, 32 KB, 8 ways, 64 byte lines - case 0x60: l1 = 16; break; // 60h data L1 cache, 16 KB, 8 ways, 64 byte lines, sectored - case 0x66: l1 = 8; break; // 66h data L1 cache, 8 KB, 4 ways, 64 byte lines, sectored - case 0x67: l1 = 16; break; // 67h data L1 cache, 16 KB, 4 ways, 64 byte lines, sectored - case 0x68: l1 = 32; break; // 68h data L1 cache, 32 KB, 4 ways, 64 byte lines, sectored - case 0x1A: l2 = 96; break; // code and data L2 cache, 96 KB, 6 ways, 64 byte lines (IA-64) - case 0x22: l3 = 512; break; // code and data L3 cache, 512 KB, 4 ways (!), 64 byte lines, dual-sectored - case 0x23: l3 = 1024; break; // code and data L3 cache, 1024 KB, 8 ways, 64 byte lines, dual-sectored - case 0x25: l3 = 2048; break; // code and data L3 cache, 2048 KB, 8 ways, 64 byte lines, dual-sectored - case 0x29: l3 = 4096; break; // code and data L3 cache, 4096 KB, 8 ways, 64 byte lines, dual-sectored - case 0x39: l2 = 128; break; // code and data L2 cache, 128 KB, 4 ways, 64 byte lines, sectored - case 0x3A: l2 = 192; break; // code and data L2 cache, 192 KB, 6 ways, 64 byte lines, sectored - case 0x3B: l2 = 128; break; // code and data L2 cache, 128 KB, 2 ways, 64 byte lines, sectored - case 0x3C: l2 = 256; break; // code and data L2 cache, 256 KB, 4 ways, 64 byte lines, sectored - case 0x3D: l2 = 384; break; // code and data L2 cache, 384 KB, 6 ways, 64 byte lines, sectored - case 0x3E: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 64 byte lines, sectored - case 0x40: l2 = 0; break; // no integrated L2 cache (P6 core) or L3 cache (P4 core) - case 0x41: l2 = 128; break; // code and data L2 cache, 128 KB, 4 ways, 32 byte lines - case 0x42: l2 = 256; break; // code and data L2 cache, 256 KB, 4 ways, 32 byte lines - case 0x43: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 32 byte lines - case 0x44: l2 = 1024; break; // code and data L2 cache, 1024 KB, 4 ways, 32 byte lines - case 0x45: l2 = 2048; break; // code and data L2 cache, 2048 KB, 4 ways, 32 byte lines - case 0x46: l3 = 4096; break; // code and data L3 cache, 4096 KB, 4 ways, 64 byte lines - case 0x47: l3 = 8192; break; // code and data L3 cache, 8192 KB, 8 ways, 64 byte lines - case 0x48: l2 = 3072; break; // code and data L2 cache, 3072 KB, 12 ways, 64 byte lines - case 0x49: if(l2!=0) l3 = 4096; else {check_for_p2_core2=true; l3 = l2 = 4096;} break;// code and data L3 cache, 4096 KB, 16 ways, 64 byte lines (P4) or L2 for core2 - case 0x4A: l3 = 6144; break; // code and data L3 cache, 6144 KB, 12 ways, 64 byte lines - case 0x4B: l3 = 8192; break; // code and data L3 cache, 8192 KB, 16 ways, 64 byte lines - case 0x4C: l3 = 12288; break; // code and data L3 cache, 12288 KB, 12 ways, 64 byte lines - case 0x4D: l3 = 16384; break; // code and data L3 cache, 16384 KB, 16 ways, 64 byte lines - case 0x4E: l2 = 6144; break; // code and data L2 cache, 6144 KB, 24 ways, 64 byte lines - case 0x78: l2 = 1024; break; // code and data L2 cache, 1024 KB, 4 ways, 64 byte lines - case 0x79: l2 = 128; break; // code and data L2 cache, 128 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7A: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7B: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7C: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7D: l2 = 2048; break; // code and data L2 cache, 2048 KB, 8 ways, 64 byte lines - case 0x7E: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 128 byte lines, sect. (IA-64) - case 0x7F: l2 = 512; break; // code and data L2 cache, 512 KB, 2 ways, 64 byte lines - case 0x80: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 64 byte lines - case 0x81: l2 = 128; break; // code and data L2 cache, 128 KB, 8 ways, 32 byte lines - case 0x82: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 32 byte lines - case 0x83: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 32 byte lines - case 0x84: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 32 byte lines - case 0x85: l2 = 2048; break; // code and data L2 cache, 2048 KB, 8 ways, 32 byte lines - case 0x86: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 64 byte lines - case 0x87: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 64 byte lines - case 0x88: l3 = 2048; break; // code and data L3 cache, 2048 KB, 4 ways, 64 byte lines (IA-64) - case 0x89: l3 = 4096; break; // code and data L3 cache, 4096 KB, 4 ways, 64 byte lines (IA-64) - case 0x8A: l3 = 8192; break; // code and data L3 cache, 8192 KB, 4 ways, 64 byte lines (IA-64) - case 0x8D: l3 = 3072; break; // code and data L3 cache, 3072 KB, 12 ways, 128 byte lines (IA-64) - - default: break; - } - } - if(check_for_p2_core2 && l2 == l3) - l3 = 0; - l1 *= 1024; - l2 *= 1024; - l3 *= 1024; -} - -inline void queryCacheSizes_intel(int& l1, int& l2, int& l3, int max_std_funcs) -{ - if(max_std_funcs>=4) - queryCacheSizes_intel_direct(l1,l2,l3); - else - queryCacheSizes_intel_codes(l1,l2,l3); -} - -inline void queryCacheSizes_amd(int& l1, int& l2, int& l3) -{ - int abcd[4]; - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x80000005,0); - l1 = (abcd[2] >> 24) * 1024; // C[31:24] = L1 size in KB - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x80000006,0); - l2 = (abcd[2] >> 16) * 1024; // C[31;16] = l2 cache size in KB - l3 = ((abcd[3] & 0xFFFC000) >> 18) * 512 * 1024; // D[31;18] = l3 cache size in 512KB -} -#endif - -/** \internal - * Queries and returns the cache sizes in Bytes of the L1, L2, and L3 data caches respectively */ -inline void queryCacheSizes(int& l1, int& l2, int& l3) -{ - #ifdef EIGEN_CPUID - int abcd[4]; - const int GenuineIntel[] = {0x756e6547, 0x49656e69, 0x6c65746e}; - const int AuthenticAMD[] = {0x68747541, 0x69746e65, 0x444d4163}; - const int AMDisbetter_[] = {0x69444d41, 0x74656273, 0x21726574}; // "AMDisbetter!" - - // identify the CPU vendor - EIGEN_CPUID(abcd,0x0,0); - int max_std_funcs = abcd[1]; - if(cpuid_is_vendor(abcd,GenuineIntel)) - queryCacheSizes_intel(l1,l2,l3,max_std_funcs); - else if(cpuid_is_vendor(abcd,AuthenticAMD) || cpuid_is_vendor(abcd,AMDisbetter_)) - queryCacheSizes_amd(l1,l2,l3); - else - // by default let's use Intel's API - queryCacheSizes_intel(l1,l2,l3,max_std_funcs); - - // here is the list of other vendors: -// ||cpuid_is_vendor(abcd,"VIA VIA VIA ") -// ||cpuid_is_vendor(abcd,"CyrixInstead") -// ||cpuid_is_vendor(abcd,"CentaurHauls") -// ||cpuid_is_vendor(abcd,"GenuineTMx86") -// ||cpuid_is_vendor(abcd,"TransmetaCPU") -// ||cpuid_is_vendor(abcd,"RiseRiseRise") -// ||cpuid_is_vendor(abcd,"Geode by NSC") -// ||cpuid_is_vendor(abcd,"SiS SiS SiS ") -// ||cpuid_is_vendor(abcd,"UMC UMC UMC ") -// ||cpuid_is_vendor(abcd,"NexGenDriven") - #else - l1 = l2 = l3 = -1; - #endif -} - -/** \internal - * \returns the size in Bytes of the L1 data cache */ -inline int queryL1CacheSize() -{ - int l1(-1), l2, l3; - queryCacheSizes(l1,l2,l3); - return l1; -} - -/** \internal - * \returns the size in Bytes of the L2 or L3 cache if this later is present */ -inline int queryTopLevelCacheSize() -{ - int l1, l2(-1), l3(-1); - queryCacheSizes(l1,l2,l3); - return (std::max)(l2,l3); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MEMORY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h deleted file mode 100644 index 9b61ff037aa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h +++ /dev/null @@ -1,568 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_META_H -#define EIGEN_META_H - -#if defined(__CUDA_ARCH__) -#include -#include -#endif - -#if EIGEN_COMP_ICC>=1600 && __cplusplus >= 201103L -#include -#endif - -namespace Eigen { - -typedef EIGEN_DEFAULT_DENSE_INDEX_TYPE DenseIndex; - -/** - * \brief The Index type as used for the API. - * \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE. - * \sa \blank \ref TopicPreprocessorDirectives, StorageIndex. - */ - -typedef EIGEN_DEFAULT_DENSE_INDEX_TYPE Index; - -namespace internal { - -/** \internal - * \file Meta.h - * This file contains generic metaprogramming classes which are not specifically related to Eigen. - * \note In case you wonder, yes we're aware that Boost already provides all these features, - * we however don't want to add a dependency to Boost. - */ - -// Only recent versions of ICC complain about using ptrdiff_t to hold pointers, -// and older versions do not provide *intptr_t types. -#if EIGEN_COMP_ICC>=1600 && __cplusplus >= 201103L -typedef std::intptr_t IntPtr; -typedef std::uintptr_t UIntPtr; -#else -typedef std::ptrdiff_t IntPtr; -typedef std::size_t UIntPtr; -#endif - -struct true_type { enum { value = 1 }; }; -struct false_type { enum { value = 0 }; }; - -template -struct conditional { typedef Then type; }; - -template -struct conditional { typedef Else type; }; - -template struct is_same { enum { value = 0 }; }; -template struct is_same { enum { value = 1 }; }; - -template struct remove_reference { typedef T type; }; -template struct remove_reference { typedef T type; }; - -template struct remove_pointer { typedef T type; }; -template struct remove_pointer { typedef T type; }; -template struct remove_pointer { typedef T type; }; - -template struct remove_const { typedef T type; }; -template struct remove_const { typedef T type; }; -template struct remove_const { typedef T type[]; }; -template struct remove_const { typedef T type[Size]; }; - -template struct remove_all { typedef T type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; - -template struct is_arithmetic { enum { value = false }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic{ enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; - -#if EIGEN_HAS_CXX11 -using std::is_integral; -#else -template struct is_integral { enum { value = false }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -#if EIGEN_COMP_MSVC -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral{ enum { value = true }; }; -#endif -#endif - -#if EIGEN_HAS_CXX11 -using std::make_unsigned; -#else -// TODO: Possibly improve this implementation of make_unsigned. -// It is currently used only by -// template struct random_default_impl. -template struct make_unsigned; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned short type; }; -template<> struct make_unsigned { typedef unsigned short type; }; -template<> struct make_unsigned { typedef unsigned int type; }; -template<> struct make_unsigned { typedef unsigned int type; }; -template<> struct make_unsigned { typedef unsigned long type; }; -template<> struct make_unsigned { typedef unsigned long type; }; -#if EIGEN_COMP_MSVC -template<> struct make_unsigned { typedef unsigned __int64 type; }; -template<> struct make_unsigned { typedef unsigned __int64 type; }; -#endif -#endif - -template struct add_const { typedef const T type; }; -template struct add_const { typedef T& type; }; - -template struct is_const { enum { value = 0 }; }; -template struct is_const { enum { value = 1 }; }; - -template struct add_const_on_value_type { typedef const T type; }; -template struct add_const_on_value_type { typedef T const& type; }; -template struct add_const_on_value_type { typedef T const* type; }; -template struct add_const_on_value_type { typedef T const* const type; }; -template struct add_const_on_value_type { typedef T const* const type; }; - - -template -struct is_convertible_impl -{ -private: - struct any_conversion - { - template any_conversion(const volatile T&); - template any_conversion(T&); - }; - struct yes {int a[1];}; - struct no {int a[2];}; - - static yes test(const To&, int); - static no test(any_conversion, ...); - -public: - static From ms_from; -#ifdef __INTEL_COMPILER - #pragma warning push - #pragma warning ( disable : 2259 ) -#endif - enum { value = sizeof(test(ms_from, 0))==sizeof(yes) }; -#ifdef __INTEL_COMPILER - #pragma warning pop -#endif -}; - -template -struct is_convertible -{ - enum { value = is_convertible_impl::type, - typename remove_all::type>::value }; -}; - -/** \internal Allows to enable/disable an overload - * according to a compile time condition. - */ -template struct enable_if; - -template struct enable_if -{ typedef T type; }; - -#if defined(__CUDA_ARCH__) -#if !defined(__FLT_EPSILON__) -#define __FLT_EPSILON__ FLT_EPSILON -#define __DBL_EPSILON__ DBL_EPSILON -#endif - -namespace device { - -template struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static T epsilon() { return 0; } - static T (max)() { assert(false && "Highest not supported for this type"); } - static T (min)() { assert(false && "Lowest not supported for this type"); } - static T infinity() { assert(false && "Infinity not supported for this type"); } - static T quiet_NaN() { assert(false && "quiet_NaN not supported for this type"); } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static float epsilon() { return __FLT_EPSILON__; } - EIGEN_DEVICE_FUNC - static float (max)() { return CUDART_MAX_NORMAL_F; } - EIGEN_DEVICE_FUNC - static float (min)() { return FLT_MIN; } - EIGEN_DEVICE_FUNC - static float infinity() { return CUDART_INF_F; } - EIGEN_DEVICE_FUNC - static float quiet_NaN() { return CUDART_NAN_F; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static double epsilon() { return __DBL_EPSILON__; } - EIGEN_DEVICE_FUNC - static double (max)() { return DBL_MAX; } - EIGEN_DEVICE_FUNC - static double (min)() { return DBL_MIN; } - EIGEN_DEVICE_FUNC - static double infinity() { return CUDART_INF; } - EIGEN_DEVICE_FUNC - static double quiet_NaN() { return CUDART_NAN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static int epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static int (max)() { return INT_MAX; } - EIGEN_DEVICE_FUNC - static int (min)() { return INT_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned int epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned int (max)() { return UINT_MAX; } - EIGEN_DEVICE_FUNC - static unsigned int (min)() { return 0; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static long (max)() { return LONG_MAX; } - EIGEN_DEVICE_FUNC - static long (min)() { return LONG_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned long (max)() { return ULONG_MAX; } - EIGEN_DEVICE_FUNC - static unsigned long (min)() { return 0; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static long long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static long long (max)() { return LLONG_MAX; } - EIGEN_DEVICE_FUNC - static long long (min)() { return LLONG_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned long long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned long long (max)() { return ULLONG_MAX; } - EIGEN_DEVICE_FUNC - static unsigned long long (min)() { return 0; } -}; - -} - -#endif - -/** \internal - * A base class do disable default copy ctor and copy assignement operator. - */ -class noncopyable -{ - EIGEN_DEVICE_FUNC noncopyable(const noncopyable&); - EIGEN_DEVICE_FUNC const noncopyable& operator=(const noncopyable&); -protected: - EIGEN_DEVICE_FUNC noncopyable() {} - EIGEN_DEVICE_FUNC ~noncopyable() {} -}; - -/** \internal - * Convenient struct to get the result type of a unary or binary functor. - * - * It supports both the current STL mechanism (using the result_type member) as well as - * upcoming next STL generation (using a templated result member). - * If none of these members is provided, then the type of the first argument is returned. FIXME, that behavior is a pretty bad hack. - */ -#if EIGEN_HAS_STD_RESULT_OF -template struct result_of { - typedef typename std::result_of::type type1; - typedef typename remove_all::type type; -}; -#else -template struct result_of { }; - -struct has_none {int a[1];}; -struct has_std_result_type {int a[2];}; -struct has_tr1_result {int a[3];}; - -template -struct unary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct unary_result_of_select {typedef typename Func::result_type type;}; - -template -struct unary_result_of_select {typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename unary_result_of_select::type type; -}; - -template -struct binary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct binary_result_of_select -{typedef typename Func::result_type type;}; - -template -struct binary_result_of_select -{typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename binary_result_of_select::type type; -}; - -template -struct ternary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct ternary_result_of_select -{typedef typename Func::result_type type;}; - -template -struct ternary_result_of_select -{typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename ternary_result_of_select::type type; -}; -#endif - -struct meta_yes { char a[1]; }; -struct meta_no { char a[2]; }; - -// Check whether T::ReturnType does exist -template -struct has_ReturnType -{ - template static meta_yes testFunctor(typename C::ReturnType const *); - template static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(0)) == sizeof(meta_yes) }; -}; - -template const T* return_ptr(); - -template -struct has_nullary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()())>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -template -struct has_unary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()(IndexType(0)))>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -template -struct has_binary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()(IndexType(0),IndexType(0)))>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -/** \internal In short, it computes int(sqrt(\a Y)) with \a Y an integer. - * Usage example: \code meta_sqrt<1023>::ret \endcode - */ -template Y))) > - // use ?: instead of || just to shut up a stupid gcc 4.3 warning -class meta_sqrt -{ - enum { - MidX = (InfX+SupX)/2, - TakeInf = MidX*MidX > Y ? 1 : 0, - NewInf = int(TakeInf) ? InfX : int(MidX), - NewSup = int(TakeInf) ? int(MidX) : SupX - }; - public: - enum { ret = meta_sqrt::ret }; -}; - -template -class meta_sqrt { public: enum { ret = (SupX*SupX <= Y) ? SupX : InfX }; }; - - -/** \internal Computes the least common multiple of two positive integer A and B - * at compile-time. It implements a naive algorithm testing all multiples of A. - * It thus works better if A>=B. - */ -template -struct meta_least_common_multiple -{ - enum { ret = meta_least_common_multiple::ret }; -}; -template -struct meta_least_common_multiple -{ - enum { ret = A*K }; -}; - -/** \internal determines whether the product of two numeric types is allowed and what the return type is */ -template struct scalar_product_traits -{ - enum { Defined = 0 }; -}; - -// FIXME quick workaround around current limitation of result_of -// template -// struct result_of(ArgType0,ArgType1)> { -// typedef typename scalar_product_traits::type, typename remove_all::type>::ReturnType type; -// }; - -} // end namespace internal - -namespace numext { - -#if defined(__CUDA_ARCH__) -template EIGEN_DEVICE_FUNC void swap(T &a, T &b) { T tmp = b; b = a; a = tmp; } -#else -template EIGEN_STRONG_INLINE void swap(T &a, T &b) { std::swap(a,b); } -#endif - -#if defined(__CUDA_ARCH__) -using internal::device::numeric_limits; -#else -using std::numeric_limits; -#endif - -// Integer division with rounding up. -// T is assumed to be an integer type with a>=0, and b>0 -template -T div_ceil(const T &a, const T &b) -{ - return (a+b-1) / b; -} - -// The aim of the following functions is to bypass -Wfloat-equal warnings -// when we really want a strict equality comparison on floating points. -template EIGEN_STRONG_INLINE -bool equal_strict(const X& x,const Y& y) { return x == y; } - -template<> EIGEN_STRONG_INLINE -bool equal_strict(const float& x,const float& y) { return std::equal_to()(x,y); } - -template<> EIGEN_STRONG_INLINE -bool equal_strict(const double& x,const double& y) { return std::equal_to()(x,y); } - -template EIGEN_STRONG_INLINE -bool not_equal_strict(const X& x,const Y& y) { return x != y; } - -template<> EIGEN_STRONG_INLINE -bool not_equal_strict(const float& x,const float& y) { return std::not_equal_to()(x,y); } - -template<> EIGEN_STRONG_INLINE -bool not_equal_strict(const double& x,const double& y) { return std::not_equal_to()(x,y); } - -} // end namespace numext - -} // end namespace Eigen - -// Define portable (u)int{32,64} types -#if EIGEN_HAS_CXX11 -#include -namespace Eigen { -namespace numext { -typedef std::uint32_t uint32_t; -typedef std::int32_t int32_t; -typedef std::uint64_t uint64_t; -typedef std::int64_t int64_t; -} -} -#else -// Without c++11, all compilers able to compile Eigen also -// provides the C99 stdint.h header file. -#include -namespace Eigen { -namespace numext { -typedef ::uint32_t uint32_t; -typedef ::int32_t int32_t; -typedef ::uint64_t uint64_t; -typedef ::int64_t int64_t; -} -} -#endif - - -#endif // EIGEN_META_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/NonMPL2.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/NonMPL2.h deleted file mode 100644 index 1af67cf18c7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/NonMPL2.h +++ /dev/null @@ -1,3 +0,0 @@ -#ifdef EIGEN_MPL2_ONLY -#error Including non-MPL2 code in EIGEN_MPL2_ONLY mode -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h deleted file mode 100644 index 1ce6fd1b009..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h +++ /dev/null @@ -1,31 +0,0 @@ -#ifdef EIGEN_WARNINGS_DISABLED_2 -// "DisableStupidWarnings.h" was included twice recursively: Do not reenable warnings yet! -# undef EIGEN_WARNINGS_DISABLED_2 - -#elif defined(EIGEN_WARNINGS_DISABLED) -#undef EIGEN_WARNINGS_DISABLED - -#ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #ifdef _MSC_VER - #pragma warning( pop ) - #elif defined __INTEL_COMPILER - #pragma warning pop - #elif defined __clang__ - #pragma clang diagnostic pop - #elif defined __GNUC__ && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) - #pragma GCC diagnostic pop - #endif - - #if defined __NVCC__ -// Don't reenable the diagnostic messages, as it turns out these messages need -// to be disabled at the point of the template instantiation (i.e the user code) -// otherwise they'll be triggered by nvcc. -// #pragma diag_default code_is_unreachable -// #pragma diag_default initialization_not_reachable -// #pragma diag_default 2651 -// #pragma diag_default 2653 - #endif - -#endif - -#endif // EIGEN_WARNINGS_DISABLED diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h deleted file mode 100644 index 500e47792a4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h +++ /dev/null @@ -1,218 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STATIC_ASSERT_H -#define EIGEN_STATIC_ASSERT_H - -/* Some notes on Eigen's static assertion mechanism: - * - * - in EIGEN_STATIC_ASSERT(CONDITION,MSG) the parameter CONDITION must be a compile time boolean - * expression, and MSG an enum listed in struct internal::static_assertion - * - * - define EIGEN_NO_STATIC_ASSERT to disable them (and save compilation time) - * in that case, the static assertion is converted to the following runtime assert: - * eigen_assert(CONDITION && "MSG") - * - * - currently EIGEN_STATIC_ASSERT can only be used in function scope - * - */ - -#ifndef EIGEN_STATIC_ASSERT -#ifndef EIGEN_NO_STATIC_ASSERT - - #if EIGEN_MAX_CPP_VER>=11 && (__has_feature(cxx_static_assert) || (defined(__cplusplus) && __cplusplus >= 201103L) || (EIGEN_COMP_MSVC >= 1600)) - - // if native static_assert is enabled, let's use it - #define EIGEN_STATIC_ASSERT(X,MSG) static_assert(X,#MSG); - - #else // not CXX0X - - namespace Eigen { - - namespace internal { - - template - struct static_assertion {}; - - template<> - struct static_assertion - { - enum { - YOU_TRIED_CALLING_A_VECTOR_METHOD_ON_A_MATRIX=1, - YOU_MIXED_VECTORS_OF_DIFFERENT_SIZES=1, - YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES=1, - THIS_METHOD_IS_ONLY_FOR_VECTORS_OF_A_SPECIFIC_SIZE=1, - THIS_METHOD_IS_ONLY_FOR_MATRICES_OF_A_SPECIFIC_SIZE=1, - THIS_METHOD_IS_ONLY_FOR_OBJECTS_OF_A_SPECIFIC_SIZE=1, - OUT_OF_RANGE_ACCESS=1, - YOU_MADE_A_PROGRAMMING_MISTAKE=1, - EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT=1, - EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE=1, - YOU_CALLED_A_FIXED_SIZE_METHOD_ON_A_DYNAMIC_SIZE_MATRIX_OR_VECTOR=1, - YOU_CALLED_A_DYNAMIC_SIZE_METHOD_ON_A_FIXED_SIZE_MATRIX_OR_VECTOR=1, - UNALIGNED_LOAD_AND_STORE_OPERATIONS_UNIMPLEMENTED_ON_ALTIVEC=1, - THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES=1, - FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED=1, - NUMERIC_TYPE_MUST_BE_REAL=1, - COEFFICIENT_WRITE_ACCESS_TO_SELFADJOINT_NOT_SUPPORTED=1, - WRITING_TO_TRIANGULAR_PART_WITH_UNIT_DIAGONAL_IS_NOT_SUPPORTED=1, - THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE=1, - INVALID_MATRIX_PRODUCT=1, - INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS=1, - INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION=1, - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY=1, - THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES=1, - THIS_METHOD_IS_ONLY_FOR_ROW_MAJOR_MATRICES=1, - INVALID_MATRIX_TEMPLATE_PARAMETERS=1, - INVALID_MATRIXBASE_TEMPLATE_PARAMETERS=1, - BOTH_MATRICES_MUST_HAVE_THE_SAME_STORAGE_ORDER=1, - THIS_METHOD_IS_ONLY_FOR_DIAGONAL_MATRIX=1, - THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE=1, - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_WITH_DIRECT_MEMORY_ACCESS_SUCH_AS_MAP_OR_PLAIN_MATRICES=1, - YOU_ALREADY_SPECIFIED_THIS_STRIDE=1, - INVALID_STORAGE_ORDER_FOR_THIS_VECTOR_EXPRESSION=1, - THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD=1, - PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1=1, - THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS=1, - YOU_CANNOT_MIX_ARRAYS_AND_MATRICES=1, - YOU_PERFORMED_AN_INVALID_TRANSFORMATION_CONVERSION=1, - THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY=1, - YOU_ARE_TRYING_TO_USE_AN_INDEX_BASED_ACCESSOR_ON_AN_EXPRESSION_THAT_DOES_NOT_SUPPORT_THAT=1, - THIS_METHOD_IS_ONLY_FOR_1x1_EXPRESSIONS=1, - THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS=1, - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL=1, - THIS_METHOD_IS_ONLY_FOR_ARRAYS_NOT_MATRICES=1, - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED=1, - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED=1, - THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE=1, - THE_STORAGE_ORDER_OF_BOTH_SIDES_MUST_MATCH=1, - OBJECT_ALLOCATED_ON_STACK_IS_TOO_BIG=1, - IMPLICIT_CONVERSION_TO_SCALAR_IS_FOR_INNER_PRODUCT_ONLY=1, - STORAGE_LAYOUT_DOES_NOT_MATCH=1, - EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT__INVALID_COST_VALUE=1, - THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS=1, - MATRIX_FREE_CONJUGATE_GRADIENT_IS_COMPATIBLE_WITH_UPPER_UNION_LOWER_MODE_ONLY=1, - THIS_TYPE_IS_NOT_SUPPORTED=1, - STORAGE_KIND_MUST_MATCH=1, - STORAGE_INDEX_MUST_MATCH=1, - CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY=1, - SELFADJOINTVIEW_ACCEPTS_UPPER_AND_LOWER_MODE_ONLY=1 - }; - }; - - } // end namespace internal - - } // end namespace Eigen - - // Specialized implementation for MSVC to avoid "conditional - // expression is constant" warnings. This implementation doesn't - // appear to work under GCC, hence the multiple implementations. - #if EIGEN_COMP_MSVC - - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) \ - {Eigen::internal::static_assertion::MSG;} - - #else - // In some cases clang interprets bool(CONDITION) as function declaration - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) \ - if (Eigen::internal::static_assertion(CONDITION)>::MSG) {} - - #endif - - #endif // not CXX0X - -#else // EIGEN_NO_STATIC_ASSERT - - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) eigen_assert((CONDITION) && #MSG); - -#endif // EIGEN_NO_STATIC_ASSERT -#endif // EIGEN_STATIC_ASSERT - -// static assertion failing if the type \a TYPE is not a vector type -#define EIGEN_STATIC_ASSERT_VECTOR_ONLY(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::IsVectorAtCompileTime, \ - YOU_TRIED_CALLING_A_VECTOR_METHOD_ON_A_MATRIX) - -// static assertion failing if the type \a TYPE is not fixed-size -#define EIGEN_STATIC_ASSERT_FIXED_SIZE(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::SizeAtCompileTime!=Eigen::Dynamic, \ - YOU_CALLED_A_FIXED_SIZE_METHOD_ON_A_DYNAMIC_SIZE_MATRIX_OR_VECTOR) - -// static assertion failing if the type \a TYPE is not dynamic-size -#define EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::SizeAtCompileTime==Eigen::Dynamic, \ - YOU_CALLED_A_DYNAMIC_SIZE_METHOD_ON_A_FIXED_SIZE_MATRIX_OR_VECTOR) - -// static assertion failing if the type \a TYPE is not a vector type of the given size -#define EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(TYPE, SIZE) \ - EIGEN_STATIC_ASSERT(TYPE::IsVectorAtCompileTime && TYPE::SizeAtCompileTime==SIZE, \ - THIS_METHOD_IS_ONLY_FOR_VECTORS_OF_A_SPECIFIC_SIZE) - -// static assertion failing if the type \a TYPE is not a vector type of the given size -#define EIGEN_STATIC_ASSERT_MATRIX_SPECIFIC_SIZE(TYPE, ROWS, COLS) \ - EIGEN_STATIC_ASSERT(TYPE::RowsAtCompileTime==ROWS && TYPE::ColsAtCompileTime==COLS, \ - THIS_METHOD_IS_ONLY_FOR_MATRICES_OF_A_SPECIFIC_SIZE) - -// static assertion failing if the two vector expression types are not compatible (same fixed-size or dynamic size) -#define EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(TYPE0,TYPE1) \ - EIGEN_STATIC_ASSERT( \ - (int(TYPE0::SizeAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::SizeAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::SizeAtCompileTime)==int(TYPE1::SizeAtCompileTime)),\ - YOU_MIXED_VECTORS_OF_DIFFERENT_SIZES) - -#define EIGEN_PREDICATE_SAME_MATRIX_SIZE(TYPE0,TYPE1) \ - ( \ - (int(Eigen::internal::size_of_xpr_at_compile_time::ret)==0 && int(Eigen::internal::size_of_xpr_at_compile_time::ret)==0) \ - || (\ - (int(TYPE0::RowsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::RowsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::RowsAtCompileTime)==int(TYPE1::RowsAtCompileTime)) \ - && (int(TYPE0::ColsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::ColsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::ColsAtCompileTime)==int(TYPE1::ColsAtCompileTime))\ - ) \ - ) - -#define EIGEN_STATIC_ASSERT_NON_INTEGER(TYPE) \ - EIGEN_STATIC_ASSERT(!NumTraits::IsInteger, THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES) - - -// static assertion failing if it is guaranteed at compile-time that the two matrix expression types have different sizes -#define EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(TYPE0,TYPE1) \ - EIGEN_STATIC_ASSERT( \ - EIGEN_PREDICATE_SAME_MATRIX_SIZE(TYPE0,TYPE1),\ - YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES) - -#define EIGEN_STATIC_ASSERT_SIZE_1x1(TYPE) \ - EIGEN_STATIC_ASSERT((TYPE::RowsAtCompileTime == 1 || TYPE::RowsAtCompileTime == Dynamic) && \ - (TYPE::ColsAtCompileTime == 1 || TYPE::ColsAtCompileTime == Dynamic), \ - THIS_METHOD_IS_ONLY_FOR_1x1_EXPRESSIONS) - -#define EIGEN_STATIC_ASSERT_LVALUE(Derived) \ - EIGEN_STATIC_ASSERT(Eigen::internal::is_lvalue::value, \ - THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY) - -#define EIGEN_STATIC_ASSERT_ARRAYXPR(Derived) \ - EIGEN_STATIC_ASSERT((Eigen::internal::is_same::XprKind, ArrayXpr>::value), \ - THIS_METHOD_IS_ONLY_FOR_ARRAYS_NOT_MATRICES) - -#define EIGEN_STATIC_ASSERT_SAME_XPR_KIND(Derived1, Derived2) \ - EIGEN_STATIC_ASSERT((Eigen::internal::is_same::XprKind, \ - typename Eigen::internal::traits::XprKind \ - >::value), \ - YOU_CANNOT_MIX_ARRAYS_AND_MATRICES) - -// Check that a cost value is positive, and that is stay within a reasonable range -// TODO this check could be enabled for internal debugging only -#define EIGEN_INTERNAL_CHECK_COST_VALUE(C) \ - EIGEN_STATIC_ASSERT((C)>=0 && (C)<=HugeCost*HugeCost, EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT__INVALID_COST_VALUE); - -#endif // EIGEN_STATIC_ASSERT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h deleted file mode 100644 index 6bb4970828e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h +++ /dev/null @@ -1,838 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_XPRHELPER_H -#define EIGEN_XPRHELPER_H - -// just a workaround because GCC seems to not really like empty structs -// FIXME: gcc 4.3 generates bad code when strict-aliasing is enabled -// so currently we simply disable this optimization for gcc 4.3 -#if EIGEN_COMP_GNUC && !EIGEN_GNUC_AT(4,3) - #define EIGEN_EMPTY_STRUCT_CTOR(X) \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE X() {} \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE X(const X& ) {} -#else - #define EIGEN_EMPTY_STRUCT_CTOR(X) -#endif - -namespace Eigen { - -namespace internal { - -template -EIGEN_DEVICE_FUNC -inline IndexDest convert_index(const IndexSrc& idx) { - // for sizeof(IndexDest)>=sizeof(IndexSrc) compilers should be able to optimize this away: - eigen_internal_assert(idx <= NumTraits::highest() && "Index value to big for target type"); - return IndexDest(idx); -} - -// true if T can be considered as an integral index (i.e., and integral type or enum) -template struct is_valid_index_type -{ - enum { value = -#if EIGEN_HAS_TYPE_TRAITS - internal::is_integral::value || std::is_enum::value -#elif EIGEN_COMP_MSVC - internal::is_integral::value || __is_enum(T) -#else - // without C++11, we use is_convertible to Index instead of is_integral in order to treat enums as Index. - internal::is_convertible::value && !internal::is_same::value && !is_same::value -#endif - }; -}; - -// promote_scalar_arg is an helper used in operation between an expression and a scalar, like: -// expression * scalar -// Its role is to determine how the type T of the scalar operand should be promoted given the scalar type ExprScalar of the given expression. -// The IsSupported template parameter must be provided by the caller as: internal::has_ReturnType >::value using the proper order for ExprScalar and T. -// Then the logic is as follows: -// - if the operation is natively supported as defined by IsSupported, then the scalar type is not promoted, and T is returned. -// - otherwise, NumTraits::Literal is returned if T is implicitly convertible to NumTraits::Literal AND that this does not imply a float to integer conversion. -// - otherwise, ExprScalar is returned if T is implicitly convertible to ExprScalar AND that this does not imply a float to integer conversion. -// - In all other cases, the promoted type is not defined, and the respective operation is thus invalid and not available (SFINAE). -template -struct promote_scalar_arg; - -template -struct promote_scalar_arg -{ - typedef T type; -}; - -// Recursively check safe conversion to PromotedType, and then ExprScalar if they are different. -template::value, - bool IsSafe = NumTraits::IsInteger || !NumTraits::IsInteger> -struct promote_scalar_arg_unsupported; - -// Start recursion with NumTraits::Literal -template -struct promote_scalar_arg : promote_scalar_arg_unsupported::Literal> {}; - -// We found a match! -template -struct promote_scalar_arg_unsupported -{ - typedef PromotedType type; -}; - -// No match, but no real-to-integer issues, and ExprScalar and current PromotedType are different, -// so let's try to promote to ExprScalar -template -struct promote_scalar_arg_unsupported - : promote_scalar_arg_unsupported -{}; - -// Unsafe real-to-integer, let's stop. -template -struct promote_scalar_arg_unsupported {}; - -// T is not even convertible to ExprScalar, let's stop. -template -struct promote_scalar_arg_unsupported {}; - -//classes inheriting no_assignment_operator don't generate a default operator=. -class no_assignment_operator -{ - private: - no_assignment_operator& operator=(const no_assignment_operator&); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(no_assignment_operator) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(no_assignment_operator) -}; - -/** \internal return the index type with the largest number of bits */ -template -struct promote_index_type -{ - typedef typename conditional<(sizeof(I1)::type type; -}; - -/** \internal If the template parameter Value is Dynamic, this class is just a wrapper around a T variable that - * can be accessed using value() and setValue(). - * Otherwise, this class is an empty structure and value() just returns the template parameter Value. - */ -template class variable_if_dynamic -{ - public: - EIGEN_EMPTY_STRUCT_CTOR(variable_if_dynamic) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamic(T v) { EIGEN_ONLY_USED_FOR_DEBUG(v); eigen_assert(v == T(Value)); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE T value() { return T(Value); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T) {} -}; - -template class variable_if_dynamic -{ - T m_value; - EIGEN_DEVICE_FUNC variable_if_dynamic() { eigen_assert(false); } - public: - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamic(T value) : m_value(value) {} - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T value() const { return m_value; } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T value) { m_value = value; } -}; - -/** \internal like variable_if_dynamic but for DynamicIndex - */ -template class variable_if_dynamicindex -{ - public: - EIGEN_EMPTY_STRUCT_CTOR(variable_if_dynamicindex) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamicindex(T v) { EIGEN_ONLY_USED_FOR_DEBUG(v); eigen_assert(v == T(Value)); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE T value() { return T(Value); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T) {} -}; - -template class variable_if_dynamicindex -{ - T m_value; - EIGEN_DEVICE_FUNC variable_if_dynamicindex() { eigen_assert(false); } - public: - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamicindex(T value) : m_value(value) {} - EIGEN_DEVICE_FUNC T EIGEN_STRONG_INLINE value() const { return m_value; } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T value) { m_value = value; } -}; - -template struct functor_traits -{ - enum - { - Cost = 10, - PacketAccess = false, - IsRepeatable = false - }; -}; - -template struct packet_traits; - -template struct unpacket_traits -{ - typedef T type; - typedef T half; - enum - { - size = 1, - alignment = 1 - }; -}; - -template::size)==0 || is_same::half>::value> -struct find_best_packet_helper; - -template< int Size, typename PacketType> -struct find_best_packet_helper -{ - typedef PacketType type; -}; - -template -struct find_best_packet_helper -{ - typedef typename find_best_packet_helper::half>::type type; -}; - -template -struct find_best_packet -{ - typedef typename find_best_packet_helper::type>::type type; -}; - -#if EIGEN_MAX_STATIC_ALIGN_BYTES>0 -template -struct compute_default_alignment_helper -{ - enum { value = 0 }; -}; - -template -struct compute_default_alignment_helper // Match -{ - enum { value = AlignmentBytes }; -}; - -template -struct compute_default_alignment_helper // Try-half -{ - // current packet too large, try with an half-packet - enum { value = compute_default_alignment_helper::value }; -}; -#else -// If static alignment is disabled, no need to bother. -// This also avoids a division by zero in "bool Match = bool((ArrayBytes%AlignmentBytes)==0)" -template -struct compute_default_alignment_helper -{ - enum { value = 0 }; -}; -#endif - -template struct compute_default_alignment { - enum { value = compute_default_alignment_helper::value }; -}; - -template struct compute_default_alignment { - enum { value = EIGEN_MAX_ALIGN_BYTES }; -}; - -template class make_proper_matrix_type -{ - enum { - IsColVector = _Cols==1 && _Rows!=1, - IsRowVector = _Rows==1 && _Cols!=1, - Options = IsColVector ? (_Options | ColMajor) & ~RowMajor - : IsRowVector ? (_Options | RowMajor) & ~ColMajor - : _Options - }; - public: - typedef Matrix<_Scalar, _Rows, _Cols, Options, _MaxRows, _MaxCols> type; -}; - -template -class compute_matrix_flags -{ - enum { row_major_bit = Options&RowMajor ? RowMajorBit : 0 }; - public: - // FIXME currently we still have to handle DirectAccessBit at the expression level to handle DenseCoeffsBase<> - // and then propagate this information to the evaluator's flags. - // However, I (Gael) think that DirectAccessBit should only matter at the evaluation stage. - enum { ret = DirectAccessBit | LvalueBit | NestByRefBit | row_major_bit }; -}; - -template struct size_at_compile_time -{ - enum { ret = (_Rows==Dynamic || _Cols==Dynamic) ? Dynamic : _Rows * _Cols }; -}; - -template struct size_of_xpr_at_compile_time -{ - enum { ret = size_at_compile_time::RowsAtCompileTime,traits::ColsAtCompileTime>::ret }; -}; - -/* plain_matrix_type : the difference from eval is that plain_matrix_type is always a plain matrix type, - * whereas eval is a const reference in the case of a matrix - */ - -template::StorageKind> struct plain_matrix_type; -template struct plain_matrix_type_dense; -template struct plain_matrix_type -{ - typedef typename plain_matrix_type_dense::XprKind, traits::Flags>::type type; -}; -template struct plain_matrix_type -{ - typedef typename T::PlainObject type; -}; - -template struct plain_matrix_type_dense -{ - typedef Matrix::Scalar, - traits::RowsAtCompileTime, - traits::ColsAtCompileTime, - AutoAlign | (Flags&RowMajorBit ? RowMajor : ColMajor), - traits::MaxRowsAtCompileTime, - traits::MaxColsAtCompileTime - > type; -}; - -template struct plain_matrix_type_dense -{ - typedef Array::Scalar, - traits::RowsAtCompileTime, - traits::ColsAtCompileTime, - AutoAlign | (Flags&RowMajorBit ? RowMajor : ColMajor), - traits::MaxRowsAtCompileTime, - traits::MaxColsAtCompileTime - > type; -}; - -/* eval : the return type of eval(). For matrices, this is just a const reference - * in order to avoid a useless copy - */ - -template::StorageKind> struct eval; - -template struct eval -{ - typedef typename plain_matrix_type::type type; -// typedef typename T::PlainObject type; -// typedef T::Matrix::Scalar, -// traits::RowsAtCompileTime, -// traits::ColsAtCompileTime, -// AutoAlign | (traits::Flags&RowMajorBit ? RowMajor : ColMajor), -// traits::MaxRowsAtCompileTime, -// traits::MaxColsAtCompileTime -// > type; -}; - -template struct eval -{ - typedef typename plain_matrix_type::type type; -}; - -// for matrices, no need to evaluate, just use a const reference to avoid a useless copy -template -struct eval, Dense> -{ - typedef const Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& type; -}; - -template -struct eval, Dense> -{ - typedef const Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& type; -}; - - -/* similar to plain_matrix_type, but using the evaluator's Flags */ -template::StorageKind> struct plain_object_eval; - -template -struct plain_object_eval -{ - typedef typename plain_matrix_type_dense::XprKind, evaluator::Flags>::type type; -}; - - -/* plain_matrix_type_column_major : same as plain_matrix_type but guaranteed to be column-major - */ -template struct plain_matrix_type_column_major -{ - enum { Rows = traits::RowsAtCompileTime, - Cols = traits::ColsAtCompileTime, - MaxRows = traits::MaxRowsAtCompileTime, - MaxCols = traits::MaxColsAtCompileTime - }; - typedef Matrix::Scalar, - Rows, - Cols, - (MaxRows==1&&MaxCols!=1) ? RowMajor : ColMajor, - MaxRows, - MaxCols - > type; -}; - -/* plain_matrix_type_row_major : same as plain_matrix_type but guaranteed to be row-major - */ -template struct plain_matrix_type_row_major -{ - enum { Rows = traits::RowsAtCompileTime, - Cols = traits::ColsAtCompileTime, - MaxRows = traits::MaxRowsAtCompileTime, - MaxCols = traits::MaxColsAtCompileTime - }; - typedef Matrix::Scalar, - Rows, - Cols, - (MaxCols==1&&MaxRows!=1) ? RowMajor : ColMajor, - MaxRows, - MaxCols - > type; -}; - -/** \internal The reference selector for template expressions. The idea is that we don't - * need to use references for expressions since they are light weight proxy - * objects which should generate no copying overhead. */ -template -struct ref_selector -{ - typedef typename conditional< - bool(traits::Flags & NestByRefBit), - T const&, - const T - >::type type; - - typedef typename conditional< - bool(traits::Flags & NestByRefBit), - T &, - T - >::type non_const_type; -}; - -/** \internal Adds the const qualifier on the value-type of T2 if and only if T1 is a const type */ -template -struct transfer_constness -{ - typedef typename conditional< - bool(internal::is_const::value), - typename internal::add_const_on_value_type::type, - T2 - >::type type; -}; - - -// However, we still need a mechanism to detect whether an expression which is evaluated multiple time -// has to be evaluated into a temporary. -// That's the purpose of this new nested_eval helper: -/** \internal Determines how a given expression should be nested when evaluated multiple times. - * For example, when you do a * (b+c), Eigen will determine how the expression b+c should be - * evaluated into the bigger product expression. The choice is between nesting the expression b+c as-is, or - * evaluating that expression b+c into a temporary variable d, and nest d so that the resulting expression is - * a*d. Evaluating can be beneficial for example if every coefficient access in the resulting expression causes - * many coefficient accesses in the nested expressions -- as is the case with matrix product for example. - * - * \tparam T the type of the expression being nested. - * \tparam n the number of coefficient accesses in the nested expression for each coefficient access in the bigger expression. - * \tparam PlainObject the type of the temporary if needed. - */ -template::type> struct nested_eval -{ - enum { - ScalarReadCost = NumTraits::Scalar>::ReadCost, - CoeffReadCost = evaluator::CoeffReadCost, // NOTE What if an evaluator evaluate itself into a tempory? - // Then CoeffReadCost will be small (e.g., 1) but we still have to evaluate, especially if n>1. - // This situation is already taken care by the EvalBeforeNestingBit flag, which is turned ON - // for all evaluator creating a temporary. This flag is then propagated by the parent evaluators. - // Another solution could be to count the number of temps? - NAsInteger = n == Dynamic ? HugeCost : n, - CostEval = (NAsInteger+1) * ScalarReadCost + CoeffReadCost, - CostNoEval = NAsInteger * CoeffReadCost, - Evaluate = (int(evaluator::Flags) & EvalBeforeNestingBit) || (int(CostEval) < int(CostNoEval)) - }; - - typedef typename conditional::type>::type type; -}; - -template -EIGEN_DEVICE_FUNC -inline T* const_cast_ptr(const T* ptr) -{ - return const_cast(ptr); -} - -template::XprKind> -struct dense_xpr_base -{ - /* dense_xpr_base should only ever be used on dense expressions, thus falling either into the MatrixXpr or into the ArrayXpr cases */ -}; - -template -struct dense_xpr_base -{ - typedef MatrixBase type; -}; - -template -struct dense_xpr_base -{ - typedef ArrayBase type; -}; - -template::XprKind, typename StorageKind = typename traits::StorageKind> -struct generic_xpr_base; - -template -struct generic_xpr_base -{ - typedef typename dense_xpr_base::type type; -}; - -template struct cast_return_type -{ - typedef typename XprType::Scalar CurrentScalarType; - typedef typename remove_all::type _CastType; - typedef typename _CastType::Scalar NewScalarType; - typedef typename conditional::value, - const XprType&,CastType>::type type; -}; - -template struct promote_storage_type; - -template struct promote_storage_type -{ - typedef A ret; -}; -template struct promote_storage_type -{ - typedef A ret; -}; -template struct promote_storage_type -{ - typedef A ret; -}; - -/** \internal Specify the "storage kind" of applying a coefficient-wise - * binary operations between two expressions of kinds A and B respectively. - * The template parameter Functor permits to specialize the resulting storage kind wrt to - * the functor. - * The default rules are as follows: - * \code - * A op A -> A - * A op dense -> dense - * dense op B -> dense - * sparse op dense -> sparse - * dense op sparse -> sparse - * \endcode - */ -template struct cwise_promote_storage_type; - -template struct cwise_promote_storage_type { typedef A ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Sparse ret; }; -template struct cwise_promote_storage_type { typedef Sparse ret; }; - -template struct cwise_promote_storage_order { - enum { value = LhsOrder }; -}; - -template struct cwise_promote_storage_order { enum { value = RhsOrder }; }; -template struct cwise_promote_storage_order { enum { value = LhsOrder }; }; -template struct cwise_promote_storage_order { enum { value = Order }; }; - - -/** \internal Specify the "storage kind" of multiplying an expression of kind A with kind B. - * The template parameter ProductTag permits to specialize the resulting storage kind wrt to - * some compile-time properties of the product: GemmProduct, GemvProduct, OuterProduct, InnerProduct. - * The default rules are as follows: - * \code - * K * K -> K - * dense * K -> dense - * K * dense -> dense - * diag * K -> K - * K * diag -> K - * Perm * K -> K - * K * Perm -> K - * \endcode - */ -template struct product_promote_storage_type; - -template struct product_promote_storage_type { typedef A ret;}; -template struct product_promote_storage_type { typedef Dense ret;}; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -template struct product_promote_storage_type { typedef A ret; }; -template struct product_promote_storage_type { typedef B ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -template struct product_promote_storage_type { typedef A ret; }; -template struct product_promote_storage_type { typedef B ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -/** \internal gives the plain matrix or array type to store a row/column/diagonal of a matrix type. - * \tparam Scalar optional parameter allowing to pass a different scalar type than the one of the MatrixType. - */ -template -struct plain_row_type -{ - typedef Matrix MatrixRowType; - typedef Array ArrayRowType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixRowType, - ArrayRowType - >::type type; -}; - -template -struct plain_col_type -{ - typedef Matrix MatrixColType; - typedef Array ArrayColType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixColType, - ArrayColType - >::type type; -}; - -template -struct plain_diag_type -{ - enum { diag_size = EIGEN_SIZE_MIN_PREFER_DYNAMIC(ExpressionType::RowsAtCompileTime, ExpressionType::ColsAtCompileTime), - max_diag_size = EIGEN_SIZE_MIN_PREFER_FIXED(ExpressionType::MaxRowsAtCompileTime, ExpressionType::MaxColsAtCompileTime) - }; - typedef Matrix MatrixDiagType; - typedef Array ArrayDiagType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixDiagType, - ArrayDiagType - >::type type; -}; - -template -struct plain_constant_type -{ - enum { Options = (traits::Flags&RowMajorBit)?RowMajor:0 }; - - typedef Array::RowsAtCompileTime, traits::ColsAtCompileTime, - Options, traits::MaxRowsAtCompileTime,traits::MaxColsAtCompileTime> array_type; - - typedef Matrix::RowsAtCompileTime, traits::ColsAtCompileTime, - Options, traits::MaxRowsAtCompileTime,traits::MaxColsAtCompileTime> matrix_type; - - typedef CwiseNullaryOp, const typename conditional::XprKind, MatrixXpr >::value, matrix_type, array_type>::type > type; -}; - -template -struct is_lvalue -{ - enum { value = (!bool(is_const::value)) && - bool(traits::Flags & LvalueBit) }; -}; - -template struct is_diagonal -{ enum { ret = false }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct glue_shapes; -template<> struct glue_shapes { typedef TriangularShape type; }; - -template -bool is_same_dense(const T1 &mat1, const T2 &mat2, typename enable_if::ret&&has_direct_access::ret, T1>::type * = 0) -{ - return (mat1.data()==mat2.data()) && (mat1.innerStride()==mat2.innerStride()) && (mat1.outerStride()==mat2.outerStride()); -} - -template -bool is_same_dense(const T1 &, const T2 &, typename enable_if::ret&&has_direct_access::ret), T1>::type * = 0) -{ - return false; -} - -// Internal helper defining the cost of a scalar division for the type T. -// The default heuristic can be specialized for each scalar type and architecture. -template -struct scalar_div_cost { - enum { value = 8*NumTraits::MulCost }; -}; - -template -struct scalar_div_cost, Vectorized> { - enum { value = 2*scalar_div_cost::value - + 6*NumTraits::MulCost - + 3*NumTraits::AddCost - }; -}; - - -template -struct scalar_div_cost::type> { enum { value = 24 }; }; -template -struct scalar_div_cost::type> { enum { value = 21 }; }; - - -#ifdef EIGEN_DEBUG_ASSIGN -std::string demangle_traversal(int t) -{ - if(t==DefaultTraversal) return "DefaultTraversal"; - if(t==LinearTraversal) return "LinearTraversal"; - if(t==InnerVectorizedTraversal) return "InnerVectorizedTraversal"; - if(t==LinearVectorizedTraversal) return "LinearVectorizedTraversal"; - if(t==SliceVectorizedTraversal) return "SliceVectorizedTraversal"; - return "?"; -} -std::string demangle_unrolling(int t) -{ - if(t==NoUnrolling) return "NoUnrolling"; - if(t==InnerUnrolling) return "InnerUnrolling"; - if(t==CompleteUnrolling) return "CompleteUnrolling"; - return "?"; -} -std::string demangle_flags(int f) -{ - std::string res; - if(f&RowMajorBit) res += " | RowMajor"; - if(f&PacketAccessBit) res += " | Packet"; - if(f&LinearAccessBit) res += " | Linear"; - if(f&LvalueBit) res += " | Lvalue"; - if(f&DirectAccessBit) res += " | Direct"; - if(f&NestByRefBit) res += " | NestByRef"; - if(f&NoPreferredStorageOrderBit) res += " | NoPreferredStorageOrderBit"; - - return res; -} -#endif - -} // end namespace internal - - -/** \class ScalarBinaryOpTraits - * \ingroup Core_Module - * - * \brief Determines whether the given binary operation of two numeric types is allowed and what the scalar return type is. - * - * This class permits to control the scalar return type of any binary operation performed on two different scalar types through (partial) template specializations. - * - * For instance, let \c U1, \c U2 and \c U3 be three user defined scalar types for which most operations between instances of \c U1 and \c U2 returns an \c U3. - * You can let %Eigen knows that by defining: - \code - template - struct ScalarBinaryOpTraits { typedef U3 ReturnType; }; - template - struct ScalarBinaryOpTraits { typedef U3 ReturnType; }; - \endcode - * You can then explicitly disable some particular operations to get more explicit error messages: - \code - template<> - struct ScalarBinaryOpTraits > {}; - \endcode - * Or customize the return type for individual operation: - \code - template<> - struct ScalarBinaryOpTraits > { typedef U1 ReturnType; }; - \endcode - * - * By default, the following generic combinations are supported: - - - - - -
ScalarAScalarBBinaryOpReturnTypeNote
\c T \c T \c * \c T
\c NumTraits::Real \c T \c * \c T Only if \c NumTraits::IsComplex
\c T \c NumTraits::Real \c * \c T Only if \c NumTraits::IsComplex
- * - * \sa CwiseBinaryOp - */ -template > -struct ScalarBinaryOpTraits -#ifndef EIGEN_PARSED_BY_DOXYGEN - // for backward compatibility, use the hints given by the (deprecated) internal::scalar_product_traits class. - : internal::scalar_product_traits -#endif // EIGEN_PARSED_BY_DOXYGEN -{}; - -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -template -struct ScalarBinaryOpTraits::IsComplex,T>::type>::Real, BinaryOp> -{ - typedef T ReturnType; -}; -template -struct ScalarBinaryOpTraits::IsComplex,T>::type>::Real, T, BinaryOp> -{ - typedef T ReturnType; -}; - -// For Matrix * Permutation -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -// For Permutation * Matrix -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -// for Permutation*Permutation -template -struct ScalarBinaryOpTraits -{ - typedef void ReturnType; -}; - -// We require Lhs and Rhs to have "compatible" scalar types. -// It is tempting to always allow mixing different types but remember that this is often impossible in the vectorized paths. -// So allowing mixing different types gives very unexpected errors when enabling vectorization, when the user tries to -// add together a float matrix and a double matrix. -#define EIGEN_CHECK_BINARY_COMPATIBILIY(BINOP,LHS,RHS) \ - EIGEN_STATIC_ASSERT((Eigen::internal::has_ReturnType >::value), \ - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - -} // end namespace Eigen - -#endif // EIGEN_XPRHELPER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexEigenSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexEigenSolver.h deleted file mode 100644 index dc5fae06a3c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/ComplexEigenSolver.h +++ /dev/null @@ -1,346 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Claire Maurice -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2010,2012 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_EIGEN_SOLVER_H -#define EIGEN_COMPLEX_EIGEN_SOLVER_H - -#include "./ComplexSchur.h" - -namespace Eigen { - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class ComplexEigenSolver - * - * \brief Computes eigenvalues and eigenvectors of general complex matrices - * - * \tparam _MatrixType the type of the matrix of which we are - * computing the eigendecomposition; this is expected to be an - * instantiation of the Matrix class template. - * - * The eigenvalues and eigenvectors of a matrix \f$ A \f$ are scalars - * \f$ \lambda \f$ and vectors \f$ v \f$ such that \f$ Av = \lambda v - * \f$. If \f$ D \f$ is a diagonal matrix with the eigenvalues on - * the diagonal, and \f$ V \f$ is a matrix with the eigenvectors as - * its columns, then \f$ A V = V D \f$. The matrix \f$ V \f$ is - * almost always invertible, in which case we have \f$ A = V D V^{-1} - * \f$. This is called the eigendecomposition. - * - * The main function in this class is compute(), which computes the - * eigenvalues and eigenvectors of a given function. The - * documentation for that function contains an example showing the - * main features of the class. - * - * \sa class EigenSolver, class SelfAdjointEigenSolver - */ -template class ComplexEigenSolver -{ - public: - - /** \brief Synonym for the template parameter \p _MatrixType. */ - typedef _MatrixType MatrixType; - - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - /** \brief Scalar type for matrices of type #MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - /** \brief Complex scalar type for #MatrixType. - * - * This is \c std::complex if #Scalar is real (e.g., - * \c float or \c double) and just \c Scalar if #Scalar is - * complex. - */ - typedef std::complex ComplexScalar; - - /** \brief Type for vector of eigenvalues as returned by eigenvalues(). - * - * This is a column vector with entries of type #ComplexScalar. - * The length of the vector is the size of #MatrixType. - */ - typedef Matrix EigenvalueType; - - /** \brief Type for matrix of eigenvectors as returned by eigenvectors(). - * - * This is a square matrix with entries of type #ComplexScalar. - * The size is the same as the size of #MatrixType. - */ - typedef Matrix EigenvectorType; - - /** \brief Default constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). - */ - ComplexEigenSolver() - : m_eivec(), - m_eivalues(), - m_schur(), - m_isInitialized(false), - m_eigenvectorsOk(false), - m_matX() - {} - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa ComplexEigenSolver() - */ - explicit ComplexEigenSolver(Index size) - : m_eivec(size, size), - m_eivalues(size), - m_schur(size), - m_isInitialized(false), - m_eigenvectorsOk(false), - m_matX(size, size) - {} - - /** \brief Constructor; computes eigendecomposition of given matrix. - * - * \param[in] matrix Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are - * computed. - * - * This constructor calls compute() to compute the eigendecomposition. - */ - template - explicit ComplexEigenSolver(const EigenBase& matrix, bool computeEigenvectors = true) - : m_eivec(matrix.rows(),matrix.cols()), - m_eivalues(matrix.cols()), - m_schur(matrix.rows()), - m_isInitialized(false), - m_eigenvectorsOk(false), - m_matX(matrix.rows(),matrix.cols()) - { - compute(matrix.derived(), computeEigenvectors); - } - - /** \brief Returns the eigenvectors of given matrix. - * - * \returns A const reference to the matrix whose columns are the eigenvectors. - * - * \pre Either the constructor - * ComplexEigenSolver(const MatrixType& matrix, bool) or the member - * function compute(const MatrixType& matrix, bool) has been called before - * to compute the eigendecomposition of a matrix, and - * \p computeEigenvectors was set to true (the default). - * - * This function returns a matrix whose columns are the eigenvectors. Column - * \f$ k \f$ is an eigenvector corresponding to eigenvalue number \f$ k - * \f$ as returned by eigenvalues(). The eigenvectors are normalized to - * have (Euclidean) norm equal to one. The matrix returned by this - * function is the matrix \f$ V \f$ in the eigendecomposition \f$ A = V D - * V^{-1} \f$, if it exists. - * - * Example: \include ComplexEigenSolver_eigenvectors.cpp - * Output: \verbinclude ComplexEigenSolver_eigenvectors.out - */ - const EigenvectorType& eigenvectors() const - { - eigen_assert(m_isInitialized && "ComplexEigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - return m_eivec; - } - - /** \brief Returns the eigenvalues of given matrix. - * - * \returns A const reference to the column vector containing the eigenvalues. - * - * \pre Either the constructor - * ComplexEigenSolver(const MatrixType& matrix, bool) or the member - * function compute(const MatrixType& matrix, bool) has been called before - * to compute the eigendecomposition of a matrix. - * - * This function returns a column vector containing the - * eigenvalues. Eigenvalues are repeated according to their - * algebraic multiplicity, so there are as many eigenvalues as - * rows in the matrix. The eigenvalues are not sorted in any particular - * order. - * - * Example: \include ComplexEigenSolver_eigenvalues.cpp - * Output: \verbinclude ComplexEigenSolver_eigenvalues.out - */ - const EigenvalueType& eigenvalues() const - { - eigen_assert(m_isInitialized && "ComplexEigenSolver is not initialized."); - return m_eivalues; - } - - /** \brief Computes eigendecomposition of given matrix. - * - * \param[in] matrix Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are - * computed. - * \returns Reference to \c *this - * - * This function computes the eigenvalues of the complex matrix \p matrix. - * The eigenvalues() function can be used to retrieve them. If - * \p computeEigenvectors is true, then the eigenvectors are also computed - * and can be retrieved by calling eigenvectors(). - * - * The matrix is first reduced to Schur form using the - * ComplexSchur class. The Schur decomposition is then used to - * compute the eigenvalues and eigenvectors. - * - * The cost of the computation is dominated by the cost of the - * Schur decomposition, which is \f$ O(n^3) \f$ where \f$ n \f$ - * is the size of the matrix. - * - * Example: \include ComplexEigenSolver_compute.cpp - * Output: \verbinclude ComplexEigenSolver_compute.out - */ - template - ComplexEigenSolver& compute(const EigenBase& matrix, bool computeEigenvectors = true); - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, \c NoConvergence otherwise. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "ComplexEigenSolver is not initialized."); - return m_schur.info(); - } - - /** \brief Sets the maximum number of iterations allowed. */ - ComplexEigenSolver& setMaxIterations(Index maxIters) - { - m_schur.setMaxIterations(maxIters); - return *this; - } - - /** \brief Returns the maximum number of iterations. */ - Index getMaxIterations() - { - return m_schur.getMaxIterations(); - } - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - EigenvectorType m_eivec; - EigenvalueType m_eivalues; - ComplexSchur m_schur; - bool m_isInitialized; - bool m_eigenvectorsOk; - EigenvectorType m_matX; - - private: - void doComputeEigenvectors(RealScalar matrixnorm); - void sortEigenvalues(bool computeEigenvectors); -}; - - -template -template -ComplexEigenSolver& -ComplexEigenSolver::compute(const EigenBase& matrix, bool computeEigenvectors) -{ - check_template_parameters(); - - // this code is inspired from Jampack - eigen_assert(matrix.cols() == matrix.rows()); - - // Do a complex Schur decomposition, A = U T U^* - // The eigenvalues are on the diagonal of T. - m_schur.compute(matrix.derived(), computeEigenvectors); - - if(m_schur.info() == Success) - { - m_eivalues = m_schur.matrixT().diagonal(); - if(computeEigenvectors) - doComputeEigenvectors(m_schur.matrixT().norm()); - sortEigenvalues(computeEigenvectors); - } - - m_isInitialized = true; - m_eigenvectorsOk = computeEigenvectors; - return *this; -} - - -template -void ComplexEigenSolver::doComputeEigenvectors(RealScalar matrixnorm) -{ - const Index n = m_eivalues.size(); - - matrixnorm = numext::maxi(matrixnorm,(std::numeric_limits::min)()); - - // Compute X such that T = X D X^(-1), where D is the diagonal of T. - // The matrix X is unit triangular. - m_matX = EigenvectorType::Zero(n, n); - for(Index k=n-1 ; k>=0 ; k--) - { - m_matX.coeffRef(k,k) = ComplexScalar(1.0,0.0); - // Compute X(i,k) using the (i,k) entry of the equation X T = D X - for(Index i=k-1 ; i>=0 ; i--) - { - m_matX.coeffRef(i,k) = -m_schur.matrixT().coeff(i,k); - if(k-i-1>0) - m_matX.coeffRef(i,k) -= (m_schur.matrixT().row(i).segment(i+1,k-i-1) * m_matX.col(k).segment(i+1,k-i-1)).value(); - ComplexScalar z = m_schur.matrixT().coeff(i,i) - m_schur.matrixT().coeff(k,k); - if(z==ComplexScalar(0)) - { - // If the i-th and k-th eigenvalue are equal, then z equals 0. - // Use a small value instead, to prevent division by zero. - numext::real_ref(z) = NumTraits::epsilon() * matrixnorm; - } - m_matX.coeffRef(i,k) = m_matX.coeff(i,k) / z; - } - } - - // Compute V as V = U X; now A = U T U^* = U X D X^(-1) U^* = V D V^(-1) - m_eivec.noalias() = m_schur.matrixU() * m_matX; - // .. and normalize the eigenvectors - for(Index k=0 ; k -void ComplexEigenSolver::sortEigenvalues(bool computeEigenvectors) -{ - const Index n = m_eivalues.size(); - for (Index i=0; i -// Copyright (C) 2010,2012 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_SCHUR_H -#define EIGEN_COMPLEX_SCHUR_H - -#include "./HessenbergDecomposition.h" - -namespace Eigen { - -namespace internal { -template struct complex_schur_reduce_to_hessenberg; -} - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class ComplexSchur - * - * \brief Performs a complex Schur decomposition of a real or complex square matrix - * - * \tparam _MatrixType the type of the matrix of which we are - * computing the Schur decomposition; this is expected to be an - * instantiation of the Matrix class template. - * - * Given a real or complex square matrix A, this class computes the - * Schur decomposition: \f$ A = U T U^*\f$ where U is a unitary - * complex matrix, and T is a complex upper triangular matrix. The - * diagonal of the matrix T corresponds to the eigenvalues of the - * matrix A. - * - * Call the function compute() to compute the Schur decomposition of - * a given matrix. Alternatively, you can use the - * ComplexSchur(const MatrixType&, bool) constructor which computes - * the Schur decomposition at construction time. Once the - * decomposition is computed, you can use the matrixU() and matrixT() - * functions to retrieve the matrices U and V in the decomposition. - * - * \note This code is inspired from Jampack - * - * \sa class RealSchur, class EigenSolver, class ComplexEigenSolver - */ -template class ComplexSchur -{ - public: - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - /** \brief Scalar type for matrices of type \p _MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - /** \brief Complex scalar type for \p _MatrixType. - * - * This is \c std::complex if #Scalar is real (e.g., - * \c float or \c double) and just \c Scalar if #Scalar is - * complex. - */ - typedef std::complex ComplexScalar; - - /** \brief Type for the matrices in the Schur decomposition. - * - * This is a square matrix with entries of type #ComplexScalar. - * The size is the same as the size of \p _MatrixType. - */ - typedef Matrix ComplexMatrixType; - - /** \brief Default constructor. - * - * \param [in] size Positive integer, size of the matrix whose Schur decomposition will be computed. - * - * The default constructor is useful in cases in which the user - * intends to perform decompositions via compute(). The \p size - * parameter is only used as a hint. It is not an error to give a - * wrong \p size, but it may impair performance. - * - * \sa compute() for an example. - */ - explicit ComplexSchur(Index size = RowsAtCompileTime==Dynamic ? 1 : RowsAtCompileTime) - : m_matT(size,size), - m_matU(size,size), - m_hess(size), - m_isInitialized(false), - m_matUisUptodate(false), - m_maxIters(-1) - {} - - /** \brief Constructor; computes Schur decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Schur decomposition is to be computed. - * \param[in] computeU If true, both T and U are computed; if false, only T is computed. - * - * This constructor calls compute() to compute the Schur decomposition. - * - * \sa matrixT() and matrixU() for examples. - */ - template - explicit ComplexSchur(const EigenBase& matrix, bool computeU = true) - : m_matT(matrix.rows(),matrix.cols()), - m_matU(matrix.rows(),matrix.cols()), - m_hess(matrix.rows()), - m_isInitialized(false), - m_matUisUptodate(false), - m_maxIters(-1) - { - compute(matrix.derived(), computeU); - } - - /** \brief Returns the unitary matrix in the Schur decomposition. - * - * \returns A const reference to the matrix U. - * - * It is assumed that either the constructor - * ComplexSchur(const MatrixType& matrix, bool computeU) or the - * member function compute(const MatrixType& matrix, bool computeU) - * has been called before to compute the Schur decomposition of a - * matrix, and that \p computeU was set to true (the default - * value). - * - * Example: \include ComplexSchur_matrixU.cpp - * Output: \verbinclude ComplexSchur_matrixU.out - */ - const ComplexMatrixType& matrixU() const - { - eigen_assert(m_isInitialized && "ComplexSchur is not initialized."); - eigen_assert(m_matUisUptodate && "The matrix U has not been computed during the ComplexSchur decomposition."); - return m_matU; - } - - /** \brief Returns the triangular matrix in the Schur decomposition. - * - * \returns A const reference to the matrix T. - * - * It is assumed that either the constructor - * ComplexSchur(const MatrixType& matrix, bool computeU) or the - * member function compute(const MatrixType& matrix, bool computeU) - * has been called before to compute the Schur decomposition of a - * matrix. - * - * Note that this function returns a plain square matrix. If you want to reference - * only the upper triangular part, use: - * \code schur.matrixT().triangularView() \endcode - * - * Example: \include ComplexSchur_matrixT.cpp - * Output: \verbinclude ComplexSchur_matrixT.out - */ - const ComplexMatrixType& matrixT() const - { - eigen_assert(m_isInitialized && "ComplexSchur is not initialized."); - return m_matT; - } - - /** \brief Computes Schur decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Schur decomposition is to be computed. - * \param[in] computeU If true, both T and U are computed; if false, only T is computed. - - * \returns Reference to \c *this - * - * The Schur decomposition is computed by first reducing the - * matrix to Hessenberg form using the class - * HessenbergDecomposition. The Hessenberg matrix is then reduced - * to triangular form by performing QR iterations with a single - * shift. The cost of computing the Schur decomposition depends - * on the number of iterations; as a rough guide, it may be taken - * on the number of iterations; as a rough guide, it may be taken - * to be \f$25n^3\f$ complex flops, or \f$10n^3\f$ complex flops - * if \a computeU is false. - * - * Example: \include ComplexSchur_compute.cpp - * Output: \verbinclude ComplexSchur_compute.out - * - * \sa compute(const MatrixType&, bool, Index) - */ - template - ComplexSchur& compute(const EigenBase& matrix, bool computeU = true); - - /** \brief Compute Schur decomposition from a given Hessenberg matrix - * \param[in] matrixH Matrix in Hessenberg form H - * \param[in] matrixQ orthogonal matrix Q that transform a matrix A to H : A = Q H Q^T - * \param computeU Computes the matriX U of the Schur vectors - * \return Reference to \c *this - * - * This routine assumes that the matrix is already reduced in Hessenberg form matrixH - * using either the class HessenbergDecomposition or another mean. - * It computes the upper quasi-triangular matrix T of the Schur decomposition of H - * When computeU is true, this routine computes the matrix U such that - * A = U T U^T = (QZ) T (QZ)^T = Q H Q^T where A is the initial matrix - * - * NOTE Q is referenced if computeU is true; so, if the initial orthogonal matrix - * is not available, the user should give an identity matrix (Q.setIdentity()) - * - * \sa compute(const MatrixType&, bool) - */ - template - ComplexSchur& computeFromHessenberg(const HessMatrixType& matrixH, const OrthMatrixType& matrixQ, bool computeU=true); - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, \c NoConvergence otherwise. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "ComplexSchur is not initialized."); - return m_info; - } - - /** \brief Sets the maximum number of iterations allowed. - * - * If not specified by the user, the maximum number of iterations is m_maxIterationsPerRow times the size - * of the matrix. - */ - ComplexSchur& setMaxIterations(Index maxIters) - { - m_maxIters = maxIters; - return *this; - } - - /** \brief Returns the maximum number of iterations. */ - Index getMaxIterations() - { - return m_maxIters; - } - - /** \brief Maximum number of iterations per row. - * - * If not otherwise specified, the maximum number of iterations is this number times the size of the - * matrix. It is currently set to 30. - */ - static const int m_maxIterationsPerRow = 30; - - protected: - ComplexMatrixType m_matT, m_matU; - HessenbergDecomposition m_hess; - ComputationInfo m_info; - bool m_isInitialized; - bool m_matUisUptodate; - Index m_maxIters; - - private: - bool subdiagonalEntryIsNeglegible(Index i); - ComplexScalar computeShift(Index iu, Index iter); - void reduceToTriangularForm(bool computeU); - friend struct internal::complex_schur_reduce_to_hessenberg::IsComplex>; -}; - -/** If m_matT(i+1,i) is neglegible in floating point arithmetic - * compared to m_matT(i,i) and m_matT(j,j), then set it to zero and - * return true, else return false. */ -template -inline bool ComplexSchur::subdiagonalEntryIsNeglegible(Index i) -{ - RealScalar d = numext::norm1(m_matT.coeff(i,i)) + numext::norm1(m_matT.coeff(i+1,i+1)); - RealScalar sd = numext::norm1(m_matT.coeff(i+1,i)); - if (internal::isMuchSmallerThan(sd, d, NumTraits::epsilon())) - { - m_matT.coeffRef(i+1,i) = ComplexScalar(0); - return true; - } - return false; -} - - -/** Compute the shift in the current QR iteration. */ -template -typename ComplexSchur::ComplexScalar ComplexSchur::computeShift(Index iu, Index iter) -{ - using std::abs; - if (iter == 10 || iter == 20) - { - // exceptional shift, taken from http://www.netlib.org/eispack/comqr.f - return abs(numext::real(m_matT.coeff(iu,iu-1))) + abs(numext::real(m_matT.coeff(iu-1,iu-2))); - } - - // compute the shift as one of the eigenvalues of t, the 2x2 - // diagonal block on the bottom of the active submatrix - Matrix t = m_matT.template block<2,2>(iu-1,iu-1); - RealScalar normt = t.cwiseAbs().sum(); - t /= normt; // the normalization by sf is to avoid under/overflow - - ComplexScalar b = t.coeff(0,1) * t.coeff(1,0); - ComplexScalar c = t.coeff(0,0) - t.coeff(1,1); - ComplexScalar disc = sqrt(c*c + RealScalar(4)*b); - ComplexScalar det = t.coeff(0,0) * t.coeff(1,1) - b; - ComplexScalar trace = t.coeff(0,0) + t.coeff(1,1); - ComplexScalar eival1 = (trace + disc) / RealScalar(2); - ComplexScalar eival2 = (trace - disc) / RealScalar(2); - RealScalar eival1_norm = numext::norm1(eival1); - RealScalar eival2_norm = numext::norm1(eival2); - // A division by zero can only occur if eival1==eival2==0. - // In this case, det==0, and all we have to do is checking that eival2_norm!=0 - if(eival1_norm > eival2_norm) - eival2 = det / eival1; - else if(eival2_norm!=RealScalar(0)) - eival1 = det / eival2; - - // choose the eigenvalue closest to the bottom entry of the diagonal - if(numext::norm1(eival1-t.coeff(1,1)) < numext::norm1(eival2-t.coeff(1,1))) - return normt * eival1; - else - return normt * eival2; -} - - -template -template -ComplexSchur& ComplexSchur::compute(const EigenBase& matrix, bool computeU) -{ - m_matUisUptodate = false; - eigen_assert(matrix.cols() == matrix.rows()); - - if(matrix.cols() == 1) - { - m_matT = matrix.derived().template cast(); - if(computeU) m_matU = ComplexMatrixType::Identity(1,1); - m_info = Success; - m_isInitialized = true; - m_matUisUptodate = computeU; - return *this; - } - - internal::complex_schur_reduce_to_hessenberg::IsComplex>::run(*this, matrix.derived(), computeU); - computeFromHessenberg(m_matT, m_matU, computeU); - return *this; -} - -template -template -ComplexSchur& ComplexSchur::computeFromHessenberg(const HessMatrixType& matrixH, const OrthMatrixType& matrixQ, bool computeU) -{ - m_matT = matrixH; - if(computeU) - m_matU = matrixQ; - reduceToTriangularForm(computeU); - return *this; -} -namespace internal { - -/* Reduce given matrix to Hessenberg form */ -template -struct complex_schur_reduce_to_hessenberg -{ - // this is the implementation for the case IsComplex = true - static void run(ComplexSchur& _this, const MatrixType& matrix, bool computeU) - { - _this.m_hess.compute(matrix); - _this.m_matT = _this.m_hess.matrixH(); - if(computeU) _this.m_matU = _this.m_hess.matrixQ(); - } -}; - -template -struct complex_schur_reduce_to_hessenberg -{ - static void run(ComplexSchur& _this, const MatrixType& matrix, bool computeU) - { - typedef typename ComplexSchur::ComplexScalar ComplexScalar; - - // Note: m_hess is over RealScalar; m_matT and m_matU is over ComplexScalar - _this.m_hess.compute(matrix); - _this.m_matT = _this.m_hess.matrixH().template cast(); - if(computeU) - { - // This may cause an allocation which seems to be avoidable - MatrixType Q = _this.m_hess.matrixQ(); - _this.m_matU = Q.template cast(); - } - } -}; - -} // end namespace internal - -// Reduce the Hessenberg matrix m_matT to triangular form by QR iteration. -template -void ComplexSchur::reduceToTriangularForm(bool computeU) -{ - Index maxIters = m_maxIters; - if (maxIters == -1) - maxIters = m_maxIterationsPerRow * m_matT.rows(); - - // The matrix m_matT is divided in three parts. - // Rows 0,...,il-1 are decoupled from the rest because m_matT(il,il-1) is zero. - // Rows il,...,iu is the part we are working on (the active submatrix). - // Rows iu+1,...,end are already brought in triangular form. - Index iu = m_matT.cols() - 1; - Index il; - Index iter = 0; // number of iterations we are working on the (iu,iu) element - Index totalIter = 0; // number of iterations for whole matrix - - while(true) - { - // find iu, the bottom row of the active submatrix - while(iu > 0) - { - if(!subdiagonalEntryIsNeglegible(iu-1)) break; - iter = 0; - --iu; - } - - // if iu is zero then we are done; the whole matrix is triangularized - if(iu==0) break; - - // if we spent too many iterations, we give up - iter++; - totalIter++; - if(totalIter > maxIters) break; - - // find il, the top row of the active submatrix - il = iu-1; - while(il > 0 && !subdiagonalEntryIsNeglegible(il-1)) - { - --il; - } - - /* perform the QR step using Givens rotations. The first rotation - creates a bulge; the (il+2,il) element becomes nonzero. This - bulge is chased down to the bottom of the active submatrix. */ - - ComplexScalar shift = computeShift(iu, iter); - JacobiRotation rot; - rot.makeGivens(m_matT.coeff(il,il) - shift, m_matT.coeff(il+1,il)); - m_matT.rightCols(m_matT.cols()-il).applyOnTheLeft(il, il+1, rot.adjoint()); - m_matT.topRows((std::min)(il+2,iu)+1).applyOnTheRight(il, il+1, rot); - if(computeU) m_matU.applyOnTheRight(il, il+1, rot); - - for(Index i=il+1 ; i template inline \ -ComplexSchur >& \ -ComplexSchur >::compute(const EigenBase& matrix, bool computeU) \ -{ \ - typedef Matrix MatrixType; \ - typedef MatrixType::RealScalar RealScalar; \ - typedef std::complex ComplexScalar; \ -\ - eigen_assert(matrix.cols() == matrix.rows()); \ -\ - m_matUisUptodate = false; \ - if(matrix.cols() == 1) \ - { \ - m_matT = matrix.derived().template cast(); \ - if(computeU) m_matU = ComplexMatrixType::Identity(1,1); \ - m_info = Success; \ - m_isInitialized = true; \ - m_matUisUptodate = computeU; \ - return *this; \ - } \ - lapack_int n = internal::convert_index(matrix.cols()), sdim, info; \ - lapack_int matrix_order = LAPACKE_COLROW; \ - char jobvs, sort='N'; \ - LAPACK_##LAPACKE_PREFIX_U##_SELECT1 select = 0; \ - jobvs = (computeU) ? 'V' : 'N'; \ - m_matU.resize(n, n); \ - lapack_int ldvs = internal::convert_index(m_matU.outerStride()); \ - m_matT = matrix; \ - lapack_int lda = internal::convert_index(m_matT.outerStride()); \ - Matrix w; \ - w.resize(n, 1);\ - info = LAPACKE_##LAPACKE_PREFIX##gees( matrix_order, jobvs, sort, select, n, (LAPACKE_TYPE*)m_matT.data(), lda, &sdim, (LAPACKE_TYPE*)w.data(), (LAPACKE_TYPE*)m_matU.data(), ldvs ); \ - if(info == 0) \ - m_info = Success; \ - else \ - m_info = NoConvergence; \ -\ - m_isInitialized = true; \ - m_matUisUptodate = computeU; \ - return *this; \ -\ -} - -EIGEN_LAPACKE_SCHUR_COMPLEX(dcomplex, lapack_complex_double, z, Z, ColMajor, LAPACK_COL_MAJOR) -EIGEN_LAPACKE_SCHUR_COMPLEX(scomplex, lapack_complex_float, c, C, ColMajor, LAPACK_COL_MAJOR) -EIGEN_LAPACKE_SCHUR_COMPLEX(dcomplex, lapack_complex_double, z, Z, RowMajor, LAPACK_ROW_MAJOR) -EIGEN_LAPACKE_SCHUR_COMPLEX(scomplex, lapack_complex_float, c, C, RowMajor, LAPACK_ROW_MAJOR) - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_SCHUR_LAPACKE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/EigenSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/EigenSolver.h deleted file mode 100644 index f205b185dee..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/EigenSolver.h +++ /dev/null @@ -1,622 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2010,2012 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_EIGENSOLVER_H -#define EIGEN_EIGENSOLVER_H - -#include "./RealSchur.h" - -namespace Eigen { - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class EigenSolver - * - * \brief Computes eigenvalues and eigenvectors of general matrices - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * eigendecomposition; this is expected to be an instantiation of the Matrix - * class template. Currently, only real matrices are supported. - * - * The eigenvalues and eigenvectors of a matrix \f$ A \f$ are scalars - * \f$ \lambda \f$ and vectors \f$ v \f$ such that \f$ Av = \lambda v \f$. If - * \f$ D \f$ is a diagonal matrix with the eigenvalues on the diagonal, and - * \f$ V \f$ is a matrix with the eigenvectors as its columns, then \f$ A V = - * V D \f$. The matrix \f$ V \f$ is almost always invertible, in which case we - * have \f$ A = V D V^{-1} \f$. This is called the eigendecomposition. - * - * The eigenvalues and eigenvectors of a matrix may be complex, even when the - * matrix is real. However, we can choose real matrices \f$ V \f$ and \f$ D - * \f$ satisfying \f$ A V = V D \f$, just like the eigendecomposition, if the - * matrix \f$ D \f$ is not required to be diagonal, but if it is allowed to - * have blocks of the form - * \f[ \begin{bmatrix} u & v \\ -v & u \end{bmatrix} \f] - * (where \f$ u \f$ and \f$ v \f$ are real numbers) on the diagonal. These - * blocks correspond to complex eigenvalue pairs \f$ u \pm iv \f$. We call - * this variant of the eigendecomposition the pseudo-eigendecomposition. - * - * Call the function compute() to compute the eigenvalues and eigenvectors of - * a given matrix. Alternatively, you can use the - * EigenSolver(const MatrixType&, bool) constructor which computes the - * eigenvalues and eigenvectors at construction time. Once the eigenvalue and - * eigenvectors are computed, they can be retrieved with the eigenvalues() and - * eigenvectors() functions. The pseudoEigenvalueMatrix() and - * pseudoEigenvectors() methods allow the construction of the - * pseudo-eigendecomposition. - * - * The documentation for EigenSolver(const MatrixType&, bool) contains an - * example of the typical use of this class. - * - * \note The implementation is adapted from - * JAMA (public domain). - * Their code is based on EISPACK. - * - * \sa MatrixBase::eigenvalues(), class ComplexEigenSolver, class SelfAdjointEigenSolver - */ -template class EigenSolver -{ - public: - - /** \brief Synonym for the template parameter \p _MatrixType. */ - typedef _MatrixType MatrixType; - - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - /** \brief Scalar type for matrices of type #MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - /** \brief Complex scalar type for #MatrixType. - * - * This is \c std::complex if #Scalar is real (e.g., - * \c float or \c double) and just \c Scalar if #Scalar is - * complex. - */ - typedef std::complex ComplexScalar; - - /** \brief Type for vector of eigenvalues as returned by eigenvalues(). - * - * This is a column vector with entries of type #ComplexScalar. - * The length of the vector is the size of #MatrixType. - */ - typedef Matrix EigenvalueType; - - /** \brief Type for matrix of eigenvectors as returned by eigenvectors(). - * - * This is a square matrix with entries of type #ComplexScalar. - * The size is the same as the size of #MatrixType. - */ - typedef Matrix EigenvectorsType; - - /** \brief Default constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via EigenSolver::compute(const MatrixType&, bool). - * - * \sa compute() for an example. - */ - EigenSolver() : m_eivec(), m_eivalues(), m_isInitialized(false), m_realSchur(), m_matT(), m_tmp() {} - - /** \brief Default constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa EigenSolver() - */ - explicit EigenSolver(Index size) - : m_eivec(size, size), - m_eivalues(size), - m_isInitialized(false), - m_eigenvectorsOk(false), - m_realSchur(size), - m_matT(size, size), - m_tmp(size) - {} - - /** \brief Constructor; computes eigendecomposition of given matrix. - * - * \param[in] matrix Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are - * computed. - * - * This constructor calls compute() to compute the eigenvalues - * and eigenvectors. - * - * Example: \include EigenSolver_EigenSolver_MatrixType.cpp - * Output: \verbinclude EigenSolver_EigenSolver_MatrixType.out - * - * \sa compute() - */ - template - explicit EigenSolver(const EigenBase& matrix, bool computeEigenvectors = true) - : m_eivec(matrix.rows(), matrix.cols()), - m_eivalues(matrix.cols()), - m_isInitialized(false), - m_eigenvectorsOk(false), - m_realSchur(matrix.cols()), - m_matT(matrix.rows(), matrix.cols()), - m_tmp(matrix.cols()) - { - compute(matrix.derived(), computeEigenvectors); - } - - /** \brief Returns the eigenvectors of given matrix. - * - * \returns %Matrix whose columns are the (possibly complex) eigenvectors. - * - * \pre Either the constructor - * EigenSolver(const MatrixType&,bool) or the member function - * compute(const MatrixType&, bool) has been called before, and - * \p computeEigenvectors was set to true (the default). - * - * Column \f$ k \f$ of the returned matrix is an eigenvector corresponding - * to eigenvalue number \f$ k \f$ as returned by eigenvalues(). The - * eigenvectors are normalized to have (Euclidean) norm equal to one. The - * matrix returned by this function is the matrix \f$ V \f$ in the - * eigendecomposition \f$ A = V D V^{-1} \f$, if it exists. - * - * Example: \include EigenSolver_eigenvectors.cpp - * Output: \verbinclude EigenSolver_eigenvectors.out - * - * \sa eigenvalues(), pseudoEigenvectors() - */ - EigenvectorsType eigenvectors() const; - - /** \brief Returns the pseudo-eigenvectors of given matrix. - * - * \returns Const reference to matrix whose columns are the pseudo-eigenvectors. - * - * \pre Either the constructor - * EigenSolver(const MatrixType&,bool) or the member function - * compute(const MatrixType&, bool) has been called before, and - * \p computeEigenvectors was set to true (the default). - * - * The real matrix \f$ V \f$ returned by this function and the - * block-diagonal matrix \f$ D \f$ returned by pseudoEigenvalueMatrix() - * satisfy \f$ AV = VD \f$. - * - * Example: \include EigenSolver_pseudoEigenvectors.cpp - * Output: \verbinclude EigenSolver_pseudoEigenvectors.out - * - * \sa pseudoEigenvalueMatrix(), eigenvectors() - */ - const MatrixType& pseudoEigenvectors() const - { - eigen_assert(m_isInitialized && "EigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - return m_eivec; - } - - /** \brief Returns the block-diagonal matrix in the pseudo-eigendecomposition. - * - * \returns A block-diagonal matrix. - * - * \pre Either the constructor - * EigenSolver(const MatrixType&,bool) or the member function - * compute(const MatrixType&, bool) has been called before. - * - * The matrix \f$ D \f$ returned by this function is real and - * block-diagonal. The blocks on the diagonal are either 1-by-1 or 2-by-2 - * blocks of the form - * \f$ \begin{bmatrix} u & v \\ -v & u \end{bmatrix} \f$. - * These blocks are not sorted in any particular order. - * The matrix \f$ D \f$ and the matrix \f$ V \f$ returned by - * pseudoEigenvectors() satisfy \f$ AV = VD \f$. - * - * \sa pseudoEigenvectors() for an example, eigenvalues() - */ - MatrixType pseudoEigenvalueMatrix() const; - - /** \brief Returns the eigenvalues of given matrix. - * - * \returns A const reference to the column vector containing the eigenvalues. - * - * \pre Either the constructor - * EigenSolver(const MatrixType&,bool) or the member function - * compute(const MatrixType&, bool) has been called before. - * - * The eigenvalues are repeated according to their algebraic multiplicity, - * so there are as many eigenvalues as rows in the matrix. The eigenvalues - * are not sorted in any particular order. - * - * Example: \include EigenSolver_eigenvalues.cpp - * Output: \verbinclude EigenSolver_eigenvalues.out - * - * \sa eigenvectors(), pseudoEigenvalueMatrix(), - * MatrixBase::eigenvalues() - */ - const EigenvalueType& eigenvalues() const - { - eigen_assert(m_isInitialized && "EigenSolver is not initialized."); - return m_eivalues; - } - - /** \brief Computes eigendecomposition of given matrix. - * - * \param[in] matrix Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are - * computed. - * \returns Reference to \c *this - * - * This function computes the eigenvalues of the real matrix \p matrix. - * The eigenvalues() function can be used to retrieve them. If - * \p computeEigenvectors is true, then the eigenvectors are also computed - * and can be retrieved by calling eigenvectors(). - * - * The matrix is first reduced to real Schur form using the RealSchur - * class. The Schur decomposition is then used to compute the eigenvalues - * and eigenvectors. - * - * The cost of the computation is dominated by the cost of the - * Schur decomposition, which is very approximately \f$ 25n^3 \f$ - * (where \f$ n \f$ is the size of the matrix) if \p computeEigenvectors - * is true, and \f$ 10n^3 \f$ if \p computeEigenvectors is false. - * - * This method reuses of the allocated data in the EigenSolver object. - * - * Example: \include EigenSolver_compute.cpp - * Output: \verbinclude EigenSolver_compute.out - */ - template - EigenSolver& compute(const EigenBase& matrix, bool computeEigenvectors = true); - - /** \returns NumericalIssue if the input contains INF or NaN values or overflow occured. Returns Success otherwise. */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "EigenSolver is not initialized."); - return m_info; - } - - /** \brief Sets the maximum number of iterations allowed. */ - EigenSolver& setMaxIterations(Index maxIters) - { - m_realSchur.setMaxIterations(maxIters); - return *this; - } - - /** \brief Returns the maximum number of iterations. */ - Index getMaxIterations() - { - return m_realSchur.getMaxIterations(); - } - - private: - void doComputeEigenvectors(); - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - EIGEN_STATIC_ASSERT(!NumTraits::IsComplex, NUMERIC_TYPE_MUST_BE_REAL); - } - - MatrixType m_eivec; - EigenvalueType m_eivalues; - bool m_isInitialized; - bool m_eigenvectorsOk; - ComputationInfo m_info; - RealSchur m_realSchur; - MatrixType m_matT; - - typedef Matrix ColumnVectorType; - ColumnVectorType m_tmp; -}; - -template -MatrixType EigenSolver::pseudoEigenvalueMatrix() const -{ - eigen_assert(m_isInitialized && "EigenSolver is not initialized."); - const RealScalar precision = RealScalar(2)*NumTraits::epsilon(); - Index n = m_eivalues.rows(); - MatrixType matD = MatrixType::Zero(n,n); - for (Index i=0; i(i,i) << numext::real(m_eivalues.coeff(i)), numext::imag(m_eivalues.coeff(i)), - -numext::imag(m_eivalues.coeff(i)), numext::real(m_eivalues.coeff(i)); - ++i; - } - } - return matD; -} - -template -typename EigenSolver::EigenvectorsType EigenSolver::eigenvectors() const -{ - eigen_assert(m_isInitialized && "EigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - const RealScalar precision = RealScalar(2)*NumTraits::epsilon(); - Index n = m_eivec.cols(); - EigenvectorsType matV(n,n); - for (Index j=0; j(); - matV.col(j).normalize(); - } - else - { - // we have a pair of complex eigen values - for (Index i=0; i -template -EigenSolver& -EigenSolver::compute(const EigenBase& matrix, bool computeEigenvectors) -{ - check_template_parameters(); - - using std::sqrt; - using std::abs; - using numext::isfinite; - eigen_assert(matrix.cols() == matrix.rows()); - - // Reduce to real Schur form. - m_realSchur.compute(matrix.derived(), computeEigenvectors); - - m_info = m_realSchur.info(); - - if (m_info == Success) - { - m_matT = m_realSchur.matrixT(); - if (computeEigenvectors) - m_eivec = m_realSchur.matrixU(); - - // Compute eigenvalues from matT - m_eivalues.resize(matrix.cols()); - Index i = 0; - while (i < matrix.cols()) - { - if (i == matrix.cols() - 1 || m_matT.coeff(i+1, i) == Scalar(0)) - { - m_eivalues.coeffRef(i) = m_matT.coeff(i, i); - if(!(isfinite)(m_eivalues.coeffRef(i))) - { - m_isInitialized = true; - m_eigenvectorsOk = false; - m_info = NumericalIssue; - return *this; - } - ++i; - } - else - { - Scalar p = Scalar(0.5) * (m_matT.coeff(i, i) - m_matT.coeff(i+1, i+1)); - Scalar z; - // Compute z = sqrt(abs(p * p + m_matT.coeff(i+1, i) * m_matT.coeff(i, i+1))); - // without overflow - { - Scalar t0 = m_matT.coeff(i+1, i); - Scalar t1 = m_matT.coeff(i, i+1); - Scalar maxval = numext::maxi(abs(p),numext::maxi(abs(t0),abs(t1))); - t0 /= maxval; - t1 /= maxval; - Scalar p0 = p/maxval; - z = maxval * sqrt(abs(p0 * p0 + t0 * t1)); - } - - m_eivalues.coeffRef(i) = ComplexScalar(m_matT.coeff(i+1, i+1) + p, z); - m_eivalues.coeffRef(i+1) = ComplexScalar(m_matT.coeff(i+1, i+1) + p, -z); - if(!((isfinite)(m_eivalues.coeffRef(i)) && (isfinite)(m_eivalues.coeffRef(i+1)))) - { - m_isInitialized = true; - m_eigenvectorsOk = false; - m_info = NumericalIssue; - return *this; - } - i += 2; - } - } - - // Compute eigenvectors. - if (computeEigenvectors) - doComputeEigenvectors(); - } - - m_isInitialized = true; - m_eigenvectorsOk = computeEigenvectors; - - return *this; -} - - -template -void EigenSolver::doComputeEigenvectors() -{ - using std::abs; - const Index size = m_eivec.cols(); - const Scalar eps = NumTraits::epsilon(); - - // inefficient! this is already computed in RealSchur - Scalar norm(0); - for (Index j = 0; j < size; ++j) - { - norm += m_matT.row(j).segment((std::max)(j-1,Index(0)), size-(std::max)(j-1,Index(0))).cwiseAbs().sum(); - } - - // Backsubstitute to find vectors of upper triangular form - if (norm == Scalar(0)) - { - return; - } - - for (Index n = size-1; n >= 0; n--) - { - Scalar p = m_eivalues.coeff(n).real(); - Scalar q = m_eivalues.coeff(n).imag(); - - // Scalar vector - if (q == Scalar(0)) - { - Scalar lastr(0), lastw(0); - Index l = n; - - m_matT.coeffRef(n,n) = Scalar(1); - for (Index i = n-1; i >= 0; i--) - { - Scalar w = m_matT.coeff(i,i) - p; - Scalar r = m_matT.row(i).segment(l,n-l+1).dot(m_matT.col(n).segment(l, n-l+1)); - - if (m_eivalues.coeff(i).imag() < Scalar(0)) - { - lastw = w; - lastr = r; - } - else - { - l = i; - if (m_eivalues.coeff(i).imag() == Scalar(0)) - { - if (w != Scalar(0)) - m_matT.coeffRef(i,n) = -r / w; - else - m_matT.coeffRef(i,n) = -r / (eps * norm); - } - else // Solve real equations - { - Scalar x = m_matT.coeff(i,i+1); - Scalar y = m_matT.coeff(i+1,i); - Scalar denom = (m_eivalues.coeff(i).real() - p) * (m_eivalues.coeff(i).real() - p) + m_eivalues.coeff(i).imag() * m_eivalues.coeff(i).imag(); - Scalar t = (x * lastr - lastw * r) / denom; - m_matT.coeffRef(i,n) = t; - if (abs(x) > abs(lastw)) - m_matT.coeffRef(i+1,n) = (-r - w * t) / x; - else - m_matT.coeffRef(i+1,n) = (-lastr - y * t) / lastw; - } - - // Overflow control - Scalar t = abs(m_matT.coeff(i,n)); - if ((eps * t) * t > Scalar(1)) - m_matT.col(n).tail(size-i) /= t; - } - } - } - else if (q < Scalar(0) && n > 0) // Complex vector - { - Scalar lastra(0), lastsa(0), lastw(0); - Index l = n-1; - - // Last vector component imaginary so matrix is triangular - if (abs(m_matT.coeff(n,n-1)) > abs(m_matT.coeff(n-1,n))) - { - m_matT.coeffRef(n-1,n-1) = q / m_matT.coeff(n,n-1); - m_matT.coeffRef(n-1,n) = -(m_matT.coeff(n,n) - p) / m_matT.coeff(n,n-1); - } - else - { - ComplexScalar cc = ComplexScalar(Scalar(0),-m_matT.coeff(n-1,n)) / ComplexScalar(m_matT.coeff(n-1,n-1)-p,q); - m_matT.coeffRef(n-1,n-1) = numext::real(cc); - m_matT.coeffRef(n-1,n) = numext::imag(cc); - } - m_matT.coeffRef(n,n-1) = Scalar(0); - m_matT.coeffRef(n,n) = Scalar(1); - for (Index i = n-2; i >= 0; i--) - { - Scalar ra = m_matT.row(i).segment(l, n-l+1).dot(m_matT.col(n-1).segment(l, n-l+1)); - Scalar sa = m_matT.row(i).segment(l, n-l+1).dot(m_matT.col(n).segment(l, n-l+1)); - Scalar w = m_matT.coeff(i,i) - p; - - if (m_eivalues.coeff(i).imag() < Scalar(0)) - { - lastw = w; - lastra = ra; - lastsa = sa; - } - else - { - l = i; - if (m_eivalues.coeff(i).imag() == RealScalar(0)) - { - ComplexScalar cc = ComplexScalar(-ra,-sa) / ComplexScalar(w,q); - m_matT.coeffRef(i,n-1) = numext::real(cc); - m_matT.coeffRef(i,n) = numext::imag(cc); - } - else - { - // Solve complex equations - Scalar x = m_matT.coeff(i,i+1); - Scalar y = m_matT.coeff(i+1,i); - Scalar vr = (m_eivalues.coeff(i).real() - p) * (m_eivalues.coeff(i).real() - p) + m_eivalues.coeff(i).imag() * m_eivalues.coeff(i).imag() - q * q; - Scalar vi = (m_eivalues.coeff(i).real() - p) * Scalar(2) * q; - if ((vr == Scalar(0)) && (vi == Scalar(0))) - vr = eps * norm * (abs(w) + abs(q) + abs(x) + abs(y) + abs(lastw)); - - ComplexScalar cc = ComplexScalar(x*lastra-lastw*ra+q*sa,x*lastsa-lastw*sa-q*ra) / ComplexScalar(vr,vi); - m_matT.coeffRef(i,n-1) = numext::real(cc); - m_matT.coeffRef(i,n) = numext::imag(cc); - if (abs(x) > (abs(lastw) + abs(q))) - { - m_matT.coeffRef(i+1,n-1) = (-ra - w * m_matT.coeff(i,n-1) + q * m_matT.coeff(i,n)) / x; - m_matT.coeffRef(i+1,n) = (-sa - w * m_matT.coeff(i,n) - q * m_matT.coeff(i,n-1)) / x; - } - else - { - cc = ComplexScalar(-lastra-y*m_matT.coeff(i,n-1),-lastsa-y*m_matT.coeff(i,n)) / ComplexScalar(lastw,q); - m_matT.coeffRef(i+1,n-1) = numext::real(cc); - m_matT.coeffRef(i+1,n) = numext::imag(cc); - } - } - - // Overflow control - Scalar t = numext::maxi(abs(m_matT.coeff(i,n-1)),abs(m_matT.coeff(i,n))); - if ((eps * t) * t > Scalar(1)) - m_matT.block(i, n-1, size-i, 2) /= t; - - } - } - - // We handled a pair of complex conjugate eigenvalues, so need to skip them both - n--; - } - else - { - eigen_assert(0 && "Internal bug in EigenSolver (INF or NaN has not been detected)"); // this should not happen - } - } - - // Back transformation to get eigenvectors of original matrix - for (Index j = size-1; j >= 0; j--) - { - m_tmp.noalias() = m_eivec.leftCols(j+1) * m_matT.col(j).segment(0, j+1); - m_eivec.col(j) = m_tmp; - } -} - -} // end namespace Eigen - -#endif // EIGEN_EIGENSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedEigenSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedEigenSolver.h deleted file mode 100644 index 87d789b3f4a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedEigenSolver.h +++ /dev/null @@ -1,418 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012-2016 Gael Guennebaud -// Copyright (C) 2010,2012 Jitse Niesen -// Copyright (C) 2016 Tobias Wood -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERALIZEDEIGENSOLVER_H -#define EIGEN_GENERALIZEDEIGENSOLVER_H - -#include "./RealQZ.h" - -namespace Eigen { - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class GeneralizedEigenSolver - * - * \brief Computes the generalized eigenvalues and eigenvectors of a pair of general matrices - * - * \tparam _MatrixType the type of the matrices of which we are computing the - * eigen-decomposition; this is expected to be an instantiation of the Matrix - * class template. Currently, only real matrices are supported. - * - * The generalized eigenvalues and eigenvectors of a matrix pair \f$ A \f$ and \f$ B \f$ are scalars - * \f$ \lambda \f$ and vectors \f$ v \f$ such that \f$ Av = \lambda Bv \f$. If - * \f$ D \f$ is a diagonal matrix with the eigenvalues on the diagonal, and - * \f$ V \f$ is a matrix with the eigenvectors as its columns, then \f$ A V = - * B V D \f$. The matrix \f$ V \f$ is almost always invertible, in which case we - * have \f$ A = B V D V^{-1} \f$. This is called the generalized eigen-decomposition. - * - * The generalized eigenvalues and eigenvectors of a matrix pair may be complex, even when the - * matrices are real. Moreover, the generalized eigenvalue might be infinite if the matrix B is - * singular. To workaround this difficulty, the eigenvalues are provided as a pair of complex \f$ \alpha \f$ - * and real \f$ \beta \f$ such that: \f$ \lambda_i = \alpha_i / \beta_i \f$. If \f$ \beta_i \f$ is (nearly) zero, - * then one can consider the well defined left eigenvalue \f$ \mu = \beta_i / \alpha_i\f$ such that: - * \f$ \mu_i A v_i = B v_i \f$, or even \f$ \mu_i u_i^T A = u_i^T B \f$ where \f$ u_i \f$ is - * called the left eigenvector. - * - * Call the function compute() to compute the generalized eigenvalues and eigenvectors of - * a given matrix pair. Alternatively, you can use the - * GeneralizedEigenSolver(const MatrixType&, const MatrixType&, bool) constructor which computes the - * eigenvalues and eigenvectors at construction time. Once the eigenvalue and - * eigenvectors are computed, they can be retrieved with the eigenvalues() and - * eigenvectors() functions. - * - * Here is an usage example of this class: - * Example: \include GeneralizedEigenSolver.cpp - * Output: \verbinclude GeneralizedEigenSolver.out - * - * \sa MatrixBase::eigenvalues(), class ComplexEigenSolver, class SelfAdjointEigenSolver - */ -template class GeneralizedEigenSolver -{ - public: - - /** \brief Synonym for the template parameter \p _MatrixType. */ - typedef _MatrixType MatrixType; - - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - /** \brief Scalar type for matrices of type #MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - /** \brief Complex scalar type for #MatrixType. - * - * This is \c std::complex if #Scalar is real (e.g., - * \c float or \c double) and just \c Scalar if #Scalar is - * complex. - */ - typedef std::complex ComplexScalar; - - /** \brief Type for vector of real scalar values eigenvalues as returned by betas(). - * - * This is a column vector with entries of type #Scalar. - * The length of the vector is the size of #MatrixType. - */ - typedef Matrix VectorType; - - /** \brief Type for vector of complex scalar values eigenvalues as returned by alphas(). - * - * This is a column vector with entries of type #ComplexScalar. - * The length of the vector is the size of #MatrixType. - */ - typedef Matrix ComplexVectorType; - - /** \brief Expression type for the eigenvalues as returned by eigenvalues(). - */ - typedef CwiseBinaryOp,ComplexVectorType,VectorType> EigenvalueType; - - /** \brief Type for matrix of eigenvectors as returned by eigenvectors(). - * - * This is a square matrix with entries of type #ComplexScalar. - * The size is the same as the size of #MatrixType. - */ - typedef Matrix EigenvectorsType; - - /** \brief Default constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via EigenSolver::compute(const MatrixType&, bool). - * - * \sa compute() for an example. - */ - GeneralizedEigenSolver() - : m_eivec(), - m_alphas(), - m_betas(), - m_valuesOkay(false), - m_vectorsOkay(false), - m_realQZ() - {} - - /** \brief Default constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa GeneralizedEigenSolver() - */ - explicit GeneralizedEigenSolver(Index size) - : m_eivec(size, size), - m_alphas(size), - m_betas(size), - m_valuesOkay(false), - m_vectorsOkay(false), - m_realQZ(size), - m_tmp(size) - {} - - /** \brief Constructor; computes the generalized eigendecomposition of given matrix pair. - * - * \param[in] A Square matrix whose eigendecomposition is to be computed. - * \param[in] B Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are computed. - * - * This constructor calls compute() to compute the generalized eigenvalues - * and eigenvectors. - * - * \sa compute() - */ - GeneralizedEigenSolver(const MatrixType& A, const MatrixType& B, bool computeEigenvectors = true) - : m_eivec(A.rows(), A.cols()), - m_alphas(A.cols()), - m_betas(A.cols()), - m_valuesOkay(false), - m_vectorsOkay(false), - m_realQZ(A.cols()), - m_tmp(A.cols()) - { - compute(A, B, computeEigenvectors); - } - - /* \brief Returns the computed generalized eigenvectors. - * - * \returns %Matrix whose columns are the (possibly complex) right eigenvectors. - * i.e. the eigenvectors that solve (A - l*B)x = 0. The ordering matches the eigenvalues. - * - * \pre Either the constructor - * GeneralizedEigenSolver(const MatrixType&,const MatrixType&, bool) or the member function - * compute(const MatrixType&, const MatrixType& bool) has been called before, and - * \p computeEigenvectors was set to true (the default). - * - * \sa eigenvalues() - */ - EigenvectorsType eigenvectors() const { - eigen_assert(m_vectorsOkay && "Eigenvectors for GeneralizedEigenSolver were not calculated."); - return m_eivec; - } - - /** \brief Returns an expression of the computed generalized eigenvalues. - * - * \returns An expression of the column vector containing the eigenvalues. - * - * It is a shortcut for \code this->alphas().cwiseQuotient(this->betas()); \endcode - * Not that betas might contain zeros. It is therefore not recommended to use this function, - * but rather directly deal with the alphas and betas vectors. - * - * \pre Either the constructor - * GeneralizedEigenSolver(const MatrixType&,const MatrixType&,bool) or the member function - * compute(const MatrixType&,const MatrixType&,bool) has been called before. - * - * The eigenvalues are repeated according to their algebraic multiplicity, - * so there are as many eigenvalues as rows in the matrix. The eigenvalues - * are not sorted in any particular order. - * - * \sa alphas(), betas(), eigenvectors() - */ - EigenvalueType eigenvalues() const - { - eigen_assert(m_valuesOkay && "GeneralizedEigenSolver is not initialized."); - return EigenvalueType(m_alphas,m_betas); - } - - /** \returns A const reference to the vectors containing the alpha values - * - * This vector permits to reconstruct the j-th eigenvalues as alphas(i)/betas(j). - * - * \sa betas(), eigenvalues() */ - ComplexVectorType alphas() const - { - eigen_assert(m_valuesOkay && "GeneralizedEigenSolver is not initialized."); - return m_alphas; - } - - /** \returns A const reference to the vectors containing the beta values - * - * This vector permits to reconstruct the j-th eigenvalues as alphas(i)/betas(j). - * - * \sa alphas(), eigenvalues() */ - VectorType betas() const - { - eigen_assert(m_valuesOkay && "GeneralizedEigenSolver is not initialized."); - return m_betas; - } - - /** \brief Computes generalized eigendecomposition of given matrix. - * - * \param[in] A Square matrix whose eigendecomposition is to be computed. - * \param[in] B Square matrix whose eigendecomposition is to be computed. - * \param[in] computeEigenvectors If true, both the eigenvectors and the - * eigenvalues are computed; if false, only the eigenvalues are - * computed. - * \returns Reference to \c *this - * - * This function computes the eigenvalues of the real matrix \p matrix. - * The eigenvalues() function can be used to retrieve them. If - * \p computeEigenvectors is true, then the eigenvectors are also computed - * and can be retrieved by calling eigenvectors(). - * - * The matrix is first reduced to real generalized Schur form using the RealQZ - * class. The generalized Schur decomposition is then used to compute the eigenvalues - * and eigenvectors. - * - * The cost of the computation is dominated by the cost of the - * generalized Schur decomposition. - * - * This method reuses of the allocated data in the GeneralizedEigenSolver object. - */ - GeneralizedEigenSolver& compute(const MatrixType& A, const MatrixType& B, bool computeEigenvectors = true); - - ComputationInfo info() const - { - eigen_assert(m_valuesOkay && "EigenSolver is not initialized."); - return m_realQZ.info(); - } - - /** Sets the maximal number of iterations allowed. - */ - GeneralizedEigenSolver& setMaxIterations(Index maxIters) - { - m_realQZ.setMaxIterations(maxIters); - return *this; - } - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - EIGEN_STATIC_ASSERT(!NumTraits::IsComplex, NUMERIC_TYPE_MUST_BE_REAL); - } - - EigenvectorsType m_eivec; - ComplexVectorType m_alphas; - VectorType m_betas; - bool m_valuesOkay, m_vectorsOkay; - RealQZ m_realQZ; - ComplexVectorType m_tmp; -}; - -template -GeneralizedEigenSolver& -GeneralizedEigenSolver::compute(const MatrixType& A, const MatrixType& B, bool computeEigenvectors) -{ - check_template_parameters(); - - using std::sqrt; - using std::abs; - eigen_assert(A.cols() == A.rows() && B.cols() == A.rows() && B.cols() == B.rows()); - Index size = A.cols(); - m_valuesOkay = false; - m_vectorsOkay = false; - // Reduce to generalized real Schur form: - // A = Q S Z and B = Q T Z - m_realQZ.compute(A, B, computeEigenvectors); - if (m_realQZ.info() == Success) - { - // Resize storage - m_alphas.resize(size); - m_betas.resize(size); - if (computeEigenvectors) - { - m_eivec.resize(size,size); - m_tmp.resize(size); - } - - // Aliases: - Map v(reinterpret_cast(m_tmp.data()), size); - ComplexVectorType &cv = m_tmp; - const MatrixType &mS = m_realQZ.matrixS(); - const MatrixType &mT = m_realQZ.matrixT(); - - Index i = 0; - while (i < size) - { - if (i == size - 1 || mS.coeff(i+1, i) == Scalar(0)) - { - // Real eigenvalue - m_alphas.coeffRef(i) = mS.diagonal().coeff(i); - m_betas.coeffRef(i) = mT.diagonal().coeff(i); - if (computeEigenvectors) - { - v.setConstant(Scalar(0.0)); - v.coeffRef(i) = Scalar(1.0); - // For singular eigenvalues do nothing more - if(abs(m_betas.coeffRef(i)) >= (std::numeric_limits::min)()) - { - // Non-singular eigenvalue - const Scalar alpha = real(m_alphas.coeffRef(i)); - const Scalar beta = m_betas.coeffRef(i); - for (Index j = i-1; j >= 0; j--) - { - const Index st = j+1; - const Index sz = i-j; - if (j > 0 && mS.coeff(j, j-1) != Scalar(0)) - { - // 2x2 block - Matrix rhs = (alpha*mT.template block<2,Dynamic>(j-1,st,2,sz) - beta*mS.template block<2,Dynamic>(j-1,st,2,sz)) .lazyProduct( v.segment(st,sz) ); - Matrix lhs = beta * mS.template block<2,2>(j-1,j-1) - alpha * mT.template block<2,2>(j-1,j-1); - v.template segment<2>(j-1) = lhs.partialPivLu().solve(rhs); - j--; - } - else - { - v.coeffRef(j) = -v.segment(st,sz).transpose().cwiseProduct(beta*mS.block(j,st,1,sz) - alpha*mT.block(j,st,1,sz)).sum() / (beta*mS.coeffRef(j,j) - alpha*mT.coeffRef(j,j)); - } - } - } - m_eivec.col(i).real().noalias() = m_realQZ.matrixZ().transpose() * v; - m_eivec.col(i).real().normalize(); - m_eivec.col(i).imag().setConstant(0); - } - ++i; - } - else - { - // We need to extract the generalized eigenvalues of the pair of a general 2x2 block S and a positive diagonal 2x2 block T - // Then taking beta=T_00*T_11, we can avoid any division, and alpha is the eigenvalues of A = (U^-1 * S * U) * diag(T_11,T_00): - - // T = [a 0] - // [0 b] - RealScalar a = mT.diagonal().coeff(i), - b = mT.diagonal().coeff(i+1); - const RealScalar beta = m_betas.coeffRef(i) = m_betas.coeffRef(i+1) = a*b; - - // ^^ NOTE: using diagonal()(i) instead of coeff(i,i) workarounds a MSVC bug. - Matrix S2 = mS.template block<2,2>(i,i) * Matrix(b,a).asDiagonal(); - - Scalar p = Scalar(0.5) * (S2.coeff(0,0) - S2.coeff(1,1)); - Scalar z = sqrt(abs(p * p + S2.coeff(1,0) * S2.coeff(0,1))); - const ComplexScalar alpha = ComplexScalar(S2.coeff(1,1) + p, (beta > 0) ? z : -z); - m_alphas.coeffRef(i) = conj(alpha); - m_alphas.coeffRef(i+1) = alpha; - - if (computeEigenvectors) { - // Compute eigenvector in position (i+1) and then position (i) is just the conjugate - cv.setZero(); - cv.coeffRef(i+1) = Scalar(1.0); - // here, the "static_cast" workaound expression template issues. - cv.coeffRef(i) = -(static_cast(beta*mS.coeffRef(i,i+1)) - alpha*mT.coeffRef(i,i+1)) - / (static_cast(beta*mS.coeffRef(i,i)) - alpha*mT.coeffRef(i,i)); - for (Index j = i-1; j >= 0; j--) - { - const Index st = j+1; - const Index sz = i+1-j; - if (j > 0 && mS.coeff(j, j-1) != Scalar(0)) - { - // 2x2 block - Matrix rhs = (alpha*mT.template block<2,Dynamic>(j-1,st,2,sz) - beta*mS.template block<2,Dynamic>(j-1,st,2,sz)) .lazyProduct( cv.segment(st,sz) ); - Matrix lhs = beta * mS.template block<2,2>(j-1,j-1) - alpha * mT.template block<2,2>(j-1,j-1); - cv.template segment<2>(j-1) = lhs.partialPivLu().solve(rhs); - j--; - } else { - cv.coeffRef(j) = cv.segment(st,sz).transpose().cwiseProduct(beta*mS.block(j,st,1,sz) - alpha*mT.block(j,st,1,sz)).sum() - / (alpha*mT.coeffRef(j,j) - static_cast(beta*mS.coeffRef(j,j))); - } - } - m_eivec.col(i+1).noalias() = (m_realQZ.matrixZ().transpose() * cv); - m_eivec.col(i+1).normalize(); - m_eivec.col(i) = m_eivec.col(i+1).conjugate(); - } - i += 2; - } - } - - m_valuesOkay = true; - m_vectorsOkay = computeEigenvectors; - } - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_GENERALIZEDEIGENSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h deleted file mode 100644 index 5f6bb82898b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/GeneralizedSelfAdjointEigenSolver.h +++ /dev/null @@ -1,226 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2010 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERALIZEDSELFADJOINTEIGENSOLVER_H -#define EIGEN_GENERALIZEDSELFADJOINTEIGENSOLVER_H - -#include "./Tridiagonalization.h" - -namespace Eigen { - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class GeneralizedSelfAdjointEigenSolver - * - * \brief Computes eigenvalues and eigenvectors of the generalized selfadjoint eigen problem - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * eigendecomposition; this is expected to be an instantiation of the Matrix - * class template. - * - * This class solves the generalized eigenvalue problem - * \f$ Av = \lambda Bv \f$. In this case, the matrix \f$ A \f$ should be - * selfadjoint and the matrix \f$ B \f$ should be positive definite. - * - * Only the \b lower \b triangular \b part of the input matrix is referenced. - * - * Call the function compute() to compute the eigenvalues and eigenvectors of - * a given matrix. Alternatively, you can use the - * GeneralizedSelfAdjointEigenSolver(const MatrixType&, const MatrixType&, int) - * constructor which computes the eigenvalues and eigenvectors at construction time. - * Once the eigenvalue and eigenvectors are computed, they can be retrieved with the eigenvalues() - * and eigenvectors() functions. - * - * The documentation for GeneralizedSelfAdjointEigenSolver(const MatrixType&, const MatrixType&, int) - * contains an example of the typical use of this class. - * - * \sa class SelfAdjointEigenSolver, class EigenSolver, class ComplexEigenSolver - */ -template -class GeneralizedSelfAdjointEigenSolver : public SelfAdjointEigenSolver<_MatrixType> -{ - typedef SelfAdjointEigenSolver<_MatrixType> Base; - public: - - typedef _MatrixType MatrixType; - - /** \brief Default constructor for fixed-size matrices. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). This constructor - * can only be used if \p _MatrixType is a fixed-size matrix; use - * GeneralizedSelfAdjointEigenSolver(Index) for dynamic-size matrices. - */ - GeneralizedSelfAdjointEigenSolver() : Base() {} - - /** \brief Constructor, pre-allocates memory for dynamic-size matrices. - * - * \param [in] size Positive integer, size of the matrix whose - * eigenvalues and eigenvectors will be computed. - * - * This constructor is useful for dynamic-size matrices, when the user - * intends to perform decompositions via compute(). The \p size - * parameter is only used as a hint. It is not an error to give a wrong - * \p size, but it may impair performance. - * - * \sa compute() for an example - */ - explicit GeneralizedSelfAdjointEigenSolver(Index size) - : Base(size) - {} - - /** \brief Constructor; computes generalized eigendecomposition of given matrix pencil. - * - * \param[in] matA Selfadjoint matrix in matrix pencil. - * Only the lower triangular part of the matrix is referenced. - * \param[in] matB Positive-definite matrix in matrix pencil. - * Only the lower triangular part of the matrix is referenced. - * \param[in] options A or-ed set of flags {#ComputeEigenvectors,#EigenvaluesOnly} | {#Ax_lBx,#ABx_lx,#BAx_lx}. - * Default is #ComputeEigenvectors|#Ax_lBx. - * - * This constructor calls compute(const MatrixType&, const MatrixType&, int) - * to compute the eigenvalues and (if requested) the eigenvectors of the - * generalized eigenproblem \f$ Ax = \lambda B x \f$ with \a matA the - * selfadjoint matrix \f$ A \f$ and \a matB the positive definite matrix - * \f$ B \f$. Each eigenvector \f$ x \f$ satisfies the property - * \f$ x^* B x = 1 \f$. The eigenvectors are computed if - * \a options contains ComputeEigenvectors. - * - * In addition, the two following variants can be solved via \p options: - * - \c ABx_lx: \f$ ABx = \lambda x \f$ - * - \c BAx_lx: \f$ BAx = \lambda x \f$ - * - * Example: \include SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType2.cpp - * Output: \verbinclude SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType2.out - * - * \sa compute(const MatrixType&, const MatrixType&, int) - */ - GeneralizedSelfAdjointEigenSolver(const MatrixType& matA, const MatrixType& matB, - int options = ComputeEigenvectors|Ax_lBx) - : Base(matA.cols()) - { - compute(matA, matB, options); - } - - /** \brief Computes generalized eigendecomposition of given matrix pencil. - * - * \param[in] matA Selfadjoint matrix in matrix pencil. - * Only the lower triangular part of the matrix is referenced. - * \param[in] matB Positive-definite matrix in matrix pencil. - * Only the lower triangular part of the matrix is referenced. - * \param[in] options A or-ed set of flags {#ComputeEigenvectors,#EigenvaluesOnly} | {#Ax_lBx,#ABx_lx,#BAx_lx}. - * Default is #ComputeEigenvectors|#Ax_lBx. - * - * \returns Reference to \c *this - * - * Accoring to \p options, this function computes eigenvalues and (if requested) - * the eigenvectors of one of the following three generalized eigenproblems: - * - \c Ax_lBx: \f$ Ax = \lambda B x \f$ - * - \c ABx_lx: \f$ ABx = \lambda x \f$ - * - \c BAx_lx: \f$ BAx = \lambda x \f$ - * with \a matA the selfadjoint matrix \f$ A \f$ and \a matB the positive definite - * matrix \f$ B \f$. - * In addition, each eigenvector \f$ x \f$ satisfies the property \f$ x^* B x = 1 \f$. - * - * The eigenvalues() function can be used to retrieve - * the eigenvalues. If \p options contains ComputeEigenvectors, then the - * eigenvectors are also computed and can be retrieved by calling - * eigenvectors(). - * - * The implementation uses LLT to compute the Cholesky decomposition - * \f$ B = LL^* \f$ and computes the classical eigendecomposition - * of the selfadjoint matrix \f$ L^{-1} A (L^*)^{-1} \f$ if \p options contains Ax_lBx - * and of \f$ L^{*} A L \f$ otherwise. This solves the - * generalized eigenproblem, because any solution of the generalized - * eigenproblem \f$ Ax = \lambda B x \f$ corresponds to a solution - * \f$ L^{-1} A (L^*)^{-1} (L^* x) = \lambda (L^* x) \f$ of the - * eigenproblem for \f$ L^{-1} A (L^*)^{-1} \f$. Similar statements - * can be made for the two other variants. - * - * Example: \include SelfAdjointEigenSolver_compute_MatrixType2.cpp - * Output: \verbinclude SelfAdjointEigenSolver_compute_MatrixType2.out - * - * \sa GeneralizedSelfAdjointEigenSolver(const MatrixType&, const MatrixType&, int) - */ - GeneralizedSelfAdjointEigenSolver& compute(const MatrixType& matA, const MatrixType& matB, - int options = ComputeEigenvectors|Ax_lBx); - - protected: - -}; - - -template -GeneralizedSelfAdjointEigenSolver& GeneralizedSelfAdjointEigenSolver:: -compute(const MatrixType& matA, const MatrixType& matB, int options) -{ - eigen_assert(matA.cols()==matA.rows() && matB.rows()==matA.rows() && matB.cols()==matB.rows()); - eigen_assert((options&~(EigVecMask|GenEigMask))==0 - && (options&EigVecMask)!=EigVecMask - && ((options&GenEigMask)==0 || (options&GenEigMask)==Ax_lBx - || (options&GenEigMask)==ABx_lx || (options&GenEigMask)==BAx_lx) - && "invalid option parameter"); - - bool computeEigVecs = ((options&EigVecMask)==0) || ((options&EigVecMask)==ComputeEigenvectors); - - // Compute the cholesky decomposition of matB = L L' = U'U - LLT cholB(matB); - - int type = (options&GenEigMask); - if(type==0) - type = Ax_lBx; - - if(type==Ax_lBx) - { - // compute C = inv(L) A inv(L') - MatrixType matC = matA.template selfadjointView(); - cholB.matrixL().template solveInPlace(matC); - cholB.matrixU().template solveInPlace(matC); - - Base::compute(matC, computeEigVecs ? ComputeEigenvectors : EigenvaluesOnly ); - - // transform back the eigen vectors: evecs = inv(U) * evecs - if(computeEigVecs) - cholB.matrixU().solveInPlace(Base::m_eivec); - } - else if(type==ABx_lx) - { - // compute C = L' A L - MatrixType matC = matA.template selfadjointView(); - matC = matC * cholB.matrixL(); - matC = cholB.matrixU() * matC; - - Base::compute(matC, computeEigVecs ? ComputeEigenvectors : EigenvaluesOnly); - - // transform back the eigen vectors: evecs = inv(U) * evecs - if(computeEigVecs) - cholB.matrixU().solveInPlace(Base::m_eivec); - } - else if(type==BAx_lx) - { - // compute C = L' A L - MatrixType matC = matA.template selfadjointView(); - matC = matC * cholB.matrixL(); - matC = cholB.matrixU() * matC; - - Base::compute(matC, computeEigVecs ? ComputeEigenvectors : EigenvaluesOnly); - - // transform back the eigen vectors: evecs = L * evecs - if(computeEigVecs) - Base::m_eivec = cholB.matrixL() * Base::m_eivec; - } - - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_GENERALIZEDSELFADJOINTEIGENSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/HessenbergDecomposition.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/HessenbergDecomposition.h deleted file mode 100644 index f647f69b060..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/HessenbergDecomposition.h +++ /dev/null @@ -1,374 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2010 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HESSENBERGDECOMPOSITION_H -#define EIGEN_HESSENBERGDECOMPOSITION_H - -namespace Eigen { - -namespace internal { - -template struct HessenbergDecompositionMatrixHReturnType; -template -struct traits > -{ - typedef MatrixType ReturnType; -}; - -} - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class HessenbergDecomposition - * - * \brief Reduces a square matrix to Hessenberg form by an orthogonal similarity transformation - * - * \tparam _MatrixType the type of the matrix of which we are computing the Hessenberg decomposition - * - * This class performs an Hessenberg decomposition of a matrix \f$ A \f$. In - * the real case, the Hessenberg decomposition consists of an orthogonal - * matrix \f$ Q \f$ and a Hessenberg matrix \f$ H \f$ such that \f$ A = Q H - * Q^T \f$. An orthogonal matrix is a matrix whose inverse equals its - * transpose (\f$ Q^{-1} = Q^T \f$). A Hessenberg matrix has zeros below the - * subdiagonal, so it is almost upper triangular. The Hessenberg decomposition - * of a complex matrix is \f$ A = Q H Q^* \f$ with \f$ Q \f$ unitary (that is, - * \f$ Q^{-1} = Q^* \f$). - * - * Call the function compute() to compute the Hessenberg decomposition of a - * given matrix. Alternatively, you can use the - * HessenbergDecomposition(const MatrixType&) constructor which computes the - * Hessenberg decomposition at construction time. Once the decomposition is - * computed, you can use the matrixH() and matrixQ() functions to construct - * the matrices H and Q in the decomposition. - * - * The documentation for matrixH() contains an example of the typical use of - * this class. - * - * \sa class ComplexSchur, class Tridiagonalization, \ref QR_Module "QR Module" - */ -template class HessenbergDecomposition -{ - public: - - /** \brief Synonym for the template parameter \p _MatrixType. */ - typedef _MatrixType MatrixType; - - enum { - Size = MatrixType::RowsAtCompileTime, - SizeMinusOne = Size == Dynamic ? Dynamic : Size - 1, - Options = MatrixType::Options, - MaxSize = MatrixType::MaxRowsAtCompileTime, - MaxSizeMinusOne = MaxSize == Dynamic ? Dynamic : MaxSize - 1 - }; - - /** \brief Scalar type for matrices of type #MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - /** \brief Type for vector of Householder coefficients. - * - * This is column vector with entries of type #Scalar. The length of the - * vector is one less than the size of #MatrixType, if it is a fixed-side - * type. - */ - typedef Matrix CoeffVectorType; - - /** \brief Return type of matrixQ() */ - typedef HouseholderSequence::type> HouseholderSequenceType; - - typedef internal::HessenbergDecompositionMatrixHReturnType MatrixHReturnType; - - /** \brief Default constructor; the decomposition will be computed later. - * - * \param [in] size The size of the matrix whose Hessenberg decomposition will be computed. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). The \p size parameter is only - * used as a hint. It is not an error to give a wrong \p size, but it may - * impair performance. - * - * \sa compute() for an example. - */ - explicit HessenbergDecomposition(Index size = Size==Dynamic ? 2 : Size) - : m_matrix(size,size), - m_temp(size), - m_isInitialized(false) - { - if(size>1) - m_hCoeffs.resize(size-1); - } - - /** \brief Constructor; computes Hessenberg decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Hessenberg decomposition is to be computed. - * - * This constructor calls compute() to compute the Hessenberg - * decomposition. - * - * \sa matrixH() for an example. - */ - template - explicit HessenbergDecomposition(const EigenBase& matrix) - : m_matrix(matrix.derived()), - m_temp(matrix.rows()), - m_isInitialized(false) - { - if(matrix.rows()<2) - { - m_isInitialized = true; - return; - } - m_hCoeffs.resize(matrix.rows()-1,1); - _compute(m_matrix, m_hCoeffs, m_temp); - m_isInitialized = true; - } - - /** \brief Computes Hessenberg decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Hessenberg decomposition is to be computed. - * \returns Reference to \c *this - * - * The Hessenberg decomposition is computed by bringing the columns of the - * matrix successively in the required form using Householder reflections - * (see, e.g., Algorithm 7.4.2 in Golub \& Van Loan, %Matrix - * Computations). The cost is \f$ 10n^3/3 \f$ flops, where \f$ n \f$ - * denotes the size of the given matrix. - * - * This method reuses of the allocated data in the HessenbergDecomposition - * object. - * - * Example: \include HessenbergDecomposition_compute.cpp - * Output: \verbinclude HessenbergDecomposition_compute.out - */ - template - HessenbergDecomposition& compute(const EigenBase& matrix) - { - m_matrix = matrix.derived(); - if(matrix.rows()<2) - { - m_isInitialized = true; - return *this; - } - m_hCoeffs.resize(matrix.rows()-1,1); - _compute(m_matrix, m_hCoeffs, m_temp); - m_isInitialized = true; - return *this; - } - - /** \brief Returns the Householder coefficients. - * - * \returns a const reference to the vector of Householder coefficients - * - * \pre Either the constructor HessenbergDecomposition(const MatrixType&) - * or the member function compute(const MatrixType&) has been called - * before to compute the Hessenberg decomposition of a matrix. - * - * The Householder coefficients allow the reconstruction of the matrix - * \f$ Q \f$ in the Hessenberg decomposition from the packed data. - * - * \sa packedMatrix(), \ref Householder_Module "Householder module" - */ - const CoeffVectorType& householderCoefficients() const - { - eigen_assert(m_isInitialized && "HessenbergDecomposition is not initialized."); - return m_hCoeffs; - } - - /** \brief Returns the internal representation of the decomposition - * - * \returns a const reference to a matrix with the internal representation - * of the decomposition. - * - * \pre Either the constructor HessenbergDecomposition(const MatrixType&) - * or the member function compute(const MatrixType&) has been called - * before to compute the Hessenberg decomposition of a matrix. - * - * The returned matrix contains the following information: - * - the upper part and lower sub-diagonal represent the Hessenberg matrix H - * - the rest of the lower part contains the Householder vectors that, combined with - * Householder coefficients returned by householderCoefficients(), - * allows to reconstruct the matrix Q as - * \f$ Q = H_{N-1} \ldots H_1 H_0 \f$. - * Here, the matrices \f$ H_i \f$ are the Householder transformations - * \f$ H_i = (I - h_i v_i v_i^T) \f$ - * where \f$ h_i \f$ is the \f$ i \f$th Householder coefficient and - * \f$ v_i \f$ is the Householder vector defined by - * \f$ v_i = [ 0, \ldots, 0, 1, M(i+2,i), \ldots, M(N-1,i) ]^T \f$ - * with M the matrix returned by this function. - * - * See LAPACK for further details on this packed storage. - * - * Example: \include HessenbergDecomposition_packedMatrix.cpp - * Output: \verbinclude HessenbergDecomposition_packedMatrix.out - * - * \sa householderCoefficients() - */ - const MatrixType& packedMatrix() const - { - eigen_assert(m_isInitialized && "HessenbergDecomposition is not initialized."); - return m_matrix; - } - - /** \brief Reconstructs the orthogonal matrix Q in the decomposition - * - * \returns object representing the matrix Q - * - * \pre Either the constructor HessenbergDecomposition(const MatrixType&) - * or the member function compute(const MatrixType&) has been called - * before to compute the Hessenberg decomposition of a matrix. - * - * This function returns a light-weight object of template class - * HouseholderSequence. You can either apply it directly to a matrix or - * you can convert it to a matrix of type #MatrixType. - * - * \sa matrixH() for an example, class HouseholderSequence - */ - HouseholderSequenceType matrixQ() const - { - eigen_assert(m_isInitialized && "HessenbergDecomposition is not initialized."); - return HouseholderSequenceType(m_matrix, m_hCoeffs.conjugate()) - .setLength(m_matrix.rows() - 1) - .setShift(1); - } - - /** \brief Constructs the Hessenberg matrix H in the decomposition - * - * \returns expression object representing the matrix H - * - * \pre Either the constructor HessenbergDecomposition(const MatrixType&) - * or the member function compute(const MatrixType&) has been called - * before to compute the Hessenberg decomposition of a matrix. - * - * The object returned by this function constructs the Hessenberg matrix H - * when it is assigned to a matrix or otherwise evaluated. The matrix H is - * constructed from the packed matrix as returned by packedMatrix(): The - * upper part (including the subdiagonal) of the packed matrix contains - * the matrix H. It may sometimes be better to directly use the packed - * matrix instead of constructing the matrix H. - * - * Example: \include HessenbergDecomposition_matrixH.cpp - * Output: \verbinclude HessenbergDecomposition_matrixH.out - * - * \sa matrixQ(), packedMatrix() - */ - MatrixHReturnType matrixH() const - { - eigen_assert(m_isInitialized && "HessenbergDecomposition is not initialized."); - return MatrixHReturnType(*this); - } - - private: - - typedef Matrix VectorType; - typedef typename NumTraits::Real RealScalar; - static void _compute(MatrixType& matA, CoeffVectorType& hCoeffs, VectorType& temp); - - protected: - MatrixType m_matrix; - CoeffVectorType m_hCoeffs; - VectorType m_temp; - bool m_isInitialized; -}; - -/** \internal - * Performs a tridiagonal decomposition of \a matA in place. - * - * \param matA the input selfadjoint matrix - * \param hCoeffs returned Householder coefficients - * - * The result is written in the lower triangular part of \a matA. - * - * Implemented from Golub's "%Matrix Computations", algorithm 8.3.1. - * - * \sa packedMatrix() - */ -template -void HessenbergDecomposition::_compute(MatrixType& matA, CoeffVectorType& hCoeffs, VectorType& temp) -{ - eigen_assert(matA.rows()==matA.cols()); - Index n = matA.rows(); - temp.resize(n); - for (Index i = 0; i struct HessenbergDecompositionMatrixHReturnType -: public ReturnByValue > -{ - public: - /** \brief Constructor. - * - * \param[in] hess Hessenberg decomposition - */ - HessenbergDecompositionMatrixHReturnType(const HessenbergDecomposition& hess) : m_hess(hess) { } - - /** \brief Hessenberg matrix in decomposition. - * - * \param[out] result Hessenberg matrix in decomposition \p hess which - * was passed to the constructor - */ - template - inline void evalTo(ResultType& result) const - { - result = m_hess.packedMatrix(); - Index n = result.rows(); - if (n>2) - result.bottomLeftCorner(n-2, n-2).template triangularView().setZero(); - } - - Index rows() const { return m_hess.packedMatrix().rows(); } - Index cols() const { return m_hess.packedMatrix().cols(); } - - protected: - const HessenbergDecomposition& m_hess; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_HESSENBERGDECOMPOSITION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h deleted file mode 100644 index e4e42607113..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h +++ /dev/null @@ -1,158 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2010 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MATRIXBASEEIGENVALUES_H -#define EIGEN_MATRIXBASEEIGENVALUES_H - -namespace Eigen { - -namespace internal { - -template -struct eigenvalues_selector -{ - // this is the implementation for the case IsComplex = true - static inline typename MatrixBase::EigenvaluesReturnType const - run(const MatrixBase& m) - { - typedef typename Derived::PlainObject PlainObject; - PlainObject m_eval(m); - return ComplexEigenSolver(m_eval, false).eigenvalues(); - } -}; - -template -struct eigenvalues_selector -{ - static inline typename MatrixBase::EigenvaluesReturnType const - run(const MatrixBase& m) - { - typedef typename Derived::PlainObject PlainObject; - PlainObject m_eval(m); - return EigenSolver(m_eval, false).eigenvalues(); - } -}; - -} // end namespace internal - -/** \brief Computes the eigenvalues of a matrix - * \returns Column vector containing the eigenvalues. - * - * \eigenvalues_module - * This function computes the eigenvalues with the help of the EigenSolver - * class (for real matrices) or the ComplexEigenSolver class (for complex - * matrices). - * - * The eigenvalues are repeated according to their algebraic multiplicity, - * so there are as many eigenvalues as rows in the matrix. - * - * The SelfAdjointView class provides a better algorithm for selfadjoint - * matrices. - * - * Example: \include MatrixBase_eigenvalues.cpp - * Output: \verbinclude MatrixBase_eigenvalues.out - * - * \sa EigenSolver::eigenvalues(), ComplexEigenSolver::eigenvalues(), - * SelfAdjointView::eigenvalues() - */ -template -inline typename MatrixBase::EigenvaluesReturnType -MatrixBase::eigenvalues() const -{ - return internal::eigenvalues_selector::IsComplex>::run(derived()); -} - -/** \brief Computes the eigenvalues of a matrix - * \returns Column vector containing the eigenvalues. - * - * \eigenvalues_module - * This function computes the eigenvalues with the help of the - * SelfAdjointEigenSolver class. The eigenvalues are repeated according to - * their algebraic multiplicity, so there are as many eigenvalues as rows in - * the matrix. - * - * Example: \include SelfAdjointView_eigenvalues.cpp - * Output: \verbinclude SelfAdjointView_eigenvalues.out - * - * \sa SelfAdjointEigenSolver::eigenvalues(), MatrixBase::eigenvalues() - */ -template -inline typename SelfAdjointView::EigenvaluesReturnType -SelfAdjointView::eigenvalues() const -{ - PlainObject thisAsMatrix(*this); - return SelfAdjointEigenSolver(thisAsMatrix, false).eigenvalues(); -} - - - -/** \brief Computes the L2 operator norm - * \returns Operator norm of the matrix. - * - * \eigenvalues_module - * This function computes the L2 operator norm of a matrix, which is also - * known as the spectral norm. The norm of a matrix \f$ A \f$ is defined to be - * \f[ \|A\|_2 = \max_x \frac{\|Ax\|_2}{\|x\|_2} \f] - * where the maximum is over all vectors and the norm on the right is the - * Euclidean vector norm. The norm equals the largest singular value, which is - * the square root of the largest eigenvalue of the positive semi-definite - * matrix \f$ A^*A \f$. - * - * The current implementation uses the eigenvalues of \f$ A^*A \f$, as computed - * by SelfAdjointView::eigenvalues(), to compute the operator norm of a - * matrix. The SelfAdjointView class provides a better algorithm for - * selfadjoint matrices. - * - * Example: \include MatrixBase_operatorNorm.cpp - * Output: \verbinclude MatrixBase_operatorNorm.out - * - * \sa SelfAdjointView::eigenvalues(), SelfAdjointView::operatorNorm() - */ -template -inline typename MatrixBase::RealScalar -MatrixBase::operatorNorm() const -{ - using std::sqrt; - typename Derived::PlainObject m_eval(derived()); - // FIXME if it is really guaranteed that the eigenvalues are already sorted, - // then we don't need to compute a maxCoeff() here, comparing the 1st and last ones is enough. - return sqrt((m_eval*m_eval.adjoint()) - .eval() - .template selfadjointView() - .eigenvalues() - .maxCoeff() - ); -} - -/** \brief Computes the L2 operator norm - * \returns Operator norm of the matrix. - * - * \eigenvalues_module - * This function computes the L2 operator norm of a self-adjoint matrix. For a - * self-adjoint matrix, the operator norm is the largest eigenvalue. - * - * The current implementation uses the eigenvalues of the matrix, as computed - * by eigenvalues(), to compute the operator norm of the matrix. - * - * Example: \include SelfAdjointView_operatorNorm.cpp - * Output: \verbinclude SelfAdjointView_operatorNorm.out - * - * \sa eigenvalues(), MatrixBase::operatorNorm() - */ -template -inline typename SelfAdjointView::RealScalar -SelfAdjointView::operatorNorm() const -{ - return eigenvalues().cwiseAbs().maxCoeff(); -} - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealQZ.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealQZ.h deleted file mode 100644 index b3a910dd9f3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealQZ.h +++ /dev/null @@ -1,654 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Alexey Korepanov -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_REAL_QZ_H -#define EIGEN_REAL_QZ_H - -namespace Eigen { - - /** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class RealQZ - * - * \brief Performs a real QZ decomposition of a pair of square matrices - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * real QZ decomposition; this is expected to be an instantiation of the - * Matrix class template. - * - * Given a real square matrices A and B, this class computes the real QZ - * decomposition: \f$ A = Q S Z \f$, \f$ B = Q T Z \f$ where Q and Z are - * real orthogonal matrixes, T is upper-triangular matrix, and S is upper - * quasi-triangular matrix. An orthogonal matrix is a matrix whose - * inverse is equal to its transpose, \f$ U^{-1} = U^T \f$. A quasi-triangular - * matrix is a block-triangular matrix whose diagonal consists of 1-by-1 - * blocks and 2-by-2 blocks where further reduction is impossible due to - * complex eigenvalues. - * - * The eigenvalues of the pencil \f$ A - z B \f$ can be obtained from - * 1x1 and 2x2 blocks on the diagonals of S and T. - * - * Call the function compute() to compute the real QZ decomposition of a - * given pair of matrices. Alternatively, you can use the - * RealQZ(const MatrixType& B, const MatrixType& B, bool computeQZ) - * constructor which computes the real QZ decomposition at construction - * time. Once the decomposition is computed, you can use the matrixS(), - * matrixT(), matrixQ() and matrixZ() functions to retrieve the matrices - * S, T, Q and Z in the decomposition. If computeQZ==false, some time - * is saved by not computing matrices Q and Z. - * - * Example: \include RealQZ_compute.cpp - * Output: \include RealQZ_compute.out - * - * \note The implementation is based on the algorithm in "Matrix Computations" - * by Gene H. Golub and Charles F. Van Loan, and a paper "An algorithm for - * generalized eigenvalue problems" by C.B.Moler and G.W.Stewart. - * - * \sa class RealSchur, class ComplexSchur, class EigenSolver, class ComplexEigenSolver - */ - - template class RealQZ - { - public: - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef std::complex::Real> ComplexScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - typedef Matrix EigenvalueType; - typedef Matrix ColumnVectorType; - - /** \brief Default constructor. - * - * \param [in] size Positive integer, size of the matrix whose QZ decomposition will be computed. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). The \p size parameter is only - * used as a hint. It is not an error to give a wrong \p size, but it may - * impair performance. - * - * \sa compute() for an example. - */ - explicit RealQZ(Index size = RowsAtCompileTime==Dynamic ? 1 : RowsAtCompileTime) : - m_S(size, size), - m_T(size, size), - m_Q(size, size), - m_Z(size, size), - m_workspace(size*2), - m_maxIters(400), - m_isInitialized(false) - { } - - /** \brief Constructor; computes real QZ decomposition of given matrices - * - * \param[in] A Matrix A. - * \param[in] B Matrix B. - * \param[in] computeQZ If false, A and Z are not computed. - * - * This constructor calls compute() to compute the QZ decomposition. - */ - RealQZ(const MatrixType& A, const MatrixType& B, bool computeQZ = true) : - m_S(A.rows(),A.cols()), - m_T(A.rows(),A.cols()), - m_Q(A.rows(),A.cols()), - m_Z(A.rows(),A.cols()), - m_workspace(A.rows()*2), - m_maxIters(400), - m_isInitialized(false) { - compute(A, B, computeQZ); - } - - /** \brief Returns matrix Q in the QZ decomposition. - * - * \returns A const reference to the matrix Q. - */ - const MatrixType& matrixQ() const { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - eigen_assert(m_computeQZ && "The matrices Q and Z have not been computed during the QZ decomposition."); - return m_Q; - } - - /** \brief Returns matrix Z in the QZ decomposition. - * - * \returns A const reference to the matrix Z. - */ - const MatrixType& matrixZ() const { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - eigen_assert(m_computeQZ && "The matrices Q and Z have not been computed during the QZ decomposition."); - return m_Z; - } - - /** \brief Returns matrix S in the QZ decomposition. - * - * \returns A const reference to the matrix S. - */ - const MatrixType& matrixS() const { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - return m_S; - } - - /** \brief Returns matrix S in the QZ decomposition. - * - * \returns A const reference to the matrix S. - */ - const MatrixType& matrixT() const { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - return m_T; - } - - /** \brief Computes QZ decomposition of given matrix. - * - * \param[in] A Matrix A. - * \param[in] B Matrix B. - * \param[in] computeQZ If false, A and Z are not computed. - * \returns Reference to \c *this - */ - RealQZ& compute(const MatrixType& A, const MatrixType& B, bool computeQZ = true); - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, \c NoConvergence otherwise. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - return m_info; - } - - /** \brief Returns number of performed QR-like iterations. - */ - Index iterations() const - { - eigen_assert(m_isInitialized && "RealQZ is not initialized."); - return m_global_iter; - } - - /** Sets the maximal number of iterations allowed to converge to one eigenvalue - * or decouple the problem. - */ - RealQZ& setMaxIterations(Index maxIters) - { - m_maxIters = maxIters; - return *this; - } - - private: - - MatrixType m_S, m_T, m_Q, m_Z; - Matrix m_workspace; - ComputationInfo m_info; - Index m_maxIters; - bool m_isInitialized; - bool m_computeQZ; - Scalar m_normOfT, m_normOfS; - Index m_global_iter; - - typedef Matrix Vector3s; - typedef Matrix Vector2s; - typedef Matrix Matrix2s; - typedef JacobiRotation JRs; - - void hessenbergTriangular(); - void computeNorms(); - Index findSmallSubdiagEntry(Index iu); - Index findSmallDiagEntry(Index f, Index l); - void splitOffTwoRows(Index i); - void pushDownZero(Index z, Index f, Index l); - void step(Index f, Index l, Index iter); - - }; // RealQZ - - /** \internal Reduces S and T to upper Hessenberg - triangular form */ - template - void RealQZ::hessenbergTriangular() - { - - const Index dim = m_S.cols(); - - // perform QR decomposition of T, overwrite T with R, save Q - HouseholderQR qrT(m_T); - m_T = qrT.matrixQR(); - m_T.template triangularView().setZero(); - m_Q = qrT.householderQ(); - // overwrite S with Q* S - m_S.applyOnTheLeft(m_Q.adjoint()); - // init Z as Identity - if (m_computeQZ) - m_Z = MatrixType::Identity(dim,dim); - // reduce S to upper Hessenberg with Givens rotations - for (Index j=0; j<=dim-3; j++) { - for (Index i=dim-1; i>=j+2; i--) { - JRs G; - // kill S(i,j) - if(m_S.coeff(i,j) != 0) - { - G.makeGivens(m_S.coeff(i-1,j), m_S.coeff(i,j), &m_S.coeffRef(i-1, j)); - m_S.coeffRef(i,j) = Scalar(0.0); - m_S.rightCols(dim-j-1).applyOnTheLeft(i-1,i,G.adjoint()); - m_T.rightCols(dim-i+1).applyOnTheLeft(i-1,i,G.adjoint()); - // update Q - if (m_computeQZ) - m_Q.applyOnTheRight(i-1,i,G); - } - // kill T(i,i-1) - if(m_T.coeff(i,i-1)!=Scalar(0)) - { - G.makeGivens(m_T.coeff(i,i), m_T.coeff(i,i-1), &m_T.coeffRef(i,i)); - m_T.coeffRef(i,i-1) = Scalar(0.0); - m_S.applyOnTheRight(i,i-1,G); - m_T.topRows(i).applyOnTheRight(i,i-1,G); - // update Z - if (m_computeQZ) - m_Z.applyOnTheLeft(i,i-1,G.adjoint()); - } - } - } - } - - /** \internal Computes vector L1 norms of S and T when in Hessenberg-Triangular form already */ - template - inline void RealQZ::computeNorms() - { - const Index size = m_S.cols(); - m_normOfS = Scalar(0.0); - m_normOfT = Scalar(0.0); - for (Index j = 0; j < size; ++j) - { - m_normOfS += m_S.col(j).segment(0, (std::min)(size,j+2)).cwiseAbs().sum(); - m_normOfT += m_T.row(j).segment(j, size - j).cwiseAbs().sum(); - } - } - - - /** \internal Look for single small sub-diagonal element S(res, res-1) and return res (or 0) */ - template - inline Index RealQZ::findSmallSubdiagEntry(Index iu) - { - using std::abs; - Index res = iu; - while (res > 0) - { - Scalar s = abs(m_S.coeff(res-1,res-1)) + abs(m_S.coeff(res,res)); - if (s == Scalar(0.0)) - s = m_normOfS; - if (abs(m_S.coeff(res,res-1)) < NumTraits::epsilon() * s) - break; - res--; - } - return res; - } - - /** \internal Look for single small diagonal element T(res, res) for res between f and l, and return res (or f-1) */ - template - inline Index RealQZ::findSmallDiagEntry(Index f, Index l) - { - using std::abs; - Index res = l; - while (res >= f) { - if (abs(m_T.coeff(res,res)) <= NumTraits::epsilon() * m_normOfT) - break; - res--; - } - return res; - } - - /** \internal decouple 2x2 diagonal block in rows i, i+1 if eigenvalues are real */ - template - inline void RealQZ::splitOffTwoRows(Index i) - { - using std::abs; - using std::sqrt; - const Index dim=m_S.cols(); - if (abs(m_S.coeff(i+1,i))==Scalar(0)) - return; - Index j = findSmallDiagEntry(i,i+1); - if (j==i-1) - { - // block of (S T^{-1}) - Matrix2s STi = m_T.template block<2,2>(i,i).template triangularView(). - template solve(m_S.template block<2,2>(i,i)); - Scalar p = Scalar(0.5)*(STi(0,0)-STi(1,1)); - Scalar q = p*p + STi(1,0)*STi(0,1); - if (q>=0) { - Scalar z = sqrt(q); - // one QR-like iteration for ABi - lambda I - // is enough - when we know exact eigenvalue in advance, - // convergence is immediate - JRs G; - if (p>=0) - G.makeGivens(p + z, STi(1,0)); - else - G.makeGivens(p - z, STi(1,0)); - m_S.rightCols(dim-i).applyOnTheLeft(i,i+1,G.adjoint()); - m_T.rightCols(dim-i).applyOnTheLeft(i,i+1,G.adjoint()); - // update Q - if (m_computeQZ) - m_Q.applyOnTheRight(i,i+1,G); - - G.makeGivens(m_T.coeff(i+1,i+1), m_T.coeff(i+1,i)); - m_S.topRows(i+2).applyOnTheRight(i+1,i,G); - m_T.topRows(i+2).applyOnTheRight(i+1,i,G); - // update Z - if (m_computeQZ) - m_Z.applyOnTheLeft(i+1,i,G.adjoint()); - - m_S.coeffRef(i+1,i) = Scalar(0.0); - m_T.coeffRef(i+1,i) = Scalar(0.0); - } - } - else - { - pushDownZero(j,i,i+1); - } - } - - /** \internal use zero in T(z,z) to zero S(l,l-1), working in block f..l */ - template - inline void RealQZ::pushDownZero(Index z, Index f, Index l) - { - JRs G; - const Index dim = m_S.cols(); - for (Index zz=z; zzf ? (zz-1) : zz; - G.makeGivens(m_T.coeff(zz, zz+1), m_T.coeff(zz+1, zz+1)); - m_S.rightCols(dim-firstColS).applyOnTheLeft(zz,zz+1,G.adjoint()); - m_T.rightCols(dim-zz).applyOnTheLeft(zz,zz+1,G.adjoint()); - m_T.coeffRef(zz+1,zz+1) = Scalar(0.0); - // update Q - if (m_computeQZ) - m_Q.applyOnTheRight(zz,zz+1,G); - // kill S(zz+1, zz-1) - if (zz>f) - { - G.makeGivens(m_S.coeff(zz+1, zz), m_S.coeff(zz+1,zz-1)); - m_S.topRows(zz+2).applyOnTheRight(zz, zz-1,G); - m_T.topRows(zz+1).applyOnTheRight(zz, zz-1,G); - m_S.coeffRef(zz+1,zz-1) = Scalar(0.0); - // update Z - if (m_computeQZ) - m_Z.applyOnTheLeft(zz,zz-1,G.adjoint()); - } - } - // finally kill S(l,l-1) - G.makeGivens(m_S.coeff(l,l), m_S.coeff(l,l-1)); - m_S.applyOnTheRight(l,l-1,G); - m_T.applyOnTheRight(l,l-1,G); - m_S.coeffRef(l,l-1)=Scalar(0.0); - // update Z - if (m_computeQZ) - m_Z.applyOnTheLeft(l,l-1,G.adjoint()); - } - - /** \internal QR-like iterative step for block f..l */ - template - inline void RealQZ::step(Index f, Index l, Index iter) - { - using std::abs; - const Index dim = m_S.cols(); - - // x, y, z - Scalar x, y, z; - if (iter==10) - { - // Wilkinson ad hoc shift - const Scalar - a11=m_S.coeff(f+0,f+0), a12=m_S.coeff(f+0,f+1), - a21=m_S.coeff(f+1,f+0), a22=m_S.coeff(f+1,f+1), a32=m_S.coeff(f+2,f+1), - b12=m_T.coeff(f+0,f+1), - b11i=Scalar(1.0)/m_T.coeff(f+0,f+0), - b22i=Scalar(1.0)/m_T.coeff(f+1,f+1), - a87=m_S.coeff(l-1,l-2), - a98=m_S.coeff(l-0,l-1), - b77i=Scalar(1.0)/m_T.coeff(l-2,l-2), - b88i=Scalar(1.0)/m_T.coeff(l-1,l-1); - Scalar ss = abs(a87*b77i) + abs(a98*b88i), - lpl = Scalar(1.5)*ss, - ll = ss*ss; - x = ll + a11*a11*b11i*b11i - lpl*a11*b11i + a12*a21*b11i*b22i - - a11*a21*b12*b11i*b11i*b22i; - y = a11*a21*b11i*b11i - lpl*a21*b11i + a21*a22*b11i*b22i - - a21*a21*b12*b11i*b11i*b22i; - z = a21*a32*b11i*b22i; - } - else if (iter==16) - { - // another exceptional shift - x = m_S.coeff(f,f)/m_T.coeff(f,f)-m_S.coeff(l,l)/m_T.coeff(l,l) + m_S.coeff(l,l-1)*m_T.coeff(l-1,l) / - (m_T.coeff(l-1,l-1)*m_T.coeff(l,l)); - y = m_S.coeff(f+1,f)/m_T.coeff(f,f); - z = 0; - } - else if (iter>23 && !(iter%8)) - { - // extremely exceptional shift - x = internal::random(-1.0,1.0); - y = internal::random(-1.0,1.0); - z = internal::random(-1.0,1.0); - } - else - { - // Compute the shifts: (x,y,z,0...) = (AB^-1 - l1 I) (AB^-1 - l2 I) e1 - // where l1 and l2 are the eigenvalues of the 2x2 matrix C = U V^-1 where - // U and V are 2x2 bottom right sub matrices of A and B. Thus: - // = AB^-1AB^-1 + l1 l2 I - (l1+l2)(AB^-1) - // = AB^-1AB^-1 + det(M) - tr(M)(AB^-1) - // Since we are only interested in having x, y, z with a correct ratio, we have: - const Scalar - a11 = m_S.coeff(f,f), a12 = m_S.coeff(f,f+1), - a21 = m_S.coeff(f+1,f), a22 = m_S.coeff(f+1,f+1), - a32 = m_S.coeff(f+2,f+1), - - a88 = m_S.coeff(l-1,l-1), a89 = m_S.coeff(l-1,l), - a98 = m_S.coeff(l,l-1), a99 = m_S.coeff(l,l), - - b11 = m_T.coeff(f,f), b12 = m_T.coeff(f,f+1), - b22 = m_T.coeff(f+1,f+1), - - b88 = m_T.coeff(l-1,l-1), b89 = m_T.coeff(l-1,l), - b99 = m_T.coeff(l,l); - - x = ( (a88/b88 - a11/b11)*(a99/b99 - a11/b11) - (a89/b99)*(a98/b88) + (a98/b88)*(b89/b99)*(a11/b11) ) * (b11/a21) - + a12/b22 - (a11/b11)*(b12/b22); - y = (a22/b22-a11/b11) - (a21/b11)*(b12/b22) - (a88/b88-a11/b11) - (a99/b99-a11/b11) + (a98/b88)*(b89/b99); - z = a32/b22; - } - - JRs G; - - for (Index k=f; k<=l-2; k++) - { - // variables for Householder reflections - Vector2s essential2; - Scalar tau, beta; - - Vector3s hr(x,y,z); - - // Q_k to annihilate S(k+1,k-1) and S(k+2,k-1) - hr.makeHouseholderInPlace(tau, beta); - essential2 = hr.template bottomRows<2>(); - Index fc=(std::max)(k-1,Index(0)); // first col to update - m_S.template middleRows<3>(k).rightCols(dim-fc).applyHouseholderOnTheLeft(essential2, tau, m_workspace.data()); - m_T.template middleRows<3>(k).rightCols(dim-fc).applyHouseholderOnTheLeft(essential2, tau, m_workspace.data()); - if (m_computeQZ) - m_Q.template middleCols<3>(k).applyHouseholderOnTheRight(essential2, tau, m_workspace.data()); - if (k>f) - m_S.coeffRef(k+2,k-1) = m_S.coeffRef(k+1,k-1) = Scalar(0.0); - - // Z_{k1} to annihilate T(k+2,k+1) and T(k+2,k) - hr << m_T.coeff(k+2,k+2),m_T.coeff(k+2,k),m_T.coeff(k+2,k+1); - hr.makeHouseholderInPlace(tau, beta); - essential2 = hr.template bottomRows<2>(); - { - Index lr = (std::min)(k+4,dim); // last row to update - Map > tmp(m_workspace.data(),lr); - // S - tmp = m_S.template middleCols<2>(k).topRows(lr) * essential2; - tmp += m_S.col(k+2).head(lr); - m_S.col(k+2).head(lr) -= tau*tmp; - m_S.template middleCols<2>(k).topRows(lr) -= (tau*tmp) * essential2.adjoint(); - // T - tmp = m_T.template middleCols<2>(k).topRows(lr) * essential2; - tmp += m_T.col(k+2).head(lr); - m_T.col(k+2).head(lr) -= tau*tmp; - m_T.template middleCols<2>(k).topRows(lr) -= (tau*tmp) * essential2.adjoint(); - } - if (m_computeQZ) - { - // Z - Map > tmp(m_workspace.data(),dim); - tmp = essential2.adjoint()*(m_Z.template middleRows<2>(k)); - tmp += m_Z.row(k+2); - m_Z.row(k+2) -= tau*tmp; - m_Z.template middleRows<2>(k) -= essential2 * (tau*tmp); - } - m_T.coeffRef(k+2,k) = m_T.coeffRef(k+2,k+1) = Scalar(0.0); - - // Z_{k2} to annihilate T(k+1,k) - G.makeGivens(m_T.coeff(k+1,k+1), m_T.coeff(k+1,k)); - m_S.applyOnTheRight(k+1,k,G); - m_T.applyOnTheRight(k+1,k,G); - // update Z - if (m_computeQZ) - m_Z.applyOnTheLeft(k+1,k,G.adjoint()); - m_T.coeffRef(k+1,k) = Scalar(0.0); - - // update x,y,z - x = m_S.coeff(k+1,k); - y = m_S.coeff(k+2,k); - if (k < l-2) - z = m_S.coeff(k+3,k); - } // loop over k - - // Q_{n-1} to annihilate y = S(l,l-2) - G.makeGivens(x,y); - m_S.applyOnTheLeft(l-1,l,G.adjoint()); - m_T.applyOnTheLeft(l-1,l,G.adjoint()); - if (m_computeQZ) - m_Q.applyOnTheRight(l-1,l,G); - m_S.coeffRef(l,l-2) = Scalar(0.0); - - // Z_{n-1} to annihilate T(l,l-1) - G.makeGivens(m_T.coeff(l,l),m_T.coeff(l,l-1)); - m_S.applyOnTheRight(l,l-1,G); - m_T.applyOnTheRight(l,l-1,G); - if (m_computeQZ) - m_Z.applyOnTheLeft(l,l-1,G.adjoint()); - m_T.coeffRef(l,l-1) = Scalar(0.0); - } - - template - RealQZ& RealQZ::compute(const MatrixType& A_in, const MatrixType& B_in, bool computeQZ) - { - - const Index dim = A_in.cols(); - - eigen_assert (A_in.rows()==dim && A_in.cols()==dim - && B_in.rows()==dim && B_in.cols()==dim - && "Need square matrices of the same dimension"); - - m_isInitialized = true; - m_computeQZ = computeQZ; - m_S = A_in; m_T = B_in; - m_workspace.resize(dim*2); - m_global_iter = 0; - - // entrance point: hessenberg triangular decomposition - hessenbergTriangular(); - // compute L1 vector norms of T, S into m_normOfS, m_normOfT - computeNorms(); - - Index l = dim-1, - f, - local_iter = 0; - - while (l>0 && local_iter0) m_S.coeffRef(f,f-1) = Scalar(0.0); - if (f == l) // One root found - { - l--; - local_iter = 0; - } - else if (f == l-1) // Two roots found - { - splitOffTwoRows(f); - l -= 2; - local_iter = 0; - } - else // No convergence yet - { - // if there's zero on diagonal of T, we can isolate an eigenvalue with Givens rotations - Index z = findSmallDiagEntry(f,l); - if (z>=f) - { - // zero found - pushDownZero(z,f,l); - } - else - { - // We are sure now that S.block(f,f, l-f+1,l-f+1) is underuced upper-Hessenberg - // and T.block(f,f, l-f+1,l-f+1) is invertible uper-triangular, which allows to - // apply a QR-like iteration to rows and columns f..l. - step(f,l, local_iter); - local_iter++; - m_global_iter++; - } - } - } - // check if we converged before reaching iterations limit - m_info = (local_iter j_left, j_right; - internal::real_2x2_jacobi_svd(m_T, i, i+1, &j_left, &j_right); - - // Apply resulting Jacobi rotations - m_S.applyOnTheLeft(i,i+1,j_left); - m_S.applyOnTheRight(i,i+1,j_right); - m_T.applyOnTheLeft(i,i+1,j_left); - m_T.applyOnTheRight(i,i+1,j_right); - m_T(i+1,i) = m_T(i,i+1) = Scalar(0); - - if(m_computeQZ) { - m_Q.applyOnTheRight(i,i+1,j_left.transpose()); - m_Z.applyOnTheLeft(i,i+1,j_right.transpose()); - } - - i++; - } - } - } - - return *this; - } // end compute - -} // end namespace Eigen - -#endif //EIGEN_REAL_QZ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur.h deleted file mode 100644 index 9191519abe9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur.h +++ /dev/null @@ -1,553 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2010,2012 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_REAL_SCHUR_H -#define EIGEN_REAL_SCHUR_H - -#include "./HessenbergDecomposition.h" - -namespace Eigen { - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class RealSchur - * - * \brief Performs a real Schur decomposition of a square matrix - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * real Schur decomposition; this is expected to be an instantiation of the - * Matrix class template. - * - * Given a real square matrix A, this class computes the real Schur - * decomposition: \f$ A = U T U^T \f$ where U is a real orthogonal matrix and - * T is a real quasi-triangular matrix. An orthogonal matrix is a matrix whose - * inverse is equal to its transpose, \f$ U^{-1} = U^T \f$. A quasi-triangular - * matrix is a block-triangular matrix whose diagonal consists of 1-by-1 - * blocks and 2-by-2 blocks with complex eigenvalues. The eigenvalues of the - * blocks on the diagonal of T are the same as the eigenvalues of the matrix - * A, and thus the real Schur decomposition is used in EigenSolver to compute - * the eigendecomposition of a matrix. - * - * Call the function compute() to compute the real Schur decomposition of a - * given matrix. Alternatively, you can use the RealSchur(const MatrixType&, bool) - * constructor which computes the real Schur decomposition at construction - * time. Once the decomposition is computed, you can use the matrixU() and - * matrixT() functions to retrieve the matrices U and T in the decomposition. - * - * The documentation of RealSchur(const MatrixType&, bool) contains an example - * of the typical use of this class. - * - * \note The implementation is adapted from - * JAMA (public domain). - * Their code is based on EISPACK. - * - * \sa class ComplexSchur, class EigenSolver, class ComplexEigenSolver - */ -template class RealSchur -{ - public: - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef std::complex::Real> ComplexScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - typedef Matrix EigenvalueType; - typedef Matrix ColumnVectorType; - - /** \brief Default constructor. - * - * \param [in] size Positive integer, size of the matrix whose Schur decomposition will be computed. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). The \p size parameter is only - * used as a hint. It is not an error to give a wrong \p size, but it may - * impair performance. - * - * \sa compute() for an example. - */ - explicit RealSchur(Index size = RowsAtCompileTime==Dynamic ? 1 : RowsAtCompileTime) - : m_matT(size, size), - m_matU(size, size), - m_workspaceVector(size), - m_hess(size), - m_isInitialized(false), - m_matUisUptodate(false), - m_maxIters(-1) - { } - - /** \brief Constructor; computes real Schur decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Schur decomposition is to be computed. - * \param[in] computeU If true, both T and U are computed; if false, only T is computed. - * - * This constructor calls compute() to compute the Schur decomposition. - * - * Example: \include RealSchur_RealSchur_MatrixType.cpp - * Output: \verbinclude RealSchur_RealSchur_MatrixType.out - */ - template - explicit RealSchur(const EigenBase& matrix, bool computeU = true) - : m_matT(matrix.rows(),matrix.cols()), - m_matU(matrix.rows(),matrix.cols()), - m_workspaceVector(matrix.rows()), - m_hess(matrix.rows()), - m_isInitialized(false), - m_matUisUptodate(false), - m_maxIters(-1) - { - compute(matrix.derived(), computeU); - } - - /** \brief Returns the orthogonal matrix in the Schur decomposition. - * - * \returns A const reference to the matrix U. - * - * \pre Either the constructor RealSchur(const MatrixType&, bool) or the - * member function compute(const MatrixType&, bool) has been called before - * to compute the Schur decomposition of a matrix, and \p computeU was set - * to true (the default value). - * - * \sa RealSchur(const MatrixType&, bool) for an example - */ - const MatrixType& matrixU() const - { - eigen_assert(m_isInitialized && "RealSchur is not initialized."); - eigen_assert(m_matUisUptodate && "The matrix U has not been computed during the RealSchur decomposition."); - return m_matU; - } - - /** \brief Returns the quasi-triangular matrix in the Schur decomposition. - * - * \returns A const reference to the matrix T. - * - * \pre Either the constructor RealSchur(const MatrixType&, bool) or the - * member function compute(const MatrixType&, bool) has been called before - * to compute the Schur decomposition of a matrix. - * - * \sa RealSchur(const MatrixType&, bool) for an example - */ - const MatrixType& matrixT() const - { - eigen_assert(m_isInitialized && "RealSchur is not initialized."); - return m_matT; - } - - /** \brief Computes Schur decomposition of given matrix. - * - * \param[in] matrix Square matrix whose Schur decomposition is to be computed. - * \param[in] computeU If true, both T and U are computed; if false, only T is computed. - * \returns Reference to \c *this - * - * The Schur decomposition is computed by first reducing the matrix to - * Hessenberg form using the class HessenbergDecomposition. The Hessenberg - * matrix is then reduced to triangular form by performing Francis QR - * iterations with implicit double shift. The cost of computing the Schur - * decomposition depends on the number of iterations; as a rough guide, it - * may be taken to be \f$25n^3\f$ flops if \a computeU is true and - * \f$10n^3\f$ flops if \a computeU is false. - * - * Example: \include RealSchur_compute.cpp - * Output: \verbinclude RealSchur_compute.out - * - * \sa compute(const MatrixType&, bool, Index) - */ - template - RealSchur& compute(const EigenBase& matrix, bool computeU = true); - - /** \brief Computes Schur decomposition of a Hessenberg matrix H = Z T Z^T - * \param[in] matrixH Matrix in Hessenberg form H - * \param[in] matrixQ orthogonal matrix Q that transform a matrix A to H : A = Q H Q^T - * \param computeU Computes the matriX U of the Schur vectors - * \return Reference to \c *this - * - * This routine assumes that the matrix is already reduced in Hessenberg form matrixH - * using either the class HessenbergDecomposition or another mean. - * It computes the upper quasi-triangular matrix T of the Schur decomposition of H - * When computeU is true, this routine computes the matrix U such that - * A = U T U^T = (QZ) T (QZ)^T = Q H Q^T where A is the initial matrix - * - * NOTE Q is referenced if computeU is true; so, if the initial orthogonal matrix - * is not available, the user should give an identity matrix (Q.setIdentity()) - * - * \sa compute(const MatrixType&, bool) - */ - template - RealSchur& computeFromHessenberg(const HessMatrixType& matrixH, const OrthMatrixType& matrixQ, bool computeU); - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, \c NoConvergence otherwise. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "RealSchur is not initialized."); - return m_info; - } - - /** \brief Sets the maximum number of iterations allowed. - * - * If not specified by the user, the maximum number of iterations is m_maxIterationsPerRow times the size - * of the matrix. - */ - RealSchur& setMaxIterations(Index maxIters) - { - m_maxIters = maxIters; - return *this; - } - - /** \brief Returns the maximum number of iterations. */ - Index getMaxIterations() - { - return m_maxIters; - } - - /** \brief Maximum number of iterations per row. - * - * If not otherwise specified, the maximum number of iterations is this number times the size of the - * matrix. It is currently set to 40. - */ - static const int m_maxIterationsPerRow = 40; - - private: - - MatrixType m_matT; - MatrixType m_matU; - ColumnVectorType m_workspaceVector; - HessenbergDecomposition m_hess; - ComputationInfo m_info; - bool m_isInitialized; - bool m_matUisUptodate; - Index m_maxIters; - - typedef Matrix Vector3s; - - Scalar computeNormOfT(); - Index findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero); - void splitOffTwoRows(Index iu, bool computeU, const Scalar& exshift); - void computeShift(Index iu, Index iter, Scalar& exshift, Vector3s& shiftInfo); - void initFrancisQRStep(Index il, Index iu, const Vector3s& shiftInfo, Index& im, Vector3s& firstHouseholderVector); - void performFrancisQRStep(Index il, Index im, Index iu, bool computeU, const Vector3s& firstHouseholderVector, Scalar* workspace); -}; - - -template -template -RealSchur& RealSchur::compute(const EigenBase& matrix, bool computeU) -{ - const Scalar considerAsZero = (std::numeric_limits::min)(); - - eigen_assert(matrix.cols() == matrix.rows()); - Index maxIters = m_maxIters; - if (maxIters == -1) - maxIters = m_maxIterationsPerRow * matrix.rows(); - - Scalar scale = matrix.derived().cwiseAbs().maxCoeff(); - if(scale -template -RealSchur& RealSchur::computeFromHessenberg(const HessMatrixType& matrixH, const OrthMatrixType& matrixQ, bool computeU) -{ - using std::abs; - - m_matT = matrixH; - if(computeU) - m_matU = matrixQ; - - Index maxIters = m_maxIters; - if (maxIters == -1) - maxIters = m_maxIterationsPerRow * matrixH.rows(); - m_workspaceVector.resize(m_matT.cols()); - Scalar* workspace = &m_workspaceVector.coeffRef(0); - - // The matrix m_matT is divided in three parts. - // Rows 0,...,il-1 are decoupled from the rest because m_matT(il,il-1) is zero. - // Rows il,...,iu is the part we are working on (the active window). - // Rows iu+1,...,end are already brought in triangular form. - Index iu = m_matT.cols() - 1; - Index iter = 0; // iteration count for current eigenvalue - Index totalIter = 0; // iteration count for whole matrix - Scalar exshift(0); // sum of exceptional shifts - Scalar norm = computeNormOfT(); - // sub-diagonal entries smaller than considerAsZero will be treated as zero. - // We use eps^2 to enable more precision in small eigenvalues. - Scalar considerAsZero = numext::maxi( norm * numext::abs2(NumTraits::epsilon()), - (std::numeric_limits::min)() ); - - if(norm!=Scalar(0)) - { - while (iu >= 0) - { - Index il = findSmallSubdiagEntry(iu,considerAsZero); - - // Check for convergence - if (il == iu) // One root found - { - m_matT.coeffRef(iu,iu) = m_matT.coeff(iu,iu) + exshift; - if (iu > 0) - m_matT.coeffRef(iu, iu-1) = Scalar(0); - iu--; - iter = 0; - } - else if (il == iu-1) // Two roots found - { - splitOffTwoRows(iu, computeU, exshift); - iu -= 2; - iter = 0; - } - else // No convergence yet - { - // The firstHouseholderVector vector has to be initialized to something to get rid of a silly GCC warning (-O1 -Wall -DNDEBUG ) - Vector3s firstHouseholderVector = Vector3s::Zero(), shiftInfo; - computeShift(iu, iter, exshift, shiftInfo); - iter = iter + 1; - totalIter = totalIter + 1; - if (totalIter > maxIters) break; - Index im; - initFrancisQRStep(il, iu, shiftInfo, im, firstHouseholderVector); - performFrancisQRStep(il, im, iu, computeU, firstHouseholderVector, workspace); - } - } - } - if(totalIter <= maxIters) - m_info = Success; - else - m_info = NoConvergence; - - m_isInitialized = true; - m_matUisUptodate = computeU; - return *this; -} - -/** \internal Computes and returns vector L1 norm of T */ -template -inline typename MatrixType::Scalar RealSchur::computeNormOfT() -{ - const Index size = m_matT.cols(); - // FIXME to be efficient the following would requires a triangular reduxion code - // Scalar norm = m_matT.upper().cwiseAbs().sum() - // + m_matT.bottomLeftCorner(size-1,size-1).diagonal().cwiseAbs().sum(); - Scalar norm(0); - for (Index j = 0; j < size; ++j) - norm += m_matT.col(j).segment(0, (std::min)(size,j+2)).cwiseAbs().sum(); - return norm; -} - -/** \internal Look for single small sub-diagonal element and returns its index */ -template -inline Index RealSchur::findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero) -{ - using std::abs; - Index res = iu; - while (res > 0) - { - Scalar s = abs(m_matT.coeff(res-1,res-1)) + abs(m_matT.coeff(res,res)); - - s = numext::maxi(s * NumTraits::epsilon(), considerAsZero); - - if (abs(m_matT.coeff(res,res-1)) <= s) - break; - res--; - } - return res; -} - -/** \internal Update T given that rows iu-1 and iu decouple from the rest. */ -template -inline void RealSchur::splitOffTwoRows(Index iu, bool computeU, const Scalar& exshift) -{ - using std::sqrt; - using std::abs; - const Index size = m_matT.cols(); - - // The eigenvalues of the 2x2 matrix [a b; c d] are - // trace +/- sqrt(discr/4) where discr = tr^2 - 4*det, tr = a + d, det = ad - bc - Scalar p = Scalar(0.5) * (m_matT.coeff(iu-1,iu-1) - m_matT.coeff(iu,iu)); - Scalar q = p * p + m_matT.coeff(iu,iu-1) * m_matT.coeff(iu-1,iu); // q = tr^2 / 4 - det = discr/4 - m_matT.coeffRef(iu,iu) += exshift; - m_matT.coeffRef(iu-1,iu-1) += exshift; - - if (q >= Scalar(0)) // Two real eigenvalues - { - Scalar z = sqrt(abs(q)); - JacobiRotation rot; - if (p >= Scalar(0)) - rot.makeGivens(p + z, m_matT.coeff(iu, iu-1)); - else - rot.makeGivens(p - z, m_matT.coeff(iu, iu-1)); - - m_matT.rightCols(size-iu+1).applyOnTheLeft(iu-1, iu, rot.adjoint()); - m_matT.topRows(iu+1).applyOnTheRight(iu-1, iu, rot); - m_matT.coeffRef(iu, iu-1) = Scalar(0); - if (computeU) - m_matU.applyOnTheRight(iu-1, iu, rot); - } - - if (iu > 1) - m_matT.coeffRef(iu-1, iu-2) = Scalar(0); -} - -/** \internal Form shift in shiftInfo, and update exshift if an exceptional shift is performed. */ -template -inline void RealSchur::computeShift(Index iu, Index iter, Scalar& exshift, Vector3s& shiftInfo) -{ - using std::sqrt; - using std::abs; - shiftInfo.coeffRef(0) = m_matT.coeff(iu,iu); - shiftInfo.coeffRef(1) = m_matT.coeff(iu-1,iu-1); - shiftInfo.coeffRef(2) = m_matT.coeff(iu,iu-1) * m_matT.coeff(iu-1,iu); - - // Wilkinson's original ad hoc shift - if (iter == 10) - { - exshift += shiftInfo.coeff(0); - for (Index i = 0; i <= iu; ++i) - m_matT.coeffRef(i,i) -= shiftInfo.coeff(0); - Scalar s = abs(m_matT.coeff(iu,iu-1)) + abs(m_matT.coeff(iu-1,iu-2)); - shiftInfo.coeffRef(0) = Scalar(0.75) * s; - shiftInfo.coeffRef(1) = Scalar(0.75) * s; - shiftInfo.coeffRef(2) = Scalar(-0.4375) * s * s; - } - - // MATLAB's new ad hoc shift - if (iter == 30) - { - Scalar s = (shiftInfo.coeff(1) - shiftInfo.coeff(0)) / Scalar(2.0); - s = s * s + shiftInfo.coeff(2); - if (s > Scalar(0)) - { - s = sqrt(s); - if (shiftInfo.coeff(1) < shiftInfo.coeff(0)) - s = -s; - s = s + (shiftInfo.coeff(1) - shiftInfo.coeff(0)) / Scalar(2.0); - s = shiftInfo.coeff(0) - shiftInfo.coeff(2) / s; - exshift += s; - for (Index i = 0; i <= iu; ++i) - m_matT.coeffRef(i,i) -= s; - shiftInfo.setConstant(Scalar(0.964)); - } - } -} - -/** \internal Compute index im at which Francis QR step starts and the first Householder vector. */ -template -inline void RealSchur::initFrancisQRStep(Index il, Index iu, const Vector3s& shiftInfo, Index& im, Vector3s& firstHouseholderVector) -{ - using std::abs; - Vector3s& v = firstHouseholderVector; // alias to save typing - - for (im = iu-2; im >= il; --im) - { - const Scalar Tmm = m_matT.coeff(im,im); - const Scalar r = shiftInfo.coeff(0) - Tmm; - const Scalar s = shiftInfo.coeff(1) - Tmm; - v.coeffRef(0) = (r * s - shiftInfo.coeff(2)) / m_matT.coeff(im+1,im) + m_matT.coeff(im,im+1); - v.coeffRef(1) = m_matT.coeff(im+1,im+1) - Tmm - r - s; - v.coeffRef(2) = m_matT.coeff(im+2,im+1); - if (im == il) { - break; - } - const Scalar lhs = m_matT.coeff(im,im-1) * (abs(v.coeff(1)) + abs(v.coeff(2))); - const Scalar rhs = v.coeff(0) * (abs(m_matT.coeff(im-1,im-1)) + abs(Tmm) + abs(m_matT.coeff(im+1,im+1))); - if (abs(lhs) < NumTraits::epsilon() * rhs) - break; - } -} - -/** \internal Perform a Francis QR step involving rows il:iu and columns im:iu. */ -template -inline void RealSchur::performFrancisQRStep(Index il, Index im, Index iu, bool computeU, const Vector3s& firstHouseholderVector, Scalar* workspace) -{ - eigen_assert(im >= il); - eigen_assert(im <= iu-2); - - const Index size = m_matT.cols(); - - for (Index k = im; k <= iu-2; ++k) - { - bool firstIteration = (k == im); - - Vector3s v; - if (firstIteration) - v = firstHouseholderVector; - else - v = m_matT.template block<3,1>(k,k-1); - - Scalar tau, beta; - Matrix ess; - v.makeHouseholder(ess, tau, beta); - - if (beta != Scalar(0)) // if v is not zero - { - if (firstIteration && k > il) - m_matT.coeffRef(k,k-1) = -m_matT.coeff(k,k-1); - else if (!firstIteration) - m_matT.coeffRef(k,k-1) = beta; - - // These Householder transformations form the O(n^3) part of the algorithm - m_matT.block(k, k, 3, size-k).applyHouseholderOnTheLeft(ess, tau, workspace); - m_matT.block(0, k, (std::min)(iu,k+3) + 1, 3).applyHouseholderOnTheRight(ess, tau, workspace); - if (computeU) - m_matU.block(0, k, size, 3).applyHouseholderOnTheRight(ess, tau, workspace); - } - } - - Matrix v = m_matT.template block<2,1>(iu-1, iu-2); - Scalar tau, beta; - Matrix ess; - v.makeHouseholder(ess, tau, beta); - - if (beta != Scalar(0)) // if v is not zero - { - m_matT.coeffRef(iu-1, iu-2) = beta; - m_matT.block(iu-1, iu-1, 2, size-iu+1).applyHouseholderOnTheLeft(ess, tau, workspace); - m_matT.block(0, iu-1, iu+1, 2).applyHouseholderOnTheRight(ess, tau, workspace); - if (computeU) - m_matU.block(0, iu-1, size, 2).applyHouseholderOnTheRight(ess, tau, workspace); - } - - // clean up pollution due to round-off errors - for (Index i = im+2; i <= iu; ++i) - { - m_matT.coeffRef(i,i-2) = Scalar(0); - if (i > im+2) - m_matT.coeffRef(i,i-3) = Scalar(0); - } -} - -} // end namespace Eigen - -#endif // EIGEN_REAL_SCHUR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur_LAPACKE.h deleted file mode 100644 index 2c22517155e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/RealSchur_LAPACKE.h +++ /dev/null @@ -1,77 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * Real Schur needed to real unsymmetrical eigenvalues/eigenvectors. - ******************************************************************************** -*/ - -#ifndef EIGEN_REAL_SCHUR_LAPACKE_H -#define EIGEN_REAL_SCHUR_LAPACKE_H - -namespace Eigen { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_SCHUR_REAL(EIGTYPE, LAPACKE_TYPE, LAPACKE_PREFIX, LAPACKE_PREFIX_U, EIGCOLROW, LAPACKE_COLROW) \ -template<> template inline \ -RealSchur >& \ -RealSchur >::compute(const EigenBase& matrix, bool computeU) \ -{ \ - eigen_assert(matrix.cols() == matrix.rows()); \ -\ - lapack_int n = internal::convert_index(matrix.cols()), sdim, info; \ - lapack_int matrix_order = LAPACKE_COLROW; \ - char jobvs, sort='N'; \ - LAPACK_##LAPACKE_PREFIX_U##_SELECT2 select = 0; \ - jobvs = (computeU) ? 'V' : 'N'; \ - m_matU.resize(n, n); \ - lapack_int ldvs = internal::convert_index(m_matU.outerStride()); \ - m_matT = matrix; \ - lapack_int lda = internal::convert_index(m_matT.outerStride()); \ - Matrix wr, wi; \ - wr.resize(n, 1); wi.resize(n, 1); \ - info = LAPACKE_##LAPACKE_PREFIX##gees( matrix_order, jobvs, sort, select, n, (LAPACKE_TYPE*)m_matT.data(), lda, &sdim, (LAPACKE_TYPE*)wr.data(), (LAPACKE_TYPE*)wi.data(), (LAPACKE_TYPE*)m_matU.data(), ldvs ); \ - if(info == 0) \ - m_info = Success; \ - else \ - m_info = NoConvergence; \ -\ - m_isInitialized = true; \ - m_matUisUptodate = computeU; \ - return *this; \ -\ -} - -EIGEN_LAPACKE_SCHUR_REAL(double, double, d, D, ColMajor, LAPACK_COL_MAJOR) -EIGEN_LAPACKE_SCHUR_REAL(float, float, s, S, ColMajor, LAPACK_COL_MAJOR) -EIGEN_LAPACKE_SCHUR_REAL(double, double, d, D, RowMajor, LAPACK_ROW_MAJOR) -EIGEN_LAPACKE_SCHUR_REAL(float, float, s, S, RowMajor, LAPACK_ROW_MAJOR) - -} // end namespace Eigen - -#endif // EIGEN_REAL_SCHUR_LAPACKE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h deleted file mode 100644 index d37656fa207..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h +++ /dev/null @@ -1,871 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2010 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINTEIGENSOLVER_H -#define EIGEN_SELFADJOINTEIGENSOLVER_H - -#include "./Tridiagonalization.h" - -namespace Eigen { - -template -class GeneralizedSelfAdjointEigenSolver; - -namespace internal { -template struct direct_selfadjoint_eigenvalues; -template -ComputationInfo computeFromTridiagonal_impl(DiagType& diag, SubDiagType& subdiag, const Index maxIterations, bool computeEigenvectors, MatrixType& eivec); -} - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class SelfAdjointEigenSolver - * - * \brief Computes eigenvalues and eigenvectors of selfadjoint matrices - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * eigendecomposition; this is expected to be an instantiation of the Matrix - * class template. - * - * A matrix \f$ A \f$ is selfadjoint if it equals its adjoint. For real - * matrices, this means that the matrix is symmetric: it equals its - * transpose. This class computes the eigenvalues and eigenvectors of a - * selfadjoint matrix. These are the scalars \f$ \lambda \f$ and vectors - * \f$ v \f$ such that \f$ Av = \lambda v \f$. The eigenvalues of a - * selfadjoint matrix are always real. If \f$ D \f$ is a diagonal matrix with - * the eigenvalues on the diagonal, and \f$ V \f$ is a matrix with the - * eigenvectors as its columns, then \f$ A = V D V^{-1} \f$ (for selfadjoint - * matrices, the matrix \f$ V \f$ is always invertible). This is called the - * eigendecomposition. - * - * The algorithm exploits the fact that the matrix is selfadjoint, making it - * faster and more accurate than the general purpose eigenvalue algorithms - * implemented in EigenSolver and ComplexEigenSolver. - * - * Only the \b lower \b triangular \b part of the input matrix is referenced. - * - * Call the function compute() to compute the eigenvalues and eigenvectors of - * a given matrix. Alternatively, you can use the - * SelfAdjointEigenSolver(const MatrixType&, int) constructor which computes - * the eigenvalues and eigenvectors at construction time. Once the eigenvalue - * and eigenvectors are computed, they can be retrieved with the eigenvalues() - * and eigenvectors() functions. - * - * The documentation for SelfAdjointEigenSolver(const MatrixType&, int) - * contains an example of the typical use of this class. - * - * To solve the \em generalized eigenvalue problem \f$ Av = \lambda Bv \f$ and - * the likes, see the class GeneralizedSelfAdjointEigenSolver. - * - * \sa MatrixBase::eigenvalues(), class EigenSolver, class ComplexEigenSolver - */ -template class SelfAdjointEigenSolver -{ - public: - - typedef _MatrixType MatrixType; - enum { - Size = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - Options = MatrixType::Options, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - /** \brief Scalar type for matrices of type \p _MatrixType. */ - typedef typename MatrixType::Scalar Scalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - typedef Matrix EigenvectorsType; - - /** \brief Real scalar type for \p _MatrixType. - * - * This is just \c Scalar if #Scalar is real (e.g., \c float or - * \c double), and the type of the real part of \c Scalar if #Scalar is - * complex. - */ - typedef typename NumTraits::Real RealScalar; - - friend struct internal::direct_selfadjoint_eigenvalues::IsComplex>; - - /** \brief Type for vector of eigenvalues as returned by eigenvalues(). - * - * This is a column vector with entries of type #RealScalar. - * The length of the vector is the size of \p _MatrixType. - */ - typedef typename internal::plain_col_type::type RealVectorType; - typedef Tridiagonalization TridiagonalizationType; - typedef typename TridiagonalizationType::SubDiagonalType SubDiagonalType; - - /** \brief Default constructor for fixed-size matrices. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). This constructor - * can only be used if \p _MatrixType is a fixed-size matrix; use - * SelfAdjointEigenSolver(Index) for dynamic-size matrices. - * - * Example: \include SelfAdjointEigenSolver_SelfAdjointEigenSolver.cpp - * Output: \verbinclude SelfAdjointEigenSolver_SelfAdjointEigenSolver.out - */ - EIGEN_DEVICE_FUNC - SelfAdjointEigenSolver() - : m_eivec(), - m_eivalues(), - m_subdiag(), - m_isInitialized(false) - { } - - /** \brief Constructor, pre-allocates memory for dynamic-size matrices. - * - * \param [in] size Positive integer, size of the matrix whose - * eigenvalues and eigenvectors will be computed. - * - * This constructor is useful for dynamic-size matrices, when the user - * intends to perform decompositions via compute(). The \p size - * parameter is only used as a hint. It is not an error to give a wrong - * \p size, but it may impair performance. - * - * \sa compute() for an example - */ - EIGEN_DEVICE_FUNC - explicit SelfAdjointEigenSolver(Index size) - : m_eivec(size, size), - m_eivalues(size), - m_subdiag(size > 1 ? size - 1 : 1), - m_isInitialized(false) - {} - - /** \brief Constructor; computes eigendecomposition of given matrix. - * - * \param[in] matrix Selfadjoint matrix whose eigendecomposition is to - * be computed. Only the lower triangular part of the matrix is referenced. - * \param[in] options Can be #ComputeEigenvectors (default) or #EigenvaluesOnly. - * - * This constructor calls compute(const MatrixType&, int) to compute the - * eigenvalues of the matrix \p matrix. The eigenvectors are computed if - * \p options equals #ComputeEigenvectors. - * - * Example: \include SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType.cpp - * Output: \verbinclude SelfAdjointEigenSolver_SelfAdjointEigenSolver_MatrixType.out - * - * \sa compute(const MatrixType&, int) - */ - template - EIGEN_DEVICE_FUNC - explicit SelfAdjointEigenSolver(const EigenBase& matrix, int options = ComputeEigenvectors) - : m_eivec(matrix.rows(), matrix.cols()), - m_eivalues(matrix.cols()), - m_subdiag(matrix.rows() > 1 ? matrix.rows() - 1 : 1), - m_isInitialized(false) - { - compute(matrix.derived(), options); - } - - /** \brief Computes eigendecomposition of given matrix. - * - * \param[in] matrix Selfadjoint matrix whose eigendecomposition is to - * be computed. Only the lower triangular part of the matrix is referenced. - * \param[in] options Can be #ComputeEigenvectors (default) or #EigenvaluesOnly. - * \returns Reference to \c *this - * - * This function computes the eigenvalues of \p matrix. The eigenvalues() - * function can be used to retrieve them. If \p options equals #ComputeEigenvectors, - * then the eigenvectors are also computed and can be retrieved by - * calling eigenvectors(). - * - * This implementation uses a symmetric QR algorithm. The matrix is first - * reduced to tridiagonal form using the Tridiagonalization class. The - * tridiagonal matrix is then brought to diagonal form with implicit - * symmetric QR steps with Wilkinson shift. Details can be found in - * Section 8.3 of Golub \& Van Loan, %Matrix Computations. - * - * The cost of the computation is about \f$ 9n^3 \f$ if the eigenvectors - * are required and \f$ 4n^3/3 \f$ if they are not required. - * - * This method reuses the memory in the SelfAdjointEigenSolver object that - * was allocated when the object was constructed, if the size of the - * matrix does not change. - * - * Example: \include SelfAdjointEigenSolver_compute_MatrixType.cpp - * Output: \verbinclude SelfAdjointEigenSolver_compute_MatrixType.out - * - * \sa SelfAdjointEigenSolver(const MatrixType&, int) - */ - template - EIGEN_DEVICE_FUNC - SelfAdjointEigenSolver& compute(const EigenBase& matrix, int options = ComputeEigenvectors); - - /** \brief Computes eigendecomposition of given matrix using a closed-form algorithm - * - * This is a variant of compute(const MatrixType&, int options) which - * directly solves the underlying polynomial equation. - * - * Currently only 2x2 and 3x3 matrices for which the sizes are known at compile time are supported (e.g., Matrix3d). - * - * This method is usually significantly faster than the QR iterative algorithm - * but it might also be less accurate. It is also worth noting that - * for 3x3 matrices it involves trigonometric operations which are - * not necessarily available for all scalar types. - * - * For the 3x3 case, we observed the following worst case relative error regarding the eigenvalues: - * - double: 1e-8 - * - float: 1e-3 - * - * \sa compute(const MatrixType&, int options) - */ - EIGEN_DEVICE_FUNC - SelfAdjointEigenSolver& computeDirect(const MatrixType& matrix, int options = ComputeEigenvectors); - - /** - *\brief Computes the eigen decomposition from a tridiagonal symmetric matrix - * - * \param[in] diag The vector containing the diagonal of the matrix. - * \param[in] subdiag The subdiagonal of the matrix. - * \param[in] options Can be #ComputeEigenvectors (default) or #EigenvaluesOnly. - * \returns Reference to \c *this - * - * This function assumes that the matrix has been reduced to tridiagonal form. - * - * \sa compute(const MatrixType&, int) for more information - */ - SelfAdjointEigenSolver& computeFromTridiagonal(const RealVectorType& diag, const SubDiagonalType& subdiag , int options=ComputeEigenvectors); - - /** \brief Returns the eigenvectors of given matrix. - * - * \returns A const reference to the matrix whose columns are the eigenvectors. - * - * \pre The eigenvectors have been computed before. - * - * Column \f$ k \f$ of the returned matrix is an eigenvector corresponding - * to eigenvalue number \f$ k \f$ as returned by eigenvalues(). The - * eigenvectors are normalized to have (Euclidean) norm equal to one. If - * this object was used to solve the eigenproblem for the selfadjoint - * matrix \f$ A \f$, then the matrix returned by this function is the - * matrix \f$ V \f$ in the eigendecomposition \f$ A = V D V^{-1} \f$. - * - * Example: \include SelfAdjointEigenSolver_eigenvectors.cpp - * Output: \verbinclude SelfAdjointEigenSolver_eigenvectors.out - * - * \sa eigenvalues() - */ - EIGEN_DEVICE_FUNC - const EigenvectorsType& eigenvectors() const - { - eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - return m_eivec; - } - - /** \brief Returns the eigenvalues of given matrix. - * - * \returns A const reference to the column vector containing the eigenvalues. - * - * \pre The eigenvalues have been computed before. - * - * The eigenvalues are repeated according to their algebraic multiplicity, - * so there are as many eigenvalues as rows in the matrix. The eigenvalues - * are sorted in increasing order. - * - * Example: \include SelfAdjointEigenSolver_eigenvalues.cpp - * Output: \verbinclude SelfAdjointEigenSolver_eigenvalues.out - * - * \sa eigenvectors(), MatrixBase::eigenvalues() - */ - EIGEN_DEVICE_FUNC - const RealVectorType& eigenvalues() const - { - eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); - return m_eivalues; - } - - /** \brief Computes the positive-definite square root of the matrix. - * - * \returns the positive-definite square root of the matrix - * - * \pre The eigenvalues and eigenvectors of a positive-definite matrix - * have been computed before. - * - * The square root of a positive-definite matrix \f$ A \f$ is the - * positive-definite matrix whose square equals \f$ A \f$. This function - * uses the eigendecomposition \f$ A = V D V^{-1} \f$ to compute the - * square root as \f$ A^{1/2} = V D^{1/2} V^{-1} \f$. - * - * Example: \include SelfAdjointEigenSolver_operatorSqrt.cpp - * Output: \verbinclude SelfAdjointEigenSolver_operatorSqrt.out - * - * \sa operatorInverseSqrt(), MatrixFunctions Module - */ - EIGEN_DEVICE_FUNC - MatrixType operatorSqrt() const - { - eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - return m_eivec * m_eivalues.cwiseSqrt().asDiagonal() * m_eivec.adjoint(); - } - - /** \brief Computes the inverse square root of the matrix. - * - * \returns the inverse positive-definite square root of the matrix - * - * \pre The eigenvalues and eigenvectors of a positive-definite matrix - * have been computed before. - * - * This function uses the eigendecomposition \f$ A = V D V^{-1} \f$ to - * compute the inverse square root as \f$ V D^{-1/2} V^{-1} \f$. This is - * cheaper than first computing the square root with operatorSqrt() and - * then its inverse with MatrixBase::inverse(). - * - * Example: \include SelfAdjointEigenSolver_operatorInverseSqrt.cpp - * Output: \verbinclude SelfAdjointEigenSolver_operatorInverseSqrt.out - * - * \sa operatorSqrt(), MatrixBase::inverse(), MatrixFunctions Module - */ - EIGEN_DEVICE_FUNC - MatrixType operatorInverseSqrt() const - { - eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); - eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); - return m_eivec * m_eivalues.cwiseInverse().cwiseSqrt().asDiagonal() * m_eivec.adjoint(); - } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, \c NoConvergence otherwise. - */ - EIGEN_DEVICE_FUNC - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); - return m_info; - } - - /** \brief Maximum number of iterations. - * - * The algorithm terminates if it does not converge within m_maxIterations * n iterations, where n - * denotes the size of the matrix. This value is currently set to 30 (copied from LAPACK). - */ - static const int m_maxIterations = 30; - - protected: - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - EigenvectorsType m_eivec; - RealVectorType m_eivalues; - typename TridiagonalizationType::SubDiagonalType m_subdiag; - ComputationInfo m_info; - bool m_isInitialized; - bool m_eigenvectorsOk; -}; - -namespace internal { -/** \internal - * - * \eigenvalues_module \ingroup Eigenvalues_Module - * - * Performs a QR step on a tridiagonal symmetric matrix represented as a - * pair of two vectors \a diag and \a subdiag. - * - * \param diag the diagonal part of the input selfadjoint tridiagonal matrix - * \param subdiag the sub-diagonal part of the input selfadjoint tridiagonal matrix - * \param start starting index of the submatrix to work on - * \param end last+1 index of the submatrix to work on - * \param matrixQ pointer to the column-major matrix holding the eigenvectors, can be 0 - * \param n size of the input matrix - * - * For compilation efficiency reasons, this procedure does not use eigen expression - * for its arguments. - * - * Implemented from Golub's "Matrix Computations", algorithm 8.3.2: - * "implicit symmetric QR step with Wilkinson shift" - */ -template -EIGEN_DEVICE_FUNC -static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n); -} - -template -template -EIGEN_DEVICE_FUNC -SelfAdjointEigenSolver& SelfAdjointEigenSolver -::compute(const EigenBase& a_matrix, int options) -{ - check_template_parameters(); - - const InputType &matrix(a_matrix.derived()); - - using std::abs; - eigen_assert(matrix.cols() == matrix.rows()); - eigen_assert((options&~(EigVecMask|GenEigMask))==0 - && (options&EigVecMask)!=EigVecMask - && "invalid option parameter"); - bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors; - Index n = matrix.cols(); - m_eivalues.resize(n,1); - - if(n==1) - { - m_eivec = matrix; - m_eivalues.coeffRef(0,0) = numext::real(m_eivec.coeff(0,0)); - if(computeEigenvectors) - m_eivec.setOnes(n,n); - m_info = Success; - m_isInitialized = true; - m_eigenvectorsOk = computeEigenvectors; - return *this; - } - - // declare some aliases - RealVectorType& diag = m_eivalues; - EigenvectorsType& mat = m_eivec; - - // map the matrix coefficients to [-1:1] to avoid over- and underflow. - mat = matrix.template triangularView(); - RealScalar scale = mat.cwiseAbs().maxCoeff(); - if(scale==RealScalar(0)) scale = RealScalar(1); - mat.template triangularView() /= scale; - m_subdiag.resize(n-1); - internal::tridiagonalization_inplace(mat, diag, m_subdiag, computeEigenvectors); - - m_info = internal::computeFromTridiagonal_impl(diag, m_subdiag, m_maxIterations, computeEigenvectors, m_eivec); - - // scale back the eigen values - m_eivalues *= scale; - - m_isInitialized = true; - m_eigenvectorsOk = computeEigenvectors; - return *this; -} - -template -SelfAdjointEigenSolver& SelfAdjointEigenSolver -::computeFromTridiagonal(const RealVectorType& diag, const SubDiagonalType& subdiag , int options) -{ - //TODO : Add an option to scale the values beforehand - bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors; - - m_eivalues = diag; - m_subdiag = subdiag; - if (computeEigenvectors) - { - m_eivec.setIdentity(diag.size(), diag.size()); - } - m_info = internal::computeFromTridiagonal_impl(m_eivalues, m_subdiag, m_maxIterations, computeEigenvectors, m_eivec); - - m_isInitialized = true; - m_eigenvectorsOk = computeEigenvectors; - return *this; -} - -namespace internal { -/** - * \internal - * \brief Compute the eigendecomposition from a tridiagonal matrix - * - * \param[in,out] diag : On input, the diagonal of the matrix, on output the eigenvalues - * \param[in,out] subdiag : The subdiagonal part of the matrix (entries are modified during the decomposition) - * \param[in] maxIterations : the maximum number of iterations - * \param[in] computeEigenvectors : whether the eigenvectors have to be computed or not - * \param[out] eivec : The matrix to store the eigenvectors if computeEigenvectors==true. Must be allocated on input. - * \returns \c Success or \c NoConvergence - */ -template -ComputationInfo computeFromTridiagonal_impl(DiagType& diag, SubDiagType& subdiag, const Index maxIterations, bool computeEigenvectors, MatrixType& eivec) -{ - using std::abs; - - ComputationInfo info; - typedef typename MatrixType::Scalar Scalar; - - Index n = diag.size(); - Index end = n-1; - Index start = 0; - Index iter = 0; // total number of iterations - - typedef typename DiagType::RealScalar RealScalar; - const RealScalar considerAsZero = (std::numeric_limits::min)(); - const RealScalar precision = RealScalar(2)*NumTraits::epsilon(); - - while (end>0) - { - for (Index i = start; i0 && subdiag[end-1]==RealScalar(0)) - { - end--; - } - if (end<=0) - break; - - // if we spent too many iterations, we give up - iter++; - if(iter > maxIterations * n) break; - - start = end - 1; - while (start>0 && subdiag[start-1]!=0) - start--; - - internal::tridiagonal_qr_step(diag.data(), subdiag.data(), start, end, computeEigenvectors ? eivec.data() : (Scalar*)0, n); - } - if (iter <= maxIterations * n) - info = Success; - else - info = NoConvergence; - - // Sort eigenvalues and corresponding vectors. - // TODO make the sort optional ? - // TODO use a better sort algorithm !! - if (info == Success) - { - for (Index i = 0; i < n-1; ++i) - { - Index k; - diag.segment(i,n-i).minCoeff(&k); - if (k > 0) - { - std::swap(diag[i], diag[k+i]); - if(computeEigenvectors) - eivec.col(i).swap(eivec.col(k+i)); - } - } - } - return info; -} - -template struct direct_selfadjoint_eigenvalues -{ - EIGEN_DEVICE_FUNC - static inline void run(SolverType& eig, const typename SolverType::MatrixType& A, int options) - { eig.compute(A,options); } -}; - -template struct direct_selfadjoint_eigenvalues -{ - typedef typename SolverType::MatrixType MatrixType; - typedef typename SolverType::RealVectorType VectorType; - typedef typename SolverType::Scalar Scalar; - typedef typename SolverType::EigenvectorsType EigenvectorsType; - - - /** \internal - * Computes the roots of the characteristic polynomial of \a m. - * For numerical stability m.trace() should be near zero and to avoid over- or underflow m should be normalized. - */ - EIGEN_DEVICE_FUNC - static inline void computeRoots(const MatrixType& m, VectorType& roots) - { - EIGEN_USING_STD_MATH(sqrt) - EIGEN_USING_STD_MATH(atan2) - EIGEN_USING_STD_MATH(cos) - EIGEN_USING_STD_MATH(sin) - const Scalar s_inv3 = Scalar(1)/Scalar(3); - const Scalar s_sqrt3 = sqrt(Scalar(3)); - - // The characteristic equation is x^3 - c2*x^2 + c1*x - c0 = 0. The - // eigenvalues are the roots to this equation, all guaranteed to be - // real-valued, because the matrix is symmetric. - Scalar c0 = m(0,0)*m(1,1)*m(2,2) + Scalar(2)*m(1,0)*m(2,0)*m(2,1) - m(0,0)*m(2,1)*m(2,1) - m(1,1)*m(2,0)*m(2,0) - m(2,2)*m(1,0)*m(1,0); - Scalar c1 = m(0,0)*m(1,1) - m(1,0)*m(1,0) + m(0,0)*m(2,2) - m(2,0)*m(2,0) + m(1,1)*m(2,2) - m(2,1)*m(2,1); - Scalar c2 = m(0,0) + m(1,1) + m(2,2); - - // Construct the parameters used in classifying the roots of the equation - // and in solving the equation for the roots in closed form. - Scalar c2_over_3 = c2*s_inv3; - Scalar a_over_3 = (c2*c2_over_3 - c1)*s_inv3; - a_over_3 = numext::maxi(a_over_3, Scalar(0)); - - Scalar half_b = Scalar(0.5)*(c0 + c2_over_3*(Scalar(2)*c2_over_3*c2_over_3 - c1)); - - Scalar q = a_over_3*a_over_3*a_over_3 - half_b*half_b; - q = numext::maxi(q, Scalar(0)); - - // Compute the eigenvalues by solving for the roots of the polynomial. - Scalar rho = sqrt(a_over_3); - Scalar theta = atan2(sqrt(q),half_b)*s_inv3; // since sqrt(q) > 0, atan2 is in [0, pi] and theta is in [0, pi/3] - Scalar cos_theta = cos(theta); - Scalar sin_theta = sin(theta); - // roots are already sorted, since cos is monotonically decreasing on [0, pi] - roots(0) = c2_over_3 - rho*(cos_theta + s_sqrt3*sin_theta); // == 2*rho*cos(theta+2pi/3) - roots(1) = c2_over_3 - rho*(cos_theta - s_sqrt3*sin_theta); // == 2*rho*cos(theta+ pi/3) - roots(2) = c2_over_3 + Scalar(2)*rho*cos_theta; - } - - EIGEN_DEVICE_FUNC - static inline bool extract_kernel(MatrixType& mat, Ref res, Ref representative) - { - EIGEN_USING_STD_MATH(sqrt) - EIGEN_USING_STD_MATH(abs) - Index i0; - // Find non-zero column i0 (by construction, there must exist a non zero coefficient on the diagonal): - mat.diagonal().cwiseAbs().maxCoeff(&i0); - // mat.col(i0) is a good candidate for an orthogonal vector to the current eigenvector, - // so let's save it: - representative = mat.col(i0); - Scalar n0, n1; - VectorType c0, c1; - n0 = (c0 = representative.cross(mat.col((i0+1)%3))).squaredNorm(); - n1 = (c1 = representative.cross(mat.col((i0+2)%3))).squaredNorm(); - if(n0>n1) res = c0/sqrt(n0); - else res = c1/sqrt(n1); - - return true; - } - - EIGEN_DEVICE_FUNC - static inline void run(SolverType& solver, const MatrixType& mat, int options) - { - eigen_assert(mat.cols() == 3 && mat.cols() == mat.rows()); - eigen_assert((options&~(EigVecMask|GenEigMask))==0 - && (options&EigVecMask)!=EigVecMask - && "invalid option parameter"); - bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors; - - EigenvectorsType& eivecs = solver.m_eivec; - VectorType& eivals = solver.m_eivalues; - - // Shift the matrix to the mean eigenvalue and map the matrix coefficients to [-1:1] to avoid over- and underflow. - Scalar shift = mat.trace() / Scalar(3); - // TODO Avoid this copy. Currently it is necessary to suppress bogus values when determining maxCoeff and for computing the eigenvectors later - MatrixType scaledMat = mat.template selfadjointView(); - scaledMat.diagonal().array() -= shift; - Scalar scale = scaledMat.cwiseAbs().maxCoeff(); - if(scale > 0) scaledMat /= scale; // TODO for scale==0 we could save the remaining operations - - // compute the eigenvalues - computeRoots(scaledMat,eivals); - - // compute the eigenvectors - if(computeEigenvectors) - { - if((eivals(2)-eivals(0))<=Eigen::NumTraits::epsilon()) - { - // All three eigenvalues are numerically the same - eivecs.setIdentity(); - } - else - { - MatrixType tmp; - tmp = scaledMat; - - // Compute the eigenvector of the most distinct eigenvalue - Scalar d0 = eivals(2) - eivals(1); - Scalar d1 = eivals(1) - eivals(0); - Index k(0), l(2); - if(d0 > d1) - { - numext::swap(k,l); - d0 = d1; - } - - // Compute the eigenvector of index k - { - tmp.diagonal().array () -= eivals(k); - // By construction, 'tmp' is of rank 2, and its kernel corresponds to the respective eigenvector. - extract_kernel(tmp, eivecs.col(k), eivecs.col(l)); - } - - // Compute eigenvector of index l - if(d0<=2*Eigen::NumTraits::epsilon()*d1) - { - // If d0 is too small, then the two other eigenvalues are numerically the same, - // and thus we only have to ortho-normalize the near orthogonal vector we saved above. - eivecs.col(l) -= eivecs.col(k).dot(eivecs.col(l))*eivecs.col(l); - eivecs.col(l).normalize(); - } - else - { - tmp = scaledMat; - tmp.diagonal().array () -= eivals(l); - - VectorType dummy; - extract_kernel(tmp, eivecs.col(l), dummy); - } - - // Compute last eigenvector from the other two - eivecs.col(1) = eivecs.col(2).cross(eivecs.col(0)).normalized(); - } - } - - // Rescale back to the original size. - eivals *= scale; - eivals.array() += shift; - - solver.m_info = Success; - solver.m_isInitialized = true; - solver.m_eigenvectorsOk = computeEigenvectors; - } -}; - -// 2x2 direct eigenvalues decomposition, code from Hauke Heibel -template -struct direct_selfadjoint_eigenvalues -{ - typedef typename SolverType::MatrixType MatrixType; - typedef typename SolverType::RealVectorType VectorType; - typedef typename SolverType::Scalar Scalar; - typedef typename SolverType::EigenvectorsType EigenvectorsType; - - EIGEN_DEVICE_FUNC - static inline void computeRoots(const MatrixType& m, VectorType& roots) - { - using std::sqrt; - const Scalar t0 = Scalar(0.5) * sqrt( numext::abs2(m(0,0)-m(1,1)) + Scalar(4)*numext::abs2(m(1,0))); - const Scalar t1 = Scalar(0.5) * (m(0,0) + m(1,1)); - roots(0) = t1 - t0; - roots(1) = t1 + t0; - } - - EIGEN_DEVICE_FUNC - static inline void run(SolverType& solver, const MatrixType& mat, int options) - { - EIGEN_USING_STD_MATH(sqrt); - EIGEN_USING_STD_MATH(abs); - - eigen_assert(mat.cols() == 2 && mat.cols() == mat.rows()); - eigen_assert((options&~(EigVecMask|GenEigMask))==0 - && (options&EigVecMask)!=EigVecMask - && "invalid option parameter"); - bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors; - - EigenvectorsType& eivecs = solver.m_eivec; - VectorType& eivals = solver.m_eivalues; - - // Shift the matrix to the mean eigenvalue and map the matrix coefficients to [-1:1] to avoid over- and underflow. - Scalar shift = mat.trace() / Scalar(2); - MatrixType scaledMat = mat; - scaledMat.coeffRef(0,1) = mat.coeff(1,0); - scaledMat.diagonal().array() -= shift; - Scalar scale = scaledMat.cwiseAbs().maxCoeff(); - if(scale > Scalar(0)) - scaledMat /= scale; - - // Compute the eigenvalues - computeRoots(scaledMat,eivals); - - // compute the eigen vectors - if(computeEigenvectors) - { - if((eivals(1)-eivals(0))<=abs(eivals(1))*Eigen::NumTraits::epsilon()) - { - eivecs.setIdentity(); - } - else - { - scaledMat.diagonal().array () -= eivals(1); - Scalar a2 = numext::abs2(scaledMat(0,0)); - Scalar c2 = numext::abs2(scaledMat(1,1)); - Scalar b2 = numext::abs2(scaledMat(1,0)); - if(a2>c2) - { - eivecs.col(1) << -scaledMat(1,0), scaledMat(0,0); - eivecs.col(1) /= sqrt(a2+b2); - } - else - { - eivecs.col(1) << -scaledMat(1,1), scaledMat(1,0); - eivecs.col(1) /= sqrt(c2+b2); - } - - eivecs.col(0) << eivecs.col(1).unitOrthogonal(); - } - } - - // Rescale back to the original size. - eivals *= scale; - eivals.array() += shift; - - solver.m_info = Success; - solver.m_isInitialized = true; - solver.m_eigenvectorsOk = computeEigenvectors; - } -}; - -} - -template -EIGEN_DEVICE_FUNC -SelfAdjointEigenSolver& SelfAdjointEigenSolver -::computeDirect(const MatrixType& matrix, int options) -{ - internal::direct_selfadjoint_eigenvalues::IsComplex>::run(*this,matrix,options); - return *this; -} - -namespace internal { -template -EIGEN_DEVICE_FUNC -static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n) -{ - using std::abs; - RealScalar td = (diag[end-1] - diag[end])*RealScalar(0.5); - RealScalar e = subdiag[end-1]; - // Note that thanks to scaling, e^2 or td^2 cannot overflow, however they can still - // underflow thus leading to inf/NaN values when using the following commented code: -// RealScalar e2 = numext::abs2(subdiag[end-1]); -// RealScalar mu = diag[end] - e2 / (td + (td>0 ? 1 : -1) * sqrt(td*td + e2)); - // This explain the following, somewhat more complicated, version: - RealScalar mu = diag[end]; - if(td==RealScalar(0)) - mu -= abs(e); - else - { - RealScalar e2 = numext::abs2(subdiag[end-1]); - RealScalar h = numext::hypot(td,e); - if(e2==RealScalar(0)) mu -= (e / (td + (td>RealScalar(0) ? RealScalar(1) : RealScalar(-1)))) * (e / h); - else mu -= e2 / (td + (td>RealScalar(0) ? h : -h)); - } - - RealScalar x = diag[start] - mu; - RealScalar z = subdiag[start]; - for (Index k = start; k < end; ++k) - { - JacobiRotation rot; - rot.makeGivens(x, z); - - // do T = G' T G - RealScalar sdk = rot.s() * diag[k] + rot.c() * subdiag[k]; - RealScalar dkp1 = rot.s() * subdiag[k] + rot.c() * diag[k+1]; - - diag[k] = rot.c() * (rot.c() * diag[k] - rot.s() * subdiag[k]) - rot.s() * (rot.c() * subdiag[k] - rot.s() * diag[k+1]); - diag[k+1] = rot.s() * sdk + rot.c() * dkp1; - subdiag[k] = rot.c() * sdk - rot.s() * dkp1; - - - if (k > start) - subdiag[k - 1] = rot.c() * subdiag[k-1] - rot.s() * z; - - x = subdiag[k]; - - if (k < end - 1) - { - z = -rot.s() * subdiag[k+1]; - subdiag[k + 1] = rot.c() * subdiag[k+1]; - } - - // apply the givens rotation to the unit matrix Q = Q * G - if (matrixQ) - { - // FIXME if StorageOrder == RowMajor this operation is not very efficient - Map > q(matrixQ,n,n); - q.applyOnTheRight(k,k+1,rot); - } - } -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINTEIGENSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h deleted file mode 100644 index b0c947dc07b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h +++ /dev/null @@ -1,87 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * Self-adjoint eigenvalues/eigenvectors. - ******************************************************************************** -*/ - -#ifndef EIGEN_SAEIGENSOLVER_LAPACKE_H -#define EIGEN_SAEIGENSOLVER_LAPACKE_H - -namespace Eigen { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_EIG_SELFADJ_2(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_NAME, EIGCOLROW ) \ -template<> template inline \ -SelfAdjointEigenSolver >& \ -SelfAdjointEigenSolver >::compute(const EigenBase& matrix, int options) \ -{ \ - eigen_assert(matrix.cols() == matrix.rows()); \ - eigen_assert((options&~(EigVecMask|GenEigMask))==0 \ - && (options&EigVecMask)!=EigVecMask \ - && "invalid option parameter"); \ - bool computeEigenvectors = (options&ComputeEigenvectors)==ComputeEigenvectors; \ - lapack_int n = internal::convert_index(matrix.cols()), lda, info; \ - m_eivalues.resize(n,1); \ - m_subdiag.resize(n-1); \ - m_eivec = matrix; \ -\ - if(n==1) \ - { \ - m_eivalues.coeffRef(0,0) = numext::real(m_eivec.coeff(0,0)); \ - if(computeEigenvectors) m_eivec.setOnes(n,n); \ - m_info = Success; \ - m_isInitialized = true; \ - m_eigenvectorsOk = computeEigenvectors; \ - return *this; \ - } \ -\ - lda = internal::convert_index(m_eivec.outerStride()); \ - char jobz, uplo='L'/*, range='A'*/; \ - jobz = computeEigenvectors ? 'V' : 'N'; \ -\ - info = LAPACKE_##LAPACKE_NAME( LAPACK_COL_MAJOR, jobz, uplo, n, (LAPACKE_TYPE*)m_eivec.data(), lda, (LAPACKE_RTYPE*)m_eivalues.data() ); \ - m_info = (info==0) ? Success : NoConvergence; \ - m_isInitialized = true; \ - m_eigenvectorsOk = computeEigenvectors; \ - return *this; \ -} - -#define EIGEN_LAPACKE_EIG_SELFADJ(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_NAME ) \ - EIGEN_LAPACKE_EIG_SELFADJ_2(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_NAME, ColMajor ) \ - EIGEN_LAPACKE_EIG_SELFADJ_2(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_NAME, RowMajor ) - -EIGEN_LAPACKE_EIG_SELFADJ(double, double, double, dsyev) -EIGEN_LAPACKE_EIG_SELFADJ(float, float, float, ssyev) -EIGEN_LAPACKE_EIG_SELFADJ(dcomplex, lapack_complex_double, double, zheev) -EIGEN_LAPACKE_EIG_SELFADJ(scomplex, lapack_complex_float, float, cheev) - -} // end namespace Eigen - -#endif // EIGEN_SAEIGENSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/Tridiagonalization.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/Tridiagonalization.h deleted file mode 100644 index 1d102c17bc1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Eigenvalues/Tridiagonalization.h +++ /dev/null @@ -1,556 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2010 Jitse Niesen -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIDIAGONALIZATION_H -#define EIGEN_TRIDIAGONALIZATION_H - -namespace Eigen { - -namespace internal { - -template struct TridiagonalizationMatrixTReturnType; -template -struct traits > - : public traits -{ - typedef typename MatrixType::PlainObject ReturnType; // FIXME shall it be a BandMatrix? - enum { Flags = 0 }; -}; - -template -void tridiagonalization_inplace(MatrixType& matA, CoeffVectorType& hCoeffs); -} - -/** \eigenvalues_module \ingroup Eigenvalues_Module - * - * - * \class Tridiagonalization - * - * \brief Tridiagonal decomposition of a selfadjoint matrix - * - * \tparam _MatrixType the type of the matrix of which we are computing the - * tridiagonal decomposition; this is expected to be an instantiation of the - * Matrix class template. - * - * This class performs a tridiagonal decomposition of a selfadjoint matrix \f$ A \f$ such that: - * \f$ A = Q T Q^* \f$ where \f$ Q \f$ is unitary and \f$ T \f$ a real symmetric tridiagonal matrix. - * - * A tridiagonal matrix is a matrix which has nonzero elements only on the - * main diagonal and the first diagonal below and above it. The Hessenberg - * decomposition of a selfadjoint matrix is in fact a tridiagonal - * decomposition. This class is used in SelfAdjointEigenSolver to compute the - * eigenvalues and eigenvectors of a selfadjoint matrix. - * - * Call the function compute() to compute the tridiagonal decomposition of a - * given matrix. Alternatively, you can use the Tridiagonalization(const MatrixType&) - * constructor which computes the tridiagonal Schur decomposition at - * construction time. Once the decomposition is computed, you can use the - * matrixQ() and matrixT() functions to retrieve the matrices Q and T in the - * decomposition. - * - * The documentation of Tridiagonalization(const MatrixType&) contains an - * example of the typical use of this class. - * - * \sa class HessenbergDecomposition, class SelfAdjointEigenSolver - */ -template class Tridiagonalization -{ - public: - - /** \brief Synonym for the template parameter \p _MatrixType. */ - typedef _MatrixType MatrixType; - - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - enum { - Size = MatrixType::RowsAtCompileTime, - SizeMinusOne = Size == Dynamic ? Dynamic : (Size > 1 ? Size - 1 : 1), - Options = MatrixType::Options, - MaxSize = MatrixType::MaxRowsAtCompileTime, - MaxSizeMinusOne = MaxSize == Dynamic ? Dynamic : (MaxSize > 1 ? MaxSize - 1 : 1) - }; - - typedef Matrix CoeffVectorType; - typedef typename internal::plain_col_type::type DiagonalType; - typedef Matrix SubDiagonalType; - typedef typename internal::remove_all::type MatrixTypeRealView; - typedef internal::TridiagonalizationMatrixTReturnType MatrixTReturnType; - - typedef typename internal::conditional::IsComplex, - typename internal::add_const_on_value_type::RealReturnType>::type, - const Diagonal - >::type DiagonalReturnType; - - typedef typename internal::conditional::IsComplex, - typename internal::add_const_on_value_type::RealReturnType>::type, - const Diagonal - >::type SubDiagonalReturnType; - - /** \brief Return type of matrixQ() */ - typedef HouseholderSequence::type> HouseholderSequenceType; - - /** \brief Default constructor. - * - * \param [in] size Positive integer, size of the matrix whose tridiagonal - * decomposition will be computed. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via compute(). The \p size parameter is only - * used as a hint. It is not an error to give a wrong \p size, but it may - * impair performance. - * - * \sa compute() for an example. - */ - explicit Tridiagonalization(Index size = Size==Dynamic ? 2 : Size) - : m_matrix(size,size), - m_hCoeffs(size > 1 ? size-1 : 1), - m_isInitialized(false) - {} - - /** \brief Constructor; computes tridiagonal decomposition of given matrix. - * - * \param[in] matrix Selfadjoint matrix whose tridiagonal decomposition - * is to be computed. - * - * This constructor calls compute() to compute the tridiagonal decomposition. - * - * Example: \include Tridiagonalization_Tridiagonalization_MatrixType.cpp - * Output: \verbinclude Tridiagonalization_Tridiagonalization_MatrixType.out - */ - template - explicit Tridiagonalization(const EigenBase& matrix) - : m_matrix(matrix.derived()), - m_hCoeffs(matrix.cols() > 1 ? matrix.cols()-1 : 1), - m_isInitialized(false) - { - internal::tridiagonalization_inplace(m_matrix, m_hCoeffs); - m_isInitialized = true; - } - - /** \brief Computes tridiagonal decomposition of given matrix. - * - * \param[in] matrix Selfadjoint matrix whose tridiagonal decomposition - * is to be computed. - * \returns Reference to \c *this - * - * The tridiagonal decomposition is computed by bringing the columns of - * the matrix successively in the required form using Householder - * reflections. The cost is \f$ 4n^3/3 \f$ flops, where \f$ n \f$ denotes - * the size of the given matrix. - * - * This method reuses of the allocated data in the Tridiagonalization - * object, if the size of the matrix does not change. - * - * Example: \include Tridiagonalization_compute.cpp - * Output: \verbinclude Tridiagonalization_compute.out - */ - template - Tridiagonalization& compute(const EigenBase& matrix) - { - m_matrix = matrix.derived(); - m_hCoeffs.resize(matrix.rows()-1, 1); - internal::tridiagonalization_inplace(m_matrix, m_hCoeffs); - m_isInitialized = true; - return *this; - } - - /** \brief Returns the Householder coefficients. - * - * \returns a const reference to the vector of Householder coefficients - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * The Householder coefficients allow the reconstruction of the matrix - * \f$ Q \f$ in the tridiagonal decomposition from the packed data. - * - * Example: \include Tridiagonalization_householderCoefficients.cpp - * Output: \verbinclude Tridiagonalization_householderCoefficients.out - * - * \sa packedMatrix(), \ref Householder_Module "Householder module" - */ - inline CoeffVectorType householderCoefficients() const - { - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return m_hCoeffs; - } - - /** \brief Returns the internal representation of the decomposition - * - * \returns a const reference to a matrix with the internal representation - * of the decomposition. - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * The returned matrix contains the following information: - * - the strict upper triangular part is equal to the input matrix A. - * - the diagonal and lower sub-diagonal represent the real tridiagonal - * symmetric matrix T. - * - the rest of the lower part contains the Householder vectors that, - * combined with Householder coefficients returned by - * householderCoefficients(), allows to reconstruct the matrix Q as - * \f$ Q = H_{N-1} \ldots H_1 H_0 \f$. - * Here, the matrices \f$ H_i \f$ are the Householder transformations - * \f$ H_i = (I - h_i v_i v_i^T) \f$ - * where \f$ h_i \f$ is the \f$ i \f$th Householder coefficient and - * \f$ v_i \f$ is the Householder vector defined by - * \f$ v_i = [ 0, \ldots, 0, 1, M(i+2,i), \ldots, M(N-1,i) ]^T \f$ - * with M the matrix returned by this function. - * - * See LAPACK for further details on this packed storage. - * - * Example: \include Tridiagonalization_packedMatrix.cpp - * Output: \verbinclude Tridiagonalization_packedMatrix.out - * - * \sa householderCoefficients() - */ - inline const MatrixType& packedMatrix() const - { - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return m_matrix; - } - - /** \brief Returns the unitary matrix Q in the decomposition - * - * \returns object representing the matrix Q - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * This function returns a light-weight object of template class - * HouseholderSequence. You can either apply it directly to a matrix or - * you can convert it to a matrix of type #MatrixType. - * - * \sa Tridiagonalization(const MatrixType&) for an example, - * matrixT(), class HouseholderSequence - */ - HouseholderSequenceType matrixQ() const - { - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return HouseholderSequenceType(m_matrix, m_hCoeffs.conjugate()) - .setLength(m_matrix.rows() - 1) - .setShift(1); - } - - /** \brief Returns an expression of the tridiagonal matrix T in the decomposition - * - * \returns expression object representing the matrix T - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * Currently, this function can be used to extract the matrix T from internal - * data and copy it to a dense matrix object. In most cases, it may be - * sufficient to directly use the packed matrix or the vector expressions - * returned by diagonal() and subDiagonal() instead of creating a new - * dense copy matrix with this function. - * - * \sa Tridiagonalization(const MatrixType&) for an example, - * matrixQ(), packedMatrix(), diagonal(), subDiagonal() - */ - MatrixTReturnType matrixT() const - { - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return MatrixTReturnType(m_matrix.real()); - } - - /** \brief Returns the diagonal of the tridiagonal matrix T in the decomposition. - * - * \returns expression representing the diagonal of T - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * Example: \include Tridiagonalization_diagonal.cpp - * Output: \verbinclude Tridiagonalization_diagonal.out - * - * \sa matrixT(), subDiagonal() - */ - DiagonalReturnType diagonal() const; - - /** \brief Returns the subdiagonal of the tridiagonal matrix T in the decomposition. - * - * \returns expression representing the subdiagonal of T - * - * \pre Either the constructor Tridiagonalization(const MatrixType&) or - * the member function compute(const MatrixType&) has been called before - * to compute the tridiagonal decomposition of a matrix. - * - * \sa diagonal() for an example, matrixT() - */ - SubDiagonalReturnType subDiagonal() const; - - protected: - - MatrixType m_matrix; - CoeffVectorType m_hCoeffs; - bool m_isInitialized; -}; - -template -typename Tridiagonalization::DiagonalReturnType -Tridiagonalization::diagonal() const -{ - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return m_matrix.diagonal().real(); -} - -template -typename Tridiagonalization::SubDiagonalReturnType -Tridiagonalization::subDiagonal() const -{ - eigen_assert(m_isInitialized && "Tridiagonalization is not initialized."); - return m_matrix.template diagonal<-1>().real(); -} - -namespace internal { - -/** \internal - * Performs a tridiagonal decomposition of the selfadjoint matrix \a matA in-place. - * - * \param[in,out] matA On input the selfadjoint matrix. Only the \b lower triangular part is referenced. - * On output, the strict upper part is left unchanged, and the lower triangular part - * represents the T and Q matrices in packed format has detailed below. - * \param[out] hCoeffs returned Householder coefficients (see below) - * - * On output, the tridiagonal selfadjoint matrix T is stored in the diagonal - * and lower sub-diagonal of the matrix \a matA. - * The unitary matrix Q is represented in a compact way as a product of - * Householder reflectors \f$ H_i \f$ such that: - * \f$ Q = H_{N-1} \ldots H_1 H_0 \f$. - * The Householder reflectors are defined as - * \f$ H_i = (I - h_i v_i v_i^T) \f$ - * where \f$ h_i = hCoeffs[i]\f$ is the \f$ i \f$th Householder coefficient and - * \f$ v_i \f$ is the Householder vector defined by - * \f$ v_i = [ 0, \ldots, 0, 1, matA(i+2,i), \ldots, matA(N-1,i) ]^T \f$. - * - * Implemented from Golub's "Matrix Computations", algorithm 8.3.1. - * - * \sa Tridiagonalization::packedMatrix() - */ -template -void tridiagonalization_inplace(MatrixType& matA, CoeffVectorType& hCoeffs) -{ - using numext::conj; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - Index n = matA.rows(); - eigen_assert(n==matA.cols()); - eigen_assert(n==hCoeffs.size()+1 || n==1); - - for (Index i = 0; i() - * (conj(h) * matA.col(i).tail(remainingSize))); - - hCoeffs.tail(n-i-1) += (conj(h)*RealScalar(-0.5)*(hCoeffs.tail(remainingSize).dot(matA.col(i).tail(remainingSize)))) * matA.col(i).tail(n-i-1); - - matA.bottomRightCorner(remainingSize, remainingSize).template selfadjointView() - .rankUpdate(matA.col(i).tail(remainingSize), hCoeffs.tail(remainingSize), Scalar(-1)); - - matA.col(i).coeffRef(i+1) = beta; - hCoeffs.coeffRef(i) = h; - } -} - -// forward declaration, implementation at the end of this file -template::IsComplex> -struct tridiagonalization_inplace_selector; - -/** \brief Performs a full tridiagonalization in place - * - * \param[in,out] mat On input, the selfadjoint matrix whose tridiagonal - * decomposition is to be computed. Only the lower triangular part referenced. - * The rest is left unchanged. On output, the orthogonal matrix Q - * in the decomposition if \p extractQ is true. - * \param[out] diag The diagonal of the tridiagonal matrix T in the - * decomposition. - * \param[out] subdiag The subdiagonal of the tridiagonal matrix T in - * the decomposition. - * \param[in] extractQ If true, the orthogonal matrix Q in the - * decomposition is computed and stored in \p mat. - * - * Computes the tridiagonal decomposition of the selfadjoint matrix \p mat in place - * such that \f$ mat = Q T Q^* \f$ where \f$ Q \f$ is unitary and \f$ T \f$ a real - * symmetric tridiagonal matrix. - * - * The tridiagonal matrix T is passed to the output parameters \p diag and \p subdiag. If - * \p extractQ is true, then the orthogonal matrix Q is passed to \p mat. Otherwise the lower - * part of the matrix \p mat is destroyed. - * - * The vectors \p diag and \p subdiag are not resized. The function - * assumes that they are already of the correct size. The length of the - * vector \p diag should equal the number of rows in \p mat, and the - * length of the vector \p subdiag should be one left. - * - * This implementation contains an optimized path for 3-by-3 matrices - * which is especially useful for plane fitting. - * - * \note Currently, it requires two temporary vectors to hold the intermediate - * Householder coefficients, and to reconstruct the matrix Q from the Householder - * reflectors. - * - * Example (this uses the same matrix as the example in - * Tridiagonalization::Tridiagonalization(const MatrixType&)): - * \include Tridiagonalization_decomposeInPlace.cpp - * Output: \verbinclude Tridiagonalization_decomposeInPlace.out - * - * \sa class Tridiagonalization - */ -template -void tridiagonalization_inplace(MatrixType& mat, DiagonalType& diag, SubDiagonalType& subdiag, bool extractQ) -{ - eigen_assert(mat.cols()==mat.rows() && diag.size()==mat.rows() && subdiag.size()==mat.rows()-1); - tridiagonalization_inplace_selector::run(mat, diag, subdiag, extractQ); -} - -/** \internal - * General full tridiagonalization - */ -template -struct tridiagonalization_inplace_selector -{ - typedef typename Tridiagonalization::CoeffVectorType CoeffVectorType; - typedef typename Tridiagonalization::HouseholderSequenceType HouseholderSequenceType; - template - static void run(MatrixType& mat, DiagonalType& diag, SubDiagonalType& subdiag, bool extractQ) - { - CoeffVectorType hCoeffs(mat.cols()-1); - tridiagonalization_inplace(mat,hCoeffs); - diag = mat.diagonal().real(); - subdiag = mat.template diagonal<-1>().real(); - if(extractQ) - mat = HouseholderSequenceType(mat, hCoeffs.conjugate()) - .setLength(mat.rows() - 1) - .setShift(1); - } -}; - -/** \internal - * Specialization for 3x3 real matrices. - * Especially useful for plane fitting. - */ -template -struct tridiagonalization_inplace_selector -{ - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - - template - static void run(MatrixType& mat, DiagonalType& diag, SubDiagonalType& subdiag, bool extractQ) - { - using std::sqrt; - const RealScalar tol = (std::numeric_limits::min)(); - diag[0] = mat(0,0); - RealScalar v1norm2 = numext::abs2(mat(2,0)); - if(v1norm2 <= tol) - { - diag[1] = mat(1,1); - diag[2] = mat(2,2); - subdiag[0] = mat(1,0); - subdiag[1] = mat(2,1); - if (extractQ) - mat.setIdentity(); - } - else - { - RealScalar beta = sqrt(numext::abs2(mat(1,0)) + v1norm2); - RealScalar invBeta = RealScalar(1)/beta; - Scalar m01 = mat(1,0) * invBeta; - Scalar m02 = mat(2,0) * invBeta; - Scalar q = RealScalar(2)*m01*mat(2,1) + m02*(mat(2,2) - mat(1,1)); - diag[1] = mat(1,1) + m02*q; - diag[2] = mat(2,2) - m02*q; - subdiag[0] = beta; - subdiag[1] = mat(2,1) - m01 * q; - if (extractQ) - { - mat << 1, 0, 0, - 0, m01, m02, - 0, m02, -m01; - } - } - } -}; - -/** \internal - * Trivial specialization for 1x1 matrices - */ -template -struct tridiagonalization_inplace_selector -{ - typedef typename MatrixType::Scalar Scalar; - - template - static void run(MatrixType& mat, DiagonalType& diag, SubDiagonalType&, bool extractQ) - { - diag(0,0) = numext::real(mat(0,0)); - if(extractQ) - mat(0,0) = Scalar(1); - } -}; - -/** \internal - * \eigenvalues_module \ingroup Eigenvalues_Module - * - * \brief Expression type for return value of Tridiagonalization::matrixT() - * - * \tparam MatrixType type of underlying dense matrix - */ -template struct TridiagonalizationMatrixTReturnType -: public ReturnByValue > -{ - public: - /** \brief Constructor. - * - * \param[in] mat The underlying dense matrix - */ - TridiagonalizationMatrixTReturnType(const MatrixType& mat) : m_matrix(mat) { } - - template - inline void evalTo(ResultType& result) const - { - result.setZero(); - result.template diagonal<1>() = m_matrix.template diagonal<-1>().conjugate(); - result.diagonal() = m_matrix.diagonal(); - result.template diagonal<-1>() = m_matrix.template diagonal<-1>(); - } - - Index rows() const { return m_matrix.rows(); } - Index cols() const { return m_matrix.cols(); } - - protected: - typename MatrixType::Nested m_matrix; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIDIAGONALIZATION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AlignedBox.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AlignedBox.h deleted file mode 100644 index 066eae4f921..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AlignedBox.h +++ /dev/null @@ -1,392 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ALIGNEDBOX_H -#define EIGEN_ALIGNEDBOX_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * - * \class AlignedBox - * - * \brief An axis aligned box - * - * \tparam _Scalar the type of the scalar coefficients - * \tparam _AmbientDim the dimension of the ambient space, can be a compile time value or Dynamic. - * - * This class represents an axis aligned box as a pair of the minimal and maximal corners. - * \warning The result of most methods is undefined when applied to an empty box. You can check for empty boxes using isEmpty(). - * \sa alignedboxtypedefs - */ -template -class AlignedBox -{ -public: -EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_AmbientDim) - enum { AmbientDimAtCompileTime = _AmbientDim }; - typedef _Scalar Scalar; - typedef NumTraits ScalarTraits; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef typename ScalarTraits::Real RealScalar; - typedef typename ScalarTraits::NonInteger NonInteger; - typedef Matrix VectorType; - typedef CwiseBinaryOp, const VectorType, const VectorType> VectorTypeSum; - - /** Define constants to name the corners of a 1D, 2D or 3D axis aligned bounding box */ - enum CornerType - { - /** 1D names @{ */ - Min=0, Max=1, - /** @} */ - - /** Identifier for 2D corner @{ */ - BottomLeft=0, BottomRight=1, - TopLeft=2, TopRight=3, - /** @} */ - - /** Identifier for 3D corner @{ */ - BottomLeftFloor=0, BottomRightFloor=1, - TopLeftFloor=2, TopRightFloor=3, - BottomLeftCeil=4, BottomRightCeil=5, - TopLeftCeil=6, TopRightCeil=7 - /** @} */ - }; - - - /** Default constructor initializing a null box. */ - EIGEN_DEVICE_FUNC inline AlignedBox() - { if (AmbientDimAtCompileTime!=Dynamic) setEmpty(); } - - /** Constructs a null box with \a _dim the dimension of the ambient space. */ - EIGEN_DEVICE_FUNC inline explicit AlignedBox(Index _dim) : m_min(_dim), m_max(_dim) - { setEmpty(); } - - /** Constructs a box with extremities \a _min and \a _max. - * \warning If either component of \a _min is larger than the same component of \a _max, the constructed box is empty. */ - template - EIGEN_DEVICE_FUNC inline AlignedBox(const OtherVectorType1& _min, const OtherVectorType2& _max) : m_min(_min), m_max(_max) {} - - /** Constructs a box containing a single point \a p. */ - template - EIGEN_DEVICE_FUNC inline explicit AlignedBox(const MatrixBase& p) : m_min(p), m_max(m_min) - { } - - EIGEN_DEVICE_FUNC ~AlignedBox() {} - - /** \returns the dimension in which the box holds */ - EIGEN_DEVICE_FUNC inline Index dim() const { return AmbientDimAtCompileTime==Dynamic ? m_min.size() : Index(AmbientDimAtCompileTime); } - - /** \deprecated use isEmpty() */ - EIGEN_DEVICE_FUNC inline bool isNull() const { return isEmpty(); } - - /** \deprecated use setEmpty() */ - EIGEN_DEVICE_FUNC inline void setNull() { setEmpty(); } - - /** \returns true if the box is empty. - * \sa setEmpty */ - EIGEN_DEVICE_FUNC inline bool isEmpty() const { return (m_min.array() > m_max.array()).any(); } - - /** Makes \c *this an empty box. - * \sa isEmpty */ - EIGEN_DEVICE_FUNC inline void setEmpty() - { - m_min.setConstant( ScalarTraits::highest() ); - m_max.setConstant( ScalarTraits::lowest() ); - } - - /** \returns the minimal corner */ - EIGEN_DEVICE_FUNC inline const VectorType& (min)() const { return m_min; } - /** \returns a non const reference to the minimal corner */ - EIGEN_DEVICE_FUNC inline VectorType& (min)() { return m_min; } - /** \returns the maximal corner */ - EIGEN_DEVICE_FUNC inline const VectorType& (max)() const { return m_max; } - /** \returns a non const reference to the maximal corner */ - EIGEN_DEVICE_FUNC inline VectorType& (max)() { return m_max; } - - /** \returns the center of the box */ - EIGEN_DEVICE_FUNC inline const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(VectorTypeSum, RealScalar, quotient) - center() const - { return (m_min+m_max)/RealScalar(2); } - - /** \returns the lengths of the sides of the bounding box. - * Note that this function does not get the same - * result for integral or floating scalar types: see - */ - EIGEN_DEVICE_FUNC inline const CwiseBinaryOp< internal::scalar_difference_op, const VectorType, const VectorType> sizes() const - { return m_max - m_min; } - - /** \returns the volume of the bounding box */ - EIGEN_DEVICE_FUNC inline Scalar volume() const - { return sizes().prod(); } - - /** \returns an expression for the bounding box diagonal vector - * if the length of the diagonal is needed: diagonal().norm() - * will provide it. - */ - EIGEN_DEVICE_FUNC inline CwiseBinaryOp< internal::scalar_difference_op, const VectorType, const VectorType> diagonal() const - { return sizes(); } - - /** \returns the vertex of the bounding box at the corner defined by - * the corner-id corner. It works only for a 1D, 2D or 3D bounding box. - * For 1D bounding boxes corners are named by 2 enum constants: - * BottomLeft and BottomRight. - * For 2D bounding boxes, corners are named by 4 enum constants: - * BottomLeft, BottomRight, TopLeft, TopRight. - * For 3D bounding boxes, the following names are added: - * BottomLeftCeil, BottomRightCeil, TopLeftCeil, TopRightCeil. - */ - EIGEN_DEVICE_FUNC inline VectorType corner(CornerType corner) const - { - EIGEN_STATIC_ASSERT(_AmbientDim <= 3, THIS_METHOD_IS_ONLY_FOR_VECTORS_OF_A_SPECIFIC_SIZE); - - VectorType res; - - Index mult = 1; - for(Index d=0; d(Scalar(0), Scalar(1)); - } - else - r[d] = internal::random(m_min[d], m_max[d]); - } - return r; - } - - /** \returns true if the point \a p is inside the box \c *this. */ - template - EIGEN_DEVICE_FUNC inline bool contains(const MatrixBase& p) const - { - typename internal::nested_eval::type p_n(p.derived()); - return (m_min.array()<=p_n.array()).all() && (p_n.array()<=m_max.array()).all(); - } - - /** \returns true if the box \a b is entirely inside the box \c *this. */ - EIGEN_DEVICE_FUNC inline bool contains(const AlignedBox& b) const - { return (m_min.array()<=(b.min)().array()).all() && ((b.max)().array()<=m_max.array()).all(); } - - /** \returns true if the box \a b is intersecting the box \c *this. - * \sa intersection, clamp */ - EIGEN_DEVICE_FUNC inline bool intersects(const AlignedBox& b) const - { return (m_min.array()<=(b.max)().array()).all() && ((b.min)().array()<=m_max.array()).all(); } - - /** Extends \c *this such that it contains the point \a p and returns a reference to \c *this. - * \sa extend(const AlignedBox&) */ - template - EIGEN_DEVICE_FUNC inline AlignedBox& extend(const MatrixBase& p) - { - typename internal::nested_eval::type p_n(p.derived()); - m_min = m_min.cwiseMin(p_n); - m_max = m_max.cwiseMax(p_n); - return *this; - } - - /** Extends \c *this such that it contains the box \a b and returns a reference to \c *this. - * \sa merged, extend(const MatrixBase&) */ - EIGEN_DEVICE_FUNC inline AlignedBox& extend(const AlignedBox& b) - { - m_min = m_min.cwiseMin(b.m_min); - m_max = m_max.cwiseMax(b.m_max); - return *this; - } - - /** Clamps \c *this by the box \a b and returns a reference to \c *this. - * \note If the boxes don't intersect, the resulting box is empty. - * \sa intersection(), intersects() */ - EIGEN_DEVICE_FUNC inline AlignedBox& clamp(const AlignedBox& b) - { - m_min = m_min.cwiseMax(b.m_min); - m_max = m_max.cwiseMin(b.m_max); - return *this; - } - - /** Returns an AlignedBox that is the intersection of \a b and \c *this - * \note If the boxes don't intersect, the resulting box is empty. - * \sa intersects(), clamp, contains() */ - EIGEN_DEVICE_FUNC inline AlignedBox intersection(const AlignedBox& b) const - {return AlignedBox(m_min.cwiseMax(b.m_min), m_max.cwiseMin(b.m_max)); } - - /** Returns an AlignedBox that is the union of \a b and \c *this. - * \note Merging with an empty box may result in a box bigger than \c *this. - * \sa extend(const AlignedBox&) */ - EIGEN_DEVICE_FUNC inline AlignedBox merged(const AlignedBox& b) const - { return AlignedBox(m_min.cwiseMin(b.m_min), m_max.cwiseMax(b.m_max)); } - - /** Translate \c *this by the vector \a t and returns a reference to \c *this. */ - template - EIGEN_DEVICE_FUNC inline AlignedBox& translate(const MatrixBase& a_t) - { - const typename internal::nested_eval::type t(a_t.derived()); - m_min += t; - m_max += t; - return *this; - } - - /** \returns the squared distance between the point \a p and the box \c *this, - * and zero if \a p is inside the box. - * \sa exteriorDistance(const MatrixBase&), squaredExteriorDistance(const AlignedBox&) - */ - template - EIGEN_DEVICE_FUNC inline Scalar squaredExteriorDistance(const MatrixBase& p) const; - - /** \returns the squared distance between the boxes \a b and \c *this, - * and zero if the boxes intersect. - * \sa exteriorDistance(const AlignedBox&), squaredExteriorDistance(const MatrixBase&) - */ - EIGEN_DEVICE_FUNC inline Scalar squaredExteriorDistance(const AlignedBox& b) const; - - /** \returns the distance between the point \a p and the box \c *this, - * and zero if \a p is inside the box. - * \sa squaredExteriorDistance(const MatrixBase&), exteriorDistance(const AlignedBox&) - */ - template - EIGEN_DEVICE_FUNC inline NonInteger exteriorDistance(const MatrixBase& p) const - { EIGEN_USING_STD_MATH(sqrt) return sqrt(NonInteger(squaredExteriorDistance(p))); } - - /** \returns the distance between the boxes \a b and \c *this, - * and zero if the boxes intersect. - * \sa squaredExteriorDistance(const AlignedBox&), exteriorDistance(const MatrixBase&) - */ - EIGEN_DEVICE_FUNC inline NonInteger exteriorDistance(const AlignedBox& b) const - { EIGEN_USING_STD_MATH(sqrt) return sqrt(NonInteger(squaredExteriorDistance(b))); } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { - return typename internal::cast_return_type >::type(*this); - } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit AlignedBox(const AlignedBox& other) - { - m_min = (other.min)().template cast(); - m_max = (other.max)().template cast(); - } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const AlignedBox& other, const RealScalar& prec = ScalarTraits::dummy_precision()) const - { return m_min.isApprox(other.m_min, prec) && m_max.isApprox(other.m_max, prec); } - -protected: - - VectorType m_min, m_max; -}; - - - -template -template -EIGEN_DEVICE_FUNC inline Scalar AlignedBox::squaredExteriorDistance(const MatrixBase& a_p) const -{ - typename internal::nested_eval::type p(a_p.derived()); - Scalar dist2(0); - Scalar aux; - for (Index k=0; k p[k] ) - { - aux = m_min[k] - p[k]; - dist2 += aux*aux; - } - else if( p[k] > m_max[k] ) - { - aux = p[k] - m_max[k]; - dist2 += aux*aux; - } - } - return dist2; -} - -template -EIGEN_DEVICE_FUNC inline Scalar AlignedBox::squaredExteriorDistance(const AlignedBox& b) const -{ - Scalar dist2(0); - Scalar aux; - for (Index k=0; k b.m_max[k] ) - { - aux = m_min[k] - b.m_max[k]; - dist2 += aux*aux; - } - else if( b.m_min[k] > m_max[k] ) - { - aux = b.m_min[k] - m_max[k]; - dist2 += aux*aux; - } - } - return dist2; -} - -/** \defgroup alignedboxtypedefs Global aligned box typedefs - * - * \ingroup Geometry_Module - * - * Eigen defines several typedef shortcuts for most common aligned box types. - * - * The general patterns are the following: - * - * \c AlignedBoxSizeType where \c Size can be \c 1, \c 2,\c 3,\c 4 for fixed size boxes or \c X for dynamic size, - * and where \c Type can be \c i for integer, \c f for float, \c d for double. - * - * For example, \c AlignedBox3d is a fixed-size 3x3 aligned box type of doubles, and \c AlignedBoxXf is a dynamic-size aligned box of floats. - * - * \sa class AlignedBox - */ - -#define EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Size, SizeSuffix) \ -/** \ingroup alignedboxtypedefs */ \ -typedef AlignedBox AlignedBox##SizeSuffix##TypeSuffix; - -#define EIGEN_MAKE_TYPEDEFS_ALL_SIZES(Type, TypeSuffix) \ -EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 1, 1) \ -EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 2, 2) \ -EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 3, 3) \ -EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 4, 4) \ -EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Dynamic, X) - -EIGEN_MAKE_TYPEDEFS_ALL_SIZES(int, i) -EIGEN_MAKE_TYPEDEFS_ALL_SIZES(float, f) -EIGEN_MAKE_TYPEDEFS_ALL_SIZES(double, d) - -#undef EIGEN_MAKE_TYPEDEFS_ALL_SIZES -#undef EIGEN_MAKE_TYPEDEFS - -} // end namespace Eigen - -#endif // EIGEN_ALIGNEDBOX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AngleAxis.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AngleAxis.h deleted file mode 100644 index 83ee1be4616..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/AngleAxis.h +++ /dev/null @@ -1,247 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ANGLEAXIS_H -#define EIGEN_ANGLEAXIS_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class AngleAxis - * - * \brief Represents a 3D rotation as a rotation angle around an arbitrary 3D axis - * - * \param _Scalar the scalar type, i.e., the type of the coefficients. - * - * \warning When setting up an AngleAxis object, the axis vector \b must \b be \b normalized. - * - * The following two typedefs are provided for convenience: - * \li \c AngleAxisf for \c float - * \li \c AngleAxisd for \c double - * - * Combined with MatrixBase::Unit{X,Y,Z}, AngleAxis can be used to easily - * mimic Euler-angles. Here is an example: - * \include AngleAxis_mimic_euler.cpp - * Output: \verbinclude AngleAxis_mimic_euler.out - * - * \note This class is not aimed to be used to store a rotation transformation, - * but rather to make easier the creation of other rotation (Quaternion, rotation Matrix) - * and transformation objects. - * - * \sa class Quaternion, class Transform, MatrixBase::UnitX() - */ - -namespace internal { -template struct traits > -{ - typedef _Scalar Scalar; -}; -} - -template -class AngleAxis : public RotationBase,3> -{ - typedef RotationBase,3> Base; - -public: - - using Base::operator*; - - enum { Dim = 3 }; - /** the scalar type of the coefficients */ - typedef _Scalar Scalar; - typedef Matrix Matrix3; - typedef Matrix Vector3; - typedef Quaternion QuaternionType; - -protected: - - Vector3 m_axis; - Scalar m_angle; - -public: - - /** Default constructor without initialization. */ - EIGEN_DEVICE_FUNC AngleAxis() {} - /** Constructs and initialize the angle-axis rotation from an \a angle in radian - * and an \a axis which \b must \b be \b normalized. - * - * \warning If the \a axis vector is not normalized, then the angle-axis object - * represents an invalid rotation. */ - template - EIGEN_DEVICE_FUNC - inline AngleAxis(const Scalar& angle, const MatrixBase& axis) : m_axis(axis), m_angle(angle) {} - /** Constructs and initialize the angle-axis rotation from a quaternion \a q. - * This function implicitly normalizes the quaternion \a q. - */ - template - EIGEN_DEVICE_FUNC inline explicit AngleAxis(const QuaternionBase& q) { *this = q; } - /** Constructs and initialize the angle-axis rotation from a 3x3 rotation matrix. */ - template - EIGEN_DEVICE_FUNC inline explicit AngleAxis(const MatrixBase& m) { *this = m; } - - /** \returns the value of the rotation angle in radian */ - EIGEN_DEVICE_FUNC Scalar angle() const { return m_angle; } - /** \returns a read-write reference to the stored angle in radian */ - EIGEN_DEVICE_FUNC Scalar& angle() { return m_angle; } - - /** \returns the rotation axis */ - EIGEN_DEVICE_FUNC const Vector3& axis() const { return m_axis; } - /** \returns a read-write reference to the stored rotation axis. - * - * \warning The rotation axis must remain a \b unit vector. - */ - EIGEN_DEVICE_FUNC Vector3& axis() { return m_axis; } - - /** Concatenates two rotations */ - EIGEN_DEVICE_FUNC inline QuaternionType operator* (const AngleAxis& other) const - { return QuaternionType(*this) * QuaternionType(other); } - - /** Concatenates two rotations */ - EIGEN_DEVICE_FUNC inline QuaternionType operator* (const QuaternionType& other) const - { return QuaternionType(*this) * other; } - - /** Concatenates two rotations */ - friend EIGEN_DEVICE_FUNC inline QuaternionType operator* (const QuaternionType& a, const AngleAxis& b) - { return a * QuaternionType(b); } - - /** \returns the inverse rotation, i.e., an angle-axis with opposite rotation angle */ - EIGEN_DEVICE_FUNC AngleAxis inverse() const - { return AngleAxis(-m_angle, m_axis); } - - template - EIGEN_DEVICE_FUNC AngleAxis& operator=(const QuaternionBase& q); - template - EIGEN_DEVICE_FUNC AngleAxis& operator=(const MatrixBase& m); - - template - EIGEN_DEVICE_FUNC AngleAxis& fromRotationMatrix(const MatrixBase& m); - EIGEN_DEVICE_FUNC Matrix3 toRotationMatrix(void) const; - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { return typename internal::cast_return_type >::type(*this); } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit AngleAxis(const AngleAxis& other) - { - m_axis = other.axis().template cast(); - m_angle = Scalar(other.angle()); - } - - EIGEN_DEVICE_FUNC static inline const AngleAxis Identity() { return AngleAxis(Scalar(0), Vector3::UnitX()); } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const AngleAxis& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return m_axis.isApprox(other.m_axis, prec) && internal::isApprox(m_angle,other.m_angle, prec); } -}; - -/** \ingroup Geometry_Module - * single precision angle-axis type */ -typedef AngleAxis AngleAxisf; -/** \ingroup Geometry_Module - * double precision angle-axis type */ -typedef AngleAxis AngleAxisd; - -/** Set \c *this from a \b unit quaternion. - * - * The resulting axis is normalized, and the computed angle is in the [0,pi] range. - * - * This function implicitly normalizes the quaternion \a q. - */ -template -template -EIGEN_DEVICE_FUNC AngleAxis& AngleAxis::operator=(const QuaternionBase& q) -{ - EIGEN_USING_STD_MATH(atan2) - EIGEN_USING_STD_MATH(abs) - Scalar n = q.vec().norm(); - if(n::epsilon()) - n = q.vec().stableNorm(); - - if (n != Scalar(0)) - { - m_angle = Scalar(2)*atan2(n, abs(q.w())); - if(q.w() < Scalar(0)) - n = -n; - m_axis = q.vec() / n; - } - else - { - m_angle = Scalar(0); - m_axis << Scalar(1), Scalar(0), Scalar(0); - } - return *this; -} - -/** Set \c *this from a 3x3 rotation matrix \a mat. - */ -template -template -EIGEN_DEVICE_FUNC AngleAxis& AngleAxis::operator=(const MatrixBase& mat) -{ - // Since a direct conversion would not be really faster, - // let's use the robust Quaternion implementation: - return *this = QuaternionType(mat); -} - -/** -* \brief Sets \c *this from a 3x3 rotation matrix. -**/ -template -template -EIGEN_DEVICE_FUNC AngleAxis& AngleAxis::fromRotationMatrix(const MatrixBase& mat) -{ - return *this = QuaternionType(mat); -} - -/** Constructs and \returns an equivalent 3x3 rotation matrix. - */ -template -typename AngleAxis::Matrix3 -EIGEN_DEVICE_FUNC AngleAxis::toRotationMatrix(void) const -{ - EIGEN_USING_STD_MATH(sin) - EIGEN_USING_STD_MATH(cos) - Matrix3 res; - Vector3 sin_axis = sin(m_angle) * m_axis; - Scalar c = cos(m_angle); - Vector3 cos1_axis = (Scalar(1)-c) * m_axis; - - Scalar tmp; - tmp = cos1_axis.x() * m_axis.y(); - res.coeffRef(0,1) = tmp - sin_axis.z(); - res.coeffRef(1,0) = tmp + sin_axis.z(); - - tmp = cos1_axis.x() * m_axis.z(); - res.coeffRef(0,2) = tmp + sin_axis.y(); - res.coeffRef(2,0) = tmp - sin_axis.y(); - - tmp = cos1_axis.y() * m_axis.z(); - res.coeffRef(1,2) = tmp - sin_axis.x(); - res.coeffRef(2,1) = tmp + sin_axis.x(); - - res.diagonal() = (cos1_axis.cwiseProduct(m_axis)).array() + c; - - return res; -} - -} // end namespace Eigen - -#endif // EIGEN_ANGLEAXIS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/EulerAngles.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/EulerAngles.h deleted file mode 100644 index c633268af2c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/EulerAngles.h +++ /dev/null @@ -1,114 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_EULERANGLES_H -#define EIGEN_EULERANGLES_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * - * \returns the Euler-angles of the rotation matrix \c *this using the convention defined by the triplet (\a a0,\a a1,\a a2) - * - * Each of the three parameters \a a0,\a a1,\a a2 represents the respective rotation axis as an integer in {0,1,2}. - * For instance, in: - * \code Vector3f ea = mat.eulerAngles(2, 0, 2); \endcode - * "2" represents the z axis and "0" the x axis, etc. The returned angles are such that - * we have the following equality: - * \code - * mat == AngleAxisf(ea[0], Vector3f::UnitZ()) - * * AngleAxisf(ea[1], Vector3f::UnitX()) - * * AngleAxisf(ea[2], Vector3f::UnitZ()); \endcode - * This corresponds to the right-multiply conventions (with right hand side frames). - * - * The returned angles are in the ranges [0:pi]x[-pi:pi]x[-pi:pi]. - * - * \sa class AngleAxis - */ -template -EIGEN_DEVICE_FUNC inline Matrix::Scalar,3,1> -MatrixBase::eulerAngles(Index a0, Index a1, Index a2) const -{ - EIGEN_USING_STD_MATH(atan2) - EIGEN_USING_STD_MATH(sin) - EIGEN_USING_STD_MATH(cos) - /* Implemented from Graphics Gems IV */ - EIGEN_STATIC_ASSERT_MATRIX_SPECIFIC_SIZE(Derived,3,3) - - Matrix res; - typedef Matrix Vector2; - - const Index odd = ((a0+1)%3 == a1) ? 0 : 1; - const Index i = a0; - const Index j = (a0 + 1 + odd)%3; - const Index k = (a0 + 2 - odd)%3; - - if (a0==a2) - { - res[0] = atan2(coeff(j,i), coeff(k,i)); - if((odd && res[0]Scalar(0))) - { - if(res[0] > Scalar(0)) { - res[0] -= Scalar(EIGEN_PI); - } - else { - res[0] += Scalar(EIGEN_PI); - } - Scalar s2 = Vector2(coeff(j,i), coeff(k,i)).norm(); - res[1] = -atan2(s2, coeff(i,i)); - } - else - { - Scalar s2 = Vector2(coeff(j,i), coeff(k,i)).norm(); - res[1] = atan2(s2, coeff(i,i)); - } - - // With a=(0,1,0), we have i=0; j=1; k=2, and after computing the first two angles, - // we can compute their respective rotation, and apply its inverse to M. Since the result must - // be a rotation around x, we have: - // - // c2 s1.s2 c1.s2 1 0 0 - // 0 c1 -s1 * M = 0 c3 s3 - // -s2 s1.c2 c1.c2 0 -s3 c3 - // - // Thus: m11.c1 - m21.s1 = c3 & m12.c1 - m22.s1 = s3 - - Scalar s1 = sin(res[0]); - Scalar c1 = cos(res[0]); - res[2] = atan2(c1*coeff(j,k)-s1*coeff(k,k), c1*coeff(j,j) - s1 * coeff(k,j)); - } - else - { - res[0] = atan2(coeff(j,k), coeff(k,k)); - Scalar c2 = Vector2(coeff(i,i), coeff(i,j)).norm(); - if((odd && res[0]Scalar(0))) { - if(res[0] > Scalar(0)) { - res[0] -= Scalar(EIGEN_PI); - } - else { - res[0] += Scalar(EIGEN_PI); - } - res[1] = atan2(-coeff(i,k), -c2); - } - else - res[1] = atan2(-coeff(i,k), c2); - Scalar s1 = sin(res[0]); - Scalar c1 = cos(res[0]); - res[2] = atan2(s1*coeff(k,i)-c1*coeff(j,i), c1*coeff(j,j) - s1 * coeff(k,j)); - } - if (!odd) - res = -res; - - return res; -} - -} // end namespace Eigen - -#endif // EIGEN_EULERANGLES_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Homogeneous.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Homogeneous.h deleted file mode 100644 index 5f0da1a9e86..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Homogeneous.h +++ /dev/null @@ -1,497 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HOMOGENEOUS_H -#define EIGEN_HOMOGENEOUS_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class Homogeneous - * - * \brief Expression of one (or a set of) homogeneous vector(s) - * - * \param MatrixType the type of the object in which we are making homogeneous - * - * This class represents an expression of one (or a set of) homogeneous vector(s). - * It is the return type of MatrixBase::homogeneous() and most of the time - * this is the only way it is used. - * - * \sa MatrixBase::homogeneous() - */ - -namespace internal { - -template -struct traits > - : traits -{ - typedef typename traits::StorageKind StorageKind; - typedef typename ref_selector::type MatrixTypeNested; - typedef typename remove_reference::type _MatrixTypeNested; - enum { - RowsPlusOne = (MatrixType::RowsAtCompileTime != Dynamic) ? - int(MatrixType::RowsAtCompileTime) + 1 : Dynamic, - ColsPlusOne = (MatrixType::ColsAtCompileTime != Dynamic) ? - int(MatrixType::ColsAtCompileTime) + 1 : Dynamic, - RowsAtCompileTime = Direction==Vertical ? RowsPlusOne : MatrixType::RowsAtCompileTime, - ColsAtCompileTime = Direction==Horizontal ? ColsPlusOne : MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = RowsAtCompileTime, - MaxColsAtCompileTime = ColsAtCompileTime, - TmpFlags = _MatrixTypeNested::Flags & HereditaryBits, - Flags = ColsAtCompileTime==1 ? (TmpFlags & ~RowMajorBit) - : RowsAtCompileTime==1 ? (TmpFlags | RowMajorBit) - : TmpFlags - }; -}; - -template struct homogeneous_left_product_impl; -template struct homogeneous_right_product_impl; - -} // end namespace internal - -template class Homogeneous - : public MatrixBase >, internal::no_assignment_operator -{ - public: - - typedef MatrixType NestedExpression; - enum { Direction = _Direction }; - - typedef MatrixBase Base; - EIGEN_DENSE_PUBLIC_INTERFACE(Homogeneous) - - EIGEN_DEVICE_FUNC explicit inline Homogeneous(const MatrixType& matrix) - : m_matrix(matrix) - {} - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_matrix.rows() + (int(Direction)==Vertical ? 1 : 0); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_matrix.cols() + (int(Direction)==Horizontal ? 1 : 0); } - - EIGEN_DEVICE_FUNC const NestedExpression& nestedExpression() const { return m_matrix; } - - template - EIGEN_DEVICE_FUNC inline const Product - operator* (const MatrixBase& rhs) const - { - eigen_assert(int(Direction)==Horizontal); - return Product(*this,rhs.derived()); - } - - template friend - EIGEN_DEVICE_FUNC inline const Product - operator* (const MatrixBase& lhs, const Homogeneous& rhs) - { - eigen_assert(int(Direction)==Vertical); - return Product(lhs.derived(),rhs); - } - - template friend - EIGEN_DEVICE_FUNC inline const Product, Homogeneous > - operator* (const Transform& lhs, const Homogeneous& rhs) - { - eigen_assert(int(Direction)==Vertical); - return Product, Homogeneous>(lhs,rhs); - } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::result_of::type - redux(const Func& func) const - { - return func(m_matrix.redux(func), Scalar(1)); - } - - protected: - typename MatrixType::Nested m_matrix; -}; - -/** \geometry_module \ingroup Geometry_Module - * - * \returns a vector expression that is one longer than the vector argument, with the value 1 symbolically appended as the last coefficient. - * - * This can be used to convert affine coordinates to homogeneous coordinates. - * - * \only_for_vectors - * - * Example: \include MatrixBase_homogeneous.cpp - * Output: \verbinclude MatrixBase_homogeneous.out - * - * \sa VectorwiseOp::homogeneous(), class Homogeneous - */ -template -EIGEN_DEVICE_FUNC inline typename MatrixBase::HomogeneousReturnType -MatrixBase::homogeneous() const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - return HomogeneousReturnType(derived()); -} - -/** \geometry_module \ingroup Geometry_Module - * - * \returns an expression where the value 1 is symbolically appended as the final coefficient to each column (or row) of the matrix. - * - * This can be used to convert affine coordinates to homogeneous coordinates. - * - * Example: \include VectorwiseOp_homogeneous.cpp - * Output: \verbinclude VectorwiseOp_homogeneous.out - * - * \sa MatrixBase::homogeneous(), class Homogeneous */ -template -EIGEN_DEVICE_FUNC inline Homogeneous -VectorwiseOp::homogeneous() const -{ - return HomogeneousReturnType(_expression()); -} - -/** \geometry_module \ingroup Geometry_Module - * - * \brief homogeneous normalization - * - * \returns a vector expression of the N-1 first coefficients of \c *this divided by that last coefficient. - * - * This can be used to convert homogeneous coordinates to affine coordinates. - * - * It is essentially a shortcut for: - * \code - this->head(this->size()-1)/this->coeff(this->size()-1); - \endcode - * - * Example: \include MatrixBase_hnormalized.cpp - * Output: \verbinclude MatrixBase_hnormalized.out - * - * \sa VectorwiseOp::hnormalized() */ -template -EIGEN_DEVICE_FUNC inline const typename MatrixBase::HNormalizedReturnType -MatrixBase::hnormalized() const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - return ConstStartMinusOne(derived(),0,0, - ColsAtCompileTime==1?size()-1:1, - ColsAtCompileTime==1?1:size()-1) / coeff(size()-1); -} - -/** \geometry_module \ingroup Geometry_Module - * - * \brief column or row-wise homogeneous normalization - * - * \returns an expression of the first N-1 coefficients of each column (or row) of \c *this divided by the last coefficient of each column (or row). - * - * This can be used to convert homogeneous coordinates to affine coordinates. - * - * It is conceptually equivalent to calling MatrixBase::hnormalized() to each column (or row) of \c *this. - * - * Example: \include DirectionWise_hnormalized.cpp - * Output: \verbinclude DirectionWise_hnormalized.out - * - * \sa MatrixBase::hnormalized() */ -template -EIGEN_DEVICE_FUNC inline const typename VectorwiseOp::HNormalizedReturnType -VectorwiseOp::hnormalized() const -{ - return HNormalized_Block(_expression(),0,0, - Direction==Vertical ? _expression().rows()-1 : _expression().rows(), - Direction==Horizontal ? _expression().cols()-1 : _expression().cols()).cwiseQuotient( - Replicate - (HNormalized_Factors(_expression(), - Direction==Vertical ? _expression().rows()-1:0, - Direction==Horizontal ? _expression().cols()-1:0, - Direction==Vertical ? 1 : _expression().rows(), - Direction==Horizontal ? 1 : _expression().cols()), - Direction==Vertical ? _expression().rows()-1 : 1, - Direction==Horizontal ? _expression().cols()-1 : 1)); -} - -namespace internal { - -template -struct take_matrix_for_product -{ - typedef MatrixOrTransformType type; - EIGEN_DEVICE_FUNC static const type& run(const type &x) { return x; } -}; - -template -struct take_matrix_for_product > -{ - typedef Transform TransformType; - typedef typename internal::add_const::type type; - EIGEN_DEVICE_FUNC static type run (const TransformType& x) { return x.affine(); } -}; - -template -struct take_matrix_for_product > -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType type; - EIGEN_DEVICE_FUNC static const type& run (const TransformType& x) { return x.matrix(); } -}; - -template -struct traits,Lhs> > -{ - typedef typename take_matrix_for_product::type LhsMatrixType; - typedef typename remove_all::type MatrixTypeCleaned; - typedef typename remove_all::type LhsMatrixTypeCleaned; - typedef typename make_proper_matrix_type< - typename traits::Scalar, - LhsMatrixTypeCleaned::RowsAtCompileTime, - MatrixTypeCleaned::ColsAtCompileTime, - MatrixTypeCleaned::PlainObject::Options, - LhsMatrixTypeCleaned::MaxRowsAtCompileTime, - MatrixTypeCleaned::MaxColsAtCompileTime>::type ReturnType; -}; - -template -struct homogeneous_left_product_impl,Lhs> - : public ReturnByValue,Lhs> > -{ - typedef typename traits::LhsMatrixType LhsMatrixType; - typedef typename remove_all::type LhsMatrixTypeCleaned; - typedef typename remove_all::type LhsMatrixTypeNested; - EIGEN_DEVICE_FUNC homogeneous_left_product_impl(const Lhs& lhs, const MatrixType& rhs) - : m_lhs(take_matrix_for_product::run(lhs)), - m_rhs(rhs) - {} - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_lhs.rows(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_rhs.cols(); } - - template EIGEN_DEVICE_FUNC void evalTo(Dest& dst) const - { - // FIXME investigate how to allow lazy evaluation of this product when possible - dst = Block - (m_lhs,0,0,m_lhs.rows(),m_lhs.cols()-1) * m_rhs; - dst += m_lhs.col(m_lhs.cols()-1).rowwise() - .template replicate(m_rhs.cols()); - } - - typename LhsMatrixTypeCleaned::Nested m_lhs; - typename MatrixType::Nested m_rhs; -}; - -template -struct traits,Rhs> > -{ - typedef typename make_proper_matrix_type::Scalar, - MatrixType::RowsAtCompileTime, - Rhs::ColsAtCompileTime, - MatrixType::PlainObject::Options, - MatrixType::MaxRowsAtCompileTime, - Rhs::MaxColsAtCompileTime>::type ReturnType; -}; - -template -struct homogeneous_right_product_impl,Rhs> - : public ReturnByValue,Rhs> > -{ - typedef typename remove_all::type RhsNested; - EIGEN_DEVICE_FUNC homogeneous_right_product_impl(const MatrixType& lhs, const Rhs& rhs) - : m_lhs(lhs), m_rhs(rhs) - {} - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_lhs.rows(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_rhs.cols(); } - - template EIGEN_DEVICE_FUNC void evalTo(Dest& dst) const - { - // FIXME investigate how to allow lazy evaluation of this product when possible - dst = m_lhs * Block - (m_rhs,0,0,m_rhs.rows()-1,m_rhs.cols()); - dst += m_rhs.row(m_rhs.rows()-1).colwise() - .template replicate(m_lhs.rows()); - } - - typename MatrixType::Nested m_lhs; - typename Rhs::Nested m_rhs; -}; - -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef HomogeneousShape Shape; -}; - -template<> struct AssignmentKind { typedef Dense2Dense Kind; }; - - -template -struct unary_evaluator, IndexBased> - : evaluator::PlainObject > -{ - typedef Homogeneous XprType; - typedef typename XprType::PlainObject PlainObject; - typedef evaluator Base; - - EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& op) - : Base(), m_temp(op) - { - ::new (static_cast(this)) Base(m_temp); - } - -protected: - PlainObject m_temp; -}; - -// dense = homogeneous -template< typename DstXprType, typename ArgType, typename Scalar> -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Homogeneous SrcXprType; - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - dst.template topRows(src.nestedExpression().rows()) = src.nestedExpression(); - dst.row(dst.rows()-1).setOnes(); - } -}; - -// dense = homogeneous -template< typename DstXprType, typename ArgType, typename Scalar> -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Homogeneous SrcXprType; - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - dst.template leftCols(src.nestedExpression().cols()) = src.nestedExpression(); - dst.col(dst.cols()-1).setOnes(); - } -}; - -template -struct generic_product_impl, Rhs, HomogeneousShape, DenseShape, ProductTag> -{ - template - EIGEN_DEVICE_FUNC static void evalTo(Dest& dst, const Homogeneous& lhs, const Rhs& rhs) - { - homogeneous_right_product_impl, Rhs>(lhs.nestedExpression(), rhs).evalTo(dst); - } -}; - -template -struct homogeneous_right_product_refactoring_helper -{ - enum { - Dim = Lhs::ColsAtCompileTime, - Rows = Lhs::RowsAtCompileTime - }; - typedef typename Rhs::template ConstNRowsBlockXpr::Type LinearBlockConst; - typedef typename remove_const::type LinearBlock; - typedef typename Rhs::ConstRowXpr ConstantColumn; - typedef Replicate ConstantBlock; - typedef Product LinearProduct; - typedef CwiseBinaryOp, const LinearProduct, const ConstantBlock> Xpr; -}; - -template -struct product_evaluator, ProductTag, HomogeneousShape, DenseShape> - : public evaluator::Xpr> -{ - typedef Product XprType; - typedef homogeneous_right_product_refactoring_helper helper; - typedef typename helper::ConstantBlock ConstantBlock; - typedef typename helper::Xpr RefactoredXpr; - typedef evaluator Base; - - EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) - : Base( xpr.lhs().nestedExpression() .lazyProduct( xpr.rhs().template topRows(xpr.lhs().nestedExpression().cols()) ) - + ConstantBlock(xpr.rhs().row(xpr.rhs().rows()-1),xpr.lhs().rows(), 1) ) - {} -}; - -template -struct generic_product_impl, DenseShape, HomogeneousShape, ProductTag> -{ - template - EIGEN_DEVICE_FUNC static void evalTo(Dest& dst, const Lhs& lhs, const Homogeneous& rhs) - { - homogeneous_left_product_impl, Lhs>(lhs, rhs.nestedExpression()).evalTo(dst); - } -}; - -// TODO: the following specialization is to address a regression from 3.2 to 3.3 -// In the future, this path should be optimized. -template -struct generic_product_impl, TriangularShape, HomogeneousShape, ProductTag> -{ - template - static void evalTo(Dest& dst, const Lhs& lhs, const Homogeneous& rhs) - { - dst.noalias() = lhs * rhs.eval(); - } -}; - -template -struct homogeneous_left_product_refactoring_helper -{ - enum { - Dim = Rhs::RowsAtCompileTime, - Cols = Rhs::ColsAtCompileTime - }; - typedef typename Lhs::template ConstNColsBlockXpr::Type LinearBlockConst; - typedef typename remove_const::type LinearBlock; - typedef typename Lhs::ConstColXpr ConstantColumn; - typedef Replicate ConstantBlock; - typedef Product LinearProduct; - typedef CwiseBinaryOp, const LinearProduct, const ConstantBlock> Xpr; -}; - -template -struct product_evaluator, ProductTag, DenseShape, HomogeneousShape> - : public evaluator::Xpr> -{ - typedef Product XprType; - typedef homogeneous_left_product_refactoring_helper helper; - typedef typename helper::ConstantBlock ConstantBlock; - typedef typename helper::Xpr RefactoredXpr; - typedef evaluator Base; - - EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) - : Base( xpr.lhs().template leftCols(xpr.rhs().nestedExpression().rows()) .lazyProduct( xpr.rhs().nestedExpression() ) - + ConstantBlock(xpr.lhs().col(xpr.lhs().cols()-1),1,xpr.rhs().cols()) ) - {} -}; - -template -struct generic_product_impl, Homogeneous, DenseShape, HomogeneousShape, ProductTag> -{ - typedef Transform TransformType; - template - EIGEN_DEVICE_FUNC static void evalTo(Dest& dst, const TransformType& lhs, const Homogeneous& rhs) - { - homogeneous_left_product_impl, TransformType>(lhs, rhs.nestedExpression()).evalTo(dst); - } -}; - -template -struct permutation_matrix_product - : public permutation_matrix_product -{}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_HOMOGENEOUS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Hyperplane.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Hyperplane.h deleted file mode 100644 index 05929b29941..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Hyperplane.h +++ /dev/null @@ -1,282 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HYPERPLANE_H -#define EIGEN_HYPERPLANE_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class Hyperplane - * - * \brief A hyperplane - * - * A hyperplane is an affine subspace of dimension n-1 in a space of dimension n. - * For example, a hyperplane in a plane is a line; a hyperplane in 3-space is a plane. - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients - * \tparam _AmbientDim the dimension of the ambient space, can be a compile time value or Dynamic. - * Notice that the dimension of the hyperplane is _AmbientDim-1. - * - * This class represents an hyperplane as the zero set of the implicit equation - * \f$ n \cdot x + d = 0 \f$ where \f$ n \f$ is a unit normal vector of the plane (linear part) - * and \f$ d \f$ is the distance (offset) to the origin. - */ -template -class Hyperplane -{ -public: - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_AmbientDim==Dynamic ? Dynamic : _AmbientDim+1) - enum { - AmbientDimAtCompileTime = _AmbientDim, - Options = _Options - }; - typedef _Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef Matrix VectorType; - typedef Matrix Coefficients; - typedef Block NormalReturnType; - typedef const Block ConstNormalReturnType; - - /** Default constructor without initialization */ - EIGEN_DEVICE_FUNC inline Hyperplane() {} - - template - EIGEN_DEVICE_FUNC Hyperplane(const Hyperplane& other) - : m_coeffs(other.coeffs()) - {} - - /** Constructs a dynamic-size hyperplane with \a _dim the dimension - * of the ambient space */ - EIGEN_DEVICE_FUNC inline explicit Hyperplane(Index _dim) : m_coeffs(_dim+1) {} - - /** Construct a plane from its normal \a n and a point \a e onto the plane. - * \warning the vector normal is assumed to be normalized. - */ - EIGEN_DEVICE_FUNC inline Hyperplane(const VectorType& n, const VectorType& e) - : m_coeffs(n.size()+1) - { - normal() = n; - offset() = -n.dot(e); - } - - /** Constructs a plane from its normal \a n and distance to the origin \a d - * such that the algebraic equation of the plane is \f$ n \cdot x + d = 0 \f$. - * \warning the vector normal is assumed to be normalized. - */ - EIGEN_DEVICE_FUNC inline Hyperplane(const VectorType& n, const Scalar& d) - : m_coeffs(n.size()+1) - { - normal() = n; - offset() = d; - } - - /** Constructs a hyperplane passing through the two points. If the dimension of the ambient space - * is greater than 2, then there isn't uniqueness, so an arbitrary choice is made. - */ - EIGEN_DEVICE_FUNC static inline Hyperplane Through(const VectorType& p0, const VectorType& p1) - { - Hyperplane result(p0.size()); - result.normal() = (p1 - p0).unitOrthogonal(); - result.offset() = -p0.dot(result.normal()); - return result; - } - - /** Constructs a hyperplane passing through the three points. The dimension of the ambient space - * is required to be exactly 3. - */ - EIGEN_DEVICE_FUNC static inline Hyperplane Through(const VectorType& p0, const VectorType& p1, const VectorType& p2) - { - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(VectorType, 3) - Hyperplane result(p0.size()); - VectorType v0(p2 - p0), v1(p1 - p0); - result.normal() = v0.cross(v1); - RealScalar norm = result.normal().norm(); - if(norm <= v0.norm() * v1.norm() * NumTraits::epsilon()) - { - Matrix m; m << v0.transpose(), v1.transpose(); - JacobiSVD > svd(m, ComputeFullV); - result.normal() = svd.matrixV().col(2); - } - else - result.normal() /= norm; - result.offset() = -p0.dot(result.normal()); - return result; - } - - /** Constructs a hyperplane passing through the parametrized line \a parametrized. - * If the dimension of the ambient space is greater than 2, then there isn't uniqueness, - * so an arbitrary choice is made. - */ - // FIXME to be consitent with the rest this could be implemented as a static Through function ?? - EIGEN_DEVICE_FUNC explicit Hyperplane(const ParametrizedLine& parametrized) - { - normal() = parametrized.direction().unitOrthogonal(); - offset() = -parametrized.origin().dot(normal()); - } - - EIGEN_DEVICE_FUNC ~Hyperplane() {} - - /** \returns the dimension in which the plane holds */ - EIGEN_DEVICE_FUNC inline Index dim() const { return AmbientDimAtCompileTime==Dynamic ? m_coeffs.size()-1 : Index(AmbientDimAtCompileTime); } - - /** normalizes \c *this */ - EIGEN_DEVICE_FUNC void normalize(void) - { - m_coeffs /= normal().norm(); - } - - /** \returns the signed distance between the plane \c *this and a point \a p. - * \sa absDistance() - */ - EIGEN_DEVICE_FUNC inline Scalar signedDistance(const VectorType& p) const { return normal().dot(p) + offset(); } - - /** \returns the absolute distance between the plane \c *this and a point \a p. - * \sa signedDistance() - */ - EIGEN_DEVICE_FUNC inline Scalar absDistance(const VectorType& p) const { return numext::abs(signedDistance(p)); } - - /** \returns the projection of a point \a p onto the plane \c *this. - */ - EIGEN_DEVICE_FUNC inline VectorType projection(const VectorType& p) const { return p - signedDistance(p) * normal(); } - - /** \returns a constant reference to the unit normal vector of the plane, which corresponds - * to the linear part of the implicit equation. - */ - EIGEN_DEVICE_FUNC inline ConstNormalReturnType normal() const { return ConstNormalReturnType(m_coeffs,0,0,dim(),1); } - - /** \returns a non-constant reference to the unit normal vector of the plane, which corresponds - * to the linear part of the implicit equation. - */ - EIGEN_DEVICE_FUNC inline NormalReturnType normal() { return NormalReturnType(m_coeffs,0,0,dim(),1); } - - /** \returns the distance to the origin, which is also the "constant term" of the implicit equation - * \warning the vector normal is assumed to be normalized. - */ - EIGEN_DEVICE_FUNC inline const Scalar& offset() const { return m_coeffs.coeff(dim()); } - - /** \returns a non-constant reference to the distance to the origin, which is also the constant part - * of the implicit equation */ - EIGEN_DEVICE_FUNC inline Scalar& offset() { return m_coeffs(dim()); } - - /** \returns a constant reference to the coefficients c_i of the plane equation: - * \f$ c_0*x_0 + ... + c_{d-1}*x_{d-1} + c_d = 0 \f$ - */ - EIGEN_DEVICE_FUNC inline const Coefficients& coeffs() const { return m_coeffs; } - - /** \returns a non-constant reference to the coefficients c_i of the plane equation: - * \f$ c_0*x_0 + ... + c_{d-1}*x_{d-1} + c_d = 0 \f$ - */ - EIGEN_DEVICE_FUNC inline Coefficients& coeffs() { return m_coeffs; } - - /** \returns the intersection of *this with \a other. - * - * \warning The ambient space must be a plane, i.e. have dimension 2, so that \c *this and \a other are lines. - * - * \note If \a other is approximately parallel to *this, this method will return any point on *this. - */ - EIGEN_DEVICE_FUNC VectorType intersection(const Hyperplane& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(VectorType, 2) - Scalar det = coeffs().coeff(0) * other.coeffs().coeff(1) - coeffs().coeff(1) * other.coeffs().coeff(0); - // since the line equations ax+by=c are normalized with a^2+b^2=1, the following tests - // whether the two lines are approximately parallel. - if(internal::isMuchSmallerThan(det, Scalar(1))) - { // special case where the two lines are approximately parallel. Pick any point on the first line. - if(numext::abs(coeffs().coeff(1))>numext::abs(coeffs().coeff(0))) - return VectorType(coeffs().coeff(1), -coeffs().coeff(2)/coeffs().coeff(1)-coeffs().coeff(0)); - else - return VectorType(-coeffs().coeff(2)/coeffs().coeff(0)-coeffs().coeff(1), coeffs().coeff(0)); - } - else - { // general case - Scalar invdet = Scalar(1) / det; - return VectorType(invdet*(coeffs().coeff(1)*other.coeffs().coeff(2)-other.coeffs().coeff(1)*coeffs().coeff(2)), - invdet*(other.coeffs().coeff(0)*coeffs().coeff(2)-coeffs().coeff(0)*other.coeffs().coeff(2))); - } - } - - /** Applies the transformation matrix \a mat to \c *this and returns a reference to \c *this. - * - * \param mat the Dim x Dim transformation matrix - * \param traits specifies whether the matrix \a mat represents an #Isometry - * or a more generic #Affine transformation. The default is #Affine. - */ - template - EIGEN_DEVICE_FUNC inline Hyperplane& transform(const MatrixBase& mat, TransformTraits traits = Affine) - { - if (traits==Affine) - { - normal() = mat.inverse().transpose() * normal(); - m_coeffs /= normal().norm(); - } - else if (traits==Isometry) - normal() = mat * normal(); - else - { - eigen_assert(0 && "invalid traits value in Hyperplane::transform()"); - } - return *this; - } - - /** Applies the transformation \a t to \c *this and returns a reference to \c *this. - * - * \param t the transformation of dimension Dim - * \param traits specifies whether the transformation \a t represents an #Isometry - * or a more generic #Affine transformation. The default is #Affine. - * Other kind of transformations are not supported. - */ - template - EIGEN_DEVICE_FUNC inline Hyperplane& transform(const Transform& t, - TransformTraits traits = Affine) - { - transform(t.linear(), traits); - offset() -= normal().dot(t.translation()); - return *this; - } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { - return typename internal::cast_return_type >::type(*this); - } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit Hyperplane(const Hyperplane& other) - { m_coeffs = other.coeffs().template cast(); } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - template - EIGEN_DEVICE_FUNC bool isApprox(const Hyperplane& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return m_coeffs.isApprox(other.m_coeffs, prec); } - -protected: - - Coefficients m_coeffs; -}; - -} // end namespace Eigen - -#endif // EIGEN_HYPERPLANE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/OrthoMethods.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/OrthoMethods.h deleted file mode 100644 index a035e6310a7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/OrthoMethods.h +++ /dev/null @@ -1,234 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ORTHOMETHODS_H -#define EIGEN_ORTHOMETHODS_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \returns the cross product of \c *this and \a other - * - * Here is a very good explanation of cross-product: http://xkcd.com/199/ - * - * With complex numbers, the cross product is implemented as - * \f$ (\mathbf{a}+i\mathbf{b}) \times (\mathbf{c}+i\mathbf{d}) = (\mathbf{a} \times \mathbf{c} - \mathbf{b} \times \mathbf{d}) - i(\mathbf{a} \times \mathbf{d} - \mathbf{b} \times \mathbf{c})\f$ - * - * \sa MatrixBase::cross3() - */ -template -template -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_DEVICE_FUNC inline typename MatrixBase::template cross_product_return_type::type -#else -inline typename MatrixBase::PlainObject -#endif -MatrixBase::cross(const MatrixBase& other) const -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Derived,3) - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,3) - - // Note that there is no need for an expression here since the compiler - // optimize such a small temporary very well (even within a complex expression) - typename internal::nested_eval::type lhs(derived()); - typename internal::nested_eval::type rhs(other.derived()); - return typename cross_product_return_type::type( - numext::conj(lhs.coeff(1) * rhs.coeff(2) - lhs.coeff(2) * rhs.coeff(1)), - numext::conj(lhs.coeff(2) * rhs.coeff(0) - lhs.coeff(0) * rhs.coeff(2)), - numext::conj(lhs.coeff(0) * rhs.coeff(1) - lhs.coeff(1) * rhs.coeff(0)) - ); -} - -namespace internal { - -template< int Arch,typename VectorLhs,typename VectorRhs, - typename Scalar = typename VectorLhs::Scalar, - bool Vectorizable = bool((VectorLhs::Flags&VectorRhs::Flags)&PacketAccessBit)> -struct cross3_impl { - EIGEN_DEVICE_FUNC static inline typename internal::plain_matrix_type::type - run(const VectorLhs& lhs, const VectorRhs& rhs) - { - return typename internal::plain_matrix_type::type( - numext::conj(lhs.coeff(1) * rhs.coeff(2) - lhs.coeff(2) * rhs.coeff(1)), - numext::conj(lhs.coeff(2) * rhs.coeff(0) - lhs.coeff(0) * rhs.coeff(2)), - numext::conj(lhs.coeff(0) * rhs.coeff(1) - lhs.coeff(1) * rhs.coeff(0)), - 0 - ); - } -}; - -} - -/** \geometry_module \ingroup Geometry_Module - * - * \returns the cross product of \c *this and \a other using only the x, y, and z coefficients - * - * The size of \c *this and \a other must be four. This function is especially useful - * when using 4D vectors instead of 3D ones to get advantage of SSE/AltiVec vectorization. - * - * \sa MatrixBase::cross() - */ -template -template -EIGEN_DEVICE_FUNC inline typename MatrixBase::PlainObject -MatrixBase::cross3(const MatrixBase& other) const -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Derived,4) - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,4) - - typedef typename internal::nested_eval::type DerivedNested; - typedef typename internal::nested_eval::type OtherDerivedNested; - DerivedNested lhs(derived()); - OtherDerivedNested rhs(other.derived()); - - return internal::cross3_impl::type, - typename internal::remove_all::type>::run(lhs,rhs); -} - -/** \geometry_module \ingroup Geometry_Module - * - * \returns a matrix expression of the cross product of each column or row - * of the referenced expression with the \a other vector. - * - * The referenced matrix must have one dimension equal to 3. - * The result matrix has the same dimensions than the referenced one. - * - * \sa MatrixBase::cross() */ -template -template -EIGEN_DEVICE_FUNC -const typename VectorwiseOp::CrossReturnType -VectorwiseOp::cross(const MatrixBase& other) const -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,3) - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - - typename internal::nested_eval::type mat(_expression()); - typename internal::nested_eval::type vec(other.derived()); - - CrossReturnType res(_expression().rows(),_expression().cols()); - if(Direction==Vertical) - { - eigen_assert(CrossReturnType::RowsAtCompileTime==3 && "the matrix must have exactly 3 rows"); - res.row(0) = (mat.row(1) * vec.coeff(2) - mat.row(2) * vec.coeff(1)).conjugate(); - res.row(1) = (mat.row(2) * vec.coeff(0) - mat.row(0) * vec.coeff(2)).conjugate(); - res.row(2) = (mat.row(0) * vec.coeff(1) - mat.row(1) * vec.coeff(0)).conjugate(); - } - else - { - eigen_assert(CrossReturnType::ColsAtCompileTime==3 && "the matrix must have exactly 3 columns"); - res.col(0) = (mat.col(1) * vec.coeff(2) - mat.col(2) * vec.coeff(1)).conjugate(); - res.col(1) = (mat.col(2) * vec.coeff(0) - mat.col(0) * vec.coeff(2)).conjugate(); - res.col(2) = (mat.col(0) * vec.coeff(1) - mat.col(1) * vec.coeff(0)).conjugate(); - } - return res; -} - -namespace internal { - -template -struct unitOrthogonal_selector -{ - typedef typename plain_matrix_type::type VectorType; - typedef typename traits::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Matrix Vector2; - EIGEN_DEVICE_FUNC - static inline VectorType run(const Derived& src) - { - VectorType perp = VectorType::Zero(src.size()); - Index maxi = 0; - Index sndi = 0; - src.cwiseAbs().maxCoeff(&maxi); - if (maxi==0) - sndi = 1; - RealScalar invnm = RealScalar(1)/(Vector2() << src.coeff(sndi),src.coeff(maxi)).finished().norm(); - perp.coeffRef(maxi) = -numext::conj(src.coeff(sndi)) * invnm; - perp.coeffRef(sndi) = numext::conj(src.coeff(maxi)) * invnm; - - return perp; - } -}; - -template -struct unitOrthogonal_selector -{ - typedef typename plain_matrix_type::type VectorType; - typedef typename traits::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - EIGEN_DEVICE_FUNC - static inline VectorType run(const Derived& src) - { - VectorType perp; - /* Let us compute the crossed product of *this with a vector - * that is not too close to being colinear to *this. - */ - - /* unless the x and y coords are both close to zero, we can - * simply take ( -y, x, 0 ) and normalize it. - */ - if((!isMuchSmallerThan(src.x(), src.z())) - || (!isMuchSmallerThan(src.y(), src.z()))) - { - RealScalar invnm = RealScalar(1)/src.template head<2>().norm(); - perp.coeffRef(0) = -numext::conj(src.y())*invnm; - perp.coeffRef(1) = numext::conj(src.x())*invnm; - perp.coeffRef(2) = 0; - } - /* if both x and y are close to zero, then the vector is close - * to the z-axis, so it's far from colinear to the x-axis for instance. - * So we take the crossed product with (1,0,0) and normalize it. - */ - else - { - RealScalar invnm = RealScalar(1)/src.template tail<2>().norm(); - perp.coeffRef(0) = 0; - perp.coeffRef(1) = -numext::conj(src.z())*invnm; - perp.coeffRef(2) = numext::conj(src.y())*invnm; - } - - return perp; - } -}; - -template -struct unitOrthogonal_selector -{ - typedef typename plain_matrix_type::type VectorType; - EIGEN_DEVICE_FUNC - static inline VectorType run(const Derived& src) - { return VectorType(-numext::conj(src.y()), numext::conj(src.x())).normalized(); } -}; - -} // end namespace internal - -/** \geometry_module \ingroup Geometry_Module - * - * \returns a unit vector which is orthogonal to \c *this - * - * The size of \c *this must be at least 2. If the size is exactly 2, - * then the returned vector is a counter clock wise rotation of \c *this, i.e., (-y,x).normalized(). - * - * \sa cross() - */ -template -EIGEN_DEVICE_FUNC typename MatrixBase::PlainObject -MatrixBase::unitOrthogonal() const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return internal::unitOrthogonal_selector::run(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_ORTHOMETHODS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/ParametrizedLine.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/ParametrizedLine.h deleted file mode 100644 index 1e985d8cde2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/ParametrizedLine.h +++ /dev/null @@ -1,195 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARAMETRIZEDLINE_H -#define EIGEN_PARAMETRIZEDLINE_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class ParametrizedLine - * - * \brief A parametrized line - * - * A parametrized line is defined by an origin point \f$ \mathbf{o} \f$ and a unit - * direction vector \f$ \mathbf{d} \f$ such that the line corresponds to - * the set \f$ l(t) = \mathbf{o} + t \mathbf{d} \f$, \f$ t \in \mathbf{R} \f$. - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients - * \tparam _AmbientDim the dimension of the ambient space, can be a compile time value or Dynamic. - */ -template -class ParametrizedLine -{ -public: - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_AmbientDim) - enum { - AmbientDimAtCompileTime = _AmbientDim, - Options = _Options - }; - typedef _Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef Matrix VectorType; - - /** Default constructor without initialization */ - EIGEN_DEVICE_FUNC inline ParametrizedLine() {} - - template - EIGEN_DEVICE_FUNC ParametrizedLine(const ParametrizedLine& other) - : m_origin(other.origin()), m_direction(other.direction()) - {} - - /** Constructs a dynamic-size line with \a _dim the dimension - * of the ambient space */ - EIGEN_DEVICE_FUNC inline explicit ParametrizedLine(Index _dim) : m_origin(_dim), m_direction(_dim) {} - - /** Initializes a parametrized line of direction \a direction and origin \a origin. - * \warning the vector direction is assumed to be normalized. - */ - EIGEN_DEVICE_FUNC ParametrizedLine(const VectorType& origin, const VectorType& direction) - : m_origin(origin), m_direction(direction) {} - - template - EIGEN_DEVICE_FUNC explicit ParametrizedLine(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane); - - /** Constructs a parametrized line going from \a p0 to \a p1. */ - EIGEN_DEVICE_FUNC static inline ParametrizedLine Through(const VectorType& p0, const VectorType& p1) - { return ParametrizedLine(p0, (p1-p0).normalized()); } - - EIGEN_DEVICE_FUNC ~ParametrizedLine() {} - - /** \returns the dimension in which the line holds */ - EIGEN_DEVICE_FUNC inline Index dim() const { return m_direction.size(); } - - EIGEN_DEVICE_FUNC const VectorType& origin() const { return m_origin; } - EIGEN_DEVICE_FUNC VectorType& origin() { return m_origin; } - - EIGEN_DEVICE_FUNC const VectorType& direction() const { return m_direction; } - EIGEN_DEVICE_FUNC VectorType& direction() { return m_direction; } - - /** \returns the squared distance of a point \a p to its projection onto the line \c *this. - * \sa distance() - */ - EIGEN_DEVICE_FUNC RealScalar squaredDistance(const VectorType& p) const - { - VectorType diff = p - origin(); - return (diff - direction().dot(diff) * direction()).squaredNorm(); - } - /** \returns the distance of a point \a p to its projection onto the line \c *this. - * \sa squaredDistance() - */ - EIGEN_DEVICE_FUNC RealScalar distance(const VectorType& p) const { EIGEN_USING_STD_MATH(sqrt) return sqrt(squaredDistance(p)); } - - /** \returns the projection of a point \a p onto the line \c *this. */ - EIGEN_DEVICE_FUNC VectorType projection(const VectorType& p) const - { return origin() + direction().dot(p-origin()) * direction(); } - - EIGEN_DEVICE_FUNC VectorType pointAt(const Scalar& t) const; - - template - EIGEN_DEVICE_FUNC Scalar intersectionParameter(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const; - - template - EIGEN_DEVICE_FUNC Scalar intersection(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const; - - template - EIGEN_DEVICE_FUNC VectorType intersectionPoint(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const; - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { - return typename internal::cast_return_type >::type(*this); - } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit ParametrizedLine(const ParametrizedLine& other) - { - m_origin = other.origin().template cast(); - m_direction = other.direction().template cast(); - } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const ParametrizedLine& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return m_origin.isApprox(other.m_origin, prec) && m_direction.isApprox(other.m_direction, prec); } - -protected: - - VectorType m_origin, m_direction; -}; - -/** Constructs a parametrized line from a 2D hyperplane - * - * \warning the ambient space must have dimension 2 such that the hyperplane actually describes a line - */ -template -template -EIGEN_DEVICE_FUNC inline ParametrizedLine<_Scalar, _AmbientDim,_Options>::ParametrizedLine(const Hyperplane<_Scalar, _AmbientDim,OtherOptions>& hyperplane) -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(VectorType, 2) - direction() = hyperplane.normal().unitOrthogonal(); - origin() = -hyperplane.normal()*hyperplane.offset(); -} - -/** \returns the point at \a t along this line - */ -template -EIGEN_DEVICE_FUNC inline typename ParametrizedLine<_Scalar, _AmbientDim,_Options>::VectorType -ParametrizedLine<_Scalar, _AmbientDim,_Options>::pointAt(const _Scalar& t) const -{ - return origin() + (direction()*t); -} - -/** \returns the parameter value of the intersection between \c *this and the given \a hyperplane - */ -template -template -EIGEN_DEVICE_FUNC inline _Scalar ParametrizedLine<_Scalar, _AmbientDim,_Options>::intersectionParameter(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const -{ - return -(hyperplane.offset()+hyperplane.normal().dot(origin())) - / hyperplane.normal().dot(direction()); -} - - -/** \deprecated use intersectionParameter() - * \returns the parameter value of the intersection between \c *this and the given \a hyperplane - */ -template -template -EIGEN_DEVICE_FUNC inline _Scalar ParametrizedLine<_Scalar, _AmbientDim,_Options>::intersection(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const -{ - return intersectionParameter(hyperplane); -} - -/** \returns the point of the intersection between \c *this and the given hyperplane - */ -template -template -EIGEN_DEVICE_FUNC inline typename ParametrizedLine<_Scalar, _AmbientDim,_Options>::VectorType -ParametrizedLine<_Scalar, _AmbientDim,_Options>::intersectionPoint(const Hyperplane<_Scalar, _AmbientDim, OtherOptions>& hyperplane) const -{ - return pointAt(intersectionParameter(hyperplane)); -} - -} // end namespace Eigen - -#endif // EIGEN_PARAMETRIZEDLINE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Quaternion.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Quaternion.h deleted file mode 100644 index b818206568e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Quaternion.h +++ /dev/null @@ -1,832 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2009 Mathieu Gautier -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_QUATERNION_H -#define EIGEN_QUATERNION_H -namespace Eigen { - - -/*************************************************************************** -* Definition of QuaternionBase -* The implementation is at the end of the file -***************************************************************************/ - -namespace internal { -template -struct quaternionbase_assign_impl; -} - -/** \geometry_module \ingroup Geometry_Module - * \class QuaternionBase - * \brief Base class for quaternion expressions - * \tparam Derived derived type (CRTP) - * \sa class Quaternion - */ -template -class QuaternionBase : public RotationBase -{ - public: - typedef RotationBase Base; - - using Base::operator*; - using Base::derived; - - typedef typename internal::traits::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef typename internal::traits::Coefficients Coefficients; - typedef typename Coefficients::CoeffReturnType CoeffReturnType; - typedef typename internal::conditional::Flags&LvalueBit), - Scalar&, CoeffReturnType>::type NonConstCoeffReturnType; - - - enum { - Flags = Eigen::internal::traits::Flags - }; - - // typedef typename Matrix Coefficients; - /** the type of a 3D vector */ - typedef Matrix Vector3; - /** the equivalent rotation matrix type */ - typedef Matrix Matrix3; - /** the equivalent angle-axis type */ - typedef AngleAxis AngleAxisType; - - - - /** \returns the \c x coefficient */ - EIGEN_DEVICE_FUNC inline CoeffReturnType x() const { return this->derived().coeffs().coeff(0); } - /** \returns the \c y coefficient */ - EIGEN_DEVICE_FUNC inline CoeffReturnType y() const { return this->derived().coeffs().coeff(1); } - /** \returns the \c z coefficient */ - EIGEN_DEVICE_FUNC inline CoeffReturnType z() const { return this->derived().coeffs().coeff(2); } - /** \returns the \c w coefficient */ - EIGEN_DEVICE_FUNC inline CoeffReturnType w() const { return this->derived().coeffs().coeff(3); } - - /** \returns a reference to the \c x coefficient (if Derived is a non-const lvalue) */ - EIGEN_DEVICE_FUNC inline NonConstCoeffReturnType x() { return this->derived().coeffs().x(); } - /** \returns a reference to the \c y coefficient (if Derived is a non-const lvalue) */ - EIGEN_DEVICE_FUNC inline NonConstCoeffReturnType y() { return this->derived().coeffs().y(); } - /** \returns a reference to the \c z coefficient (if Derived is a non-const lvalue) */ - EIGEN_DEVICE_FUNC inline NonConstCoeffReturnType z() { return this->derived().coeffs().z(); } - /** \returns a reference to the \c w coefficient (if Derived is a non-const lvalue) */ - EIGEN_DEVICE_FUNC inline NonConstCoeffReturnType w() { return this->derived().coeffs().w(); } - - /** \returns a read-only vector expression of the imaginary part (x,y,z) */ - EIGEN_DEVICE_FUNC inline const VectorBlock vec() const { return coeffs().template head<3>(); } - - /** \returns a vector expression of the imaginary part (x,y,z) */ - EIGEN_DEVICE_FUNC inline VectorBlock vec() { return coeffs().template head<3>(); } - - /** \returns a read-only vector expression of the coefficients (x,y,z,w) */ - EIGEN_DEVICE_FUNC inline const typename internal::traits::Coefficients& coeffs() const { return derived().coeffs(); } - - /** \returns a vector expression of the coefficients (x,y,z,w) */ - EIGEN_DEVICE_FUNC inline typename internal::traits::Coefficients& coeffs() { return derived().coeffs(); } - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE QuaternionBase& operator=(const QuaternionBase& other); - template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const QuaternionBase& other); - -// disabled this copy operator as it is giving very strange compilation errors when compiling -// test_stdvector with GCC 4.4.2. This looks like a GCC bug though, so feel free to re-enable it if it's -// useful; however notice that we already have the templated operator= above and e.g. in MatrixBase -// we didn't have to add, in addition to templated operator=, such a non-templated copy operator. -// Derived& operator=(const QuaternionBase& other) -// { return operator=(other); } - - EIGEN_DEVICE_FUNC Derived& operator=(const AngleAxisType& aa); - template EIGEN_DEVICE_FUNC Derived& operator=(const MatrixBase& m); - - /** \returns a quaternion representing an identity rotation - * \sa MatrixBase::Identity() - */ - EIGEN_DEVICE_FUNC static inline Quaternion Identity() { return Quaternion(Scalar(1), Scalar(0), Scalar(0), Scalar(0)); } - - /** \sa QuaternionBase::Identity(), MatrixBase::setIdentity() - */ - EIGEN_DEVICE_FUNC inline QuaternionBase& setIdentity() { coeffs() << Scalar(0), Scalar(0), Scalar(0), Scalar(1); return *this; } - - /** \returns the squared norm of the quaternion's coefficients - * \sa QuaternionBase::norm(), MatrixBase::squaredNorm() - */ - EIGEN_DEVICE_FUNC inline Scalar squaredNorm() const { return coeffs().squaredNorm(); } - - /** \returns the norm of the quaternion's coefficients - * \sa QuaternionBase::squaredNorm(), MatrixBase::norm() - */ - EIGEN_DEVICE_FUNC inline Scalar norm() const { return coeffs().norm(); } - - /** Normalizes the quaternion \c *this - * \sa normalized(), MatrixBase::normalize() */ - EIGEN_DEVICE_FUNC inline void normalize() { coeffs().normalize(); } - /** \returns a normalized copy of \c *this - * \sa normalize(), MatrixBase::normalized() */ - EIGEN_DEVICE_FUNC inline Quaternion normalized() const { return Quaternion(coeffs().normalized()); } - - /** \returns the dot product of \c *this and \a other - * Geometrically speaking, the dot product of two unit quaternions - * corresponds to the cosine of half the angle between the two rotations. - * \sa angularDistance() - */ - template EIGEN_DEVICE_FUNC inline Scalar dot(const QuaternionBase& other) const { return coeffs().dot(other.coeffs()); } - - template EIGEN_DEVICE_FUNC Scalar angularDistance(const QuaternionBase& other) const; - - /** \returns an equivalent 3x3 rotation matrix */ - EIGEN_DEVICE_FUNC Matrix3 toRotationMatrix() const; - - /** \returns the quaternion which transform \a a into \a b through a rotation */ - template - EIGEN_DEVICE_FUNC Derived& setFromTwoVectors(const MatrixBase& a, const MatrixBase& b); - - template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Quaternion operator* (const QuaternionBase& q) const; - template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator*= (const QuaternionBase& q); - - /** \returns the quaternion describing the inverse rotation */ - EIGEN_DEVICE_FUNC Quaternion inverse() const; - - /** \returns the conjugated quaternion */ - EIGEN_DEVICE_FUNC Quaternion conjugate() const; - - template EIGEN_DEVICE_FUNC Quaternion slerp(const Scalar& t, const QuaternionBase& other) const; - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - template - EIGEN_DEVICE_FUNC bool isApprox(const QuaternionBase& other, const RealScalar& prec = NumTraits::dummy_precision()) const - { return coeffs().isApprox(other.coeffs(), prec); } - - /** return the result vector of \a v through the rotation*/ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Vector3 _transformVector(const Vector3& v) const; - - #ifdef EIGEN_PARSED_BY_DOXYGEN - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const; - - #else - - template - EIGEN_DEVICE_FUNC inline - typename internal::enable_if::value,const Derived&>::type cast() const - { - return derived(); - } - - template - EIGEN_DEVICE_FUNC inline - typename internal::enable_if::value,Quaternion >::type cast() const - { - return Quaternion(coeffs().template cast()); - } - #endif - -#ifdef EIGEN_QUATERNIONBASE_PLUGIN -# include EIGEN_QUATERNIONBASE_PLUGIN -#endif -protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(QuaternionBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(QuaternionBase) -}; - -/*************************************************************************** -* Definition/implementation of Quaternion -***************************************************************************/ - -/** \geometry_module \ingroup Geometry_Module - * - * \class Quaternion - * - * \brief The quaternion class used to represent 3D orientations and rotations - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients - * \tparam _Options controls the memory alignment of the coefficients. Can be \# AutoAlign or \# DontAlign. Default is AutoAlign. - * - * This class represents a quaternion \f$ w+xi+yj+zk \f$ that is a convenient representation of - * orientations and rotations of objects in three dimensions. Compared to other representations - * like Euler angles or 3x3 matrices, quaternions offer the following advantages: - * \li \b compact storage (4 scalars) - * \li \b efficient to compose (28 flops), - * \li \b stable spherical interpolation - * - * The following two typedefs are provided for convenience: - * \li \c Quaternionf for \c float - * \li \c Quaterniond for \c double - * - * \warning Operations interpreting the quaternion as rotation have undefined behavior if the quaternion is not normalized. - * - * \sa class AngleAxis, class Transform - */ - -namespace internal { -template -struct traits > -{ - typedef Quaternion<_Scalar,_Options> PlainObject; - typedef _Scalar Scalar; - typedef Matrix<_Scalar,4,1,_Options> Coefficients; - enum{ - Alignment = internal::traits::Alignment, - Flags = LvalueBit - }; -}; -} - -template -class Quaternion : public QuaternionBase > -{ -public: - typedef QuaternionBase > Base; - enum { NeedsAlignment = internal::traits::Alignment>0 }; - - typedef _Scalar Scalar; - - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Quaternion) - using Base::operator*=; - - typedef typename internal::traits::Coefficients Coefficients; - typedef typename Base::AngleAxisType AngleAxisType; - - /** Default constructor leaving the quaternion uninitialized. */ - EIGEN_DEVICE_FUNC inline Quaternion() {} - - /** Constructs and initializes the quaternion \f$ w+xi+yj+zk \f$ from - * its four coefficients \a w, \a x, \a y and \a z. - * - * \warning Note the order of the arguments: the real \a w coefficient first, - * while internally the coefficients are stored in the following order: - * [\c x, \c y, \c z, \c w] - */ - EIGEN_DEVICE_FUNC inline Quaternion(const Scalar& w, const Scalar& x, const Scalar& y, const Scalar& z) : m_coeffs(x, y, z, w){} - - /** Constructs and initialize a quaternion from the array data */ - EIGEN_DEVICE_FUNC explicit inline Quaternion(const Scalar* data) : m_coeffs(data) {} - - /** Copy constructor */ - template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Quaternion(const QuaternionBase& other) { this->Base::operator=(other); } - - /** Constructs and initializes a quaternion from the angle-axis \a aa */ - EIGEN_DEVICE_FUNC explicit inline Quaternion(const AngleAxisType& aa) { *this = aa; } - - /** Constructs and initializes a quaternion from either: - * - a rotation matrix expression, - * - a 4D vector expression representing quaternion coefficients. - */ - template - EIGEN_DEVICE_FUNC explicit inline Quaternion(const MatrixBase& other) { *this = other; } - - /** Explicit copy constructor with scalar conversion */ - template - EIGEN_DEVICE_FUNC explicit inline Quaternion(const Quaternion& other) - { m_coeffs = other.coeffs().template cast(); } - - EIGEN_DEVICE_FUNC static Quaternion UnitRandom(); - - template - EIGEN_DEVICE_FUNC static Quaternion FromTwoVectors(const MatrixBase& a, const MatrixBase& b); - - EIGEN_DEVICE_FUNC inline Coefficients& coeffs() { return m_coeffs;} - EIGEN_DEVICE_FUNC inline const Coefficients& coeffs() const { return m_coeffs;} - - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(bool(NeedsAlignment)) - -#ifdef EIGEN_QUATERNION_PLUGIN -# include EIGEN_QUATERNION_PLUGIN -#endif - -protected: - Coefficients m_coeffs; - -#ifndef EIGEN_PARSED_BY_DOXYGEN - static EIGEN_STRONG_INLINE void _check_template_params() - { - EIGEN_STATIC_ASSERT( (_Options & DontAlign) == _Options, - INVALID_MATRIX_TEMPLATE_PARAMETERS) - } -#endif -}; - -/** \ingroup Geometry_Module - * single precision quaternion type */ -typedef Quaternion Quaternionf; -/** \ingroup Geometry_Module - * double precision quaternion type */ -typedef Quaternion Quaterniond; - -/*************************************************************************** -* Specialization of Map> -***************************************************************************/ - -namespace internal { - template - struct traits, _Options> > : traits > - { - typedef Map, _Options> Coefficients; - }; -} - -namespace internal { - template - struct traits, _Options> > : traits > - { - typedef Map, _Options> Coefficients; - typedef traits > TraitsBase; - enum { - Flags = TraitsBase::Flags & ~LvalueBit - }; - }; -} - -/** \ingroup Geometry_Module - * \brief Quaternion expression mapping a constant memory buffer - * - * \tparam _Scalar the type of the Quaternion coefficients - * \tparam _Options see class Map - * - * This is a specialization of class Map for Quaternion. This class allows to view - * a 4 scalar memory buffer as an Eigen's Quaternion object. - * - * \sa class Map, class Quaternion, class QuaternionBase - */ -template -class Map, _Options > - : public QuaternionBase, _Options> > -{ - public: - typedef QuaternionBase, _Options> > Base; - - typedef _Scalar Scalar; - typedef typename internal::traits::Coefficients Coefficients; - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Map) - using Base::operator*=; - - /** Constructs a Mapped Quaternion object from the pointer \a coeffs - * - * The pointer \a coeffs must reference the four coefficients of Quaternion in the following order: - * \code *coeffs == {x, y, z, w} \endcode - * - * If the template parameter _Options is set to #Aligned, then the pointer coeffs must be aligned. */ - EIGEN_DEVICE_FUNC explicit EIGEN_STRONG_INLINE Map(const Scalar* coeffs) : m_coeffs(coeffs) {} - - EIGEN_DEVICE_FUNC inline const Coefficients& coeffs() const { return m_coeffs;} - - protected: - const Coefficients m_coeffs; -}; - -/** \ingroup Geometry_Module - * \brief Expression of a quaternion from a memory buffer - * - * \tparam _Scalar the type of the Quaternion coefficients - * \tparam _Options see class Map - * - * This is a specialization of class Map for Quaternion. This class allows to view - * a 4 scalar memory buffer as an Eigen's Quaternion object. - * - * \sa class Map, class Quaternion, class QuaternionBase - */ -template -class Map, _Options > - : public QuaternionBase, _Options> > -{ - public: - typedef QuaternionBase, _Options> > Base; - - typedef _Scalar Scalar; - typedef typename internal::traits::Coefficients Coefficients; - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Map) - using Base::operator*=; - - /** Constructs a Mapped Quaternion object from the pointer \a coeffs - * - * The pointer \a coeffs must reference the four coefficients of Quaternion in the following order: - * \code *coeffs == {x, y, z, w} \endcode - * - * If the template parameter _Options is set to #Aligned, then the pointer coeffs must be aligned. */ - EIGEN_DEVICE_FUNC explicit EIGEN_STRONG_INLINE Map(Scalar* coeffs) : m_coeffs(coeffs) {} - - EIGEN_DEVICE_FUNC inline Coefficients& coeffs() { return m_coeffs; } - EIGEN_DEVICE_FUNC inline const Coefficients& coeffs() const { return m_coeffs; } - - protected: - Coefficients m_coeffs; -}; - -/** \ingroup Geometry_Module - * Map an unaligned array of single precision scalars as a quaternion */ -typedef Map, 0> QuaternionMapf; -/** \ingroup Geometry_Module - * Map an unaligned array of double precision scalars as a quaternion */ -typedef Map, 0> QuaternionMapd; -/** \ingroup Geometry_Module - * Map a 16-byte aligned array of single precision scalars as a quaternion */ -typedef Map, Aligned> QuaternionMapAlignedf; -/** \ingroup Geometry_Module - * Map a 16-byte aligned array of double precision scalars as a quaternion */ -typedef Map, Aligned> QuaternionMapAlignedd; - -/*************************************************************************** -* Implementation of QuaternionBase methods -***************************************************************************/ - -// Generic Quaternion * Quaternion product -// This product can be specialized for a given architecture via the Arch template argument. -namespace internal { -template struct quat_product -{ - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Quaternion run(const QuaternionBase& a, const QuaternionBase& b){ - return Quaternion - ( - a.w() * b.w() - a.x() * b.x() - a.y() * b.y() - a.z() * b.z(), - a.w() * b.x() + a.x() * b.w() + a.y() * b.z() - a.z() * b.y(), - a.w() * b.y() + a.y() * b.w() + a.z() * b.x() - a.x() * b.z(), - a.w() * b.z() + a.z() * b.w() + a.x() * b.y() - a.y() * b.x() - ); - } -}; -} - -/** \returns the concatenation of two rotations as a quaternion-quaternion product */ -template -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Quaternion::Scalar> -QuaternionBase::operator* (const QuaternionBase& other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - return internal::quat_product::Scalar>::run(*this, other); -} - -/** \sa operator*(Quaternion) */ -template -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& QuaternionBase::operator*= (const QuaternionBase& other) -{ - derived() = derived() * other.derived(); - return derived(); -} - -/** Rotation of a vector by a quaternion. - * \remarks If the quaternion is used to rotate several points (>1) - * then it is much more efficient to first convert it to a 3x3 Matrix. - * Comparison of the operation cost for n transformations: - * - Quaternion2: 30n - * - Via a Matrix3: 24 + 15n - */ -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename QuaternionBase::Vector3 -QuaternionBase::_transformVector(const Vector3& v) const -{ - // Note that this algorithm comes from the optimization by hand - // of the conversion to a Matrix followed by a Matrix/Vector product. - // It appears to be much faster than the common algorithm found - // in the literature (30 versus 39 flops). It also requires two - // Vector3 as temporaries. - Vector3 uv = this->vec().cross(v); - uv += uv; - return v + this->w() * uv + this->vec().cross(uv); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE QuaternionBase& QuaternionBase::operator=(const QuaternionBase& other) -{ - coeffs() = other.coeffs(); - return derived(); -} - -template -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& QuaternionBase::operator=(const QuaternionBase& other) -{ - coeffs() = other.coeffs(); - return derived(); -} - -/** Set \c *this from an angle-axis \a aa and returns a reference to \c *this - */ -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& QuaternionBase::operator=(const AngleAxisType& aa) -{ - EIGEN_USING_STD_MATH(cos) - EIGEN_USING_STD_MATH(sin) - Scalar ha = Scalar(0.5)*aa.angle(); // Scalar(0.5) to suppress precision loss warnings - this->w() = cos(ha); - this->vec() = sin(ha) * aa.axis(); - return derived(); -} - -/** Set \c *this from the expression \a xpr: - * - if \a xpr is a 4x1 vector, then \a xpr is assumed to be a quaternion - * - if \a xpr is a 3x3 matrix, then \a xpr is assumed to be rotation matrix - * and \a xpr is converted to a quaternion - */ - -template -template -EIGEN_DEVICE_FUNC inline Derived& QuaternionBase::operator=(const MatrixBase& xpr) -{ - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - internal::quaternionbase_assign_impl::run(*this, xpr.derived()); - return derived(); -} - -/** Convert the quaternion to a 3x3 rotation matrix. The quaternion is required to - * be normalized, otherwise the result is undefined. - */ -template -EIGEN_DEVICE_FUNC inline typename QuaternionBase::Matrix3 -QuaternionBase::toRotationMatrix(void) const -{ - // NOTE if inlined, then gcc 4.2 and 4.4 get rid of the temporary (not gcc 4.3 !!) - // if not inlined then the cost of the return by value is huge ~ +35%, - // however, not inlining this function is an order of magnitude slower, so - // it has to be inlined, and so the return by value is not an issue - Matrix3 res; - - const Scalar tx = Scalar(2)*this->x(); - const Scalar ty = Scalar(2)*this->y(); - const Scalar tz = Scalar(2)*this->z(); - const Scalar twx = tx*this->w(); - const Scalar twy = ty*this->w(); - const Scalar twz = tz*this->w(); - const Scalar txx = tx*this->x(); - const Scalar txy = ty*this->x(); - const Scalar txz = tz*this->x(); - const Scalar tyy = ty*this->y(); - const Scalar tyz = tz*this->y(); - const Scalar tzz = tz*this->z(); - - res.coeffRef(0,0) = Scalar(1)-(tyy+tzz); - res.coeffRef(0,1) = txy-twz; - res.coeffRef(0,2) = txz+twy; - res.coeffRef(1,0) = txy+twz; - res.coeffRef(1,1) = Scalar(1)-(txx+tzz); - res.coeffRef(1,2) = tyz-twx; - res.coeffRef(2,0) = txz-twy; - res.coeffRef(2,1) = tyz+twx; - res.coeffRef(2,2) = Scalar(1)-(txx+tyy); - - return res; -} - -/** Sets \c *this to be a quaternion representing a rotation between - * the two arbitrary vectors \a a and \a b. In other words, the built - * rotation represent a rotation sending the line of direction \a a - * to the line of direction \a b, both lines passing through the origin. - * - * \returns a reference to \c *this. - * - * Note that the two input vectors do \b not have to be normalized, and - * do not need to have the same norm. - */ -template -template -EIGEN_DEVICE_FUNC inline Derived& QuaternionBase::setFromTwoVectors(const MatrixBase& a, const MatrixBase& b) -{ - EIGEN_USING_STD_MATH(sqrt) - Vector3 v0 = a.normalized(); - Vector3 v1 = b.normalized(); - Scalar c = v1.dot(v0); - - // if dot == -1, vectors are nearly opposites - // => accurately compute the rotation axis by computing the - // intersection of the two planes. This is done by solving: - // x^T v0 = 0 - // x^T v1 = 0 - // under the constraint: - // ||x|| = 1 - // which yields a singular value problem - if (c < Scalar(-1)+NumTraits::dummy_precision()) - { - c = numext::maxi(c,Scalar(-1)); - Matrix m; m << v0.transpose(), v1.transpose(); - JacobiSVD > svd(m, ComputeFullV); - Vector3 axis = svd.matrixV().col(2); - - Scalar w2 = (Scalar(1)+c)*Scalar(0.5); - this->w() = sqrt(w2); - this->vec() = axis * sqrt(Scalar(1) - w2); - return derived(); - } - Vector3 axis = v0.cross(v1); - Scalar s = sqrt((Scalar(1)+c)*Scalar(2)); - Scalar invs = Scalar(1)/s; - this->vec() = axis * invs; - this->w() = s * Scalar(0.5); - - return derived(); -} - -/** \returns a random unit quaternion following a uniform distribution law on SO(3) - * - * \note The implementation is based on http://planning.cs.uiuc.edu/node198.html - */ -template -EIGEN_DEVICE_FUNC Quaternion Quaternion::UnitRandom() -{ - EIGEN_USING_STD_MATH(sqrt) - EIGEN_USING_STD_MATH(sin) - EIGEN_USING_STD_MATH(cos) - const Scalar u1 = internal::random(0, 1), - u2 = internal::random(0, 2*EIGEN_PI), - u3 = internal::random(0, 2*EIGEN_PI); - const Scalar a = sqrt(1 - u1), - b = sqrt(u1); - return Quaternion (a * sin(u2), a * cos(u2), b * sin(u3), b * cos(u3)); -} - - -/** Returns a quaternion representing a rotation between - * the two arbitrary vectors \a a and \a b. In other words, the built - * rotation represent a rotation sending the line of direction \a a - * to the line of direction \a b, both lines passing through the origin. - * - * \returns resulting quaternion - * - * Note that the two input vectors do \b not have to be normalized, and - * do not need to have the same norm. - */ -template -template -EIGEN_DEVICE_FUNC Quaternion Quaternion::FromTwoVectors(const MatrixBase& a, const MatrixBase& b) -{ - Quaternion quat; - quat.setFromTwoVectors(a, b); - return quat; -} - - -/** \returns the multiplicative inverse of \c *this - * Note that in most cases, i.e., if you simply want the opposite rotation, - * and/or the quaternion is normalized, then it is enough to use the conjugate. - * - * \sa QuaternionBase::conjugate() - */ -template -EIGEN_DEVICE_FUNC inline Quaternion::Scalar> QuaternionBase::inverse() const -{ - // FIXME should this function be called multiplicativeInverse and conjugate() be called inverse() or opposite() ?? - Scalar n2 = this->squaredNorm(); - if (n2 > Scalar(0)) - return Quaternion(conjugate().coeffs() / n2); - else - { - // return an invalid result to flag the error - return Quaternion(Coefficients::Zero()); - } -} - -// Generic conjugate of a Quaternion -namespace internal { -template struct quat_conj -{ - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Quaternion run(const QuaternionBase& q){ - return Quaternion(q.w(),-q.x(),-q.y(),-q.z()); - } -}; -} - -/** \returns the conjugate of the \c *this which is equal to the multiplicative inverse - * if the quaternion is normalized. - * The conjugate of a quaternion represents the opposite rotation. - * - * \sa Quaternion2::inverse() - */ -template -EIGEN_DEVICE_FUNC inline Quaternion::Scalar> -QuaternionBase::conjugate() const -{ - return internal::quat_conj::Scalar>::run(*this); - -} - -/** \returns the angle (in radian) between two rotations - * \sa dot() - */ -template -template -EIGEN_DEVICE_FUNC inline typename internal::traits::Scalar -QuaternionBase::angularDistance(const QuaternionBase& other) const -{ - EIGEN_USING_STD_MATH(atan2) - Quaternion d = (*this) * other.conjugate(); - return Scalar(2) * atan2( d.vec().norm(), numext::abs(d.w()) ); -} - - - -/** \returns the spherical linear interpolation between the two quaternions - * \c *this and \a other at the parameter \a t in [0;1]. - * - * This represents an interpolation for a constant motion between \c *this and \a other, - * see also http://en.wikipedia.org/wiki/Slerp. - */ -template -template -EIGEN_DEVICE_FUNC Quaternion::Scalar> -QuaternionBase::slerp(const Scalar& t, const QuaternionBase& other) const -{ - EIGEN_USING_STD_MATH(acos) - EIGEN_USING_STD_MATH(sin) - const Scalar one = Scalar(1) - NumTraits::epsilon(); - Scalar d = this->dot(other); - Scalar absD = numext::abs(d); - - Scalar scale0; - Scalar scale1; - - if(absD>=one) - { - scale0 = Scalar(1) - t; - scale1 = t; - } - else - { - // theta is the angle between the 2 quaternions - Scalar theta = acos(absD); - Scalar sinTheta = sin(theta); - - scale0 = sin( ( Scalar(1) - t ) * theta) / sinTheta; - scale1 = sin( ( t * theta) ) / sinTheta; - } - if(d(scale0 * coeffs() + scale1 * other.coeffs()); -} - -namespace internal { - -// set from a rotation matrix -template -struct quaternionbase_assign_impl -{ - typedef typename Other::Scalar Scalar; - template EIGEN_DEVICE_FUNC static inline void run(QuaternionBase& q, const Other& a_mat) - { - const typename internal::nested_eval::type mat(a_mat); - EIGEN_USING_STD_MATH(sqrt) - // This algorithm comes from "Quaternion Calculus and Fast Animation", - // Ken Shoemake, 1987 SIGGRAPH course notes - Scalar t = mat.trace(); - if (t > Scalar(0)) - { - t = sqrt(t + Scalar(1.0)); - q.w() = Scalar(0.5)*t; - t = Scalar(0.5)/t; - q.x() = (mat.coeff(2,1) - mat.coeff(1,2)) * t; - q.y() = (mat.coeff(0,2) - mat.coeff(2,0)) * t; - q.z() = (mat.coeff(1,0) - mat.coeff(0,1)) * t; - } - else - { - Index i = 0; - if (mat.coeff(1,1) > mat.coeff(0,0)) - i = 1; - if (mat.coeff(2,2) > mat.coeff(i,i)) - i = 2; - Index j = (i+1)%3; - Index k = (j+1)%3; - - t = sqrt(mat.coeff(i,i)-mat.coeff(j,j)-mat.coeff(k,k) + Scalar(1.0)); - q.coeffs().coeffRef(i) = Scalar(0.5) * t; - t = Scalar(0.5)/t; - q.w() = (mat.coeff(k,j)-mat.coeff(j,k))*t; - q.coeffs().coeffRef(j) = (mat.coeff(j,i)+mat.coeff(i,j))*t; - q.coeffs().coeffRef(k) = (mat.coeff(k,i)+mat.coeff(i,k))*t; - } - } -}; - -// set from a vector of coefficients assumed to be a quaternion -template -struct quaternionbase_assign_impl -{ - typedef typename Other::Scalar Scalar; - template EIGEN_DEVICE_FUNC static inline void run(QuaternionBase& q, const Other& vec) - { - q.coeffs() = vec; - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_QUATERNION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Rotation2D.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Rotation2D.h deleted file mode 100644 index 884b7d0ee95..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Rotation2D.h +++ /dev/null @@ -1,199 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ROTATION2D_H -#define EIGEN_ROTATION2D_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class Rotation2D - * - * \brief Represents a rotation/orientation in a 2 dimensional space. - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients - * - * This class is equivalent to a single scalar representing a counter clock wise rotation - * as a single angle in radian. It provides some additional features such as the automatic - * conversion from/to a 2x2 rotation matrix. Moreover this class aims to provide a similar - * interface to Quaternion in order to facilitate the writing of generic algorithms - * dealing with rotations. - * - * \sa class Quaternion, class Transform - */ - -namespace internal { - -template struct traits > -{ - typedef _Scalar Scalar; -}; -} // end namespace internal - -template -class Rotation2D : public RotationBase,2> -{ - typedef RotationBase,2> Base; - -public: - - using Base::operator*; - - enum { Dim = 2 }; - /** the scalar type of the coefficients */ - typedef _Scalar Scalar; - typedef Matrix Vector2; - typedef Matrix Matrix2; - -protected: - - Scalar m_angle; - -public: - - /** Construct a 2D counter clock wise rotation from the angle \a a in radian. */ - EIGEN_DEVICE_FUNC explicit inline Rotation2D(const Scalar& a) : m_angle(a) {} - - /** Default constructor wihtout initialization. The represented rotation is undefined. */ - EIGEN_DEVICE_FUNC Rotation2D() {} - - /** Construct a 2D rotation from a 2x2 rotation matrix \a mat. - * - * \sa fromRotationMatrix() - */ - template - EIGEN_DEVICE_FUNC explicit Rotation2D(const MatrixBase& m) - { - fromRotationMatrix(m.derived()); - } - - /** \returns the rotation angle */ - EIGEN_DEVICE_FUNC inline Scalar angle() const { return m_angle; } - - /** \returns a read-write reference to the rotation angle */ - EIGEN_DEVICE_FUNC inline Scalar& angle() { return m_angle; } - - /** \returns the rotation angle in [0,2pi] */ - EIGEN_DEVICE_FUNC inline Scalar smallestPositiveAngle() const { - Scalar tmp = numext::fmod(m_angle,Scalar(2*EIGEN_PI)); - return tmpScalar(EIGEN_PI)) tmp -= Scalar(2*EIGEN_PI); - else if(tmp<-Scalar(EIGEN_PI)) tmp += Scalar(2*EIGEN_PI); - return tmp; - } - - /** \returns the inverse rotation */ - EIGEN_DEVICE_FUNC inline Rotation2D inverse() const { return Rotation2D(-m_angle); } - - /** Concatenates two rotations */ - EIGEN_DEVICE_FUNC inline Rotation2D operator*(const Rotation2D& other) const - { return Rotation2D(m_angle + other.m_angle); } - - /** Concatenates two rotations */ - EIGEN_DEVICE_FUNC inline Rotation2D& operator*=(const Rotation2D& other) - { m_angle += other.m_angle; return *this; } - - /** Applies the rotation to a 2D vector */ - EIGEN_DEVICE_FUNC Vector2 operator* (const Vector2& vec) const - { return toRotationMatrix() * vec; } - - template - EIGEN_DEVICE_FUNC Rotation2D& fromRotationMatrix(const MatrixBase& m); - EIGEN_DEVICE_FUNC Matrix2 toRotationMatrix() const; - - /** Set \c *this from a 2x2 rotation matrix \a mat. - * In other words, this function extract the rotation angle from the rotation matrix. - * - * This method is an alias for fromRotationMatrix() - * - * \sa fromRotationMatrix() - */ - template - EIGEN_DEVICE_FUNC Rotation2D& operator=(const MatrixBase& m) - { return fromRotationMatrix(m.derived()); } - - /** \returns the spherical interpolation between \c *this and \a other using - * parameter \a t. It is in fact equivalent to a linear interpolation. - */ - EIGEN_DEVICE_FUNC inline Rotation2D slerp(const Scalar& t, const Rotation2D& other) const - { - Scalar dist = Rotation2D(other.m_angle-m_angle).smallestAngle(); - return Rotation2D(m_angle + dist*t); - } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { return typename internal::cast_return_type >::type(*this); } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit Rotation2D(const Rotation2D& other) - { - m_angle = Scalar(other.angle()); - } - - EIGEN_DEVICE_FUNC static inline Rotation2D Identity() { return Rotation2D(0); } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const Rotation2D& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return internal::isApprox(m_angle,other.m_angle, prec); } - -}; - -/** \ingroup Geometry_Module - * single precision 2D rotation type */ -typedef Rotation2D Rotation2Df; -/** \ingroup Geometry_Module - * double precision 2D rotation type */ -typedef Rotation2D Rotation2Dd; - -/** Set \c *this from a 2x2 rotation matrix \a mat. - * In other words, this function extract the rotation angle - * from the rotation matrix. - */ -template -template -EIGEN_DEVICE_FUNC Rotation2D& Rotation2D::fromRotationMatrix(const MatrixBase& mat) -{ - EIGEN_USING_STD_MATH(atan2) - EIGEN_STATIC_ASSERT(Derived::RowsAtCompileTime==2 && Derived::ColsAtCompileTime==2,YOU_MADE_A_PROGRAMMING_MISTAKE) - m_angle = atan2(mat.coeff(1,0), mat.coeff(0,0)); - return *this; -} - -/** Constructs and \returns an equivalent 2x2 rotation matrix. - */ -template -typename Rotation2D::Matrix2 -EIGEN_DEVICE_FUNC Rotation2D::toRotationMatrix(void) const -{ - EIGEN_USING_STD_MATH(sin) - EIGEN_USING_STD_MATH(cos) - Scalar sinA = sin(m_angle); - Scalar cosA = cos(m_angle); - return (Matrix2() << cosA, -sinA, sinA, cosA).finished(); -} - -} // end namespace Eigen - -#endif // EIGEN_ROTATION2D_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/RotationBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/RotationBase.h deleted file mode 100644 index f0ee0bd03c5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/RotationBase.h +++ /dev/null @@ -1,206 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ROTATIONBASE_H -#define EIGEN_ROTATIONBASE_H - -namespace Eigen { - -// forward declaration -namespace internal { -template -struct rotation_base_generic_product_selector; -} - -/** \class RotationBase - * - * \brief Common base class for compact rotation representations - * - * \tparam Derived is the derived type, i.e., a rotation type - * \tparam _Dim the dimension of the space - */ -template -class RotationBase -{ - public: - enum { Dim = _Dim }; - /** the scalar type of the coefficients */ - typedef typename internal::traits::Scalar Scalar; - - /** corresponding linear transformation matrix type */ - typedef Matrix RotationMatrixType; - typedef Matrix VectorType; - - public: - EIGEN_DEVICE_FUNC inline const Derived& derived() const { return *static_cast(this); } - EIGEN_DEVICE_FUNC inline Derived& derived() { return *static_cast(this); } - - /** \returns an equivalent rotation matrix */ - EIGEN_DEVICE_FUNC inline RotationMatrixType toRotationMatrix() const { return derived().toRotationMatrix(); } - - /** \returns an equivalent rotation matrix - * This function is added to be conform with the Transform class' naming scheme. - */ - EIGEN_DEVICE_FUNC inline RotationMatrixType matrix() const { return derived().toRotationMatrix(); } - - /** \returns the inverse rotation */ - EIGEN_DEVICE_FUNC inline Derived inverse() const { return derived().inverse(); } - - /** \returns the concatenation of the rotation \c *this with a translation \a t */ - EIGEN_DEVICE_FUNC inline Transform operator*(const Translation& t) const - { return Transform(*this) * t; } - - /** \returns the concatenation of the rotation \c *this with a uniform scaling \a s */ - EIGEN_DEVICE_FUNC inline RotationMatrixType operator*(const UniformScaling& s) const - { return toRotationMatrix() * s.factor(); } - - /** \returns the concatenation of the rotation \c *this with a generic expression \a e - * \a e can be: - * - a DimxDim linear transformation matrix - * - a DimxDim diagonal matrix (axis aligned scaling) - * - a vector of size Dim - */ - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::rotation_base_generic_product_selector::ReturnType - operator*(const EigenBase& e) const - { return internal::rotation_base_generic_product_selector::run(derived(), e.derived()); } - - /** \returns the concatenation of a linear transformation \a l with the rotation \a r */ - template friend - EIGEN_DEVICE_FUNC inline RotationMatrixType operator*(const EigenBase& l, const Derived& r) - { return l.derived() * r.toRotationMatrix(); } - - /** \returns the concatenation of a scaling \a l with the rotation \a r */ - EIGEN_DEVICE_FUNC friend inline Transform operator*(const DiagonalMatrix& l, const Derived& r) - { - Transform res(r); - res.linear().applyOnTheLeft(l); - return res; - } - - /** \returns the concatenation of the rotation \c *this with a transformation \a t */ - template - EIGEN_DEVICE_FUNC inline Transform operator*(const Transform& t) const - { return toRotationMatrix() * t; } - - template - EIGEN_DEVICE_FUNC inline VectorType _transformVector(const OtherVectorType& v) const - { return toRotationMatrix() * v; } -}; - -namespace internal { - -// implementation of the generic product rotation * matrix -template -struct rotation_base_generic_product_selector -{ - enum { Dim = RotationDerived::Dim }; - typedef Matrix ReturnType; - EIGEN_DEVICE_FUNC static inline ReturnType run(const RotationDerived& r, const MatrixType& m) - { return r.toRotationMatrix() * m; } -}; - -template -struct rotation_base_generic_product_selector< RotationDerived, DiagonalMatrix, false > -{ - typedef Transform ReturnType; - EIGEN_DEVICE_FUNC static inline ReturnType run(const RotationDerived& r, const DiagonalMatrix& m) - { - ReturnType res(r); - res.linear() *= m; - return res; - } -}; - -template -struct rotation_base_generic_product_selector -{ - enum { Dim = RotationDerived::Dim }; - typedef Matrix ReturnType; - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE ReturnType run(const RotationDerived& r, const OtherVectorType& v) - { - return r._transformVector(v); - } -}; - -} // end namespace internal - -/** \geometry_module - * - * \brief Constructs a Dim x Dim rotation matrix from the rotation \a r - */ -template -template -EIGEN_DEVICE_FUNC Matrix<_Scalar, _Rows, _Cols, _Storage, _MaxRows, _MaxCols> -::Matrix(const RotationBase& r) -{ - EIGEN_STATIC_ASSERT_MATRIX_SPECIFIC_SIZE(Matrix,int(OtherDerived::Dim),int(OtherDerived::Dim)) - *this = r.toRotationMatrix(); -} - -/** \geometry_module - * - * \brief Set a Dim x Dim rotation matrix from the rotation \a r - */ -template -template -EIGEN_DEVICE_FUNC Matrix<_Scalar, _Rows, _Cols, _Storage, _MaxRows, _MaxCols>& -Matrix<_Scalar, _Rows, _Cols, _Storage, _MaxRows, _MaxCols> -::operator=(const RotationBase& r) -{ - EIGEN_STATIC_ASSERT_MATRIX_SPECIFIC_SIZE(Matrix,int(OtherDerived::Dim),int(OtherDerived::Dim)) - return *this = r.toRotationMatrix(); -} - -namespace internal { - -/** \internal - * - * Helper function to return an arbitrary rotation object to a rotation matrix. - * - * \tparam Scalar the numeric type of the matrix coefficients - * \tparam Dim the dimension of the current space - * - * It returns a Dim x Dim fixed size matrix. - * - * Default specializations are provided for: - * - any scalar type (2D), - * - any matrix expression, - * - any type based on RotationBase (e.g., Quaternion, AngleAxis, Rotation2D) - * - * Currently toRotationMatrix is only used by Transform. - * - * \sa class Transform, class Rotation2D, class Quaternion, class AngleAxis - */ -template -EIGEN_DEVICE_FUNC static inline Matrix toRotationMatrix(const Scalar& s) -{ - EIGEN_STATIC_ASSERT(Dim==2,YOU_MADE_A_PROGRAMMING_MISTAKE) - return Rotation2D(s).toRotationMatrix(); -} - -template -EIGEN_DEVICE_FUNC static inline Matrix toRotationMatrix(const RotationBase& r) -{ - return r.toRotationMatrix(); -} - -template -EIGEN_DEVICE_FUNC static inline const MatrixBase& toRotationMatrix(const MatrixBase& mat) -{ - EIGEN_STATIC_ASSERT(OtherDerived::RowsAtCompileTime==Dim && OtherDerived::ColsAtCompileTime==Dim, - YOU_MADE_A_PROGRAMMING_MISTAKE) - return mat; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_ROTATIONBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Scaling.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Scaling.h deleted file mode 100644 index 33eabd81a4c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Scaling.h +++ /dev/null @@ -1,170 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SCALING_H -#define EIGEN_SCALING_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class UniformScaling - * - * \brief Represents a generic uniform scaling transformation - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients. - * - * This class represent a uniform scaling transformation. It is the return - * type of Scaling(Scalar), and most of the time this is the only way it - * is used. In particular, this class is not aimed to be used to store a scaling transformation, - * but rather to make easier the constructions and updates of Transform objects. - * - * To represent an axis aligned scaling, use the DiagonalMatrix class. - * - * \sa Scaling(), class DiagonalMatrix, MatrixBase::asDiagonal(), class Translation, class Transform - */ -template -class UniformScaling -{ -public: - /** the scalar type of the coefficients */ - typedef _Scalar Scalar; - -protected: - - Scalar m_factor; - -public: - - /** Default constructor without initialization. */ - UniformScaling() {} - /** Constructs and initialize a uniform scaling transformation */ - explicit inline UniformScaling(const Scalar& s) : m_factor(s) {} - - inline const Scalar& factor() const { return m_factor; } - inline Scalar& factor() { return m_factor; } - - /** Concatenates two uniform scaling */ - inline UniformScaling operator* (const UniformScaling& other) const - { return UniformScaling(m_factor * other.factor()); } - - /** Concatenates a uniform scaling and a translation */ - template - inline Transform operator* (const Translation& t) const; - - /** Concatenates a uniform scaling and an affine transformation */ - template - inline Transform operator* (const Transform& t) const - { - Transform res = t; - res.prescale(factor()); - return res; - } - - /** Concatenates a uniform scaling and a linear transformation matrix */ - // TODO returns an expression - template - inline typename internal::plain_matrix_type::type operator* (const MatrixBase& other) const - { return other * m_factor; } - - template - inline Matrix operator*(const RotationBase& r) const - { return r.toRotationMatrix() * m_factor; } - - /** \returns the inverse scaling */ - inline UniformScaling inverse() const - { return UniformScaling(Scalar(1)/m_factor); } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - inline UniformScaling cast() const - { return UniformScaling(NewScalarType(m_factor)); } - - /** Copy constructor with scalar type conversion */ - template - inline explicit UniformScaling(const UniformScaling& other) - { m_factor = Scalar(other.factor()); } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - bool isApprox(const UniformScaling& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return internal::isApprox(m_factor, other.factor(), prec); } - -}; - -/** \addtogroup Geometry_Module */ -//@{ - -/** Concatenates a linear transformation matrix and a uniform scaling - * \relates UniformScaling - */ -// NOTE this operator is defiend in MatrixBase and not as a friend function -// of UniformScaling to fix an internal crash of Intel's ICC -template -EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,Scalar,product) -operator*(const MatrixBase& matrix, const UniformScaling& s) -{ return matrix.derived() * s.factor(); } - -/** Constructs a uniform scaling from scale factor \a s */ -inline UniformScaling Scaling(float s) { return UniformScaling(s); } -/** Constructs a uniform scaling from scale factor \a s */ -inline UniformScaling Scaling(double s) { return UniformScaling(s); } -/** Constructs a uniform scaling from scale factor \a s */ -template -inline UniformScaling > Scaling(const std::complex& s) -{ return UniformScaling >(s); } - -/** Constructs a 2D axis aligned scaling */ -template -inline DiagonalMatrix Scaling(const Scalar& sx, const Scalar& sy) -{ return DiagonalMatrix(sx, sy); } -/** Constructs a 3D axis aligned scaling */ -template -inline DiagonalMatrix Scaling(const Scalar& sx, const Scalar& sy, const Scalar& sz) -{ return DiagonalMatrix(sx, sy, sz); } - -/** Constructs an axis aligned scaling expression from vector expression \a coeffs - * This is an alias for coeffs.asDiagonal() - */ -template -inline const DiagonalWrapper Scaling(const MatrixBase& coeffs) -{ return coeffs.asDiagonal(); } - -/** \deprecated */ -typedef DiagonalMatrix AlignedScaling2f; -/** \deprecated */ -typedef DiagonalMatrix AlignedScaling2d; -/** \deprecated */ -typedef DiagonalMatrix AlignedScaling3f; -/** \deprecated */ -typedef DiagonalMatrix AlignedScaling3d; -//@} - -template -template -inline Transform -UniformScaling::operator* (const Translation& t) const -{ - Transform res; - res.matrix().setZero(); - res.linear().diagonal().fill(factor()); - res.translation() = factor() * t.vector(); - res(Dim,Dim) = Scalar(1); - return res; -} - -} // end namespace Eigen - -#endif // EIGEN_SCALING_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Transform.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Transform.h deleted file mode 100644 index c21d9e550e8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Transform.h +++ /dev/null @@ -1,1542 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// Copyright (C) 2010 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSFORM_H -#define EIGEN_TRANSFORM_H - -namespace Eigen { - -namespace internal { - -template -struct transform_traits -{ - enum - { - Dim = Transform::Dim, - HDim = Transform::HDim, - Mode = Transform::Mode, - IsProjective = (int(Mode)==int(Projective)) - }; -}; - -template< typename TransformType, - typename MatrixType, - int Case = transform_traits::IsProjective ? 0 - : int(MatrixType::RowsAtCompileTime) == int(transform_traits::HDim) ? 1 - : 2, - int RhsCols = MatrixType::ColsAtCompileTime> -struct transform_right_product_impl; - -template< typename Other, - int Mode, - int Options, - int Dim, - int HDim, - int OtherRows=Other::RowsAtCompileTime, - int OtherCols=Other::ColsAtCompileTime> -struct transform_left_product_impl; - -template< typename Lhs, - typename Rhs, - bool AnyProjective = - transform_traits::IsProjective || - transform_traits::IsProjective> -struct transform_transform_product_impl; - -template< typename Other, - int Mode, - int Options, - int Dim, - int HDim, - int OtherRows=Other::RowsAtCompileTime, - int OtherCols=Other::ColsAtCompileTime> -struct transform_construct_from_matrix; - -template struct transform_take_affine_part; - -template -struct traits > -{ - typedef _Scalar Scalar; - typedef Eigen::Index StorageIndex; - typedef Dense StorageKind; - enum { - Dim1 = _Dim==Dynamic ? _Dim : _Dim + 1, - RowsAtCompileTime = _Mode==Projective ? Dim1 : _Dim, - ColsAtCompileTime = Dim1, - MaxRowsAtCompileTime = RowsAtCompileTime, - MaxColsAtCompileTime = ColsAtCompileTime, - Flags = 0 - }; -}; - -template struct transform_make_affine; - -} // end namespace internal - -/** \geometry_module \ingroup Geometry_Module - * - * \class Transform - * - * \brief Represents an homogeneous transformation in a N dimensional space - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients - * \tparam _Dim the dimension of the space - * \tparam _Mode the type of the transformation. Can be: - * - #Affine: the transformation is stored as a (Dim+1)^2 matrix, - * where the last row is assumed to be [0 ... 0 1]. - * - #AffineCompact: the transformation is stored as a (Dim)x(Dim+1) matrix. - * - #Projective: the transformation is stored as a (Dim+1)^2 matrix - * without any assumption. - * \tparam _Options has the same meaning as in class Matrix. It allows to specify DontAlign and/or RowMajor. - * These Options are passed directly to the underlying matrix type. - * - * The homography is internally represented and stored by a matrix which - * is available through the matrix() method. To understand the behavior of - * this class you have to think a Transform object as its internal - * matrix representation. The chosen convention is right multiply: - * - * \code v' = T * v \endcode - * - * Therefore, an affine transformation matrix M is shaped like this: - * - * \f$ \left( \begin{array}{cc} - * linear & translation\\ - * 0 ... 0 & 1 - * \end{array} \right) \f$ - * - * Note that for a projective transformation the last row can be anything, - * and then the interpretation of different parts might be sightly different. - * - * However, unlike a plain matrix, the Transform class provides many features - * simplifying both its assembly and usage. In particular, it can be composed - * with any other transformations (Transform,Translation,RotationBase,DiagonalMatrix) - * and can be directly used to transform implicit homogeneous vectors. All these - * operations are handled via the operator*. For the composition of transformations, - * its principle consists to first convert the right/left hand sides of the product - * to a compatible (Dim+1)^2 matrix and then perform a pure matrix product. - * Of course, internally, operator* tries to perform the minimal number of operations - * according to the nature of each terms. Likewise, when applying the transform - * to points, the latters are automatically promoted to homogeneous vectors - * before doing the matrix product. The conventions to homogeneous representations - * are performed as follow: - * - * \b Translation t (Dim)x(1): - * \f$ \left( \begin{array}{cc} - * I & t \\ - * 0\,...\,0 & 1 - * \end{array} \right) \f$ - * - * \b Rotation R (Dim)x(Dim): - * \f$ \left( \begin{array}{cc} - * R & 0\\ - * 0\,...\,0 & 1 - * \end{array} \right) \f$ - * - * \b Scaling \b DiagonalMatrix S (Dim)x(Dim): - * \f$ \left( \begin{array}{cc} - * S & 0\\ - * 0\,...\,0 & 1 - * \end{array} \right) \f$ - * - * \b Column \b point v (Dim)x(1): - * \f$ \left( \begin{array}{c} - * v\\ - * 1 - * \end{array} \right) \f$ - * - * \b Set \b of \b column \b points V1...Vn (Dim)x(n): - * \f$ \left( \begin{array}{ccc} - * v_1 & ... & v_n\\ - * 1 & ... & 1 - * \end{array} \right) \f$ - * - * The concatenation of a Transform object with any kind of other transformation - * always returns a Transform object. - * - * A little exception to the "as pure matrix product" rule is the case of the - * transformation of non homogeneous vectors by an affine transformation. In - * that case the last matrix row can be ignored, and the product returns non - * homogeneous vectors. - * - * Since, for instance, a Dim x Dim matrix is interpreted as a linear transformation, - * it is not possible to directly transform Dim vectors stored in a Dim x Dim matrix. - * The solution is either to use a Dim x Dynamic matrix or explicitly request a - * vector transformation by making the vector homogeneous: - * \code - * m' = T * m.colwise().homogeneous(); - * \endcode - * Note that there is zero overhead. - * - * Conversion methods from/to Qt's QMatrix and QTransform are available if the - * preprocessor token EIGEN_QT_SUPPORT is defined. - * - * This class can be extended with the help of the plugin mechanism described on the page - * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_TRANSFORM_PLUGIN. - * - * \sa class Matrix, class Quaternion - */ -template -class Transform -{ -public: - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_Dim==Dynamic ? Dynamic : (_Dim+1)*(_Dim+1)) - enum { - Mode = _Mode, - Options = _Options, - Dim = _Dim, ///< space dimension in which the transformation holds - HDim = _Dim+1, ///< size of a respective homogeneous vector - Rows = int(Mode)==(AffineCompact) ? Dim : HDim - }; - /** the scalar type of the coefficients */ - typedef _Scalar Scalar; - typedef Eigen::Index StorageIndex; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - /** type of the matrix used to represent the transformation */ - typedef typename internal::make_proper_matrix_type::type MatrixType; - /** constified MatrixType */ - typedef const MatrixType ConstMatrixType; - /** type of the matrix used to represent the linear part of the transformation */ - typedef Matrix LinearMatrixType; - /** type of read/write reference to the linear part of the transformation */ - typedef Block LinearPart; - /** type of read reference to the linear part of the transformation */ - typedef const Block ConstLinearPart; - /** type of read/write reference to the affine part of the transformation */ - typedef typename internal::conditional >::type AffinePart; - /** type of read reference to the affine part of the transformation */ - typedef typename internal::conditional >::type ConstAffinePart; - /** type of a vector */ - typedef Matrix VectorType; - /** type of a read/write reference to the translation part of the rotation */ - typedef Block::Flags & RowMajorBit)> TranslationPart; - /** type of a read reference to the translation part of the rotation */ - typedef const Block::Flags & RowMajorBit)> ConstTranslationPart; - /** corresponding translation type */ - typedef Translation TranslationType; - - // this intermediate enum is needed to avoid an ICE with gcc 3.4 and 4.0 - enum { TransformTimeDiagonalMode = ((Mode==int(Isometry))?Affine:int(Mode)) }; - /** The return type of the product between a diagonal matrix and a transform */ - typedef Transform TransformTimeDiagonalReturnType; - -protected: - - MatrixType m_matrix; - -public: - - /** Default constructor without initialization of the meaningful coefficients. - * If Mode==Affine or Mode==Isometry, then the last row is set to [0 ... 0 1] */ - EIGEN_DEVICE_FUNC inline Transform() - { - check_template_params(); - internal::transform_make_affine<(int(Mode)==Affine || int(Mode)==Isometry) ? Affine : AffineCompact>::run(m_matrix); - } - - EIGEN_DEVICE_FUNC inline Transform(const Transform& other) - { - check_template_params(); - m_matrix = other.m_matrix; - } - - EIGEN_DEVICE_FUNC inline explicit Transform(const TranslationType& t) - { - check_template_params(); - *this = t; - } - EIGEN_DEVICE_FUNC inline explicit Transform(const UniformScaling& s) - { - check_template_params(); - *this = s; - } - template - EIGEN_DEVICE_FUNC inline explicit Transform(const RotationBase& r) - { - check_template_params(); - *this = r; - } - - EIGEN_DEVICE_FUNC inline Transform& operator=(const Transform& other) - { m_matrix = other.m_matrix; return *this; } - - typedef internal::transform_take_affine_part take_affine_part; - - /** Constructs and initializes a transformation from a Dim^2 or a (Dim+1)^2 matrix. */ - template - EIGEN_DEVICE_FUNC inline explicit Transform(const EigenBase& other) - { - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY); - - check_template_params(); - internal::transform_construct_from_matrix::run(this, other.derived()); - } - - /** Set \c *this from a Dim^2 or (Dim+1)^2 matrix. */ - template - EIGEN_DEVICE_FUNC inline Transform& operator=(const EigenBase& other) - { - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY); - - internal::transform_construct_from_matrix::run(this, other.derived()); - return *this; - } - - template - EIGEN_DEVICE_FUNC inline Transform(const Transform& other) - { - check_template_params(); - // only the options change, we can directly copy the matrices - m_matrix = other.matrix(); - } - - template - EIGEN_DEVICE_FUNC inline Transform(const Transform& other) - { - check_template_params(); - // prevent conversions as: - // Affine | AffineCompact | Isometry = Projective - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(OtherMode==int(Projective), Mode==int(Projective)), - YOU_PERFORMED_AN_INVALID_TRANSFORMATION_CONVERSION) - - // prevent conversions as: - // Isometry = Affine | AffineCompact - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(OtherMode==int(Affine)||OtherMode==int(AffineCompact), Mode!=int(Isometry)), - YOU_PERFORMED_AN_INVALID_TRANSFORMATION_CONVERSION) - - enum { ModeIsAffineCompact = Mode == int(AffineCompact), - OtherModeIsAffineCompact = OtherMode == int(AffineCompact) - }; - - if(ModeIsAffineCompact == OtherModeIsAffineCompact) - { - // We need the block expression because the code is compiled for all - // combinations of transformations and will trigger a compile time error - // if one tries to assign the matrices directly - m_matrix.template block(0,0) = other.matrix().template block(0,0); - makeAffine(); - } - else if(OtherModeIsAffineCompact) - { - typedef typename Transform::MatrixType OtherMatrixType; - internal::transform_construct_from_matrix::run(this, other.matrix()); - } - else - { - // here we know that Mode == AffineCompact and OtherMode != AffineCompact. - // if OtherMode were Projective, the static assert above would already have caught it. - // So the only possibility is that OtherMode == Affine - linear() = other.linear(); - translation() = other.translation(); - } - } - - template - EIGEN_DEVICE_FUNC Transform(const ReturnByValue& other) - { - check_template_params(); - other.evalTo(*this); - } - - template - EIGEN_DEVICE_FUNC Transform& operator=(const ReturnByValue& other) - { - other.evalTo(*this); - return *this; - } - - #ifdef EIGEN_QT_SUPPORT - inline Transform(const QMatrix& other); - inline Transform& operator=(const QMatrix& other); - inline QMatrix toQMatrix(void) const; - inline Transform(const QTransform& other); - inline Transform& operator=(const QTransform& other); - inline QTransform toQTransform(void) const; - #endif - - EIGEN_DEVICE_FUNC Index rows() const { return int(Mode)==int(Projective) ? m_matrix.cols() : (m_matrix.cols()-1); } - EIGEN_DEVICE_FUNC Index cols() const { return m_matrix.cols(); } - - /** shortcut for m_matrix(row,col); - * \sa MatrixBase::operator(Index,Index) const */ - EIGEN_DEVICE_FUNC inline Scalar operator() (Index row, Index col) const { return m_matrix(row,col); } - /** shortcut for m_matrix(row,col); - * \sa MatrixBase::operator(Index,Index) */ - EIGEN_DEVICE_FUNC inline Scalar& operator() (Index row, Index col) { return m_matrix(row,col); } - - /** \returns a read-only expression of the transformation matrix */ - EIGEN_DEVICE_FUNC inline const MatrixType& matrix() const { return m_matrix; } - /** \returns a writable expression of the transformation matrix */ - EIGEN_DEVICE_FUNC inline MatrixType& matrix() { return m_matrix; } - - /** \returns a read-only expression of the linear part of the transformation */ - EIGEN_DEVICE_FUNC inline ConstLinearPart linear() const { return ConstLinearPart(m_matrix,0,0); } - /** \returns a writable expression of the linear part of the transformation */ - EIGEN_DEVICE_FUNC inline LinearPart linear() { return LinearPart(m_matrix,0,0); } - - /** \returns a read-only expression of the Dim x HDim affine part of the transformation */ - EIGEN_DEVICE_FUNC inline ConstAffinePart affine() const { return take_affine_part::run(m_matrix); } - /** \returns a writable expression of the Dim x HDim affine part of the transformation */ - EIGEN_DEVICE_FUNC inline AffinePart affine() { return take_affine_part::run(m_matrix); } - - /** \returns a read-only expression of the translation vector of the transformation */ - EIGEN_DEVICE_FUNC inline ConstTranslationPart translation() const { return ConstTranslationPart(m_matrix,0,Dim); } - /** \returns a writable expression of the translation vector of the transformation */ - EIGEN_DEVICE_FUNC inline TranslationPart translation() { return TranslationPart(m_matrix,0,Dim); } - - /** \returns an expression of the product between the transform \c *this and a matrix expression \a other. - * - * The right-hand-side \a other can be either: - * \li an homogeneous vector of size Dim+1, - * \li a set of homogeneous vectors of size Dim+1 x N, - * \li a transformation matrix of size Dim+1 x Dim+1. - * - * Moreover, if \c *this represents an affine transformation (i.e., Mode!=Projective), then \a other can also be: - * \li a point of size Dim (computes: \code this->linear() * other + this->translation()\endcode), - * \li a set of N points as a Dim x N matrix (computes: \code (this->linear() * other).colwise() + this->translation()\endcode), - * - * In all cases, the return type is a matrix or vector of same sizes as the right-hand-side \a other. - * - * If you want to interpret \a other as a linear or affine transformation, then first convert it to a Transform<> type, - * or do your own cooking. - * - * Finally, if you want to apply Affine transformations to vectors, then explicitly apply the linear part only: - * \code - * Affine3f A; - * Vector3f v1, v2; - * v2 = A.linear() * v1; - * \endcode - * - */ - // note: this function is defined here because some compilers cannot find the respective declaration - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename internal::transform_right_product_impl::ResultType - operator * (const EigenBase &other) const - { return internal::transform_right_product_impl::run(*this,other.derived()); } - - /** \returns the product expression of a transformation matrix \a a times a transform \a b - * - * The left hand side \a other can be either: - * \li a linear transformation matrix of size Dim x Dim, - * \li an affine transformation matrix of size Dim x Dim+1, - * \li a general transformation matrix of size Dim+1 x Dim+1. - */ - template friend - EIGEN_DEVICE_FUNC inline const typename internal::transform_left_product_impl::ResultType - operator * (const EigenBase &a, const Transform &b) - { return internal::transform_left_product_impl::run(a.derived(),b); } - - /** \returns The product expression of a transform \a a times a diagonal matrix \a b - * - * The rhs diagonal matrix is interpreted as an affine scaling transformation. The - * product results in a Transform of the same type (mode) as the lhs only if the lhs - * mode is no isometry. In that case, the returned transform is an affinity. - */ - template - EIGEN_DEVICE_FUNC inline const TransformTimeDiagonalReturnType - operator * (const DiagonalBase &b) const - { - TransformTimeDiagonalReturnType res(*this); - res.linearExt() *= b; - return res; - } - - /** \returns The product expression of a diagonal matrix \a a times a transform \a b - * - * The lhs diagonal matrix is interpreted as an affine scaling transformation. The - * product results in a Transform of the same type (mode) as the lhs only if the lhs - * mode is no isometry. In that case, the returned transform is an affinity. - */ - template - EIGEN_DEVICE_FUNC friend inline TransformTimeDiagonalReturnType - operator * (const DiagonalBase &a, const Transform &b) - { - TransformTimeDiagonalReturnType res; - res.linear().noalias() = a*b.linear(); - res.translation().noalias() = a*b.translation(); - if (Mode!=int(AffineCompact)) - res.matrix().row(Dim) = b.matrix().row(Dim); - return res; - } - - template - EIGEN_DEVICE_FUNC inline Transform& operator*=(const EigenBase& other) { return *this = *this * other; } - - /** Concatenates two transformations */ - EIGEN_DEVICE_FUNC inline const Transform operator * (const Transform& other) const - { - return internal::transform_transform_product_impl::run(*this,other); - } - - #if EIGEN_COMP_ICC -private: - // this intermediate structure permits to workaround a bug in ICC 11: - // error: template instantiation resulted in unexpected function type of "Eigen::Transform - // (const Eigen::Transform &) const" - // (the meaning of a name may have changed since the template declaration -- the type of the template is: - // "Eigen::internal::transform_transform_product_impl, - // Eigen::Transform, >::ResultType (const Eigen::Transform &) const") - // - template struct icc_11_workaround - { - typedef internal::transform_transform_product_impl > ProductType; - typedef typename ProductType::ResultType ResultType; - }; - -public: - /** Concatenates two different transformations */ - template - inline typename icc_11_workaround::ResultType - operator * (const Transform& other) const - { - typedef typename icc_11_workaround::ProductType ProductType; - return ProductType::run(*this,other); - } - #else - /** Concatenates two different transformations */ - template - EIGEN_DEVICE_FUNC inline typename internal::transform_transform_product_impl >::ResultType - operator * (const Transform& other) const - { - return internal::transform_transform_product_impl >::run(*this,other); - } - #endif - - /** \sa MatrixBase::setIdentity() */ - EIGEN_DEVICE_FUNC void setIdentity() { m_matrix.setIdentity(); } - - /** - * \brief Returns an identity transformation. - * \todo In the future this function should be returning a Transform expression. - */ - EIGEN_DEVICE_FUNC static const Transform Identity() - { - return Transform(MatrixType::Identity()); - } - - template - EIGEN_DEVICE_FUNC - inline Transform& scale(const MatrixBase &other); - - template - EIGEN_DEVICE_FUNC - inline Transform& prescale(const MatrixBase &other); - - EIGEN_DEVICE_FUNC inline Transform& scale(const Scalar& s); - EIGEN_DEVICE_FUNC inline Transform& prescale(const Scalar& s); - - template - EIGEN_DEVICE_FUNC - inline Transform& translate(const MatrixBase &other); - - template - EIGEN_DEVICE_FUNC - inline Transform& pretranslate(const MatrixBase &other); - - template - EIGEN_DEVICE_FUNC - inline Transform& rotate(const RotationType& rotation); - - template - EIGEN_DEVICE_FUNC - inline Transform& prerotate(const RotationType& rotation); - - EIGEN_DEVICE_FUNC Transform& shear(const Scalar& sx, const Scalar& sy); - EIGEN_DEVICE_FUNC Transform& preshear(const Scalar& sx, const Scalar& sy); - - EIGEN_DEVICE_FUNC inline Transform& operator=(const TranslationType& t); - - EIGEN_DEVICE_FUNC - inline Transform& operator*=(const TranslationType& t) { return translate(t.vector()); } - - EIGEN_DEVICE_FUNC inline Transform operator*(const TranslationType& t) const; - - EIGEN_DEVICE_FUNC - inline Transform& operator=(const UniformScaling& t); - - EIGEN_DEVICE_FUNC - inline Transform& operator*=(const UniformScaling& s) { return scale(s.factor()); } - - EIGEN_DEVICE_FUNC - inline TransformTimeDiagonalReturnType operator*(const UniformScaling& s) const - { - TransformTimeDiagonalReturnType res = *this; - res.scale(s.factor()); - return res; - } - - EIGEN_DEVICE_FUNC - inline Transform& operator*=(const DiagonalMatrix& s) { linearExt() *= s; return *this; } - - template - EIGEN_DEVICE_FUNC inline Transform& operator=(const RotationBase& r); - template - EIGEN_DEVICE_FUNC inline Transform& operator*=(const RotationBase& r) { return rotate(r.toRotationMatrix()); } - template - EIGEN_DEVICE_FUNC inline Transform operator*(const RotationBase& r) const; - - EIGEN_DEVICE_FUNC const LinearMatrixType rotation() const; - template - EIGEN_DEVICE_FUNC - void computeRotationScaling(RotationMatrixType *rotation, ScalingMatrixType *scaling) const; - template - EIGEN_DEVICE_FUNC - void computeScalingRotation(ScalingMatrixType *scaling, RotationMatrixType *rotation) const; - - template - EIGEN_DEVICE_FUNC - Transform& fromPositionOrientationScale(const MatrixBase &position, - const OrientationType& orientation, const MatrixBase &scale); - - EIGEN_DEVICE_FUNC - inline Transform inverse(TransformTraits traits = (TransformTraits)Mode) const; - - /** \returns a const pointer to the column major internal matrix */ - EIGEN_DEVICE_FUNC const Scalar* data() const { return m_matrix.data(); } - /** \returns a non-const pointer to the column major internal matrix */ - EIGEN_DEVICE_FUNC Scalar* data() { return m_matrix.data(); } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { return typename internal::cast_return_type >::type(*this); } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit Transform(const Transform& other) - { - check_template_params(); - m_matrix = other.matrix().template cast(); - } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const Transform& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return m_matrix.isApprox(other.m_matrix, prec); } - - /** Sets the last row to [0 ... 0 1] - */ - EIGEN_DEVICE_FUNC void makeAffine() - { - internal::transform_make_affine::run(m_matrix); - } - - /** \internal - * \returns the Dim x Dim linear part if the transformation is affine, - * and the HDim x Dim part for projective transformations. - */ - EIGEN_DEVICE_FUNC inline Block linearExt() - { return m_matrix.template block(0,0); } - /** \internal - * \returns the Dim x Dim linear part if the transformation is affine, - * and the HDim x Dim part for projective transformations. - */ - EIGEN_DEVICE_FUNC inline const Block linearExt() const - { return m_matrix.template block(0,0); } - - /** \internal - * \returns the translation part if the transformation is affine, - * and the last column for projective transformations. - */ - EIGEN_DEVICE_FUNC inline Block translationExt() - { return m_matrix.template block(0,Dim); } - /** \internal - * \returns the translation part if the transformation is affine, - * and the last column for projective transformations. - */ - EIGEN_DEVICE_FUNC inline const Block translationExt() const - { return m_matrix.template block(0,Dim); } - - - #ifdef EIGEN_TRANSFORM_PLUGIN - #include EIGEN_TRANSFORM_PLUGIN - #endif - -protected: - #ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void check_template_params() - { - EIGEN_STATIC_ASSERT((Options & (DontAlign|RowMajor)) == Options, INVALID_MATRIX_TEMPLATE_PARAMETERS) - } - #endif - -}; - -/** \ingroup Geometry_Module */ -typedef Transform Isometry2f; -/** \ingroup Geometry_Module */ -typedef Transform Isometry3f; -/** \ingroup Geometry_Module */ -typedef Transform Isometry2d; -/** \ingroup Geometry_Module */ -typedef Transform Isometry3d; - -/** \ingroup Geometry_Module */ -typedef Transform Affine2f; -/** \ingroup Geometry_Module */ -typedef Transform Affine3f; -/** \ingroup Geometry_Module */ -typedef Transform Affine2d; -/** \ingroup Geometry_Module */ -typedef Transform Affine3d; - -/** \ingroup Geometry_Module */ -typedef Transform AffineCompact2f; -/** \ingroup Geometry_Module */ -typedef Transform AffineCompact3f; -/** \ingroup Geometry_Module */ -typedef Transform AffineCompact2d; -/** \ingroup Geometry_Module */ -typedef Transform AffineCompact3d; - -/** \ingroup Geometry_Module */ -typedef Transform Projective2f; -/** \ingroup Geometry_Module */ -typedef Transform Projective3f; -/** \ingroup Geometry_Module */ -typedef Transform Projective2d; -/** \ingroup Geometry_Module */ -typedef Transform Projective3d; - -/************************** -*** Optional QT support *** -**************************/ - -#ifdef EIGEN_QT_SUPPORT -/** Initializes \c *this from a QMatrix assuming the dimension is 2. - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -Transform::Transform(const QMatrix& other) -{ - check_template_params(); - *this = other; -} - -/** Set \c *this from a QMatrix assuming the dimension is 2. - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -Transform& Transform::operator=(const QMatrix& other) -{ - EIGEN_STATIC_ASSERT(Dim==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - if (Mode == int(AffineCompact)) - m_matrix << other.m11(), other.m21(), other.dx(), - other.m12(), other.m22(), other.dy(); - else - m_matrix << other.m11(), other.m21(), other.dx(), - other.m12(), other.m22(), other.dy(), - 0, 0, 1; - return *this; -} - -/** \returns a QMatrix from \c *this assuming the dimension is 2. - * - * \warning this conversion might loss data if \c *this is not affine - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -QMatrix Transform::toQMatrix(void) const -{ - check_template_params(); - EIGEN_STATIC_ASSERT(Dim==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - return QMatrix(m_matrix.coeff(0,0), m_matrix.coeff(1,0), - m_matrix.coeff(0,1), m_matrix.coeff(1,1), - m_matrix.coeff(0,2), m_matrix.coeff(1,2)); -} - -/** Initializes \c *this from a QTransform assuming the dimension is 2. - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -Transform::Transform(const QTransform& other) -{ - check_template_params(); - *this = other; -} - -/** Set \c *this from a QTransform assuming the dimension is 2. - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -Transform& Transform::operator=(const QTransform& other) -{ - check_template_params(); - EIGEN_STATIC_ASSERT(Dim==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - if (Mode == int(AffineCompact)) - m_matrix << other.m11(), other.m21(), other.dx(), - other.m12(), other.m22(), other.dy(); - else - m_matrix << other.m11(), other.m21(), other.dx(), - other.m12(), other.m22(), other.dy(), - other.m13(), other.m23(), other.m33(); - return *this; -} - -/** \returns a QTransform from \c *this assuming the dimension is 2. - * - * This function is available only if the token EIGEN_QT_SUPPORT is defined. - */ -template -QTransform Transform::toQTransform(void) const -{ - EIGEN_STATIC_ASSERT(Dim==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - if (Mode == int(AffineCompact)) - return QTransform(m_matrix.coeff(0,0), m_matrix.coeff(1,0), - m_matrix.coeff(0,1), m_matrix.coeff(1,1), - m_matrix.coeff(0,2), m_matrix.coeff(1,2)); - else - return QTransform(m_matrix.coeff(0,0), m_matrix.coeff(1,0), m_matrix.coeff(2,0), - m_matrix.coeff(0,1), m_matrix.coeff(1,1), m_matrix.coeff(2,1), - m_matrix.coeff(0,2), m_matrix.coeff(1,2), m_matrix.coeff(2,2)); -} -#endif - -/********************* -*** Procedural API *** -*********************/ - -/** Applies on the right the non uniform scale transformation represented - * by the vector \a other to \c *this and returns a reference to \c *this. - * \sa prescale() - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::scale(const MatrixBase &other) -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,int(Dim)) - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - linearExt().noalias() = (linearExt() * other.asDiagonal()); - return *this; -} - -/** Applies on the right a uniform scale of a factor \a c to \c *this - * and returns a reference to \c *this. - * \sa prescale(Scalar) - */ -template -EIGEN_DEVICE_FUNC inline Transform& Transform::scale(const Scalar& s) -{ - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - linearExt() *= s; - return *this; -} - -/** Applies on the left the non uniform scale transformation represented - * by the vector \a other to \c *this and returns a reference to \c *this. - * \sa scale() - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::prescale(const MatrixBase &other) -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,int(Dim)) - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - affine().noalias() = (other.asDiagonal() * affine()); - return *this; -} - -/** Applies on the left a uniform scale of a factor \a c to \c *this - * and returns a reference to \c *this. - * \sa scale(Scalar) - */ -template -EIGEN_DEVICE_FUNC inline Transform& Transform::prescale(const Scalar& s) -{ - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - m_matrix.template topRows() *= s; - return *this; -} - -/** Applies on the right the translation matrix represented by the vector \a other - * to \c *this and returns a reference to \c *this. - * \sa pretranslate() - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::translate(const MatrixBase &other) -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,int(Dim)) - translationExt() += linearExt() * other; - return *this; -} - -/** Applies on the left the translation matrix represented by the vector \a other - * to \c *this and returns a reference to \c *this. - * \sa translate() - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::pretranslate(const MatrixBase &other) -{ - EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(OtherDerived,int(Dim)) - if(int(Mode)==int(Projective)) - affine() += other * m_matrix.row(Dim); - else - translation() += other; - return *this; -} - -/** Applies on the right the rotation represented by the rotation \a rotation - * to \c *this and returns a reference to \c *this. - * - * The template parameter \a RotationType is the type of the rotation which - * must be known by internal::toRotationMatrix<>. - * - * Natively supported types includes: - * - any scalar (2D), - * - a Dim x Dim matrix expression, - * - a Quaternion (3D), - * - a AngleAxis (3D) - * - * This mechanism is easily extendable to support user types such as Euler angles, - * or a pair of Quaternion for 4D rotations. - * - * \sa rotate(Scalar), class Quaternion, class AngleAxis, prerotate(RotationType) - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::rotate(const RotationType& rotation) -{ - linearExt() *= internal::toRotationMatrix(rotation); - return *this; -} - -/** Applies on the left the rotation represented by the rotation \a rotation - * to \c *this and returns a reference to \c *this. - * - * See rotate() for further details. - * - * \sa rotate() - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::prerotate(const RotationType& rotation) -{ - m_matrix.template block(0,0) = internal::toRotationMatrix(rotation) - * m_matrix.template block(0,0); - return *this; -} - -/** Applies on the right the shear transformation represented - * by the vector \a other to \c *this and returns a reference to \c *this. - * \warning 2D only. - * \sa preshear() - */ -template -EIGEN_DEVICE_FUNC Transform& -Transform::shear(const Scalar& sx, const Scalar& sy) -{ - EIGEN_STATIC_ASSERT(int(Dim)==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - VectorType tmp = linear().col(0)*sy + linear().col(1); - linear() << linear().col(0) + linear().col(1)*sx, tmp; - return *this; -} - -/** Applies on the left the shear transformation represented - * by the vector \a other to \c *this and returns a reference to \c *this. - * \warning 2D only. - * \sa shear() - */ -template -EIGEN_DEVICE_FUNC Transform& -Transform::preshear(const Scalar& sx, const Scalar& sy) -{ - EIGEN_STATIC_ASSERT(int(Dim)==2, YOU_MADE_A_PROGRAMMING_MISTAKE) - EIGEN_STATIC_ASSERT(Mode!=int(Isometry), THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS) - m_matrix.template block(0,0) = LinearMatrixType(1, sx, sy, 1) * m_matrix.template block(0,0); - return *this; -} - -/****************************************************** -*** Scaling, Translation and Rotation compatibility *** -******************************************************/ - -template -EIGEN_DEVICE_FUNC inline Transform& Transform::operator=(const TranslationType& t) -{ - linear().setIdentity(); - translation() = t.vector(); - makeAffine(); - return *this; -} - -template -EIGEN_DEVICE_FUNC inline Transform Transform::operator*(const TranslationType& t) const -{ - Transform res = *this; - res.translate(t.vector()); - return res; -} - -template -EIGEN_DEVICE_FUNC inline Transform& Transform::operator=(const UniformScaling& s) -{ - m_matrix.setZero(); - linear().diagonal().fill(s.factor()); - makeAffine(); - return *this; -} - -template -template -EIGEN_DEVICE_FUNC inline Transform& Transform::operator=(const RotationBase& r) -{ - linear() = internal::toRotationMatrix(r); - translation().setZero(); - makeAffine(); - return *this; -} - -template -template -EIGEN_DEVICE_FUNC inline Transform Transform::operator*(const RotationBase& r) const -{ - Transform res = *this; - res.rotate(r.derived()); - return res; -} - -/************************ -*** Special functions *** -************************/ - -/** \returns the rotation part of the transformation - * - * - * \svd_module - * - * \sa computeRotationScaling(), computeScalingRotation(), class SVD - */ -template -EIGEN_DEVICE_FUNC const typename Transform::LinearMatrixType -Transform::rotation() const -{ - LinearMatrixType result; - computeRotationScaling(&result, (LinearMatrixType*)0); - return result; -} - - -/** decomposes the linear part of the transformation as a product rotation x scaling, the scaling being - * not necessarily positive. - * - * If either pointer is zero, the corresponding computation is skipped. - * - * - * - * \svd_module - * - * \sa computeScalingRotation(), rotation(), class SVD - */ -template -template -EIGEN_DEVICE_FUNC void Transform::computeRotationScaling(RotationMatrixType *rotation, ScalingMatrixType *scaling) const -{ - JacobiSVD svd(linear(), ComputeFullU | ComputeFullV); - - Scalar x = (svd.matrixU() * svd.matrixV().adjoint()).determinant(); // so x has absolute value 1 - VectorType sv(svd.singularValues()); - sv.coeffRef(0) *= x; - if(scaling) scaling->lazyAssign(svd.matrixV() * sv.asDiagonal() * svd.matrixV().adjoint()); - if(rotation) - { - LinearMatrixType m(svd.matrixU()); - m.col(0) /= x; - rotation->lazyAssign(m * svd.matrixV().adjoint()); - } -} - -/** decomposes the linear part of the transformation as a product scaling x rotation, the scaling being - * not necessarily positive. - * - * If either pointer is zero, the corresponding computation is skipped. - * - * - * - * \svd_module - * - * \sa computeRotationScaling(), rotation(), class SVD - */ -template -template -EIGEN_DEVICE_FUNC void Transform::computeScalingRotation(ScalingMatrixType *scaling, RotationMatrixType *rotation) const -{ - JacobiSVD svd(linear(), ComputeFullU | ComputeFullV); - - Scalar x = (svd.matrixU() * svd.matrixV().adjoint()).determinant(); // so x has absolute value 1 - VectorType sv(svd.singularValues()); - sv.coeffRef(0) *= x; - if(scaling) scaling->lazyAssign(svd.matrixU() * sv.asDiagonal() * svd.matrixU().adjoint()); - if(rotation) - { - LinearMatrixType m(svd.matrixU()); - m.col(0) /= x; - rotation->lazyAssign(m * svd.matrixV().adjoint()); - } -} - -/** Convenient method to set \c *this from a position, orientation and scale - * of a 3D object. - */ -template -template -EIGEN_DEVICE_FUNC Transform& -Transform::fromPositionOrientationScale(const MatrixBase &position, - const OrientationType& orientation, const MatrixBase &scale) -{ - linear() = internal::toRotationMatrix(orientation); - linear() *= scale.asDiagonal(); - translation() = position; - makeAffine(); - return *this; -} - -namespace internal { - -template -struct transform_make_affine -{ - template - EIGEN_DEVICE_FUNC static void run(MatrixType &mat) - { - static const int Dim = MatrixType::ColsAtCompileTime-1; - mat.template block<1,Dim>(Dim,0).setZero(); - mat.coeffRef(Dim,Dim) = typename MatrixType::Scalar(1); - } -}; - -template<> -struct transform_make_affine -{ - template EIGEN_DEVICE_FUNC static void run(MatrixType &) { } -}; - -// selector needed to avoid taking the inverse of a 3x4 matrix -template -struct projective_transform_inverse -{ - EIGEN_DEVICE_FUNC static inline void run(const TransformType&, TransformType&) - {} -}; - -template -struct projective_transform_inverse -{ - EIGEN_DEVICE_FUNC static inline void run(const TransformType& m, TransformType& res) - { - res.matrix() = m.matrix().inverse(); - } -}; - -} // end namespace internal - - -/** - * - * \returns the inverse transformation according to some given knowledge - * on \c *this. - * - * \param hint allows to optimize the inversion process when the transformation - * is known to be not a general transformation (optional). The possible values are: - * - #Projective if the transformation is not necessarily affine, i.e., if the - * last row is not guaranteed to be [0 ... 0 1] - * - #Affine if the last row can be assumed to be [0 ... 0 1] - * - #Isometry if the transformation is only a concatenations of translations - * and rotations. - * The default is the template class parameter \c Mode. - * - * \warning unless \a traits is always set to NoShear or NoScaling, this function - * requires the generic inverse method of MatrixBase defined in the LU module. If - * you forget to include this module, then you will get hard to debug linking errors. - * - * \sa MatrixBase::inverse() - */ -template -EIGEN_DEVICE_FUNC Transform -Transform::inverse(TransformTraits hint) const -{ - Transform res; - if (hint == Projective) - { - internal::projective_transform_inverse::run(*this, res); - } - else - { - if (hint == Isometry) - { - res.matrix().template topLeftCorner() = linear().transpose(); - } - else if(hint&Affine) - { - res.matrix().template topLeftCorner() = linear().inverse(); - } - else - { - eigen_assert(false && "Invalid transform traits in Transform::Inverse"); - } - // translation and remaining parts - res.matrix().template topRightCorner() - = - res.matrix().template topLeftCorner() * translation(); - res.makeAffine(); // we do need this, because in the beginning res is uninitialized - } - return res; -} - -namespace internal { - -/***************************************************** -*** Specializations of take affine part *** -*****************************************************/ - -template struct transform_take_affine_part { - typedef typename TransformType::MatrixType MatrixType; - typedef typename TransformType::AffinePart AffinePart; - typedef typename TransformType::ConstAffinePart ConstAffinePart; - static inline AffinePart run(MatrixType& m) - { return m.template block(0,0); } - static inline ConstAffinePart run(const MatrixType& m) - { return m.template block(0,0); } -}; - -template -struct transform_take_affine_part > { - typedef typename Transform::MatrixType MatrixType; - static inline MatrixType& run(MatrixType& m) { return m; } - static inline const MatrixType& run(const MatrixType& m) { return m; } -}; - -/***************************************************** -*** Specializations of construct from matrix *** -*****************************************************/ - -template -struct transform_construct_from_matrix -{ - static inline void run(Transform *transform, const Other& other) - { - transform->linear() = other; - transform->translation().setZero(); - transform->makeAffine(); - } -}; - -template -struct transform_construct_from_matrix -{ - static inline void run(Transform *transform, const Other& other) - { - transform->affine() = other; - transform->makeAffine(); - } -}; - -template -struct transform_construct_from_matrix -{ - static inline void run(Transform *transform, const Other& other) - { transform->matrix() = other; } -}; - -template -struct transform_construct_from_matrix -{ - static inline void run(Transform *transform, const Other& other) - { transform->matrix() = other.template block(0,0); } -}; - -/********************************************************** -*** Specializations of operator* with rhs EigenBase *** -**********************************************************/ - -template -struct transform_product_result -{ - enum - { - Mode = - (LhsMode == (int)Projective || RhsMode == (int)Projective ) ? Projective : - (LhsMode == (int)Affine || RhsMode == (int)Affine ) ? Affine : - (LhsMode == (int)AffineCompact || RhsMode == (int)AffineCompact ) ? AffineCompact : - (LhsMode == (int)Isometry || RhsMode == (int)Isometry ) ? Isometry : Projective - }; -}; - -template< typename TransformType, typename MatrixType, int RhsCols> -struct transform_right_product_impl< TransformType, MatrixType, 0, RhsCols> -{ - typedef typename MatrixType::PlainObject ResultType; - - static EIGEN_STRONG_INLINE ResultType run(const TransformType& T, const MatrixType& other) - { - return T.matrix() * other; - } -}; - -template< typename TransformType, typename MatrixType, int RhsCols> -struct transform_right_product_impl< TransformType, MatrixType, 1, RhsCols> -{ - enum { - Dim = TransformType::Dim, - HDim = TransformType::HDim, - OtherRows = MatrixType::RowsAtCompileTime, - OtherCols = MatrixType::ColsAtCompileTime - }; - - typedef typename MatrixType::PlainObject ResultType; - - static EIGEN_STRONG_INLINE ResultType run(const TransformType& T, const MatrixType& other) - { - EIGEN_STATIC_ASSERT(OtherRows==HDim, YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES); - - typedef Block TopLeftLhs; - - ResultType res(other.rows(),other.cols()); - TopLeftLhs(res, 0, 0, Dim, other.cols()).noalias() = T.affine() * other; - res.row(OtherRows-1) = other.row(OtherRows-1); - - return res; - } -}; - -template< typename TransformType, typename MatrixType, int RhsCols> -struct transform_right_product_impl< TransformType, MatrixType, 2, RhsCols> -{ - enum { - Dim = TransformType::Dim, - HDim = TransformType::HDim, - OtherRows = MatrixType::RowsAtCompileTime, - OtherCols = MatrixType::ColsAtCompileTime - }; - - typedef typename MatrixType::PlainObject ResultType; - - static EIGEN_STRONG_INLINE ResultType run(const TransformType& T, const MatrixType& other) - { - EIGEN_STATIC_ASSERT(OtherRows==Dim, YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES); - - typedef Block TopLeftLhs; - ResultType res(Replicate(T.translation(),1,other.cols())); - TopLeftLhs(res, 0, 0, Dim, other.cols()).noalias() += T.linear() * other; - - return res; - } -}; - -template< typename TransformType, typename MatrixType > -struct transform_right_product_impl< TransformType, MatrixType, 2, 1> // rhs is a vector of size Dim -{ - typedef typename TransformType::MatrixType TransformMatrix; - enum { - Dim = TransformType::Dim, - HDim = TransformType::HDim, - OtherRows = MatrixType::RowsAtCompileTime, - WorkingRows = EIGEN_PLAIN_ENUM_MIN(TransformMatrix::RowsAtCompileTime,HDim) - }; - - typedef typename MatrixType::PlainObject ResultType; - - static EIGEN_STRONG_INLINE ResultType run(const TransformType& T, const MatrixType& other) - { - EIGEN_STATIC_ASSERT(OtherRows==Dim, YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES); - - Matrix rhs; - rhs.template head() = other; rhs[Dim] = typename ResultType::Scalar(1); - Matrix res(T.matrix() * rhs); - return res.template head(); - } -}; - -/********************************************************** -*** Specializations of operator* with lhs EigenBase *** -**********************************************************/ - -// generic HDim x HDim matrix * T => Projective -template -struct transform_left_product_impl -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType MatrixType; - typedef Transform ResultType; - static ResultType run(const Other& other,const TransformType& tr) - { return ResultType(other * tr.matrix()); } -}; - -// generic HDim x HDim matrix * AffineCompact => Projective -template -struct transform_left_product_impl -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType MatrixType; - typedef Transform ResultType; - static ResultType run(const Other& other,const TransformType& tr) - { - ResultType res; - res.matrix().noalias() = other.template block(0,0) * tr.matrix(); - res.matrix().col(Dim) += other.col(Dim); - return res; - } -}; - -// affine matrix * T -template -struct transform_left_product_impl -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType MatrixType; - typedef TransformType ResultType; - static ResultType run(const Other& other,const TransformType& tr) - { - ResultType res; - res.affine().noalias() = other * tr.matrix(); - res.matrix().row(Dim) = tr.matrix().row(Dim); - return res; - } -}; - -// affine matrix * AffineCompact -template -struct transform_left_product_impl -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType MatrixType; - typedef TransformType ResultType; - static ResultType run(const Other& other,const TransformType& tr) - { - ResultType res; - res.matrix().noalias() = other.template block(0,0) * tr.matrix(); - res.translation() += other.col(Dim); - return res; - } -}; - -// linear matrix * T -template -struct transform_left_product_impl -{ - typedef Transform TransformType; - typedef typename TransformType::MatrixType MatrixType; - typedef TransformType ResultType; - static ResultType run(const Other& other, const TransformType& tr) - { - TransformType res; - if(Mode!=int(AffineCompact)) - res.matrix().row(Dim) = tr.matrix().row(Dim); - res.matrix().template topRows().noalias() - = other * tr.matrix().template topRows(); - return res; - } -}; - -/********************************************************** -*** Specializations of operator* with another Transform *** -**********************************************************/ - -template -struct transform_transform_product_impl,Transform,false > -{ - enum { ResultMode = transform_product_result::Mode }; - typedef Transform Lhs; - typedef Transform Rhs; - typedef Transform ResultType; - static ResultType run(const Lhs& lhs, const Rhs& rhs) - { - ResultType res; - res.linear() = lhs.linear() * rhs.linear(); - res.translation() = lhs.linear() * rhs.translation() + lhs.translation(); - res.makeAffine(); - return res; - } -}; - -template -struct transform_transform_product_impl,Transform,true > -{ - typedef Transform Lhs; - typedef Transform Rhs; - typedef Transform ResultType; - static ResultType run(const Lhs& lhs, const Rhs& rhs) - { - return ResultType( lhs.matrix() * rhs.matrix() ); - } -}; - -template -struct transform_transform_product_impl,Transform,true > -{ - typedef Transform Lhs; - typedef Transform Rhs; - typedef Transform ResultType; - static ResultType run(const Lhs& lhs, const Rhs& rhs) - { - ResultType res; - res.matrix().template topRows() = lhs.matrix() * rhs.matrix(); - res.matrix().row(Dim) = rhs.matrix().row(Dim); - return res; - } -}; - -template -struct transform_transform_product_impl,Transform,true > -{ - typedef Transform Lhs; - typedef Transform Rhs; - typedef Transform ResultType; - static ResultType run(const Lhs& lhs, const Rhs& rhs) - { - ResultType res(lhs.matrix().template leftCols() * rhs.matrix()); - res.matrix().col(Dim) += lhs.matrix().col(Dim); - return res; - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRANSFORM_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Translation.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Translation.h deleted file mode 100644 index 0e99ce68e26..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Translation.h +++ /dev/null @@ -1,202 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSLATION_H -#define EIGEN_TRANSLATION_H - -namespace Eigen { - -/** \geometry_module \ingroup Geometry_Module - * - * \class Translation - * - * \brief Represents a translation transformation - * - * \tparam _Scalar the scalar type, i.e., the type of the coefficients. - * \tparam _Dim the dimension of the space, can be a compile time value or Dynamic - * - * \note This class is not aimed to be used to store a translation transformation, - * but rather to make easier the constructions and updates of Transform objects. - * - * \sa class Scaling, class Transform - */ -template -class Translation -{ -public: - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_Dim) - /** dimension of the space */ - enum { Dim = _Dim }; - /** the scalar type of the coefficients */ - typedef _Scalar Scalar; - /** corresponding vector type */ - typedef Matrix VectorType; - /** corresponding linear transformation matrix type */ - typedef Matrix LinearMatrixType; - /** corresponding affine transformation type */ - typedef Transform AffineTransformType; - /** corresponding isometric transformation type */ - typedef Transform IsometryTransformType; - -protected: - - VectorType m_coeffs; - -public: - - /** Default constructor without initialization. */ - EIGEN_DEVICE_FUNC Translation() {} - /** */ - EIGEN_DEVICE_FUNC inline Translation(const Scalar& sx, const Scalar& sy) - { - eigen_assert(Dim==2); - m_coeffs.x() = sx; - m_coeffs.y() = sy; - } - /** */ - EIGEN_DEVICE_FUNC inline Translation(const Scalar& sx, const Scalar& sy, const Scalar& sz) - { - eigen_assert(Dim==3); - m_coeffs.x() = sx; - m_coeffs.y() = sy; - m_coeffs.z() = sz; - } - /** Constructs and initialize the translation transformation from a vector of translation coefficients */ - EIGEN_DEVICE_FUNC explicit inline Translation(const VectorType& vector) : m_coeffs(vector) {} - - /** \brief Retruns the x-translation by value. **/ - EIGEN_DEVICE_FUNC inline Scalar x() const { return m_coeffs.x(); } - /** \brief Retruns the y-translation by value. **/ - EIGEN_DEVICE_FUNC inline Scalar y() const { return m_coeffs.y(); } - /** \brief Retruns the z-translation by value. **/ - EIGEN_DEVICE_FUNC inline Scalar z() const { return m_coeffs.z(); } - - /** \brief Retruns the x-translation as a reference. **/ - EIGEN_DEVICE_FUNC inline Scalar& x() { return m_coeffs.x(); } - /** \brief Retruns the y-translation as a reference. **/ - EIGEN_DEVICE_FUNC inline Scalar& y() { return m_coeffs.y(); } - /** \brief Retruns the z-translation as a reference. **/ - EIGEN_DEVICE_FUNC inline Scalar& z() { return m_coeffs.z(); } - - EIGEN_DEVICE_FUNC const VectorType& vector() const { return m_coeffs; } - EIGEN_DEVICE_FUNC VectorType& vector() { return m_coeffs; } - - EIGEN_DEVICE_FUNC const VectorType& translation() const { return m_coeffs; } - EIGEN_DEVICE_FUNC VectorType& translation() { return m_coeffs; } - - /** Concatenates two translation */ - EIGEN_DEVICE_FUNC inline Translation operator* (const Translation& other) const - { return Translation(m_coeffs + other.m_coeffs); } - - /** Concatenates a translation and a uniform scaling */ - EIGEN_DEVICE_FUNC inline AffineTransformType operator* (const UniformScaling& other) const; - - /** Concatenates a translation and a linear transformation */ - template - EIGEN_DEVICE_FUNC inline AffineTransformType operator* (const EigenBase& linear) const; - - /** Concatenates a translation and a rotation */ - template - EIGEN_DEVICE_FUNC inline IsometryTransformType operator*(const RotationBase& r) const - { return *this * IsometryTransformType(r); } - - /** \returns the concatenation of a linear transformation \a l with the translation \a t */ - // its a nightmare to define a templated friend function outside its declaration - template friend - EIGEN_DEVICE_FUNC inline AffineTransformType operator*(const EigenBase& linear, const Translation& t) - { - AffineTransformType res; - res.matrix().setZero(); - res.linear() = linear.derived(); - res.translation() = linear.derived() * t.m_coeffs; - res.matrix().row(Dim).setZero(); - res(Dim,Dim) = Scalar(1); - return res; - } - - /** Concatenates a translation and a transformation */ - template - EIGEN_DEVICE_FUNC inline Transform operator* (const Transform& t) const - { - Transform res = t; - res.pretranslate(m_coeffs); - return res; - } - - /** Applies translation to vector */ - template - inline typename internal::enable_if::type - operator* (const MatrixBase& vec) const - { return m_coeffs + vec.derived(); } - - /** \returns the inverse translation (opposite) */ - Translation inverse() const { return Translation(-m_coeffs); } - - static const Translation Identity() { return Translation(VectorType::Zero()); } - - /** \returns \c *this with scalar type casted to \a NewScalarType - * - * Note that if \a NewScalarType is equal to the current scalar type of \c *this - * then this function smartly returns a const reference to \c *this. - */ - template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const - { return typename internal::cast_return_type >::type(*this); } - - /** Copy constructor with scalar type conversion */ - template - EIGEN_DEVICE_FUNC inline explicit Translation(const Translation& other) - { m_coeffs = other.vector().template cast(); } - - /** \returns \c true if \c *this is approximately equal to \a other, within the precision - * determined by \a prec. - * - * \sa MatrixBase::isApprox() */ - EIGEN_DEVICE_FUNC bool isApprox(const Translation& other, const typename NumTraits::Real& prec = NumTraits::dummy_precision()) const - { return m_coeffs.isApprox(other.m_coeffs, prec); } - -}; - -/** \addtogroup Geometry_Module */ -//@{ -typedef Translation Translation2f; -typedef Translation Translation2d; -typedef Translation Translation3f; -typedef Translation Translation3d; -//@} - -template -EIGEN_DEVICE_FUNC inline typename Translation::AffineTransformType -Translation::operator* (const UniformScaling& other) const -{ - AffineTransformType res; - res.matrix().setZero(); - res.linear().diagonal().fill(other.factor()); - res.translation() = m_coeffs; - res(Dim,Dim) = Scalar(1); - return res; -} - -template -template -EIGEN_DEVICE_FUNC inline typename Translation::AffineTransformType -Translation::operator* (const EigenBase& linear) const -{ - AffineTransformType res; - res.matrix().setZero(); - res.linear() = linear.derived(); - res.translation() = m_coeffs; - res.matrix().row(Dim).setZero(); - res(Dim,Dim) = Scalar(1); - return res; -} - -} // end namespace Eigen - -#endif // EIGEN_TRANSLATION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Umeyama.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Umeyama.h deleted file mode 100644 index 6b755008fdc..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/Umeyama.h +++ /dev/null @@ -1,166 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_UMEYAMA_H -#define EIGEN_UMEYAMA_H - -// This file requires the user to include -// * Eigen/Core -// * Eigen/LU -// * Eigen/SVD -// * Eigen/Array - -namespace Eigen { - -#ifndef EIGEN_PARSED_BY_DOXYGEN - -// These helpers are required since it allows to use mixed types as parameters -// for the Umeyama. The problem with mixed parameters is that the return type -// cannot trivially be deduced when float and double types are mixed. -namespace internal { - -// Compile time return type deduction for different MatrixBase types. -// Different means here different alignment and parameters but the same underlying -// real scalar type. -template -struct umeyama_transform_matrix_type -{ - enum { - MinRowsAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(MatrixType::RowsAtCompileTime, OtherMatrixType::RowsAtCompileTime), - - // When possible we want to choose some small fixed size value since the result - // is likely to fit on the stack. So here, EIGEN_SIZE_MIN_PREFER_DYNAMIC is not what we want. - HomogeneousDimension = int(MinRowsAtCompileTime) == Dynamic ? Dynamic : int(MinRowsAtCompileTime)+1 - }; - - typedef Matrix::Scalar, - HomogeneousDimension, - HomogeneousDimension, - AutoAlign | (traits::Flags & RowMajorBit ? RowMajor : ColMajor), - HomogeneousDimension, - HomogeneousDimension - > type; -}; - -} - -#endif - -/** -* \geometry_module \ingroup Geometry_Module -* -* \brief Returns the transformation between two point sets. -* -* The algorithm is based on: -* "Least-squares estimation of transformation parameters between two point patterns", -* Shinji Umeyama, PAMI 1991, DOI: 10.1109/34.88573 -* -* It estimates parameters \f$ c, \mathbf{R}, \f$ and \f$ \mathbf{t} \f$ such that -* \f{align*} -* \frac{1}{n} \sum_{i=1}^n \vert\vert y_i - (c\mathbf{R}x_i + \mathbf{t}) \vert\vert_2^2 -* \f} -* is minimized. -* -* The algorithm is based on the analysis of the covariance matrix -* \f$ \Sigma_{\mathbf{x}\mathbf{y}} \in \mathbb{R}^{d \times d} \f$ -* of the input point sets \f$ \mathbf{x} \f$ and \f$ \mathbf{y} \f$ where -* \f$d\f$ is corresponding to the dimension (which is typically small). -* The analysis is involving the SVD having a complexity of \f$O(d^3)\f$ -* though the actual computational effort lies in the covariance -* matrix computation which has an asymptotic lower bound of \f$O(dm)\f$ when -* the input point sets have dimension \f$d \times m\f$. -* -* Currently the method is working only for floating point matrices. -* -* \todo Should the return type of umeyama() become a Transform? -* -* \param src Source points \f$ \mathbf{x} = \left( x_1, \hdots, x_n \right) \f$. -* \param dst Destination points \f$ \mathbf{y} = \left( y_1, \hdots, y_n \right) \f$. -* \param with_scaling Sets \f$ c=1 \f$ when false is passed. -* \return The homogeneous transformation -* \f{align*} -* T = \begin{bmatrix} c\mathbf{R} & \mathbf{t} \\ \mathbf{0} & 1 \end{bmatrix} -* \f} -* minimizing the residual above. This transformation is always returned as an -* Eigen::Matrix. -*/ -template -typename internal::umeyama_transform_matrix_type::type -umeyama(const MatrixBase& src, const MatrixBase& dst, bool with_scaling = true) -{ - typedef typename internal::umeyama_transform_matrix_type::type TransformationMatrixType; - typedef typename internal::traits::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - - EIGEN_STATIC_ASSERT(!NumTraits::IsComplex, NUMERIC_TYPE_MUST_BE_REAL) - EIGEN_STATIC_ASSERT((internal::is_same::Scalar>::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - - enum { Dimension = EIGEN_SIZE_MIN_PREFER_DYNAMIC(Derived::RowsAtCompileTime, OtherDerived::RowsAtCompileTime) }; - - typedef Matrix VectorType; - typedef Matrix MatrixType; - typedef typename internal::plain_matrix_type_row_major::type RowMajorMatrixType; - - const Index m = src.rows(); // dimension - const Index n = src.cols(); // number of measurements - - // required for demeaning ... - const RealScalar one_over_n = RealScalar(1) / static_cast(n); - - // computation of mean - const VectorType src_mean = src.rowwise().sum() * one_over_n; - const VectorType dst_mean = dst.rowwise().sum() * one_over_n; - - // demeaning of src and dst points - const RowMajorMatrixType src_demean = src.colwise() - src_mean; - const RowMajorMatrixType dst_demean = dst.colwise() - dst_mean; - - // Eq. (36)-(37) - const Scalar src_var = src_demean.rowwise().squaredNorm().sum() * one_over_n; - - // Eq. (38) - const MatrixType sigma = one_over_n * dst_demean * src_demean.transpose(); - - JacobiSVD svd(sigma, ComputeFullU | ComputeFullV); - - // Initialize the resulting transformation with an identity matrix... - TransformationMatrixType Rt = TransformationMatrixType::Identity(m+1,m+1); - - // Eq. (39) - VectorType S = VectorType::Ones(m); - - if ( svd.matrixU().determinant() * svd.matrixV().determinant() < 0 ) - S(m-1) = -1; - - // Eq. (40) and (43) - Rt.block(0,0,m,m).noalias() = svd.matrixU() * S.asDiagonal() * svd.matrixV().transpose(); - - if (with_scaling) - { - // Eq. (42) - const Scalar c = Scalar(1)/src_var * svd.singularValues().dot(S); - - // Eq. (41) - Rt.col(m).head(m) = dst_mean; - Rt.col(m).head(m).noalias() -= c*Rt.topLeftCorner(m,m)*src_mean; - Rt.block(0,0,m,m) *= c; - } - else - { - Rt.col(m).head(m) = dst_mean; - Rt.col(m).head(m).noalias() -= Rt.topLeftCorner(m,m)*src_mean; - } - - return Rt; -} - -} // end namespace Eigen - -#endif // EIGEN_UMEYAMA_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/arch/Geometry_SSE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/arch/Geometry_SSE.h deleted file mode 100644 index f68cab5834f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Geometry/arch/Geometry_SSE.h +++ /dev/null @@ -1,161 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Rohit Garg -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GEOMETRY_SSE_H -#define EIGEN_GEOMETRY_SSE_H - -namespace Eigen { - -namespace internal { - -template -struct quat_product -{ - enum { - AAlignment = traits::Alignment, - BAlignment = traits::Alignment, - ResAlignment = traits >::Alignment - }; - static inline Quaternion run(const QuaternionBase& _a, const QuaternionBase& _b) - { - Quaternion res; - const __m128 mask = _mm_setr_ps(0.f,0.f,0.f,-0.f); - __m128 a = _a.coeffs().template packet(0); - __m128 b = _b.coeffs().template packet(0); - __m128 s1 = _mm_mul_ps(vec4f_swizzle1(a,1,2,0,2),vec4f_swizzle1(b,2,0,1,2)); - __m128 s2 = _mm_mul_ps(vec4f_swizzle1(a,3,3,3,1),vec4f_swizzle1(b,0,1,2,1)); - pstoret( - &res.x(), - _mm_add_ps(_mm_sub_ps(_mm_mul_ps(a,vec4f_swizzle1(b,3,3,3,3)), - _mm_mul_ps(vec4f_swizzle1(a,2,0,1,0), - vec4f_swizzle1(b,1,2,0,0))), - _mm_xor_ps(mask,_mm_add_ps(s1,s2)))); - - return res; - } -}; - -template -struct quat_conj -{ - enum { - ResAlignment = traits >::Alignment - }; - static inline Quaternion run(const QuaternionBase& q) - { - Quaternion res; - const __m128 mask = _mm_setr_ps(-0.f,-0.f,-0.f,0.f); - pstoret(&res.x(), _mm_xor_ps(mask, q.coeffs().template packet::Alignment>(0))); - return res; - } -}; - - -template -struct cross3_impl -{ - enum { - ResAlignment = traits::type>::Alignment - }; - static inline typename plain_matrix_type::type - run(const VectorLhs& lhs, const VectorRhs& rhs) - { - __m128 a = lhs.template packet::Alignment>(0); - __m128 b = rhs.template packet::Alignment>(0); - __m128 mul1=_mm_mul_ps(vec4f_swizzle1(a,1,2,0,3),vec4f_swizzle1(b,2,0,1,3)); - __m128 mul2=_mm_mul_ps(vec4f_swizzle1(a,2,0,1,3),vec4f_swizzle1(b,1,2,0,3)); - typename plain_matrix_type::type res; - pstoret(&res.x(),_mm_sub_ps(mul1,mul2)); - return res; - } -}; - - - - -template -struct quat_product -{ - enum { - BAlignment = traits::Alignment, - ResAlignment = traits >::Alignment - }; - - static inline Quaternion run(const QuaternionBase& _a, const QuaternionBase& _b) - { - const Packet2d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); - - Quaternion res; - - const double* a = _a.coeffs().data(); - Packet2d b_xy = _b.coeffs().template packet(0); - Packet2d b_zw = _b.coeffs().template packet(2); - Packet2d a_xx = pset1(a[0]); - Packet2d a_yy = pset1(a[1]); - Packet2d a_zz = pset1(a[2]); - Packet2d a_ww = pset1(a[3]); - - // two temporaries: - Packet2d t1, t2; - - /* - * t1 = ww*xy + yy*zw - * t2 = zz*xy - xx*zw - * res.xy = t1 +/- swap(t2) - */ - t1 = padd(pmul(a_ww, b_xy), pmul(a_yy, b_zw)); - t2 = psub(pmul(a_zz, b_xy), pmul(a_xx, b_zw)); -#ifdef EIGEN_VECTORIZE_SSE3 - EIGEN_UNUSED_VARIABLE(mask) - pstoret(&res.x(), _mm_addsub_pd(t1, preverse(t2))); -#else - pstoret(&res.x(), padd(t1, pxor(mask,preverse(t2)))); -#endif - - /* - * t1 = ww*zw - yy*xy - * t2 = zz*zw + xx*xy - * res.zw = t1 -/+ swap(t2) = swap( swap(t1) +/- t2) - */ - t1 = psub(pmul(a_ww, b_zw), pmul(a_yy, b_xy)); - t2 = padd(pmul(a_zz, b_zw), pmul(a_xx, b_xy)); -#ifdef EIGEN_VECTORIZE_SSE3 - EIGEN_UNUSED_VARIABLE(mask) - pstoret(&res.z(), preverse(_mm_addsub_pd(preverse(t1), t2))); -#else - pstoret(&res.z(), psub(t1, pxor(mask,preverse(t2)))); -#endif - - return res; -} -}; - -template -struct quat_conj -{ - enum { - ResAlignment = traits >::Alignment - }; - static inline Quaternion run(const QuaternionBase& q) - { - Quaternion res; - const __m128d mask0 = _mm_setr_pd(-0.,-0.); - const __m128d mask2 = _mm_setr_pd(-0.,0.); - pstoret(&res.x(), _mm_xor_pd(mask0, q.coeffs().template packet::Alignment>(0))); - pstoret(&res.z(), _mm_xor_pd(mask2, q.coeffs().template packet::Alignment>(2))); - return res; - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_GEOMETRY_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h deleted file mode 100644 index 01a7ed1884d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h +++ /dev/null @@ -1,103 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Vincent Lejeune -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BLOCK_HOUSEHOLDER_H -#define EIGEN_BLOCK_HOUSEHOLDER_H - -// This file contains some helper function to deal with block householder reflectors - -namespace Eigen { - -namespace internal { - -/** \internal */ -// template -// void make_block_householder_triangular_factor(TriangularFactorType& triFactor, const VectorsType& vectors, const CoeffsType& hCoeffs) -// { -// typedef typename VectorsType::Scalar Scalar; -// const Index nbVecs = vectors.cols(); -// eigen_assert(triFactor.rows() == nbVecs && triFactor.cols() == nbVecs && vectors.rows()>=nbVecs); -// -// for(Index i = 0; i < nbVecs; i++) -// { -// Index rs = vectors.rows() - i; -// // Warning, note that hCoeffs may alias with vectors. -// // It is then necessary to copy it before modifying vectors(i,i). -// typename CoeffsType::Scalar h = hCoeffs(i); -// // This hack permits to pass trough nested Block<> and Transpose<> expressions. -// Scalar *Vii_ptr = const_cast(vectors.data() + vectors.outerStride()*i + vectors.innerStride()*i); -// Scalar Vii = *Vii_ptr; -// *Vii_ptr = Scalar(1); -// triFactor.col(i).head(i).noalias() = -h * vectors.block(i, 0, rs, i).adjoint() -// * vectors.col(i).tail(rs); -// *Vii_ptr = Vii; -// // FIXME add .noalias() once the triangular product can work inplace -// triFactor.col(i).head(i) = triFactor.block(0,0,i,i).template triangularView() -// * triFactor.col(i).head(i); -// triFactor(i,i) = hCoeffs(i); -// } -// } - -/** \internal */ -// This variant avoid modifications in vectors -template -void make_block_householder_triangular_factor(TriangularFactorType& triFactor, const VectorsType& vectors, const CoeffsType& hCoeffs) -{ - const Index nbVecs = vectors.cols(); - eigen_assert(triFactor.rows() == nbVecs && triFactor.cols() == nbVecs && vectors.rows()>=nbVecs); - - for(Index i = nbVecs-1; i >=0 ; --i) - { - Index rs = vectors.rows() - i - 1; - Index rt = nbVecs-i-1; - - if(rt>0) - { - triFactor.row(i).tail(rt).noalias() = -hCoeffs(i) * vectors.col(i).tail(rs).adjoint() - * vectors.bottomRightCorner(rs, rt).template triangularView(); - - // FIXME add .noalias() once the triangular product can work inplace - triFactor.row(i).tail(rt) = triFactor.row(i).tail(rt) * triFactor.bottomRightCorner(rt,rt).template triangularView(); - - } - triFactor(i,i) = hCoeffs(i); - } -} - -/** \internal - * if forward then perform mat = H0 * H1 * H2 * mat - * otherwise perform mat = H2 * H1 * H0 * mat - */ -template -void apply_block_householder_on_the_left(MatrixType& mat, const VectorsType& vectors, const CoeffsType& hCoeffs, bool forward) -{ - enum { TFactorSize = MatrixType::ColsAtCompileTime }; - Index nbVecs = vectors.cols(); - Matrix T(nbVecs,nbVecs); - - if(forward) make_block_householder_triangular_factor(T, vectors, hCoeffs); - else make_block_householder_triangular_factor(T, vectors, hCoeffs.conjugate()); - const TriangularView V(vectors); - - // A -= V T V^* A - Matrix tmp = V.adjoint() * mat; - // FIXME add .noalias() once the triangular product can work inplace - if(forward) tmp = T.template triangularView() * tmp; - else tmp = T.template triangularView().adjoint() * tmp; - mat.noalias() -= V * tmp; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BLOCK_HOUSEHOLDER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h deleted file mode 100644 index 80de2c3052c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h +++ /dev/null @@ -1,172 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HOUSEHOLDER_H -#define EIGEN_HOUSEHOLDER_H - -namespace Eigen { - -namespace internal { -template struct decrement_size -{ - enum { - ret = n==Dynamic ? n : n-1 - }; -}; -} - -/** Computes the elementary reflector H such that: - * \f$ H *this = [ beta 0 ... 0]^T \f$ - * where the transformation H is: - * \f$ H = I - tau v v^*\f$ - * and the vector v is: - * \f$ v^T = [1 essential^T] \f$ - * - * The essential part of the vector \c v is stored in *this. - * - * On output: - * \param tau the scaling factor of the Householder transformation - * \param beta the result of H * \c *this - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::applyHouseholderOnTheLeft(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -void MatrixBase::makeHouseholderInPlace(Scalar& tau, RealScalar& beta) -{ - VectorBlock::ret> essentialPart(derived(), 1, size()-1); - makeHouseholder(essentialPart, tau, beta); -} - -/** Computes the elementary reflector H such that: - * \f$ H *this = [ beta 0 ... 0]^T \f$ - * where the transformation H is: - * \f$ H = I - tau v v^*\f$ - * and the vector v is: - * \f$ v^T = [1 essential^T] \f$ - * - * On output: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param beta the result of H * \c *this - * - * \sa MatrixBase::makeHouseholderInPlace(), MatrixBase::applyHouseholderOnTheLeft(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -template -void MatrixBase::makeHouseholder( - EssentialPart& essential, - Scalar& tau, - RealScalar& beta) const -{ - using std::sqrt; - using numext::conj; - - EIGEN_STATIC_ASSERT_VECTOR_ONLY(EssentialPart) - VectorBlock tail(derived(), 1, size()-1); - - RealScalar tailSqNorm = size()==1 ? RealScalar(0) : tail.squaredNorm(); - Scalar c0 = coeff(0); - const RealScalar tol = (std::numeric_limits::min)(); - - if(tailSqNorm <= tol && numext::abs2(numext::imag(c0))<=tol) - { - tau = RealScalar(0); - beta = numext::real(c0); - essential.setZero(); - } - else - { - beta = sqrt(numext::abs2(c0) + tailSqNorm); - if (numext::real(c0)>=RealScalar(0)) - beta = -beta; - essential = tail / (c0 - beta); - tau = conj((beta - c0) / beta); - } -} - -/** Apply the elementary reflector H given by - * \f$ H = I - tau v v^*\f$ - * with - * \f$ v^T = [1 essential^T] \f$ - * from the left to a vector or matrix. - * - * On input: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param workspace a pointer to working space with at least - * this->cols() * essential.size() entries - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::makeHouseholderInPlace(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -template -void MatrixBase::applyHouseholderOnTheLeft( - const EssentialPart& essential, - const Scalar& tau, - Scalar* workspace) -{ - if(rows() == 1) - { - *this *= Scalar(1)-tau; - } - else if(tau!=Scalar(0)) - { - Map::type> tmp(workspace,cols()); - Block bottom(derived(), 1, 0, rows()-1, cols()); - tmp.noalias() = essential.adjoint() * bottom; - tmp += this->row(0); - this->row(0) -= tau * tmp; - bottom.noalias() -= tau * essential * tmp; - } -} - -/** Apply the elementary reflector H given by - * \f$ H = I - tau v v^*\f$ - * with - * \f$ v^T = [1 essential^T] \f$ - * from the right to a vector or matrix. - * - * On input: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param workspace a pointer to working space with at least - * this->cols() * essential.size() entries - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::makeHouseholderInPlace(), - * MatrixBase::applyHouseholderOnTheLeft() - */ -template -template -void MatrixBase::applyHouseholderOnTheRight( - const EssentialPart& essential, - const Scalar& tau, - Scalar* workspace) -{ - if(cols() == 1) - { - *this *= Scalar(1)-tau; - } - else if(tau!=Scalar(0)) - { - Map::type> tmp(workspace,rows()); - Block right(derived(), 0, 1, rows(), cols()-1); - tmp.noalias() = right * essential.conjugate(); - tmp += this->col(0); - this->col(0) -= tau * tmp; - right.noalias() -= tau * tmp * essential.transpose(); - } -} - -} // end namespace Eigen - -#endif // EIGEN_HOUSEHOLDER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h deleted file mode 100644 index 3ce0a693d9b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h +++ /dev/null @@ -1,470 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HOUSEHOLDER_SEQUENCE_H -#define EIGEN_HOUSEHOLDER_SEQUENCE_H - -namespace Eigen { - -/** \ingroup Householder_Module - * \householder_module - * \class HouseholderSequence - * \brief Sequence of Householder reflections acting on subspaces with decreasing size - * \tparam VectorsType type of matrix containing the Householder vectors - * \tparam CoeffsType type of vector containing the Householder coefficients - * \tparam Side either OnTheLeft (the default) or OnTheRight - * - * This class represents a product sequence of Householder reflections where the first Householder reflection - * acts on the whole space, the second Householder reflection leaves the one-dimensional subspace spanned by - * the first unit vector invariant, the third Householder reflection leaves the two-dimensional subspace - * spanned by the first two unit vectors invariant, and so on up to the last reflection which leaves all but - * one dimensions invariant and acts only on the last dimension. Such sequences of Householder reflections - * are used in several algorithms to zero out certain parts of a matrix. Indeed, the methods - * HessenbergDecomposition::matrixQ(), Tridiagonalization::matrixQ(), HouseholderQR::householderQ(), - * and ColPivHouseholderQR::householderQ() all return a %HouseholderSequence. - * - * More precisely, the class %HouseholderSequence represents an \f$ n \times n \f$ matrix \f$ H \f$ of the - * form \f$ H = \prod_{i=0}^{n-1} H_i \f$ where the i-th Householder reflection is \f$ H_i = I - h_i v_i - * v_i^* \f$. The i-th Householder coefficient \f$ h_i \f$ is a scalar and the i-th Householder vector \f$ - * v_i \f$ is a vector of the form - * \f[ - * v_i = [\underbrace{0, \ldots, 0}_{i-1\mbox{ zeros}}, 1, \underbrace{*, \ldots,*}_{n-i\mbox{ arbitrary entries}} ]. - * \f] - * The last \f$ n-i \f$ entries of \f$ v_i \f$ are called the essential part of the Householder vector. - * - * Typical usages are listed below, where H is a HouseholderSequence: - * \code - * A.applyOnTheRight(H); // A = A * H - * A.applyOnTheLeft(H); // A = H * A - * A.applyOnTheRight(H.adjoint()); // A = A * H^* - * A.applyOnTheLeft(H.adjoint()); // A = H^* * A - * MatrixXd Q = H; // conversion to a dense matrix - * \endcode - * In addition to the adjoint, you can also apply the inverse (=adjoint), the transpose, and the conjugate operators. - * - * See the documentation for HouseholderSequence(const VectorsType&, const CoeffsType&) for an example. - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ - -namespace internal { - -template -struct traits > -{ - typedef typename VectorsType::Scalar Scalar; - typedef typename VectorsType::StorageIndex StorageIndex; - typedef typename VectorsType::StorageKind StorageKind; - enum { - RowsAtCompileTime = Side==OnTheLeft ? traits::RowsAtCompileTime - : traits::ColsAtCompileTime, - ColsAtCompileTime = RowsAtCompileTime, - MaxRowsAtCompileTime = Side==OnTheLeft ? traits::MaxRowsAtCompileTime - : traits::MaxColsAtCompileTime, - MaxColsAtCompileTime = MaxRowsAtCompileTime, - Flags = 0 - }; -}; - -struct HouseholderSequenceShape {}; - -template -struct evaluator_traits > - : public evaluator_traits_base > -{ - typedef HouseholderSequenceShape Shape; -}; - -template -struct hseq_side_dependent_impl -{ - typedef Block EssentialVectorType; - typedef HouseholderSequence HouseholderSequenceType; - static inline const EssentialVectorType essentialVector(const HouseholderSequenceType& h, Index k) - { - Index start = k+1+h.m_shift; - return Block(h.m_vectors, start, k, h.rows()-start, 1); - } -}; - -template -struct hseq_side_dependent_impl -{ - typedef Transpose > EssentialVectorType; - typedef HouseholderSequence HouseholderSequenceType; - static inline const EssentialVectorType essentialVector(const HouseholderSequenceType& h, Index k) - { - Index start = k+1+h.m_shift; - return Block(h.m_vectors, k, start, 1, h.rows()-start).transpose(); - } -}; - -template struct matrix_type_times_scalar_type -{ - typedef typename ScalarBinaryOpTraits::ReturnType - ResultScalar; - typedef Matrix Type; -}; - -} // end namespace internal - -template class HouseholderSequence - : public EigenBase > -{ - typedef typename internal::hseq_side_dependent_impl::EssentialVectorType EssentialVectorType; - - public: - enum { - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime - }; - typedef typename internal::traits::Scalar Scalar; - - typedef HouseholderSequence< - typename internal::conditional::IsComplex, - typename internal::remove_all::type, - VectorsType>::type, - typename internal::conditional::IsComplex, - typename internal::remove_all::type, - CoeffsType>::type, - Side - > ConjugateReturnType; - - /** \brief Constructor. - * \param[in] v %Matrix containing the essential parts of the Householder vectors - * \param[in] h Vector containing the Householder coefficients - * - * Constructs the Householder sequence with coefficients given by \p h and vectors given by \p v. The - * i-th Householder coefficient \f$ h_i \f$ is given by \p h(i) and the essential part of the i-th - * Householder vector \f$ v_i \f$ is given by \p v(k,i) with \p k > \p i (the subdiagonal part of the - * i-th column). If \p v has fewer columns than rows, then the Householder sequence contains as many - * Householder reflections as there are columns. - * - * \note The %HouseholderSequence object stores \p v and \p h by reference. - * - * Example: \include HouseholderSequence_HouseholderSequence.cpp - * Output: \verbinclude HouseholderSequence_HouseholderSequence.out - * - * \sa setLength(), setShift() - */ - HouseholderSequence(const VectorsType& v, const CoeffsType& h) - : m_vectors(v), m_coeffs(h), m_trans(false), m_length(v.diagonalSize()), - m_shift(0) - { - } - - /** \brief Copy constructor. */ - HouseholderSequence(const HouseholderSequence& other) - : m_vectors(other.m_vectors), - m_coeffs(other.m_coeffs), - m_trans(other.m_trans), - m_length(other.m_length), - m_shift(other.m_shift) - { - } - - /** \brief Number of rows of transformation viewed as a matrix. - * \returns Number of rows - * \details This equals the dimension of the space that the transformation acts on. - */ - Index rows() const { return Side==OnTheLeft ? m_vectors.rows() : m_vectors.cols(); } - - /** \brief Number of columns of transformation viewed as a matrix. - * \returns Number of columns - * \details This equals the dimension of the space that the transformation acts on. - */ - Index cols() const { return rows(); } - - /** \brief Essential part of a Householder vector. - * \param[in] k Index of Householder reflection - * \returns Vector containing non-trivial entries of k-th Householder vector - * - * This function returns the essential part of the Householder vector \f$ v_i \f$. This is a vector of - * length \f$ n-i \f$ containing the last \f$ n-i \f$ entries of the vector - * \f[ - * v_i = [\underbrace{0, \ldots, 0}_{i-1\mbox{ zeros}}, 1, \underbrace{*, \ldots,*}_{n-i\mbox{ arbitrary entries}} ]. - * \f] - * The index \f$ i \f$ equals \p k + shift(), corresponding to the k-th column of the matrix \p v - * passed to the constructor. - * - * \sa setShift(), shift() - */ - const EssentialVectorType essentialVector(Index k) const - { - eigen_assert(k >= 0 && k < m_length); - return internal::hseq_side_dependent_impl::essentialVector(*this, k); - } - - /** \brief %Transpose of the Householder sequence. */ - HouseholderSequence transpose() const - { - return HouseholderSequence(*this).setTrans(!m_trans); - } - - /** \brief Complex conjugate of the Householder sequence. */ - ConjugateReturnType conjugate() const - { - return ConjugateReturnType(m_vectors.conjugate(), m_coeffs.conjugate()) - .setTrans(m_trans) - .setLength(m_length) - .setShift(m_shift); - } - - /** \brief Adjoint (conjugate transpose) of the Householder sequence. */ - ConjugateReturnType adjoint() const - { - return conjugate().setTrans(!m_trans); - } - - /** \brief Inverse of the Householder sequence (equals the adjoint). */ - ConjugateReturnType inverse() const { return adjoint(); } - - /** \internal */ - template inline void evalTo(DestType& dst) const - { - Matrix workspace(rows()); - evalTo(dst, workspace); - } - - /** \internal */ - template - void evalTo(Dest& dst, Workspace& workspace) const - { - workspace.resize(rows()); - Index vecs = m_length; - if(internal::is_same_dense(dst,m_vectors)) - { - // in-place - dst.diagonal().setOnes(); - dst.template triangularView().setZero(); - for(Index k = vecs-1; k >= 0; --k) - { - Index cornerSize = rows() - k - m_shift; - if(m_trans) - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheRight(essentialVector(k), m_coeffs.coeff(k), workspace.data()); - else - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheLeft(essentialVector(k), m_coeffs.coeff(k), workspace.data()); - - // clear the off diagonal vector - dst.col(k).tail(rows()-k-1).setZero(); - } - // clear the remaining columns if needed - for(Index k = 0; k= 0; --k) - { - Index cornerSize = rows() - k - m_shift; - if(m_trans) - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheRight(essentialVector(k), m_coeffs.coeff(k), &workspace.coeffRef(0)); - else - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheLeft(essentialVector(k), m_coeffs.coeff(k), &workspace.coeffRef(0)); - } - } - } - - /** \internal */ - template inline void applyThisOnTheRight(Dest& dst) const - { - Matrix workspace(dst.rows()); - applyThisOnTheRight(dst, workspace); - } - - /** \internal */ - template - inline void applyThisOnTheRight(Dest& dst, Workspace& workspace) const - { - workspace.resize(dst.rows()); - for(Index k = 0; k < m_length; ++k) - { - Index actual_k = m_trans ? m_length-k-1 : k; - dst.rightCols(rows()-m_shift-actual_k) - .applyHouseholderOnTheRight(essentialVector(actual_k), m_coeffs.coeff(actual_k), workspace.data()); - } - } - - /** \internal */ - template inline void applyThisOnTheLeft(Dest& dst) const - { - Matrix workspace; - applyThisOnTheLeft(dst, workspace); - } - - /** \internal */ - template - inline void applyThisOnTheLeft(Dest& dst, Workspace& workspace) const - { - const Index BlockSize = 48; - // if the entries are large enough, then apply the reflectors by block - if(m_length>=BlockSize && dst.cols()>1) - { - for(Index i = 0; i < m_length; i+=BlockSize) - { - Index end = m_trans ? (std::min)(m_length,i+BlockSize) : m_length-i; - Index k = m_trans ? i : (std::max)(Index(0),end-BlockSize); - Index bs = end-k; - Index start = k + m_shift; - - typedef Block::type,Dynamic,Dynamic> SubVectorsType; - SubVectorsType sub_vecs1(m_vectors.const_cast_derived(), Side==OnTheRight ? k : start, - Side==OnTheRight ? start : k, - Side==OnTheRight ? bs : m_vectors.rows()-start, - Side==OnTheRight ? m_vectors.cols()-start : bs); - typename internal::conditional, SubVectorsType&>::type sub_vecs(sub_vecs1); - Block sub_dst(dst,dst.rows()-rows()+m_shift+k,0, rows()-m_shift-k,dst.cols()); - apply_block_householder_on_the_left(sub_dst, sub_vecs, m_coeffs.segment(k, bs), !m_trans); - } - } - else - { - workspace.resize(dst.cols()); - for(Index k = 0; k < m_length; ++k) - { - Index actual_k = m_trans ? k : m_length-k-1; - dst.bottomRows(rows()-m_shift-actual_k) - .applyHouseholderOnTheLeft(essentialVector(actual_k), m_coeffs.coeff(actual_k), workspace.data()); - } - } - } - - /** \brief Computes the product of a Householder sequence with a matrix. - * \param[in] other %Matrix being multiplied. - * \returns Expression object representing the product. - * - * This function computes \f$ HM \f$ where \f$ H \f$ is the Householder sequence represented by \p *this - * and \f$ M \f$ is the matrix \p other. - */ - template - typename internal::matrix_type_times_scalar_type::Type operator*(const MatrixBase& other) const - { - typename internal::matrix_type_times_scalar_type::Type - res(other.template cast::ResultScalar>()); - applyThisOnTheLeft(res); - return res; - } - - template friend struct internal::hseq_side_dependent_impl; - - /** \brief Sets the length of the Householder sequence. - * \param [in] length New value for the length. - * - * By default, the length \f$ n \f$ of the Householder sequence \f$ H = H_0 H_1 \ldots H_{n-1} \f$ is set - * to the number of columns of the matrix \p v passed to the constructor, or the number of rows if that - * is smaller. After this function is called, the length equals \p length. - * - * \sa length() - */ - HouseholderSequence& setLength(Index length) - { - m_length = length; - return *this; - } - - /** \brief Sets the shift of the Householder sequence. - * \param [in] shift New value for the shift. - * - * By default, a %HouseholderSequence object represents \f$ H = H_0 H_1 \ldots H_{n-1} \f$ and the i-th - * column of the matrix \p v passed to the constructor corresponds to the i-th Householder - * reflection. After this function is called, the object represents \f$ H = H_{\mathrm{shift}} - * H_{\mathrm{shift}+1} \ldots H_{n-1} \f$ and the i-th column of \p v corresponds to the (shift+i)-th - * Householder reflection. - * - * \sa shift() - */ - HouseholderSequence& setShift(Index shift) - { - m_shift = shift; - return *this; - } - - Index length() const { return m_length; } /**< \brief Returns the length of the Householder sequence. */ - Index shift() const { return m_shift; } /**< \brief Returns the shift of the Householder sequence. */ - - /* Necessary for .adjoint() and .conjugate() */ - template friend class HouseholderSequence; - - protected: - - /** \brief Sets the transpose flag. - * \param [in] trans New value of the transpose flag. - * - * By default, the transpose flag is not set. If the transpose flag is set, then this object represents - * \f$ H^T = H_{n-1}^T \ldots H_1^T H_0^T \f$ instead of \f$ H = H_0 H_1 \ldots H_{n-1} \f$. - * - * \sa trans() - */ - HouseholderSequence& setTrans(bool trans) - { - m_trans = trans; - return *this; - } - - bool trans() const { return m_trans; } /**< \brief Returns the transpose flag. */ - - typename VectorsType::Nested m_vectors; - typename CoeffsType::Nested m_coeffs; - bool m_trans; - Index m_length; - Index m_shift; -}; - -/** \brief Computes the product of a matrix with a Householder sequence. - * \param[in] other %Matrix being multiplied. - * \param[in] h %HouseholderSequence being multiplied. - * \returns Expression object representing the product. - * - * This function computes \f$ MH \f$ where \f$ M \f$ is the matrix \p other and \f$ H \f$ is the - * Householder sequence represented by \p h. - */ -template -typename internal::matrix_type_times_scalar_type::Type operator*(const MatrixBase& other, const HouseholderSequence& h) -{ - typename internal::matrix_type_times_scalar_type::Type - res(other.template cast::ResultScalar>()); - h.applyThisOnTheRight(res); - return res; -} - -/** \ingroup Householder_Module \householder_module - * \brief Convenience function for constructing a Householder sequence. - * \returns A HouseholderSequence constructed from the specified arguments. - */ -template -HouseholderSequence householderSequence(const VectorsType& v, const CoeffsType& h) -{ - return HouseholderSequence(v, h); -} - -/** \ingroup Householder_Module \householder_module - * \brief Convenience function for constructing a Householder sequence. - * \returns A HouseholderSequence constructed from the specified arguments. - * \details This function differs from householderSequence() in that the template argument \p OnTheSide of - * the constructed HouseholderSequence is set to OnTheRight, instead of the default OnTheLeft. - */ -template -HouseholderSequence rightHouseholderSequence(const VectorsType& v, const CoeffsType& h) -{ - return HouseholderSequence(v, h); -} - -} // end namespace Eigen - -#endif // EIGEN_HOUSEHOLDER_SEQUENCE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h deleted file mode 100644 index f66c846ef79..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h +++ /dev/null @@ -1,226 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2011-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BASIC_PRECONDITIONERS_H -#define EIGEN_BASIC_PRECONDITIONERS_H - -namespace Eigen { - -/** \ingroup IterativeLinearSolvers_Module - * \brief A preconditioner based on the digonal entries - * - * This class allows to approximately solve for A.x = b problems assuming A is a diagonal matrix. - * In other words, this preconditioner neglects all off diagonal entries and, in Eigen's language, solves for: - \code - A.diagonal().asDiagonal() . x = b - \endcode - * - * \tparam _Scalar the type of the scalar. - * - * \implsparsesolverconcept - * - * This preconditioner is suitable for both selfadjoint and general problems. - * The diagonal entries are pre-inverted and stored into a dense vector. - * - * \note A variant that has yet to be implemented would attempt to preserve the norm of each column. - * - * \sa class LeastSquareDiagonalPreconditioner, class ConjugateGradient - */ -template -class DiagonalPreconditioner -{ - typedef _Scalar Scalar; - typedef Matrix Vector; - public: - typedef typename Vector::StorageIndex StorageIndex; - enum { - ColsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic - }; - - DiagonalPreconditioner() : m_isInitialized(false) {} - - template - explicit DiagonalPreconditioner(const MatType& mat) : m_invdiag(mat.cols()) - { - compute(mat); - } - - Index rows() const { return m_invdiag.size(); } - Index cols() const { return m_invdiag.size(); } - - template - DiagonalPreconditioner& analyzePattern(const MatType& ) - { - return *this; - } - - template - DiagonalPreconditioner& factorize(const MatType& mat) - { - m_invdiag.resize(mat.cols()); - for(int j=0; j - DiagonalPreconditioner& compute(const MatType& mat) - { - return factorize(mat); - } - - /** \internal */ - template - void _solve_impl(const Rhs& b, Dest& x) const - { - x = m_invdiag.array() * b.array() ; - } - - template inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "DiagonalPreconditioner is not initialized."); - eigen_assert(m_invdiag.size()==b.rows() - && "DiagonalPreconditioner::solve(): invalid number of rows of the right hand side matrix b"); - return Solve(*this, b.derived()); - } - - ComputationInfo info() { return Success; } - - protected: - Vector m_invdiag; - bool m_isInitialized; -}; - -/** \ingroup IterativeLinearSolvers_Module - * \brief Jacobi preconditioner for LeastSquaresConjugateGradient - * - * This class allows to approximately solve for A' A x = A' b problems assuming A' A is a diagonal matrix. - * In other words, this preconditioner neglects all off diagonal entries and, in Eigen's language, solves for: - \code - (A.adjoint() * A).diagonal().asDiagonal() * x = b - \endcode - * - * \tparam _Scalar the type of the scalar. - * - * \implsparsesolverconcept - * - * The diagonal entries are pre-inverted and stored into a dense vector. - * - * \sa class LeastSquaresConjugateGradient, class DiagonalPreconditioner - */ -template -class LeastSquareDiagonalPreconditioner : public DiagonalPreconditioner<_Scalar> -{ - typedef _Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef DiagonalPreconditioner<_Scalar> Base; - using Base::m_invdiag; - public: - - LeastSquareDiagonalPreconditioner() : Base() {} - - template - explicit LeastSquareDiagonalPreconditioner(const MatType& mat) : Base() - { - compute(mat); - } - - template - LeastSquareDiagonalPreconditioner& analyzePattern(const MatType& ) - { - return *this; - } - - template - LeastSquareDiagonalPreconditioner& factorize(const MatType& mat) - { - // Compute the inverse squared-norm of each column of mat - m_invdiag.resize(mat.cols()); - if(MatType::IsRowMajor) - { - m_invdiag.setZero(); - for(Index j=0; jRealScalar(0)) - m_invdiag(j) = RealScalar(1)/numext::real(m_invdiag(j)); - } - else - { - for(Index j=0; jRealScalar(0)) - m_invdiag(j) = RealScalar(1)/sum; - else - m_invdiag(j) = RealScalar(1); - } - } - Base::m_isInitialized = true; - return *this; - } - - template - LeastSquareDiagonalPreconditioner& compute(const MatType& mat) - { - return factorize(mat); - } - - ComputationInfo info() { return Success; } - - protected: -}; - -/** \ingroup IterativeLinearSolvers_Module - * \brief A naive preconditioner which approximates any matrix as the identity matrix - * - * \implsparsesolverconcept - * - * \sa class DiagonalPreconditioner - */ -class IdentityPreconditioner -{ - public: - - IdentityPreconditioner() {} - - template - explicit IdentityPreconditioner(const MatrixType& ) {} - - template - IdentityPreconditioner& analyzePattern(const MatrixType& ) { return *this; } - - template - IdentityPreconditioner& factorize(const MatrixType& ) { return *this; } - - template - IdentityPreconditioner& compute(const MatrixType& ) { return *this; } - - template - inline const Rhs& solve(const Rhs& b) const { return b; } - - ComputationInfo info() { return Success; } -}; - -} // end namespace Eigen - -#endif // EIGEN_BASIC_PRECONDITIONERS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BiCGSTAB.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BiCGSTAB.h deleted file mode 100644 index 454f4681497..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/BiCGSTAB.h +++ /dev/null @@ -1,228 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2011-2014 Gael Guennebaud -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BICGSTAB_H -#define EIGEN_BICGSTAB_H - -namespace Eigen { - -namespace internal { - -/** \internal Low-level bi conjugate gradient stabilized algorithm - * \param mat The matrix A - * \param rhs The right hand side vector b - * \param x On input and initial solution, on output the computed solution. - * \param precond A preconditioner being able to efficiently solve for an - * approximation of Ax=b (regardless of b) - * \param iters On input the max number of iteration, on output the number of performed iterations. - * \param tol_error On input the tolerance error, on output an estimation of the relative error. - * \return false in the case of numerical issue, for example a break down of BiCGSTAB. - */ -template -bool bicgstab(const MatrixType& mat, const Rhs& rhs, Dest& x, - const Preconditioner& precond, Index& iters, - typename Dest::RealScalar& tol_error) -{ - using std::sqrt; - using std::abs; - typedef typename Dest::RealScalar RealScalar; - typedef typename Dest::Scalar Scalar; - typedef Matrix VectorType; - RealScalar tol = tol_error; - Index maxIters = iters; - - Index n = mat.cols(); - VectorType r = rhs - mat * x; - VectorType r0 = r; - - RealScalar r0_sqnorm = r0.squaredNorm(); - RealScalar rhs_sqnorm = rhs.squaredNorm(); - if(rhs_sqnorm == 0) - { - x.setZero(); - return true; - } - Scalar rho = 1; - Scalar alpha = 1; - Scalar w = 1; - - VectorType v = VectorType::Zero(n), p = VectorType::Zero(n); - VectorType y(n), z(n); - VectorType kt(n), ks(n); - - VectorType s(n), t(n); - - RealScalar tol2 = tol*tol*rhs_sqnorm; - RealScalar eps2 = NumTraits::epsilon()*NumTraits::epsilon(); - Index i = 0; - Index restarts = 0; - - while ( r.squaredNorm() > tol2 && iRealScalar(0)) - w = t.dot(s) / tmp; - else - w = Scalar(0); - x += alpha * y + w * z; - r = s - w * t; - ++i; - } - tol_error = sqrt(r.squaredNorm()/rhs_sqnorm); - iters = i; - return true; -} - -} - -template< typename _MatrixType, - typename _Preconditioner = DiagonalPreconditioner > -class BiCGSTAB; - -namespace internal { - -template< typename _MatrixType, typename _Preconditioner> -struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Preconditioner Preconditioner; -}; - -} - -/** \ingroup IterativeLinearSolvers_Module - * \brief A bi conjugate gradient stabilized solver for sparse square problems - * - * This class allows to solve for A.x = b sparse linear problems using a bi conjugate gradient - * stabilized algorithm. The vectors x and b can be either dense or sparse. - * - * \tparam _MatrixType the type of the sparse matrix A, can be a dense or a sparse matrix. - * \tparam _Preconditioner the type of the preconditioner. Default is DiagonalPreconditioner - * - * \implsparsesolverconcept - * - * The maximal number of iterations and tolerance value can be controlled via the setMaxIterations() - * and setTolerance() methods. The defaults are the size of the problem for the maximal number of iterations - * and NumTraits::epsilon() for the tolerance. - * - * The tolerance corresponds to the relative residual error: |Ax-b|/|b| - * - * \b Performance: when using sparse matrices, best performance is achied for a row-major sparse matrix format. - * Moreover, in this case multi-threading can be exploited if the user code is compiled with OpenMP enabled. - * See \ref TopicMultiThreading for details. - * - * This class can be used as the direct solver classes. Here is a typical usage example: - * \include BiCGSTAB_simple.cpp - * - * By default the iterations start with x=0 as an initial guess of the solution. - * One can control the start using the solveWithGuess() method. - * - * BiCGSTAB can also be used in a matrix-free context, see the following \link MatrixfreeSolverExample example \endlink. - * - * \sa class SimplicialCholesky, DiagonalPreconditioner, IdentityPreconditioner - */ -template< typename _MatrixType, typename _Preconditioner> -class BiCGSTAB : public IterativeSolverBase > -{ - typedef IterativeSolverBase Base; - using Base::matrix; - using Base::m_error; - using Base::m_iterations; - using Base::m_info; - using Base::m_isInitialized; -public: - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef _Preconditioner Preconditioner; - -public: - - /** Default constructor. */ - BiCGSTAB() : Base() {} - - /** Initialize the solver with matrix \a A for further \c Ax=b solving. - * - * This constructor is a shortcut for the default constructor followed - * by a call to compute(). - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - explicit BiCGSTAB(const EigenBase& A) : Base(A.derived()) {} - - ~BiCGSTAB() {} - - /** \internal */ - template - void _solve_with_guess_impl(const Rhs& b, Dest& x) const - { - bool failed = false; - for(Index j=0; j - void _solve_impl(const MatrixBase& b, Dest& x) const - { - x.resize(this->rows(),b.cols()); - x.setZero(); - _solve_with_guess_impl(b,x); - } - -protected: - -}; - -} // end namespace Eigen - -#endif // EIGEN_BICGSTAB_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h deleted file mode 100644 index f7ce471349a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/ConjugateGradient.h +++ /dev/null @@ -1,246 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2011-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_CONJUGATE_GRADIENT_H -#define EIGEN_CONJUGATE_GRADIENT_H - -namespace Eigen { - -namespace internal { - -/** \internal Low-level conjugate gradient algorithm - * \param mat The matrix A - * \param rhs The right hand side vector b - * \param x On input and initial solution, on output the computed solution. - * \param precond A preconditioner being able to efficiently solve for an - * approximation of Ax=b (regardless of b) - * \param iters On input the max number of iteration, on output the number of performed iterations. - * \param tol_error On input the tolerance error, on output an estimation of the relative error. - */ -template -EIGEN_DONT_INLINE -void conjugate_gradient(const MatrixType& mat, const Rhs& rhs, Dest& x, - const Preconditioner& precond, Index& iters, - typename Dest::RealScalar& tol_error) -{ - using std::sqrt; - using std::abs; - typedef typename Dest::RealScalar RealScalar; - typedef typename Dest::Scalar Scalar; - typedef Matrix VectorType; - - RealScalar tol = tol_error; - Index maxIters = iters; - - Index n = mat.cols(); - - VectorType residual = rhs - mat * x; //initial residual - - RealScalar rhsNorm2 = rhs.squaredNorm(); - if(rhsNorm2 == 0) - { - x.setZero(); - iters = 0; - tol_error = 0; - return; - } - const RealScalar considerAsZero = (std::numeric_limits::min)(); - RealScalar threshold = numext::maxi(tol*tol*rhsNorm2,considerAsZero); - RealScalar residualNorm2 = residual.squaredNorm(); - if (residualNorm2 < threshold) - { - iters = 0; - tol_error = sqrt(residualNorm2 / rhsNorm2); - return; - } - - VectorType p(n); - p = precond.solve(residual); // initial search direction - - VectorType z(n), tmp(n); - RealScalar absNew = numext::real(residual.dot(p)); // the square of the absolute value of r scaled by invM - Index i = 0; - while(i < maxIters) - { - tmp.noalias() = mat * p; // the bottleneck of the algorithm - - Scalar alpha = absNew / p.dot(tmp); // the amount we travel on dir - x += alpha * p; // update solution - residual -= alpha * tmp; // update residual - - residualNorm2 = residual.squaredNorm(); - if(residualNorm2 < threshold) - break; - - z = precond.solve(residual); // approximately solve for "A z = residual" - - RealScalar absOld = absNew; - absNew = numext::real(residual.dot(z)); // update the absolute value of r - RealScalar beta = absNew / absOld; // calculate the Gram-Schmidt value used to create the new search direction - p = z + beta * p; // update search direction - i++; - } - tol_error = sqrt(residualNorm2 / rhsNorm2); - iters = i; -} - -} - -template< typename _MatrixType, int _UpLo=Lower, - typename _Preconditioner = DiagonalPreconditioner > -class ConjugateGradient; - -namespace internal { - -template< typename _MatrixType, int _UpLo, typename _Preconditioner> -struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Preconditioner Preconditioner; -}; - -} - -/** \ingroup IterativeLinearSolvers_Module - * \brief A conjugate gradient solver for sparse (or dense) self-adjoint problems - * - * This class allows to solve for A.x = b linear problems using an iterative conjugate gradient algorithm. - * The matrix A must be selfadjoint. The matrix A and the vectors x and b can be either dense or sparse. - * - * \tparam _MatrixType the type of the matrix A, can be a dense or a sparse matrix. - * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower, - * \c Upper, or \c Lower|Upper in which the full matrix entries will be considered. - * Default is \c Lower, best performance is \c Lower|Upper. - * \tparam _Preconditioner the type of the preconditioner. Default is DiagonalPreconditioner - * - * \implsparsesolverconcept - * - * The maximal number of iterations and tolerance value can be controlled via the setMaxIterations() - * and setTolerance() methods. The defaults are the size of the problem for the maximal number of iterations - * and NumTraits::epsilon() for the tolerance. - * - * The tolerance corresponds to the relative residual error: |Ax-b|/|b| - * - * \b Performance: Even though the default value of \c _UpLo is \c Lower, significantly higher performance is - * achieved when using a complete matrix and \b Lower|Upper as the \a _UpLo template parameter. Moreover, in this - * case multi-threading can be exploited if the user code is compiled with OpenMP enabled. - * See \ref TopicMultiThreading for details. - * - * This class can be used as the direct solver classes. Here is a typical usage example: - \code - int n = 10000; - VectorXd x(n), b(n); - SparseMatrix A(n,n); - // fill A and b - ConjugateGradient, Lower|Upper> cg; - cg.compute(A); - x = cg.solve(b); - std::cout << "#iterations: " << cg.iterations() << std::endl; - std::cout << "estimated error: " << cg.error() << std::endl; - // update b, and solve again - x = cg.solve(b); - \endcode - * - * By default the iterations start with x=0 as an initial guess of the solution. - * One can control the start using the solveWithGuess() method. - * - * ConjugateGradient can also be used in a matrix-free context, see the following \link MatrixfreeSolverExample example \endlink. - * - * \sa class LeastSquaresConjugateGradient, class SimplicialCholesky, DiagonalPreconditioner, IdentityPreconditioner - */ -template< typename _MatrixType, int _UpLo, typename _Preconditioner> -class ConjugateGradient : public IterativeSolverBase > -{ - typedef IterativeSolverBase Base; - using Base::matrix; - using Base::m_error; - using Base::m_iterations; - using Base::m_info; - using Base::m_isInitialized; -public: - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef _Preconditioner Preconditioner; - - enum { - UpLo = _UpLo - }; - -public: - - /** Default constructor. */ - ConjugateGradient() : Base() {} - - /** Initialize the solver with matrix \a A for further \c Ax=b solving. - * - * This constructor is a shortcut for the default constructor followed - * by a call to compute(). - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - explicit ConjugateGradient(const EigenBase& A) : Base(A.derived()) {} - - ~ConjugateGradient() {} - - /** \internal */ - template - void _solve_with_guess_impl(const Rhs& b, Dest& x) const - { - typedef typename Base::MatrixWrapper MatrixWrapper; - typedef typename Base::ActualMatrixType ActualMatrixType; - enum { - TransposeInput = (!MatrixWrapper::MatrixFree) - && (UpLo==(Lower|Upper)) - && (!MatrixType::IsRowMajor) - && (!NumTraits::IsComplex) - }; - typedef typename internal::conditional, ActualMatrixType const&>::type RowMajorWrapper; - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(MatrixWrapper::MatrixFree,UpLo==(Lower|Upper)),MATRIX_FREE_CONJUGATE_GRADIENT_IS_COMPATIBLE_WITH_UPPER_UNION_LOWER_MODE_ONLY); - typedef typename internal::conditional::Type - >::type SelfAdjointWrapper; - m_iterations = Base::maxIterations(); - m_error = Base::m_tolerance; - - for(Index j=0; j - void _solve_impl(const MatrixBase& b, Dest& x) const - { - x.setZero(); - _solve_with_guess_impl(b.derived(),x); - } - -protected: - -}; - -} // end namespace Eigen - -#endif // EIGEN_CONJUGATE_GRADIENT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteCholesky.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteCholesky.h deleted file mode 100644 index e45c272b4c4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteCholesky.h +++ /dev/null @@ -1,400 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_INCOMPLETE_CHOlESKY_H -#define EIGEN_INCOMPLETE_CHOlESKY_H - -#include -#include - -namespace Eigen { -/** - * \brief Modified Incomplete Cholesky with dual threshold - * - * References : C-J. Lin and J. J. Moré, Incomplete Cholesky Factorizations with - * Limited memory, SIAM J. Sci. Comput. 21(1), pp. 24-45, 1999 - * - * \tparam Scalar the scalar type of the input matrices - * \tparam _UpLo The triangular part that will be used for the computations. It can be Lower - * or Upper. Default is Lower. - * \tparam _OrderingType The ordering method to use, either AMDOrdering<> or NaturalOrdering<>. Default is AMDOrdering, - * unless EIGEN_MPL2_ONLY is defined, in which case the default is NaturalOrdering. - * - * \implsparsesolverconcept - * - * It performs the following incomplete factorization: \f$ S P A P' S \approx L L' \f$ - * where L is a lower triangular factor, S is a diagonal scaling matrix, and P is a - * fill-in reducing permutation as computed by the ordering method. - * - * \b Shifting \b strategy: Let \f$ B = S P A P' S \f$ be the scaled matrix on which the factorization is carried out, - * and \f$ \beta \f$ be the minimum value of the diagonal. If \f$ \beta > 0 \f$ then, the factorization is directly performed - * on the matrix B. Otherwise, the factorization is performed on the shifted matrix \f$ B + (\sigma+|\beta| I \f$ where - * \f$ \sigma \f$ is the initial shift value as returned and set by setInitialShift() method. The default value is \f$ \sigma = 10^{-3} \f$. - * If the factorization fails, then the shift in doubled until it succeed or a maximum of ten attempts. If it still fails, as returned by - * the info() method, then you can either increase the initial shift, or better use another preconditioning technique. - * - */ -template -#else -NaturalOrdering -#endif -> -class IncompleteCholesky : public SparseSolverBase > -{ - protected: - typedef SparseSolverBase > Base; - using Base::m_isInitialized; - public: - typedef typename NumTraits::Real RealScalar; - typedef _OrderingType OrderingType; - typedef typename OrderingType::PermutationType PermutationType; - typedef typename PermutationType::StorageIndex StorageIndex; - typedef SparseMatrix FactorType; - typedef Matrix VectorSx; - typedef Matrix VectorRx; - typedef Matrix VectorIx; - typedef std::vector > VectorList; - enum { UpLo = _UpLo }; - enum { - ColsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic - }; - public: - - /** Default constructor leaving the object in a partly non-initialized stage. - * - * You must call compute() or the pair analyzePattern()/factorize() to make it valid. - * - * \sa IncompleteCholesky(const MatrixType&) - */ - IncompleteCholesky() : m_initialShift(1e-3),m_factorizationIsOk(false) {} - - /** Constructor computing the incomplete factorization for the given matrix \a matrix. - */ - template - IncompleteCholesky(const MatrixType& matrix) : m_initialShift(1e-3),m_factorizationIsOk(false) - { - compute(matrix); - } - - /** \returns number of rows of the factored matrix */ - Index rows() const { return m_L.rows(); } - - /** \returns number of columns of the factored matrix */ - Index cols() const { return m_L.cols(); } - - - /** \brief Reports whether previous computation was successful. - * - * It triggers an assertion if \c *this has not been initialized through the respective constructor, - * or a call to compute() or analyzePattern(). - * - * \returns \c Success if computation was successful, - * \c NumericalIssue if the matrix appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "IncompleteCholesky is not initialized."); - return m_info; - } - - /** \brief Set the initial shift parameter \f$ \sigma \f$. - */ - void setInitialShift(RealScalar shift) { m_initialShift = shift; } - - /** \brief Computes the fill reducing permutation vector using the sparsity pattern of \a mat - */ - template - void analyzePattern(const MatrixType& mat) - { - OrderingType ord; - PermutationType pinv; - ord(mat.template selfadjointView(), pinv); - if(pinv.size()>0) m_perm = pinv.inverse(); - else m_perm.resize(0); - m_L.resize(mat.rows(), mat.cols()); - m_analysisIsOk = true; - m_isInitialized = true; - m_info = Success; - } - - /** \brief Performs the numerical factorization of the input matrix \a mat - * - * The method analyzePattern() or compute() must have been called beforehand - * with a matrix having the same pattern. - * - * \sa compute(), analyzePattern() - */ - template - void factorize(const MatrixType& mat); - - /** Computes or re-computes the incomplete Cholesky factorization of the input matrix \a mat - * - * It is a shortcut for a sequential call to the analyzePattern() and factorize() methods. - * - * \sa analyzePattern(), factorize() - */ - template - void compute(const MatrixType& mat) - { - analyzePattern(mat); - factorize(mat); - } - - // internal - template - void _solve_impl(const Rhs& b, Dest& x) const - { - eigen_assert(m_factorizationIsOk && "factorize() should be called first"); - if (m_perm.rows() == b.rows()) x = m_perm * b; - else x = b; - x = m_scale.asDiagonal() * x; - x = m_L.template triangularView().solve(x); - x = m_L.adjoint().template triangularView().solve(x); - x = m_scale.asDiagonal() * x; - if (m_perm.rows() == b.rows()) - x = m_perm.inverse() * x; - } - - /** \returns the sparse lower triangular factor L */ - const FactorType& matrixL() const { eigen_assert("m_factorizationIsOk"); return m_L; } - - /** \returns a vector representing the scaling factor S */ - const VectorRx& scalingS() const { eigen_assert("m_factorizationIsOk"); return m_scale; } - - /** \returns the fill-in reducing permutation P (can be empty for a natural ordering) */ - const PermutationType& permutationP() const { eigen_assert("m_analysisIsOk"); return m_perm; } - - protected: - FactorType m_L; // The lower part stored in CSC - VectorRx m_scale; // The vector for scaling the matrix - RealScalar m_initialShift; // The initial shift parameter - bool m_analysisIsOk; - bool m_factorizationIsOk; - ComputationInfo m_info; - PermutationType m_perm; - - private: - inline void updateList(Ref colPtr, Ref rowIdx, Ref vals, const Index& col, const Index& jk, VectorIx& firstElt, VectorList& listCol); -}; - -// Based on the following paper: -// C-J. Lin and J. J. Moré, Incomplete Cholesky Factorizations with -// Limited memory, SIAM J. Sci. Comput. 21(1), pp. 24-45, 1999 -// http://ftp.mcs.anl.gov/pub/tech_reports/reports/P682.pdf -template -template -void IncompleteCholesky::factorize(const _MatrixType& mat) -{ - using std::sqrt; - eigen_assert(m_analysisIsOk && "analyzePattern() should be called first"); - - // Dropping strategy : Keep only the p largest elements per column, where p is the number of elements in the column of the original matrix. Other strategies will be added - - // Apply the fill-reducing permutation computed in analyzePattern() - if (m_perm.rows() == mat.rows() ) // To detect the null permutation - { - // The temporary is needed to make sure that the diagonal entry is properly sorted - FactorType tmp(mat.rows(), mat.cols()); - tmp = mat.template selfadjointView<_UpLo>().twistedBy(m_perm); - m_L.template selfadjointView() = tmp.template selfadjointView(); - } - else - { - m_L.template selfadjointView() = mat.template selfadjointView<_UpLo>(); - } - - Index n = m_L.cols(); - Index nnz = m_L.nonZeros(); - Map vals(m_L.valuePtr(), nnz); //values - Map rowIdx(m_L.innerIndexPtr(), nnz); //Row indices - Map colPtr( m_L.outerIndexPtr(), n+1); // Pointer to the beginning of each row - VectorIx firstElt(n-1); // for each j, points to the next entry in vals that will be used in the factorization - VectorList listCol(n); // listCol(j) is a linked list of columns to update column j - VectorSx col_vals(n); // Store a nonzero values in each column - VectorIx col_irow(n); // Row indices of nonzero elements in each column - VectorIx col_pattern(n); - col_pattern.fill(-1); - StorageIndex col_nnz; - - - // Computes the scaling factors - m_scale.resize(n); - m_scale.setZero(); - for (Index j = 0; j < n; j++) - for (Index k = colPtr[j]; k < colPtr[j+1]; k++) - { - m_scale(j) += numext::abs2(vals(k)); - if(rowIdx[k]!=j) - m_scale(rowIdx[k]) += numext::abs2(vals(k)); - } - - m_scale = m_scale.cwiseSqrt().cwiseSqrt(); - - for (Index j = 0; j < n; ++j) - if(m_scale(j)>(std::numeric_limits::min)()) - m_scale(j) = RealScalar(1)/m_scale(j); - else - m_scale(j) = 1; - - // TODO disable scaling if not needed, i.e., if it is roughly uniform? (this will make solve() faster) - - // Scale and compute the shift for the matrix - RealScalar mindiag = NumTraits::highest(); - for (Index j = 0; j < n; j++) - { - for (Index k = colPtr[j]; k < colPtr[j+1]; k++) - vals[k] *= (m_scale(j)*m_scale(rowIdx[k])); - eigen_internal_assert(rowIdx[colPtr[j]]==j && "IncompleteCholesky: only the lower triangular part must be stored"); - mindiag = numext::mini(numext::real(vals[colPtr[j]]), mindiag); - } - - FactorType L_save = m_L; - - RealScalar shift = 0; - if(mindiag <= RealScalar(0.)) - shift = m_initialShift - mindiag; - - m_info = NumericalIssue; - - // Try to perform the incomplete factorization using the current shift - int iter = 0; - do - { - // Apply the shift to the diagonal elements of the matrix - for (Index j = 0; j < n; j++) - vals[colPtr[j]] += shift; - - // jki version of the Cholesky factorization - Index j=0; - for (; j < n; ++j) - { - // Left-looking factorization of the j-th column - // First, load the j-th column into col_vals - Scalar diag = vals[colPtr[j]]; // It is assumed that only the lower part is stored - col_nnz = 0; - for (Index i = colPtr[j] + 1; i < colPtr[j+1]; i++) - { - StorageIndex l = rowIdx[i]; - col_vals(col_nnz) = vals[i]; - col_irow(col_nnz) = l; - col_pattern(l) = col_nnz; - col_nnz++; - } - { - typename std::list::iterator k; - // Browse all previous columns that will update column j - for(k = listCol[j].begin(); k != listCol[j].end(); k++) - { - Index jk = firstElt(*k); // First element to use in the column - eigen_internal_assert(rowIdx[jk]==j); - Scalar v_j_jk = numext::conj(vals[jk]); - - jk += 1; - for (Index i = jk; i < colPtr[*k+1]; i++) - { - StorageIndex l = rowIdx[i]; - if(col_pattern[l]<0) - { - col_vals(col_nnz) = vals[i] * v_j_jk; - col_irow[col_nnz] = l; - col_pattern(l) = col_nnz; - col_nnz++; - } - else - col_vals(col_pattern[l]) -= vals[i] * v_j_jk; - } - updateList(colPtr,rowIdx,vals, *k, jk, firstElt, listCol); - } - } - - // Scale the current column - if(numext::real(diag) <= 0) - { - if(++iter>=10) - return; - - // increase shift - shift = numext::maxi(m_initialShift,RealScalar(2)*shift); - // restore m_L, col_pattern, and listCol - vals = Map(L_save.valuePtr(), nnz); - rowIdx = Map(L_save.innerIndexPtr(), nnz); - colPtr = Map(L_save.outerIndexPtr(), n+1); - col_pattern.fill(-1); - for(Index i=0; i cvals = col_vals.head(col_nnz); - Ref cirow = col_irow.head(col_nnz); - internal::QuickSplit(cvals,cirow, p); - // Insert the largest p elements in the matrix - Index cpt = 0; - for (Index i = colPtr[j]+1; i < colPtr[j+1]; i++) - { - vals[i] = col_vals(cpt); - rowIdx[i] = col_irow(cpt); - // restore col_pattern: - col_pattern(col_irow(cpt)) = -1; - cpt++; - } - // Get the first smallest row index and put it after the diagonal element - Index jk = colPtr(j)+1; - updateList(colPtr,rowIdx,vals,j,jk,firstElt,listCol); - } - - if(j==n) - { - m_factorizationIsOk = true; - m_info = Success; - } - } while(m_info!=Success); -} - -template -inline void IncompleteCholesky::updateList(Ref colPtr, Ref rowIdx, Ref vals, const Index& col, const Index& jk, VectorIx& firstElt, VectorList& listCol) -{ - if (jk < colPtr(col+1) ) - { - Index p = colPtr(col+1) - jk; - Index minpos; - rowIdx.segment(jk,p).minCoeff(&minpos); - minpos += jk; - if (rowIdx(minpos) != rowIdx(jk)) - { - //Swap - std::swap(rowIdx(jk),rowIdx(minpos)); - std::swap(vals(jk),vals(minpos)); - } - firstElt(col) = internal::convert_index(jk); - listCol[rowIdx(jk)].push_back(internal::convert_index(col)); - } -} - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteLUT.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteLUT.h deleted file mode 100644 index 338e6f10a87..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IncompleteLUT.h +++ /dev/null @@ -1,462 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_INCOMPLETE_LUT_H -#define EIGEN_INCOMPLETE_LUT_H - - -namespace Eigen { - -namespace internal { - -/** \internal - * Compute a quick-sort split of a vector - * On output, the vector row is permuted such that its elements satisfy - * abs(row(i)) >= abs(row(ncut)) if incut - * \param row The vector of values - * \param ind The array of index for the elements in @p row - * \param ncut The number of largest elements to keep - **/ -template -Index QuickSplit(VectorV &row, VectorI &ind, Index ncut) -{ - typedef typename VectorV::RealScalar RealScalar; - using std::swap; - using std::abs; - Index mid; - Index n = row.size(); /* length of the vector */ - Index first, last ; - - ncut--; /* to fit the zero-based indices */ - first = 0; - last = n-1; - if (ncut < first || ncut > last ) return 0; - - do { - mid = first; - RealScalar abskey = abs(row(mid)); - for (Index j = first + 1; j <= last; j++) { - if ( abs(row(j)) > abskey) { - ++mid; - swap(row(mid), row(j)); - swap(ind(mid), ind(j)); - } - } - /* Interchange for the pivot element */ - swap(row(mid), row(first)); - swap(ind(mid), ind(first)); - - if (mid > ncut) last = mid - 1; - else if (mid < ncut ) first = mid + 1; - } while (mid != ncut ); - - return 0; /* mid is equal to ncut */ -} - -}// end namespace internal - -/** \ingroup IterativeLinearSolvers_Module - * \class IncompleteLUT - * \brief Incomplete LU factorization with dual-threshold strategy - * - * \implsparsesolverconcept - * - * During the numerical factorization, two dropping rules are used : - * 1) any element whose magnitude is less than some tolerance is dropped. - * This tolerance is obtained by multiplying the input tolerance @p droptol - * by the average magnitude of all the original elements in the current row. - * 2) After the elimination of the row, only the @p fill largest elements in - * the L part and the @p fill largest elements in the U part are kept - * (in addition to the diagonal element ). Note that @p fill is computed from - * the input parameter @p fillfactor which is used the ratio to control the fill_in - * relatively to the initial number of nonzero elements. - * - * The two extreme cases are when @p droptol=0 (to keep all the @p fill*2 largest elements) - * and when @p fill=n/2 with @p droptol being different to zero. - * - * References : Yousef Saad, ILUT: A dual threshold incomplete LU factorization, - * Numerical Linear Algebra with Applications, 1(4), pp 387-402, 1994. - * - * NOTE : The following implementation is derived from the ILUT implementation - * in the SPARSKIT package, Copyright (C) 2005, the Regents of the University of Minnesota - * released under the terms of the GNU LGPL: - * http://www-users.cs.umn.edu/~saad/software/SPARSKIT/README - * However, Yousef Saad gave us permission to relicense his ILUT code to MPL2. - * See the Eigen mailing list archive, thread: ILUT, date: July 8, 2012: - * http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2012/07/msg00064.html - * alternatively, on GMANE: - * http://comments.gmane.org/gmane.comp.lib.eigen/3302 - */ -template -class IncompleteLUT : public SparseSolverBase > -{ - protected: - typedef SparseSolverBase Base; - using Base::m_isInitialized; - public: - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - typedef typename NumTraits::Real RealScalar; - typedef Matrix Vector; - typedef Matrix VectorI; - typedef SparseMatrix FactorType; - - enum { - ColsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic - }; - - public: - - IncompleteLUT() - : m_droptol(NumTraits::dummy_precision()), m_fillfactor(10), - m_analysisIsOk(false), m_factorizationIsOk(false) - {} - - template - explicit IncompleteLUT(const MatrixType& mat, const RealScalar& droptol=NumTraits::dummy_precision(), int fillfactor = 10) - : m_droptol(droptol),m_fillfactor(fillfactor), - m_analysisIsOk(false),m_factorizationIsOk(false) - { - eigen_assert(fillfactor != 0); - compute(mat); - } - - Index rows() const { return m_lu.rows(); } - - Index cols() const { return m_lu.cols(); } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the matrix.appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "IncompleteLUT is not initialized."); - return m_info; - } - - template - void analyzePattern(const MatrixType& amat); - - template - void factorize(const MatrixType& amat); - - /** - * Compute an incomplete LU factorization with dual threshold on the matrix mat - * No pivoting is done in this version - * - **/ - template - IncompleteLUT& compute(const MatrixType& amat) - { - analyzePattern(amat); - factorize(amat); - return *this; - } - - void setDroptol(const RealScalar& droptol); - void setFillfactor(int fillfactor); - - template - void _solve_impl(const Rhs& b, Dest& x) const - { - x = m_Pinv * b; - x = m_lu.template triangularView().solve(x); - x = m_lu.template triangularView().solve(x); - x = m_P * x; - } - -protected: - - /** keeps off-diagonal entries; drops diagonal entries */ - struct keep_diag { - inline bool operator() (const Index& row, const Index& col, const Scalar&) const - { - return row!=col; - } - }; - -protected: - - FactorType m_lu; - RealScalar m_droptol; - int m_fillfactor; - bool m_analysisIsOk; - bool m_factorizationIsOk; - ComputationInfo m_info; - PermutationMatrix m_P; // Fill-reducing permutation - PermutationMatrix m_Pinv; // Inverse permutation -}; - -/** - * Set control parameter droptol - * \param droptol Drop any element whose magnitude is less than this tolerance - **/ -template -void IncompleteLUT::setDroptol(const RealScalar& droptol) -{ - this->m_droptol = droptol; -} - -/** - * Set control parameter fillfactor - * \param fillfactor This is used to compute the number @p fill_in of largest elements to keep on each row. - **/ -template -void IncompleteLUT::setFillfactor(int fillfactor) -{ - this->m_fillfactor = fillfactor; -} - -template -template -void IncompleteLUT::analyzePattern(const _MatrixType& amat) -{ - // Compute the Fill-reducing permutation - // Since ILUT does not perform any numerical pivoting, - // it is highly preferable to keep the diagonal through symmetric permutations. -#ifndef EIGEN_MPL2_ONLY - // To this end, let's symmetrize the pattern and perform AMD on it. - SparseMatrix mat1 = amat; - SparseMatrix mat2 = amat.transpose(); - // FIXME for a matrix with nearly symmetric pattern, mat2+mat1 is the appropriate choice. - // on the other hand for a really non-symmetric pattern, mat2*mat1 should be prefered... - SparseMatrix AtA = mat2 + mat1; - AMDOrdering ordering; - ordering(AtA,m_P); - m_Pinv = m_P.inverse(); // cache the inverse permutation -#else - // If AMD is not available, (MPL2-only), then let's use the slower COLAMD routine. - SparseMatrix mat1 = amat; - COLAMDOrdering ordering; - ordering(mat1,m_Pinv); - m_P = m_Pinv.inverse(); -#endif - - m_analysisIsOk = true; - m_factorizationIsOk = false; - m_isInitialized = true; -} - -template -template -void IncompleteLUT::factorize(const _MatrixType& amat) -{ - using std::sqrt; - using std::swap; - using std::abs; - using internal::convert_index; - - eigen_assert((amat.rows() == amat.cols()) && "The factorization should be done on a square matrix"); - Index n = amat.cols(); // Size of the matrix - m_lu.resize(n,n); - // Declare Working vectors and variables - Vector u(n) ; // real values of the row -- maximum size is n -- - VectorI ju(n); // column position of the values in u -- maximum size is n - VectorI jr(n); // Indicate the position of the nonzero elements in the vector u -- A zero location is indicated by -1 - - // Apply the fill-reducing permutation - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - SparseMatrix mat; - mat = amat.twistedBy(m_Pinv); - - // Initialization - jr.fill(-1); - ju.fill(0); - u.fill(0); - - // number of largest elements to keep in each row: - Index fill_in = (amat.nonZeros()*m_fillfactor)/n + 1; - if (fill_in > n) fill_in = n; - - // number of largest nonzero elements to keep in the L and the U part of the current row: - Index nnzL = fill_in/2; - Index nnzU = nnzL; - m_lu.reserve(n * (nnzL + nnzU + 1)); - - // global loop over the rows of the sparse matrix - for (Index ii = 0; ii < n; ii++) - { - // 1 - copy the lower and the upper part of the row i of mat in the working vector u - - Index sizeu = 1; // number of nonzero elements in the upper part of the current row - Index sizel = 0; // number of nonzero elements in the lower part of the current row - ju(ii) = convert_index(ii); - u(ii) = 0; - jr(ii) = convert_index(ii); - RealScalar rownorm = 0; - - typename FactorType::InnerIterator j_it(mat, ii); // Iterate through the current row ii - for (; j_it; ++j_it) - { - Index k = j_it.index(); - if (k < ii) - { - // copy the lower part - ju(sizel) = convert_index(k); - u(sizel) = j_it.value(); - jr(k) = convert_index(sizel); - ++sizel; - } - else if (k == ii) - { - u(ii) = j_it.value(); - } - else - { - // copy the upper part - Index jpos = ii + sizeu; - ju(jpos) = convert_index(k); - u(jpos) = j_it.value(); - jr(k) = convert_index(jpos); - ++sizeu; - } - rownorm += numext::abs2(j_it.value()); - } - - // 2 - detect possible zero row - if(rownorm==0) - { - m_info = NumericalIssue; - return; - } - // Take the 2-norm of the current row as a relative tolerance - rownorm = sqrt(rownorm); - - // 3 - eliminate the previous nonzero rows - Index jj = 0; - Index len = 0; - while (jj < sizel) - { - // In order to eliminate in the correct order, - // we must select first the smallest column index among ju(jj:sizel) - Index k; - Index minrow = ju.segment(jj,sizel-jj).minCoeff(&k); // k is relative to the segment - k += jj; - if (minrow != ju(jj)) - { - // swap the two locations - Index j = ju(jj); - swap(ju(jj), ju(k)); - jr(minrow) = convert_index(jj); - jr(j) = convert_index(k); - swap(u(jj), u(k)); - } - // Reset this location - jr(minrow) = -1; - - // Start elimination - typename FactorType::InnerIterator ki_it(m_lu, minrow); - while (ki_it && ki_it.index() < minrow) ++ki_it; - eigen_internal_assert(ki_it && ki_it.col()==minrow); - Scalar fact = u(jj) / ki_it.value(); - - // drop too small elements - if(abs(fact) <= m_droptol) - { - jj++; - continue; - } - - // linear combination of the current row ii and the row minrow - ++ki_it; - for (; ki_it; ++ki_it) - { - Scalar prod = fact * ki_it.value(); - Index j = ki_it.index(); - Index jpos = jr(j); - if (jpos == -1) // fill-in element - { - Index newpos; - if (j >= ii) // dealing with the upper part - { - newpos = ii + sizeu; - sizeu++; - eigen_internal_assert(sizeu<=n); - } - else // dealing with the lower part - { - newpos = sizel; - sizel++; - eigen_internal_assert(sizel<=ii); - } - ju(newpos) = convert_index(j); - u(newpos) = -prod; - jr(j) = convert_index(newpos); - } - else - u(jpos) -= prod; - } - // store the pivot element - u(len) = fact; - ju(len) = convert_index(minrow); - ++len; - - jj++; - } // end of the elimination on the row ii - - // reset the upper part of the pointer jr to zero - for(Index k = 0; k m_droptol * rownorm ) - { - ++len; - u(ii + len) = u(ii + k); - ju(ii + len) = ju(ii + k); - } - } - sizeu = len + 1; // +1 to take into account the diagonal element - len = (std::min)(sizeu, nnzU); - typename Vector::SegmentReturnType uu(u.segment(ii+1, sizeu-1)); - typename VectorI::SegmentReturnType juu(ju.segment(ii+1, sizeu-1)); - internal::QuickSplit(uu, juu, len); - - // store the largest elements of the U part - for(Index k = ii + 1; k < ii + len; k++) - m_lu.insertBackByOuterInnerUnordered(ii,ju(k)) = u(k); - } - m_lu.finalize(); - m_lu.makeCompressed(); - - m_factorizationIsOk = true; - m_info = Success; -} - -} // end namespace Eigen - -#endif // EIGEN_INCOMPLETE_LUT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IterativeSolverBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IterativeSolverBase.h deleted file mode 100644 index 7c2326eb7fd..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/IterativeSolverBase.h +++ /dev/null @@ -1,394 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2011-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ITERATIVE_SOLVER_BASE_H -#define EIGEN_ITERATIVE_SOLVER_BASE_H - -namespace Eigen { - -namespace internal { - -template -struct is_ref_compatible_impl -{ -private: - template - struct any_conversion - { - template any_conversion(const volatile T&); - template any_conversion(T&); - }; - struct yes {int a[1];}; - struct no {int a[2];}; - - template - static yes test(const Ref&, int); - template - static no test(any_conversion, ...); - -public: - static MatrixType ms_from; - enum { value = sizeof(test(ms_from, 0))==sizeof(yes) }; -}; - -template -struct is_ref_compatible -{ - enum { value = is_ref_compatible_impl::type>::value }; -}; - -template::value> -class generic_matrix_wrapper; - -// We have an explicit matrix at hand, compatible with Ref<> -template -class generic_matrix_wrapper -{ -public: - typedef Ref ActualMatrixType; - template struct ConstSelfAdjointViewReturnType { - typedef typename ActualMatrixType::template ConstSelfAdjointViewReturnType::Type Type; - }; - - enum { - MatrixFree = false - }; - - generic_matrix_wrapper() - : m_dummy(0,0), m_matrix(m_dummy) - {} - - template - generic_matrix_wrapper(const InputType &mat) - : m_matrix(mat) - {} - - const ActualMatrixType& matrix() const - { - return m_matrix; - } - - template - void grab(const EigenBase &mat) - { - m_matrix.~Ref(); - ::new (&m_matrix) Ref(mat.derived()); - } - - void grab(const Ref &mat) - { - if(&(mat.derived()) != &m_matrix) - { - m_matrix.~Ref(); - ::new (&m_matrix) Ref(mat); - } - } - -protected: - MatrixType m_dummy; // used to default initialize the Ref<> object - ActualMatrixType m_matrix; -}; - -// MatrixType is not compatible with Ref<> -> matrix-free wrapper -template -class generic_matrix_wrapper -{ -public: - typedef MatrixType ActualMatrixType; - template struct ConstSelfAdjointViewReturnType - { - typedef ActualMatrixType Type; - }; - - enum { - MatrixFree = true - }; - - generic_matrix_wrapper() - : mp_matrix(0) - {} - - generic_matrix_wrapper(const MatrixType &mat) - : mp_matrix(&mat) - {} - - const ActualMatrixType& matrix() const - { - return *mp_matrix; - } - - void grab(const MatrixType &mat) - { - mp_matrix = &mat; - } - -protected: - const ActualMatrixType *mp_matrix; -}; - -} - -/** \ingroup IterativeLinearSolvers_Module - * \brief Base class for linear iterative solvers - * - * \sa class SimplicialCholesky, DiagonalPreconditioner, IdentityPreconditioner - */ -template< typename Derived> -class IterativeSolverBase : public SparseSolverBase -{ -protected: - typedef SparseSolverBase Base; - using Base::m_isInitialized; - -public: - typedef typename internal::traits::MatrixType MatrixType; - typedef typename internal::traits::Preconditioner Preconditioner; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename MatrixType::RealScalar RealScalar; - - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - -public: - - using Base::derived; - - /** Default constructor. */ - IterativeSolverBase() - { - init(); - } - - /** Initialize the solver with matrix \a A for further \c Ax=b solving. - * - * This constructor is a shortcut for the default constructor followed - * by a call to compute(). - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - explicit IterativeSolverBase(const EigenBase& A) - : m_matrixWrapper(A.derived()) - { - init(); - compute(matrix()); - } - - ~IterativeSolverBase() {} - - /** Initializes the iterative solver for the sparsity pattern of the matrix \a A for further solving \c Ax=b problems. - * - * Currently, this function mostly calls analyzePattern on the preconditioner. In the future - * we might, for instance, implement column reordering for faster matrix vector products. - */ - template - Derived& analyzePattern(const EigenBase& A) - { - grab(A.derived()); - m_preconditioner.analyzePattern(matrix()); - m_isInitialized = true; - m_analysisIsOk = true; - m_info = m_preconditioner.info(); - return derived(); - } - - /** Initializes the iterative solver with the numerical values of the matrix \a A for further solving \c Ax=b problems. - * - * Currently, this function mostly calls factorize on the preconditioner. - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - Derived& factorize(const EigenBase& A) - { - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - grab(A.derived()); - m_preconditioner.factorize(matrix()); - m_factorizationIsOk = true; - m_info = m_preconditioner.info(); - return derived(); - } - - /** Initializes the iterative solver with the matrix \a A for further solving \c Ax=b problems. - * - * Currently, this function mostly initializes/computes the preconditioner. In the future - * we might, for instance, implement column reordering for faster matrix vector products. - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - Derived& compute(const EigenBase& A) - { - grab(A.derived()); - m_preconditioner.compute(matrix()); - m_isInitialized = true; - m_analysisIsOk = true; - m_factorizationIsOk = true; - m_info = m_preconditioner.info(); - return derived(); - } - - /** \internal */ - Index rows() const { return matrix().rows(); } - - /** \internal */ - Index cols() const { return matrix().cols(); } - - /** \returns the tolerance threshold used by the stopping criteria. - * \sa setTolerance() - */ - RealScalar tolerance() const { return m_tolerance; } - - /** Sets the tolerance threshold used by the stopping criteria. - * - * This value is used as an upper bound to the relative residual error: |Ax-b|/|b|. - * The default value is the machine precision given by NumTraits::epsilon() - */ - Derived& setTolerance(const RealScalar& tolerance) - { - m_tolerance = tolerance; - return derived(); - } - - /** \returns a read-write reference to the preconditioner for custom configuration. */ - Preconditioner& preconditioner() { return m_preconditioner; } - - /** \returns a read-only reference to the preconditioner. */ - const Preconditioner& preconditioner() const { return m_preconditioner; } - - /** \returns the max number of iterations. - * It is either the value setted by setMaxIterations or, by default, - * twice the number of columns of the matrix. - */ - Index maxIterations() const - { - return (m_maxIterations<0) ? 2*matrix().cols() : m_maxIterations; - } - - /** Sets the max number of iterations. - * Default is twice the number of columns of the matrix. - */ - Derived& setMaxIterations(Index maxIters) - { - m_maxIterations = maxIters; - return derived(); - } - - /** \returns the number of iterations performed during the last solve */ - Index iterations() const - { - eigen_assert(m_isInitialized && "ConjugateGradient is not initialized."); - return m_iterations; - } - - /** \returns the tolerance error reached during the last solve. - * It is a close approximation of the true relative residual error |Ax-b|/|b|. - */ - RealScalar error() const - { - eigen_assert(m_isInitialized && "ConjugateGradient is not initialized."); - return m_error; - } - - /** \returns the solution x of \f$ A x = b \f$ using the current decomposition of A - * and \a x0 as an initial solution. - * - * \sa solve(), compute() - */ - template - inline const SolveWithGuess - solveWithGuess(const MatrixBase& b, const Guess& x0) const - { - eigen_assert(m_isInitialized && "Solver is not initialized."); - eigen_assert(derived().rows()==b.rows() && "solve(): invalid number of rows of the right hand side matrix b"); - return SolveWithGuess(derived(), b.derived(), x0); - } - - /** \returns Success if the iterations converged, and NoConvergence otherwise. */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "IterativeSolverBase is not initialized."); - return m_info; - } - - /** \internal */ - template - void _solve_impl(const Rhs& b, SparseMatrixBase &aDest) const - { - eigen_assert(rows()==b.rows()); - - Index rhsCols = b.cols(); - Index size = b.rows(); - DestDerived& dest(aDest.derived()); - typedef typename DestDerived::Scalar DestScalar; - Eigen::Matrix tb(size); - Eigen::Matrix tx(cols()); - // We do not directly fill dest because sparse expressions have to be free of aliasing issue. - // For non square least-square problems, b and dest might not have the same size whereas they might alias each-other. - typename DestDerived::PlainObject tmp(cols(),rhsCols); - for(Index k=0; k::epsilon(); - } - - typedef internal::generic_matrix_wrapper MatrixWrapper; - typedef typename MatrixWrapper::ActualMatrixType ActualMatrixType; - - const ActualMatrixType& matrix() const - { - return m_matrixWrapper.matrix(); - } - - template - void grab(const InputType &A) - { - m_matrixWrapper.grab(A); - } - - MatrixWrapper m_matrixWrapper; - Preconditioner m_preconditioner; - - Index m_maxIterations; - RealScalar m_tolerance; - - mutable RealScalar m_error; - mutable Index m_iterations; - mutable ComputationInfo m_info; - mutable bool m_analysisIsOk, m_factorizationIsOk; -}; - -} // end namespace Eigen - -#endif // EIGEN_ITERATIVE_SOLVER_BASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/LeastSquareConjugateGradient.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/LeastSquareConjugateGradient.h deleted file mode 100644 index 0aea0e099d2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/LeastSquareConjugateGradient.h +++ /dev/null @@ -1,216 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_LEAST_SQUARE_CONJUGATE_GRADIENT_H -#define EIGEN_LEAST_SQUARE_CONJUGATE_GRADIENT_H - -namespace Eigen { - -namespace internal { - -/** \internal Low-level conjugate gradient algorithm for least-square problems - * \param mat The matrix A - * \param rhs The right hand side vector b - * \param x On input and initial solution, on output the computed solution. - * \param precond A preconditioner being able to efficiently solve for an - * approximation of A'Ax=b (regardless of b) - * \param iters On input the max number of iteration, on output the number of performed iterations. - * \param tol_error On input the tolerance error, on output an estimation of the relative error. - */ -template -EIGEN_DONT_INLINE -void least_square_conjugate_gradient(const MatrixType& mat, const Rhs& rhs, Dest& x, - const Preconditioner& precond, Index& iters, - typename Dest::RealScalar& tol_error) -{ - using std::sqrt; - using std::abs; - typedef typename Dest::RealScalar RealScalar; - typedef typename Dest::Scalar Scalar; - typedef Matrix VectorType; - - RealScalar tol = tol_error; - Index maxIters = iters; - - Index m = mat.rows(), n = mat.cols(); - - VectorType residual = rhs - mat * x; - VectorType normal_residual = mat.adjoint() * residual; - - RealScalar rhsNorm2 = (mat.adjoint()*rhs).squaredNorm(); - if(rhsNorm2 == 0) - { - x.setZero(); - iters = 0; - tol_error = 0; - return; - } - RealScalar threshold = tol*tol*rhsNorm2; - RealScalar residualNorm2 = normal_residual.squaredNorm(); - if (residualNorm2 < threshold) - { - iters = 0; - tol_error = sqrt(residualNorm2 / rhsNorm2); - return; - } - - VectorType p(n); - p = precond.solve(normal_residual); // initial search direction - - VectorType z(n), tmp(m); - RealScalar absNew = numext::real(normal_residual.dot(p)); // the square of the absolute value of r scaled by invM - Index i = 0; - while(i < maxIters) - { - tmp.noalias() = mat * p; - - Scalar alpha = absNew / tmp.squaredNorm(); // the amount we travel on dir - x += alpha * p; // update solution - residual -= alpha * tmp; // update residual - normal_residual = mat.adjoint() * residual; // update residual of the normal equation - - residualNorm2 = normal_residual.squaredNorm(); - if(residualNorm2 < threshold) - break; - - z = precond.solve(normal_residual); // approximately solve for "A'A z = normal_residual" - - RealScalar absOld = absNew; - absNew = numext::real(normal_residual.dot(z)); // update the absolute value of r - RealScalar beta = absNew / absOld; // calculate the Gram-Schmidt value used to create the new search direction - p = z + beta * p; // update search direction - i++; - } - tol_error = sqrt(residualNorm2 / rhsNorm2); - iters = i; -} - -} - -template< typename _MatrixType, - typename _Preconditioner = LeastSquareDiagonalPreconditioner > -class LeastSquaresConjugateGradient; - -namespace internal { - -template< typename _MatrixType, typename _Preconditioner> -struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Preconditioner Preconditioner; -}; - -} - -/** \ingroup IterativeLinearSolvers_Module - * \brief A conjugate gradient solver for sparse (or dense) least-square problems - * - * This class allows to solve for A x = b linear problems using an iterative conjugate gradient algorithm. - * The matrix A can be non symmetric and rectangular, but the matrix A' A should be positive-definite to guaranty stability. - * Otherwise, the SparseLU or SparseQR classes might be preferable. - * The matrix A and the vectors x and b can be either dense or sparse. - * - * \tparam _MatrixType the type of the matrix A, can be a dense or a sparse matrix. - * \tparam _Preconditioner the type of the preconditioner. Default is LeastSquareDiagonalPreconditioner - * - * \implsparsesolverconcept - * - * The maximal number of iterations and tolerance value can be controlled via the setMaxIterations() - * and setTolerance() methods. The defaults are the size of the problem for the maximal number of iterations - * and NumTraits::epsilon() for the tolerance. - * - * This class can be used as the direct solver classes. Here is a typical usage example: - \code - int m=1000000, n = 10000; - VectorXd x(n), b(m); - SparseMatrix A(m,n); - // fill A and b - LeastSquaresConjugateGradient > lscg; - lscg.compute(A); - x = lscg.solve(b); - std::cout << "#iterations: " << lscg.iterations() << std::endl; - std::cout << "estimated error: " << lscg.error() << std::endl; - // update b, and solve again - x = lscg.solve(b); - \endcode - * - * By default the iterations start with x=0 as an initial guess of the solution. - * One can control the start using the solveWithGuess() method. - * - * \sa class ConjugateGradient, SparseLU, SparseQR - */ -template< typename _MatrixType, typename _Preconditioner> -class LeastSquaresConjugateGradient : public IterativeSolverBase > -{ - typedef IterativeSolverBase Base; - using Base::matrix; - using Base::m_error; - using Base::m_iterations; - using Base::m_info; - using Base::m_isInitialized; -public: - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef _Preconditioner Preconditioner; - -public: - - /** Default constructor. */ - LeastSquaresConjugateGradient() : Base() {} - - /** Initialize the solver with matrix \a A for further \c Ax=b solving. - * - * This constructor is a shortcut for the default constructor followed - * by a call to compute(). - * - * \warning this class stores a reference to the matrix A as well as some - * precomputed values that depend on it. Therefore, if \a A is changed - * this class becomes invalid. Call compute() to update it with the new - * matrix A, or modify a copy of A. - */ - template - explicit LeastSquaresConjugateGradient(const EigenBase& A) : Base(A.derived()) {} - - ~LeastSquaresConjugateGradient() {} - - /** \internal */ - template - void _solve_with_guess_impl(const Rhs& b, Dest& x) const - { - m_iterations = Base::maxIterations(); - m_error = Base::m_tolerance; - - for(Index j=0; j - void _solve_impl(const MatrixBase& b, Dest& x) const - { - x.setZero(); - _solve_with_guess_impl(b.derived(),x); - } - -}; - -} // end namespace Eigen - -#endif // EIGEN_LEAST_SQUARE_CONJUGATE_GRADIENT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/SolveWithGuess.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/SolveWithGuess.h deleted file mode 100644 index 0ace4517717..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/IterativeLinearSolvers/SolveWithGuess.h +++ /dev/null @@ -1,115 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVEWITHGUESS_H -#define EIGEN_SOLVEWITHGUESS_H - -namespace Eigen { - -template class SolveWithGuess; - -/** \class SolveWithGuess - * \ingroup IterativeLinearSolvers_Module - * - * \brief Pseudo expression representing a solving operation - * - * \tparam Decomposition the type of the matrix or decomposion object - * \tparam Rhstype the type of the right-hand side - * - * This class represents an expression of A.solve(B) - * and most of the time this is the only way it is used. - * - */ -namespace internal { - - -template -struct traits > - : traits > -{}; - -} - - -template -class SolveWithGuess : public internal::generic_xpr_base, MatrixXpr, typename internal::traits::StorageKind>::type -{ -public: - typedef typename internal::traits::Scalar Scalar; - typedef typename internal::traits::PlainObject PlainObject; - typedef typename internal::generic_xpr_base, MatrixXpr, typename internal::traits::StorageKind>::type Base; - typedef typename internal::ref_selector::type Nested; - - SolveWithGuess(const Decomposition &dec, const RhsType &rhs, const GuessType &guess) - : m_dec(dec), m_rhs(rhs), m_guess(guess) - {} - - EIGEN_DEVICE_FUNC Index rows() const { return m_dec.cols(); } - EIGEN_DEVICE_FUNC Index cols() const { return m_rhs.cols(); } - - EIGEN_DEVICE_FUNC const Decomposition& dec() const { return m_dec; } - EIGEN_DEVICE_FUNC const RhsType& rhs() const { return m_rhs; } - EIGEN_DEVICE_FUNC const GuessType& guess() const { return m_guess; } - -protected: - const Decomposition &m_dec; - const RhsType &m_rhs; - const GuessType &m_guess; - -private: - Scalar coeff(Index row, Index col) const; - Scalar coeff(Index i) const; -}; - -namespace internal { - -// Evaluator of SolveWithGuess -> eval into a temporary -template -struct evaluator > - : public evaluator::PlainObject> -{ - typedef SolveWithGuess SolveType; - typedef typename SolveType::PlainObject PlainObject; - typedef evaluator Base; - - evaluator(const SolveType& solve) - : m_result(solve.rows(), solve.cols()) - { - ::new (static_cast(this)) Base(m_result); - m_result = solve.guess(); - solve.dec()._solve_with_guess_impl(solve.rhs(), m_result); - } - -protected: - PlainObject m_result; -}; - -// Specialization for "dst = dec.solveWithGuess(rhs)" -// NOTE we need to specialize it for Dense2Dense to avoid ambiguous specialization error and a Sparse2Sparse specialization must exist somewhere -template -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef SolveWithGuess SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - dst = src.guess(); - src.dec()._solve_with_guess_impl(src.rhs(), dst/*, src.guess()*/); - } -}; - -} // end namepsace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVEWITHGUESS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h deleted file mode 100644 index 1998c632274..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h +++ /dev/null @@ -1,462 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_JACOBI_H -#define EIGEN_JACOBI_H - -namespace Eigen { - -/** \ingroup Jacobi_Module - * \jacobi_module - * \class JacobiRotation - * \brief Rotation given by a cosine-sine pair. - * - * This class represents a Jacobi or Givens rotation. - * This is a 2D rotation in the plane \c J of angle \f$ \theta \f$ defined by - * its cosine \c c and sine \c s as follow: - * \f$ J = \left ( \begin{array}{cc} c & \overline s \\ -s & \overline c \end{array} \right ) \f$ - * - * You can apply the respective counter-clockwise rotation to a column vector \c v by - * applying its adjoint on the left: \f$ v = J^* v \f$ that translates to the following Eigen code: - * \code - * v.applyOnTheLeft(J.adjoint()); - * \endcode - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template class JacobiRotation -{ - public: - typedef typename NumTraits::Real RealScalar; - - /** Default constructor without any initialization. */ - JacobiRotation() {} - - /** Construct a planar rotation from a cosine-sine pair (\a c, \c s). */ - JacobiRotation(const Scalar& c, const Scalar& s) : m_c(c), m_s(s) {} - - Scalar& c() { return m_c; } - Scalar c() const { return m_c; } - Scalar& s() { return m_s; } - Scalar s() const { return m_s; } - - /** Concatenates two planar rotation */ - JacobiRotation operator*(const JacobiRotation& other) - { - using numext::conj; - return JacobiRotation(m_c * other.m_c - conj(m_s) * other.m_s, - conj(m_c * conj(other.m_s) + conj(m_s) * conj(other.m_c))); - } - - /** Returns the transposed transformation */ - JacobiRotation transpose() const { using numext::conj; return JacobiRotation(m_c, -conj(m_s)); } - - /** Returns the adjoint transformation */ - JacobiRotation adjoint() const { using numext::conj; return JacobiRotation(conj(m_c), -m_s); } - - template - bool makeJacobi(const MatrixBase&, Index p, Index q); - bool makeJacobi(const RealScalar& x, const Scalar& y, const RealScalar& z); - - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r=0); - - protected: - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::true_type); - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::false_type); - - Scalar m_c, m_s; -}; - -/** Makes \c *this as a Jacobi rotation \a J such that applying \a J on both the right and left sides of the selfadjoint 2x2 matrix - * \f$ B = \left ( \begin{array}{cc} x & y \\ \overline y & z \end{array} \right )\f$ yields a diagonal matrix \f$ A = J^* B J \f$ - * - * \sa MatrixBase::makeJacobi(const MatrixBase&, Index, Index), MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -bool JacobiRotation::makeJacobi(const RealScalar& x, const Scalar& y, const RealScalar& z) -{ - using std::sqrt; - using std::abs; - RealScalar deno = RealScalar(2)*abs(y); - if(deno < (std::numeric_limits::min)()) - { - m_c = Scalar(1); - m_s = Scalar(0); - return false; - } - else - { - RealScalar tau = (x-z)/deno; - RealScalar w = sqrt(numext::abs2(tau) + RealScalar(1)); - RealScalar t; - if(tau>RealScalar(0)) - { - t = RealScalar(1) / (tau + w); - } - else - { - t = RealScalar(1) / (tau - w); - } - RealScalar sign_t = t > RealScalar(0) ? RealScalar(1) : RealScalar(-1); - RealScalar n = RealScalar(1) / sqrt(numext::abs2(t)+RealScalar(1)); - m_s = - sign_t * (numext::conj(y) / abs(y)) * abs(t) * n; - m_c = n; - return true; - } -} - -/** Makes \c *this as a Jacobi rotation \c J such that applying \a J on both the right and left sides of the 2x2 selfadjoint matrix - * \f$ B = \left ( \begin{array}{cc} \text{this}_{pp} & \text{this}_{pq} \\ (\text{this}_{pq})^* & \text{this}_{qq} \end{array} \right )\f$ yields - * a diagonal matrix \f$ A = J^* B J \f$ - * - * Example: \include Jacobi_makeJacobi.cpp - * Output: \verbinclude Jacobi_makeJacobi.out - * - * \sa JacobiRotation::makeJacobi(RealScalar, Scalar, RealScalar), MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -template -inline bool JacobiRotation::makeJacobi(const MatrixBase& m, Index p, Index q) -{ - return makeJacobi(numext::real(m.coeff(p,p)), m.coeff(p,q), numext::real(m.coeff(q,q))); -} - -/** Makes \c *this as a Givens rotation \c G such that applying \f$ G^* \f$ to the left of the vector - * \f$ V = \left ( \begin{array}{c} p \\ q \end{array} \right )\f$ yields: - * \f$ G^* V = \left ( \begin{array}{c} r \\ 0 \end{array} \right )\f$. - * - * The value of \a r is returned if \a r is not null (the default is null). - * Also note that G is built such that the cosine is always real. - * - * Example: \include Jacobi_makeGivens.cpp - * Output: \verbinclude Jacobi_makeGivens.out - * - * This function implements the continuous Givens rotation generation algorithm - * found in Anderson (2000), Discontinuous Plane Rotations and the Symmetric Eigenvalue Problem. - * LAPACK Working Note 150, University of Tennessee, UT-CS-00-454, December 4, 2000. - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r) -{ - makeGivens(p, q, r, typename internal::conditional::IsComplex, internal::true_type, internal::false_type>::type()); -} - - -// specialization for complexes -template -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::true_type) -{ - using std::sqrt; - using std::abs; - using numext::conj; - - if(q==Scalar(0)) - { - m_c = numext::real(p)<0 ? Scalar(-1) : Scalar(1); - m_s = 0; - if(r) *r = m_c * p; - } - else if(p==Scalar(0)) - { - m_c = 0; - m_s = -q/abs(q); - if(r) *r = abs(q); - } - else - { - RealScalar p1 = numext::norm1(p); - RealScalar q1 = numext::norm1(q); - if(p1>=q1) - { - Scalar ps = p / p1; - RealScalar p2 = numext::abs2(ps); - Scalar qs = q / p1; - RealScalar q2 = numext::abs2(qs); - - RealScalar u = sqrt(RealScalar(1) + q2/p2); - if(numext::real(p) -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::false_type) -{ - using std::sqrt; - using std::abs; - if(q==Scalar(0)) - { - m_c = p abs(q)) - { - Scalar t = q/p; - Scalar u = sqrt(Scalar(1) + numext::abs2(t)); - if(p -void apply_rotation_in_the_plane(DenseBase& xpr_x, DenseBase& xpr_y, const JacobiRotation& j); -} - -/** \jacobi_module - * Applies the rotation in the plane \a j to the rows \a p and \a q of \c *this, i.e., it computes B = J * B, - * with \f$ B = \left ( \begin{array}{cc} \text{*this.row}(p) \\ \text{*this.row}(q) \end{array} \right ) \f$. - * - * \sa class JacobiRotation, MatrixBase::applyOnTheRight(), internal::apply_rotation_in_the_plane() - */ -template -template -inline void MatrixBase::applyOnTheLeft(Index p, Index q, const JacobiRotation& j) -{ - RowXpr x(this->row(p)); - RowXpr y(this->row(q)); - internal::apply_rotation_in_the_plane(x, y, j); -} - -/** \ingroup Jacobi_Module - * Applies the rotation in the plane \a j to the columns \a p and \a q of \c *this, i.e., it computes B = B * J - * with \f$ B = \left ( \begin{array}{cc} \text{*this.col}(p) & \text{*this.col}(q) \end{array} \right ) \f$. - * - * \sa class JacobiRotation, MatrixBase::applyOnTheLeft(), internal::apply_rotation_in_the_plane() - */ -template -template -inline void MatrixBase::applyOnTheRight(Index p, Index q, const JacobiRotation& j) -{ - ColXpr x(this->col(p)); - ColXpr y(this->col(q)); - internal::apply_rotation_in_the_plane(x, y, j.transpose()); -} - -namespace internal { - -template -struct apply_rotation_in_the_plane_selector -{ - static inline void run(Scalar *x, Index incrx, Scalar *y, Index incry, Index size, OtherScalar c, OtherScalar s) - { - for(Index i=0; i -struct apply_rotation_in_the_plane_selector -{ - static inline void run(Scalar *x, Index incrx, Scalar *y, Index incry, Index size, OtherScalar c, OtherScalar s) - { - enum { - PacketSize = packet_traits::size, - OtherPacketSize = packet_traits::size - }; - typedef typename packet_traits::type Packet; - typedef typename packet_traits::type OtherPacket; - - /*** dynamic-size vectorized paths ***/ - if(SizeAtCompileTime == Dynamic && ((incrx==1 && incry==1) || PacketSize == 1)) - { - // both vectors are sequentially stored in memory => vectorization - enum { Peeling = 2 }; - - Index alignedStart = internal::first_default_aligned(y, size); - Index alignedEnd = alignedStart + ((size-alignedStart)/PacketSize)*PacketSize; - - const OtherPacket pc = pset1(c); - const OtherPacket ps = pset1(s); - conj_helper::IsComplex,false> pcj; - conj_helper pm; - - for(Index i=0; i(px); - Packet yi = pload(py); - pstore(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore(py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - px += PacketSize; - py += PacketSize; - } - } - else - { - Index peelingEnd = alignedStart + ((size-alignedStart)/(Peeling*PacketSize))*(Peeling*PacketSize); - for(Index i=alignedStart; i(px); - Packet xi1 = ploadu(px+PacketSize); - Packet yi = pload (py); - Packet yi1 = pload (py+PacketSize); - pstoreu(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstoreu(px+PacketSize, padd(pm.pmul(pc,xi1),pcj.pmul(ps,yi1))); - pstore (py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - pstore (py+PacketSize, psub(pcj.pmul(pc,yi1),pm.pmul(ps,xi1))); - px += Peeling*PacketSize; - py += Peeling*PacketSize; - } - if(alignedEnd!=peelingEnd) - { - Packet xi = ploadu(x+peelingEnd); - Packet yi = pload (y+peelingEnd); - pstoreu(x+peelingEnd, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore (y+peelingEnd, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - } - } - - for(Index i=alignedEnd; i0) // FIXME should be compared to the required alignment - { - const OtherPacket pc = pset1(c); - const OtherPacket ps = pset1(s); - conj_helper::IsComplex,false> pcj; - conj_helper pm; - Scalar* EIGEN_RESTRICT px = x; - Scalar* EIGEN_RESTRICT py = y; - for(Index i=0; i(px); - Packet yi = pload(py); - pstore(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore(py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - px += PacketSize; - py += PacketSize; - } - } - - /*** non-vectorized path ***/ - else - { - apply_rotation_in_the_plane_selector::run(x,incrx,y,incry,size,c,s); - } - } -}; - -template -void /*EIGEN_DONT_INLINE*/ apply_rotation_in_the_plane(DenseBase& xpr_x, DenseBase& xpr_y, const JacobiRotation& j) -{ - typedef typename VectorX::Scalar Scalar; - const bool Vectorizable = (VectorX::Flags & VectorY::Flags & PacketAccessBit) - && (int(packet_traits::size) == int(packet_traits::size)); - - eigen_assert(xpr_x.size() == xpr_y.size()); - Index size = xpr_x.size(); - Index incrx = xpr_x.derived().innerStride(); - Index incry = xpr_y.derived().innerStride(); - - Scalar* EIGEN_RESTRICT x = &xpr_x.derived().coeffRef(0); - Scalar* EIGEN_RESTRICT y = &xpr_y.derived().coeffRef(0); - - OtherScalar c = j.c(); - OtherScalar s = j.s(); - if (c==OtherScalar(1) && s==OtherScalar(0)) - return; - - apply_rotation_in_the_plane_selector< - Scalar,OtherScalar, - VectorX::SizeAtCompileTime, - EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment), - Vectorizable>::run(x,incrx,y,incry,size,c,s); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_JACOBI_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h deleted file mode 100644 index d6a3c1e5a54..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h +++ /dev/null @@ -1,101 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_DETERMINANT_H -#define EIGEN_DETERMINANT_H - -namespace Eigen { - -namespace internal { - -template -inline const typename Derived::Scalar bruteforce_det3_helper -(const MatrixBase& matrix, int a, int b, int c) -{ - return matrix.coeff(0,a) - * (matrix.coeff(1,b) * matrix.coeff(2,c) - matrix.coeff(1,c) * matrix.coeff(2,b)); -} - -template -const typename Derived::Scalar bruteforce_det4_helper -(const MatrixBase& matrix, int j, int k, int m, int n) -{ - return (matrix.coeff(j,0) * matrix.coeff(k,1) - matrix.coeff(k,0) * matrix.coeff(j,1)) - * (matrix.coeff(m,2) * matrix.coeff(n,3) - matrix.coeff(n,2) * matrix.coeff(m,3)); -} - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - if(Derived::ColsAtCompileTime==Dynamic && m.rows()==0) - return typename traits::Scalar(1); - return m.partialPivLu().determinant(); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return m.coeff(0,0); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return m.coeff(0,0) * m.coeff(1,1) - m.coeff(1,0) * m.coeff(0,1); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return bruteforce_det3_helper(m,0,1,2) - - bruteforce_det3_helper(m,1,0,2) - + bruteforce_det3_helper(m,2,0,1); - } -}; - -template struct determinant_impl -{ - static typename traits::Scalar run(const Derived& m) - { - // trick by Martin Costabel to compute 4x4 det with only 30 muls - return bruteforce_det4_helper(m,0,1,2,3) - - bruteforce_det4_helper(m,0,2,1,3) - + bruteforce_det4_helper(m,0,3,1,2) - + bruteforce_det4_helper(m,1,2,0,3) - - bruteforce_det4_helper(m,1,3,0,2) - + bruteforce_det4_helper(m,2,3,0,1); - } -}; - -} // end namespace internal - -/** \lu_module - * - * \returns the determinant of this matrix - */ -template -inline typename internal::traits::Scalar MatrixBase::determinant() const -{ - eigen_assert(rows() == cols()); - typedef typename internal::nested_eval::type Nested; - return internal::determinant_impl::type>::run(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_DETERMINANT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h deleted file mode 100644 index 03b6af70613..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h +++ /dev/null @@ -1,891 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_LU_H -#define EIGEN_LU_H - -namespace Eigen { - -namespace internal { -template struct traits > - : traits<_MatrixType> -{ - typedef MatrixXpr XprKind; - typedef SolverStorage StorageKind; - enum { Flags = 0 }; -}; - -} // end namespace internal - -/** \ingroup LU_Module - * - * \class FullPivLU - * - * \brief LU decomposition of a matrix with complete pivoting, and related features - * - * \tparam _MatrixType the type of the matrix of which we are computing the LU decomposition - * - * This class represents a LU decomposition of any matrix, with complete pivoting: the matrix A is - * decomposed as \f$ A = P^{-1} L U Q^{-1} \f$ where L is unit-lower-triangular, U is - * upper-triangular, and P and Q are permutation matrices. This is a rank-revealing LU - * decomposition. The eigenvalues (diagonal coefficients) of U are sorted in such a way that any - * zeros are at the end. - * - * This decomposition provides the generic approach to solving systems of linear equations, computing - * the rank, invertibility, inverse, kernel, and determinant. - * - * This LU decomposition is very stable and well tested with large matrices. However there are use cases where the SVD - * decomposition is inherently more stable and/or flexible. For example, when computing the kernel of a matrix, - * working with the SVD allows to select the smallest singular values of the matrix, something that - * the LU decomposition doesn't see. - * - * The data of the LU decomposition can be directly accessed through the methods matrixLU(), - * permutationP(), permutationQ(). - * - * As an exemple, here is how the original matrix can be retrieved: - * \include class_FullPivLU.cpp - * Output: \verbinclude class_FullPivLU.out - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::fullPivLu(), MatrixBase::determinant(), MatrixBase::inverse() - */ -template class FullPivLU - : public SolverBase > -{ - public: - typedef _MatrixType MatrixType; - typedef SolverBase Base; - - EIGEN_GENERIC_PUBLIC_INTERFACE(FullPivLU) - // FIXME StorageIndex defined in EIGEN_GENERIC_PUBLIC_INTERFACE should be int - enum { - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename internal::plain_row_type::type IntRowVectorType; - typedef typename internal::plain_col_type::type IntColVectorType; - typedef PermutationMatrix PermutationQType; - typedef PermutationMatrix PermutationPType; - typedef typename MatrixType::PlainObject PlainObject; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via LU::compute(const MatrixType&). - */ - FullPivLU(); - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa FullPivLU() - */ - FullPivLU(Index rows, Index cols); - - /** Constructor. - * - * \param matrix the matrix of which to compute the LU decomposition. - * It is required to be nonzero. - */ - template - explicit FullPivLU(const EigenBase& matrix); - - /** \brief Constructs a LU factorization from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. - * - * \sa FullPivLU(const EigenBase&) - */ - template - explicit FullPivLU(EigenBase& matrix); - - /** Computes the LU decomposition of the given matrix. - * - * \param matrix the matrix of which to compute the LU decomposition. - * It is required to be nonzero. - * - * \returns a reference to *this - */ - template - FullPivLU& compute(const EigenBase& matrix) { - m_lu = matrix.derived(); - computeInPlace(); - return *this; - } - - /** \returns the LU decomposition matrix: the upper-triangular part is U, the - * unit-lower-triangular part is L (at least for square matrices; in the non-square - * case, special care is needed, see the documentation of class FullPivLU). - * - * \sa matrixL(), matrixU() - */ - inline const MatrixType& matrixLU() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_lu; - } - - /** \returns the number of nonzero pivots in the LU decomposition. - * Here nonzero is meant in the exact sense, not in a fuzzy sense. - * So that notion isn't really intrinsically interesting, but it is - * still useful when implementing algorithms. - * - * \sa rank() - */ - inline Index nonzeroPivots() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_nonzero_pivots; - } - - /** \returns the absolute value of the biggest pivot, i.e. the biggest - * diagonal coefficient of U. - */ - RealScalar maxPivot() const { return m_maxpivot; } - - /** \returns the permutation matrix P - * - * \sa permutationQ() - */ - EIGEN_DEVICE_FUNC inline const PermutationPType& permutationP() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_p; - } - - /** \returns the permutation matrix Q - * - * \sa permutationP() - */ - inline const PermutationQType& permutationQ() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_q; - } - - /** \returns the kernel of the matrix, also called its null-space. The columns of the returned matrix - * will form a basis of the kernel. - * - * \note If the kernel has dimension zero, then the returned matrix is a column-vector filled with zeros. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - * - * Example: \include FullPivLU_kernel.cpp - * Output: \verbinclude FullPivLU_kernel.out - * - * \sa image() - */ - inline const internal::kernel_retval kernel() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return internal::kernel_retval(*this); - } - - /** \returns the image of the matrix, also called its column-space. The columns of the returned matrix - * will form a basis of the image (column-space). - * - * \param originalMatrix the original matrix, of which *this is the LU decomposition. - * The reason why it is needed to pass it here, is that this allows - * a large optimization, as otherwise this method would need to reconstruct it - * from the LU decomposition. - * - * \note If the image has dimension zero, then the returned matrix is a column-vector filled with zeros. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - * - * Example: \include FullPivLU_image.cpp - * Output: \verbinclude FullPivLU_image.out - * - * \sa kernel() - */ - inline const internal::image_retval - image(const MatrixType& originalMatrix) const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return internal::image_retval(*this, originalMatrix); - } - - /** \return a solution x to the equation Ax=b, where A is the matrix of which - * *this is the LU decomposition. - * - * \param b the right-hand-side of the equation to solve. Can be a vector or a matrix, - * the only requirement in order for the equation to make sense is that - * b.rows()==A.rows(), where A is the matrix of which *this is the LU decomposition. - * - * \returns a solution. - * - * \note_about_checking_solutions - * - * \note_about_arbitrary_choice_of_solution - * \note_about_using_kernel_to_study_multiple_solutions - * - * Example: \include FullPivLU_solve.cpp - * Output: \verbinclude FullPivLU_solve.out - * - * \sa TriangularView::solve(), kernel(), inverse() - */ - // FIXME this is a copy-paste of the base-class member to add the isInitialized assertion. - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return Solve(*this, b.derived()); - } - - /** \returns an estimate of the reciprocal condition number of the matrix of which \c *this is - the LU decomposition. - */ - inline RealScalar rcond() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return internal::rcond_estimate_helper(m_l1_norm, *this); - } - - /** \returns the determinant of the matrix of which - * *this is the LU decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the LU decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \note For fixed-size matrices of size up to 4, MatrixBase::determinant() offers - * optimized paths. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * - * \sa MatrixBase::determinant() - */ - typename internal::traits::Scalar determinant() const; - - /** Allows to prescribe a threshold to be used by certain methods, such as rank(), - * who need to determine when pivots are to be considered nonzero. This is not used for the - * LU decomposition itself. - * - * When it needs to get the threshold value, Eigen calls threshold(). By default, this - * uses a formula to automatically determine a reasonable threshold. - * Once you have called the present method setThreshold(const RealScalar&), - * your value is used instead. - * - * \param threshold The new value to use as the threshold. - * - * A pivot will be considered nonzero if its absolute value is strictly greater than - * \f$ \vert pivot \vert \leqslant threshold \times \vert maxpivot \vert \f$ - * where maxpivot is the biggest pivot. - * - * If you want to come back to the default behavior, call setThreshold(Default_t) - */ - FullPivLU& setThreshold(const RealScalar& threshold) - { - m_usePrescribedThreshold = true; - m_prescribedThreshold = threshold; - return *this; - } - - /** Allows to come back to the default behavior, letting Eigen use its default formula for - * determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code lu.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - FullPivLU& setThreshold(Default_t) - { - m_usePrescribedThreshold = false; - return *this; - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const - { - eigen_assert(m_isInitialized || m_usePrescribedThreshold); - return m_usePrescribedThreshold ? m_prescribedThreshold - // this formula comes from experimenting (see "LU precision tuning" thread on the list) - // and turns out to be identical to Higham's formula used already in LDLt. - : NumTraits::epsilon() * m_lu.diagonalSize(); - } - - /** \returns the rank of the matrix of which *this is the LU decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index rank() const - { - using std::abs; - eigen_assert(m_isInitialized && "LU is not initialized."); - RealScalar premultiplied_threshold = abs(m_maxpivot) * threshold(); - Index result = 0; - for(Index i = 0; i < m_nonzero_pivots; ++i) - result += (abs(m_lu.coeff(i,i)) > premultiplied_threshold); - return result; - } - - /** \returns the dimension of the kernel of the matrix of which *this is the LU decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index dimensionOfKernel() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return cols() - rank(); - } - - /** \returns true if the matrix of which *this is the LU decomposition represents an injective - * linear map, i.e. has trivial kernel; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInjective() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return rank() == cols(); - } - - /** \returns true if the matrix of which *this is the LU decomposition represents a surjective - * linear map; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isSurjective() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return rank() == rows(); - } - - /** \returns true if the matrix of which *this is the LU decomposition is invertible. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInvertible() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return isInjective() && (m_lu.rows() == m_lu.cols()); - } - - /** \returns the inverse of the matrix of which *this is the LU decomposition. - * - * \note If this matrix is not invertible, the returned matrix has undefined coefficients. - * Use isInvertible() to first determine whether this matrix is invertible. - * - * \sa MatrixBase::inverse() - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - eigen_assert(m_lu.rows() == m_lu.cols() && "You can't take the inverse of a non-square matrix!"); - return Inverse(*this); - } - - MatrixType reconstructedMatrix() const; - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_lu.rows(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_lu.cols(); } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - - template - EIGEN_DEVICE_FUNC - void _solve_impl_transposed(const RhsType &rhs, DstType &dst) const; - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - MatrixType m_lu; - PermutationPType m_p; - PermutationQType m_q; - IntColVectorType m_rowsTranspositions; - IntRowVectorType m_colsTranspositions; - Index m_nonzero_pivots; - RealScalar m_l1_norm; - RealScalar m_maxpivot, m_prescribedThreshold; - signed char m_det_pq; - bool m_isInitialized, m_usePrescribedThreshold; -}; - -template -FullPivLU::FullPivLU() - : m_isInitialized(false), m_usePrescribedThreshold(false) -{ -} - -template -FullPivLU::FullPivLU(Index rows, Index cols) - : m_lu(rows, cols), - m_p(rows), - m_q(cols), - m_rowsTranspositions(rows), - m_colsTranspositions(cols), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ -} - -template -template -FullPivLU::FullPivLU(const EigenBase& matrix) - : m_lu(matrix.rows(), matrix.cols()), - m_p(matrix.rows()), - m_q(matrix.cols()), - m_rowsTranspositions(matrix.rows()), - m_colsTranspositions(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ - compute(matrix.derived()); -} - -template -template -FullPivLU::FullPivLU(EigenBase& matrix) - : m_lu(matrix.derived()), - m_p(matrix.rows()), - m_q(matrix.cols()), - m_rowsTranspositions(matrix.rows()), - m_colsTranspositions(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ - computeInPlace(); -} - -template -void FullPivLU::computeInPlace() -{ - check_template_parameters(); - - // the permutations are stored as int indices, so just to be sure: - eigen_assert(m_lu.rows()<=NumTraits::highest() && m_lu.cols()<=NumTraits::highest()); - - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); - - const Index size = m_lu.diagonalSize(); - const Index rows = m_lu.rows(); - const Index cols = m_lu.cols(); - - // will store the transpositions, before we accumulate them at the end. - // can't accumulate on-the-fly because that will be done in reverse order for the rows. - m_rowsTranspositions.resize(m_lu.rows()); - m_colsTranspositions.resize(m_lu.cols()); - Index number_of_transpositions = 0; // number of NONTRIVIAL transpositions, i.e. m_rowsTranspositions[i]!=i - - m_nonzero_pivots = size; // the generic case is that in which all pivots are nonzero (invertible case) - m_maxpivot = RealScalar(0); - - for(Index k = 0; k < size; ++k) - { - // First, we need to find the pivot. - - // biggest coefficient in the remaining bottom-right corner (starting at row k, col k) - Index row_of_biggest_in_corner, col_of_biggest_in_corner; - typedef internal::scalar_score_coeff_op Scoring; - typedef typename Scoring::result_type Score; - Score biggest_in_corner; - biggest_in_corner = m_lu.bottomRightCorner(rows-k, cols-k) - .unaryExpr(Scoring()) - .maxCoeff(&row_of_biggest_in_corner, &col_of_biggest_in_corner); - row_of_biggest_in_corner += k; // correct the values! since they were computed in the corner, - col_of_biggest_in_corner += k; // need to add k to them. - - if(biggest_in_corner==Score(0)) - { - // before exiting, make sure to initialize the still uninitialized transpositions - // in a sane state without destroying what we already have. - m_nonzero_pivots = k; - for(Index i = k; i < size; ++i) - { - m_rowsTranspositions.coeffRef(i) = i; - m_colsTranspositions.coeffRef(i) = i; - } - break; - } - - RealScalar abs_pivot = internal::abs_knowing_score()(m_lu(row_of_biggest_in_corner, col_of_biggest_in_corner), biggest_in_corner); - if(abs_pivot > m_maxpivot) m_maxpivot = abs_pivot; - - // Now that we've found the pivot, we need to apply the row/col swaps to - // bring it to the location (k,k). - - m_rowsTranspositions.coeffRef(k) = row_of_biggest_in_corner; - m_colsTranspositions.coeffRef(k) = col_of_biggest_in_corner; - if(k != row_of_biggest_in_corner) { - m_lu.row(k).swap(m_lu.row(row_of_biggest_in_corner)); - ++number_of_transpositions; - } - if(k != col_of_biggest_in_corner) { - m_lu.col(k).swap(m_lu.col(col_of_biggest_in_corner)); - ++number_of_transpositions; - } - - // Now that the pivot is at the right location, we update the remaining - // bottom-right corner by Gaussian elimination. - - if(k= 0; --k) - m_p.applyTranspositionOnTheRight(k, m_rowsTranspositions.coeff(k)); - - m_q.setIdentity(cols); - for(Index k = 0; k < size; ++k) - m_q.applyTranspositionOnTheRight(k, m_colsTranspositions.coeff(k)); - - m_det_pq = (number_of_transpositions%2) ? -1 : 1; - - m_isInitialized = true; -} - -template -typename internal::traits::Scalar FullPivLU::determinant() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - eigen_assert(m_lu.rows() == m_lu.cols() && "You can't take the determinant of a non-square matrix!"); - return Scalar(m_det_pq) * Scalar(m_lu.diagonal().prod()); -} - -/** \returns the matrix represented by the decomposition, - * i.e., it returns the product: \f$ P^{-1} L U Q^{-1} \f$. - * This function is provided for debug purposes. */ -template -MatrixType FullPivLU::reconstructedMatrix() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - const Index smalldim = (std::min)(m_lu.rows(), m_lu.cols()); - // LU - MatrixType res(m_lu.rows(),m_lu.cols()); - // FIXME the .toDenseMatrix() should not be needed... - res = m_lu.leftCols(smalldim) - .template triangularView().toDenseMatrix() - * m_lu.topRows(smalldim) - .template triangularView().toDenseMatrix(); - - // P^{-1}(LU) - res = m_p.inverse() * res; - - // (P^{-1}LU)Q^{-1} - res = res * m_q.inverse(); - - return res; -} - -/********* Implementation of kernel() **************************************************/ - -namespace internal { -template -struct kernel_retval > - : kernel_retval_base > -{ - EIGEN_MAKE_KERNEL_HELPERS(FullPivLU<_MatrixType>) - - enum { MaxSmallDimAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED( - MatrixType::MaxColsAtCompileTime, - MatrixType::MaxRowsAtCompileTime) - }; - - template void evalTo(Dest& dst) const - { - using std::abs; - const Index cols = dec().matrixLU().cols(), dimker = cols - rank(); - if(dimker == 0) - { - // The Kernel is just {0}, so it doesn't have a basis properly speaking, but let's - // avoid crashing/asserting as that depends on floating point calculations. Let's - // just return a single column vector filled with zeros. - dst.setZero(); - return; - } - - /* Let us use the following lemma: - * - * Lemma: If the matrix A has the LU decomposition PAQ = LU, - * then Ker A = Q(Ker U). - * - * Proof: trivial: just keep in mind that P, Q, L are invertible. - */ - - /* Thus, all we need to do is to compute Ker U, and then apply Q. - * - * U is upper triangular, with eigenvalues sorted so that any zeros appear at the end. - * Thus, the diagonal of U ends with exactly - * dimKer zero's. Let us use that to construct dimKer linearly - * independent vectors in Ker U. - */ - - Matrix pivots(rank()); - RealScalar premultiplied_threshold = dec().maxPivot() * dec().threshold(); - Index p = 0; - for(Index i = 0; i < dec().nonzeroPivots(); ++i) - if(abs(dec().matrixLU().coeff(i,i)) > premultiplied_threshold) - pivots.coeffRef(p++) = i; - eigen_internal_assert(p == rank()); - - // we construct a temporaty trapezoid matrix m, by taking the U matrix and - // permuting the rows and cols to bring the nonnegligible pivots to the top of - // the main diagonal. We need that to be able to apply our triangular solvers. - // FIXME when we get triangularView-for-rectangular-matrices, this can be simplified - Matrix - m(dec().matrixLU().block(0, 0, rank(), cols)); - for(Index i = 0; i < rank(); ++i) - { - if(i) m.row(i).head(i).setZero(); - m.row(i).tail(cols-i) = dec().matrixLU().row(pivots.coeff(i)).tail(cols-i); - } - m.block(0, 0, rank(), rank()); - m.block(0, 0, rank(), rank()).template triangularView().setZero(); - for(Index i = 0; i < rank(); ++i) - m.col(i).swap(m.col(pivots.coeff(i))); - - // ok, we have our trapezoid matrix, we can apply the triangular solver. - // notice that the math behind this suggests that we should apply this to the - // negative of the RHS, but for performance we just put the negative sign elsewhere, see below. - m.topLeftCorner(rank(), rank()) - .template triangularView().solveInPlace( - m.topRightCorner(rank(), dimker) - ); - - // now we must undo the column permutation that we had applied! - for(Index i = rank()-1; i >= 0; --i) - m.col(i).swap(m.col(pivots.coeff(i))); - - // see the negative sign in the next line, that's what we were talking about above. - for(Index i = 0; i < rank(); ++i) dst.row(dec().permutationQ().indices().coeff(i)) = -m.row(i).tail(dimker); - for(Index i = rank(); i < cols; ++i) dst.row(dec().permutationQ().indices().coeff(i)).setZero(); - for(Index k = 0; k < dimker; ++k) dst.coeffRef(dec().permutationQ().indices().coeff(rank()+k), k) = Scalar(1); - } -}; - -/***** Implementation of image() *****************************************************/ - -template -struct image_retval > - : image_retval_base > -{ - EIGEN_MAKE_IMAGE_HELPERS(FullPivLU<_MatrixType>) - - enum { MaxSmallDimAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED( - MatrixType::MaxColsAtCompileTime, - MatrixType::MaxRowsAtCompileTime) - }; - - template void evalTo(Dest& dst) const - { - using std::abs; - if(rank() == 0) - { - // The Image is just {0}, so it doesn't have a basis properly speaking, but let's - // avoid crashing/asserting as that depends on floating point calculations. Let's - // just return a single column vector filled with zeros. - dst.setZero(); - return; - } - - Matrix pivots(rank()); - RealScalar premultiplied_threshold = dec().maxPivot() * dec().threshold(); - Index p = 0; - for(Index i = 0; i < dec().nonzeroPivots(); ++i) - if(abs(dec().matrixLU().coeff(i,i)) > premultiplied_threshold) - pivots.coeffRef(p++) = i; - eigen_internal_assert(p == rank()); - - for(Index i = 0; i < rank(); ++i) - dst.col(i) = originalMatrix().col(dec().permutationQ().indices().coeff(pivots.coeff(i))); - } -}; - -/***** Implementation of solve() *****************************************************/ - -} // end namespace internal - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void FullPivLU<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - /* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1}. - * So we proceed as follows: - * Step 1: compute c = P * rhs. - * Step 2: replace c by the solution x to Lx = c. Exists because L is invertible. - * Step 3: replace c by the solution x to Ux = c. May or may not exist. - * Step 4: result = Q * c; - */ - - const Index rows = this->rows(), - cols = this->cols(), - nonzero_pivots = this->rank(); - eigen_assert(rhs.rows() == rows); - const Index smalldim = (std::min)(rows, cols); - - if(nonzero_pivots == 0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs.rows(), rhs.cols()); - - // Step 1 - c = permutationP() * rhs; - - // Step 2 - m_lu.topLeftCorner(smalldim,smalldim) - .template triangularView() - .solveInPlace(c.topRows(smalldim)); - if(rows>cols) - c.bottomRows(rows-cols) -= m_lu.bottomRows(rows-cols) * c.topRows(cols); - - // Step 3 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .solveInPlace(c.topRows(nonzero_pivots)); - - // Step 4 - for(Index i = 0; i < nonzero_pivots; ++i) - dst.row(permutationQ().indices().coeff(i)) = c.row(i); - for(Index i = nonzero_pivots; i < m_lu.cols(); ++i) - dst.row(permutationQ().indices().coeff(i)).setZero(); -} - -template -template -void FullPivLU<_MatrixType>::_solve_impl_transposed(const RhsType &rhs, DstType &dst) const -{ - /* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1}, - * and since permutations are real and unitary, we can write this - * as A^T = Q U^T L^T P, - * So we proceed as follows: - * Step 1: compute c = Q^T rhs. - * Step 2: replace c by the solution x to U^T x = c. May or may not exist. - * Step 3: replace c by the solution x to L^T x = c. - * Step 4: result = P^T c. - * If Conjugate is true, replace "^T" by "^*" above. - */ - - const Index rows = this->rows(), cols = this->cols(), - nonzero_pivots = this->rank(); - eigen_assert(rhs.rows() == cols); - const Index smalldim = (std::min)(rows, cols); - - if(nonzero_pivots == 0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs.rows(), rhs.cols()); - - // Step 1 - c = permutationQ().inverse() * rhs; - - if (Conjugate) { - // Step 2 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .adjoint() - .solveInPlace(c.topRows(nonzero_pivots)); - // Step 3 - m_lu.topLeftCorner(smalldim, smalldim) - .template triangularView() - .adjoint() - .solveInPlace(c.topRows(smalldim)); - } else { - // Step 2 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .transpose() - .solveInPlace(c.topRows(nonzero_pivots)); - // Step 3 - m_lu.topLeftCorner(smalldim, smalldim) - .template triangularView() - .transpose() - .solveInPlace(c.topRows(smalldim)); - } - - // Step 4 - PermutationPType invp = permutationP().inverse().eval(); - for(Index i = 0; i < smalldim; ++i) - dst.row(invp.indices().coeff(i)) = c.row(i); - for(Index i = smalldim; i < rows; ++i) - dst.row(invp.indices().coeff(i)).setZero(); -} - -#endif - -namespace internal { - - -/***** Implementation of inverse() *****************************************************/ -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef FullPivLU LuType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; -} // end namespace internal - -/******* MatrixBase methods *****************************************************************/ - -/** \lu_module - * - * \return the full-pivoting LU decomposition of \c *this. - * - * \sa class FullPivLU - */ -template -inline const FullPivLU::PlainObject> -MatrixBase::fullPivLu() const -{ - return FullPivLU(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_LU_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h deleted file mode 100644 index f49f2336004..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h +++ /dev/null @@ -1,415 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Benoit Jacob -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_INVERSE_IMPL_H -#define EIGEN_INVERSE_IMPL_H - -namespace Eigen { - -namespace internal { - -/********************************** -*** General case implementation *** -**********************************/ - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - result = matrix.partialPivLu().inverse(); - } -}; - -template -struct compute_inverse_and_det_with_check { /* nothing! general case not supported. */ }; - -/**************************** -*** Size 1 implementation *** -****************************/ - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename MatrixType::Scalar Scalar; - internal::evaluator matrixEval(matrix); - result.coeffRef(0,0) = Scalar(1) / matrixEval.coeff(0,0); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& result, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - determinant = matrix.coeff(0,0); - invertible = abs(determinant) > absDeterminantThreshold; - if(invertible) result.coeffRef(0,0) = typename ResultType::Scalar(1) / determinant; - } -}; - -/**************************** -*** Size 2 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline void compute_inverse_size2_helper( - const MatrixType& matrix, const typename ResultType::Scalar& invdet, - ResultType& result) -{ - result.coeffRef(0,0) = matrix.coeff(1,1) * invdet; - result.coeffRef(1,0) = -matrix.coeff(1,0) * invdet; - result.coeffRef(0,1) = -matrix.coeff(0,1) * invdet; - result.coeffRef(1,1) = matrix.coeff(0,0) * invdet; -} - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename ResultType::Scalar Scalar; - const Scalar invdet = typename MatrixType::Scalar(1) / matrix.determinant(); - compute_inverse_size2_helper(matrix, invdet, result); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - typedef typename ResultType::Scalar Scalar; - determinant = matrix.determinant(); - invertible = abs(determinant) > absDeterminantThreshold; - if(!invertible) return; - const Scalar invdet = Scalar(1) / determinant; - compute_inverse_size2_helper(matrix, invdet, inverse); - } -}; - -/**************************** -*** Size 3 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline typename MatrixType::Scalar cofactor_3x3(const MatrixType& m) -{ - enum { - i1 = (i+1) % 3, - i2 = (i+2) % 3, - j1 = (j+1) % 3, - j2 = (j+2) % 3 - }; - return m.coeff(i1, j1) * m.coeff(i2, j2) - - m.coeff(i1, j2) * m.coeff(i2, j1); -} - -template -EIGEN_DEVICE_FUNC -inline void compute_inverse_size3_helper( - const MatrixType& matrix, - const typename ResultType::Scalar& invdet, - const Matrix& cofactors_col0, - ResultType& result) -{ - result.row(0) = cofactors_col0 * invdet; - result.coeffRef(1,0) = cofactor_3x3(matrix) * invdet; - result.coeffRef(1,1) = cofactor_3x3(matrix) * invdet; - result.coeffRef(1,2) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,0) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,1) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,2) = cofactor_3x3(matrix) * invdet; -} - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename ResultType::Scalar Scalar; - Matrix cofactors_col0; - cofactors_col0.coeffRef(0) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(1) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(2) = cofactor_3x3(matrix); - const Scalar det = (cofactors_col0.cwiseProduct(matrix.col(0))).sum(); - const Scalar invdet = Scalar(1) / det; - compute_inverse_size3_helper(matrix, invdet, cofactors_col0, result); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - typedef typename ResultType::Scalar Scalar; - Matrix cofactors_col0; - cofactors_col0.coeffRef(0) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(1) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(2) = cofactor_3x3(matrix); - determinant = (cofactors_col0.cwiseProduct(matrix.col(0))).sum(); - invertible = abs(determinant) > absDeterminantThreshold; - if(!invertible) return; - const Scalar invdet = Scalar(1) / determinant; - compute_inverse_size3_helper(matrix, invdet, cofactors_col0, inverse); - } -}; - -/**************************** -*** Size 4 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline const typename Derived::Scalar general_det3_helper -(const MatrixBase& matrix, int i1, int i2, int i3, int j1, int j2, int j3) -{ - return matrix.coeff(i1,j1) - * (matrix.coeff(i2,j2) * matrix.coeff(i3,j3) - matrix.coeff(i2,j3) * matrix.coeff(i3,j2)); -} - -template -EIGEN_DEVICE_FUNC -inline typename MatrixType::Scalar cofactor_4x4(const MatrixType& matrix) -{ - enum { - i1 = (i+1) % 4, - i2 = (i+2) % 4, - i3 = (i+3) % 4, - j1 = (j+1) % 4, - j2 = (j+2) % 4, - j3 = (j+3) % 4 - }; - return general_det3_helper(matrix, i1, i2, i3, j1, j2, j3) - + general_det3_helper(matrix, i2, i3, i1, j1, j2, j3) - + general_det3_helper(matrix, i3, i1, i2, j1, j2, j3); -} - -template -struct compute_inverse_size4 -{ - EIGEN_DEVICE_FUNC - static void run(const MatrixType& matrix, ResultType& result) - { - result.coeffRef(0,0) = cofactor_4x4(matrix); - result.coeffRef(1,0) = -cofactor_4x4(matrix); - result.coeffRef(2,0) = cofactor_4x4(matrix); - result.coeffRef(3,0) = -cofactor_4x4(matrix); - result.coeffRef(0,2) = cofactor_4x4(matrix); - result.coeffRef(1,2) = -cofactor_4x4(matrix); - result.coeffRef(2,2) = cofactor_4x4(matrix); - result.coeffRef(3,2) = -cofactor_4x4(matrix); - result.coeffRef(0,1) = -cofactor_4x4(matrix); - result.coeffRef(1,1) = cofactor_4x4(matrix); - result.coeffRef(2,1) = -cofactor_4x4(matrix); - result.coeffRef(3,1) = cofactor_4x4(matrix); - result.coeffRef(0,3) = -cofactor_4x4(matrix); - result.coeffRef(1,3) = cofactor_4x4(matrix); - result.coeffRef(2,3) = -cofactor_4x4(matrix); - result.coeffRef(3,3) = cofactor_4x4(matrix); - result /= (matrix.col(0).cwiseProduct(result.row(0).transpose())).sum(); - } -}; - -template -struct compute_inverse - : compute_inverse_size4 -{ -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - determinant = matrix.determinant(); - invertible = abs(determinant) > absDeterminantThreshold; - if(invertible) compute_inverse::run(matrix, inverse); - } -}; - -/************************* -*** MatrixBase methods *** -*************************/ - -} // end namespace internal - -namespace internal { - -// Specialization for "dense = dense_xpr.inverse()" -template -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - const int Size = EIGEN_PLAIN_ENUM_MIN(XprType::ColsAtCompileTime,DstXprType::ColsAtCompileTime); - EIGEN_ONLY_USED_FOR_DEBUG(Size); - eigen_assert(( (Size<=1) || (Size>4) || (extract_data(src.nestedExpression())!=extract_data(dst))) - && "Aliasing problem detected in inverse(), you need to do inverse().eval() here."); - - typedef typename internal::nested_eval::type ActualXprType; - typedef typename internal::remove_all::type ActualXprTypeCleanded; - - ActualXprType actual_xpr(src.nestedExpression()); - - compute_inverse::run(actual_xpr, dst); - } -}; - - -} // end namespace internal - -/** \lu_module - * - * \returns the matrix inverse of this matrix. - * - * For small fixed sizes up to 4x4, this method uses cofactors. - * In the general case, this method uses class PartialPivLU. - * - * \note This matrix must be invertible, otherwise the result is undefined. If you need an - * invertibility check, do the following: - * \li for fixed sizes up to 4x4, use computeInverseAndDetWithCheck(). - * \li for the general case, use class FullPivLU. - * - * Example: \include MatrixBase_inverse.cpp - * Output: \verbinclude MatrixBase_inverse.out - * - * \sa computeInverseAndDetWithCheck() - */ -template -inline const Inverse MatrixBase::inverse() const -{ - EIGEN_STATIC_ASSERT(!NumTraits::IsInteger,THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES) - eigen_assert(rows() == cols()); - return Inverse(derived()); -} - -/** \lu_module - * - * Computation of matrix inverse and determinant, with invertibility check. - * - * This is only for fixed-size square matrices of size up to 4x4. - * - * \param inverse Reference to the matrix in which to store the inverse. - * \param determinant Reference to the variable in which to store the determinant. - * \param invertible Reference to the bool variable in which to store whether the matrix is invertible. - * \param absDeterminantThreshold Optional parameter controlling the invertibility check. - * The matrix will be declared invertible if the absolute value of its - * determinant is greater than this threshold. - * - * Example: \include MatrixBase_computeInverseAndDetWithCheck.cpp - * Output: \verbinclude MatrixBase_computeInverseAndDetWithCheck.out - * - * \sa inverse(), computeInverseWithCheck() - */ -template -template -inline void MatrixBase::computeInverseAndDetWithCheck( - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible, - const RealScalar& absDeterminantThreshold - ) const -{ - // i'd love to put some static assertions there, but SFINAE means that they have no effect... - eigen_assert(rows() == cols()); - // for 2x2, it's worth giving a chance to avoid evaluating. - // for larger sizes, evaluating has negligible cost and limits code size. - typedef typename internal::conditional< - RowsAtCompileTime == 2, - typename internal::remove_all::type>::type, - PlainObject - >::type MatrixType; - internal::compute_inverse_and_det_with_check::run - (derived(), absDeterminantThreshold, inverse, determinant, invertible); -} - -/** \lu_module - * - * Computation of matrix inverse, with invertibility check. - * - * This is only for fixed-size square matrices of size up to 4x4. - * - * \param inverse Reference to the matrix in which to store the inverse. - * \param invertible Reference to the bool variable in which to store whether the matrix is invertible. - * \param absDeterminantThreshold Optional parameter controlling the invertibility check. - * The matrix will be declared invertible if the absolute value of its - * determinant is greater than this threshold. - * - * Example: \include MatrixBase_computeInverseWithCheck.cpp - * Output: \verbinclude MatrixBase_computeInverseWithCheck.out - * - * \sa inverse(), computeInverseAndDetWithCheck() - */ -template -template -inline void MatrixBase::computeInverseWithCheck( - ResultType& inverse, - bool& invertible, - const RealScalar& absDeterminantThreshold - ) const -{ - Scalar determinant; - // i'd love to put some static assertions there, but SFINAE means that they have no effect... - eigen_assert(rows() == cols()); - computeInverseAndDetWithCheck(inverse,determinant,invertible,absDeterminantThreshold); -} - -} // end namespace Eigen - -#endif // EIGEN_INVERSE_IMPL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h deleted file mode 100644 index 6b10f39fab4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h +++ /dev/null @@ -1,614 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2009 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARTIALLU_H -#define EIGEN_PARTIALLU_H - -namespace Eigen { - -namespace internal { -template struct traits > - : traits<_MatrixType> -{ - typedef MatrixXpr XprKind; - typedef SolverStorage StorageKind; - typedef traits<_MatrixType> BaseTraits; - enum { - Flags = BaseTraits::Flags & RowMajorBit, - CoeffReadCost = Dynamic - }; -}; - -template -struct enable_if_ref; -// { -// typedef Derived type; -// }; - -template -struct enable_if_ref,Derived> { - typedef Derived type; -}; - -} // end namespace internal - -/** \ingroup LU_Module - * - * \class PartialPivLU - * - * \brief LU decomposition of a matrix with partial pivoting, and related features - * - * \tparam _MatrixType the type of the matrix of which we are computing the LU decomposition - * - * This class represents a LU decomposition of a \b square \b invertible matrix, with partial pivoting: the matrix A - * is decomposed as A = PLU where L is unit-lower-triangular, U is upper-triangular, and P - * is a permutation matrix. - * - * Typically, partial pivoting LU decomposition is only considered numerically stable for square invertible - * matrices. Thus LAPACK's dgesv and dgesvx require the matrix to be square and invertible. The present class - * does the same. It will assert that the matrix is square, but it won't (actually it can't) check that the - * matrix is invertible: it is your task to check that you only use this decomposition on invertible matrices. - * - * The guaranteed safe alternative, working for all matrices, is the full pivoting LU decomposition, provided - * by class FullPivLU. - * - * This is \b not a rank-revealing LU decomposition. Many features are intentionally absent from this class, - * such as rank computation. If you need these features, use class FullPivLU. - * - * This LU decomposition is suitable to invert invertible matrices. It is what MatrixBase::inverse() uses - * in the general case. - * On the other hand, it is \b not suitable to determine whether a given matrix is invertible. - * - * The data of the LU decomposition can be directly accessed through the methods matrixLU(), permutationP(). - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::partialPivLu(), MatrixBase::determinant(), MatrixBase::inverse(), MatrixBase::computeInverse(), class FullPivLU - */ -template class PartialPivLU - : public SolverBase > -{ - public: - - typedef _MatrixType MatrixType; - typedef SolverBase Base; - EIGEN_GENERIC_PUBLIC_INTERFACE(PartialPivLU) - // FIXME StorageIndex defined in EIGEN_GENERIC_PUBLIC_INTERFACE should be int - enum { - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef PermutationMatrix PermutationType; - typedef Transpositions TranspositionType; - typedef typename MatrixType::PlainObject PlainObject; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via PartialPivLU::compute(const MatrixType&). - */ - PartialPivLU(); - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa PartialPivLU() - */ - explicit PartialPivLU(Index size); - - /** Constructor. - * - * \param matrix the matrix of which to compute the LU decomposition. - * - * \warning The matrix should have full rank (e.g. if it's square, it should be invertible). - * If you need to deal with non-full rank, use class FullPivLU instead. - */ - template - explicit PartialPivLU(const EigenBase& matrix); - - /** Constructor for \link InplaceDecomposition inplace decomposition \endlink - * - * \param matrix the matrix of which to compute the LU decomposition. - * - * \warning The matrix should have full rank (e.g. if it's square, it should be invertible). - * If you need to deal with non-full rank, use class FullPivLU instead. - */ - template - explicit PartialPivLU(EigenBase& matrix); - - template - PartialPivLU& compute(const EigenBase& matrix) { - m_lu = matrix.derived(); - compute(); - return *this; - } - - /** \returns the LU decomposition matrix: the upper-triangular part is U, the - * unit-lower-triangular part is L (at least for square matrices; in the non-square - * case, special care is needed, see the documentation of class FullPivLU). - * - * \sa matrixL(), matrixU() - */ - inline const MatrixType& matrixLU() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return m_lu; - } - - /** \returns the permutation matrix P. - */ - inline const PermutationType& permutationP() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return m_p; - } - - /** This method returns the solution x to the equation Ax=b, where A is the matrix of which - * *this is the LU decomposition. - * - * \param b the right-hand-side of the equation to solve. Can be a vector or a matrix, - * the only requirement in order for the equation to make sense is that - * b.rows()==A.rows(), where A is the matrix of which *this is the LU decomposition. - * - * \returns the solution. - * - * Example: \include PartialPivLU_solve.cpp - * Output: \verbinclude PartialPivLU_solve.out - * - * Since this PartialPivLU class assumes anyway that the matrix A is invertible, the solution - * theoretically exists and is unique regardless of b. - * - * \sa TriangularView::solve(), inverse(), computeInverse() - */ - // FIXME this is a copy-paste of the base-class member to add the isInitialized assertion. - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Solve(*this, b.derived()); - } - - /** \returns an estimate of the reciprocal condition number of the matrix of which \c *this is - the LU decomposition. - */ - inline RealScalar rcond() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return internal::rcond_estimate_helper(m_l1_norm, *this); - } - - /** \returns the inverse of the matrix of which *this is the LU decomposition. - * - * \warning The matrix being decomposed here is assumed to be invertible. If you need to check for - * invertibility, use class FullPivLU instead. - * - * \sa MatrixBase::inverse(), LU::inverse() - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Inverse(*this); - } - - /** \returns the determinant of the matrix of which - * *this is the LU decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the LU decomposition has already been computed. - * - * \note For fixed-size matrices of size up to 4, MatrixBase::determinant() offers - * optimized paths. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * - * \sa MatrixBase::determinant() - */ - Scalar determinant() const; - - MatrixType reconstructedMatrix() const; - - inline Index rows() const { return m_lu.rows(); } - inline Index cols() const { return m_lu.cols(); } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const { - /* The decomposition PA = LU can be rewritten as A = P^{-1} L U. - * So we proceed as follows: - * Step 1: compute c = Pb. - * Step 2: replace c by the solution x to Lx = c. - * Step 3: replace c by the solution x to Ux = c. - */ - - eigen_assert(rhs.rows() == m_lu.rows()); - - // Step 1 - dst = permutationP() * rhs; - - // Step 2 - m_lu.template triangularView().solveInPlace(dst); - - // Step 3 - m_lu.template triangularView().solveInPlace(dst); - } - - template - EIGEN_DEVICE_FUNC - void _solve_impl_transposed(const RhsType &rhs, DstType &dst) const { - /* The decomposition PA = LU can be rewritten as A = P^{-1} L U. - * So we proceed as follows: - * Step 1: compute c = Pb. - * Step 2: replace c by the solution x to Lx = c. - * Step 3: replace c by the solution x to Ux = c. - */ - - eigen_assert(rhs.rows() == m_lu.cols()); - - if (Conjugate) { - // Step 1 - dst = m_lu.template triangularView().adjoint().solve(rhs); - // Step 2 - m_lu.template triangularView().adjoint().solveInPlace(dst); - } else { - // Step 1 - dst = m_lu.template triangularView().transpose().solve(rhs); - // Step 2 - m_lu.template triangularView().transpose().solveInPlace(dst); - } - // Step 3 - dst = permutationP().transpose() * dst; - } - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void compute(); - - MatrixType m_lu; - PermutationType m_p; - TranspositionType m_rowsTranspositions; - RealScalar m_l1_norm; - signed char m_det_p; - bool m_isInitialized; -}; - -template -PartialPivLU::PartialPivLU() - : m_lu(), - m_p(), - m_rowsTranspositions(), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ -} - -template -PartialPivLU::PartialPivLU(Index size) - : m_lu(size, size), - m_p(size), - m_rowsTranspositions(size), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ -} - -template -template -PartialPivLU::PartialPivLU(const EigenBase& matrix) - : m_lu(matrix.rows(),matrix.cols()), - m_p(matrix.rows()), - m_rowsTranspositions(matrix.rows()), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ - compute(matrix.derived()); -} - -template -template -PartialPivLU::PartialPivLU(EigenBase& matrix) - : m_lu(matrix.derived()), - m_p(matrix.rows()), - m_rowsTranspositions(matrix.rows()), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ - compute(); -} - -namespace internal { - -/** \internal This is the blocked version of fullpivlu_unblocked() */ -template -struct partial_lu_impl -{ - // FIXME add a stride to Map, so that the following mapping becomes easier, - // another option would be to create an expression being able to automatically - // warp any Map, Matrix, and Block expressions as a unique type, but since that's exactly - // a Map + stride, why not adding a stride to Map, and convenient ctors from a Matrix, - // and Block. - typedef Map > MapLU; - typedef Block MatrixType; - typedef Block BlockType; - typedef typename MatrixType::RealScalar RealScalar; - - /** \internal performs the LU decomposition in-place of the matrix \a lu - * using an unblocked algorithm. - * - * In addition, this function returns the row transpositions in the - * vector \a row_transpositions which must have a size equal to the number - * of columns of the matrix \a lu, and an integer \a nb_transpositions - * which returns the actual number of transpositions. - * - * \returns The index of the first pivot which is exactly zero if any, or a negative number otherwise. - */ - static Index unblocked_lu(MatrixType& lu, PivIndex* row_transpositions, PivIndex& nb_transpositions) - { - typedef scalar_score_coeff_op Scoring; - typedef typename Scoring::result_type Score; - const Index rows = lu.rows(); - const Index cols = lu.cols(); - const Index size = (std::min)(rows,cols); - nb_transpositions = 0; - Index first_zero_pivot = -1; - for(Index k = 0; k < size; ++k) - { - Index rrows = rows-k-1; - Index rcols = cols-k-1; - - Index row_of_biggest_in_col; - Score biggest_in_corner - = lu.col(k).tail(rows-k).unaryExpr(Scoring()).maxCoeff(&row_of_biggest_in_col); - row_of_biggest_in_col += k; - - row_transpositions[k] = PivIndex(row_of_biggest_in_col); - - if(biggest_in_corner != Score(0)) - { - if(k != row_of_biggest_in_col) - { - lu.row(k).swap(lu.row(row_of_biggest_in_col)); - ++nb_transpositions; - } - - // FIXME shall we introduce a safe quotient expression in cas 1/lu.coeff(k,k) - // overflow but not the actual quotient? - lu.col(k).tail(rrows) /= lu.coeff(k,k); - } - else if(first_zero_pivot==-1) - { - // the pivot is exactly zero, we record the index of the first pivot which is exactly 0, - // and continue the factorization such we still have A = PLU - first_zero_pivot = k; - } - - if(k > > - */ - static Index blocked_lu(Index rows, Index cols, Scalar* lu_data, Index luStride, PivIndex* row_transpositions, PivIndex& nb_transpositions, Index maxBlockSize=256) - { - MapLU lu1(lu_data,StorageOrder==RowMajor?rows:luStride,StorageOrder==RowMajor?luStride:cols); - MatrixType lu(lu1,0,0,rows,cols); - - const Index size = (std::min)(rows,cols); - - // if the matrix is too small, no blocking: - if(size<=16) - { - return unblocked_lu(lu, row_transpositions, nb_transpositions); - } - - // automatically adjust the number of subdivisions to the size - // of the matrix so that there is enough sub blocks: - Index blockSize; - { - blockSize = size/8; - blockSize = (blockSize/16)*16; - blockSize = (std::min)((std::max)(blockSize,Index(8)), maxBlockSize); - } - - nb_transpositions = 0; - Index first_zero_pivot = -1; - for(Index k = 0; k < size; k+=blockSize) - { - Index bs = (std::min)(size-k,blockSize); // actual size of the block - Index trows = rows - k - bs; // trailing rows - Index tsize = size - k - bs; // trailing size - - // partition the matrix: - // A00 | A01 | A02 - // lu = A_0 | A_1 | A_2 = A10 | A11 | A12 - // A20 | A21 | A22 - BlockType A_0(lu,0,0,rows,k); - BlockType A_2(lu,0,k+bs,rows,tsize); - BlockType A11(lu,k,k,bs,bs); - BlockType A12(lu,k,k+bs,bs,tsize); - BlockType A21(lu,k+bs,k,trows,bs); - BlockType A22(lu,k+bs,k+bs,trows,tsize); - - PivIndex nb_transpositions_in_panel; - // recursively call the blocked LU algorithm on [A11^T A21^T]^T - // with a very small blocking size: - Index ret = blocked_lu(trows+bs, bs, &lu.coeffRef(k,k), luStride, - row_transpositions+k, nb_transpositions_in_panel, 16); - if(ret>=0 && first_zero_pivot==-1) - first_zero_pivot = k+ret; - - nb_transpositions += nb_transpositions_in_panel; - // update permutations and apply them to A_0 - for(Index i=k; i(k)); - A_0.row(i).swap(A_0.row(piv)); - } - - if(trows) - { - // apply permutations to A_2 - for(Index i=k;i().solveInPlace(A12); - - A22.noalias() -= A21 * A12; - } - } - return first_zero_pivot; - } -}; - -/** \internal performs the LU decomposition with partial pivoting in-place. - */ -template -void partial_lu_inplace(MatrixType& lu, TranspositionType& row_transpositions, typename TranspositionType::StorageIndex& nb_transpositions) -{ - eigen_assert(lu.cols() == row_transpositions.size()); - eigen_assert((&row_transpositions.coeffRef(1)-&row_transpositions.coeffRef(0)) == 1); - - partial_lu_impl - - ::blocked_lu(lu.rows(), lu.cols(), &lu.coeffRef(0,0), lu.outerStride(), &row_transpositions.coeffRef(0), nb_transpositions); -} - -} // end namespace internal - -template -void PartialPivLU::compute() -{ - check_template_parameters(); - - // the row permutation is stored as int indices, so just to be sure: - eigen_assert(m_lu.rows()::highest()); - - if(m_lu.cols()>0) - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); - else - m_l1_norm = RealScalar(0); - - eigen_assert(m_lu.rows() == m_lu.cols() && "PartialPivLU is only for square (and moreover invertible) matrices"); - const Index size = m_lu.rows(); - - m_rowsTranspositions.resize(size); - - typename TranspositionType::StorageIndex nb_transpositions; - internal::partial_lu_inplace(m_lu, m_rowsTranspositions, nb_transpositions); - m_det_p = (nb_transpositions%2) ? -1 : 1; - - m_p = m_rowsTranspositions; - - m_isInitialized = true; -} - -template -typename PartialPivLU::Scalar PartialPivLU::determinant() const -{ - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Scalar(m_det_p) * m_lu.diagonal().prod(); -} - -/** \returns the matrix represented by the decomposition, - * i.e., it returns the product: P^{-1} L U. - * This function is provided for debug purpose. */ -template -MatrixType PartialPivLU::reconstructedMatrix() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - // LU - MatrixType res = m_lu.template triangularView().toDenseMatrix() - * m_lu.template triangularView(); - - // P^{-1}(LU) - res = m_p.inverse() * res; - - return res; -} - -/***** Implementation details *****************************************************/ - -namespace internal { - -/***** Implementation of inverse() *****************************************************/ -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef PartialPivLU LuType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; -} // end namespace internal - -/******** MatrixBase methods *******/ - -/** \lu_module - * - * \return the partial-pivoting LU decomposition of \c *this. - * - * \sa class PartialPivLU - */ -template -inline const PartialPivLU::PlainObject> -MatrixBase::partialPivLu() const -{ - return PartialPivLU(eval()); -} - -/** \lu_module - * - * Synonym of partialPivLu(). - * - * \return the partial-pivoting LU decomposition of \c *this. - * - * \sa class PartialPivLU - */ -template -inline const PartialPivLU::PlainObject> -MatrixBase::lu() const -{ - return PartialPivLU(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_PARTIALLU_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h deleted file mode 100644 index 755168a9460..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h +++ /dev/null @@ -1,83 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * LU decomposition with partial pivoting based on LAPACKE_?getrf function. - ******************************************************************************** -*/ - -#ifndef EIGEN_PARTIALLU_LAPACK_H -#define EIGEN_PARTIALLU_LAPACK_H - -namespace Eigen { - -namespace internal { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_LU_PARTPIV(EIGTYPE, LAPACKE_TYPE, LAPACKE_PREFIX) \ -template \ -struct partial_lu_impl \ -{ \ - /* \internal performs the LU decomposition in-place of the matrix represented */ \ - static lapack_int blocked_lu(Index rows, Index cols, EIGTYPE* lu_data, Index luStride, lapack_int* row_transpositions, lapack_int& nb_transpositions, lapack_int maxBlockSize=256) \ - { \ - EIGEN_UNUSED_VARIABLE(maxBlockSize);\ - lapack_int matrix_order, first_zero_pivot; \ - lapack_int m, n, lda, *ipiv, info; \ - EIGTYPE* a; \ -/* Set up parameters for ?getrf */ \ - matrix_order = StorageOrder==RowMajor ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \ - lda = convert_index(luStride); \ - a = lu_data; \ - ipiv = row_transpositions; \ - m = convert_index(rows); \ - n = convert_index(cols); \ - nb_transpositions = 0; \ -\ - info = LAPACKE_##LAPACKE_PREFIX##getrf( matrix_order, m, n, (LAPACKE_TYPE*)a, lda, ipiv ); \ -\ - for(int i=0;i= 0); \ -/* something should be done with nb_transpositions */ \ -\ - first_zero_pivot = info; \ - return first_zero_pivot; \ - } \ -}; - -EIGEN_LAPACKE_LU_PARTPIV(double, double, d) -EIGEN_LAPACKE_LU_PARTPIV(float, float, s) -EIGEN_LAPACKE_LU_PARTPIV(dcomplex, lapack_complex_double, z) -EIGEN_LAPACKE_LU_PARTPIV(scomplex, lapack_complex_float, c) - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PARTIALLU_LAPACK_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h deleted file mode 100644 index 4dce2ef20ee..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h +++ /dev/null @@ -1,338 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2001 Intel Corporation -// Copyright (C) 2010 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// The SSE code for the 4x4 float and double matrix inverse in this file -// comes from the following Intel's library: -// http://software.intel.com/en-us/articles/optimized-matrix-library-for-use-with-the-intel-pentiumr-4-processors-sse2-instructions/ -// -// Here is the respective copyright and license statement: -// -// Copyright (c) 2001 Intel Corporation. -// -// Permition is granted to use, copy, distribute and prepare derivative works -// of this library for any purpose and without fee, provided, that the above -// copyright notice and this statement appear in all copies. -// Intel makes no representations about the suitability of this software for -// any purpose, and specifically disclaims all warranties. -// See LEGAL.TXT for all the legal information. - -#ifndef EIGEN_INVERSE_SSE_H -#define EIGEN_INVERSE_SSE_H - -namespace Eigen { - -namespace internal { - -template -struct compute_inverse_size4 -{ - enum { - MatrixAlignment = traits::Alignment, - ResultAlignment = traits::Alignment, - StorageOrdersMatch = (MatrixType::Flags&RowMajorBit) == (ResultType::Flags&RowMajorBit) - }; - typedef typename conditional<(MatrixType::Flags&LinearAccessBit),MatrixType const &,typename MatrixType::PlainObject>::type ActualMatrixType; - - static void run(const MatrixType& mat, ResultType& result) - { - ActualMatrixType matrix(mat); - const Packet4f p4f_sign_PNNP = _mm_castsi128_ps(_mm_set_epi32(0x00000000, 0x80000000, 0x80000000, 0x00000000)); - - // Load the full matrix into registers - __m128 _L1 = matrix.template packet( 0); - __m128 _L2 = matrix.template packet( 4); - __m128 _L3 = matrix.template packet( 8); - __m128 _L4 = matrix.template packet(12); - - // The inverse is calculated using "Divide and Conquer" technique. The - // original matrix is divide into four 2x2 sub-matrices. Since each - // register holds four matrix element, the smaller matrices are - // represented as a registers. Hence we get a better locality of the - // calculations. - - __m128 A, B, C, D; // the four sub-matrices - if(!StorageOrdersMatch) - { - A = _mm_unpacklo_ps(_L1, _L2); - B = _mm_unpacklo_ps(_L3, _L4); - C = _mm_unpackhi_ps(_L1, _L2); - D = _mm_unpackhi_ps(_L3, _L4); - } - else - { - A = _mm_movelh_ps(_L1, _L2); - B = _mm_movehl_ps(_L2, _L1); - C = _mm_movelh_ps(_L3, _L4); - D = _mm_movehl_ps(_L4, _L3); - } - - __m128 iA, iB, iC, iD, // partial inverse of the sub-matrices - DC, AB; - __m128 dA, dB, dC, dD; // determinant of the sub-matrices - __m128 det, d, d1, d2; - __m128 rd; // reciprocal of the determinant - - // AB = A# * B - AB = _mm_mul_ps(_mm_shuffle_ps(A,A,0x0F), B); - AB = _mm_sub_ps(AB,_mm_mul_ps(_mm_shuffle_ps(A,A,0xA5), _mm_shuffle_ps(B,B,0x4E))); - // DC = D# * C - DC = _mm_mul_ps(_mm_shuffle_ps(D,D,0x0F), C); - DC = _mm_sub_ps(DC,_mm_mul_ps(_mm_shuffle_ps(D,D,0xA5), _mm_shuffle_ps(C,C,0x4E))); - - // dA = |A| - dA = _mm_mul_ps(_mm_shuffle_ps(A, A, 0x5F),A); - dA = _mm_sub_ss(dA, _mm_movehl_ps(dA,dA)); - // dB = |B| - dB = _mm_mul_ps(_mm_shuffle_ps(B, B, 0x5F),B); - dB = _mm_sub_ss(dB, _mm_movehl_ps(dB,dB)); - - // dC = |C| - dC = _mm_mul_ps(_mm_shuffle_ps(C, C, 0x5F),C); - dC = _mm_sub_ss(dC, _mm_movehl_ps(dC,dC)); - // dD = |D| - dD = _mm_mul_ps(_mm_shuffle_ps(D, D, 0x5F),D); - dD = _mm_sub_ss(dD, _mm_movehl_ps(dD,dD)); - - // d = trace(AB*DC) = trace(A#*B*D#*C) - d = _mm_mul_ps(_mm_shuffle_ps(DC,DC,0xD8),AB); - - // iD = C*A#*B - iD = _mm_mul_ps(_mm_shuffle_ps(C,C,0xA0), _mm_movelh_ps(AB,AB)); - iD = _mm_add_ps(iD,_mm_mul_ps(_mm_shuffle_ps(C,C,0xF5), _mm_movehl_ps(AB,AB))); - // iA = B*D#*C - iA = _mm_mul_ps(_mm_shuffle_ps(B,B,0xA0), _mm_movelh_ps(DC,DC)); - iA = _mm_add_ps(iA,_mm_mul_ps(_mm_shuffle_ps(B,B,0xF5), _mm_movehl_ps(DC,DC))); - - // d = trace(AB*DC) = trace(A#*B*D#*C) [continue] - d = _mm_add_ps(d, _mm_movehl_ps(d, d)); - d = _mm_add_ss(d, _mm_shuffle_ps(d, d, 1)); - d1 = _mm_mul_ss(dA,dD); - d2 = _mm_mul_ss(dB,dC); - - // iD = D*|A| - C*A#*B - iD = _mm_sub_ps(_mm_mul_ps(D,_mm_shuffle_ps(dA,dA,0)), iD); - - // iA = A*|D| - B*D#*C; - iA = _mm_sub_ps(_mm_mul_ps(A,_mm_shuffle_ps(dD,dD,0)), iA); - - // det = |A|*|D| + |B|*|C| - trace(A#*B*D#*C) - det = _mm_sub_ss(_mm_add_ss(d1,d2),d); - rd = _mm_div_ss(_mm_set_ss(1.0f), det); - -// #ifdef ZERO_SINGULAR -// rd = _mm_and_ps(_mm_cmpneq_ss(det,_mm_setzero_ps()), rd); -// #endif - - // iB = D * (A#B)# = D*B#*A - iB = _mm_mul_ps(D, _mm_shuffle_ps(AB,AB,0x33)); - iB = _mm_sub_ps(iB, _mm_mul_ps(_mm_shuffle_ps(D,D,0xB1), _mm_shuffle_ps(AB,AB,0x66))); - // iC = A * (D#C)# = A*C#*D - iC = _mm_mul_ps(A, _mm_shuffle_ps(DC,DC,0x33)); - iC = _mm_sub_ps(iC, _mm_mul_ps(_mm_shuffle_ps(A,A,0xB1), _mm_shuffle_ps(DC,DC,0x66))); - - rd = _mm_shuffle_ps(rd,rd,0); - rd = _mm_xor_ps(rd, p4f_sign_PNNP); - - // iB = C*|B| - D*B#*A - iB = _mm_sub_ps(_mm_mul_ps(C,_mm_shuffle_ps(dB,dB,0)), iB); - - // iC = B*|C| - A*C#*D; - iC = _mm_sub_ps(_mm_mul_ps(B,_mm_shuffle_ps(dC,dC,0)), iC); - - // iX = iX / det - iA = _mm_mul_ps(rd,iA); - iB = _mm_mul_ps(rd,iB); - iC = _mm_mul_ps(rd,iC); - iD = _mm_mul_ps(rd,iD); - - Index res_stride = result.outerStride(); - float* res = result.data(); - pstoret(res+0, _mm_shuffle_ps(iA,iB,0x77)); - pstoret(res+res_stride, _mm_shuffle_ps(iA,iB,0x22)); - pstoret(res+2*res_stride, _mm_shuffle_ps(iC,iD,0x77)); - pstoret(res+3*res_stride, _mm_shuffle_ps(iC,iD,0x22)); - } - -}; - -template -struct compute_inverse_size4 -{ - enum { - MatrixAlignment = traits::Alignment, - ResultAlignment = traits::Alignment, - StorageOrdersMatch = (MatrixType::Flags&RowMajorBit) == (ResultType::Flags&RowMajorBit) - }; - typedef typename conditional<(MatrixType::Flags&LinearAccessBit),MatrixType const &,typename MatrixType::PlainObject>::type ActualMatrixType; - - static void run(const MatrixType& mat, ResultType& result) - { - ActualMatrixType matrix(mat); - const __m128d _Sign_NP = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); - const __m128d _Sign_PN = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - - // The inverse is calculated using "Divide and Conquer" technique. The - // original matrix is divide into four 2x2 sub-matrices. Since each - // register of the matrix holds two elements, the smaller matrices are - // consisted of two registers. Hence we get a better locality of the - // calculations. - - // the four sub-matrices - __m128d A1, A2, B1, B2, C1, C2, D1, D2; - - if(StorageOrdersMatch) - { - A1 = matrix.template packet( 0); B1 = matrix.template packet( 2); - A2 = matrix.template packet( 4); B2 = matrix.template packet( 6); - C1 = matrix.template packet( 8); D1 = matrix.template packet(10); - C2 = matrix.template packet(12); D2 = matrix.template packet(14); - } - else - { - __m128d tmp; - A1 = matrix.template packet( 0); C1 = matrix.template packet( 2); - A2 = matrix.template packet( 4); C2 = matrix.template packet( 6); - tmp = A1; - A1 = _mm_unpacklo_pd(A1,A2); - A2 = _mm_unpackhi_pd(tmp,A2); - tmp = C1; - C1 = _mm_unpacklo_pd(C1,C2); - C2 = _mm_unpackhi_pd(tmp,C2); - - B1 = matrix.template packet( 8); D1 = matrix.template packet(10); - B2 = matrix.template packet(12); D2 = matrix.template packet(14); - tmp = B1; - B1 = _mm_unpacklo_pd(B1,B2); - B2 = _mm_unpackhi_pd(tmp,B2); - tmp = D1; - D1 = _mm_unpacklo_pd(D1,D2); - D2 = _mm_unpackhi_pd(tmp,D2); - } - - __m128d iA1, iA2, iB1, iB2, iC1, iC2, iD1, iD2, // partial invese of the sub-matrices - DC1, DC2, AB1, AB2; - __m128d dA, dB, dC, dD; // determinant of the sub-matrices - __m128d det, d1, d2, rd; - - // dA = |A| - dA = _mm_shuffle_pd(A2, A2, 1); - dA = _mm_mul_pd(A1, dA); - dA = _mm_sub_sd(dA, _mm_shuffle_pd(dA,dA,3)); - // dB = |B| - dB = _mm_shuffle_pd(B2, B2, 1); - dB = _mm_mul_pd(B1, dB); - dB = _mm_sub_sd(dB, _mm_shuffle_pd(dB,dB,3)); - - // AB = A# * B - AB1 = _mm_mul_pd(B1, _mm_shuffle_pd(A2,A2,3)); - AB2 = _mm_mul_pd(B2, _mm_shuffle_pd(A1,A1,0)); - AB1 = _mm_sub_pd(AB1, _mm_mul_pd(B2, _mm_shuffle_pd(A1,A1,3))); - AB2 = _mm_sub_pd(AB2, _mm_mul_pd(B1, _mm_shuffle_pd(A2,A2,0))); - - // dC = |C| - dC = _mm_shuffle_pd(C2, C2, 1); - dC = _mm_mul_pd(C1, dC); - dC = _mm_sub_sd(dC, _mm_shuffle_pd(dC,dC,3)); - // dD = |D| - dD = _mm_shuffle_pd(D2, D2, 1); - dD = _mm_mul_pd(D1, dD); - dD = _mm_sub_sd(dD, _mm_shuffle_pd(dD,dD,3)); - - // DC = D# * C - DC1 = _mm_mul_pd(C1, _mm_shuffle_pd(D2,D2,3)); - DC2 = _mm_mul_pd(C2, _mm_shuffle_pd(D1,D1,0)); - DC1 = _mm_sub_pd(DC1, _mm_mul_pd(C2, _mm_shuffle_pd(D1,D1,3))); - DC2 = _mm_sub_pd(DC2, _mm_mul_pd(C1, _mm_shuffle_pd(D2,D2,0))); - - // rd = trace(AB*DC) = trace(A#*B*D#*C) - d1 = _mm_mul_pd(AB1, _mm_shuffle_pd(DC1, DC2, 0)); - d2 = _mm_mul_pd(AB2, _mm_shuffle_pd(DC1, DC2, 3)); - rd = _mm_add_pd(d1, d2); - rd = _mm_add_sd(rd, _mm_shuffle_pd(rd, rd,3)); - - // iD = C*A#*B - iD1 = _mm_mul_pd(AB1, _mm_shuffle_pd(C1,C1,0)); - iD2 = _mm_mul_pd(AB1, _mm_shuffle_pd(C2,C2,0)); - iD1 = _mm_add_pd(iD1, _mm_mul_pd(AB2, _mm_shuffle_pd(C1,C1,3))); - iD2 = _mm_add_pd(iD2, _mm_mul_pd(AB2, _mm_shuffle_pd(C2,C2,3))); - - // iA = B*D#*C - iA1 = _mm_mul_pd(DC1, _mm_shuffle_pd(B1,B1,0)); - iA2 = _mm_mul_pd(DC1, _mm_shuffle_pd(B2,B2,0)); - iA1 = _mm_add_pd(iA1, _mm_mul_pd(DC2, _mm_shuffle_pd(B1,B1,3))); - iA2 = _mm_add_pd(iA2, _mm_mul_pd(DC2, _mm_shuffle_pd(B2,B2,3))); - - // iD = D*|A| - C*A#*B - dA = _mm_shuffle_pd(dA,dA,0); - iD1 = _mm_sub_pd(_mm_mul_pd(D1, dA), iD1); - iD2 = _mm_sub_pd(_mm_mul_pd(D2, dA), iD2); - - // iA = A*|D| - B*D#*C; - dD = _mm_shuffle_pd(dD,dD,0); - iA1 = _mm_sub_pd(_mm_mul_pd(A1, dD), iA1); - iA2 = _mm_sub_pd(_mm_mul_pd(A2, dD), iA2); - - d1 = _mm_mul_sd(dA, dD); - d2 = _mm_mul_sd(dB, dC); - - // iB = D * (A#B)# = D*B#*A - iB1 = _mm_mul_pd(D1, _mm_shuffle_pd(AB2,AB1,1)); - iB2 = _mm_mul_pd(D2, _mm_shuffle_pd(AB2,AB1,1)); - iB1 = _mm_sub_pd(iB1, _mm_mul_pd(_mm_shuffle_pd(D1,D1,1), _mm_shuffle_pd(AB2,AB1,2))); - iB2 = _mm_sub_pd(iB2, _mm_mul_pd(_mm_shuffle_pd(D2,D2,1), _mm_shuffle_pd(AB2,AB1,2))); - - // det = |A|*|D| + |B|*|C| - trace(A#*B*D#*C) - det = _mm_add_sd(d1, d2); - det = _mm_sub_sd(det, rd); - - // iC = A * (D#C)# = A*C#*D - iC1 = _mm_mul_pd(A1, _mm_shuffle_pd(DC2,DC1,1)); - iC2 = _mm_mul_pd(A2, _mm_shuffle_pd(DC2,DC1,1)); - iC1 = _mm_sub_pd(iC1, _mm_mul_pd(_mm_shuffle_pd(A1,A1,1), _mm_shuffle_pd(DC2,DC1,2))); - iC2 = _mm_sub_pd(iC2, _mm_mul_pd(_mm_shuffle_pd(A2,A2,1), _mm_shuffle_pd(DC2,DC1,2))); - - rd = _mm_div_sd(_mm_set_sd(1.0), det); -// #ifdef ZERO_SINGULAR -// rd = _mm_and_pd(_mm_cmpneq_sd(det,_mm_setzero_pd()), rd); -// #endif - rd = _mm_shuffle_pd(rd,rd,0); - - // iB = C*|B| - D*B#*A - dB = _mm_shuffle_pd(dB,dB,0); - iB1 = _mm_sub_pd(_mm_mul_pd(C1, dB), iB1); - iB2 = _mm_sub_pd(_mm_mul_pd(C2, dB), iB2); - - d1 = _mm_xor_pd(rd, _Sign_PN); - d2 = _mm_xor_pd(rd, _Sign_NP); - - // iC = B*|C| - A*C#*D; - dC = _mm_shuffle_pd(dC,dC,0); - iC1 = _mm_sub_pd(_mm_mul_pd(B1, dC), iC1); - iC2 = _mm_sub_pd(_mm_mul_pd(B2, dC), iC2); - - Index res_stride = result.outerStride(); - double* res = result.data(); - pstoret(res+0, _mm_mul_pd(_mm_shuffle_pd(iA2, iA1, 3), d1)); - pstoret(res+res_stride, _mm_mul_pd(_mm_shuffle_pd(iA2, iA1, 0), d2)); - pstoret(res+2, _mm_mul_pd(_mm_shuffle_pd(iB2, iB1, 3), d1)); - pstoret(res+res_stride+2, _mm_mul_pd(_mm_shuffle_pd(iB2, iB1, 0), d2)); - pstoret(res+2*res_stride, _mm_mul_pd(_mm_shuffle_pd(iC2, iC1, 3), d1)); - pstoret(res+3*res_stride, _mm_mul_pd(_mm_shuffle_pd(iC2, iC1, 0), d2)); - pstoret(res+2*res_stride+2,_mm_mul_pd(_mm_shuffle_pd(iD2, iD1, 3), d1)); - pstoret(res+3*res_stride+2,_mm_mul_pd(_mm_shuffle_pd(iD2, iD1, 0), d2)); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_INVERSE_SSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/MetisSupport/MetisSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/MetisSupport/MetisSupport.h deleted file mode 100644 index 4c15304ad63..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/MetisSupport/MetisSupport.h +++ /dev/null @@ -1,137 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -#ifndef METIS_SUPPORT_H -#define METIS_SUPPORT_H - -namespace Eigen { -/** - * Get the fill-reducing ordering from the METIS package - * - * If A is the original matrix and Ap is the permuted matrix, - * the fill-reducing permutation is defined as follows : - * Row (column) i of A is the matperm(i) row (column) of Ap. - * WARNING: As computed by METIS, this corresponds to the vector iperm (instead of perm) - */ -template -class MetisOrdering -{ -public: - typedef PermutationMatrix PermutationType; - typedef Matrix IndexVector; - - template - void get_symmetrized_graph(const MatrixType& A) - { - Index m = A.cols(); - eigen_assert((A.rows() == A.cols()) && "ONLY FOR SQUARED MATRICES"); - // Get the transpose of the input matrix - MatrixType At = A.transpose(); - // Get the number of nonzeros elements in each row/col of At+A - Index TotNz = 0; - IndexVector visited(m); - visited.setConstant(-1); - for (StorageIndex j = 0; j < m; j++) - { - // Compute the union structure of of A(j,:) and At(j,:) - visited(j) = j; // Do not include the diagonal element - // Get the nonzeros in row/column j of A - for (typename MatrixType::InnerIterator it(A, j); it; ++it) - { - Index idx = it.index(); // Get the row index (for column major) or column index (for row major) - if (visited(idx) != j ) - { - visited(idx) = j; - ++TotNz; - } - } - //Get the nonzeros in row/column j of At - for (typename MatrixType::InnerIterator it(At, j); it; ++it) - { - Index idx = it.index(); - if(visited(idx) != j) - { - visited(idx) = j; - ++TotNz; - } - } - } - // Reserve place for A + At - m_indexPtr.resize(m+1); - m_innerIndices.resize(TotNz); - - // Now compute the real adjacency list of each column/row - visited.setConstant(-1); - StorageIndex CurNz = 0; - for (StorageIndex j = 0; j < m; j++) - { - m_indexPtr(j) = CurNz; - - visited(j) = j; // Do not include the diagonal element - // Add the pattern of row/column j of A to A+At - for (typename MatrixType::InnerIterator it(A,j); it; ++it) - { - StorageIndex idx = it.index(); // Get the row index (for column major) or column index (for row major) - if (visited(idx) != j ) - { - visited(idx) = j; - m_innerIndices(CurNz) = idx; - CurNz++; - } - } - //Add the pattern of row/column j of At to A+At - for (typename MatrixType::InnerIterator it(At, j); it; ++it) - { - StorageIndex idx = it.index(); - if(visited(idx) != j) - { - visited(idx) = j; - m_innerIndices(CurNz) = idx; - ++CurNz; - } - } - } - m_indexPtr(m) = CurNz; - } - - template - void operator() (const MatrixType& A, PermutationType& matperm) - { - StorageIndex m = internal::convert_index(A.cols()); // must be StorageIndex, because it is passed by address to METIS - IndexVector perm(m),iperm(m); - // First, symmetrize the matrix graph. - get_symmetrized_graph(A); - int output_error; - - // Call the fill-reducing routine from METIS - output_error = METIS_NodeND(&m, m_indexPtr.data(), m_innerIndices.data(), NULL, NULL, perm.data(), iperm.data()); - - if(output_error != METIS_OK) - { - //FIXME The ordering interface should define a class of possible errors - std::cerr << "ERROR WHILE CALLING THE METIS PACKAGE \n"; - return; - } - - // Get the fill-reducing permutation - //NOTE: If Ap is the permuted matrix then perm and iperm vectors are defined as follows - // Row (column) i of Ap is the perm(i) row(column) of A, and row (column) i of A is the iperm(i) row(column) of Ap - - matperm.resize(m); - for (int j = 0; j < m; j++) - matperm.indices()(iperm(j)) = j; - - } - - protected: - IndexVector m_indexPtr; // Pointer to the adjacenccy list of each row/column - IndexVector m_innerIndices; // Adjacency list -}; - -}// end namespace eigen -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Amd.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Amd.h deleted file mode 100644 index f91ecb24efc..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Amd.h +++ /dev/null @@ -1,445 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud - -/* - -NOTE: this routine has been adapted from the CSparse library: - -Copyright (c) 2006, Timothy A. Davis. -http://www.suitesparse.com - -CSparse is free software; you can redistribute it and/or -modify it under the terms of the GNU Lesser General Public -License as published by the Free Software Foundation; either -version 2.1 of the License, or (at your option) any later version. - -CSparse is distributed in the hope that it will be useful, -but WITHOUT ANY WARRANTY; without even the implied warranty of -MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU -Lesser General Public License for more details. - -You should have received a copy of the GNU Lesser General Public -License along with this Module; if not, write to the Free Software -Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA - -*/ - -#include "../Core/util/NonMPL2.h" - -#ifndef EIGEN_SPARSE_AMD_H -#define EIGEN_SPARSE_AMD_H - -namespace Eigen { - -namespace internal { - -template inline T amd_flip(const T& i) { return -i-2; } -template inline T amd_unflip(const T& i) { return i<0 ? amd_flip(i) : i; } -template inline bool amd_marked(const T0* w, const T1& j) { return w[j]<0; } -template inline void amd_mark(const T0* w, const T1& j) { return w[j] = amd_flip(w[j]); } - -/* clear w */ -template -static StorageIndex cs_wclear (StorageIndex mark, StorageIndex lemax, StorageIndex *w, StorageIndex n) -{ - StorageIndex k; - if(mark < 2 || (mark + lemax < 0)) - { - for(k = 0; k < n; k++) - if(w[k] != 0) - w[k] = 1; - mark = 2; - } - return (mark); /* at this point, w[0..n-1] < mark holds */ -} - -/* depth-first search and postorder of a tree rooted at node j */ -template -StorageIndex cs_tdfs(StorageIndex j, StorageIndex k, StorageIndex *head, const StorageIndex *next, StorageIndex *post, StorageIndex *stack) -{ - StorageIndex i, p, top = 0; - if(!head || !next || !post || !stack) return (-1); /* check inputs */ - stack[0] = j; /* place j on the stack */ - while (top >= 0) /* while (stack is not empty) */ - { - p = stack[top]; /* p = top of stack */ - i = head[p]; /* i = youngest child of p */ - if(i == -1) - { - top--; /* p has no unordered children left */ - post[k++] = p; /* node p is the kth postordered node */ - } - else - { - head[p] = next[i]; /* remove i from children of p */ - stack[++top] = i; /* start dfs on child node i */ - } - } - return k; -} - - -/** \internal - * \ingroup OrderingMethods_Module - * Approximate minimum degree ordering algorithm. - * - * \param[in] C the input selfadjoint matrix stored in compressed column major format. - * \param[out] perm the permutation P reducing the fill-in of the input matrix \a C - * - * Note that the input matrix \a C must be complete, that is both the upper and lower parts have to be stored, as well as the diagonal entries. - * On exit the values of C are destroyed */ -template -void minimum_degree_ordering(SparseMatrix& C, PermutationMatrix& perm) -{ - using std::sqrt; - - StorageIndex d, dk, dext, lemax = 0, e, elenk, eln, i, j, k, k1, - k2, k3, jlast, ln, dense, nzmax, mindeg = 0, nvi, nvj, nvk, mark, wnvi, - ok, nel = 0, p, p1, p2, p3, p4, pj, pk, pk1, pk2, pn, q, t, h; - - StorageIndex n = StorageIndex(C.cols()); - dense = std::max (16, StorageIndex(10 * sqrt(double(n)))); /* find dense threshold */ - dense = (std::min)(n-2, dense); - - StorageIndex cnz = StorageIndex(C.nonZeros()); - perm.resize(n+1); - t = cnz + cnz/5 + 2*n; /* add elbow room to C */ - C.resizeNonZeros(t); - - // get workspace - ei_declare_aligned_stack_constructed_variable(StorageIndex,W,8*(n+1),0); - StorageIndex* len = W; - StorageIndex* nv = W + (n+1); - StorageIndex* next = W + 2*(n+1); - StorageIndex* head = W + 3*(n+1); - StorageIndex* elen = W + 4*(n+1); - StorageIndex* degree = W + 5*(n+1); - StorageIndex* w = W + 6*(n+1); - StorageIndex* hhead = W + 7*(n+1); - StorageIndex* last = perm.indices().data(); /* use P as workspace for last */ - - /* --- Initialize quotient graph ---------------------------------------- */ - StorageIndex* Cp = C.outerIndexPtr(); - StorageIndex* Ci = C.innerIndexPtr(); - for(k = 0; k < n; k++) - len[k] = Cp[k+1] - Cp[k]; - len[n] = 0; - nzmax = t; - - for(i = 0; i <= n; i++) - { - head[i] = -1; // degree list i is empty - last[i] = -1; - next[i] = -1; - hhead[i] = -1; // hash list i is empty - nv[i] = 1; // node i is just one node - w[i] = 1; // node i is alive - elen[i] = 0; // Ek of node i is empty - degree[i] = len[i]; // degree of node i - } - mark = internal::cs_wclear(0, 0, w, n); /* clear w */ - - /* --- Initialize degree lists ------------------------------------------ */ - for(i = 0; i < n; i++) - { - bool has_diag = false; - for(p = Cp[i]; p dense || !has_diag) /* node i is dense or has no structural diagonal element */ - { - nv[i] = 0; /* absorb i into element n */ - elen[i] = -1; /* node i is dead */ - nel++; - Cp[i] = amd_flip (n); - nv[n]++; - } - else - { - if(head[d] != -1) last[head[d]] = i; - next[i] = head[d]; /* put node i in degree list d */ - head[d] = i; - } - } - - elen[n] = -2; /* n is a dead element */ - Cp[n] = -1; /* n is a root of assembly tree */ - w[n] = 0; /* n is a dead element */ - - while (nel < n) /* while (selecting pivots) do */ - { - /* --- Select node of minimum approximate degree -------------------- */ - for(k = -1; mindeg < n && (k = head[mindeg]) == -1; mindeg++) {} - if(next[k] != -1) last[next[k]] = -1; - head[mindeg] = next[k]; /* remove k from degree list */ - elenk = elen[k]; /* elenk = |Ek| */ - nvk = nv[k]; /* # of nodes k represents */ - nel += nvk; /* nv[k] nodes of A eliminated */ - - /* --- Garbage collection ------------------------------------------- */ - if(elenk > 0 && cnz + mindeg >= nzmax) - { - for(j = 0; j < n; j++) - { - if((p = Cp[j]) >= 0) /* j is a live node or element */ - { - Cp[j] = Ci[p]; /* save first entry of object */ - Ci[p] = amd_flip (j); /* first entry is now amd_flip(j) */ - } - } - for(q = 0, p = 0; p < cnz; ) /* scan all of memory */ - { - if((j = amd_flip (Ci[p++])) >= 0) /* found object j */ - { - Ci[q] = Cp[j]; /* restore first entry of object */ - Cp[j] = q++; /* new pointer to object j */ - for(k3 = 0; k3 < len[j]-1; k3++) Ci[q++] = Ci[p++]; - } - } - cnz = q; /* Ci[cnz...nzmax-1] now free */ - } - - /* --- Construct new element ---------------------------------------- */ - dk = 0; - nv[k] = -nvk; /* flag k as in Lk */ - p = Cp[k]; - pk1 = (elenk == 0) ? p : cnz; /* do in place if elen[k] == 0 */ - pk2 = pk1; - for(k1 = 1; k1 <= elenk + 1; k1++) - { - if(k1 > elenk) - { - e = k; /* search the nodes in k */ - pj = p; /* list of nodes starts at Ci[pj]*/ - ln = len[k] - elenk; /* length of list of nodes in k */ - } - else - { - e = Ci[p++]; /* search the nodes in e */ - pj = Cp[e]; - ln = len[e]; /* length of list of nodes in e */ - } - for(k2 = 1; k2 <= ln; k2++) - { - i = Ci[pj++]; - if((nvi = nv[i]) <= 0) continue; /* node i dead, or seen */ - dk += nvi; /* degree[Lk] += size of node i */ - nv[i] = -nvi; /* negate nv[i] to denote i in Lk*/ - Ci[pk2++] = i; /* place i in Lk */ - if(next[i] != -1) last[next[i]] = last[i]; - if(last[i] != -1) /* remove i from degree list */ - { - next[last[i]] = next[i]; - } - else - { - head[degree[i]] = next[i]; - } - } - if(e != k) - { - Cp[e] = amd_flip (k); /* absorb e into k */ - w[e] = 0; /* e is now a dead element */ - } - } - if(elenk != 0) cnz = pk2; /* Ci[cnz...nzmax] is free */ - degree[k] = dk; /* external degree of k - |Lk\i| */ - Cp[k] = pk1; /* element k is in Ci[pk1..pk2-1] */ - len[k] = pk2 - pk1; - elen[k] = -2; /* k is now an element */ - - /* --- Find set differences ----------------------------------------- */ - mark = internal::cs_wclear(mark, lemax, w, n); /* clear w if necessary */ - for(pk = pk1; pk < pk2; pk++) /* scan 1: find |Le\Lk| */ - { - i = Ci[pk]; - if((eln = elen[i]) <= 0) continue;/* skip if elen[i] empty */ - nvi = -nv[i]; /* nv[i] was negated */ - wnvi = mark - nvi; - for(p = Cp[i]; p <= Cp[i] + eln - 1; p++) /* scan Ei */ - { - e = Ci[p]; - if(w[e] >= mark) - { - w[e] -= nvi; /* decrement |Le\Lk| */ - } - else if(w[e] != 0) /* ensure e is a live element */ - { - w[e] = degree[e] + wnvi; /* 1st time e seen in scan 1 */ - } - } - } - - /* --- Degree update ------------------------------------------------ */ - for(pk = pk1; pk < pk2; pk++) /* scan2: degree update */ - { - i = Ci[pk]; /* consider node i in Lk */ - p1 = Cp[i]; - p2 = p1 + elen[i] - 1; - pn = p1; - for(h = 0, d = 0, p = p1; p <= p2; p++) /* scan Ei */ - { - e = Ci[p]; - if(w[e] != 0) /* e is an unabsorbed element */ - { - dext = w[e] - mark; /* dext = |Le\Lk| */ - if(dext > 0) - { - d += dext; /* sum up the set differences */ - Ci[pn++] = e; /* keep e in Ei */ - h += e; /* compute the hash of node i */ - } - else - { - Cp[e] = amd_flip (k); /* aggressive absorb. e->k */ - w[e] = 0; /* e is a dead element */ - } - } - } - elen[i] = pn - p1 + 1; /* elen[i] = |Ei| */ - p3 = pn; - p4 = p1 + len[i]; - for(p = p2 + 1; p < p4; p++) /* prune edges in Ai */ - { - j = Ci[p]; - if((nvj = nv[j]) <= 0) continue; /* node j dead or in Lk */ - d += nvj; /* degree(i) += |j| */ - Ci[pn++] = j; /* place j in node list of i */ - h += j; /* compute hash for node i */ - } - if(d == 0) /* check for mass elimination */ - { - Cp[i] = amd_flip (k); /* absorb i into k */ - nvi = -nv[i]; - dk -= nvi; /* |Lk| -= |i| */ - nvk += nvi; /* |k| += nv[i] */ - nel += nvi; - nv[i] = 0; - elen[i] = -1; /* node i is dead */ - } - else - { - degree[i] = std::min (degree[i], d); /* update degree(i) */ - Ci[pn] = Ci[p3]; /* move first node to end */ - Ci[p3] = Ci[p1]; /* move 1st el. to end of Ei */ - Ci[p1] = k; /* add k as 1st element in of Ei */ - len[i] = pn - p1 + 1; /* new len of adj. list of node i */ - h %= n; /* finalize hash of i */ - next[i] = hhead[h]; /* place i in hash bucket */ - hhead[h] = i; - last[i] = h; /* save hash of i in last[i] */ - } - } /* scan2 is done */ - degree[k] = dk; /* finalize |Lk| */ - lemax = std::max(lemax, dk); - mark = internal::cs_wclear(mark+lemax, lemax, w, n); /* clear w */ - - /* --- Supernode detection ------------------------------------------ */ - for(pk = pk1; pk < pk2; pk++) - { - i = Ci[pk]; - if(nv[i] >= 0) continue; /* skip if i is dead */ - h = last[i]; /* scan hash bucket of node i */ - i = hhead[h]; - hhead[h] = -1; /* hash bucket will be empty */ - for(; i != -1 && next[i] != -1; i = next[i], mark++) - { - ln = len[i]; - eln = elen[i]; - for(p = Cp[i]+1; p <= Cp[i] + ln-1; p++) w[Ci[p]] = mark; - jlast = i; - for(j = next[i]; j != -1; ) /* compare i with all j */ - { - ok = (len[j] == ln) && (elen[j] == eln); - for(p = Cp[j] + 1; ok && p <= Cp[j] + ln - 1; p++) - { - if(w[Ci[p]] != mark) ok = 0; /* compare i and j*/ - } - if(ok) /* i and j are identical */ - { - Cp[j] = amd_flip (i); /* absorb j into i */ - nv[i] += nv[j]; - nv[j] = 0; - elen[j] = -1; /* node j is dead */ - j = next[j]; /* delete j from hash bucket */ - next[jlast] = j; - } - else - { - jlast = j; /* j and i are different */ - j = next[j]; - } - } - } - } - - /* --- Finalize new element------------------------------------------ */ - for(p = pk1, pk = pk1; pk < pk2; pk++) /* finalize Lk */ - { - i = Ci[pk]; - if((nvi = -nv[i]) <= 0) continue;/* skip if i is dead */ - nv[i] = nvi; /* restore nv[i] */ - d = degree[i] + dk - nvi; /* compute external degree(i) */ - d = std::min (d, n - nel - nvi); - if(head[d] != -1) last[head[d]] = i; - next[i] = head[d]; /* put i back in degree list */ - last[i] = -1; - head[d] = i; - mindeg = std::min (mindeg, d); /* find new minimum degree */ - degree[i] = d; - Ci[p++] = i; /* place i in Lk */ - } - nv[k] = nvk; /* # nodes absorbed into k */ - if((len[k] = p-pk1) == 0) /* length of adj list of element k*/ - { - Cp[k] = -1; /* k is a root of the tree */ - w[k] = 0; /* k is now a dead element */ - } - if(elenk != 0) cnz = p; /* free unused space in Lk */ - } - - /* --- Postordering ----------------------------------------------------- */ - for(i = 0; i < n; i++) Cp[i] = amd_flip (Cp[i]);/* fix assembly tree */ - for(j = 0; j <= n; j++) head[j] = -1; - for(j = n; j >= 0; j--) /* place unordered nodes in lists */ - { - if(nv[j] > 0) continue; /* skip if j is an element */ - next[j] = head[Cp[j]]; /* place j in list of its parent */ - head[Cp[j]] = j; - } - for(e = n; e >= 0; e--) /* place elements in lists */ - { - if(nv[e] <= 0) continue; /* skip unless e is an element */ - if(Cp[e] != -1) - { - next[e] = head[Cp[e]]; /* place e in list of its parent */ - head[Cp[e]] = e; - } - } - for(k = 0, i = 0; i <= n; i++) /* postorder the assembly tree */ - { - if(Cp[i] == -1) k = internal::cs_tdfs(i, k, head, next, perm.indices().data(), w); - } - - perm.indices().conservativeResize(n); -} - -} // namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_AMD_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Eigen_Colamd.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Eigen_Colamd.h deleted file mode 100644 index da85b4d6ea2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Eigen_Colamd.h +++ /dev/null @@ -1,1843 +0,0 @@ -// // This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Desire Nuentsa Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is modified from the colamd/symamd library. The copyright is below - -// The authors of the code itself are Stefan I. Larimore and Timothy A. -// Davis (davis@cise.ufl.edu), University of Florida. The algorithm was -// developed in collaboration with John Gilbert, Xerox PARC, and Esmond -// Ng, Oak Ridge National Laboratory. -// -// Date: -// -// September 8, 2003. Version 2.3. -// -// Acknowledgements: -// -// This work was supported by the National Science Foundation, under -// grants DMS-9504974 and DMS-9803599. -// -// Notice: -// -// Copyright (c) 1998-2003 by the University of Florida. -// All Rights Reserved. -// -// THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY -// EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. -// -// Permission is hereby granted to use, copy, modify, and/or distribute -// this program, provided that the Copyright, this License, and the -// Availability of the original version is retained on all copies and made -// accessible to the end-user of any code or package that includes COLAMD -// or any modified version of COLAMD. -// -// Availability: -// -// The colamd/symamd library is available at -// -// http://www.suitesparse.com - - -#ifndef EIGEN_COLAMD_H -#define EIGEN_COLAMD_H - -namespace internal { -/* Ensure that debugging is turned off: */ -#ifndef COLAMD_NDEBUG -#define COLAMD_NDEBUG -#endif /* NDEBUG */ -/* ========================================================================== */ -/* === Knob and statistics definitions ====================================== */ -/* ========================================================================== */ - -/* size of the knobs [ ] array. Only knobs [0..1] are currently used. */ -#define COLAMD_KNOBS 20 - -/* number of output statistics. Only stats [0..6] are currently used. */ -#define COLAMD_STATS 20 - -/* knobs [0] and stats [0]: dense row knob and output statistic. */ -#define COLAMD_DENSE_ROW 0 - -/* knobs [1] and stats [1]: dense column knob and output statistic. */ -#define COLAMD_DENSE_COL 1 - -/* stats [2]: memory defragmentation count output statistic */ -#define COLAMD_DEFRAG_COUNT 2 - -/* stats [3]: colamd status: zero OK, > 0 warning or notice, < 0 error */ -#define COLAMD_STATUS 3 - -/* stats [4..6]: error info, or info on jumbled columns */ -#define COLAMD_INFO1 4 -#define COLAMD_INFO2 5 -#define COLAMD_INFO3 6 - -/* error codes returned in stats [3]: */ -#define COLAMD_OK (0) -#define COLAMD_OK_BUT_JUMBLED (1) -#define COLAMD_ERROR_A_not_present (-1) -#define COLAMD_ERROR_p_not_present (-2) -#define COLAMD_ERROR_nrow_negative (-3) -#define COLAMD_ERROR_ncol_negative (-4) -#define COLAMD_ERROR_nnz_negative (-5) -#define COLAMD_ERROR_p0_nonzero (-6) -#define COLAMD_ERROR_A_too_small (-7) -#define COLAMD_ERROR_col_length_negative (-8) -#define COLAMD_ERROR_row_index_out_of_bounds (-9) -#define COLAMD_ERROR_out_of_memory (-10) -#define COLAMD_ERROR_internal_error (-999) - -/* ========================================================================== */ -/* === Definitions ========================================================== */ -/* ========================================================================== */ - -#define ONES_COMPLEMENT(r) (-(r)-1) - -/* -------------------------------------------------------------------------- */ - -#define COLAMD_EMPTY (-1) - -/* Row and column status */ -#define ALIVE (0) -#define DEAD (-1) - -/* Column status */ -#define DEAD_PRINCIPAL (-1) -#define DEAD_NON_PRINCIPAL (-2) - -/* Macros for row and column status update and checking. */ -#define ROW_IS_DEAD(r) ROW_IS_MARKED_DEAD (Row[r].shared2.mark) -#define ROW_IS_MARKED_DEAD(row_mark) (row_mark < ALIVE) -#define ROW_IS_ALIVE(r) (Row [r].shared2.mark >= ALIVE) -#define COL_IS_DEAD(c) (Col [c].start < ALIVE) -#define COL_IS_ALIVE(c) (Col [c].start >= ALIVE) -#define COL_IS_DEAD_PRINCIPAL(c) (Col [c].start == DEAD_PRINCIPAL) -#define KILL_ROW(r) { Row [r].shared2.mark = DEAD ; } -#define KILL_PRINCIPAL_COL(c) { Col [c].start = DEAD_PRINCIPAL ; } -#define KILL_NON_PRINCIPAL_COL(c) { Col [c].start = DEAD_NON_PRINCIPAL ; } - -/* ========================================================================== */ -/* === Colamd reporting mechanism =========================================== */ -/* ========================================================================== */ - -// == Row and Column structures == -template -struct colamd_col -{ - IndexType start ; /* index for A of first row in this column, or DEAD */ - /* if column is dead */ - IndexType length ; /* number of rows in this column */ - union - { - IndexType thickness ; /* number of original columns represented by this */ - /* col, if the column is alive */ - IndexType parent ; /* parent in parent tree super-column structure, if */ - /* the column is dead */ - } shared1 ; - union - { - IndexType score ; /* the score used to maintain heap, if col is alive */ - IndexType order ; /* pivot ordering of this column, if col is dead */ - } shared2 ; - union - { - IndexType headhash ; /* head of a hash bucket, if col is at the head of */ - /* a degree list */ - IndexType hash ; /* hash value, if col is not in a degree list */ - IndexType prev ; /* previous column in degree list, if col is in a */ - /* degree list (but not at the head of a degree list) */ - } shared3 ; - union - { - IndexType degree_next ; /* next column, if col is in a degree list */ - IndexType hash_next ; /* next column, if col is in a hash list */ - } shared4 ; - -}; - -template -struct Colamd_Row -{ - IndexType start ; /* index for A of first col in this row */ - IndexType length ; /* number of principal columns in this row */ - union - { - IndexType degree ; /* number of principal & non-principal columns in row */ - IndexType p ; /* used as a row pointer in init_rows_cols () */ - } shared1 ; - union - { - IndexType mark ; /* for computing set differences and marking dead rows*/ - IndexType first_column ;/* first column in row (used in garbage collection) */ - } shared2 ; - -}; - -/* ========================================================================== */ -/* === Colamd recommended memory size ======================================= */ -/* ========================================================================== */ - -/* - The recommended length Alen of the array A passed to colamd is given by - the COLAMD_RECOMMENDED (nnz, n_row, n_col) macro. It returns -1 if any - argument is negative. 2*nnz space is required for the row and column - indices of the matrix. colamd_c (n_col) + colamd_r (n_row) space is - required for the Col and Row arrays, respectively, which are internal to - colamd. An additional n_col space is the minimal amount of "elbow room", - and nnz/5 more space is recommended for run time efficiency. - - This macro is not needed when using symamd. - - Explicit typecast to IndexType added Sept. 23, 2002, COLAMD version 2.2, to avoid - gcc -pedantic warning messages. -*/ -template -inline IndexType colamd_c(IndexType n_col) -{ return IndexType( ((n_col) + 1) * sizeof (colamd_col) / sizeof (IndexType) ) ; } - -template -inline IndexType colamd_r(IndexType n_row) -{ return IndexType(((n_row) + 1) * sizeof (Colamd_Row) / sizeof (IndexType)); } - -// Prototypes of non-user callable routines -template -static IndexType init_rows_cols (IndexType n_row, IndexType n_col, Colamd_Row Row [], colamd_col col [], IndexType A [], IndexType p [], IndexType stats[COLAMD_STATS] ); - -template -static void init_scoring (IndexType n_row, IndexType n_col, Colamd_Row Row [], colamd_col Col [], IndexType A [], IndexType head [], double knobs[COLAMD_KNOBS], IndexType *p_n_row2, IndexType *p_n_col2, IndexType *p_max_deg); - -template -static IndexType find_ordering (IndexType n_row, IndexType n_col, IndexType Alen, Colamd_Row Row [], colamd_col Col [], IndexType A [], IndexType head [], IndexType n_col2, IndexType max_deg, IndexType pfree); - -template -static void order_children (IndexType n_col, colamd_col Col [], IndexType p []); - -template -static void detect_super_cols (colamd_col Col [], IndexType A [], IndexType head [], IndexType row_start, IndexType row_length ) ; - -template -static IndexType garbage_collection (IndexType n_row, IndexType n_col, Colamd_Row Row [], colamd_col Col [], IndexType A [], IndexType *pfree) ; - -template -static inline IndexType clear_mark (IndexType n_row, Colamd_Row Row [] ) ; - -/* === No debugging ========================================================= */ - -#define COLAMD_DEBUG0(params) ; -#define COLAMD_DEBUG1(params) ; -#define COLAMD_DEBUG2(params) ; -#define COLAMD_DEBUG3(params) ; -#define COLAMD_DEBUG4(params) ; - -#define COLAMD_ASSERT(expression) ((void) 0) - - -/** - * \brief Returns the recommended value of Alen - * - * Returns recommended value of Alen for use by colamd. - * Returns -1 if any input argument is negative. - * The use of this routine or macro is optional. - * Note that the macro uses its arguments more than once, - * so be careful for side effects, if you pass expressions as arguments to COLAMD_RECOMMENDED. - * - * \param nnz nonzeros in A - * \param n_row number of rows in A - * \param n_col number of columns in A - * \return recommended value of Alen for use by colamd - */ -template -inline IndexType colamd_recommended ( IndexType nnz, IndexType n_row, IndexType n_col) -{ - if ((nnz) < 0 || (n_row) < 0 || (n_col) < 0) - return (-1); - else - return (2 * (nnz) + colamd_c (n_col) + colamd_r (n_row) + (n_col) + ((nnz) / 5)); -} - -/** - * \brief set default parameters The use of this routine is optional. - * - * Colamd: rows with more than (knobs [COLAMD_DENSE_ROW] * n_col) - * entries are removed prior to ordering. Columns with more than - * (knobs [COLAMD_DENSE_COL] * n_row) entries are removed prior to - * ordering, and placed last in the output column ordering. - * - * COLAMD_DENSE_ROW and COLAMD_DENSE_COL are defined as 0 and 1, - * respectively, in colamd.h. Default values of these two knobs - * are both 0.5. Currently, only knobs [0] and knobs [1] are - * used, but future versions may use more knobs. If so, they will - * be properly set to their defaults by the future version of - * colamd_set_defaults, so that the code that calls colamd will - * not need to change, assuming that you either use - * colamd_set_defaults, or pass a (double *) NULL pointer as the - * knobs array to colamd or symamd. - * - * \param knobs parameter settings for colamd - */ - -static inline void colamd_set_defaults(double knobs[COLAMD_KNOBS]) -{ - /* === Local variables ================================================== */ - - int i ; - - if (!knobs) - { - return ; /* no knobs to initialize */ - } - for (i = 0 ; i < COLAMD_KNOBS ; i++) - { - knobs [i] = 0 ; - } - knobs [COLAMD_DENSE_ROW] = 0.5 ; /* ignore rows over 50% dense */ - knobs [COLAMD_DENSE_COL] = 0.5 ; /* ignore columns over 50% dense */ -} - -/** - * \brief Computes a column ordering using the column approximate minimum degree ordering - * - * Computes a column ordering (Q) of A such that P(AQ)=LU or - * (AQ)'AQ=LL' have less fill-in and require fewer floating point - * operations than factorizing the unpermuted matrix A or A'A, - * respectively. - * - * - * \param n_row number of rows in A - * \param n_col number of columns in A - * \param Alen, size of the array A - * \param A row indices of the matrix, of size ALen - * \param p column pointers of A, of size n_col+1 - * \param knobs parameter settings for colamd - * \param stats colamd output statistics and error codes - */ -template -static bool colamd(IndexType n_row, IndexType n_col, IndexType Alen, IndexType *A, IndexType *p, double knobs[COLAMD_KNOBS], IndexType stats[COLAMD_STATS]) -{ - /* === Local variables ================================================== */ - - IndexType i ; /* loop index */ - IndexType nnz ; /* nonzeros in A */ - IndexType Row_size ; /* size of Row [], in integers */ - IndexType Col_size ; /* size of Col [], in integers */ - IndexType need ; /* minimum required length of A */ - Colamd_Row *Row ; /* pointer into A of Row [0..n_row] array */ - colamd_col *Col ; /* pointer into A of Col [0..n_col] array */ - IndexType n_col2 ; /* number of non-dense, non-empty columns */ - IndexType n_row2 ; /* number of non-dense, non-empty rows */ - IndexType ngarbage ; /* number of garbage collections performed */ - IndexType max_deg ; /* maximum row degree */ - double default_knobs [COLAMD_KNOBS] ; /* default knobs array */ - - - /* === Check the input arguments ======================================== */ - - if (!stats) - { - COLAMD_DEBUG0 (("colamd: stats not present\n")) ; - return (false) ; - } - for (i = 0 ; i < COLAMD_STATS ; i++) - { - stats [i] = 0 ; - } - stats [COLAMD_STATUS] = COLAMD_OK ; - stats [COLAMD_INFO1] = -1 ; - stats [COLAMD_INFO2] = -1 ; - - if (!A) /* A is not present */ - { - stats [COLAMD_STATUS] = COLAMD_ERROR_A_not_present ; - COLAMD_DEBUG0 (("colamd: A not present\n")) ; - return (false) ; - } - - if (!p) /* p is not present */ - { - stats [COLAMD_STATUS] = COLAMD_ERROR_p_not_present ; - COLAMD_DEBUG0 (("colamd: p not present\n")) ; - return (false) ; - } - - if (n_row < 0) /* n_row must be >= 0 */ - { - stats [COLAMD_STATUS] = COLAMD_ERROR_nrow_negative ; - stats [COLAMD_INFO1] = n_row ; - COLAMD_DEBUG0 (("colamd: nrow negative %d\n", n_row)) ; - return (false) ; - } - - if (n_col < 0) /* n_col must be >= 0 */ - { - stats [COLAMD_STATUS] = COLAMD_ERROR_ncol_negative ; - stats [COLAMD_INFO1] = n_col ; - COLAMD_DEBUG0 (("colamd: ncol negative %d\n", n_col)) ; - return (false) ; - } - - nnz = p [n_col] ; - if (nnz < 0) /* nnz must be >= 0 */ - { - stats [COLAMD_STATUS] = COLAMD_ERROR_nnz_negative ; - stats [COLAMD_INFO1] = nnz ; - COLAMD_DEBUG0 (("colamd: number of entries negative %d\n", nnz)) ; - return (false) ; - } - - if (p [0] != 0) - { - stats [COLAMD_STATUS] = COLAMD_ERROR_p0_nonzero ; - stats [COLAMD_INFO1] = p [0] ; - COLAMD_DEBUG0 (("colamd: p[0] not zero %d\n", p [0])) ; - return (false) ; - } - - /* === If no knobs, set default knobs =================================== */ - - if (!knobs) - { - colamd_set_defaults (default_knobs) ; - knobs = default_knobs ; - } - - /* === Allocate the Row and Col arrays from array A ===================== */ - - Col_size = colamd_c (n_col) ; - Row_size = colamd_r (n_row) ; - need = 2*nnz + n_col + Col_size + Row_size ; - - if (need > Alen) - { - /* not enough space in array A to perform the ordering */ - stats [COLAMD_STATUS] = COLAMD_ERROR_A_too_small ; - stats [COLAMD_INFO1] = need ; - stats [COLAMD_INFO2] = Alen ; - COLAMD_DEBUG0 (("colamd: Need Alen >= %d, given only Alen = %d\n", need,Alen)); - return (false) ; - } - - Alen -= Col_size + Row_size ; - Col = (colamd_col *) &A [Alen] ; - Row = (Colamd_Row *) &A [Alen + Col_size] ; - - /* === Construct the row and column data structures ===================== */ - - if (!Eigen::internal::init_rows_cols (n_row, n_col, Row, Col, A, p, stats)) - { - /* input matrix is invalid */ - COLAMD_DEBUG0 (("colamd: Matrix invalid\n")) ; - return (false) ; - } - - /* === Initialize scores, kill dense rows/columns ======================= */ - - Eigen::internal::init_scoring (n_row, n_col, Row, Col, A, p, knobs, - &n_row2, &n_col2, &max_deg) ; - - /* === Order the supercolumns =========================================== */ - - ngarbage = Eigen::internal::find_ordering (n_row, n_col, Alen, Row, Col, A, p, - n_col2, max_deg, 2*nnz) ; - - /* === Order the non-principal columns ================================== */ - - Eigen::internal::order_children (n_col, Col, p) ; - - /* === Return statistics in stats ======================================= */ - - stats [COLAMD_DENSE_ROW] = n_row - n_row2 ; - stats [COLAMD_DENSE_COL] = n_col - n_col2 ; - stats [COLAMD_DEFRAG_COUNT] = ngarbage ; - COLAMD_DEBUG0 (("colamd: done.\n")) ; - return (true) ; -} - -/* ========================================================================== */ -/* === NON-USER-CALLABLE ROUTINES: ========================================== */ -/* ========================================================================== */ - -/* There are no user-callable routines beyond this point in the file */ - - -/* ========================================================================== */ -/* === init_rows_cols ======================================================= */ -/* ========================================================================== */ - -/* - Takes the column form of the matrix in A and creates the row form of the - matrix. Also, row and column attributes are stored in the Col and Row - structs. If the columns are un-sorted or contain duplicate row indices, - this routine will also sort and remove duplicate row indices from the - column form of the matrix. Returns false if the matrix is invalid, - true otherwise. Not user-callable. -*/ -template -static IndexType init_rows_cols /* returns true if OK, or false otherwise */ - ( - /* === Parameters ======================================================= */ - - IndexType n_row, /* number of rows of A */ - IndexType n_col, /* number of columns of A */ - Colamd_Row Row [], /* of size n_row+1 */ - colamd_col Col [], /* of size n_col+1 */ - IndexType A [], /* row indices of A, of size Alen */ - IndexType p [], /* pointers to columns in A, of size n_col+1 */ - IndexType stats [COLAMD_STATS] /* colamd statistics */ - ) -{ - /* === Local variables ================================================== */ - - IndexType col ; /* a column index */ - IndexType row ; /* a row index */ - IndexType *cp ; /* a column pointer */ - IndexType *cp_end ; /* a pointer to the end of a column */ - IndexType *rp ; /* a row pointer */ - IndexType *rp_end ; /* a pointer to the end of a row */ - IndexType last_row ; /* previous row */ - - /* === Initialize columns, and check column pointers ==================== */ - - for (col = 0 ; col < n_col ; col++) - { - Col [col].start = p [col] ; - Col [col].length = p [col+1] - p [col] ; - - if ((Col [col].length) < 0) // extra parentheses to work-around gcc bug 10200 - { - /* column pointers must be non-decreasing */ - stats [COLAMD_STATUS] = COLAMD_ERROR_col_length_negative ; - stats [COLAMD_INFO1] = col ; - stats [COLAMD_INFO2] = Col [col].length ; - COLAMD_DEBUG0 (("colamd: col %d length %d < 0\n", col, Col [col].length)) ; - return (false) ; - } - - Col [col].shared1.thickness = 1 ; - Col [col].shared2.score = 0 ; - Col [col].shared3.prev = COLAMD_EMPTY ; - Col [col].shared4.degree_next = COLAMD_EMPTY ; - } - - /* p [0..n_col] no longer needed, used as "head" in subsequent routines */ - - /* === Scan columns, compute row degrees, and check row indices ========= */ - - stats [COLAMD_INFO3] = 0 ; /* number of duplicate or unsorted row indices*/ - - for (row = 0 ; row < n_row ; row++) - { - Row [row].length = 0 ; - Row [row].shared2.mark = -1 ; - } - - for (col = 0 ; col < n_col ; col++) - { - last_row = -1 ; - - cp = &A [p [col]] ; - cp_end = &A [p [col+1]] ; - - while (cp < cp_end) - { - row = *cp++ ; - - /* make sure row indices within range */ - if (row < 0 || row >= n_row) - { - stats [COLAMD_STATUS] = COLAMD_ERROR_row_index_out_of_bounds ; - stats [COLAMD_INFO1] = col ; - stats [COLAMD_INFO2] = row ; - stats [COLAMD_INFO3] = n_row ; - COLAMD_DEBUG0 (("colamd: row %d col %d out of bounds\n", row, col)) ; - return (false) ; - } - - if (row <= last_row || Row [row].shared2.mark == col) - { - /* row index are unsorted or repeated (or both), thus col */ - /* is jumbled. This is a notice, not an error condition. */ - stats [COLAMD_STATUS] = COLAMD_OK_BUT_JUMBLED ; - stats [COLAMD_INFO1] = col ; - stats [COLAMD_INFO2] = row ; - (stats [COLAMD_INFO3]) ++ ; - COLAMD_DEBUG1 (("colamd: row %d col %d unsorted/duplicate\n",row,col)); - } - - if (Row [row].shared2.mark != col) - { - Row [row].length++ ; - } - else - { - /* this is a repeated entry in the column, */ - /* it will be removed */ - Col [col].length-- ; - } - - /* mark the row as having been seen in this column */ - Row [row].shared2.mark = col ; - - last_row = row ; - } - } - - /* === Compute row pointers ============================================= */ - - /* row form of the matrix starts directly after the column */ - /* form of matrix in A */ - Row [0].start = p [n_col] ; - Row [0].shared1.p = Row [0].start ; - Row [0].shared2.mark = -1 ; - for (row = 1 ; row < n_row ; row++) - { - Row [row].start = Row [row-1].start + Row [row-1].length ; - Row [row].shared1.p = Row [row].start ; - Row [row].shared2.mark = -1 ; - } - - /* === Create row form ================================================== */ - - if (stats [COLAMD_STATUS] == COLAMD_OK_BUT_JUMBLED) - { - /* if cols jumbled, watch for repeated row indices */ - for (col = 0 ; col < n_col ; col++) - { - cp = &A [p [col]] ; - cp_end = &A [p [col+1]] ; - while (cp < cp_end) - { - row = *cp++ ; - if (Row [row].shared2.mark != col) - { - A [(Row [row].shared1.p)++] = col ; - Row [row].shared2.mark = col ; - } - } - } - } - else - { - /* if cols not jumbled, we don't need the mark (this is faster) */ - for (col = 0 ; col < n_col ; col++) - { - cp = &A [p [col]] ; - cp_end = &A [p [col+1]] ; - while (cp < cp_end) - { - A [(Row [*cp++].shared1.p)++] = col ; - } - } - } - - /* === Clear the row marks and set row degrees ========================== */ - - for (row = 0 ; row < n_row ; row++) - { - Row [row].shared2.mark = 0 ; - Row [row].shared1.degree = Row [row].length ; - } - - /* === See if we need to re-create columns ============================== */ - - if (stats [COLAMD_STATUS] == COLAMD_OK_BUT_JUMBLED) - { - COLAMD_DEBUG0 (("colamd: reconstructing column form, matrix jumbled\n")) ; - - - /* === Compute col pointers ========================================= */ - - /* col form of the matrix starts at A [0]. */ - /* Note, we may have a gap between the col form and the row */ - /* form if there were duplicate entries, if so, it will be */ - /* removed upon the first garbage collection */ - Col [0].start = 0 ; - p [0] = Col [0].start ; - for (col = 1 ; col < n_col ; col++) - { - /* note that the lengths here are for pruned columns, i.e. */ - /* no duplicate row indices will exist for these columns */ - Col [col].start = Col [col-1].start + Col [col-1].length ; - p [col] = Col [col].start ; - } - - /* === Re-create col form =========================================== */ - - for (row = 0 ; row < n_row ; row++) - { - rp = &A [Row [row].start] ; - rp_end = rp + Row [row].length ; - while (rp < rp_end) - { - A [(p [*rp++])++] = row ; - } - } - } - - /* === Done. Matrix is not (or no longer) jumbled ====================== */ - - return (true) ; -} - - -/* ========================================================================== */ -/* === init_scoring ========================================================= */ -/* ========================================================================== */ - -/* - Kills dense or empty columns and rows, calculates an initial score for - each column, and places all columns in the degree lists. Not user-callable. -*/ -template -static void init_scoring - ( - /* === Parameters ======================================================= */ - - IndexType n_row, /* number of rows of A */ - IndexType n_col, /* number of columns of A */ - Colamd_Row Row [], /* of size n_row+1 */ - colamd_col Col [], /* of size n_col+1 */ - IndexType A [], /* column form and row form of A */ - IndexType head [], /* of size n_col+1 */ - double knobs [COLAMD_KNOBS],/* parameters */ - IndexType *p_n_row2, /* number of non-dense, non-empty rows */ - IndexType *p_n_col2, /* number of non-dense, non-empty columns */ - IndexType *p_max_deg /* maximum row degree */ - ) -{ - /* === Local variables ================================================== */ - - IndexType c ; /* a column index */ - IndexType r, row ; /* a row index */ - IndexType *cp ; /* a column pointer */ - IndexType deg ; /* degree of a row or column */ - IndexType *cp_end ; /* a pointer to the end of a column */ - IndexType *new_cp ; /* new column pointer */ - IndexType col_length ; /* length of pruned column */ - IndexType score ; /* current column score */ - IndexType n_col2 ; /* number of non-dense, non-empty columns */ - IndexType n_row2 ; /* number of non-dense, non-empty rows */ - IndexType dense_row_count ; /* remove rows with more entries than this */ - IndexType dense_col_count ; /* remove cols with more entries than this */ - IndexType min_score ; /* smallest column score */ - IndexType max_deg ; /* maximum row degree */ - IndexType next_col ; /* Used to add to degree list.*/ - - - /* === Extract knobs ==================================================== */ - - dense_row_count = numext::maxi(IndexType(0), numext::mini(IndexType(knobs [COLAMD_DENSE_ROW] * n_col), n_col)) ; - dense_col_count = numext::maxi(IndexType(0), numext::mini(IndexType(knobs [COLAMD_DENSE_COL] * n_row), n_row)) ; - COLAMD_DEBUG1 (("colamd: densecount: %d %d\n", dense_row_count, dense_col_count)) ; - max_deg = 0 ; - n_col2 = n_col ; - n_row2 = n_row ; - - /* === Kill empty columns =============================================== */ - - /* Put the empty columns at the end in their natural order, so that LU */ - /* factorization can proceed as far as possible. */ - for (c = n_col-1 ; c >= 0 ; c--) - { - deg = Col [c].length ; - if (deg == 0) - { - /* this is a empty column, kill and order it last */ - Col [c].shared2.order = --n_col2 ; - KILL_PRINCIPAL_COL (c) ; - } - } - COLAMD_DEBUG1 (("colamd: null columns killed: %d\n", n_col - n_col2)) ; - - /* === Kill dense columns =============================================== */ - - /* Put the dense columns at the end, in their natural order */ - for (c = n_col-1 ; c >= 0 ; c--) - { - /* skip any dead columns */ - if (COL_IS_DEAD (c)) - { - continue ; - } - deg = Col [c].length ; - if (deg > dense_col_count) - { - /* this is a dense column, kill and order it last */ - Col [c].shared2.order = --n_col2 ; - /* decrement the row degrees */ - cp = &A [Col [c].start] ; - cp_end = cp + Col [c].length ; - while (cp < cp_end) - { - Row [*cp++].shared1.degree-- ; - } - KILL_PRINCIPAL_COL (c) ; - } - } - COLAMD_DEBUG1 (("colamd: Dense and null columns killed: %d\n", n_col - n_col2)) ; - - /* === Kill dense and empty rows ======================================== */ - - for (r = 0 ; r < n_row ; r++) - { - deg = Row [r].shared1.degree ; - COLAMD_ASSERT (deg >= 0 && deg <= n_col) ; - if (deg > dense_row_count || deg == 0) - { - /* kill a dense or empty row */ - KILL_ROW (r) ; - --n_row2 ; - } - else - { - /* keep track of max degree of remaining rows */ - max_deg = numext::maxi(max_deg, deg) ; - } - } - COLAMD_DEBUG1 (("colamd: Dense and null rows killed: %d\n", n_row - n_row2)) ; - - /* === Compute initial column scores ==================================== */ - - /* At this point the row degrees are accurate. They reflect the number */ - /* of "live" (non-dense) columns in each row. No empty rows exist. */ - /* Some "live" columns may contain only dead rows, however. These are */ - /* pruned in the code below. */ - - /* now find the initial matlab score for each column */ - for (c = n_col-1 ; c >= 0 ; c--) - { - /* skip dead column */ - if (COL_IS_DEAD (c)) - { - continue ; - } - score = 0 ; - cp = &A [Col [c].start] ; - new_cp = cp ; - cp_end = cp + Col [c].length ; - while (cp < cp_end) - { - /* get a row */ - row = *cp++ ; - /* skip if dead */ - if (ROW_IS_DEAD (row)) - { - continue ; - } - /* compact the column */ - *new_cp++ = row ; - /* add row's external degree */ - score += Row [row].shared1.degree - 1 ; - /* guard against integer overflow */ - score = numext::mini(score, n_col) ; - } - /* determine pruned column length */ - col_length = (IndexType) (new_cp - &A [Col [c].start]) ; - if (col_length == 0) - { - /* a newly-made null column (all rows in this col are "dense" */ - /* and have already been killed) */ - COLAMD_DEBUG2 (("Newly null killed: %d\n", c)) ; - Col [c].shared2.order = --n_col2 ; - KILL_PRINCIPAL_COL (c) ; - } - else - { - /* set column length and set score */ - COLAMD_ASSERT (score >= 0) ; - COLAMD_ASSERT (score <= n_col) ; - Col [c].length = col_length ; - Col [c].shared2.score = score ; - } - } - COLAMD_DEBUG1 (("colamd: Dense, null, and newly-null columns killed: %d\n", - n_col-n_col2)) ; - - /* At this point, all empty rows and columns are dead. All live columns */ - /* are "clean" (containing no dead rows) and simplicial (no supercolumns */ - /* yet). Rows may contain dead columns, but all live rows contain at */ - /* least one live column. */ - - /* === Initialize degree lists ========================================== */ - - - /* clear the hash buckets */ - for (c = 0 ; c <= n_col ; c++) - { - head [c] = COLAMD_EMPTY ; - } - min_score = n_col ; - /* place in reverse order, so low column indices are at the front */ - /* of the lists. This is to encourage natural tie-breaking */ - for (c = n_col-1 ; c >= 0 ; c--) - { - /* only add principal columns to degree lists */ - if (COL_IS_ALIVE (c)) - { - COLAMD_DEBUG4 (("place %d score %d minscore %d ncol %d\n", - c, Col [c].shared2.score, min_score, n_col)) ; - - /* === Add columns score to DList =============================== */ - - score = Col [c].shared2.score ; - - COLAMD_ASSERT (min_score >= 0) ; - COLAMD_ASSERT (min_score <= n_col) ; - COLAMD_ASSERT (score >= 0) ; - COLAMD_ASSERT (score <= n_col) ; - COLAMD_ASSERT (head [score] >= COLAMD_EMPTY) ; - - /* now add this column to dList at proper score location */ - next_col = head [score] ; - Col [c].shared3.prev = COLAMD_EMPTY ; - Col [c].shared4.degree_next = next_col ; - - /* if there already was a column with the same score, set its */ - /* previous pointer to this new column */ - if (next_col != COLAMD_EMPTY) - { - Col [next_col].shared3.prev = c ; - } - head [score] = c ; - - /* see if this score is less than current min */ - min_score = numext::mini(min_score, score) ; - - - } - } - - - /* === Return number of remaining columns, and max row degree =========== */ - - *p_n_col2 = n_col2 ; - *p_n_row2 = n_row2 ; - *p_max_deg = max_deg ; -} - - -/* ========================================================================== */ -/* === find_ordering ======================================================== */ -/* ========================================================================== */ - -/* - Order the principal columns of the supercolumn form of the matrix - (no supercolumns on input). Uses a minimum approximate column minimum - degree ordering method. Not user-callable. -*/ -template -static IndexType find_ordering /* return the number of garbage collections */ - ( - /* === Parameters ======================================================= */ - - IndexType n_row, /* number of rows of A */ - IndexType n_col, /* number of columns of A */ - IndexType Alen, /* size of A, 2*nnz + n_col or larger */ - Colamd_Row Row [], /* of size n_row+1 */ - colamd_col Col [], /* of size n_col+1 */ - IndexType A [], /* column form and row form of A */ - IndexType head [], /* of size n_col+1 */ - IndexType n_col2, /* Remaining columns to order */ - IndexType max_deg, /* Maximum row degree */ - IndexType pfree /* index of first free slot (2*nnz on entry) */ - ) -{ - /* === Local variables ================================================== */ - - IndexType k ; /* current pivot ordering step */ - IndexType pivot_col ; /* current pivot column */ - IndexType *cp ; /* a column pointer */ - IndexType *rp ; /* a row pointer */ - IndexType pivot_row ; /* current pivot row */ - IndexType *new_cp ; /* modified column pointer */ - IndexType *new_rp ; /* modified row pointer */ - IndexType pivot_row_start ; /* pointer to start of pivot row */ - IndexType pivot_row_degree ; /* number of columns in pivot row */ - IndexType pivot_row_length ; /* number of supercolumns in pivot row */ - IndexType pivot_col_score ; /* score of pivot column */ - IndexType needed_memory ; /* free space needed for pivot row */ - IndexType *cp_end ; /* pointer to the end of a column */ - IndexType *rp_end ; /* pointer to the end of a row */ - IndexType row ; /* a row index */ - IndexType col ; /* a column index */ - IndexType max_score ; /* maximum possible score */ - IndexType cur_score ; /* score of current column */ - unsigned int hash ; /* hash value for supernode detection */ - IndexType head_column ; /* head of hash bucket */ - IndexType first_col ; /* first column in hash bucket */ - IndexType tag_mark ; /* marker value for mark array */ - IndexType row_mark ; /* Row [row].shared2.mark */ - IndexType set_difference ; /* set difference size of row with pivot row */ - IndexType min_score ; /* smallest column score */ - IndexType col_thickness ; /* "thickness" (no. of columns in a supercol) */ - IndexType max_mark ; /* maximum value of tag_mark */ - IndexType pivot_col_thickness ; /* number of columns represented by pivot col */ - IndexType prev_col ; /* Used by Dlist operations. */ - IndexType next_col ; /* Used by Dlist operations. */ - IndexType ngarbage ; /* number of garbage collections performed */ - - - /* === Initialization and clear mark ==================================== */ - - max_mark = INT_MAX - n_col ; /* INT_MAX defined in */ - tag_mark = Eigen::internal::clear_mark (n_row, Row) ; - min_score = 0 ; - ngarbage = 0 ; - COLAMD_DEBUG1 (("colamd: Ordering, n_col2=%d\n", n_col2)) ; - - /* === Order the columns ================================================ */ - - for (k = 0 ; k < n_col2 ; /* 'k' is incremented below */) - { - - /* === Select pivot column, and order it ============================ */ - - /* make sure degree list isn't empty */ - COLAMD_ASSERT (min_score >= 0) ; - COLAMD_ASSERT (min_score <= n_col) ; - COLAMD_ASSERT (head [min_score] >= COLAMD_EMPTY) ; - - /* get pivot column from head of minimum degree list */ - while (min_score < n_col && head [min_score] == COLAMD_EMPTY) - { - min_score++ ; - } - pivot_col = head [min_score] ; - COLAMD_ASSERT (pivot_col >= 0 && pivot_col <= n_col) ; - next_col = Col [pivot_col].shared4.degree_next ; - head [min_score] = next_col ; - if (next_col != COLAMD_EMPTY) - { - Col [next_col].shared3.prev = COLAMD_EMPTY ; - } - - COLAMD_ASSERT (COL_IS_ALIVE (pivot_col)) ; - COLAMD_DEBUG3 (("Pivot col: %d\n", pivot_col)) ; - - /* remember score for defrag check */ - pivot_col_score = Col [pivot_col].shared2.score ; - - /* the pivot column is the kth column in the pivot order */ - Col [pivot_col].shared2.order = k ; - - /* increment order count by column thickness */ - pivot_col_thickness = Col [pivot_col].shared1.thickness ; - k += pivot_col_thickness ; - COLAMD_ASSERT (pivot_col_thickness > 0) ; - - /* === Garbage_collection, if necessary ============================= */ - - needed_memory = numext::mini(pivot_col_score, n_col - k) ; - if (pfree + needed_memory >= Alen) - { - pfree = Eigen::internal::garbage_collection (n_row, n_col, Row, Col, A, &A [pfree]) ; - ngarbage++ ; - /* after garbage collection we will have enough */ - COLAMD_ASSERT (pfree + needed_memory < Alen) ; - /* garbage collection has wiped out the Row[].shared2.mark array */ - tag_mark = Eigen::internal::clear_mark (n_row, Row) ; - - } - - /* === Compute pivot row pattern ==================================== */ - - /* get starting location for this new merged row */ - pivot_row_start = pfree ; - - /* initialize new row counts to zero */ - pivot_row_degree = 0 ; - - /* tag pivot column as having been visited so it isn't included */ - /* in merged pivot row */ - Col [pivot_col].shared1.thickness = -pivot_col_thickness ; - - /* pivot row is the union of all rows in the pivot column pattern */ - cp = &A [Col [pivot_col].start] ; - cp_end = cp + Col [pivot_col].length ; - while (cp < cp_end) - { - /* get a row */ - row = *cp++ ; - COLAMD_DEBUG4 (("Pivot col pattern %d %d\n", ROW_IS_ALIVE (row), row)) ; - /* skip if row is dead */ - if (ROW_IS_DEAD (row)) - { - continue ; - } - rp = &A [Row [row].start] ; - rp_end = rp + Row [row].length ; - while (rp < rp_end) - { - /* get a column */ - col = *rp++ ; - /* add the column, if alive and untagged */ - col_thickness = Col [col].shared1.thickness ; - if (col_thickness > 0 && COL_IS_ALIVE (col)) - { - /* tag column in pivot row */ - Col [col].shared1.thickness = -col_thickness ; - COLAMD_ASSERT (pfree < Alen) ; - /* place column in pivot row */ - A [pfree++] = col ; - pivot_row_degree += col_thickness ; - } - } - } - - /* clear tag on pivot column */ - Col [pivot_col].shared1.thickness = pivot_col_thickness ; - max_deg = numext::maxi(max_deg, pivot_row_degree) ; - - - /* === Kill all rows used to construct pivot row ==================== */ - - /* also kill pivot row, temporarily */ - cp = &A [Col [pivot_col].start] ; - cp_end = cp + Col [pivot_col].length ; - while (cp < cp_end) - { - /* may be killing an already dead row */ - row = *cp++ ; - COLAMD_DEBUG3 (("Kill row in pivot col: %d\n", row)) ; - KILL_ROW (row) ; - } - - /* === Select a row index to use as the new pivot row =============== */ - - pivot_row_length = pfree - pivot_row_start ; - if (pivot_row_length > 0) - { - /* pick the "pivot" row arbitrarily (first row in col) */ - pivot_row = A [Col [pivot_col].start] ; - COLAMD_DEBUG3 (("Pivotal row is %d\n", pivot_row)) ; - } - else - { - /* there is no pivot row, since it is of zero length */ - pivot_row = COLAMD_EMPTY ; - COLAMD_ASSERT (pivot_row_length == 0) ; - } - COLAMD_ASSERT (Col [pivot_col].length > 0 || pivot_row_length == 0) ; - - /* === Approximate degree computation =============================== */ - - /* Here begins the computation of the approximate degree. The column */ - /* score is the sum of the pivot row "length", plus the size of the */ - /* set differences of each row in the column minus the pattern of the */ - /* pivot row itself. The column ("thickness") itself is also */ - /* excluded from the column score (we thus use an approximate */ - /* external degree). */ - - /* The time taken by the following code (compute set differences, and */ - /* add them up) is proportional to the size of the data structure */ - /* being scanned - that is, the sum of the sizes of each column in */ - /* the pivot row. Thus, the amortized time to compute a column score */ - /* is proportional to the size of that column (where size, in this */ - /* context, is the column "length", or the number of row indices */ - /* in that column). The number of row indices in a column is */ - /* monotonically non-decreasing, from the length of the original */ - /* column on input to colamd. */ - - /* === Compute set differences ====================================== */ - - COLAMD_DEBUG3 (("** Computing set differences phase. **\n")) ; - - /* pivot row is currently dead - it will be revived later. */ - - COLAMD_DEBUG3 (("Pivot row: ")) ; - /* for each column in pivot row */ - rp = &A [pivot_row_start] ; - rp_end = rp + pivot_row_length ; - while (rp < rp_end) - { - col = *rp++ ; - COLAMD_ASSERT (COL_IS_ALIVE (col) && col != pivot_col) ; - COLAMD_DEBUG3 (("Col: %d\n", col)) ; - - /* clear tags used to construct pivot row pattern */ - col_thickness = -Col [col].shared1.thickness ; - COLAMD_ASSERT (col_thickness > 0) ; - Col [col].shared1.thickness = col_thickness ; - - /* === Remove column from degree list =========================== */ - - cur_score = Col [col].shared2.score ; - prev_col = Col [col].shared3.prev ; - next_col = Col [col].shared4.degree_next ; - COLAMD_ASSERT (cur_score >= 0) ; - COLAMD_ASSERT (cur_score <= n_col) ; - COLAMD_ASSERT (cur_score >= COLAMD_EMPTY) ; - if (prev_col == COLAMD_EMPTY) - { - head [cur_score] = next_col ; - } - else - { - Col [prev_col].shared4.degree_next = next_col ; - } - if (next_col != COLAMD_EMPTY) - { - Col [next_col].shared3.prev = prev_col ; - } - - /* === Scan the column ========================================== */ - - cp = &A [Col [col].start] ; - cp_end = cp + Col [col].length ; - while (cp < cp_end) - { - /* get a row */ - row = *cp++ ; - row_mark = Row [row].shared2.mark ; - /* skip if dead */ - if (ROW_IS_MARKED_DEAD (row_mark)) - { - continue ; - } - COLAMD_ASSERT (row != pivot_row) ; - set_difference = row_mark - tag_mark ; - /* check if the row has been seen yet */ - if (set_difference < 0) - { - COLAMD_ASSERT (Row [row].shared1.degree <= max_deg) ; - set_difference = Row [row].shared1.degree ; - } - /* subtract column thickness from this row's set difference */ - set_difference -= col_thickness ; - COLAMD_ASSERT (set_difference >= 0) ; - /* absorb this row if the set difference becomes zero */ - if (set_difference == 0) - { - COLAMD_DEBUG3 (("aggressive absorption. Row: %d\n", row)) ; - KILL_ROW (row) ; - } - else - { - /* save the new mark */ - Row [row].shared2.mark = set_difference + tag_mark ; - } - } - } - - - /* === Add up set differences for each column ======================= */ - - COLAMD_DEBUG3 (("** Adding set differences phase. **\n")) ; - - /* for each column in pivot row */ - rp = &A [pivot_row_start] ; - rp_end = rp + pivot_row_length ; - while (rp < rp_end) - { - /* get a column */ - col = *rp++ ; - COLAMD_ASSERT (COL_IS_ALIVE (col) && col != pivot_col) ; - hash = 0 ; - cur_score = 0 ; - cp = &A [Col [col].start] ; - /* compact the column */ - new_cp = cp ; - cp_end = cp + Col [col].length ; - - COLAMD_DEBUG4 (("Adding set diffs for Col: %d.\n", col)) ; - - while (cp < cp_end) - { - /* get a row */ - row = *cp++ ; - COLAMD_ASSERT(row >= 0 && row < n_row) ; - row_mark = Row [row].shared2.mark ; - /* skip if dead */ - if (ROW_IS_MARKED_DEAD (row_mark)) - { - continue ; - } - COLAMD_ASSERT (row_mark > tag_mark) ; - /* compact the column */ - *new_cp++ = row ; - /* compute hash function */ - hash += row ; - /* add set difference */ - cur_score += row_mark - tag_mark ; - /* integer overflow... */ - cur_score = numext::mini(cur_score, n_col) ; - } - - /* recompute the column's length */ - Col [col].length = (IndexType) (new_cp - &A [Col [col].start]) ; - - /* === Further mass elimination ================================= */ - - if (Col [col].length == 0) - { - COLAMD_DEBUG4 (("further mass elimination. Col: %d\n", col)) ; - /* nothing left but the pivot row in this column */ - KILL_PRINCIPAL_COL (col) ; - pivot_row_degree -= Col [col].shared1.thickness ; - COLAMD_ASSERT (pivot_row_degree >= 0) ; - /* order it */ - Col [col].shared2.order = k ; - /* increment order count by column thickness */ - k += Col [col].shared1.thickness ; - } - else - { - /* === Prepare for supercolumn detection ==================== */ - - COLAMD_DEBUG4 (("Preparing supercol detection for Col: %d.\n", col)) ; - - /* save score so far */ - Col [col].shared2.score = cur_score ; - - /* add column to hash table, for supercolumn detection */ - hash %= n_col + 1 ; - - COLAMD_DEBUG4 ((" Hash = %d, n_col = %d.\n", hash, n_col)) ; - COLAMD_ASSERT (hash <= n_col) ; - - head_column = head [hash] ; - if (head_column > COLAMD_EMPTY) - { - /* degree list "hash" is non-empty, use prev (shared3) of */ - /* first column in degree list as head of hash bucket */ - first_col = Col [head_column].shared3.headhash ; - Col [head_column].shared3.headhash = col ; - } - else - { - /* degree list "hash" is empty, use head as hash bucket */ - first_col = - (head_column + 2) ; - head [hash] = - (col + 2) ; - } - Col [col].shared4.hash_next = first_col ; - - /* save hash function in Col [col].shared3.hash */ - Col [col].shared3.hash = (IndexType) hash ; - COLAMD_ASSERT (COL_IS_ALIVE (col)) ; - } - } - - /* The approximate external column degree is now computed. */ - - /* === Supercolumn detection ======================================== */ - - COLAMD_DEBUG3 (("** Supercolumn detection phase. **\n")) ; - - Eigen::internal::detect_super_cols (Col, A, head, pivot_row_start, pivot_row_length) ; - - /* === Kill the pivotal column ====================================== */ - - KILL_PRINCIPAL_COL (pivot_col) ; - - /* === Clear mark =================================================== */ - - tag_mark += (max_deg + 1) ; - if (tag_mark >= max_mark) - { - COLAMD_DEBUG2 (("clearing tag_mark\n")) ; - tag_mark = Eigen::internal::clear_mark (n_row, Row) ; - } - - /* === Finalize the new pivot row, and column scores ================ */ - - COLAMD_DEBUG3 (("** Finalize scores phase. **\n")) ; - - /* for each column in pivot row */ - rp = &A [pivot_row_start] ; - /* compact the pivot row */ - new_rp = rp ; - rp_end = rp + pivot_row_length ; - while (rp < rp_end) - { - col = *rp++ ; - /* skip dead columns */ - if (COL_IS_DEAD (col)) - { - continue ; - } - *new_rp++ = col ; - /* add new pivot row to column */ - A [Col [col].start + (Col [col].length++)] = pivot_row ; - - /* retrieve score so far and add on pivot row's degree. */ - /* (we wait until here for this in case the pivot */ - /* row's degree was reduced due to mass elimination). */ - cur_score = Col [col].shared2.score + pivot_row_degree ; - - /* calculate the max possible score as the number of */ - /* external columns minus the 'k' value minus the */ - /* columns thickness */ - max_score = n_col - k - Col [col].shared1.thickness ; - - /* make the score the external degree of the union-of-rows */ - cur_score -= Col [col].shared1.thickness ; - - /* make sure score is less or equal than the max score */ - cur_score = numext::mini(cur_score, max_score) ; - COLAMD_ASSERT (cur_score >= 0) ; - - /* store updated score */ - Col [col].shared2.score = cur_score ; - - /* === Place column back in degree list ========================= */ - - COLAMD_ASSERT (min_score >= 0) ; - COLAMD_ASSERT (min_score <= n_col) ; - COLAMD_ASSERT (cur_score >= 0) ; - COLAMD_ASSERT (cur_score <= n_col) ; - COLAMD_ASSERT (head [cur_score] >= COLAMD_EMPTY) ; - next_col = head [cur_score] ; - Col [col].shared4.degree_next = next_col ; - Col [col].shared3.prev = COLAMD_EMPTY ; - if (next_col != COLAMD_EMPTY) - { - Col [next_col].shared3.prev = col ; - } - head [cur_score] = col ; - - /* see if this score is less than current min */ - min_score = numext::mini(min_score, cur_score) ; - - } - - /* === Resurrect the new pivot row ================================== */ - - if (pivot_row_degree > 0) - { - /* update pivot row length to reflect any cols that were killed */ - /* during super-col detection and mass elimination */ - Row [pivot_row].start = pivot_row_start ; - Row [pivot_row].length = (IndexType) (new_rp - &A[pivot_row_start]) ; - Row [pivot_row].shared1.degree = pivot_row_degree ; - Row [pivot_row].shared2.mark = 0 ; - /* pivot row is no longer dead */ - } - } - - /* === All principal columns have now been ordered ====================== */ - - return (ngarbage) ; -} - - -/* ========================================================================== */ -/* === order_children ======================================================= */ -/* ========================================================================== */ - -/* - The find_ordering routine has ordered all of the principal columns (the - representatives of the supercolumns). The non-principal columns have not - yet been ordered. This routine orders those columns by walking up the - parent tree (a column is a child of the column which absorbed it). The - final permutation vector is then placed in p [0 ... n_col-1], with p [0] - being the first column, and p [n_col-1] being the last. It doesn't look - like it at first glance, but be assured that this routine takes time linear - in the number of columns. Although not immediately obvious, the time - taken by this routine is O (n_col), that is, linear in the number of - columns. Not user-callable. -*/ -template -static inline void order_children -( - /* === Parameters ======================================================= */ - - IndexType n_col, /* number of columns of A */ - colamd_col Col [], /* of size n_col+1 */ - IndexType p [] /* p [0 ... n_col-1] is the column permutation*/ - ) -{ - /* === Local variables ================================================== */ - - IndexType i ; /* loop counter for all columns */ - IndexType c ; /* column index */ - IndexType parent ; /* index of column's parent */ - IndexType order ; /* column's order */ - - /* === Order each non-principal column ================================== */ - - for (i = 0 ; i < n_col ; i++) - { - /* find an un-ordered non-principal column */ - COLAMD_ASSERT (COL_IS_DEAD (i)) ; - if (!COL_IS_DEAD_PRINCIPAL (i) && Col [i].shared2.order == COLAMD_EMPTY) - { - parent = i ; - /* once found, find its principal parent */ - do - { - parent = Col [parent].shared1.parent ; - } while (!COL_IS_DEAD_PRINCIPAL (parent)) ; - - /* now, order all un-ordered non-principal columns along path */ - /* to this parent. collapse tree at the same time */ - c = i ; - /* get order of parent */ - order = Col [parent].shared2.order ; - - do - { - COLAMD_ASSERT (Col [c].shared2.order == COLAMD_EMPTY) ; - - /* order this column */ - Col [c].shared2.order = order++ ; - /* collaps tree */ - Col [c].shared1.parent = parent ; - - /* get immediate parent of this column */ - c = Col [c].shared1.parent ; - - /* continue until we hit an ordered column. There are */ - /* guarranteed not to be anymore unordered columns */ - /* above an ordered column */ - } while (Col [c].shared2.order == COLAMD_EMPTY) ; - - /* re-order the super_col parent to largest order for this group */ - Col [parent].shared2.order = order ; - } - } - - /* === Generate the permutation ========================================= */ - - for (c = 0 ; c < n_col ; c++) - { - p [Col [c].shared2.order] = c ; - } -} - - -/* ========================================================================== */ -/* === detect_super_cols ==================================================== */ -/* ========================================================================== */ - -/* - Detects supercolumns by finding matches between columns in the hash buckets. - Check amongst columns in the set A [row_start ... row_start + row_length-1]. - The columns under consideration are currently *not* in the degree lists, - and have already been placed in the hash buckets. - - The hash bucket for columns whose hash function is equal to h is stored - as follows: - - if head [h] is >= 0, then head [h] contains a degree list, so: - - head [h] is the first column in degree bucket h. - Col [head [h]].headhash gives the first column in hash bucket h. - - otherwise, the degree list is empty, and: - - -(head [h] + 2) is the first column in hash bucket h. - - For a column c in a hash bucket, Col [c].shared3.prev is NOT a "previous - column" pointer. Col [c].shared3.hash is used instead as the hash number - for that column. The value of Col [c].shared4.hash_next is the next column - in the same hash bucket. - - Assuming no, or "few" hash collisions, the time taken by this routine is - linear in the sum of the sizes (lengths) of each column whose score has - just been computed in the approximate degree computation. - Not user-callable. -*/ -template -static void detect_super_cols -( - /* === Parameters ======================================================= */ - - colamd_col Col [], /* of size n_col+1 */ - IndexType A [], /* row indices of A */ - IndexType head [], /* head of degree lists and hash buckets */ - IndexType row_start, /* pointer to set of columns to check */ - IndexType row_length /* number of columns to check */ -) -{ - /* === Local variables ================================================== */ - - IndexType hash ; /* hash value for a column */ - IndexType *rp ; /* pointer to a row */ - IndexType c ; /* a column index */ - IndexType super_c ; /* column index of the column to absorb into */ - IndexType *cp1 ; /* column pointer for column super_c */ - IndexType *cp2 ; /* column pointer for column c */ - IndexType length ; /* length of column super_c */ - IndexType prev_c ; /* column preceding c in hash bucket */ - IndexType i ; /* loop counter */ - IndexType *rp_end ; /* pointer to the end of the row */ - IndexType col ; /* a column index in the row to check */ - IndexType head_column ; /* first column in hash bucket or degree list */ - IndexType first_col ; /* first column in hash bucket */ - - /* === Consider each column in the row ================================== */ - - rp = &A [row_start] ; - rp_end = rp + row_length ; - while (rp < rp_end) - { - col = *rp++ ; - if (COL_IS_DEAD (col)) - { - continue ; - } - - /* get hash number for this column */ - hash = Col [col].shared3.hash ; - COLAMD_ASSERT (hash <= n_col) ; - - /* === Get the first column in this hash bucket ===================== */ - - head_column = head [hash] ; - if (head_column > COLAMD_EMPTY) - { - first_col = Col [head_column].shared3.headhash ; - } - else - { - first_col = - (head_column + 2) ; - } - - /* === Consider each column in the hash bucket ====================== */ - - for (super_c = first_col ; super_c != COLAMD_EMPTY ; - super_c = Col [super_c].shared4.hash_next) - { - COLAMD_ASSERT (COL_IS_ALIVE (super_c)) ; - COLAMD_ASSERT (Col [super_c].shared3.hash == hash) ; - length = Col [super_c].length ; - - /* prev_c is the column preceding column c in the hash bucket */ - prev_c = super_c ; - - /* === Compare super_c with all columns after it ================ */ - - for (c = Col [super_c].shared4.hash_next ; - c != COLAMD_EMPTY ; c = Col [c].shared4.hash_next) - { - COLAMD_ASSERT (c != super_c) ; - COLAMD_ASSERT (COL_IS_ALIVE (c)) ; - COLAMD_ASSERT (Col [c].shared3.hash == hash) ; - - /* not identical if lengths or scores are different */ - if (Col [c].length != length || - Col [c].shared2.score != Col [super_c].shared2.score) - { - prev_c = c ; - continue ; - } - - /* compare the two columns */ - cp1 = &A [Col [super_c].start] ; - cp2 = &A [Col [c].start] ; - - for (i = 0 ; i < length ; i++) - { - /* the columns are "clean" (no dead rows) */ - COLAMD_ASSERT (ROW_IS_ALIVE (*cp1)) ; - COLAMD_ASSERT (ROW_IS_ALIVE (*cp2)) ; - /* row indices will same order for both supercols, */ - /* no gather scatter nessasary */ - if (*cp1++ != *cp2++) - { - break ; - } - } - - /* the two columns are different if the for-loop "broke" */ - if (i != length) - { - prev_c = c ; - continue ; - } - - /* === Got it! two columns are identical =================== */ - - COLAMD_ASSERT (Col [c].shared2.score == Col [super_c].shared2.score) ; - - Col [super_c].shared1.thickness += Col [c].shared1.thickness ; - Col [c].shared1.parent = super_c ; - KILL_NON_PRINCIPAL_COL (c) ; - /* order c later, in order_children() */ - Col [c].shared2.order = COLAMD_EMPTY ; - /* remove c from hash bucket */ - Col [prev_c].shared4.hash_next = Col [c].shared4.hash_next ; - } - } - - /* === Empty this hash bucket ======================================= */ - - if (head_column > COLAMD_EMPTY) - { - /* corresponding degree list "hash" is not empty */ - Col [head_column].shared3.headhash = COLAMD_EMPTY ; - } - else - { - /* corresponding degree list "hash" is empty */ - head [hash] = COLAMD_EMPTY ; - } - } -} - - -/* ========================================================================== */ -/* === garbage_collection =================================================== */ -/* ========================================================================== */ - -/* - Defragments and compacts columns and rows in the workspace A. Used when - all avaliable memory has been used while performing row merging. Returns - the index of the first free position in A, after garbage collection. The - time taken by this routine is linear is the size of the array A, which is - itself linear in the number of nonzeros in the input matrix. - Not user-callable. -*/ -template -static IndexType garbage_collection /* returns the new value of pfree */ - ( - /* === Parameters ======================================================= */ - - IndexType n_row, /* number of rows */ - IndexType n_col, /* number of columns */ - Colamd_Row Row [], /* row info */ - colamd_col Col [], /* column info */ - IndexType A [], /* A [0 ... Alen-1] holds the matrix */ - IndexType *pfree /* &A [0] ... pfree is in use */ - ) -{ - /* === Local variables ================================================== */ - - IndexType *psrc ; /* source pointer */ - IndexType *pdest ; /* destination pointer */ - IndexType j ; /* counter */ - IndexType r ; /* a row index */ - IndexType c ; /* a column index */ - IndexType length ; /* length of a row or column */ - - /* === Defragment the columns =========================================== */ - - pdest = &A[0] ; - for (c = 0 ; c < n_col ; c++) - { - if (COL_IS_ALIVE (c)) - { - psrc = &A [Col [c].start] ; - - /* move and compact the column */ - COLAMD_ASSERT (pdest <= psrc) ; - Col [c].start = (IndexType) (pdest - &A [0]) ; - length = Col [c].length ; - for (j = 0 ; j < length ; j++) - { - r = *psrc++ ; - if (ROW_IS_ALIVE (r)) - { - *pdest++ = r ; - } - } - Col [c].length = (IndexType) (pdest - &A [Col [c].start]) ; - } - } - - /* === Prepare to defragment the rows =================================== */ - - for (r = 0 ; r < n_row ; r++) - { - if (ROW_IS_ALIVE (r)) - { - if (Row [r].length == 0) - { - /* this row is of zero length. cannot compact it, so kill it */ - COLAMD_DEBUG3 (("Defrag row kill\n")) ; - KILL_ROW (r) ; - } - else - { - /* save first column index in Row [r].shared2.first_column */ - psrc = &A [Row [r].start] ; - Row [r].shared2.first_column = *psrc ; - COLAMD_ASSERT (ROW_IS_ALIVE (r)) ; - /* flag the start of the row with the one's complement of row */ - *psrc = ONES_COMPLEMENT (r) ; - - } - } - } - - /* === Defragment the rows ============================================== */ - - psrc = pdest ; - while (psrc < pfree) - { - /* find a negative number ... the start of a row */ - if (*psrc++ < 0) - { - psrc-- ; - /* get the row index */ - r = ONES_COMPLEMENT (*psrc) ; - COLAMD_ASSERT (r >= 0 && r < n_row) ; - /* restore first column index */ - *psrc = Row [r].shared2.first_column ; - COLAMD_ASSERT (ROW_IS_ALIVE (r)) ; - - /* move and compact the row */ - COLAMD_ASSERT (pdest <= psrc) ; - Row [r].start = (IndexType) (pdest - &A [0]) ; - length = Row [r].length ; - for (j = 0 ; j < length ; j++) - { - c = *psrc++ ; - if (COL_IS_ALIVE (c)) - { - *pdest++ = c ; - } - } - Row [r].length = (IndexType) (pdest - &A [Row [r].start]) ; - - } - } - /* ensure we found all the rows */ - COLAMD_ASSERT (debug_rows == 0) ; - - /* === Return the new value of pfree ==================================== */ - - return ((IndexType) (pdest - &A [0])) ; -} - - -/* ========================================================================== */ -/* === clear_mark =========================================================== */ -/* ========================================================================== */ - -/* - Clears the Row [].shared2.mark array, and returns the new tag_mark. - Return value is the new tag_mark. Not user-callable. -*/ -template -static inline IndexType clear_mark /* return the new value for tag_mark */ - ( - /* === Parameters ======================================================= */ - - IndexType n_row, /* number of rows in A */ - Colamd_Row Row [] /* Row [0 ... n_row-1].shared2.mark is set to zero */ - ) -{ - /* === Local variables ================================================== */ - - IndexType r ; - - for (r = 0 ; r < n_row ; r++) - { - if (ROW_IS_ALIVE (r)) - { - Row [r].shared2.mark = 0 ; - } - } - return (1) ; -} - - -} // namespace internal -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Ordering.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Ordering.h deleted file mode 100644 index 7ea9b14d7eb..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/OrderingMethods/Ordering.h +++ /dev/null @@ -1,157 +0,0 @@ - -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ORDERING_H -#define EIGEN_ORDERING_H - -namespace Eigen { - -#include "Eigen_Colamd.h" - -namespace internal { - -/** \internal - * \ingroup OrderingMethods_Module - * \param[in] A the input non-symmetric matrix - * \param[out] symmat the symmetric pattern A^T+A from the input matrix \a A. - * FIXME: The values should not be considered here - */ -template -void ordering_helper_at_plus_a(const MatrixType& A, MatrixType& symmat) -{ - MatrixType C; - C = A.transpose(); // NOTE: Could be costly - for (int i = 0; i < C.rows(); i++) - { - for (typename MatrixType::InnerIterator it(C, i); it; ++it) - it.valueRef() = 0.0; - } - symmat = C + A; -} - -} - -#ifndef EIGEN_MPL2_ONLY - -/** \ingroup OrderingMethods_Module - * \class AMDOrdering - * - * Functor computing the \em approximate \em minimum \em degree ordering - * If the matrix is not structurally symmetric, an ordering of A^T+A is computed - * \tparam StorageIndex The type of indices of the matrix - * \sa COLAMDOrdering - */ -template -class AMDOrdering -{ - public: - typedef PermutationMatrix PermutationType; - - /** Compute the permutation vector from a sparse matrix - * This routine is much faster if the input matrix is column-major - */ - template - void operator()(const MatrixType& mat, PermutationType& perm) - { - // Compute the symmetric pattern - SparseMatrix symm; - internal::ordering_helper_at_plus_a(mat,symm); - - // Call the AMD routine - //m_mat.prune(keep_diag()); - internal::minimum_degree_ordering(symm, perm); - } - - /** Compute the permutation with a selfadjoint matrix */ - template - void operator()(const SparseSelfAdjointView& mat, PermutationType& perm) - { - SparseMatrix C; C = mat; - - // Call the AMD routine - // m_mat.prune(keep_diag()); //Remove the diagonal elements - internal::minimum_degree_ordering(C, perm); - } -}; - -#endif // EIGEN_MPL2_ONLY - -/** \ingroup OrderingMethods_Module - * \class NaturalOrdering - * - * Functor computing the natural ordering (identity) - * - * \note Returns an empty permutation matrix - * \tparam StorageIndex The type of indices of the matrix - */ -template -class NaturalOrdering -{ - public: - typedef PermutationMatrix PermutationType; - - /** Compute the permutation vector from a column-major sparse matrix */ - template - void operator()(const MatrixType& /*mat*/, PermutationType& perm) - { - perm.resize(0); - } - -}; - -/** \ingroup OrderingMethods_Module - * \class COLAMDOrdering - * - * \tparam StorageIndex The type of indices of the matrix - * - * Functor computing the \em column \em approximate \em minimum \em degree ordering - * The matrix should be in column-major and \b compressed format (see SparseMatrix::makeCompressed()). - */ -template -class COLAMDOrdering -{ - public: - typedef PermutationMatrix PermutationType; - typedef Matrix IndexVector; - - /** Compute the permutation vector \a perm form the sparse matrix \a mat - * \warning The input sparse matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()). - */ - template - void operator() (const MatrixType& mat, PermutationType& perm) - { - eigen_assert(mat.isCompressed() && "COLAMDOrdering requires a sparse matrix in compressed mode. Call .makeCompressed() before passing it to COLAMDOrdering"); - - StorageIndex m = StorageIndex(mat.rows()); - StorageIndex n = StorageIndex(mat.cols()); - StorageIndex nnz = StorageIndex(mat.nonZeros()); - // Get the recommended value of Alen to be used by colamd - StorageIndex Alen = internal::colamd_recommended(nnz, m, n); - // Set the default parameters - double knobs [COLAMD_KNOBS]; - StorageIndex stats [COLAMD_STATS]; - internal::colamd_set_defaults(knobs); - - IndexVector p(n+1), A(Alen); - for(StorageIndex i=0; i <= n; i++) p(i) = mat.outerIndexPtr()[i]; - for(StorageIndex i=0; i < nnz; i++) A(i) = mat.innerIndexPtr()[i]; - // Call Colamd routine to compute the ordering - StorageIndex info = internal::colamd(m, n, Alen, A.data(), p.data(), knobs, stats); - EIGEN_UNUSED_VARIABLE(info); - eigen_assert( info && "COLAMD failed " ); - - perm.resize(n); - for (StorageIndex i = 0; i < n; i++) perm.indices()(p(i)) = i; - } -}; - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PaStiXSupport/PaStiXSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PaStiXSupport/PaStiXSupport.h deleted file mode 100644 index 160d8a5234a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PaStiXSupport/PaStiXSupport.h +++ /dev/null @@ -1,678 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PASTIXSUPPORT_H -#define EIGEN_PASTIXSUPPORT_H - -namespace Eigen { - -#if defined(DCOMPLEX) - #define PASTIX_COMPLEX COMPLEX - #define PASTIX_DCOMPLEX DCOMPLEX -#else - #define PASTIX_COMPLEX std::complex - #define PASTIX_DCOMPLEX std::complex -#endif - -/** \ingroup PaStiXSupport_Module - * \brief Interface to the PaStix solver - * - * This class is used to solve the linear systems A.X = B via the PaStix library. - * The matrix can be either real or complex, symmetric or not. - * - * \sa TutorialSparseDirectSolvers - */ -template class PastixLU; -template class PastixLLT; -template class PastixLDLT; - -namespace internal -{ - - template struct pastix_traits; - - template - struct pastix_traits< PastixLU<_MatrixType> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - - template - struct pastix_traits< PastixLLT<_MatrixType,Options> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - - template - struct pastix_traits< PastixLDLT<_MatrixType,Options> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - - inline void eigen_pastix(pastix_data_t **pastix_data, int pastix_comm, int n, int *ptr, int *idx, float *vals, int *perm, int * invp, float *x, int nbrhs, int *iparm, double *dparm) - { - if (n == 0) { ptr = NULL; idx = NULL; vals = NULL; } - if (nbrhs == 0) {x = NULL; nbrhs=1;} - s_pastix(pastix_data, pastix_comm, n, ptr, idx, vals, perm, invp, x, nbrhs, iparm, dparm); - } - - inline void eigen_pastix(pastix_data_t **pastix_data, int pastix_comm, int n, int *ptr, int *idx, double *vals, int *perm, int * invp, double *x, int nbrhs, int *iparm, double *dparm) - { - if (n == 0) { ptr = NULL; idx = NULL; vals = NULL; } - if (nbrhs == 0) {x = NULL; nbrhs=1;} - d_pastix(pastix_data, pastix_comm, n, ptr, idx, vals, perm, invp, x, nbrhs, iparm, dparm); - } - - inline void eigen_pastix(pastix_data_t **pastix_data, int pastix_comm, int n, int *ptr, int *idx, std::complex *vals, int *perm, int * invp, std::complex *x, int nbrhs, int *iparm, double *dparm) - { - if (n == 0) { ptr = NULL; idx = NULL; vals = NULL; } - if (nbrhs == 0) {x = NULL; nbrhs=1;} - c_pastix(pastix_data, pastix_comm, n, ptr, idx, reinterpret_cast(vals), perm, invp, reinterpret_cast(x), nbrhs, iparm, dparm); - } - - inline void eigen_pastix(pastix_data_t **pastix_data, int pastix_comm, int n, int *ptr, int *idx, std::complex *vals, int *perm, int * invp, std::complex *x, int nbrhs, int *iparm, double *dparm) - { - if (n == 0) { ptr = NULL; idx = NULL; vals = NULL; } - if (nbrhs == 0) {x = NULL; nbrhs=1;} - z_pastix(pastix_data, pastix_comm, n, ptr, idx, reinterpret_cast(vals), perm, invp, reinterpret_cast(x), nbrhs, iparm, dparm); - } - - // Convert the matrix to Fortran-style Numbering - template - void c_to_fortran_numbering (MatrixType& mat) - { - if ( !(mat.outerIndexPtr()[0]) ) - { - int i; - for(i = 0; i <= mat.rows(); ++i) - ++mat.outerIndexPtr()[i]; - for(i = 0; i < mat.nonZeros(); ++i) - ++mat.innerIndexPtr()[i]; - } - } - - // Convert to C-style Numbering - template - void fortran_to_c_numbering (MatrixType& mat) - { - // Check the Numbering - if ( mat.outerIndexPtr()[0] == 1 ) - { // Convert to C-style numbering - int i; - for(i = 0; i <= mat.rows(); ++i) - --mat.outerIndexPtr()[i]; - for(i = 0; i < mat.nonZeros(); ++i) - --mat.innerIndexPtr()[i]; - } - } -} - -// This is the base class to interface with PaStiX functions. -// Users should not used this class directly. -template -class PastixBase : public SparseSolverBase -{ - protected: - typedef SparseSolverBase Base; - using Base::derived; - using Base::m_isInitialized; - public: - using Base::_solve_impl; - - typedef typename internal::pastix_traits::MatrixType _MatrixType; - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Matrix Vector; - typedef SparseMatrix ColSpMatrix; - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - - PastixBase() : m_initisOk(false), m_analysisIsOk(false), m_factorizationIsOk(false), m_pastixdata(0), m_size(0) - { - init(); - } - - ~PastixBase() - { - clean(); - } - - template - bool _solve_impl(const MatrixBase &b, MatrixBase &x) const; - - /** Returns a reference to the integer vector IPARM of PaStiX parameters - * to modify the default parameters. - * The statistics related to the different phases of factorization and solve are saved here as well - * \sa analyzePattern() factorize() - */ - Array& iparm() - { - return m_iparm; - } - - /** Return a reference to a particular index parameter of the IPARM vector - * \sa iparm() - */ - - int& iparm(int idxparam) - { - return m_iparm(idxparam); - } - - /** Returns a reference to the double vector DPARM of PaStiX parameters - * The statistics related to the different phases of factorization and solve are saved here as well - * \sa analyzePattern() factorize() - */ - Array& dparm() - { - return m_dparm; - } - - - /** Return a reference to a particular index parameter of the DPARM vector - * \sa dparm() - */ - double& dparm(int idxparam) - { - return m_dparm(idxparam); - } - - inline Index cols() const { return m_size; } - inline Index rows() const { return m_size; } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the PaStiX reports a problem - * \c InvalidInput if the input matrix is invalid - * - * \sa iparm() - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - protected: - - // Initialize the Pastix data structure, check the matrix - void init(); - - // Compute the ordering and the symbolic factorization - void analyzePattern(ColSpMatrix& mat); - - // Compute the numerical factorization - void factorize(ColSpMatrix& mat); - - // Free all the data allocated by Pastix - void clean() - { - eigen_assert(m_initisOk && "The Pastix structure should be allocated first"); - m_iparm(IPARM_START_TASK) = API_TASK_CLEAN; - m_iparm(IPARM_END_TASK) = API_TASK_CLEAN; - internal::eigen_pastix(&m_pastixdata, MPI_COMM_WORLD, 0, 0, 0, (Scalar*)0, - m_perm.data(), m_invp.data(), 0, 0, m_iparm.data(), m_dparm.data()); - } - - void compute(ColSpMatrix& mat); - - int m_initisOk; - int m_analysisIsOk; - int m_factorizationIsOk; - mutable ComputationInfo m_info; - mutable pastix_data_t *m_pastixdata; // Data structure for pastix - mutable int m_comm; // The MPI communicator identifier - mutable Array m_iparm; // integer vector for the input parameters - mutable Array m_dparm; // Scalar vector for the input parameters - mutable Matrix m_perm; // Permutation vector - mutable Matrix m_invp; // Inverse permutation vector - mutable int m_size; // Size of the matrix -}; - - /** Initialize the PaStiX data structure. - *A first call to this function fills iparm and dparm with the default PaStiX parameters - * \sa iparm() dparm() - */ -template -void PastixBase::init() -{ - m_size = 0; - m_iparm.setZero(IPARM_SIZE); - m_dparm.setZero(DPARM_SIZE); - - m_iparm(IPARM_MODIFY_PARAMETER) = API_NO; - pastix(&m_pastixdata, MPI_COMM_WORLD, - 0, 0, 0, 0, - 0, 0, 0, 1, m_iparm.data(), m_dparm.data()); - - m_iparm[IPARM_MATRIX_VERIFICATION] = API_NO; - m_iparm[IPARM_VERBOSE] = API_VERBOSE_NOT; - m_iparm[IPARM_ORDERING] = API_ORDER_SCOTCH; - m_iparm[IPARM_INCOMPLETE] = API_NO; - m_iparm[IPARM_OOC_LIMIT] = 2000; - m_iparm[IPARM_RHS_MAKING] = API_RHS_B; - m_iparm(IPARM_MATRIX_VERIFICATION) = API_NO; - - m_iparm(IPARM_START_TASK) = API_TASK_INIT; - m_iparm(IPARM_END_TASK) = API_TASK_INIT; - internal::eigen_pastix(&m_pastixdata, MPI_COMM_WORLD, 0, 0, 0, (Scalar*)0, - 0, 0, 0, 0, m_iparm.data(), m_dparm.data()); - - // Check the returned error - if(m_iparm(IPARM_ERROR_NUMBER)) { - m_info = InvalidInput; - m_initisOk = false; - } - else { - m_info = Success; - m_initisOk = true; - } -} - -template -void PastixBase::compute(ColSpMatrix& mat) -{ - eigen_assert(mat.rows() == mat.cols() && "The input matrix should be squared"); - - analyzePattern(mat); - factorize(mat); - - m_iparm(IPARM_MATRIX_VERIFICATION) = API_NO; -} - - -template -void PastixBase::analyzePattern(ColSpMatrix& mat) -{ - eigen_assert(m_initisOk && "The initialization of PaSTiX failed"); - - // clean previous calls - if(m_size>0) - clean(); - - m_size = internal::convert_index(mat.rows()); - m_perm.resize(m_size); - m_invp.resize(m_size); - - m_iparm(IPARM_START_TASK) = API_TASK_ORDERING; - m_iparm(IPARM_END_TASK) = API_TASK_ANALYSE; - internal::eigen_pastix(&m_pastixdata, MPI_COMM_WORLD, m_size, mat.outerIndexPtr(), mat.innerIndexPtr(), - mat.valuePtr(), m_perm.data(), m_invp.data(), 0, 0, m_iparm.data(), m_dparm.data()); - - // Check the returned error - if(m_iparm(IPARM_ERROR_NUMBER)) - { - m_info = NumericalIssue; - m_analysisIsOk = false; - } - else - { - m_info = Success; - m_analysisIsOk = true; - } -} - -template -void PastixBase::factorize(ColSpMatrix& mat) -{ -// if(&m_cpyMat != &mat) m_cpyMat = mat; - eigen_assert(m_analysisIsOk && "The analysis phase should be called before the factorization phase"); - m_iparm(IPARM_START_TASK) = API_TASK_NUMFACT; - m_iparm(IPARM_END_TASK) = API_TASK_NUMFACT; - m_size = internal::convert_index(mat.rows()); - - internal::eigen_pastix(&m_pastixdata, MPI_COMM_WORLD, m_size, mat.outerIndexPtr(), mat.innerIndexPtr(), - mat.valuePtr(), m_perm.data(), m_invp.data(), 0, 0, m_iparm.data(), m_dparm.data()); - - // Check the returned error - if(m_iparm(IPARM_ERROR_NUMBER)) - { - m_info = NumericalIssue; - m_factorizationIsOk = false; - m_isInitialized = false; - } - else - { - m_info = Success; - m_factorizationIsOk = true; - m_isInitialized = true; - } -} - -/* Solve the system */ -template -template -bool PastixBase::_solve_impl(const MatrixBase &b, MatrixBase &x) const -{ - eigen_assert(m_isInitialized && "The matrix should be factorized first"); - EIGEN_STATIC_ASSERT((Dest::Flags&RowMajorBit)==0, - THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); - int rhs = 1; - - x = b; /* on return, x is overwritten by the computed solution */ - - for (int i = 0; i < b.cols(); i++){ - m_iparm[IPARM_START_TASK] = API_TASK_SOLVE; - m_iparm[IPARM_END_TASK] = API_TASK_REFINE; - - internal::eigen_pastix(&m_pastixdata, MPI_COMM_WORLD, internal::convert_index(x.rows()), 0, 0, 0, - m_perm.data(), m_invp.data(), &x(0, i), rhs, m_iparm.data(), m_dparm.data()); - } - - // Check the returned error - m_info = m_iparm(IPARM_ERROR_NUMBER)==0 ? Success : NumericalIssue; - - return m_iparm(IPARM_ERROR_NUMBER)==0; -} - -/** \ingroup PaStiXSupport_Module - * \class PastixLU - * \brief Sparse direct LU solver based on PaStiX library - * - * This class is used to solve the linear systems A.X = B with a supernodal LU - * factorization in the PaStiX library. The matrix A should be squared and nonsingular - * PaStiX requires that the matrix A has a symmetric structural pattern. - * This interface can symmetrize the input matrix otherwise. - * The vectors or matrices X and B can be either dense or sparse. - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam IsStrSym Indicates if the input matrix has a symmetric pattern, default is false - * NOTE : Note that if the analysis and factorization phase are called separately, - * the input matrix will be symmetrized at each call, hence it is advised to - * symmetrize the matrix in a end-user program and set \p IsStrSym to true - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SparseLU - * - */ -template -class PastixLU : public PastixBase< PastixLU<_MatrixType> > -{ - public: - typedef _MatrixType MatrixType; - typedef PastixBase > Base; - typedef typename Base::ColSpMatrix ColSpMatrix; - typedef typename MatrixType::StorageIndex StorageIndex; - - public: - PastixLU() : Base() - { - init(); - } - - explicit PastixLU(const MatrixType& matrix):Base() - { - init(); - compute(matrix); - } - /** Compute the LU supernodal factorization of \p matrix. - * iparm and dparm can be used to tune the PaStiX parameters. - * see the PaStiX user's manual - * \sa analyzePattern() factorize() - */ - void compute (const MatrixType& matrix) - { - m_structureIsUptodate = false; - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::compute(temp); - } - /** Compute the LU symbolic factorization of \p matrix using its sparsity pattern. - * Several ordering methods can be used at this step. See the PaStiX user's manual. - * The result of this operation can be used with successive matrices having the same pattern as \p matrix - * \sa factorize() - */ - void analyzePattern(const MatrixType& matrix) - { - m_structureIsUptodate = false; - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::analyzePattern(temp); - } - - /** Compute the LU supernodal factorization of \p matrix - * WARNING The matrix \p matrix should have the same structural pattern - * as the same used in the analysis phase. - * \sa analyzePattern() - */ - void factorize(const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::factorize(temp); - } - protected: - - void init() - { - m_structureIsUptodate = false; - m_iparm(IPARM_SYM) = API_SYM_NO; - m_iparm(IPARM_FACTORIZATION) = API_FACT_LU; - } - - void grabMatrix(const MatrixType& matrix, ColSpMatrix& out) - { - if(IsStrSym) - out = matrix; - else - { - if(!m_structureIsUptodate) - { - // update the transposed structure - m_transposedStructure = matrix.transpose(); - - // Set the elements of the matrix to zero - for (Index j=0; j - * \tparam UpLo The part of the matrix to use : Lower or Upper. The default is Lower as required by PaStiX - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SimplicialLLT - */ -template -class PastixLLT : public PastixBase< PastixLLT<_MatrixType, _UpLo> > -{ - public: - typedef _MatrixType MatrixType; - typedef PastixBase > Base; - typedef typename Base::ColSpMatrix ColSpMatrix; - - public: - enum { UpLo = _UpLo }; - PastixLLT() : Base() - { - init(); - } - - explicit PastixLLT(const MatrixType& matrix):Base() - { - init(); - compute(matrix); - } - - /** Compute the L factor of the LL^T supernodal factorization of \p matrix - * \sa analyzePattern() factorize() - */ - void compute (const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::compute(temp); - } - - /** Compute the LL^T symbolic factorization of \p matrix using its sparsity pattern - * The result of this operation can be used with successive matrices having the same pattern as \p matrix - * \sa factorize() - */ - void analyzePattern(const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::analyzePattern(temp); - } - /** Compute the LL^T supernodal numerical factorization of \p matrix - * \sa analyzePattern() - */ - void factorize(const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::factorize(temp); - } - protected: - using Base::m_iparm; - - void init() - { - m_iparm(IPARM_SYM) = API_SYM_YES; - m_iparm(IPARM_FACTORIZATION) = API_FACT_LLT; - } - - void grabMatrix(const MatrixType& matrix, ColSpMatrix& out) - { - out.resize(matrix.rows(), matrix.cols()); - // Pastix supports only lower, column-major matrices - out.template selfadjointView() = matrix.template selfadjointView(); - internal::c_to_fortran_numbering(out); - } -}; - -/** \ingroup PaStiXSupport_Module - * \class PastixLDLT - * \brief A sparse direct supernodal Cholesky (LLT) factorization and solver based on the PaStiX library - * - * This class is used to solve the linear systems A.X = B via a LDL^T supernodal Cholesky factorization - * available in the PaStiX library. The matrix A should be symmetric and positive definite - * WARNING Selfadjoint complex matrices are not supported in the current version of PaStiX - * The vectors or matrices X and B can be either dense or sparse - * - * \tparam MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam UpLo The part of the matrix to use : Lower or Upper. The default is Lower as required by PaStiX - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SimplicialLDLT - */ -template -class PastixLDLT : public PastixBase< PastixLDLT<_MatrixType, _UpLo> > -{ - public: - typedef _MatrixType MatrixType; - typedef PastixBase > Base; - typedef typename Base::ColSpMatrix ColSpMatrix; - - public: - enum { UpLo = _UpLo }; - PastixLDLT():Base() - { - init(); - } - - explicit PastixLDLT(const MatrixType& matrix):Base() - { - init(); - compute(matrix); - } - - /** Compute the L and D factors of the LDL^T factorization of \p matrix - * \sa analyzePattern() factorize() - */ - void compute (const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::compute(temp); - } - - /** Compute the LDL^T symbolic factorization of \p matrix using its sparsity pattern - * The result of this operation can be used with successive matrices having the same pattern as \p matrix - * \sa factorize() - */ - void analyzePattern(const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::analyzePattern(temp); - } - /** Compute the LDL^T supernodal numerical factorization of \p matrix - * - */ - void factorize(const MatrixType& matrix) - { - ColSpMatrix temp; - grabMatrix(matrix, temp); - Base::factorize(temp); - } - - protected: - using Base::m_iparm; - - void init() - { - m_iparm(IPARM_SYM) = API_SYM_YES; - m_iparm(IPARM_FACTORIZATION) = API_FACT_LDLT; - } - - void grabMatrix(const MatrixType& matrix, ColSpMatrix& out) - { - // Pastix supports only lower, column-major matrices - out.resize(matrix.rows(), matrix.cols()); - out.template selfadjointView() = matrix.template selfadjointView(); - internal::c_to_fortran_numbering(out); - } -}; - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PardisoSupport/PardisoSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PardisoSupport/PardisoSupport.h deleted file mode 100644 index 98d0e3f21ab..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/PardisoSupport/PardisoSupport.h +++ /dev/null @@ -1,544 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to Intel(R) MKL PARDISO - ******************************************************************************** -*/ - -#ifndef EIGEN_PARDISOSUPPORT_H -#define EIGEN_PARDISOSUPPORT_H - -namespace Eigen { - -template class PardisoLU; -template class PardisoLLT; -template class PardisoLDLT; - -namespace internal -{ - template - struct pardiso_run_selector - { - static IndexType run( _MKL_DSS_HANDLE_t pt, IndexType maxfct, IndexType mnum, IndexType type, IndexType phase, IndexType n, void *a, - IndexType *ia, IndexType *ja, IndexType *perm, IndexType nrhs, IndexType *iparm, IndexType msglvl, void *b, void *x) - { - IndexType error = 0; - ::pardiso(pt, &maxfct, &mnum, &type, &phase, &n, a, ia, ja, perm, &nrhs, iparm, &msglvl, b, x, &error); - return error; - } - }; - template<> - struct pardiso_run_selector - { - typedef long long int IndexType; - static IndexType run( _MKL_DSS_HANDLE_t pt, IndexType maxfct, IndexType mnum, IndexType type, IndexType phase, IndexType n, void *a, - IndexType *ia, IndexType *ja, IndexType *perm, IndexType nrhs, IndexType *iparm, IndexType msglvl, void *b, void *x) - { - IndexType error = 0; - ::pardiso_64(pt, &maxfct, &mnum, &type, &phase, &n, a, ia, ja, perm, &nrhs, iparm, &msglvl, b, x, &error); - return error; - } - }; - - template struct pardiso_traits; - - template - struct pardiso_traits< PardisoLU<_MatrixType> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - - template - struct pardiso_traits< PardisoLLT<_MatrixType, Options> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - - template - struct pardiso_traits< PardisoLDLT<_MatrixType, Options> > - { - typedef _MatrixType MatrixType; - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef typename _MatrixType::StorageIndex StorageIndex; - }; - -} // end namespace internal - -template -class PardisoImpl : public SparseSolverBase -{ - protected: - typedef SparseSolverBase Base; - using Base::derived; - using Base::m_isInitialized; - - typedef internal::pardiso_traits Traits; - public: - using Base::_solve_impl; - - typedef typename Traits::MatrixType MatrixType; - typedef typename Traits::Scalar Scalar; - typedef typename Traits::RealScalar RealScalar; - typedef typename Traits::StorageIndex StorageIndex; - typedef SparseMatrix SparseMatrixType; - typedef Matrix VectorType; - typedef Matrix IntRowVectorType; - typedef Matrix IntColVectorType; - typedef Array ParameterType; - enum { - ScalarIsComplex = NumTraits::IsComplex, - ColsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic - }; - - PardisoImpl() - { - eigen_assert((sizeof(StorageIndex) >= sizeof(_INTEGER_t) && sizeof(StorageIndex) <= 8) && "Non-supported index type"); - m_iparm.setZero(); - m_msglvl = 0; // No output - m_isInitialized = false; - } - - ~PardisoImpl() - { - pardisoRelease(); - } - - inline Index cols() const { return m_size; } - inline Index rows() const { return m_size; } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the matrix appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - /** \warning for advanced usage only. - * \returns a reference to the parameter array controlling PARDISO. - * See the PARDISO manual to know how to use it. */ - ParameterType& pardisoParameterArray() - { - return m_iparm; - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - Derived& analyzePattern(const MatrixType& matrix); - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - Derived& factorize(const MatrixType& matrix); - - Derived& compute(const MatrixType& matrix); - - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const; - - protected: - void pardisoRelease() - { - if(m_isInitialized) // Factorization ran at least once - { - internal::pardiso_run_selector::run(m_pt, 1, 1, m_type, -1, internal::convert_index(m_size),0, 0, 0, m_perm.data(), 0, - m_iparm.data(), m_msglvl, NULL, NULL); - m_isInitialized = false; - } - } - - void pardisoInit(int type) - { - m_type = type; - EIGEN_USING_STD_MATH(abs); - bool symmetric = abs(m_type) < 10; - m_iparm[0] = 1; // No solver default - m_iparm[1] = 2; // use Metis for the ordering - m_iparm[2] = 0; // Reserved. Set to zero. (??Numbers of processors, value of OMP_NUM_THREADS??) - m_iparm[3] = 0; // No iterative-direct algorithm - m_iparm[4] = 0; // No user fill-in reducing permutation - m_iparm[5] = 0; // Write solution into x, b is left unchanged - m_iparm[6] = 0; // Not in use - m_iparm[7] = 2; // Max numbers of iterative refinement steps - m_iparm[8] = 0; // Not in use - m_iparm[9] = 13; // Perturb the pivot elements with 1E-13 - m_iparm[10] = symmetric ? 0 : 1; // Use nonsymmetric permutation and scaling MPS - m_iparm[11] = 0; // Not in use - m_iparm[12] = symmetric ? 0 : 1; // Maximum weighted matching algorithm is switched-off (default for symmetric). - // Try m_iparm[12] = 1 in case of inappropriate accuracy - m_iparm[13] = 0; // Output: Number of perturbed pivots - m_iparm[14] = 0; // Not in use - m_iparm[15] = 0; // Not in use - m_iparm[16] = 0; // Not in use - m_iparm[17] = -1; // Output: Number of nonzeros in the factor LU - m_iparm[18] = -1; // Output: Mflops for LU factorization - m_iparm[19] = 0; // Output: Numbers of CG Iterations - - m_iparm[20] = 0; // 1x1 pivoting - m_iparm[26] = 0; // No matrix checker - m_iparm[27] = (sizeof(RealScalar) == 4) ? 1 : 0; - m_iparm[34] = 1; // C indexing - m_iparm[36] = 0; // CSR - m_iparm[59] = 0; // 0 - In-Core ; 1 - Automatic switch between In-Core and Out-of-Core modes ; 2 - Out-of-Core - - memset(m_pt, 0, sizeof(m_pt)); - } - - protected: - // cached data to reduce reallocation, etc. - - void manageErrorCode(Index error) const - { - switch(error) - { - case 0: - m_info = Success; - break; - case -4: - case -7: - m_info = NumericalIssue; - break; - default: - m_info = InvalidInput; - } - } - - mutable SparseMatrixType m_matrix; - mutable ComputationInfo m_info; - bool m_analysisIsOk, m_factorizationIsOk; - StorageIndex m_type, m_msglvl; - mutable void *m_pt[64]; - mutable ParameterType m_iparm; - mutable IntColVectorType m_perm; - Index m_size; - -}; - -template -Derived& PardisoImpl::compute(const MatrixType& a) -{ - m_size = a.rows(); - eigen_assert(a.rows() == a.cols()); - - pardisoRelease(); - m_perm.setZero(m_size); - derived().getMatrix(a); - - Index error; - error = internal::pardiso_run_selector::run(m_pt, 1, 1, m_type, 12, internal::convert_index(m_size), - m_matrix.valuePtr(), m_matrix.outerIndexPtr(), m_matrix.innerIndexPtr(), - m_perm.data(), 0, m_iparm.data(), m_msglvl, NULL, NULL); - manageErrorCode(error); - m_analysisIsOk = true; - m_factorizationIsOk = true; - m_isInitialized = true; - return derived(); -} - -template -Derived& PardisoImpl::analyzePattern(const MatrixType& a) -{ - m_size = a.rows(); - eigen_assert(m_size == a.cols()); - - pardisoRelease(); - m_perm.setZero(m_size); - derived().getMatrix(a); - - Index error; - error = internal::pardiso_run_selector::run(m_pt, 1, 1, m_type, 11, internal::convert_index(m_size), - m_matrix.valuePtr(), m_matrix.outerIndexPtr(), m_matrix.innerIndexPtr(), - m_perm.data(), 0, m_iparm.data(), m_msglvl, NULL, NULL); - - manageErrorCode(error); - m_analysisIsOk = true; - m_factorizationIsOk = false; - m_isInitialized = true; - return derived(); -} - -template -Derived& PardisoImpl::factorize(const MatrixType& a) -{ - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - eigen_assert(m_size == a.rows() && m_size == a.cols()); - - derived().getMatrix(a); - - Index error; - error = internal::pardiso_run_selector::run(m_pt, 1, 1, m_type, 22, internal::convert_index(m_size), - m_matrix.valuePtr(), m_matrix.outerIndexPtr(), m_matrix.innerIndexPtr(), - m_perm.data(), 0, m_iparm.data(), m_msglvl, NULL, NULL); - - manageErrorCode(error); - m_factorizationIsOk = true; - return derived(); -} - -template -template -void PardisoImpl::_solve_impl(const MatrixBase &b, MatrixBase& x) const -{ - if(m_iparm[0] == 0) // Factorization was not computed - { - m_info = InvalidInput; - return; - } - - //Index n = m_matrix.rows(); - Index nrhs = Index(b.cols()); - eigen_assert(m_size==b.rows()); - eigen_assert(((MatrixBase::Flags & RowMajorBit) == 0 || nrhs == 1) && "Row-major right hand sides are not supported"); - eigen_assert(((MatrixBase::Flags & RowMajorBit) == 0 || nrhs == 1) && "Row-major matrices of unknowns are not supported"); - eigen_assert(((nrhs == 1) || b.outerStride() == b.rows())); - - -// switch (transposed) { -// case SvNoTrans : m_iparm[11] = 0 ; break; -// case SvTranspose : m_iparm[11] = 2 ; break; -// case SvAdjoint : m_iparm[11] = 1 ; break; -// default: -// //std::cerr << "Eigen: transposition option \"" << transposed << "\" not supported by the PARDISO backend\n"; -// m_iparm[11] = 0; -// } - - Scalar* rhs_ptr = const_cast(b.derived().data()); - Matrix tmp; - - // Pardiso cannot solve in-place - if(rhs_ptr == x.derived().data()) - { - tmp = b; - rhs_ptr = tmp.data(); - } - - Index error; - error = internal::pardiso_run_selector::run(m_pt, 1, 1, m_type, 33, internal::convert_index(m_size), - m_matrix.valuePtr(), m_matrix.outerIndexPtr(), m_matrix.innerIndexPtr(), - m_perm.data(), internal::convert_index(nrhs), m_iparm.data(), m_msglvl, - rhs_ptr, x.derived().data()); - - manageErrorCode(error); -} - - -/** \ingroup PardisoSupport_Module - * \class PardisoLU - * \brief A sparse direct LU factorization and solver based on the PARDISO library - * - * This class allows to solve for A.X = B sparse linear problems via a direct LU factorization - * using the Intel MKL PARDISO library. The sparse matrix A must be squared and invertible. - * The vectors or matrices X and B can be either dense or sparse. - * - * By default, it runs in in-core mode. To enable PARDISO's out-of-core feature, set: - * \code solver.pardisoParameterArray()[59] = 1; \endcode - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SparseLU - */ -template -class PardisoLU : public PardisoImpl< PardisoLU > -{ - protected: - typedef PardisoImpl Base; - typedef typename Base::Scalar Scalar; - typedef typename Base::RealScalar RealScalar; - using Base::pardisoInit; - using Base::m_matrix; - friend class PardisoImpl< PardisoLU >; - - public: - - using Base::compute; - using Base::solve; - - PardisoLU() - : Base() - { - pardisoInit(Base::ScalarIsComplex ? 13 : 11); - } - - explicit PardisoLU(const MatrixType& matrix) - : Base() - { - pardisoInit(Base::ScalarIsComplex ? 13 : 11); - compute(matrix); - } - protected: - void getMatrix(const MatrixType& matrix) - { - m_matrix = matrix; - m_matrix.makeCompressed(); - } -}; - -/** \ingroup PardisoSupport_Module - * \class PardisoLLT - * \brief A sparse direct Cholesky (LLT) factorization and solver based on the PARDISO library - * - * This class allows to solve for A.X = B sparse linear problems via a LL^T Cholesky factorization - * using the Intel MKL PARDISO library. The sparse matrix A must be selfajoint and positive definite. - * The vectors or matrices X and B can be either dense or sparse. - * - * By default, it runs in in-core mode. To enable PARDISO's out-of-core feature, set: - * \code solver.pardisoParameterArray()[59] = 1; \endcode - * - * \tparam MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam UpLo can be any bitwise combination of Upper, Lower. The default is Upper, meaning only the upper triangular part has to be used. - * Upper|Lower can be used to tell both triangular parts can be used as input. - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SimplicialLLT - */ -template -class PardisoLLT : public PardisoImpl< PardisoLLT > -{ - protected: - typedef PardisoImpl< PardisoLLT > Base; - typedef typename Base::Scalar Scalar; - typedef typename Base::RealScalar RealScalar; - using Base::pardisoInit; - using Base::m_matrix; - friend class PardisoImpl< PardisoLLT >; - - public: - - typedef typename Base::StorageIndex StorageIndex; - enum { UpLo = _UpLo }; - using Base::compute; - - PardisoLLT() - : Base() - { - pardisoInit(Base::ScalarIsComplex ? 4 : 2); - } - - explicit PardisoLLT(const MatrixType& matrix) - : Base() - { - pardisoInit(Base::ScalarIsComplex ? 4 : 2); - compute(matrix); - } - - protected: - - void getMatrix(const MatrixType& matrix) - { - // PARDISO supports only upper, row-major matrices - PermutationMatrix p_null; - m_matrix.resize(matrix.rows(), matrix.cols()); - m_matrix.template selfadjointView() = matrix.template selfadjointView().twistedBy(p_null); - m_matrix.makeCompressed(); - } -}; - -/** \ingroup PardisoSupport_Module - * \class PardisoLDLT - * \brief A sparse direct Cholesky (LDLT) factorization and solver based on the PARDISO library - * - * This class allows to solve for A.X = B sparse linear problems via a LDL^T Cholesky factorization - * using the Intel MKL PARDISO library. The sparse matrix A is assumed to be selfajoint and positive definite. - * For complex matrices, A can also be symmetric only, see the \a Options template parameter. - * The vectors or matrices X and B can be either dense or sparse. - * - * By default, it runs in in-core mode. To enable PARDISO's out-of-core feature, set: - * \code solver.pardisoParameterArray()[59] = 1; \endcode - * - * \tparam MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam Options can be any bitwise combination of Upper, Lower, and Symmetric. The default is Upper, meaning only the upper triangular part has to be used. - * Symmetric can be used for symmetric, non-selfadjoint complex matrices, the default being to assume a selfadjoint matrix. - * Upper|Lower can be used to tell both triangular parts can be used as input. - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SimplicialLDLT - */ -template -class PardisoLDLT : public PardisoImpl< PardisoLDLT > -{ - protected: - typedef PardisoImpl< PardisoLDLT > Base; - typedef typename Base::Scalar Scalar; - typedef typename Base::RealScalar RealScalar; - using Base::pardisoInit; - using Base::m_matrix; - friend class PardisoImpl< PardisoLDLT >; - - public: - - typedef typename Base::StorageIndex StorageIndex; - using Base::compute; - enum { UpLo = Options&(Upper|Lower) }; - - PardisoLDLT() - : Base() - { - pardisoInit(Base::ScalarIsComplex ? ( bool(Options&Symmetric) ? 6 : -4 ) : -2); - } - - explicit PardisoLDLT(const MatrixType& matrix) - : Base() - { - pardisoInit(Base::ScalarIsComplex ? ( bool(Options&Symmetric) ? 6 : -4 ) : -2); - compute(matrix); - } - - void getMatrix(const MatrixType& matrix) - { - // PARDISO supports only upper, row-major matrices - PermutationMatrix p_null; - m_matrix.resize(matrix.rows(), matrix.cols()); - m_matrix.template selfadjointView() = matrix.template selfadjointView().twistedBy(p_null); - m_matrix.makeCompressed(); - } -}; - -} // end namespace Eigen - -#endif // EIGEN_PARDISOSUPPORT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR.h deleted file mode 100644 index a7b47d55dc5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR.h +++ /dev/null @@ -1,653 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COLPIVOTINGHOUSEHOLDERQR_H -#define EIGEN_COLPIVOTINGHOUSEHOLDERQR_H - -namespace Eigen { - -namespace internal { -template struct traits > - : traits<_MatrixType> -{ - enum { Flags = 0 }; -}; - -} // end namespace internal - -/** \ingroup QR_Module - * - * \class ColPivHouseholderQR - * - * \brief Householder rank-revealing QR decomposition of a matrix with column-pivoting - * - * \tparam _MatrixType the type of the matrix of which we are computing the QR decomposition - * - * This class performs a rank-revealing QR decomposition of a matrix \b A into matrices \b P, \b Q and \b R - * such that - * \f[ - * \mathbf{A} \, \mathbf{P} = \mathbf{Q} \, \mathbf{R} - * \f] - * by using Householder transformations. Here, \b P is a permutation matrix, \b Q a unitary matrix and \b R an - * upper triangular matrix. - * - * This decomposition performs column pivoting in order to be rank-revealing and improve - * numerical stability. It is slower than HouseholderQR, and faster than FullPivHouseholderQR. - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::colPivHouseholderQr() - */ -template class ColPivHouseholderQR -{ - public: - - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - // FIXME should be int - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename internal::plain_diag_type::type HCoeffsType; - typedef PermutationMatrix PermutationType; - typedef typename internal::plain_row_type::type IntRowVectorType; - typedef typename internal::plain_row_type::type RowVectorType; - typedef typename internal::plain_row_type::type RealRowVectorType; - typedef HouseholderSequence::type> HouseholderSequenceType; - typedef typename MatrixType::PlainObject PlainObject; - - private: - - typedef typename PermutationType::StorageIndex PermIndexType; - - public: - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via ColPivHouseholderQR::compute(const MatrixType&). - */ - ColPivHouseholderQR() - : m_qr(), - m_hCoeffs(), - m_colsPermutation(), - m_colsTranspositions(), - m_temp(), - m_colNormsUpdated(), - m_colNormsDirect(), - m_isInitialized(false), - m_usePrescribedThreshold(false) {} - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa ColPivHouseholderQR() - */ - ColPivHouseholderQR(Index rows, Index cols) - : m_qr(rows, cols), - m_hCoeffs((std::min)(rows,cols)), - m_colsPermutation(PermIndexType(cols)), - m_colsTranspositions(cols), - m_temp(cols), - m_colNormsUpdated(cols), - m_colNormsDirect(cols), - m_isInitialized(false), - m_usePrescribedThreshold(false) {} - - /** \brief Constructs a QR factorization from a given matrix - * - * This constructor computes the QR factorization of the matrix \a matrix by calling - * the method compute(). It is a short cut for: - * - * \code - * ColPivHouseholderQR qr(matrix.rows(), matrix.cols()); - * qr.compute(matrix); - * \endcode - * - * \sa compute() - */ - template - explicit ColPivHouseholderQR(const EigenBase& matrix) - : m_qr(matrix.rows(), matrix.cols()), - m_hCoeffs((std::min)(matrix.rows(),matrix.cols())), - m_colsPermutation(PermIndexType(matrix.cols())), - m_colsTranspositions(matrix.cols()), - m_temp(matrix.cols()), - m_colNormsUpdated(matrix.cols()), - m_colNormsDirect(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) - { - compute(matrix.derived()); - } - - /** \brief Constructs a QR factorization from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. - * - * \sa ColPivHouseholderQR(const EigenBase&) - */ - template - explicit ColPivHouseholderQR(EigenBase& matrix) - : m_qr(matrix.derived()), - m_hCoeffs((std::min)(matrix.rows(),matrix.cols())), - m_colsPermutation(PermIndexType(matrix.cols())), - m_colsTranspositions(matrix.cols()), - m_temp(matrix.cols()), - m_colNormsUpdated(matrix.cols()), - m_colNormsDirect(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) - { - computeInPlace(); - } - - /** This method finds a solution x to the equation Ax=b, where A is the matrix of which - * *this is the QR decomposition, if any exists. - * - * \param b the right-hand-side of the equation to solve. - * - * \returns a solution. - * - * \note_about_checking_solutions - * - * \note_about_arbitrary_choice_of_solution - * - * Example: \include ColPivHouseholderQR_solve.cpp - * Output: \verbinclude ColPivHouseholderQR_solve.out - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return Solve(*this, b.derived()); - } - - HouseholderSequenceType householderQ() const; - HouseholderSequenceType matrixQ() const - { - return householderQ(); - } - - /** \returns a reference to the matrix where the Householder QR decomposition is stored - */ - const MatrixType& matrixQR() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return m_qr; - } - - /** \returns a reference to the matrix where the result Householder QR is stored - * \warning The strict lower part of this matrix contains internal values. - * Only the upper triangular part should be referenced. To get it, use - * \code matrixR().template triangularView() \endcode - * For rank-deficient matrices, use - * \code - * matrixR().topLeftCorner(rank(), rank()).template triangularView() - * \endcode - */ - const MatrixType& matrixR() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return m_qr; - } - - template - ColPivHouseholderQR& compute(const EigenBase& matrix); - - /** \returns a const reference to the column permutation matrix */ - const PermutationType& colsPermutation() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return m_colsPermutation; - } - - /** \returns the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * One way to work around that is to use logAbsDeterminant() instead. - * - * \sa logAbsDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar absDeterminant() const; - - /** \returns the natural log of the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \note This method is useful to work around the risk of overflow/underflow that's inherent - * to determinant computation. - * - * \sa absDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar logAbsDeterminant() const; - - /** \returns the rank of the matrix of which *this is the QR decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index rank() const - { - using std::abs; - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - RealScalar premultiplied_threshold = abs(m_maxpivot) * threshold(); - Index result = 0; - for(Index i = 0; i < m_nonzero_pivots; ++i) - result += (abs(m_qr.coeff(i,i)) > premultiplied_threshold); - return result; - } - - /** \returns the dimension of the kernel of the matrix of which *this is the QR decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index dimensionOfKernel() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return cols() - rank(); - } - - /** \returns true if the matrix of which *this is the QR decomposition represents an injective - * linear map, i.e. has trivial kernel; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInjective() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return rank() == cols(); - } - - /** \returns true if the matrix of which *this is the QR decomposition represents a surjective - * linear map; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isSurjective() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return rank() == rows(); - } - - /** \returns true if the matrix of which *this is the QR decomposition is invertible. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInvertible() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return isInjective() && isSurjective(); - } - - /** \returns the inverse of the matrix of which *this is the QR decomposition. - * - * \note If this matrix is not invertible, the returned matrix has undefined coefficients. - * Use isInvertible() to first determine whether this matrix is invertible. - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return Inverse(*this); - } - - inline Index rows() const { return m_qr.rows(); } - inline Index cols() const { return m_qr.cols(); } - - /** \returns a const reference to the vector of Householder coefficients used to represent the factor \c Q. - * - * For advanced uses only. - */ - const HCoeffsType& hCoeffs() const { return m_hCoeffs; } - - /** Allows to prescribe a threshold to be used by certain methods, such as rank(), - * who need to determine when pivots are to be considered nonzero. This is not used for the - * QR decomposition itself. - * - * When it needs to get the threshold value, Eigen calls threshold(). By default, this - * uses a formula to automatically determine a reasonable threshold. - * Once you have called the present method setThreshold(const RealScalar&), - * your value is used instead. - * - * \param threshold The new value to use as the threshold. - * - * A pivot will be considered nonzero if its absolute value is strictly greater than - * \f$ \vert pivot \vert \leqslant threshold \times \vert maxpivot \vert \f$ - * where maxpivot is the biggest pivot. - * - * If you want to come back to the default behavior, call setThreshold(Default_t) - */ - ColPivHouseholderQR& setThreshold(const RealScalar& threshold) - { - m_usePrescribedThreshold = true; - m_prescribedThreshold = threshold; - return *this; - } - - /** Allows to come back to the default behavior, letting Eigen use its default formula for - * determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code qr.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - ColPivHouseholderQR& setThreshold(Default_t) - { - m_usePrescribedThreshold = false; - return *this; - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const - { - eigen_assert(m_isInitialized || m_usePrescribedThreshold); - return m_usePrescribedThreshold ? m_prescribedThreshold - // this formula comes from experimenting (see "LU precision tuning" thread on the list) - // and turns out to be identical to Higham's formula used already in LDLt. - : NumTraits::epsilon() * RealScalar(m_qr.diagonalSize()); - } - - /** \returns the number of nonzero pivots in the QR decomposition. - * Here nonzero is meant in the exact sense, not in a fuzzy sense. - * So that notion isn't really intrinsically interesting, but it is - * still useful when implementing algorithms. - * - * \sa rank() - */ - inline Index nonzeroPivots() const - { - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return m_nonzero_pivots; - } - - /** \returns the absolute value of the biggest pivot, i.e. the biggest - * diagonal coefficient of R. - */ - RealScalar maxPivot() const { return m_maxpivot; } - - /** \brief Reports whether the QR factorization was succesful. - * - * \note This function always returns \c Success. It is provided for compatibility - * with other factorization routines. - * \returns \c Success - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return Success; - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - #endif - - protected: - - friend class CompleteOrthogonalDecomposition; - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - MatrixType m_qr; - HCoeffsType m_hCoeffs; - PermutationType m_colsPermutation; - IntRowVectorType m_colsTranspositions; - RowVectorType m_temp; - RealRowVectorType m_colNormsUpdated; - RealRowVectorType m_colNormsDirect; - bool m_isInitialized, m_usePrescribedThreshold; - RealScalar m_prescribedThreshold, m_maxpivot; - Index m_nonzero_pivots; - Index m_det_pq; -}; - -template -typename MatrixType::RealScalar ColPivHouseholderQR::absDeterminant() const -{ - using std::abs; - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return abs(m_qr.diagonal().prod()); -} - -template -typename MatrixType::RealScalar ColPivHouseholderQR::logAbsDeterminant() const -{ - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return m_qr.diagonal().cwiseAbs().array().log().sum(); -} - -/** Performs the QR factorization of the given matrix \a matrix. The result of - * the factorization is stored into \c *this, and a reference to \c *this - * is returned. - * - * \sa class ColPivHouseholderQR, ColPivHouseholderQR(const MatrixType&) - */ -template -template -ColPivHouseholderQR& ColPivHouseholderQR::compute(const EigenBase& matrix) -{ - m_qr = matrix.derived(); - computeInPlace(); - return *this; -} - -template -void ColPivHouseholderQR::computeInPlace() -{ - check_template_parameters(); - - // the column permutation is stored as int indices, so just to be sure: - eigen_assert(m_qr.cols()<=NumTraits::highest()); - - using std::abs; - - Index rows = m_qr.rows(); - Index cols = m_qr.cols(); - Index size = m_qr.diagonalSize(); - - m_hCoeffs.resize(size); - - m_temp.resize(cols); - - m_colsTranspositions.resize(m_qr.cols()); - Index number_of_transpositions = 0; - - m_colNormsUpdated.resize(cols); - m_colNormsDirect.resize(cols); - for (Index k = 0; k < cols; ++k) { - // colNormsDirect(k) caches the most recent directly computed norm of - // column k. - m_colNormsDirect.coeffRef(k) = m_qr.col(k).norm(); - m_colNormsUpdated.coeffRef(k) = m_colNormsDirect.coeffRef(k); - } - - RealScalar threshold_helper = numext::abs2(m_colNormsUpdated.maxCoeff() * NumTraits::epsilon()) / RealScalar(rows); - RealScalar norm_downdate_threshold = numext::sqrt(NumTraits::epsilon()); - - m_nonzero_pivots = size; // the generic case is that in which all pivots are nonzero (invertible case) - m_maxpivot = RealScalar(0); - - for(Index k = 0; k < size; ++k) - { - // first, we look up in our table m_colNormsUpdated which column has the biggest norm - Index biggest_col_index; - RealScalar biggest_col_sq_norm = numext::abs2(m_colNormsUpdated.tail(cols-k).maxCoeff(&biggest_col_index)); - biggest_col_index += k; - - // Track the number of meaningful pivots but do not stop the decomposition to make - // sure that the initial matrix is properly reproduced. See bug 941. - if(m_nonzero_pivots==size && biggest_col_sq_norm < threshold_helper * RealScalar(rows-k)) - m_nonzero_pivots = k; - - // apply the transposition to the columns - m_colsTranspositions.coeffRef(k) = biggest_col_index; - if(k != biggest_col_index) { - m_qr.col(k).swap(m_qr.col(biggest_col_index)); - std::swap(m_colNormsUpdated.coeffRef(k), m_colNormsUpdated.coeffRef(biggest_col_index)); - std::swap(m_colNormsDirect.coeffRef(k), m_colNormsDirect.coeffRef(biggest_col_index)); - ++number_of_transpositions; - } - - // generate the householder vector, store it below the diagonal - RealScalar beta; - m_qr.col(k).tail(rows-k).makeHouseholderInPlace(m_hCoeffs.coeffRef(k), beta); - - // apply the householder transformation to the diagonal coefficient - m_qr.coeffRef(k,k) = beta; - - // remember the maximum absolute value of diagonal coefficients - if(abs(beta) > m_maxpivot) m_maxpivot = abs(beta); - - // apply the householder transformation - m_qr.bottomRightCorner(rows-k, cols-k-1) - .applyHouseholderOnTheLeft(m_qr.col(k).tail(rows-k-1), m_hCoeffs.coeffRef(k), &m_temp.coeffRef(k+1)); - - // update our table of norms of the columns - for (Index j = k + 1; j < cols; ++j) { - // The following implements the stable norm downgrade step discussed in - // http://www.netlib.org/lapack/lawnspdf/lawn176.pdf - // and used in LAPACK routines xGEQPF and xGEQP3. - // See lines 278-297 in http://www.netlib.org/lapack/explore-html/dc/df4/sgeqpf_8f_source.html - if (m_colNormsUpdated.coeffRef(j) != RealScalar(0)) { - RealScalar temp = abs(m_qr.coeffRef(k, j)) / m_colNormsUpdated.coeffRef(j); - temp = (RealScalar(1) + temp) * (RealScalar(1) - temp); - temp = temp < RealScalar(0) ? RealScalar(0) : temp; - RealScalar temp2 = temp * numext::abs2(m_colNormsUpdated.coeffRef(j) / - m_colNormsDirect.coeffRef(j)); - if (temp2 <= norm_downdate_threshold) { - // The updated norm has become too inaccurate so re-compute the column - // norm directly. - m_colNormsDirect.coeffRef(j) = m_qr.col(j).tail(rows - k - 1).norm(); - m_colNormsUpdated.coeffRef(j) = m_colNormsDirect.coeffRef(j); - } else { - m_colNormsUpdated.coeffRef(j) *= numext::sqrt(temp); - } - } - } - } - - m_colsPermutation.setIdentity(PermIndexType(cols)); - for(PermIndexType k = 0; k < size/*m_nonzero_pivots*/; ++k) - m_colsPermutation.applyTranspositionOnTheRight(k, PermIndexType(m_colsTranspositions.coeff(k))); - - m_det_pq = (number_of_transpositions%2) ? -1 : 1; - m_isInitialized = true; -} - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void ColPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - eigen_assert(rhs.rows() == rows()); - - const Index nonzero_pivots = nonzeroPivots(); - - if(nonzero_pivots == 0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs); - - // Note that the matrix Q = H_0^* H_1^*... so its inverse is Q^* = (H_0 H_1 ...)^T - c.applyOnTheLeft(householderSequence(m_qr, m_hCoeffs) - .setLength(nonzero_pivots) - .transpose() - ); - - m_qr.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .solveInPlace(c.topRows(nonzero_pivots)); - - for(Index i = 0; i < nonzero_pivots; ++i) dst.row(m_colsPermutation.indices().coeff(i)) = c.row(i); - for(Index i = nonzero_pivots; i < cols(); ++i) dst.row(m_colsPermutation.indices().coeff(i)).setZero(); -} -#endif - -namespace internal { - -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef ColPivHouseholderQR QrType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; - -} // end namespace internal - -/** \returns the matrix Q as a sequence of householder transformations. - * You can extract the meaningful part only by using: - * \code qr.householderQ().setLength(qr.nonzeroPivots()) \endcode*/ -template -typename ColPivHouseholderQR::HouseholderSequenceType ColPivHouseholderQR - ::householderQ() const -{ - eigen_assert(m_isInitialized && "ColPivHouseholderQR is not initialized."); - return HouseholderSequenceType(m_qr, m_hCoeffs.conjugate()); -} - -/** \return the column-pivoting Householder QR decomposition of \c *this. - * - * \sa class ColPivHouseholderQR - */ -template -const ColPivHouseholderQR::PlainObject> -MatrixBase::colPivHouseholderQr() const -{ - return ColPivHouseholderQR(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_COLPIVOTINGHOUSEHOLDERQR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR_LAPACKE.h deleted file mode 100644 index 4e9651f83d5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/ColPivHouseholderQR_LAPACKE.h +++ /dev/null @@ -1,97 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * Householder QR decomposition of a matrix with column pivoting based on - * LAPACKE_?geqp3 function. - ******************************************************************************** -*/ - -#ifndef EIGEN_COLPIVOTINGHOUSEHOLDERQR_LAPACKE_H -#define EIGEN_COLPIVOTINGHOUSEHOLDERQR_LAPACKE_H - -namespace Eigen { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_QR_COLPIV(EIGTYPE, LAPACKE_TYPE, LAPACKE_PREFIX, EIGCOLROW, LAPACKE_COLROW) \ -template<> template inline \ -ColPivHouseholderQR >& \ -ColPivHouseholderQR >::compute( \ - const EigenBase& matrix) \ -\ -{ \ - using std::abs; \ - typedef Matrix MatrixType; \ - typedef MatrixType::RealScalar RealScalar; \ - Index rows = matrix.rows();\ - Index cols = matrix.cols();\ -\ - m_qr = matrix;\ - Index size = m_qr.diagonalSize();\ - m_hCoeffs.resize(size);\ -\ - m_colsTranspositions.resize(cols);\ - /*Index number_of_transpositions = 0;*/ \ -\ - m_nonzero_pivots = 0; \ - m_maxpivot = RealScalar(0);\ - m_colsPermutation.resize(cols); \ - m_colsPermutation.indices().setZero(); \ -\ - lapack_int lda = internal::convert_index(m_qr.outerStride()); \ - lapack_int matrix_order = LAPACKE_COLROW; \ - LAPACKE_##LAPACKE_PREFIX##geqp3( matrix_order, internal::convert_index(rows), internal::convert_index(cols), \ - (LAPACKE_TYPE*)m_qr.data(), lda, (lapack_int*)m_colsPermutation.indices().data(), (LAPACKE_TYPE*)m_hCoeffs.data()); \ - m_isInitialized = true; \ - m_maxpivot=m_qr.diagonal().cwiseAbs().maxCoeff(); \ - m_hCoeffs.adjointInPlace(); \ - RealScalar premultiplied_threshold = abs(m_maxpivot) * threshold(); \ - lapack_int *perm = m_colsPermutation.indices().data(); \ - for(Index i=0;i premultiplied_threshold);\ - } \ - for(Index i=0;i -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLETEORTHOGONALDECOMPOSITION_H -#define EIGEN_COMPLETEORTHOGONALDECOMPOSITION_H - -namespace Eigen { - -namespace internal { -template -struct traits > - : traits<_MatrixType> { - enum { Flags = 0 }; -}; - -} // end namespace internal - -/** \ingroup QR_Module - * - * \class CompleteOrthogonalDecomposition - * - * \brief Complete orthogonal decomposition (COD) of a matrix. - * - * \param MatrixType the type of the matrix of which we are computing the COD. - * - * This class performs a rank-revealing complete orthogonal decomposition of a - * matrix \b A into matrices \b P, \b Q, \b T, and \b Z such that - * \f[ - * \mathbf{A} \, \mathbf{P} = \mathbf{Q} \, - * \begin{bmatrix} \mathbf{T} & \mathbf{0} \\ - * \mathbf{0} & \mathbf{0} \end{bmatrix} \, \mathbf{Z} - * \f] - * by using Householder transformations. Here, \b P is a permutation matrix, - * \b Q and \b Z are unitary matrices and \b T an upper triangular matrix of - * size rank-by-rank. \b A may be rank deficient. - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::completeOrthogonalDecomposition() - */ -template -class CompleteOrthogonalDecomposition { - public: - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename internal::plain_diag_type::type HCoeffsType; - typedef PermutationMatrix - PermutationType; - typedef typename internal::plain_row_type::type - IntRowVectorType; - typedef typename internal::plain_row_type::type RowVectorType; - typedef typename internal::plain_row_type::type - RealRowVectorType; - typedef HouseholderSequence< - MatrixType, typename internal::remove_all< - typename HCoeffsType::ConjugateReturnType>::type> - HouseholderSequenceType; - typedef typename MatrixType::PlainObject PlainObject; - - private: - typedef typename PermutationType::Index PermIndexType; - - public: - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via - * \c CompleteOrthogonalDecomposition::compute(const* MatrixType&). - */ - CompleteOrthogonalDecomposition() : m_cpqr(), m_zCoeffs(), m_temp() {} - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa CompleteOrthogonalDecomposition() - */ - CompleteOrthogonalDecomposition(Index rows, Index cols) - : m_cpqr(rows, cols), m_zCoeffs((std::min)(rows, cols)), m_temp(cols) {} - - /** \brief Constructs a complete orthogonal decomposition from a given - * matrix. - * - * This constructor computes the complete orthogonal decomposition of the - * matrix \a matrix by calling the method compute(). The default - * threshold for rank determination will be used. It is a short cut for: - * - * \code - * CompleteOrthogonalDecomposition cod(matrix.rows(), - * matrix.cols()); - * cod.setThreshold(Default); - * cod.compute(matrix); - * \endcode - * - * \sa compute() - */ - template - explicit CompleteOrthogonalDecomposition(const EigenBase& matrix) - : m_cpqr(matrix.rows(), matrix.cols()), - m_zCoeffs((std::min)(matrix.rows(), matrix.cols())), - m_temp(matrix.cols()) - { - compute(matrix.derived()); - } - - /** \brief Constructs a complete orthogonal decomposition from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. - * - * \sa CompleteOrthogonalDecomposition(const EigenBase&) - */ - template - explicit CompleteOrthogonalDecomposition(EigenBase& matrix) - : m_cpqr(matrix.derived()), - m_zCoeffs((std::min)(matrix.rows(), matrix.cols())), - m_temp(matrix.cols()) - { - computeInPlace(); - } - - - /** This method computes the minimum-norm solution X to a least squares - * problem \f[\mathrm{minimize} \|A X - B\|, \f] where \b A is the matrix of - * which \c *this is the complete orthogonal decomposition. - * - * \param b the right-hand sides of the problem to solve. - * - * \returns a solution. - * - */ - template - inline const Solve solve( - const MatrixBase& b) const { - eigen_assert(m_cpqr.m_isInitialized && - "CompleteOrthogonalDecomposition is not initialized."); - return Solve(*this, b.derived()); - } - - HouseholderSequenceType householderQ(void) const; - HouseholderSequenceType matrixQ(void) const { return m_cpqr.householderQ(); } - - /** \returns the matrix \b Z. - */ - MatrixType matrixZ() const { - MatrixType Z = MatrixType::Identity(m_cpqr.cols(), m_cpqr.cols()); - applyZAdjointOnTheLeftInPlace(Z); - return Z.adjoint(); - } - - /** \returns a reference to the matrix where the complete orthogonal - * decomposition is stored - */ - const MatrixType& matrixQTZ() const { return m_cpqr.matrixQR(); } - - /** \returns a reference to the matrix where the complete orthogonal - * decomposition is stored. - * \warning The strict lower part and \code cols() - rank() \endcode right - * columns of this matrix contains internal values. - * Only the upper triangular part should be referenced. To get it, use - * \code matrixT().template triangularView() \endcode - * For rank-deficient matrices, use - * \code - * matrixR().topLeftCorner(rank(), rank()).template triangularView() - * \endcode - */ - const MatrixType& matrixT() const { return m_cpqr.matrixQR(); } - - template - CompleteOrthogonalDecomposition& compute(const EigenBase& matrix) { - // Compute the column pivoted QR factorization A P = Q R. - m_cpqr.compute(matrix); - computeInPlace(); - return *this; - } - - /** \returns a const reference to the column permutation matrix */ - const PermutationType& colsPermutation() const { - return m_cpqr.colsPermutation(); - } - - /** \returns the absolute value of the determinant of the matrix of which - * *this is the complete orthogonal decomposition. It has only linear - * complexity (that is, O(n) where n is the dimension of the square matrix) - * as the complete orthogonal decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * One way to work around that is to use logAbsDeterminant() instead. - * - * \sa logAbsDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar absDeterminant() const; - - /** \returns the natural log of the absolute value of the determinant of the - * matrix of which *this is the complete orthogonal decomposition. It has - * only linear complexity (that is, O(n) where n is the dimension of the - * square matrix) as the complete orthogonal decomposition has already been - * computed. - * - * \note This is only for square matrices. - * - * \note This method is useful to work around the risk of overflow/underflow - * that's inherent to determinant computation. - * - * \sa absDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar logAbsDeterminant() const; - - /** \returns the rank of the matrix of which *this is the complete orthogonal - * decomposition. - * - * \note This method has to determine which pivots should be considered - * nonzero. For that, it uses the threshold value that you can control by - * calling setThreshold(const RealScalar&). - */ - inline Index rank() const { return m_cpqr.rank(); } - - /** \returns the dimension of the kernel of the matrix of which *this is the - * complete orthogonal decomposition. - * - * \note This method has to determine which pivots should be considered - * nonzero. For that, it uses the threshold value that you can control by - * calling setThreshold(const RealScalar&). - */ - inline Index dimensionOfKernel() const { return m_cpqr.dimensionOfKernel(); } - - /** \returns true if the matrix of which *this is the decomposition represents - * an injective linear map, i.e. has trivial kernel; false otherwise. - * - * \note This method has to determine which pivots should be considered - * nonzero. For that, it uses the threshold value that you can control by - * calling setThreshold(const RealScalar&). - */ - inline bool isInjective() const { return m_cpqr.isInjective(); } - - /** \returns true if the matrix of which *this is the decomposition represents - * a surjective linear map; false otherwise. - * - * \note This method has to determine which pivots should be considered - * nonzero. For that, it uses the threshold value that you can control by - * calling setThreshold(const RealScalar&). - */ - inline bool isSurjective() const { return m_cpqr.isSurjective(); } - - /** \returns true if the matrix of which *this is the complete orthogonal - * decomposition is invertible. - * - * \note This method has to determine which pivots should be considered - * nonzero. For that, it uses the threshold value that you can control by - * calling setThreshold(const RealScalar&). - */ - inline bool isInvertible() const { return m_cpqr.isInvertible(); } - - /** \returns the pseudo-inverse of the matrix of which *this is the complete - * orthogonal decomposition. - * \warning: Do not compute \c this->pseudoInverse()*rhs to solve a linear systems. - * It is more efficient and numerically stable to call \c this->solve(rhs). - */ - inline const Inverse pseudoInverse() const - { - return Inverse(*this); - } - - inline Index rows() const { return m_cpqr.rows(); } - inline Index cols() const { return m_cpqr.cols(); } - - /** \returns a const reference to the vector of Householder coefficients used - * to represent the factor \c Q. - * - * For advanced uses only. - */ - inline const HCoeffsType& hCoeffs() const { return m_cpqr.hCoeffs(); } - - /** \returns a const reference to the vector of Householder coefficients - * used to represent the factor \c Z. - * - * For advanced uses only. - */ - const HCoeffsType& zCoeffs() const { return m_zCoeffs; } - - /** Allows to prescribe a threshold to be used by certain methods, such as - * rank(), who need to determine when pivots are to be considered nonzero. - * Most be called before calling compute(). - * - * When it needs to get the threshold value, Eigen calls threshold(). By - * default, this uses a formula to automatically determine a reasonable - * threshold. Once you have called the present method - * setThreshold(const RealScalar&), your value is used instead. - * - * \param threshold The new value to use as the threshold. - * - * A pivot will be considered nonzero if its absolute value is strictly - * greater than - * \f$ \vert pivot \vert \leqslant threshold \times \vert maxpivot \vert \f$ - * where maxpivot is the biggest pivot. - * - * If you want to come back to the default behavior, call - * setThreshold(Default_t) - */ - CompleteOrthogonalDecomposition& setThreshold(const RealScalar& threshold) { - m_cpqr.setThreshold(threshold); - return *this; - } - - /** Allows to come back to the default behavior, letting Eigen use its default - * formula for determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code qr.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - CompleteOrthogonalDecomposition& setThreshold(Default_t) { - m_cpqr.setThreshold(Default); - return *this; - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const { return m_cpqr.threshold(); } - - /** \returns the number of nonzero pivots in the complete orthogonal - * decomposition. Here nonzero is meant in the exact sense, not in a - * fuzzy sense. So that notion isn't really intrinsically interesting, - * but it is still useful when implementing algorithms. - * - * \sa rank() - */ - inline Index nonzeroPivots() const { return m_cpqr.nonzeroPivots(); } - - /** \returns the absolute value of the biggest pivot, i.e. the biggest - * diagonal coefficient of R. - */ - inline RealScalar maxPivot() const { return m_cpqr.maxPivot(); } - - /** \brief Reports whether the complete orthogonal decomposition was - * succesful. - * - * \note This function always returns \c Success. It is provided for - * compatibility - * with other factorization routines. - * \returns \c Success - */ - ComputationInfo info() const { - eigen_assert(m_cpqr.m_isInitialized && "Decomposition is not initialized."); - return Success; - } - -#ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC void _solve_impl(const RhsType& rhs, DstType& dst) const; -#endif - - protected: - static void check_template_parameters() { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - /** Overwrites \b rhs with \f$ \mathbf{Z}^* * \mathbf{rhs} \f$. - */ - template - void applyZAdjointOnTheLeftInPlace(Rhs& rhs) const; - - ColPivHouseholderQR m_cpqr; - HCoeffsType m_zCoeffs; - RowVectorType m_temp; -}; - -template -typename MatrixType::RealScalar -CompleteOrthogonalDecomposition::absDeterminant() const { - return m_cpqr.absDeterminant(); -} - -template -typename MatrixType::RealScalar -CompleteOrthogonalDecomposition::logAbsDeterminant() const { - return m_cpqr.logAbsDeterminant(); -} - -/** Performs the complete orthogonal decomposition of the given matrix \a - * matrix. The result of the factorization is stored into \c *this, and a - * reference to \c *this is returned. - * - * \sa class CompleteOrthogonalDecomposition, - * CompleteOrthogonalDecomposition(const MatrixType&) - */ -template -void CompleteOrthogonalDecomposition::computeInPlace() -{ - check_template_parameters(); - - // the column permutation is stored as int indices, so just to be sure: - eigen_assert(m_cpqr.cols() <= NumTraits::highest()); - - const Index rank = m_cpqr.rank(); - const Index cols = m_cpqr.cols(); - const Index rows = m_cpqr.rows(); - m_zCoeffs.resize((std::min)(rows, cols)); - m_temp.resize(cols); - - if (rank < cols) { - // We have reduced the (permuted) matrix to the form - // [R11 R12] - // [ 0 R22] - // where R11 is r-by-r (r = rank) upper triangular, R12 is - // r-by-(n-r), and R22 is empty or the norm of R22 is negligible. - // We now compute the complete orthogonal decomposition by applying - // Householder transformations from the right to the upper trapezoidal - // matrix X = [R11 R12] to zero out R12 and obtain the factorization - // [R11 R12] = [T11 0] * Z, where T11 is r-by-r upper triangular and - // Z = Z(0) * Z(1) ... Z(r-1) is an n-by-n orthogonal matrix. - // We store the data representing Z in R12 and m_zCoeffs. - for (Index k = rank - 1; k >= 0; --k) { - if (k != rank - 1) { - // Given the API for Householder reflectors, it is more convenient if - // we swap the leading parts of columns k and r-1 (zero-based) to form - // the matrix X_k = [X(0:k, k), X(0:k, r:n)] - m_cpqr.m_qr.col(k).head(k + 1).swap( - m_cpqr.m_qr.col(rank - 1).head(k + 1)); - } - // Construct Householder reflector Z(k) to zero out the last row of X_k, - // i.e. choose Z(k) such that - // [X(k, k), X(k, r:n)] * Z(k) = [beta, 0, .., 0]. - RealScalar beta; - m_cpqr.m_qr.row(k) - .tail(cols - rank + 1) - .makeHouseholderInPlace(m_zCoeffs(k), beta); - m_cpqr.m_qr(k, rank - 1) = beta; - if (k > 0) { - // Apply Z(k) to the first k rows of X_k - m_cpqr.m_qr.topRightCorner(k, cols - rank + 1) - .applyHouseholderOnTheRight( - m_cpqr.m_qr.row(k).tail(cols - rank).transpose(), m_zCoeffs(k), - &m_temp(0)); - } - if (k != rank - 1) { - // Swap X(0:k,k) back to its proper location. - m_cpqr.m_qr.col(k).head(k + 1).swap( - m_cpqr.m_qr.col(rank - 1).head(k + 1)); - } - } - } -} - -template -template -void CompleteOrthogonalDecomposition::applyZAdjointOnTheLeftInPlace( - Rhs& rhs) const { - const Index cols = this->cols(); - const Index nrhs = rhs.cols(); - const Index rank = this->rank(); - Matrix temp((std::max)(cols, nrhs)); - for (Index k = 0; k < rank; ++k) { - if (k != rank - 1) { - rhs.row(k).swap(rhs.row(rank - 1)); - } - rhs.middleRows(rank - 1, cols - rank + 1) - .applyHouseholderOnTheLeft( - matrixQTZ().row(k).tail(cols - rank).adjoint(), zCoeffs()(k), - &temp(0)); - if (k != rank - 1) { - rhs.row(k).swap(rhs.row(rank - 1)); - } - } -} - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void CompleteOrthogonalDecomposition<_MatrixType>::_solve_impl( - const RhsType& rhs, DstType& dst) const { - eigen_assert(rhs.rows() == this->rows()); - - const Index rank = this->rank(); - if (rank == 0) { - dst.setZero(); - return; - } - - // Compute c = Q^* * rhs - // Note that the matrix Q = H_0^* H_1^*... so its inverse is - // Q^* = (H_0 H_1 ...)^T - typename RhsType::PlainObject c(rhs); - c.applyOnTheLeft( - householderSequence(matrixQTZ(), hCoeffs()).setLength(rank).transpose()); - - // Solve T z = c(1:rank, :) - dst.topRows(rank) = matrixT() - .topLeftCorner(rank, rank) - .template triangularView() - .solve(c.topRows(rank)); - - const Index cols = this->cols(); - if (rank < cols) { - // Compute y = Z^* * [ z ] - // [ 0 ] - dst.bottomRows(cols - rank).setZero(); - applyZAdjointOnTheLeftInPlace(dst); - } - - // Undo permutation to get x = P^{-1} * y. - dst = colsPermutation() * dst; -} -#endif - -namespace internal { - -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef CompleteOrthogonalDecomposition CodType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.rows())); - } -}; - -} // end namespace internal - -/** \returns the matrix Q as a sequence of householder transformations */ -template -typename CompleteOrthogonalDecomposition::HouseholderSequenceType -CompleteOrthogonalDecomposition::householderQ() const { - return m_cpqr.householderQ(); -} - -/** \return the complete orthogonal decomposition of \c *this. - * - * \sa class CompleteOrthogonalDecomposition - */ -template -const CompleteOrthogonalDecomposition::PlainObject> -MatrixBase::completeOrthogonalDecomposition() const { - return CompleteOrthogonalDecomposition(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_COMPLETEORTHOGONALDECOMPOSITION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/FullPivHouseholderQR.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/FullPivHouseholderQR.h deleted file mode 100644 index e489bddc2db..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/FullPivHouseholderQR.h +++ /dev/null @@ -1,676 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_FULLPIVOTINGHOUSEHOLDERQR_H -#define EIGEN_FULLPIVOTINGHOUSEHOLDERQR_H - -namespace Eigen { - -namespace internal { - -template struct traits > - : traits<_MatrixType> -{ - enum { Flags = 0 }; -}; - -template struct FullPivHouseholderQRMatrixQReturnType; - -template -struct traits > -{ - typedef typename MatrixType::PlainObject ReturnType; -}; - -} // end namespace internal - -/** \ingroup QR_Module - * - * \class FullPivHouseholderQR - * - * \brief Householder rank-revealing QR decomposition of a matrix with full pivoting - * - * \tparam _MatrixType the type of the matrix of which we are computing the QR decomposition - * - * This class performs a rank-revealing QR decomposition of a matrix \b A into matrices \b P, \b P', \b Q and \b R - * such that - * \f[ - * \mathbf{P} \, \mathbf{A} \, \mathbf{P}' = \mathbf{Q} \, \mathbf{R} - * \f] - * by using Householder transformations. Here, \b P and \b P' are permutation matrices, \b Q a unitary matrix - * and \b R an upper triangular matrix. - * - * This decomposition performs a very prudent full pivoting in order to be rank-revealing and achieve optimal - * numerical stability. The trade-off is that it is slower than HouseholderQR and ColPivHouseholderQR. - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::fullPivHouseholderQr() - */ -template class FullPivHouseholderQR -{ - public: - - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - // FIXME should be int - typedef typename MatrixType::StorageIndex StorageIndex; - typedef internal::FullPivHouseholderQRMatrixQReturnType MatrixQReturnType; - typedef typename internal::plain_diag_type::type HCoeffsType; - typedef Matrix IntDiagSizeVectorType; - typedef PermutationMatrix PermutationType; - typedef typename internal::plain_row_type::type RowVectorType; - typedef typename internal::plain_col_type::type ColVectorType; - typedef typename MatrixType::PlainObject PlainObject; - - /** \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via FullPivHouseholderQR::compute(const MatrixType&). - */ - FullPivHouseholderQR() - : m_qr(), - m_hCoeffs(), - m_rows_transpositions(), - m_cols_transpositions(), - m_cols_permutation(), - m_temp(), - m_isInitialized(false), - m_usePrescribedThreshold(false) {} - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa FullPivHouseholderQR() - */ - FullPivHouseholderQR(Index rows, Index cols) - : m_qr(rows, cols), - m_hCoeffs((std::min)(rows,cols)), - m_rows_transpositions((std::min)(rows,cols)), - m_cols_transpositions((std::min)(rows,cols)), - m_cols_permutation(cols), - m_temp(cols), - m_isInitialized(false), - m_usePrescribedThreshold(false) {} - - /** \brief Constructs a QR factorization from a given matrix - * - * This constructor computes the QR factorization of the matrix \a matrix by calling - * the method compute(). It is a short cut for: - * - * \code - * FullPivHouseholderQR qr(matrix.rows(), matrix.cols()); - * qr.compute(matrix); - * \endcode - * - * \sa compute() - */ - template - explicit FullPivHouseholderQR(const EigenBase& matrix) - : m_qr(matrix.rows(), matrix.cols()), - m_hCoeffs((std::min)(matrix.rows(), matrix.cols())), - m_rows_transpositions((std::min)(matrix.rows(), matrix.cols())), - m_cols_transpositions((std::min)(matrix.rows(), matrix.cols())), - m_cols_permutation(matrix.cols()), - m_temp(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) - { - compute(matrix.derived()); - } - - /** \brief Constructs a QR factorization from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. - * - * \sa FullPivHouseholderQR(const EigenBase&) - */ - template - explicit FullPivHouseholderQR(EigenBase& matrix) - : m_qr(matrix.derived()), - m_hCoeffs((std::min)(matrix.rows(), matrix.cols())), - m_rows_transpositions((std::min)(matrix.rows(), matrix.cols())), - m_cols_transpositions((std::min)(matrix.rows(), matrix.cols())), - m_cols_permutation(matrix.cols()), - m_temp(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) - { - computeInPlace(); - } - - /** This method finds a solution x to the equation Ax=b, where A is the matrix of which - * \c *this is the QR decomposition. - * - * \param b the right-hand-side of the equation to solve. - * - * \returns the exact or least-square solution if the rank is greater or equal to the number of columns of A, - * and an arbitrary solution otherwise. - * - * \note_about_checking_solutions - * - * \note_about_arbitrary_choice_of_solution - * - * Example: \include FullPivHouseholderQR_solve.cpp - * Output: \verbinclude FullPivHouseholderQR_solve.out - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return Solve(*this, b.derived()); - } - - /** \returns Expression object representing the matrix Q - */ - MatrixQReturnType matrixQ(void) const; - - /** \returns a reference to the matrix where the Householder QR decomposition is stored - */ - const MatrixType& matrixQR() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return m_qr; - } - - template - FullPivHouseholderQR& compute(const EigenBase& matrix); - - /** \returns a const reference to the column permutation matrix */ - const PermutationType& colsPermutation() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return m_cols_permutation; - } - - /** \returns a const reference to the vector of indices representing the rows transpositions */ - const IntDiagSizeVectorType& rowsTranspositions() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return m_rows_transpositions; - } - - /** \returns the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * One way to work around that is to use logAbsDeterminant() instead. - * - * \sa logAbsDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar absDeterminant() const; - - /** \returns the natural log of the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \note This method is useful to work around the risk of overflow/underflow that's inherent - * to determinant computation. - * - * \sa absDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar logAbsDeterminant() const; - - /** \returns the rank of the matrix of which *this is the QR decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index rank() const - { - using std::abs; - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - RealScalar premultiplied_threshold = abs(m_maxpivot) * threshold(); - Index result = 0; - for(Index i = 0; i < m_nonzero_pivots; ++i) - result += (abs(m_qr.coeff(i,i)) > premultiplied_threshold); - return result; - } - - /** \returns the dimension of the kernel of the matrix of which *this is the QR decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index dimensionOfKernel() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return cols() - rank(); - } - - /** \returns true if the matrix of which *this is the QR decomposition represents an injective - * linear map, i.e. has trivial kernel; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInjective() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return rank() == cols(); - } - - /** \returns true if the matrix of which *this is the QR decomposition represents a surjective - * linear map; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isSurjective() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return rank() == rows(); - } - - /** \returns true if the matrix of which *this is the QR decomposition is invertible. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInvertible() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return isInjective() && isSurjective(); - } - - /** \returns the inverse of the matrix of which *this is the QR decomposition. - * - * \note If this matrix is not invertible, the returned matrix has undefined coefficients. - * Use isInvertible() to first determine whether this matrix is invertible. - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return Inverse(*this); - } - - inline Index rows() const { return m_qr.rows(); } - inline Index cols() const { return m_qr.cols(); } - - /** \returns a const reference to the vector of Householder coefficients used to represent the factor \c Q. - * - * For advanced uses only. - */ - const HCoeffsType& hCoeffs() const { return m_hCoeffs; } - - /** Allows to prescribe a threshold to be used by certain methods, such as rank(), - * who need to determine when pivots are to be considered nonzero. This is not used for the - * QR decomposition itself. - * - * When it needs to get the threshold value, Eigen calls threshold(). By default, this - * uses a formula to automatically determine a reasonable threshold. - * Once you have called the present method setThreshold(const RealScalar&), - * your value is used instead. - * - * \param threshold The new value to use as the threshold. - * - * A pivot will be considered nonzero if its absolute value is strictly greater than - * \f$ \vert pivot \vert \leqslant threshold \times \vert maxpivot \vert \f$ - * where maxpivot is the biggest pivot. - * - * If you want to come back to the default behavior, call setThreshold(Default_t) - */ - FullPivHouseholderQR& setThreshold(const RealScalar& threshold) - { - m_usePrescribedThreshold = true; - m_prescribedThreshold = threshold; - return *this; - } - - /** Allows to come back to the default behavior, letting Eigen use its default formula for - * determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code qr.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - FullPivHouseholderQR& setThreshold(Default_t) - { - m_usePrescribedThreshold = false; - return *this; - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const - { - eigen_assert(m_isInitialized || m_usePrescribedThreshold); - return m_usePrescribedThreshold ? m_prescribedThreshold - // this formula comes from experimenting (see "LU precision tuning" thread on the list) - // and turns out to be identical to Higham's formula used already in LDLt. - : NumTraits::epsilon() * RealScalar(m_qr.diagonalSize()); - } - - /** \returns the number of nonzero pivots in the QR decomposition. - * Here nonzero is meant in the exact sense, not in a fuzzy sense. - * So that notion isn't really intrinsically interesting, but it is - * still useful when implementing algorithms. - * - * \sa rank() - */ - inline Index nonzeroPivots() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_nonzero_pivots; - } - - /** \returns the absolute value of the biggest pivot, i.e. the biggest - * diagonal coefficient of U. - */ - RealScalar maxPivot() const { return m_maxpivot; } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - MatrixType m_qr; - HCoeffsType m_hCoeffs; - IntDiagSizeVectorType m_rows_transpositions; - IntDiagSizeVectorType m_cols_transpositions; - PermutationType m_cols_permutation; - RowVectorType m_temp; - bool m_isInitialized, m_usePrescribedThreshold; - RealScalar m_prescribedThreshold, m_maxpivot; - Index m_nonzero_pivots; - RealScalar m_precision; - Index m_det_pq; -}; - -template -typename MatrixType::RealScalar FullPivHouseholderQR::absDeterminant() const -{ - using std::abs; - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return abs(m_qr.diagonal().prod()); -} - -template -typename MatrixType::RealScalar FullPivHouseholderQR::logAbsDeterminant() const -{ - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return m_qr.diagonal().cwiseAbs().array().log().sum(); -} - -/** Performs the QR factorization of the given matrix \a matrix. The result of - * the factorization is stored into \c *this, and a reference to \c *this - * is returned. - * - * \sa class FullPivHouseholderQR, FullPivHouseholderQR(const MatrixType&) - */ -template -template -FullPivHouseholderQR& FullPivHouseholderQR::compute(const EigenBase& matrix) -{ - m_qr = matrix.derived(); - computeInPlace(); - return *this; -} - -template -void FullPivHouseholderQR::computeInPlace() -{ - check_template_parameters(); - - using std::abs; - Index rows = m_qr.rows(); - Index cols = m_qr.cols(); - Index size = (std::min)(rows,cols); - - - m_hCoeffs.resize(size); - - m_temp.resize(cols); - - m_precision = NumTraits::epsilon() * RealScalar(size); - - m_rows_transpositions.resize(size); - m_cols_transpositions.resize(size); - Index number_of_transpositions = 0; - - RealScalar biggest(0); - - m_nonzero_pivots = size; // the generic case is that in which all pivots are nonzero (invertible case) - m_maxpivot = RealScalar(0); - - for (Index k = 0; k < size; ++k) - { - Index row_of_biggest_in_corner, col_of_biggest_in_corner; - typedef internal::scalar_score_coeff_op Scoring; - typedef typename Scoring::result_type Score; - - Score score = m_qr.bottomRightCorner(rows-k, cols-k) - .unaryExpr(Scoring()) - .maxCoeff(&row_of_biggest_in_corner, &col_of_biggest_in_corner); - row_of_biggest_in_corner += k; - col_of_biggest_in_corner += k; - RealScalar biggest_in_corner = internal::abs_knowing_score()(m_qr(row_of_biggest_in_corner, col_of_biggest_in_corner), score); - if(k==0) biggest = biggest_in_corner; - - // if the corner is negligible, then we have less than full rank, and we can finish early - if(internal::isMuchSmallerThan(biggest_in_corner, biggest, m_precision)) - { - m_nonzero_pivots = k; - for(Index i = k; i < size; i++) - { - m_rows_transpositions.coeffRef(i) = i; - m_cols_transpositions.coeffRef(i) = i; - m_hCoeffs.coeffRef(i) = Scalar(0); - } - break; - } - - m_rows_transpositions.coeffRef(k) = row_of_biggest_in_corner; - m_cols_transpositions.coeffRef(k) = col_of_biggest_in_corner; - if(k != row_of_biggest_in_corner) { - m_qr.row(k).tail(cols-k).swap(m_qr.row(row_of_biggest_in_corner).tail(cols-k)); - ++number_of_transpositions; - } - if(k != col_of_biggest_in_corner) { - m_qr.col(k).swap(m_qr.col(col_of_biggest_in_corner)); - ++number_of_transpositions; - } - - RealScalar beta; - m_qr.col(k).tail(rows-k).makeHouseholderInPlace(m_hCoeffs.coeffRef(k), beta); - m_qr.coeffRef(k,k) = beta; - - // remember the maximum absolute value of diagonal coefficients - if(abs(beta) > m_maxpivot) m_maxpivot = abs(beta); - - m_qr.bottomRightCorner(rows-k, cols-k-1) - .applyHouseholderOnTheLeft(m_qr.col(k).tail(rows-k-1), m_hCoeffs.coeffRef(k), &m_temp.coeffRef(k+1)); - } - - m_cols_permutation.setIdentity(cols); - for(Index k = 0; k < size; ++k) - m_cols_permutation.applyTranspositionOnTheRight(k, m_cols_transpositions.coeff(k)); - - m_det_pq = (number_of_transpositions%2) ? -1 : 1; - m_isInitialized = true; -} - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void FullPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - eigen_assert(rhs.rows() == rows()); - const Index l_rank = rank(); - - // FIXME introduce nonzeroPivots() and use it here. and more generally, - // make the same improvements in this dec as in FullPivLU. - if(l_rank==0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs); - - Matrix temp(rhs.cols()); - for (Index k = 0; k < l_rank; ++k) - { - Index remainingSize = rows()-k; - c.row(k).swap(c.row(m_rows_transpositions.coeff(k))); - c.bottomRightCorner(remainingSize, rhs.cols()) - .applyHouseholderOnTheLeft(m_qr.col(k).tail(remainingSize-1), - m_hCoeffs.coeff(k), &temp.coeffRef(0)); - } - - m_qr.topLeftCorner(l_rank, l_rank) - .template triangularView() - .solveInPlace(c.topRows(l_rank)); - - for(Index i = 0; i < l_rank; ++i) dst.row(m_cols_permutation.indices().coeff(i)) = c.row(i); - for(Index i = l_rank; i < cols(); ++i) dst.row(m_cols_permutation.indices().coeff(i)).setZero(); -} -#endif - -namespace internal { - -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef FullPivHouseholderQR QrType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; - -/** \ingroup QR_Module - * - * \brief Expression type for return value of FullPivHouseholderQR::matrixQ() - * - * \tparam MatrixType type of underlying dense matrix - */ -template struct FullPivHouseholderQRMatrixQReturnType - : public ReturnByValue > -{ -public: - typedef typename FullPivHouseholderQR::IntDiagSizeVectorType IntDiagSizeVectorType; - typedef typename internal::plain_diag_type::type HCoeffsType; - typedef Matrix WorkVectorType; - - FullPivHouseholderQRMatrixQReturnType(const MatrixType& qr, - const HCoeffsType& hCoeffs, - const IntDiagSizeVectorType& rowsTranspositions) - : m_qr(qr), - m_hCoeffs(hCoeffs), - m_rowsTranspositions(rowsTranspositions) - {} - - template - void evalTo(ResultType& result) const - { - const Index rows = m_qr.rows(); - WorkVectorType workspace(rows); - evalTo(result, workspace); - } - - template - void evalTo(ResultType& result, WorkVectorType& workspace) const - { - using numext::conj; - // compute the product H'_0 H'_1 ... H'_n-1, - // where H_k is the k-th Householder transformation I - h_k v_k v_k' - // and v_k is the k-th Householder vector [1,m_qr(k+1,k), m_qr(k+2,k), ...] - const Index rows = m_qr.rows(); - const Index cols = m_qr.cols(); - const Index size = (std::min)(rows, cols); - workspace.resize(rows); - result.setIdentity(rows, rows); - for (Index k = size-1; k >= 0; k--) - { - result.block(k, k, rows-k, rows-k) - .applyHouseholderOnTheLeft(m_qr.col(k).tail(rows-k-1), conj(m_hCoeffs.coeff(k)), &workspace.coeffRef(k)); - result.row(k).swap(result.row(m_rowsTranspositions.coeff(k))); - } - } - - Index rows() const { return m_qr.rows(); } - Index cols() const { return m_qr.rows(); } - -protected: - typename MatrixType::Nested m_qr; - typename HCoeffsType::Nested m_hCoeffs; - typename IntDiagSizeVectorType::Nested m_rowsTranspositions; -}; - -// template -// struct evaluator > -// : public evaluator > > -// {}; - -} // end namespace internal - -template -inline typename FullPivHouseholderQR::MatrixQReturnType FullPivHouseholderQR::matrixQ() const -{ - eigen_assert(m_isInitialized && "FullPivHouseholderQR is not initialized."); - return MatrixQReturnType(m_qr, m_hCoeffs, m_rows_transpositions); -} - -/** \return the full-pivoting Householder QR decomposition of \c *this. - * - * \sa class FullPivHouseholderQR - */ -template -const FullPivHouseholderQR::PlainObject> -MatrixBase::fullPivHouseholderQr() const -{ - return FullPivHouseholderQR(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_FULLPIVOTINGHOUSEHOLDERQR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR.h deleted file mode 100644 index 3513d995cb6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR.h +++ /dev/null @@ -1,409 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// Copyright (C) 2010 Vincent Lejeune -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_QR_H -#define EIGEN_QR_H - -namespace Eigen { - -/** \ingroup QR_Module - * - * - * \class HouseholderQR - * - * \brief Householder QR decomposition of a matrix - * - * \tparam _MatrixType the type of the matrix of which we are computing the QR decomposition - * - * This class performs a QR decomposition of a matrix \b A into matrices \b Q and \b R - * such that - * \f[ - * \mathbf{A} = \mathbf{Q} \, \mathbf{R} - * \f] - * by using Householder transformations. Here, \b Q a unitary matrix and \b R an upper triangular matrix. - * The result is stored in a compact way compatible with LAPACK. - * - * Note that no pivoting is performed. This is \b not a rank-revealing decomposition. - * If you want that feature, use FullPivHouseholderQR or ColPivHouseholderQR instead. - * - * This Householder QR decomposition is faster, but less numerically stable and less feature-full than - * FullPivHouseholderQR or ColPivHouseholderQR. - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::householderQr() - */ -template class HouseholderQR -{ - public: - - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - // FIXME should be int - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Matrix MatrixQType; - typedef typename internal::plain_diag_type::type HCoeffsType; - typedef typename internal::plain_row_type::type RowVectorType; - typedef HouseholderSequence::type> HouseholderSequenceType; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via HouseholderQR::compute(const MatrixType&). - */ - HouseholderQR() : m_qr(), m_hCoeffs(), m_temp(), m_isInitialized(false) {} - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa HouseholderQR() - */ - HouseholderQR(Index rows, Index cols) - : m_qr(rows, cols), - m_hCoeffs((std::min)(rows,cols)), - m_temp(cols), - m_isInitialized(false) {} - - /** \brief Constructs a QR factorization from a given matrix - * - * This constructor computes the QR factorization of the matrix \a matrix by calling - * the method compute(). It is a short cut for: - * - * \code - * HouseholderQR qr(matrix.rows(), matrix.cols()); - * qr.compute(matrix); - * \endcode - * - * \sa compute() - */ - template - explicit HouseholderQR(const EigenBase& matrix) - : m_qr(matrix.rows(), matrix.cols()), - m_hCoeffs((std::min)(matrix.rows(),matrix.cols())), - m_temp(matrix.cols()), - m_isInitialized(false) - { - compute(matrix.derived()); - } - - - /** \brief Constructs a QR factorization from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when - * \c MatrixType is a Eigen::Ref. - * - * \sa HouseholderQR(const EigenBase&) - */ - template - explicit HouseholderQR(EigenBase& matrix) - : m_qr(matrix.derived()), - m_hCoeffs((std::min)(matrix.rows(),matrix.cols())), - m_temp(matrix.cols()), - m_isInitialized(false) - { - computeInPlace(); - } - - /** This method finds a solution x to the equation Ax=b, where A is the matrix of which - * *this is the QR decomposition, if any exists. - * - * \param b the right-hand-side of the equation to solve. - * - * \returns a solution. - * - * \note_about_checking_solutions - * - * \note_about_arbitrary_choice_of_solution - * - * Example: \include HouseholderQR_solve.cpp - * Output: \verbinclude HouseholderQR_solve.out - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "HouseholderQR is not initialized."); - return Solve(*this, b.derived()); - } - - /** This method returns an expression of the unitary matrix Q as a sequence of Householder transformations. - * - * The returned expression can directly be used to perform matrix products. It can also be assigned to a dense Matrix object. - * Here is an example showing how to recover the full or thin matrix Q, as well as how to perform matrix products using operator*: - * - * Example: \include HouseholderQR_householderQ.cpp - * Output: \verbinclude HouseholderQR_householderQ.out - */ - HouseholderSequenceType householderQ() const - { - eigen_assert(m_isInitialized && "HouseholderQR is not initialized."); - return HouseholderSequenceType(m_qr, m_hCoeffs.conjugate()); - } - - /** \returns a reference to the matrix where the Householder QR decomposition is stored - * in a LAPACK-compatible way. - */ - const MatrixType& matrixQR() const - { - eigen_assert(m_isInitialized && "HouseholderQR is not initialized."); - return m_qr; - } - - template - HouseholderQR& compute(const EigenBase& matrix) { - m_qr = matrix.derived(); - computeInPlace(); - return *this; - } - - /** \returns the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * One way to work around that is to use logAbsDeterminant() instead. - * - * \sa logAbsDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar absDeterminant() const; - - /** \returns the natural log of the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the QR decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \note This method is useful to work around the risk of overflow/underflow that's inherent - * to determinant computation. - * - * \sa absDeterminant(), MatrixBase::determinant() - */ - typename MatrixType::RealScalar logAbsDeterminant() const; - - inline Index rows() const { return m_qr.rows(); } - inline Index cols() const { return m_qr.cols(); } - - /** \returns a const reference to the vector of Householder coefficients used to represent the factor \c Q. - * - * For advanced uses only. - */ - const HCoeffsType& hCoeffs() const { return m_hCoeffs; } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - MatrixType m_qr; - HCoeffsType m_hCoeffs; - RowVectorType m_temp; - bool m_isInitialized; -}; - -template -typename MatrixType::RealScalar HouseholderQR::absDeterminant() const -{ - using std::abs; - eigen_assert(m_isInitialized && "HouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return abs(m_qr.diagonal().prod()); -} - -template -typename MatrixType::RealScalar HouseholderQR::logAbsDeterminant() const -{ - eigen_assert(m_isInitialized && "HouseholderQR is not initialized."); - eigen_assert(m_qr.rows() == m_qr.cols() && "You can't take the determinant of a non-square matrix!"); - return m_qr.diagonal().cwiseAbs().array().log().sum(); -} - -namespace internal { - -/** \internal */ -template -void householder_qr_inplace_unblocked(MatrixQR& mat, HCoeffs& hCoeffs, typename MatrixQR::Scalar* tempData = 0) -{ - typedef typename MatrixQR::Scalar Scalar; - typedef typename MatrixQR::RealScalar RealScalar; - Index rows = mat.rows(); - Index cols = mat.cols(); - Index size = (std::min)(rows,cols); - - eigen_assert(hCoeffs.size() == size); - - typedef Matrix TempType; - TempType tempVector; - if(tempData==0) - { - tempVector.resize(cols); - tempData = tempVector.data(); - } - - for(Index k = 0; k < size; ++k) - { - Index remainingRows = rows - k; - Index remainingCols = cols - k - 1; - - RealScalar beta; - mat.col(k).tail(remainingRows).makeHouseholderInPlace(hCoeffs.coeffRef(k), beta); - mat.coeffRef(k,k) = beta; - - // apply H to remaining part of m_qr from the left - mat.bottomRightCorner(remainingRows, remainingCols) - .applyHouseholderOnTheLeft(mat.col(k).tail(remainingRows-1), hCoeffs.coeffRef(k), tempData+k+1); - } -} - -/** \internal */ -template -struct householder_qr_inplace_blocked -{ - // This is specialized for MKL-supported Scalar types in HouseholderQR_MKL.h - static void run(MatrixQR& mat, HCoeffs& hCoeffs, Index maxBlockSize=32, - typename MatrixQR::Scalar* tempData = 0) - { - typedef typename MatrixQR::Scalar Scalar; - typedef Block BlockType; - - Index rows = mat.rows(); - Index cols = mat.cols(); - Index size = (std::min)(rows, cols); - - typedef Matrix TempType; - TempType tempVector; - if(tempData==0) - { - tempVector.resize(cols); - tempData = tempVector.data(); - } - - Index blockSize = (std::min)(maxBlockSize,size); - - Index k = 0; - for (k = 0; k < size; k += blockSize) - { - Index bs = (std::min)(size-k,blockSize); // actual size of the block - Index tcols = cols - k - bs; // trailing columns - Index brows = rows-k; // rows of the block - - // partition the matrix: - // A00 | A01 | A02 - // mat = A10 | A11 | A12 - // A20 | A21 | A22 - // and performs the qr dec of [A11^T A12^T]^T - // and update [A21^T A22^T]^T using level 3 operations. - // Finally, the algorithm continue on A22 - - BlockType A11_21 = mat.block(k,k,brows,bs); - Block hCoeffsSegment = hCoeffs.segment(k,bs); - - householder_qr_inplace_unblocked(A11_21, hCoeffsSegment, tempData); - - if(tcols) - { - BlockType A21_22 = mat.block(k,k+bs,brows,tcols); - apply_block_householder_on_the_left(A21_22,A11_21,hCoeffsSegment, false); // false == backward - } - } - } -}; - -} // end namespace internal - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void HouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - const Index rank = (std::min)(rows(), cols()); - eigen_assert(rhs.rows() == rows()); - - typename RhsType::PlainObject c(rhs); - - // Note that the matrix Q = H_0^* H_1^*... so its inverse is Q^* = (H_0 H_1 ...)^T - c.applyOnTheLeft(householderSequence( - m_qr.leftCols(rank), - m_hCoeffs.head(rank)).transpose() - ); - - m_qr.topLeftCorner(rank, rank) - .template triangularView() - .solveInPlace(c.topRows(rank)); - - dst.topRows(rank) = c.topRows(rank); - dst.bottomRows(cols()-rank).setZero(); -} -#endif - -/** Performs the QR factorization of the given matrix \a matrix. The result of - * the factorization is stored into \c *this, and a reference to \c *this - * is returned. - * - * \sa class HouseholderQR, HouseholderQR(const MatrixType&) - */ -template -void HouseholderQR::computeInPlace() -{ - check_template_parameters(); - - Index rows = m_qr.rows(); - Index cols = m_qr.cols(); - Index size = (std::min)(rows,cols); - - m_hCoeffs.resize(size); - - m_temp.resize(cols); - - internal::householder_qr_inplace_blocked::run(m_qr, m_hCoeffs, 48, m_temp.data()); - - m_isInitialized = true; -} - -/** \return the Householder QR decomposition of \c *this. - * - * \sa class HouseholderQR - */ -template -const HouseholderQR::PlainObject> -MatrixBase::householderQr() const -{ - return HouseholderQR(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_QR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR_LAPACKE.h deleted file mode 100644 index 1dc7d5363f1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/QR/HouseholderQR_LAPACKE.h +++ /dev/null @@ -1,68 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * Householder QR decomposition of a matrix w/o pivoting based on - * LAPACKE_?geqrf function. - ******************************************************************************** -*/ - -#ifndef EIGEN_QR_LAPACKE_H -#define EIGEN_QR_LAPACKE_H - -namespace Eigen { - -namespace internal { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_QR_NOPIV(EIGTYPE, LAPACKE_TYPE, LAPACKE_PREFIX) \ -template \ -struct householder_qr_inplace_blocked \ -{ \ - static void run(MatrixQR& mat, HCoeffs& hCoeffs, Index = 32, \ - typename MatrixQR::Scalar* = 0) \ - { \ - lapack_int m = (lapack_int) mat.rows(); \ - lapack_int n = (lapack_int) mat.cols(); \ - lapack_int lda = (lapack_int) mat.outerStride(); \ - lapack_int matrix_order = (MatrixQR::IsRowMajor) ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \ - LAPACKE_##LAPACKE_PREFIX##geqrf( matrix_order, m, n, (LAPACKE_TYPE*)mat.data(), lda, (LAPACKE_TYPE*)hCoeffs.data()); \ - hCoeffs.adjointInPlace(); \ - } \ -}; - -EIGEN_LAPACKE_QR_NOPIV(double, double, d) -EIGEN_LAPACKE_QR_NOPIV(float, float, s) -EIGEN_LAPACKE_QR_NOPIV(dcomplex, lapack_complex_double, z) -EIGEN_LAPACKE_QR_NOPIV(scomplex, lapack_complex_float, c) - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_QR_LAPACKE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SPQRSupport/SuiteSparseQRSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SPQRSupport/SuiteSparseQRSupport.h deleted file mode 100644 index 953d57c9d76..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SPQRSupport/SuiteSparseQRSupport.h +++ /dev/null @@ -1,313 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Desire Nuentsa -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SUITESPARSEQRSUPPORT_H -#define EIGEN_SUITESPARSEQRSUPPORT_H - -namespace Eigen { - - template class SPQR; - template struct SPQRMatrixQReturnType; - template struct SPQRMatrixQTransposeReturnType; - template struct SPQR_QProduct; - namespace internal { - template struct traits > - { - typedef typename SPQRType::MatrixType ReturnType; - }; - template struct traits > - { - typedef typename SPQRType::MatrixType ReturnType; - }; - template struct traits > - { - typedef typename Derived::PlainObject ReturnType; - }; - } // End namespace internal - -/** - * \ingroup SPQRSupport_Module - * \class SPQR - * \brief Sparse QR factorization based on SuiteSparseQR library - * - * This class is used to perform a multithreaded and multifrontal rank-revealing QR decomposition - * of sparse matrices. The result is then used to solve linear leasts_square systems. - * Clearly, a QR factorization is returned such that A*P = Q*R where : - * - * P is the column permutation. Use colsPermutation() to get it. - * - * Q is the orthogonal matrix represented as Householder reflectors. - * Use matrixQ() to get an expression and matrixQ().transpose() to get the transpose. - * You can then apply it to a vector. - * - * R is the sparse triangular factor. Use matrixQR() to get it as SparseMatrix. - * NOTE : The Index type of R is always SuiteSparse_long. You can get it with SPQR::Index - * - * \tparam _MatrixType The type of the sparse matrix A, must be a column-major SparseMatrix<> - * - * \implsparsesolverconcept - * - * - */ -template -class SPQR : public SparseSolverBase > -{ - protected: - typedef SparseSolverBase > Base; - using Base::m_isInitialized; - public: - typedef typename _MatrixType::Scalar Scalar; - typedef typename _MatrixType::RealScalar RealScalar; - typedef SuiteSparse_long StorageIndex ; - typedef SparseMatrix MatrixType; - typedef Map > PermutationType; - enum { - ColsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic - }; - public: - SPQR() - : m_ordering(SPQR_ORDERING_DEFAULT), m_allow_tol(SPQR_DEFAULT_TOL), m_tolerance (NumTraits::epsilon()), m_useDefaultThreshold(true) - { - cholmod_l_start(&m_cc); - } - - explicit SPQR(const _MatrixType& matrix) - : m_ordering(SPQR_ORDERING_DEFAULT), m_allow_tol(SPQR_DEFAULT_TOL), m_tolerance (NumTraits::epsilon()), m_useDefaultThreshold(true) - { - cholmod_l_start(&m_cc); - compute(matrix); - } - - ~SPQR() - { - SPQR_free(); - cholmod_l_finish(&m_cc); - } - void SPQR_free() - { - cholmod_l_free_sparse(&m_H, &m_cc); - cholmod_l_free_sparse(&m_cR, &m_cc); - cholmod_l_free_dense(&m_HTau, &m_cc); - std::free(m_E); - std::free(m_HPinv); - } - - void compute(const _MatrixType& matrix) - { - if(m_isInitialized) SPQR_free(); - - MatrixType mat(matrix); - - /* Compute the default threshold as in MatLab, see: - * Tim Davis, "Algorithm 915, SuiteSparseQR: Multifrontal Multithreaded Rank-Revealing - * Sparse QR Factorization, ACM Trans. on Math. Soft. 38(1), 2011, Page 8:3 - */ - RealScalar pivotThreshold = m_tolerance; - if(m_useDefaultThreshold) - { - RealScalar max2Norm = 0.0; - for (int j = 0; j < mat.cols(); j++) max2Norm = numext::maxi(max2Norm, mat.col(j).norm()); - if(max2Norm==RealScalar(0)) - max2Norm = RealScalar(1); - pivotThreshold = 20 * (mat.rows() + mat.cols()) * max2Norm * NumTraits::epsilon(); - } - cholmod_sparse A; - A = viewAsCholmod(mat); - m_rows = matrix.rows(); - Index col = matrix.cols(); - m_rank = SuiteSparseQR(m_ordering, pivotThreshold, col, &A, - &m_cR, &m_E, &m_H, &m_HPinv, &m_HTau, &m_cc); - - if (!m_cR) - { - m_info = NumericalIssue; - m_isInitialized = false; - return; - } - m_info = Success; - m_isInitialized = true; - m_isRUpToDate = false; - } - /** - * Get the number of rows of the input matrix and the Q matrix - */ - inline Index rows() const {return m_rows; } - - /** - * Get the number of columns of the input matrix. - */ - inline Index cols() const { return m_cR->ncol; } - - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const - { - eigen_assert(m_isInitialized && " The QR factorization should be computed first, call compute()"); - eigen_assert(b.cols()==1 && "This method is for vectors only"); - - //Compute Q^T * b - typename Dest::PlainObject y, y2; - y = matrixQ().transpose() * b; - - // Solves with the triangular matrix R - Index rk = this->rank(); - y2 = y; - y.resize((std::max)(cols(),Index(y.rows())),y.cols()); - y.topRows(rk) = this->matrixR().topLeftCorner(rk, rk).template triangularView().solve(y2.topRows(rk)); - - // Apply the column permutation - // colsPermutation() performs a copy of the permutation, - // so let's apply it manually: - for(Index i = 0; i < rk; ++i) dest.row(m_E[i]) = y.row(i); - for(Index i = rk; i < cols(); ++i) dest.row(m_E[i]).setZero(); - -// y.bottomRows(y.rows()-rk).setZero(); -// dest = colsPermutation() * y.topRows(cols()); - - m_info = Success; - } - - /** \returns the sparse triangular factor R. It is a sparse matrix - */ - const MatrixType matrixR() const - { - eigen_assert(m_isInitialized && " The QR factorization should be computed first, call compute()"); - if(!m_isRUpToDate) { - m_R = viewAsEigen(*m_cR); - m_isRUpToDate = true; - } - return m_R; - } - /// Get an expression of the matrix Q - SPQRMatrixQReturnType matrixQ() const - { - return SPQRMatrixQReturnType(*this); - } - /// Get the permutation that was applied to columns of A - PermutationType colsPermutation() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return PermutationType(m_E, m_cR->ncol); - } - /** - * Gets the rank of the matrix. - * It should be equal to matrixQR().cols if the matrix is full-rank - */ - Index rank() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_cc.SPQR_istat[4]; - } - /// Set the fill-reducing ordering method to be used - void setSPQROrdering(int ord) { m_ordering = ord;} - /// Set the tolerance tol to treat columns with 2-norm < =tol as zero - void setPivotThreshold(const RealScalar& tol) - { - m_useDefaultThreshold = false; - m_tolerance = tol; - } - - /** \returns a pointer to the SPQR workspace */ - cholmod_common *cholmodCommon() const { return &m_cc; } - - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the sparse QR can not be computed - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - protected: - bool m_analysisIsOk; - bool m_factorizationIsOk; - mutable bool m_isRUpToDate; - mutable ComputationInfo m_info; - int m_ordering; // Ordering method to use, see SPQR's manual - int m_allow_tol; // Allow to use some tolerance during numerical factorization. - RealScalar m_tolerance; // treat columns with 2-norm below this tolerance as zero - mutable cholmod_sparse *m_cR; // The sparse R factor in cholmod format - mutable MatrixType m_R; // The sparse matrix R in Eigen format - mutable StorageIndex *m_E; // The permutation applied to columns - mutable cholmod_sparse *m_H; //The householder vectors - mutable StorageIndex *m_HPinv; // The row permutation of H - mutable cholmod_dense *m_HTau; // The Householder coefficients - mutable Index m_rank; // The rank of the matrix - mutable cholmod_common m_cc; // Workspace and parameters - bool m_useDefaultThreshold; // Use default threshold - Index m_rows; - template friend struct SPQR_QProduct; -}; - -template -struct SPQR_QProduct : ReturnByValue > -{ - typedef typename SPQRType::Scalar Scalar; - typedef typename SPQRType::StorageIndex StorageIndex; - //Define the constructor to get reference to argument types - SPQR_QProduct(const SPQRType& spqr, const Derived& other, bool transpose) : m_spqr(spqr),m_other(other),m_transpose(transpose) {} - - inline Index rows() const { return m_transpose ? m_spqr.rows() : m_spqr.cols(); } - inline Index cols() const { return m_other.cols(); } - // Assign to a vector - template - void evalTo(ResType& res) const - { - cholmod_dense y_cd; - cholmod_dense *x_cd; - int method = m_transpose ? SPQR_QTX : SPQR_QX; - cholmod_common *cc = m_spqr.cholmodCommon(); - y_cd = viewAsCholmod(m_other.const_cast_derived()); - x_cd = SuiteSparseQR_qmult(method, m_spqr.m_H, m_spqr.m_HTau, m_spqr.m_HPinv, &y_cd, cc); - res = Matrix::Map(reinterpret_cast(x_cd->x), x_cd->nrow, x_cd->ncol); - cholmod_l_free_dense(&x_cd, cc); - } - const SPQRType& m_spqr; - const Derived& m_other; - bool m_transpose; - -}; -template -struct SPQRMatrixQReturnType{ - - SPQRMatrixQReturnType(const SPQRType& spqr) : m_spqr(spqr) {} - template - SPQR_QProduct operator*(const MatrixBase& other) - { - return SPQR_QProduct(m_spqr,other.derived(),false); - } - SPQRMatrixQTransposeReturnType adjoint() const - { - return SPQRMatrixQTransposeReturnType(m_spqr); - } - // To use for operations with the transpose of Q - SPQRMatrixQTransposeReturnType transpose() const - { - return SPQRMatrixQTransposeReturnType(m_spqr); - } - const SPQRType& m_spqr; -}; - -template -struct SPQRMatrixQTransposeReturnType{ - SPQRMatrixQTransposeReturnType(const SPQRType& spqr) : m_spqr(spqr) {} - template - SPQR_QProduct operator*(const MatrixBase& other) - { - return SPQR_QProduct(m_spqr,other.derived(), true); - } - const SPQRType& m_spqr; -}; - -}// End namespace Eigen -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/BDCSVD.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/BDCSVD.h deleted file mode 100644 index a5b73f8f21f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/BDCSVD.h +++ /dev/null @@ -1,1277 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// We used the "A Divide-And-Conquer Algorithm for the Bidiagonal SVD" -// research report written by Ming Gu and Stanley C.Eisenstat -// The code variable names correspond to the names they used in their -// report -// -// Copyright (C) 2013 Gauthier Brun -// Copyright (C) 2013 Nicolas Carre -// Copyright (C) 2013 Jean Ceccato -// Copyright (C) 2013 Pierre Zoppitelli -// Copyright (C) 2013 Jitse Niesen -// Copyright (C) 2014-2017 Gael Guennebaud -// -// Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BDCSVD_H -#define EIGEN_BDCSVD_H -// #define EIGEN_BDCSVD_DEBUG_VERBOSE -// #define EIGEN_BDCSVD_SANITY_CHECKS - -namespace Eigen { - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE -IOFormat bdcsvdfmt(8, 0, ", ", "\n", " [", "]"); -#endif - -template class BDCSVD; - -namespace internal { - -template -struct traits > -{ - typedef _MatrixType MatrixType; -}; - -} // end namespace internal - - -/** \ingroup SVD_Module - * - * - * \class BDCSVD - * - * \brief class Bidiagonal Divide and Conquer SVD - * - * \tparam _MatrixType the type of the matrix of which we are computing the SVD decomposition - * - * This class first reduces the input matrix to bi-diagonal form using class UpperBidiagonalization, - * and then performs a divide-and-conquer diagonalization. Small blocks are diagonalized using class JacobiSVD. - * You can control the switching size with the setSwitchSize() method, default is 16. - * For small matrice (<16), it is thus preferable to directly use JacobiSVD. For larger ones, BDCSVD is highly - * recommended and can several order of magnitude faster. - * - * \warning this algorithm is unlikely to provide accurate result when compiled with unsafe math optimizations. - * For instance, this concerns Intel's compiler (ICC), which perfroms such optimization by default unless - * you compile with the \c -fp-model \c precise option. Likewise, the \c -ffast-math option of GCC or clang will - * significantly degrade the accuracy. - * - * \sa class JacobiSVD - */ -template -class BDCSVD : public SVDBase > -{ - typedef SVDBase Base; - -public: - using Base::rows; - using Base::cols; - using Base::computeU; - using Base::computeV; - - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef typename NumTraits::Literal Literal; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - DiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime, ColsAtCompileTime), - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - MaxDiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(MaxRowsAtCompileTime, MaxColsAtCompileTime), - MatrixOptions = MatrixType::Options - }; - - typedef typename Base::MatrixUType MatrixUType; - typedef typename Base::MatrixVType MatrixVType; - typedef typename Base::SingularValuesType SingularValuesType; - - typedef Matrix MatrixX; - typedef Matrix MatrixXr; - typedef Matrix VectorType; - typedef Array ArrayXr; - typedef Array ArrayXi; - typedef Ref ArrayRef; - typedef Ref IndicesRef; - - /** \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via BDCSVD::compute(const MatrixType&). - */ - BDCSVD() : m_algoswap(16), m_numIters(0) - {} - - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem size. - * \sa BDCSVD() - */ - BDCSVD(Index rows, Index cols, unsigned int computationOptions = 0) - : m_algoswap(16), m_numIters(0) - { - allocate(rows, cols, computationOptions); - } - - /** \brief Constructor performing the decomposition of given matrix. - * - * \param matrix the matrix to decompose - * \param computationOptions optional parameter allowing to specify if you want full or thin U or V unitaries to be computed. - * By default, none is computed. This is a bit - field, the possible bits are #ComputeFullU, #ComputeThinU, - * #ComputeFullV, #ComputeThinV. - * - * Thin unitaries are only available if your matrix type has a Dynamic number of columns (for example MatrixXf). They also are not - * available with the (non - default) FullPivHouseholderQR preconditioner. - */ - BDCSVD(const MatrixType& matrix, unsigned int computationOptions = 0) - : m_algoswap(16), m_numIters(0) - { - compute(matrix, computationOptions); - } - - ~BDCSVD() - { - } - - /** \brief Method performing the decomposition of given matrix using custom options. - * - * \param matrix the matrix to decompose - * \param computationOptions optional parameter allowing to specify if you want full or thin U or V unitaries to be computed. - * By default, none is computed. This is a bit - field, the possible bits are #ComputeFullU, #ComputeThinU, - * #ComputeFullV, #ComputeThinV. - * - * Thin unitaries are only available if your matrix type has a Dynamic number of columns (for example MatrixXf). They also are not - * available with the (non - default) FullPivHouseholderQR preconditioner. - */ - BDCSVD& compute(const MatrixType& matrix, unsigned int computationOptions); - - /** \brief Method performing the decomposition of given matrix using current options. - * - * \param matrix the matrix to decompose - * - * This method uses the current \a computationOptions, as already passed to the constructor or to compute(const MatrixType&, unsigned int). - */ - BDCSVD& compute(const MatrixType& matrix) - { - return compute(matrix, this->m_computationOptions); - } - - void setSwitchSize(int s) - { - eigen_assert(s>3 && "BDCSVD the size of the algo switch has to be greater than 3"); - m_algoswap = s; - } - -private: - void allocate(Index rows, Index cols, unsigned int computationOptions); - void divide(Index firstCol, Index lastCol, Index firstRowW, Index firstColW, Index shift); - void computeSVDofM(Index firstCol, Index n, MatrixXr& U, VectorType& singVals, MatrixXr& V); - void computeSingVals(const ArrayRef& col0, const ArrayRef& diag, const IndicesRef& perm, VectorType& singVals, ArrayRef shifts, ArrayRef mus); - void perturbCol0(const ArrayRef& col0, const ArrayRef& diag, const IndicesRef& perm, const VectorType& singVals, const ArrayRef& shifts, const ArrayRef& mus, ArrayRef zhat); - void computeSingVecs(const ArrayRef& zhat, const ArrayRef& diag, const IndicesRef& perm, const VectorType& singVals, const ArrayRef& shifts, const ArrayRef& mus, MatrixXr& U, MatrixXr& V); - void deflation43(Index firstCol, Index shift, Index i, Index size); - void deflation44(Index firstColu , Index firstColm, Index firstRowW, Index firstColW, Index i, Index j, Index size); - void deflation(Index firstCol, Index lastCol, Index k, Index firstRowW, Index firstColW, Index shift); - template - void copyUV(const HouseholderU &householderU, const HouseholderV &householderV, const NaiveU &naiveU, const NaiveV &naivev); - void structured_update(Block A, const MatrixXr &B, Index n1); - static RealScalar secularEq(RealScalar x, const ArrayRef& col0, const ArrayRef& diag, const IndicesRef &perm, const ArrayRef& diagShifted, RealScalar shift); - -protected: - MatrixXr m_naiveU, m_naiveV; - MatrixXr m_computed; - Index m_nRec; - ArrayXr m_workspace; - ArrayXi m_workspaceI; - int m_algoswap; - bool m_isTranspose, m_compU, m_compV; - - using Base::m_singularValues; - using Base::m_diagSize; - using Base::m_computeFullU; - using Base::m_computeFullV; - using Base::m_computeThinU; - using Base::m_computeThinV; - using Base::m_matrixU; - using Base::m_matrixV; - using Base::m_isInitialized; - using Base::m_nonzeroSingularValues; - -public: - int m_numIters; -}; //end class BDCSVD - - -// Method to allocate and initialize matrix and attributes -template -void BDCSVD::allocate(Index rows, Index cols, unsigned int computationOptions) -{ - m_isTranspose = (cols > rows); - - if (Base::allocate(rows, cols, computationOptions)) - return; - - m_computed = MatrixXr::Zero(m_diagSize + 1, m_diagSize ); - m_compU = computeV(); - m_compV = computeU(); - if (m_isTranspose) - std::swap(m_compU, m_compV); - - if (m_compU) m_naiveU = MatrixXr::Zero(m_diagSize + 1, m_diagSize + 1 ); - else m_naiveU = MatrixXr::Zero(2, m_diagSize + 1 ); - - if (m_compV) m_naiveV = MatrixXr::Zero(m_diagSize, m_diagSize); - - m_workspace.resize((m_diagSize+1)*(m_diagSize+1)*3); - m_workspaceI.resize(3*m_diagSize); -}// end allocate - -template -BDCSVD& BDCSVD::compute(const MatrixType& matrix, unsigned int computationOptions) -{ -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "\n\n\n======================================================================================================================\n\n\n"; -#endif - allocate(matrix.rows(), matrix.cols(), computationOptions); - using std::abs; - - const RealScalar considerZero = (std::numeric_limits::min)(); - - //**** step -1 - If the problem is too small, directly falls back to JacobiSVD and return - if(matrix.cols() < m_algoswap) - { - // FIXME this line involves temporaries - JacobiSVD jsvd(matrix,computationOptions); - if(computeU()) m_matrixU = jsvd.matrixU(); - if(computeV()) m_matrixV = jsvd.matrixV(); - m_singularValues = jsvd.singularValues(); - m_nonzeroSingularValues = jsvd.nonzeroSingularValues(); - m_isInitialized = true; - return *this; - } - - //**** step 0 - Copy the input matrix and apply scaling to reduce over/under-flows - RealScalar scale = matrix.cwiseAbs().maxCoeff(); - if(scale==Literal(0)) scale = Literal(1); - MatrixX copy; - if (m_isTranspose) copy = matrix.adjoint()/scale; - else copy = matrix/scale; - - //**** step 1 - Bidiagonalization - // FIXME this line involves temporaries - internal::UpperBidiagonalization bid(copy); - - //**** step 2 - Divide & Conquer - m_naiveU.setZero(); - m_naiveV.setZero(); - // FIXME this line involves a temporary matrix - m_computed.topRows(m_diagSize) = bid.bidiagonal().toDenseMatrix().transpose(); - m_computed.template bottomRows<1>().setZero(); - divide(0, m_diagSize - 1, 0, 0, 0); - - //**** step 3 - Copy singular values and vectors - for (int i=0; i -template -void BDCSVD::copyUV(const HouseholderU &householderU, const HouseholderV &householderV, const NaiveU &naiveU, const NaiveV &naiveV) -{ - // Note exchange of U and V: m_matrixU is set from m_naiveV and vice versa - if (computeU()) - { - Index Ucols = m_computeThinU ? m_diagSize : householderU.cols(); - m_matrixU = MatrixX::Identity(householderU.cols(), Ucols); - m_matrixU.topLeftCorner(m_diagSize, m_diagSize) = naiveV.template cast().topLeftCorner(m_diagSize, m_diagSize); - householderU.applyThisOnTheLeft(m_matrixU); // FIXME this line involves a temporary buffer - } - if (computeV()) - { - Index Vcols = m_computeThinV ? m_diagSize : householderV.cols(); - m_matrixV = MatrixX::Identity(householderV.cols(), Vcols); - m_matrixV.topLeftCorner(m_diagSize, m_diagSize) = naiveU.template cast().topLeftCorner(m_diagSize, m_diagSize); - householderV.applyThisOnTheLeft(m_matrixV); // FIXME this line involves a temporary buffer - } -} - -/** \internal - * Performs A = A * B exploiting the special structure of the matrix A. Splitting A as: - * A = [A1] - * [A2] - * such that A1.rows()==n1, then we assume that at least half of the columns of A1 and A2 are zeros. - * We can thus pack them prior to the the matrix product. However, this is only worth the effort if the matrix is large - * enough. - */ -template -void BDCSVD::structured_update(Block A, const MatrixXr &B, Index n1) -{ - Index n = A.rows(); - if(n>100) - { - // If the matrices are large enough, let's exploit the sparse structure of A by - // splitting it in half (wrt n1), and packing the non-zero columns. - Index n2 = n - n1; - Map A1(m_workspace.data() , n1, n); - Map A2(m_workspace.data()+ n1*n, n2, n); - Map B1(m_workspace.data()+ n*n, n, n); - Map B2(m_workspace.data()+2*n*n, n, n); - Index k1=0, k2=0; - for(Index j=0; j tmp(m_workspace.data(),n,n); - tmp.noalias() = A*B; - A = tmp; - } -} - -// The divide algorithm is done "in place", we are always working on subsets of the same matrix. The divide methods takes as argument the -// place of the submatrix we are currently working on. - -//@param firstCol : The Index of the first column of the submatrix of m_computed and for m_naiveU; -//@param lastCol : The Index of the last column of the submatrix of m_computed and for m_naiveU; -// lastCol + 1 - firstCol is the size of the submatrix. -//@param firstRowW : The Index of the first row of the matrix W that we are to change. (see the reference paper section 1 for more information on W) -//@param firstRowW : Same as firstRowW with the column. -//@param shift : Each time one takes the left submatrix, one must add 1 to the shift. Why? Because! We actually want the last column of the U submatrix -// to become the first column (*coeff) and to shift all the other columns to the right. There are more details on the reference paper. -template -void BDCSVD::divide (Index firstCol, Index lastCol, Index firstRowW, Index firstColW, Index shift) -{ - // requires rows = cols + 1; - using std::pow; - using std::sqrt; - using std::abs; - const Index n = lastCol - firstCol + 1; - const Index k = n/2; - const RealScalar considerZero = (std::numeric_limits::min)(); - RealScalar alphaK; - RealScalar betaK; - RealScalar r0; - RealScalar lambda, phi, c0, s0; - VectorType l, f; - // We use the other algorithm which is more efficient for small - // matrices. - if (n < m_algoswap) - { - // FIXME this line involves temporaries - JacobiSVD b(m_computed.block(firstCol, firstCol, n + 1, n), ComputeFullU | (m_compV ? ComputeFullV : 0)); - if (m_compU) - m_naiveU.block(firstCol, firstCol, n + 1, n + 1).real() = b.matrixU(); - else - { - m_naiveU.row(0).segment(firstCol, n + 1).real() = b.matrixU().row(0); - m_naiveU.row(1).segment(firstCol, n + 1).real() = b.matrixU().row(n); - } - if (m_compV) m_naiveV.block(firstRowW, firstColW, n, n).real() = b.matrixV(); - m_computed.block(firstCol + shift, firstCol + shift, n + 1, n).setZero(); - m_computed.diagonal().segment(firstCol + shift, n) = b.singularValues().head(n); - return; - } - // We use the divide and conquer algorithm - alphaK = m_computed(firstCol + k, firstCol + k); - betaK = m_computed(firstCol + k + 1, firstCol + k); - // The divide must be done in that order in order to have good results. Divide change the data inside the submatrices - // and the divide of the right submatrice reads one column of the left submatrice. That's why we need to treat the - // right submatrix before the left one. - divide(k + 1 + firstCol, lastCol, k + 1 + firstRowW, k + 1 + firstColW, shift); - divide(firstCol, k - 1 + firstCol, firstRowW, firstColW + 1, shift + 1); - - if (m_compU) - { - lambda = m_naiveU(firstCol + k, firstCol + k); - phi = m_naiveU(firstCol + k + 1, lastCol + 1); - } - else - { - lambda = m_naiveU(1, firstCol + k); - phi = m_naiveU(0, lastCol + 1); - } - r0 = sqrt((abs(alphaK * lambda) * abs(alphaK * lambda)) + abs(betaK * phi) * abs(betaK * phi)); - if (m_compU) - { - l = m_naiveU.row(firstCol + k).segment(firstCol, k); - f = m_naiveU.row(firstCol + k + 1).segment(firstCol + k + 1, n - k - 1); - } - else - { - l = m_naiveU.row(1).segment(firstCol, k); - f = m_naiveU.row(0).segment(firstCol + k + 1, n - k - 1); - } - if (m_compV) m_naiveV(firstRowW+k, firstColW) = Literal(1); - if (r0= firstCol; i--) - m_naiveU.col(i + 1).segment(firstCol, k + 1) = m_naiveU.col(i).segment(firstCol, k + 1); - // we shift q1 at the left with a factor c0 - m_naiveU.col(firstCol).segment( firstCol, k + 1) = (q1 * c0); - // last column = q1 * - s0 - m_naiveU.col(lastCol + 1).segment(firstCol, k + 1) = (q1 * ( - s0)); - // first column = q2 * s0 - m_naiveU.col(firstCol).segment(firstCol + k + 1, n - k) = m_naiveU.col(lastCol + 1).segment(firstCol + k + 1, n - k) * s0; - // q2 *= c0 - m_naiveU.col(lastCol + 1).segment(firstCol + k + 1, n - k) *= c0; - } - else - { - RealScalar q1 = m_naiveU(0, firstCol + k); - // we shift Q1 to the right - for (Index i = firstCol + k - 1; i >= firstCol; i--) - m_naiveU(0, i + 1) = m_naiveU(0, i); - // we shift q1 at the left with a factor c0 - m_naiveU(0, firstCol) = (q1 * c0); - // last column = q1 * - s0 - m_naiveU(0, lastCol + 1) = (q1 * ( - s0)); - // first column = q2 * s0 - m_naiveU(1, firstCol) = m_naiveU(1, lastCol + 1) *s0; - // q2 *= c0 - m_naiveU(1, lastCol + 1) *= c0; - m_naiveU.row(1).segment(firstCol + 1, k).setZero(); - m_naiveU.row(0).segment(firstCol + k + 1, n - k - 1).setZero(); - } - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(m_naiveU.allFinite()); - assert(m_naiveV.allFinite()); - assert(m_computed.allFinite()); -#endif - - m_computed(firstCol + shift, firstCol + shift) = r0; - m_computed.col(firstCol + shift).segment(firstCol + shift + 1, k) = alphaK * l.transpose().real(); - m_computed.col(firstCol + shift).segment(firstCol + shift + k + 1, n - k - 1) = betaK * f.transpose().real(); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - ArrayXr tmp1 = (m_computed.block(firstCol+shift, firstCol+shift, n, n)).jacobiSvd().singularValues(); -#endif - // Second part: try to deflate singular values in combined matrix - deflation(firstCol, lastCol, k, firstRowW, firstColW, shift); -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - ArrayXr tmp2 = (m_computed.block(firstCol+shift, firstCol+shift, n, n)).jacobiSvd().singularValues(); - std::cout << "\n\nj1 = " << tmp1.transpose().format(bdcsvdfmt) << "\n"; - std::cout << "j2 = " << tmp2.transpose().format(bdcsvdfmt) << "\n\n"; - std::cout << "err: " << ((tmp1-tmp2).abs()>1e-12*tmp2.abs()).transpose() << "\n"; - static int count = 0; - std::cout << "# " << ++count << "\n\n"; - assert((tmp1-tmp2).matrix().norm() < 1e-14*tmp2.matrix().norm()); -// assert(count<681); -// assert(((tmp1-tmp2).abs()<1e-13*tmp2.abs()).all()); -#endif - - // Third part: compute SVD of combined matrix - MatrixXr UofSVD, VofSVD; - VectorType singVals; - computeSVDofM(firstCol + shift, n, UofSVD, singVals, VofSVD); - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(UofSVD.allFinite()); - assert(VofSVD.allFinite()); -#endif - - if (m_compU) - structured_update(m_naiveU.block(firstCol, firstCol, n + 1, n + 1), UofSVD, (n+2)/2); - else - { - Map,Aligned> tmp(m_workspace.data(),2,n+1); - tmp.noalias() = m_naiveU.middleCols(firstCol, n+1) * UofSVD; - m_naiveU.middleCols(firstCol, n + 1) = tmp; - } - - if (m_compV) structured_update(m_naiveV.block(firstRowW, firstColW, n, n), VofSVD, (n+1)/2); - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(m_naiveU.allFinite()); - assert(m_naiveV.allFinite()); - assert(m_computed.allFinite()); -#endif - - m_computed.block(firstCol + shift, firstCol + shift, n, n).setZero(); - m_computed.block(firstCol + shift, firstCol + shift, n, n).diagonal() = singVals; -}// end divide - -// Compute SVD of m_computed.block(firstCol, firstCol, n + 1, n); this block only has non-zeros in -// the first column and on the diagonal and has undergone deflation, so diagonal is in increasing -// order except for possibly the (0,0) entry. The computed SVD is stored U, singVals and V, except -// that if m_compV is false, then V is not computed. Singular values are sorted in decreasing order. -// -// TODO Opportunities for optimization: better root finding algo, better stopping criterion, better -// handling of round-off errors, be consistent in ordering -// For instance, to solve the secular equation using FMM, see http://www.stat.uchicago.edu/~lekheng/courses/302/classics/greengard-rokhlin.pdf -template -void BDCSVD::computeSVDofM(Index firstCol, Index n, MatrixXr& U, VectorType& singVals, MatrixXr& V) -{ - const RealScalar considerZero = (std::numeric_limits::min)(); - using std::abs; - ArrayRef col0 = m_computed.col(firstCol).segment(firstCol, n); - m_workspace.head(n) = m_computed.block(firstCol, firstCol, n, n).diagonal(); - ArrayRef diag = m_workspace.head(n); - diag(0) = Literal(0); - - // Allocate space for singular values and vectors - singVals.resize(n); - U.resize(n+1, n+1); - if (m_compV) V.resize(n, n); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - if (col0.hasNaN() || diag.hasNaN()) - std::cout << "\n\nHAS NAN\n\n"; -#endif - - // Many singular values might have been deflated, the zero ones have been moved to the end, - // but others are interleaved and we must ignore them at this stage. - // To this end, let's compute a permutation skipping them: - Index actual_n = n; - while(actual_n>1 && diag(actual_n-1)==Literal(0)) --actual_n; - Index m = 0; // size of the deflated problem - for(Index k=0;kconsiderZero) - m_workspaceI(m++) = k; - Map perm(m_workspaceI.data(),m); - - Map shifts(m_workspace.data()+1*n, n); - Map mus(m_workspace.data()+2*n, n); - Map zhat(m_workspace.data()+3*n, n); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "computeSVDofM using:\n"; - std::cout << " z: " << col0.transpose() << "\n"; - std::cout << " d: " << diag.transpose() << "\n"; -#endif - - // Compute singVals, shifts, and mus - computeSingVals(col0, diag, perm, singVals, shifts, mus); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << " j: " << (m_computed.block(firstCol, firstCol, n, n)).jacobiSvd().singularValues().transpose().reverse() << "\n\n"; - std::cout << " sing-val: " << singVals.transpose() << "\n"; - std::cout << " mu: " << mus.transpose() << "\n"; - std::cout << " shift: " << shifts.transpose() << "\n"; - - { - Index actual_n = n; - while(actual_n>1 && abs(col0(actual_n-1))0) : " << ((singVals.array()-diag) / singVals.array()).head(actual_n).transpose() << "\n\n"; - std::cout << " check3 (>0) : " << ((diag.segment(1,actual_n-1)-singVals.head(actual_n-1).array()) / singVals.head(actual_n-1).array()).transpose() << "\n\n\n"; - std::cout << " check4 (>0) : " << ((singVals.segment(1,actual_n-1)-singVals.head(actual_n-1))).transpose() << "\n\n\n"; - } -#endif - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(singVals.allFinite()); - assert(mus.allFinite()); - assert(shifts.allFinite()); -#endif - - // Compute zhat - perturbCol0(col0, diag, perm, singVals, shifts, mus, zhat); -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << " zhat: " << zhat.transpose() << "\n"; -#endif - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(zhat.allFinite()); -#endif - - computeSingVecs(zhat, diag, perm, singVals, shifts, mus, U, V); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "U^T U: " << (U.transpose() * U - MatrixXr(MatrixXr::Identity(U.cols(),U.cols()))).norm() << "\n"; - std::cout << "V^T V: " << (V.transpose() * V - MatrixXr(MatrixXr::Identity(V.cols(),V.cols()))).norm() << "\n"; -#endif - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(U.allFinite()); - assert(V.allFinite()); - assert((U.transpose() * U - MatrixXr(MatrixXr::Identity(U.cols(),U.cols()))).norm() < 1e-14 * n); - assert((V.transpose() * V - MatrixXr(MatrixXr::Identity(V.cols(),V.cols()))).norm() < 1e-14 * n); - assert(m_naiveU.allFinite()); - assert(m_naiveV.allFinite()); - assert(m_computed.allFinite()); -#endif - - // Because of deflation, the singular values might not be completely sorted. - // Fortunately, reordering them is a O(n) problem - for(Index i=0; isingVals(i+1)) - { - using std::swap; - swap(singVals(i),singVals(i+1)); - U.col(i).swap(U.col(i+1)); - if(m_compV) V.col(i).swap(V.col(i+1)); - } - } - - // Reverse order so that singular values in increased order - // Because of deflation, the zeros singular-values are already at the end - singVals.head(actual_n).reverseInPlace(); - U.leftCols(actual_n).rowwise().reverseInPlace(); - if (m_compV) V.leftCols(actual_n).rowwise().reverseInPlace(); - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - JacobiSVD jsvd(m_computed.block(firstCol, firstCol, n, n) ); - std::cout << " * j: " << jsvd.singularValues().transpose() << "\n\n"; - std::cout << " * sing-val: " << singVals.transpose() << "\n"; -// std::cout << " * err: " << ((jsvd.singularValues()-singVals)>1e-13*singVals.norm()).transpose() << "\n"; -#endif -} - -template -typename BDCSVD::RealScalar BDCSVD::secularEq(RealScalar mu, const ArrayRef& col0, const ArrayRef& diag, const IndicesRef &perm, const ArrayRef& diagShifted, RealScalar shift) -{ - Index m = perm.size(); - RealScalar res = Literal(1); - for(Index i=0; i -void BDCSVD::computeSingVals(const ArrayRef& col0, const ArrayRef& diag, const IndicesRef &perm, - VectorType& singVals, ArrayRef shifts, ArrayRef mus) -{ - using std::abs; - using std::swap; - using std::sqrt; - - Index n = col0.size(); - Index actual_n = n; - // Note that here actual_n is computed based on col0(i)==0 instead of diag(i)==0 as above - // because 1) we have diag(i)==0 => col0(i)==0 and 2) if col0(i)==0, then diag(i) is already a singular value. - while(actual_n>1 && col0(actual_n-1)==Literal(0)) --actual_n; - - for (Index k = 0; k < n; ++k) - { - if (col0(k) == Literal(0) || actual_n==1) - { - // if col0(k) == 0, then entry is deflated, so singular value is on diagonal - // if actual_n==1, then the deflated problem is already diagonalized - singVals(k) = k==0 ? col0(0) : diag(k); - mus(k) = Literal(0); - shifts(k) = k==0 ? col0(0) : diag(k); - continue; - } - - // otherwise, use secular equation to find singular value - RealScalar left = diag(k); - RealScalar right; // was: = (k != actual_n-1) ? diag(k+1) : (diag(actual_n-1) + col0.matrix().norm()); - if(k==actual_n-1) - right = (diag(actual_n-1) + col0.matrix().norm()); - else - { - // Skip deflated singular values, - // recall that at this stage we assume that z[j]!=0 and all entries for which z[j]==0 have been put aside. - // This should be equivalent to using perm[] - Index l = k+1; - while(col0(l)==Literal(0)) { ++l; eigen_internal_assert(l Literal(0)) ? left : right; - - // measure everything relative to shift - Map diagShifted(m_workspace.data()+4*n, n); - diagShifted = diag - shift; - - if(k!=actual_n-1) - { - // check that after the shift, f(mid) is still negative: - RealScalar midShifted = (right - left) / RealScalar(2); - if(shift==right) - midShifted = -midShifted; - RealScalar fMidShifted = secularEq(midShifted, col0, diag, perm, diagShifted, shift); - if(fMidShifted>0) - { - // fMid was erroneous, fix it: - shift = fMidShifted > Literal(0) ? left : right; - diagShifted = diag - shift; - } - } - - // initial guess - RealScalar muPrev, muCur; - if (shift == left) - { - muPrev = (right - left) * RealScalar(0.1); - if (k == actual_n-1) muCur = right - left; - else muCur = (right - left) * RealScalar(0.5); - } - else - { - muPrev = -(right - left) * RealScalar(0.1); - muCur = -(right - left) * RealScalar(0.5); - } - - RealScalar fPrev = secularEq(muPrev, col0, diag, perm, diagShifted, shift); - RealScalar fCur = secularEq(muCur, col0, diag, perm, diagShifted, shift); - if (abs(fPrev) < abs(fCur)) - { - swap(fPrev, fCur); - swap(muPrev, muCur); - } - - // rational interpolation: fit a function of the form a / mu + b through the two previous - // iterates and use its zero to compute the next iterate - bool useBisection = fPrev*fCur>Literal(0); - while (fCur!=Literal(0) && abs(muCur - muPrev) > Literal(8) * NumTraits::epsilon() * numext::maxi(abs(muCur), abs(muPrev)) && abs(fCur - fPrev)>NumTraits::epsilon() && !useBisection) - { - ++m_numIters; - - // Find a and b such that the function f(mu) = a / mu + b matches the current and previous samples. - RealScalar a = (fCur - fPrev) / (Literal(1)/muCur - Literal(1)/muPrev); - RealScalar b = fCur - a / muCur; - // And find mu such that f(mu)==0: - RealScalar muZero = -a/b; - RealScalar fZero = secularEq(muZero, col0, diag, perm, diagShifted, shift); - - muPrev = muCur; - fPrev = fCur; - muCur = muZero; - fCur = fZero; - - - if (shift == left && (muCur < Literal(0) || muCur > right - left)) useBisection = true; - if (shift == right && (muCur < -(right - left) || muCur > Literal(0))) useBisection = true; - if (abs(fCur)>abs(fPrev)) useBisection = true; - } - - // fall back on bisection method if rational interpolation did not work - if (useBisection) - { -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "useBisection for k = " << k << ", actual_n = " << actual_n << "\n"; -#endif - RealScalar leftShifted, rightShifted; - if (shift == left) - { - // to avoid overflow, we must have mu > max(real_min, |z(k)|/sqrt(real_max)), - // the factor 2 is to be more conservative - leftShifted = numext::maxi( (std::numeric_limits::min)(), Literal(2) * abs(col0(k)) / sqrt((std::numeric_limits::max)()) ); - - // check that we did it right: - eigen_internal_assert( (numext::isfinite)( (col0(k)/leftShifted)*(col0(k)/(diag(k)+shift+leftShifted)) ) ); - // I don't understand why the case k==0 would be special there: - // if (k == 0) rightShifted = right - left; else - rightShifted = (k==actual_n-1) ? right : ((right - left) * RealScalar(0.51)); // theoretically we can take 0.5, but let's be safe - } - else - { - leftShifted = -(right - left) * RealScalar(0.51); - if(k+1( (std::numeric_limits::min)(), abs(col0(k+1)) / sqrt((std::numeric_limits::max)()) ); - else - rightShifted = -(std::numeric_limits::min)(); - } - - RealScalar fLeft = secularEq(leftShifted, col0, diag, perm, diagShifted, shift); - eigen_internal_assert(fLeft " << leftShifted << " " << rightShifted << " shift=" << shift << "\n"; - } -#endif - eigen_internal_assert(fLeft * fRight < Literal(0)); - - if(fLeft Literal(2) * NumTraits::epsilon() * numext::maxi(abs(leftShifted), abs(rightShifted))) - { - RealScalar midShifted = (leftShifted + rightShifted) / Literal(2); - fMid = secularEq(midShifted, col0, diag, perm, diagShifted, shift); - eigen_internal_assert((numext::isfinite)(fMid)); - - if (fLeft * fMid < Literal(0)) - { - rightShifted = midShifted; - } - else - { - leftShifted = midShifted; - fLeft = fMid; - } - } - muCur = (leftShifted + rightShifted) / Literal(2); - } - else - { - // We have a problem as shifting on the left or right give either a positive or negative value - // at the middle of [left,right]... - // Instead fo abbording or entering an infinite loop, - // let's just use the middle as the estimated zero-crossing: - muCur = (right - left) * RealScalar(0.5); - if(shift == right) - muCur = -muCur; - } - } - - singVals[k] = shift + muCur; - shifts[k] = shift; - mus[k] = muCur; - - // perturb singular value slightly if it equals diagonal entry to avoid division by zero later - // (deflation is supposed to avoid this from happening) - // - this does no seem to be necessary anymore - -// if (singVals[k] == left) singVals[k] *= 1 + NumTraits::epsilon(); -// if (singVals[k] == right) singVals[k] *= 1 - NumTraits::epsilon(); - } -} - - -// zhat is perturbation of col0 for which singular vectors can be computed stably (see Section 3.1) -template -void BDCSVD::perturbCol0 - (const ArrayRef& col0, const ArrayRef& diag, const IndicesRef &perm, const VectorType& singVals, - const ArrayRef& shifts, const ArrayRef& mus, ArrayRef zhat) -{ - using std::sqrt; - Index n = col0.size(); - Index m = perm.size(); - if(m==0) - { - zhat.setZero(); - return; - } - Index last = perm(m-1); - // The offset permits to skip deflated entries while computing zhat - for (Index k = 0; k < n; ++k) - { - if (col0(k) == Literal(0)) // deflated - zhat(k) = Literal(0); - else - { - // see equation (3.6) - RealScalar dk = diag(k); - RealScalar prod = (singVals(last) + dk) * (mus(last) + (shifts(last) - dk)); - - for(Index l = 0; l 0.9 ) - std::cout << " " << ((singVals(j)+dk)*(mus(j)+(shifts(j)-dk)))/((diag(i)+dk)*(diag(i)-dk)) << " == (" << (singVals(j)+dk) << " * " << (mus(j)+(shifts(j)-dk)) - << ") / (" << (diag(i)+dk) << " * " << (diag(i)-dk) << ")\n"; -#endif - } - } -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "zhat(" << k << ") = sqrt( " << prod << ") ; " << (singVals(last) + dk) << " * " << mus(last) + shifts(last) << " - " << dk << "\n"; -#endif - RealScalar tmp = sqrt(prod); - zhat(k) = col0(k) > Literal(0) ? RealScalar(tmp) : RealScalar(-tmp); - } - } -} - -// compute singular vectors -template -void BDCSVD::computeSingVecs - (const ArrayRef& zhat, const ArrayRef& diag, const IndicesRef &perm, const VectorType& singVals, - const ArrayRef& shifts, const ArrayRef& mus, MatrixXr& U, MatrixXr& V) -{ - Index n = zhat.size(); - Index m = perm.size(); - - for (Index k = 0; k < n; ++k) - { - if (zhat(k) == Literal(0)) - { - U.col(k) = VectorType::Unit(n+1, k); - if (m_compV) V.col(k) = VectorType::Unit(n, k); - } - else - { - U.col(k).setZero(); - for(Index l=0;l= 1, di almost null and zi non null. -// We use a rotation to zero out zi applied to the left of M -template -void BDCSVD::deflation43(Index firstCol, Index shift, Index i, Index size) -{ - using std::abs; - using std::sqrt; - using std::pow; - Index start = firstCol + shift; - RealScalar c = m_computed(start, start); - RealScalar s = m_computed(start+i, start); - RealScalar r = numext::hypot(c,s); - if (r == Literal(0)) - { - m_computed(start+i, start+i) = Literal(0); - return; - } - m_computed(start,start) = r; - m_computed(start+i, start) = Literal(0); - m_computed(start+i, start+i) = Literal(0); - - JacobiRotation J(c/r,-s/r); - if (m_compU) m_naiveU.middleRows(firstCol, size+1).applyOnTheRight(firstCol, firstCol+i, J); - else m_naiveU.applyOnTheRight(firstCol, firstCol+i, J); -}// end deflation 43 - - -// page 13 -// i,j >= 1, i!=j and |di - dj| < epsilon * norm2(M) -// We apply two rotations to have zj = 0; -// TODO deflation44 is still broken and not properly tested -template -void BDCSVD::deflation44(Index firstColu , Index firstColm, Index firstRowW, Index firstColW, Index i, Index j, Index size) -{ - using std::abs; - using std::sqrt; - using std::conj; - using std::pow; - RealScalar c = m_computed(firstColm+i, firstColm); - RealScalar s = m_computed(firstColm+j, firstColm); - RealScalar r = sqrt(numext::abs2(c) + numext::abs2(s)); -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "deflation 4.4: " << i << "," << j << " -> " << c << " " << s << " " << r << " ; " - << m_computed(firstColm + i-1, firstColm) << " " - << m_computed(firstColm + i, firstColm) << " " - << m_computed(firstColm + i+1, firstColm) << " " - << m_computed(firstColm + i+2, firstColm) << "\n"; - std::cout << m_computed(firstColm + i-1, firstColm + i-1) << " " - << m_computed(firstColm + i, firstColm+i) << " " - << m_computed(firstColm + i+1, firstColm+i+1) << " " - << m_computed(firstColm + i+2, firstColm+i+2) << "\n"; -#endif - if (r==Literal(0)) - { - m_computed(firstColm + i, firstColm + i) = m_computed(firstColm + j, firstColm + j); - return; - } - c/=r; - s/=r; - m_computed(firstColm + i, firstColm) = r; - m_computed(firstColm + j, firstColm + j) = m_computed(firstColm + i, firstColm + i); - m_computed(firstColm + j, firstColm) = Literal(0); - - JacobiRotation J(c,-s); - if (m_compU) m_naiveU.middleRows(firstColu, size+1).applyOnTheRight(firstColu + i, firstColu + j, J); - else m_naiveU.applyOnTheRight(firstColu+i, firstColu+j, J); - if (m_compV) m_naiveV.middleRows(firstRowW, size).applyOnTheRight(firstColW + i, firstColW + j, J); -}// end deflation 44 - - -// acts on block from (firstCol+shift, firstCol+shift) to (lastCol+shift, lastCol+shift) [inclusive] -template -void BDCSVD::deflation(Index firstCol, Index lastCol, Index k, Index firstRowW, Index firstColW, Index shift) -{ - using std::sqrt; - using std::abs; - const Index length = lastCol + 1 - firstCol; - - Block col0(m_computed, firstCol+shift, firstCol+shift, length, 1); - Diagonal fulldiag(m_computed); - VectorBlock,Dynamic> diag(fulldiag, firstCol+shift, length); - - const RealScalar considerZero = (std::numeric_limits::min)(); - RealScalar maxDiag = diag.tail((std::max)(Index(1),length-1)).cwiseAbs().maxCoeff(); - RealScalar epsilon_strict = numext::maxi(considerZero,NumTraits::epsilon() * maxDiag); - RealScalar epsilon_coarse = Literal(8) * NumTraits::epsilon() * numext::maxi(col0.cwiseAbs().maxCoeff(), maxDiag); - -#ifdef EIGEN_BDCSVD_SANITY_CHECKS - assert(m_naiveU.allFinite()); - assert(m_naiveV.allFinite()); - assert(m_computed.allFinite()); -#endif - -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "\ndeflate:" << diag.head(k+1).transpose() << " | " << diag.segment(k+1,length-k-1).transpose() << "\n"; -#endif - - //condition 4.1 - if (diag(0) < epsilon_coarse) - { -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "deflation 4.1, because " << diag(0) << " < " << epsilon_coarse << "\n"; -#endif - diag(0) = epsilon_coarse; - } - - //condition 4.2 - for (Index i=1;i k) permutation[p] = j++; - else if (j >= length) permutation[p] = i++; - else if (diag(i) < diag(j)) permutation[p] = j++; - else permutation[p] = i++; - } - } - - // If we have a total deflation, then we have to insert diag(0) at the right place - if(total_deflation) - { - for(Index i=1; i0 && (abs(diag(i))1;--i) - if( (diag(i) - diag(i-1)) < NumTraits::epsilon()*maxDiag ) - { -#ifdef EIGEN_BDCSVD_DEBUG_VERBOSE - std::cout << "deflation 4.4 with i = " << i << " because " << (diag(i) - diag(i-1)) << " < " << NumTraits::epsilon()*diag(i) << "\n"; -#endif - eigen_internal_assert(abs(diag(i) - diag(i-1)) -BDCSVD::PlainObject> -MatrixBase::bdcSvd(unsigned int computationOptions) const -{ - return BDCSVD(*this, computationOptions); -} -#endif - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD.h deleted file mode 100644 index 43488b1e0c2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD.h +++ /dev/null @@ -1,804 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Benoit Jacob -// Copyright (C) 2013-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_JACOBISVD_H -#define EIGEN_JACOBISVD_H - -namespace Eigen { - -namespace internal { -// forward declaration (needed by ICC) -// the empty body is required by MSVC -template::IsComplex> -struct svd_precondition_2x2_block_to_be_real {}; - -/*** QR preconditioners (R-SVD) - *** - *** Their role is to reduce the problem of computing the SVD to the case of a square matrix. - *** This approach, known as R-SVD, is an optimization for rectangular-enough matrices, and is a requirement for - *** JacobiSVD which by itself is only able to work on square matrices. - ***/ - -enum { PreconditionIfMoreColsThanRows, PreconditionIfMoreRowsThanCols }; - -template -struct qr_preconditioner_should_do_anything -{ - enum { a = MatrixType::RowsAtCompileTime != Dynamic && - MatrixType::ColsAtCompileTime != Dynamic && - MatrixType::ColsAtCompileTime <= MatrixType::RowsAtCompileTime, - b = MatrixType::RowsAtCompileTime != Dynamic && - MatrixType::ColsAtCompileTime != Dynamic && - MatrixType::RowsAtCompileTime <= MatrixType::ColsAtCompileTime, - ret = !( (QRPreconditioner == NoQRPreconditioner) || - (Case == PreconditionIfMoreColsThanRows && bool(a)) || - (Case == PreconditionIfMoreRowsThanCols && bool(b)) ) - }; -}; - -template::ret -> struct qr_preconditioner_impl {}; - -template -class qr_preconditioner_impl -{ -public: - void allocate(const JacobiSVD&) {} - bool run(JacobiSVD&, const MatrixType&) - { - return false; - } -}; - -/*** preconditioner using FullPivHouseholderQR ***/ - -template -class qr_preconditioner_impl -{ -public: - typedef typename MatrixType::Scalar Scalar; - enum - { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime - }; - typedef Matrix WorkspaceType; - - void allocate(const JacobiSVD& svd) - { - if (svd.rows() != m_qr.rows() || svd.cols() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.rows(), svd.cols()); - } - if (svd.m_computeFullU) m_workspace.resize(svd.rows()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.rows() > matrix.cols()) - { - m_qr.compute(matrix); - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.cols(),matrix.cols()).template triangularView(); - if(svd.m_computeFullU) m_qr.matrixQ().evalTo(svd.m_matrixU, m_workspace); - if(svd.computeV()) svd.m_matrixV = m_qr.colsPermutation(); - return true; - } - return false; - } -private: - typedef FullPivHouseholderQR QRType; - QRType m_qr; - WorkspaceType m_workspace; -}; - -template -class qr_preconditioner_impl -{ -public: - typedef typename MatrixType::Scalar Scalar; - enum - { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - TrOptions = RowsAtCompileTime==1 ? (MatrixType::Options & ~(RowMajor)) - : ColsAtCompileTime==1 ? (MatrixType::Options | RowMajor) - : MatrixType::Options - }; - typedef Matrix - TransposeTypeWithSameStorageOrder; - - void allocate(const JacobiSVD& svd) - { - if (svd.cols() != m_qr.rows() || svd.rows() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.cols(), svd.rows()); - } - m_adjoint.resize(svd.cols(), svd.rows()); - if (svd.m_computeFullV) m_workspace.resize(svd.cols()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.cols() > matrix.rows()) - { - m_adjoint = matrix.adjoint(); - m_qr.compute(m_adjoint); - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.rows(),matrix.rows()).template triangularView().adjoint(); - if(svd.m_computeFullV) m_qr.matrixQ().evalTo(svd.m_matrixV, m_workspace); - if(svd.computeU()) svd.m_matrixU = m_qr.colsPermutation(); - return true; - } - else return false; - } -private: - typedef FullPivHouseholderQR QRType; - QRType m_qr; - TransposeTypeWithSameStorageOrder m_adjoint; - typename internal::plain_row_type::type m_workspace; -}; - -/*** preconditioner using ColPivHouseholderQR ***/ - -template -class qr_preconditioner_impl -{ -public: - void allocate(const JacobiSVD& svd) - { - if (svd.rows() != m_qr.rows() || svd.cols() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.rows(), svd.cols()); - } - if (svd.m_computeFullU) m_workspace.resize(svd.rows()); - else if (svd.m_computeThinU) m_workspace.resize(svd.cols()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.rows() > matrix.cols()) - { - m_qr.compute(matrix); - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.cols(),matrix.cols()).template triangularView(); - if(svd.m_computeFullU) m_qr.householderQ().evalTo(svd.m_matrixU, m_workspace); - else if(svd.m_computeThinU) - { - svd.m_matrixU.setIdentity(matrix.rows(), matrix.cols()); - m_qr.householderQ().applyThisOnTheLeft(svd.m_matrixU, m_workspace); - } - if(svd.computeV()) svd.m_matrixV = m_qr.colsPermutation(); - return true; - } - return false; - } - -private: - typedef ColPivHouseholderQR QRType; - QRType m_qr; - typename internal::plain_col_type::type m_workspace; -}; - -template -class qr_preconditioner_impl -{ -public: - typedef typename MatrixType::Scalar Scalar; - enum - { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - TrOptions = RowsAtCompileTime==1 ? (MatrixType::Options & ~(RowMajor)) - : ColsAtCompileTime==1 ? (MatrixType::Options | RowMajor) - : MatrixType::Options - }; - - typedef Matrix - TransposeTypeWithSameStorageOrder; - - void allocate(const JacobiSVD& svd) - { - if (svd.cols() != m_qr.rows() || svd.rows() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.cols(), svd.rows()); - } - if (svd.m_computeFullV) m_workspace.resize(svd.cols()); - else if (svd.m_computeThinV) m_workspace.resize(svd.rows()); - m_adjoint.resize(svd.cols(), svd.rows()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.cols() > matrix.rows()) - { - m_adjoint = matrix.adjoint(); - m_qr.compute(m_adjoint); - - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.rows(),matrix.rows()).template triangularView().adjoint(); - if(svd.m_computeFullV) m_qr.householderQ().evalTo(svd.m_matrixV, m_workspace); - else if(svd.m_computeThinV) - { - svd.m_matrixV.setIdentity(matrix.cols(), matrix.rows()); - m_qr.householderQ().applyThisOnTheLeft(svd.m_matrixV, m_workspace); - } - if(svd.computeU()) svd.m_matrixU = m_qr.colsPermutation(); - return true; - } - else return false; - } - -private: - typedef ColPivHouseholderQR QRType; - QRType m_qr; - TransposeTypeWithSameStorageOrder m_adjoint; - typename internal::plain_row_type::type m_workspace; -}; - -/*** preconditioner using HouseholderQR ***/ - -template -class qr_preconditioner_impl -{ -public: - void allocate(const JacobiSVD& svd) - { - if (svd.rows() != m_qr.rows() || svd.cols() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.rows(), svd.cols()); - } - if (svd.m_computeFullU) m_workspace.resize(svd.rows()); - else if (svd.m_computeThinU) m_workspace.resize(svd.cols()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.rows() > matrix.cols()) - { - m_qr.compute(matrix); - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.cols(),matrix.cols()).template triangularView(); - if(svd.m_computeFullU) m_qr.householderQ().evalTo(svd.m_matrixU, m_workspace); - else if(svd.m_computeThinU) - { - svd.m_matrixU.setIdentity(matrix.rows(), matrix.cols()); - m_qr.householderQ().applyThisOnTheLeft(svd.m_matrixU, m_workspace); - } - if(svd.computeV()) svd.m_matrixV.setIdentity(matrix.cols(), matrix.cols()); - return true; - } - return false; - } -private: - typedef HouseholderQR QRType; - QRType m_qr; - typename internal::plain_col_type::type m_workspace; -}; - -template -class qr_preconditioner_impl -{ -public: - typedef typename MatrixType::Scalar Scalar; - enum - { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - Options = MatrixType::Options - }; - - typedef Matrix - TransposeTypeWithSameStorageOrder; - - void allocate(const JacobiSVD& svd) - { - if (svd.cols() != m_qr.rows() || svd.rows() != m_qr.cols()) - { - m_qr.~QRType(); - ::new (&m_qr) QRType(svd.cols(), svd.rows()); - } - if (svd.m_computeFullV) m_workspace.resize(svd.cols()); - else if (svd.m_computeThinV) m_workspace.resize(svd.rows()); - m_adjoint.resize(svd.cols(), svd.rows()); - } - - bool run(JacobiSVD& svd, const MatrixType& matrix) - { - if(matrix.cols() > matrix.rows()) - { - m_adjoint = matrix.adjoint(); - m_qr.compute(m_adjoint); - - svd.m_workMatrix = m_qr.matrixQR().block(0,0,matrix.rows(),matrix.rows()).template triangularView().adjoint(); - if(svd.m_computeFullV) m_qr.householderQ().evalTo(svd.m_matrixV, m_workspace); - else if(svd.m_computeThinV) - { - svd.m_matrixV.setIdentity(matrix.cols(), matrix.rows()); - m_qr.householderQ().applyThisOnTheLeft(svd.m_matrixV, m_workspace); - } - if(svd.computeU()) svd.m_matrixU.setIdentity(matrix.rows(), matrix.rows()); - return true; - } - else return false; - } - -private: - typedef HouseholderQR QRType; - QRType m_qr; - TransposeTypeWithSameStorageOrder m_adjoint; - typename internal::plain_row_type::type m_workspace; -}; - -/*** 2x2 SVD implementation - *** - *** JacobiSVD consists in performing a series of 2x2 SVD subproblems - ***/ - -template -struct svd_precondition_2x2_block_to_be_real -{ - typedef JacobiSVD SVD; - typedef typename MatrixType::RealScalar RealScalar; - static bool run(typename SVD::WorkMatrixType&, SVD&, Index, Index, RealScalar&) { return true; } -}; - -template -struct svd_precondition_2x2_block_to_be_real -{ - typedef JacobiSVD SVD; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - static bool run(typename SVD::WorkMatrixType& work_matrix, SVD& svd, Index p, Index q, RealScalar& maxDiagEntry) - { - using std::sqrt; - using std::abs; - Scalar z; - JacobiRotation rot; - RealScalar n = sqrt(numext::abs2(work_matrix.coeff(p,p)) + numext::abs2(work_matrix.coeff(q,p))); - - const RealScalar considerAsZero = (std::numeric_limits::min)(); - const RealScalar precision = NumTraits::epsilon(); - - if(n==0) - { - // make sure first column is zero - work_matrix.coeffRef(p,p) = work_matrix.coeffRef(q,p) = Scalar(0); - - if(abs(numext::imag(work_matrix.coeff(p,q)))>considerAsZero) - { - // work_matrix.coeff(p,q) can be zero if work_matrix.coeff(q,p) is not zero but small enough to underflow when computing n - z = abs(work_matrix.coeff(p,q)) / work_matrix.coeff(p,q); - work_matrix.row(p) *= z; - if(svd.computeU()) svd.m_matrixU.col(p) *= conj(z); - } - if(abs(numext::imag(work_matrix.coeff(q,q)))>considerAsZero) - { - z = abs(work_matrix.coeff(q,q)) / work_matrix.coeff(q,q); - work_matrix.row(q) *= z; - if(svd.computeU()) svd.m_matrixU.col(q) *= conj(z); - } - // otherwise the second row is already zero, so we have nothing to do. - } - else - { - rot.c() = conj(work_matrix.coeff(p,p)) / n; - rot.s() = work_matrix.coeff(q,p) / n; - work_matrix.applyOnTheLeft(p,q,rot); - if(svd.computeU()) svd.m_matrixU.applyOnTheRight(p,q,rot.adjoint()); - if(abs(numext::imag(work_matrix.coeff(p,q)))>considerAsZero) - { - z = abs(work_matrix.coeff(p,q)) / work_matrix.coeff(p,q); - work_matrix.col(q) *= z; - if(svd.computeV()) svd.m_matrixV.col(q) *= z; - } - if(abs(numext::imag(work_matrix.coeff(q,q)))>considerAsZero) - { - z = abs(work_matrix.coeff(q,q)) / work_matrix.coeff(q,q); - work_matrix.row(q) *= z; - if(svd.computeU()) svd.m_matrixU.col(q) *= conj(z); - } - } - - // update largest diagonal entry - maxDiagEntry = numext::maxi(maxDiagEntry,numext::maxi(abs(work_matrix.coeff(p,p)), abs(work_matrix.coeff(q,q)))); - // and check whether the 2x2 block is already diagonal - RealScalar threshold = numext::maxi(considerAsZero, precision * maxDiagEntry); - return abs(work_matrix.coeff(p,q))>threshold || abs(work_matrix.coeff(q,p)) > threshold; - } -}; - -template -struct traits > -{ - typedef _MatrixType MatrixType; -}; - -} // end namespace internal - -/** \ingroup SVD_Module - * - * - * \class JacobiSVD - * - * \brief Two-sided Jacobi SVD decomposition of a rectangular matrix - * - * \tparam _MatrixType the type of the matrix of which we are computing the SVD decomposition - * \tparam QRPreconditioner this optional parameter allows to specify the type of QR decomposition that will be used internally - * for the R-SVD step for non-square matrices. See discussion of possible values below. - * - * SVD decomposition consists in decomposing any n-by-p matrix \a A as a product - * \f[ A = U S V^* \f] - * where \a U is a n-by-n unitary, \a V is a p-by-p unitary, and \a S is a n-by-p real positive matrix which is zero outside of its main diagonal; - * the diagonal entries of S are known as the \em singular \em values of \a A and the columns of \a U and \a V are known as the left - * and right \em singular \em vectors of \a A respectively. - * - * Singular values are always sorted in decreasing order. - * - * This JacobiSVD decomposition computes only the singular values by default. If you want \a U or \a V, you need to ask for them explicitly. - * - * You can ask for only \em thin \a U or \a V to be computed, meaning the following. In case of a rectangular n-by-p matrix, letting \a m be the - * smaller value among \a n and \a p, there are only \a m singular vectors; the remaining columns of \a U and \a V do not correspond to actual - * singular vectors. Asking for \em thin \a U or \a V means asking for only their \a m first columns to be formed. So \a U is then a n-by-m matrix, - * and \a V is then a p-by-m matrix. Notice that thin \a U and \a V are all you need for (least squares) solving. - * - * Here's an example demonstrating basic usage: - * \include JacobiSVD_basic.cpp - * Output: \verbinclude JacobiSVD_basic.out - * - * This JacobiSVD class is a two-sided Jacobi R-SVD decomposition, ensuring optimal reliability and accuracy. The downside is that it's slower than - * bidiagonalizing SVD algorithms for large square matrices; however its complexity is still \f$ O(n^2p) \f$ where \a n is the smaller dimension and - * \a p is the greater dimension, meaning that it is still of the same order of complexity as the faster bidiagonalizing R-SVD algorithms. - * In particular, like any R-SVD, it takes advantage of non-squareness in that its complexity is only linear in the greater dimension. - * - * If the input matrix has inf or nan coefficients, the result of the computation is undefined, but the computation is guaranteed to - * terminate in finite (and reasonable) time. - * - * The possible values for QRPreconditioner are: - * \li ColPivHouseholderQRPreconditioner is the default. In practice it's very safe. It uses column-pivoting QR. - * \li FullPivHouseholderQRPreconditioner, is the safest and slowest. It uses full-pivoting QR. - * Contrary to other QRs, it doesn't allow computing thin unitaries. - * \li HouseholderQRPreconditioner is the fastest, and less safe and accurate than the pivoting variants. It uses non-pivoting QR. - * This is very similar in safety and accuracy to the bidiagonalization process used by bidiagonalizing SVD algorithms (since bidiagonalization - * is inherently non-pivoting). However the resulting SVD is still more reliable than bidiagonalizing SVDs because the Jacobi-based iterarive - * process is more reliable than the optimized bidiagonal SVD iterations. - * \li NoQRPreconditioner allows not to use a QR preconditioner at all. This is useful if you know that you will only be computing - * JacobiSVD decompositions of square matrices. Non-square matrices require a QR preconditioner. Using this option will result in - * faster compilation and smaller executable code. It won't significantly speed up computation, since JacobiSVD is always checking - * if QR preconditioning is needed before applying it anyway. - * - * \sa MatrixBase::jacobiSvd() - */ -template class JacobiSVD - : public SVDBase > -{ - typedef SVDBase Base; - public: - - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - DiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime,ColsAtCompileTime), - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - MaxDiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(MaxRowsAtCompileTime,MaxColsAtCompileTime), - MatrixOptions = MatrixType::Options - }; - - typedef typename Base::MatrixUType MatrixUType; - typedef typename Base::MatrixVType MatrixVType; - typedef typename Base::SingularValuesType SingularValuesType; - - typedef typename internal::plain_row_type::type RowType; - typedef typename internal::plain_col_type::type ColType; - typedef Matrix - WorkMatrixType; - - /** \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via JacobiSVD::compute(const MatrixType&). - */ - JacobiSVD() - {} - - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem size. - * \sa JacobiSVD() - */ - JacobiSVD(Index rows, Index cols, unsigned int computationOptions = 0) - { - allocate(rows, cols, computationOptions); - } - - /** \brief Constructor performing the decomposition of given matrix. - * - * \param matrix the matrix to decompose - * \param computationOptions optional parameter allowing to specify if you want full or thin U or V unitaries to be computed. - * By default, none is computed. This is a bit-field, the possible bits are #ComputeFullU, #ComputeThinU, - * #ComputeFullV, #ComputeThinV. - * - * Thin unitaries are only available if your matrix type has a Dynamic number of columns (for example MatrixXf). They also are not - * available with the (non-default) FullPivHouseholderQR preconditioner. - */ - explicit JacobiSVD(const MatrixType& matrix, unsigned int computationOptions = 0) - { - compute(matrix, computationOptions); - } - - /** \brief Method performing the decomposition of given matrix using custom options. - * - * \param matrix the matrix to decompose - * \param computationOptions optional parameter allowing to specify if you want full or thin U or V unitaries to be computed. - * By default, none is computed. This is a bit-field, the possible bits are #ComputeFullU, #ComputeThinU, - * #ComputeFullV, #ComputeThinV. - * - * Thin unitaries are only available if your matrix type has a Dynamic number of columns (for example MatrixXf). They also are not - * available with the (non-default) FullPivHouseholderQR preconditioner. - */ - JacobiSVD& compute(const MatrixType& matrix, unsigned int computationOptions); - - /** \brief Method performing the decomposition of given matrix using current options. - * - * \param matrix the matrix to decompose - * - * This method uses the current \a computationOptions, as already passed to the constructor or to compute(const MatrixType&, unsigned int). - */ - JacobiSVD& compute(const MatrixType& matrix) - { - return compute(matrix, m_computationOptions); - } - - using Base::computeU; - using Base::computeV; - using Base::rows; - using Base::cols; - using Base::rank; - - private: - void allocate(Index rows, Index cols, unsigned int computationOptions); - - protected: - using Base::m_matrixU; - using Base::m_matrixV; - using Base::m_singularValues; - using Base::m_isInitialized; - using Base::m_isAllocated; - using Base::m_usePrescribedThreshold; - using Base::m_computeFullU; - using Base::m_computeThinU; - using Base::m_computeFullV; - using Base::m_computeThinV; - using Base::m_computationOptions; - using Base::m_nonzeroSingularValues; - using Base::m_rows; - using Base::m_cols; - using Base::m_diagSize; - using Base::m_prescribedThreshold; - WorkMatrixType m_workMatrix; - - template - friend struct internal::svd_precondition_2x2_block_to_be_real; - template - friend struct internal::qr_preconditioner_impl; - - internal::qr_preconditioner_impl m_qr_precond_morecols; - internal::qr_preconditioner_impl m_qr_precond_morerows; - MatrixType m_scaledMatrix; -}; - -template -void JacobiSVD::allocate(Index rows, Index cols, unsigned int computationOptions) -{ - eigen_assert(rows >= 0 && cols >= 0); - - if (m_isAllocated && - rows == m_rows && - cols == m_cols && - computationOptions == m_computationOptions) - { - return; - } - - m_rows = rows; - m_cols = cols; - m_isInitialized = false; - m_isAllocated = true; - m_computationOptions = computationOptions; - m_computeFullU = (computationOptions & ComputeFullU) != 0; - m_computeThinU = (computationOptions & ComputeThinU) != 0; - m_computeFullV = (computationOptions & ComputeFullV) != 0; - m_computeThinV = (computationOptions & ComputeThinV) != 0; - eigen_assert(!(m_computeFullU && m_computeThinU) && "JacobiSVD: you can't ask for both full and thin U"); - eigen_assert(!(m_computeFullV && m_computeThinV) && "JacobiSVD: you can't ask for both full and thin V"); - eigen_assert(EIGEN_IMPLIES(m_computeThinU || m_computeThinV, MatrixType::ColsAtCompileTime==Dynamic) && - "JacobiSVD: thin U and V are only available when your matrix has a dynamic number of columns."); - if (QRPreconditioner == FullPivHouseholderQRPreconditioner) - { - eigen_assert(!(m_computeThinU || m_computeThinV) && - "JacobiSVD: can't compute thin U or thin V with the FullPivHouseholderQR preconditioner. " - "Use the ColPivHouseholderQR preconditioner instead."); - } - m_diagSize = (std::min)(m_rows, m_cols); - m_singularValues.resize(m_diagSize); - if(RowsAtCompileTime==Dynamic) - m_matrixU.resize(m_rows, m_computeFullU ? m_rows - : m_computeThinU ? m_diagSize - : 0); - if(ColsAtCompileTime==Dynamic) - m_matrixV.resize(m_cols, m_computeFullV ? m_cols - : m_computeThinV ? m_diagSize - : 0); - m_workMatrix.resize(m_diagSize, m_diagSize); - - if(m_cols>m_rows) m_qr_precond_morecols.allocate(*this); - if(m_rows>m_cols) m_qr_precond_morerows.allocate(*this); - if(m_rows!=m_cols) m_scaledMatrix.resize(rows,cols); -} - -template -JacobiSVD& -JacobiSVD::compute(const MatrixType& matrix, unsigned int computationOptions) -{ - using std::abs; - allocate(matrix.rows(), matrix.cols(), computationOptions); - - // currently we stop when we reach precision 2*epsilon as the last bit of precision can require an unreasonable number of iterations, - // only worsening the precision of U and V as we accumulate more rotations - const RealScalar precision = RealScalar(2) * NumTraits::epsilon(); - - // limit for denormal numbers to be considered zero in order to avoid infinite loops (see bug 286) - const RealScalar considerAsZero = (std::numeric_limits::min)(); - - // Scaling factor to reduce over/under-flows - RealScalar scale = matrix.cwiseAbs().maxCoeff(); - if(scale==RealScalar(0)) scale = RealScalar(1); - - /*** step 1. The R-SVD step: we use a QR decomposition to reduce to the case of a square matrix */ - - if(m_rows!=m_cols) - { - m_scaledMatrix = matrix / scale; - m_qr_precond_morecols.run(*this, m_scaledMatrix); - m_qr_precond_morerows.run(*this, m_scaledMatrix); - } - else - { - m_workMatrix = matrix.block(0,0,m_diagSize,m_diagSize) / scale; - if(m_computeFullU) m_matrixU.setIdentity(m_rows,m_rows); - if(m_computeThinU) m_matrixU.setIdentity(m_rows,m_diagSize); - if(m_computeFullV) m_matrixV.setIdentity(m_cols,m_cols); - if(m_computeThinV) m_matrixV.setIdentity(m_cols, m_diagSize); - } - - /*** step 2. The main Jacobi SVD iteration. ***/ - RealScalar maxDiagEntry = m_workMatrix.cwiseAbs().diagonal().maxCoeff(); - - bool finished = false; - while(!finished) - { - finished = true; - - // do a sweep: for all index pairs (p,q), perform SVD of the corresponding 2x2 sub-matrix - - for(Index p = 1; p < m_diagSize; ++p) - { - for(Index q = 0; q < p; ++q) - { - // if this 2x2 sub-matrix is not diagonal already... - // notice that this comparison will evaluate to false if any NaN is involved, ensuring that NaN's don't - // keep us iterating forever. Similarly, small denormal numbers are considered zero. - RealScalar threshold = numext::maxi(considerAsZero, precision * maxDiagEntry); - if(abs(m_workMatrix.coeff(p,q))>threshold || abs(m_workMatrix.coeff(q,p)) > threshold) - { - finished = false; - // perform SVD decomposition of 2x2 sub-matrix corresponding to indices p,q to make it diagonal - // the complex to real operation returns true if the updated 2x2 block is not already diagonal - if(internal::svd_precondition_2x2_block_to_be_real::run(m_workMatrix, *this, p, q, maxDiagEntry)) - { - JacobiRotation j_left, j_right; - internal::real_2x2_jacobi_svd(m_workMatrix, p, q, &j_left, &j_right); - - // accumulate resulting Jacobi rotations - m_workMatrix.applyOnTheLeft(p,q,j_left); - if(computeU()) m_matrixU.applyOnTheRight(p,q,j_left.transpose()); - - m_workMatrix.applyOnTheRight(p,q,j_right); - if(computeV()) m_matrixV.applyOnTheRight(p,q,j_right); - - // keep track of the largest diagonal coefficient - maxDiagEntry = numext::maxi(maxDiagEntry,numext::maxi(abs(m_workMatrix.coeff(p,p)), abs(m_workMatrix.coeff(q,q)))); - } - } - } - } - } - - /*** step 3. The work matrix is now diagonal, so ensure it's positive so its diagonal entries are the singular values ***/ - - for(Index i = 0; i < m_diagSize; ++i) - { - // For a complex matrix, some diagonal coefficients might note have been - // treated by svd_precondition_2x2_block_to_be_real, and the imaginary part - // of some diagonal entry might not be null. - if(NumTraits::IsComplex && abs(numext::imag(m_workMatrix.coeff(i,i)))>considerAsZero) - { - RealScalar a = abs(m_workMatrix.coeff(i,i)); - m_singularValues.coeffRef(i) = abs(a); - if(computeU()) m_matrixU.col(i) *= m_workMatrix.coeff(i,i)/a; - } - else - { - // m_workMatrix.coeff(i,i) is already real, no difficulty: - RealScalar a = numext::real(m_workMatrix.coeff(i,i)); - m_singularValues.coeffRef(i) = abs(a); - if(computeU() && (a -JacobiSVD::PlainObject> -MatrixBase::jacobiSvd(unsigned int computationOptions) const -{ - return JacobiSVD(*this, computationOptions); -} - -} // end namespace Eigen - -#endif // EIGEN_JACOBISVD_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD_LAPACKE.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD_LAPACKE.h deleted file mode 100644 index ff0516f6116..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/JacobiSVD_LAPACKE.h +++ /dev/null @@ -1,91 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * Singular Value Decomposition - SVD. - ******************************************************************************** -*/ - -#ifndef EIGEN_JACOBISVD_LAPACKE_H -#define EIGEN_JACOBISVD_LAPACKE_H - -namespace Eigen { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_SVD(EIGTYPE, LAPACKE_TYPE, LAPACKE_RTYPE, LAPACKE_PREFIX, EIGCOLROW, LAPACKE_COLROW) \ -template<> inline \ -JacobiSVD, ColPivHouseholderQRPreconditioner>& \ -JacobiSVD, ColPivHouseholderQRPreconditioner>::compute(const Matrix& matrix, unsigned int computationOptions) \ -{ \ - typedef Matrix MatrixType; \ - /*typedef MatrixType::Scalar Scalar;*/ \ - /*typedef MatrixType::RealScalar RealScalar;*/ \ - allocate(matrix.rows(), matrix.cols(), computationOptions); \ -\ - /*const RealScalar precision = RealScalar(2) * NumTraits::epsilon();*/ \ - m_nonzeroSingularValues = m_diagSize; \ -\ - lapack_int lda = internal::convert_index(matrix.outerStride()), ldu, ldvt; \ - lapack_int matrix_order = LAPACKE_COLROW; \ - char jobu, jobvt; \ - LAPACKE_TYPE *u, *vt, dummy; \ - jobu = (m_computeFullU) ? 'A' : (m_computeThinU) ? 'S' : 'N'; \ - jobvt = (m_computeFullV) ? 'A' : (m_computeThinV) ? 'S' : 'N'; \ - if (computeU()) { \ - ldu = internal::convert_index(m_matrixU.outerStride()); \ - u = (LAPACKE_TYPE*)m_matrixU.data(); \ - } else { ldu=1; u=&dummy; }\ - MatrixType localV; \ - lapack_int vt_rows = (m_computeFullV) ? internal::convert_index(m_cols) : (m_computeThinV) ? internal::convert_index(m_diagSize) : 1; \ - if (computeV()) { \ - localV.resize(vt_rows, m_cols); \ - ldvt = internal::convert_index(localV.outerStride()); \ - vt = (LAPACKE_TYPE*)localV.data(); \ - } else { ldvt=1; vt=&dummy; }\ - Matrix superb; superb.resize(m_diagSize, 1); \ - MatrixType m_temp; m_temp = matrix; \ - LAPACKE_##LAPACKE_PREFIX##gesvd( matrix_order, jobu, jobvt, internal::convert_index(m_rows), internal::convert_index(m_cols), (LAPACKE_TYPE*)m_temp.data(), lda, (LAPACKE_RTYPE*)m_singularValues.data(), u, ldu, vt, ldvt, superb.data()); \ - if (computeV()) m_matrixV = localV.adjoint(); \ - /* for(int i=0;i -// Copyright (C) 2014 Gael Guennebaud -// -// Copyright (C) 2013 Gauthier Brun -// Copyright (C) 2013 Nicolas Carre -// Copyright (C) 2013 Jean Ceccato -// Copyright (C) 2013 Pierre Zoppitelli -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SVDBASE_H -#define EIGEN_SVDBASE_H - -namespace Eigen { -/** \ingroup SVD_Module - * - * - * \class SVDBase - * - * \brief Base class of SVD algorithms - * - * \tparam Derived the type of the actual SVD decomposition - * - * SVD decomposition consists in decomposing any n-by-p matrix \a A as a product - * \f[ A = U S V^* \f] - * where \a U is a n-by-n unitary, \a V is a p-by-p unitary, and \a S is a n-by-p real positive matrix which is zero outside of its main diagonal; - * the diagonal entries of S are known as the \em singular \em values of \a A and the columns of \a U and \a V are known as the left - * and right \em singular \em vectors of \a A respectively. - * - * Singular values are always sorted in decreasing order. - * - * - * You can ask for only \em thin \a U or \a V to be computed, meaning the following. In case of a rectangular n-by-p matrix, letting \a m be the - * smaller value among \a n and \a p, there are only \a m singular vectors; the remaining columns of \a U and \a V do not correspond to actual - * singular vectors. Asking for \em thin \a U or \a V means asking for only their \a m first columns to be formed. So \a U is then a n-by-m matrix, - * and \a V is then a p-by-m matrix. Notice that thin \a U and \a V are all you need for (least squares) solving. - * - * If the input matrix has inf or nan coefficients, the result of the computation is undefined, but the computation is guaranteed to - * terminate in finite (and reasonable) time. - * \sa class BDCSVD, class JacobiSVD - */ -template -class SVDBase -{ - -public: - typedef typename internal::traits::MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - DiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime,ColsAtCompileTime), - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, - MaxDiagSizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(MaxRowsAtCompileTime,MaxColsAtCompileTime), - MatrixOptions = MatrixType::Options - }; - - typedef Matrix MatrixUType; - typedef Matrix MatrixVType; - typedef typename internal::plain_diag_type::type SingularValuesType; - - Derived& derived() { return *static_cast(this); } - const Derived& derived() const { return *static_cast(this); } - - /** \returns the \a U matrix. - * - * For the SVD decomposition of a n-by-p matrix, letting \a m be the minimum of \a n and \a p, - * the U matrix is n-by-n if you asked for \link Eigen::ComputeFullU ComputeFullU \endlink, and is n-by-m if you asked for \link Eigen::ComputeThinU ComputeThinU \endlink. - * - * The \a m first columns of \a U are the left singular vectors of the matrix being decomposed. - * - * This method asserts that you asked for \a U to be computed. - */ - const MatrixUType& matrixU() const - { - eigen_assert(m_isInitialized && "SVD is not initialized."); - eigen_assert(computeU() && "This SVD decomposition didn't compute U. Did you ask for it?"); - return m_matrixU; - } - - /** \returns the \a V matrix. - * - * For the SVD decomposition of a n-by-p matrix, letting \a m be the minimum of \a n and \a p, - * the V matrix is p-by-p if you asked for \link Eigen::ComputeFullV ComputeFullV \endlink, and is p-by-m if you asked for \link Eigen::ComputeThinV ComputeThinV \endlink. - * - * The \a m first columns of \a V are the right singular vectors of the matrix being decomposed. - * - * This method asserts that you asked for \a V to be computed. - */ - const MatrixVType& matrixV() const - { - eigen_assert(m_isInitialized && "SVD is not initialized."); - eigen_assert(computeV() && "This SVD decomposition didn't compute V. Did you ask for it?"); - return m_matrixV; - } - - /** \returns the vector of singular values. - * - * For the SVD decomposition of a n-by-p matrix, letting \a m be the minimum of \a n and \a p, the - * returned vector has size \a m. Singular values are always sorted in decreasing order. - */ - const SingularValuesType& singularValues() const - { - eigen_assert(m_isInitialized && "SVD is not initialized."); - return m_singularValues; - } - - /** \returns the number of singular values that are not exactly 0 */ - Index nonzeroSingularValues() const - { - eigen_assert(m_isInitialized && "SVD is not initialized."); - return m_nonzeroSingularValues; - } - - /** \returns the rank of the matrix of which \c *this is the SVD. - * - * \note This method has to determine which singular values should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index rank() const - { - using std::abs; - eigen_assert(m_isInitialized && "JacobiSVD is not initialized."); - if(m_singularValues.size()==0) return 0; - RealScalar premultiplied_threshold = numext::maxi(m_singularValues.coeff(0) * threshold(), (std::numeric_limits::min)()); - Index i = m_nonzeroSingularValues-1; - while(i>=0 && m_singularValues.coeff(i) < premultiplied_threshold) --i; - return i+1; - } - - /** Allows to prescribe a threshold to be used by certain methods, such as rank() and solve(), - * which need to determine when singular values are to be considered nonzero. - * This is not used for the SVD decomposition itself. - * - * When it needs to get the threshold value, Eigen calls threshold(). - * The default is \c NumTraits::epsilon() - * - * \param threshold The new value to use as the threshold. - * - * A singular value will be considered nonzero if its value is strictly greater than - * \f$ \vert singular value \vert \leqslant threshold \times \vert max singular value \vert \f$. - * - * If you want to come back to the default behavior, call setThreshold(Default_t) - */ - Derived& setThreshold(const RealScalar& threshold) - { - m_usePrescribedThreshold = true; - m_prescribedThreshold = threshold; - return derived(); - } - - /** Allows to come back to the default behavior, letting Eigen use its default formula for - * determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code svd.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - Derived& setThreshold(Default_t) - { - m_usePrescribedThreshold = false; - return derived(); - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const - { - eigen_assert(m_isInitialized || m_usePrescribedThreshold); - // this temporary is needed to workaround a MSVC issue - Index diagSize = (std::max)(1,m_diagSize); - return m_usePrescribedThreshold ? m_prescribedThreshold - : RealScalar(diagSize)*NumTraits::epsilon(); - } - - /** \returns true if \a U (full or thin) is asked for in this SVD decomposition */ - inline bool computeU() const { return m_computeFullU || m_computeThinU; } - /** \returns true if \a V (full or thin) is asked for in this SVD decomposition */ - inline bool computeV() const { return m_computeFullV || m_computeThinV; } - - inline Index rows() const { return m_rows; } - inline Index cols() const { return m_cols; } - - /** \returns a (least squares) solution of \f$ A x = b \f$ using the current SVD decomposition of A. - * - * \param b the right-hand-side of the equation to solve. - * - * \note Solving requires both U and V to be computed. Thin U and V are enough, there is no need for full U or V. - * - * \note SVD solving is implicitly least-squares. Thus, this method serves both purposes of exact solving and least-squares solving. - * In other words, the returned solution is guaranteed to minimize the Euclidean norm \f$ \Vert A x - b \Vert \f$. - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "SVD is not initialized."); - eigen_assert(computeU() && computeV() && "SVD::solve() requires both unitaries U and V to be computed (thin unitaries suffice)."); - return Solve(derived(), b.derived()); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - #endif - -protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - // return true if already allocated - bool allocate(Index rows, Index cols, unsigned int computationOptions) ; - - MatrixUType m_matrixU; - MatrixVType m_matrixV; - SingularValuesType m_singularValues; - bool m_isInitialized, m_isAllocated, m_usePrescribedThreshold; - bool m_computeFullU, m_computeThinU; - bool m_computeFullV, m_computeThinV; - unsigned int m_computationOptions; - Index m_nonzeroSingularValues, m_rows, m_cols, m_diagSize; - RealScalar m_prescribedThreshold; - - /** \brief Default Constructor. - * - * Default constructor of SVDBase - */ - SVDBase() - : m_isInitialized(false), - m_isAllocated(false), - m_usePrescribedThreshold(false), - m_computationOptions(0), - m_rows(-1), m_cols(-1), m_diagSize(0) - { - check_template_parameters(); - } - - -}; - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void SVDBase::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - eigen_assert(rhs.rows() == rows()); - - // A = U S V^* - // So A^{-1} = V S^{-1} U^* - - Matrix tmp; - Index l_rank = rank(); - tmp.noalias() = m_matrixU.leftCols(l_rank).adjoint() * rhs; - tmp = m_singularValues.head(l_rank).asDiagonal().inverse() * tmp; - dst = m_matrixV.leftCols(l_rank) * tmp; -} -#endif - -template -bool SVDBase::allocate(Index rows, Index cols, unsigned int computationOptions) -{ - eigen_assert(rows >= 0 && cols >= 0); - - if (m_isAllocated && - rows == m_rows && - cols == m_cols && - computationOptions == m_computationOptions) - { - return true; - } - - m_rows = rows; - m_cols = cols; - m_isInitialized = false; - m_isAllocated = true; - m_computationOptions = computationOptions; - m_computeFullU = (computationOptions & ComputeFullU) != 0; - m_computeThinU = (computationOptions & ComputeThinU) != 0; - m_computeFullV = (computationOptions & ComputeFullV) != 0; - m_computeThinV = (computationOptions & ComputeThinV) != 0; - eigen_assert(!(m_computeFullU && m_computeThinU) && "SVDBase: you can't ask for both full and thin U"); - eigen_assert(!(m_computeFullV && m_computeThinV) && "SVDBase: you can't ask for both full and thin V"); - eigen_assert(EIGEN_IMPLIES(m_computeThinU || m_computeThinV, MatrixType::ColsAtCompileTime==Dynamic) && - "SVDBase: thin U and V are only available when your matrix has a dynamic number of columns."); - - m_diagSize = (std::min)(m_rows, m_cols); - m_singularValues.resize(m_diagSize); - if(RowsAtCompileTime==Dynamic) - m_matrixU.resize(m_rows, m_computeFullU ? m_rows : m_computeThinU ? m_diagSize : 0); - if(ColsAtCompileTime==Dynamic) - m_matrixV.resize(m_cols, m_computeFullV ? m_cols : m_computeThinV ? m_diagSize : 0); - - return false; -} - -}// end namespace - -#endif // EIGEN_SVDBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/UpperBidiagonalization.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/UpperBidiagonalization.h deleted file mode 100644 index 11ac847e1db..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SVD/UpperBidiagonalization.h +++ /dev/null @@ -1,414 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Benoit Jacob -// Copyright (C) 2013-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BIDIAGONALIZATION_H -#define EIGEN_BIDIAGONALIZATION_H - -namespace Eigen { - -namespace internal { -// UpperBidiagonalization will probably be replaced by a Bidiagonalization class, don't want to make it stable API. -// At the same time, it's useful to keep for now as it's about the only thing that is testing the BandMatrix class. - -template class UpperBidiagonalization -{ - public: - - typedef _MatrixType MatrixType; - enum { - RowsAtCompileTime = MatrixType::RowsAtCompileTime, - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - ColsAtCompileTimeMinusOne = internal::decrement_size::ret - }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef Matrix RowVectorType; - typedef Matrix ColVectorType; - typedef BandMatrix BidiagonalType; - typedef Matrix DiagVectorType; - typedef Matrix SuperDiagVectorType; - typedef HouseholderSequence< - const MatrixType, - const typename internal::remove_all::ConjugateReturnType>::type - > HouseholderUSequenceType; - typedef HouseholderSequence< - const typename internal::remove_all::type, - Diagonal, - OnTheRight - > HouseholderVSequenceType; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via Bidiagonalization::compute(const MatrixType&). - */ - UpperBidiagonalization() : m_householder(), m_bidiagonal(), m_isInitialized(false) {} - - explicit UpperBidiagonalization(const MatrixType& matrix) - : m_householder(matrix.rows(), matrix.cols()), - m_bidiagonal(matrix.cols(), matrix.cols()), - m_isInitialized(false) - { - compute(matrix); - } - - UpperBidiagonalization& compute(const MatrixType& matrix); - UpperBidiagonalization& computeUnblocked(const MatrixType& matrix); - - const MatrixType& householder() const { return m_householder; } - const BidiagonalType& bidiagonal() const { return m_bidiagonal; } - - const HouseholderUSequenceType householderU() const - { - eigen_assert(m_isInitialized && "UpperBidiagonalization is not initialized."); - return HouseholderUSequenceType(m_householder, m_householder.diagonal().conjugate()); - } - - const HouseholderVSequenceType householderV() // const here gives nasty errors and i'm lazy - { - eigen_assert(m_isInitialized && "UpperBidiagonalization is not initialized."); - return HouseholderVSequenceType(m_householder.conjugate(), m_householder.const_derived().template diagonal<1>()) - .setLength(m_householder.cols()-1) - .setShift(1); - } - - protected: - MatrixType m_householder; - BidiagonalType m_bidiagonal; - bool m_isInitialized; -}; - -// Standard upper bidiagonalization without fancy optimizations -// This version should be faster for small matrix size -template -void upperbidiagonalization_inplace_unblocked(MatrixType& mat, - typename MatrixType::RealScalar *diagonal, - typename MatrixType::RealScalar *upper_diagonal, - typename MatrixType::Scalar* tempData = 0) -{ - typedef typename MatrixType::Scalar Scalar; - - Index rows = mat.rows(); - Index cols = mat.cols(); - - typedef Matrix TempType; - TempType tempVector; - if(tempData==0) - { - tempVector.resize(rows); - tempData = tempVector.data(); - } - - for (Index k = 0; /* breaks at k==cols-1 below */ ; ++k) - { - Index remainingRows = rows - k; - Index remainingCols = cols - k - 1; - - // construct left householder transform in-place in A - mat.col(k).tail(remainingRows) - .makeHouseholderInPlace(mat.coeffRef(k,k), diagonal[k]); - // apply householder transform to remaining part of A on the left - mat.bottomRightCorner(remainingRows, remainingCols) - .applyHouseholderOnTheLeft(mat.col(k).tail(remainingRows-1), mat.coeff(k,k), tempData); - - if(k == cols-1) break; - - // construct right householder transform in-place in mat - mat.row(k).tail(remainingCols) - .makeHouseholderInPlace(mat.coeffRef(k,k+1), upper_diagonal[k]); - // apply householder transform to remaining part of mat on the left - mat.bottomRightCorner(remainingRows-1, remainingCols) - .applyHouseholderOnTheRight(mat.row(k).tail(remainingCols-1).transpose(), mat.coeff(k,k+1), tempData); - } -} - -/** \internal - * Helper routine for the block reduction to upper bidiagonal form. - * - * Let's partition the matrix A: - * - * | A00 A01 | - * A = | | - * | A10 A11 | - * - * This function reduces to bidiagonal form the left \c rows x \a blockSize vertical panel [A00/A10] - * and the \a blockSize x \c cols horizontal panel [A00 A01] of the matrix \a A. The bottom-right block A11 - * is updated using matrix-matrix products: - * A22 -= V * Y^T - X * U^T - * where V and U contains the left and right Householder vectors. U and V are stored in A10, and A01 - * respectively, and the update matrices X and Y are computed during the reduction. - * - */ -template -void upperbidiagonalization_blocked_helper(MatrixType& A, - typename MatrixType::RealScalar *diagonal, - typename MatrixType::RealScalar *upper_diagonal, - Index bs, - Ref::Flags & RowMajorBit> > X, - Ref::Flags & RowMajorBit> > Y) -{ - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename NumTraits::Literal Literal; - enum { StorageOrder = traits::Flags & RowMajorBit }; - typedef InnerStride ColInnerStride; - typedef InnerStride RowInnerStride; - typedef Ref, 0, ColInnerStride> SubColumnType; - typedef Ref, 0, RowInnerStride> SubRowType; - typedef Ref > SubMatType; - - Index brows = A.rows(); - Index bcols = A.cols(); - - Scalar tau_u, tau_u_prev(0), tau_v; - - for(Index k = 0; k < bs; ++k) - { - Index remainingRows = brows - k; - Index remainingCols = bcols - k - 1; - - SubMatType X_k1( X.block(k,0, remainingRows,k) ); - SubMatType V_k1( A.block(k,0, remainingRows,k) ); - - // 1 - update the k-th column of A - SubColumnType v_k = A.col(k).tail(remainingRows); - v_k -= V_k1 * Y.row(k).head(k).adjoint(); - if(k) v_k -= X_k1 * A.col(k).head(k); - - // 2 - construct left Householder transform in-place - v_k.makeHouseholderInPlace(tau_v, diagonal[k]); - - if(k+10) A.coeffRef(k-1,k) = tau_u_prev; - tau_u_prev = tau_u; - } - else - A.coeffRef(k-1,k) = tau_u_prev; - - A.coeffRef(k,k) = tau_v; - } - - if(bsbs && brows>bs) - { - SubMatType A11( A.bottomRightCorner(brows-bs,bcols-bs) ); - SubMatType A10( A.block(bs,0, brows-bs,bs) ); - SubMatType A01( A.block(0,bs, bs,bcols-bs) ); - Scalar tmp = A01(bs-1,0); - A01(bs-1,0) = Literal(1); - A11.noalias() -= A10 * Y.topLeftCorner(bcols,bs).bottomRows(bcols-bs).adjoint(); - A11.noalias() -= X.topLeftCorner(brows,bs).bottomRows(brows-bs) * A01; - A01(bs-1,0) = tmp; - } -} - -/** \internal - * - * Implementation of a block-bidiagonal reduction. - * It is based on the following paper: - * The Design of a Parallel Dense Linear Algebra Software Library: Reduction to Hessenberg, Tridiagonal, and Bidiagonal Form. - * by Jaeyoung Choi, Jack J. Dongarra, David W. Walker. (1995) - * section 3.3 - */ -template -void upperbidiagonalization_inplace_blocked(MatrixType& A, BidiagType& bidiagonal, - Index maxBlockSize=32, - typename MatrixType::Scalar* /*tempData*/ = 0) -{ - typedef typename MatrixType::Scalar Scalar; - typedef Block BlockType; - - Index rows = A.rows(); - Index cols = A.cols(); - Index size = (std::min)(rows, cols); - - // X and Y are work space - enum { StorageOrder = traits::Flags & RowMajorBit }; - Matrix X(rows,maxBlockSize); - Matrix Y(cols,maxBlockSize); - Index blockSize = (std::min)(maxBlockSize,size); - - Index k = 0; - for(k = 0; k < size; k += blockSize) - { - Index bs = (std::min)(size-k,blockSize); // actual size of the block - Index brows = rows - k; // rows of the block - Index bcols = cols - k; // columns of the block - - // partition the matrix A: - // - // | A00 A01 A02 | - // | | - // A = | A10 A11 A12 | - // | | - // | A20 A21 A22 | - // - // where A11 is a bs x bs diagonal block, - // and let: - // | A11 A12 | - // B = | | - // | A21 A22 | - - BlockType B = A.block(k,k,brows,bcols); - - // This stage performs the bidiagonalization of A11, A21, A12, and updating of A22. - // Finally, the algorithm continue on the updated A22. - // - // However, if B is too small, or A22 empty, then let's use an unblocked strategy - if(k+bs==cols || bcols<48) // somewhat arbitrary threshold - { - upperbidiagonalization_inplace_unblocked(B, - &(bidiagonal.template diagonal<0>().coeffRef(k)), - &(bidiagonal.template diagonal<1>().coeffRef(k)), - X.data() - ); - break; // We're done - } - else - { - upperbidiagonalization_blocked_helper( B, - &(bidiagonal.template diagonal<0>().coeffRef(k)), - &(bidiagonal.template diagonal<1>().coeffRef(k)), - bs, - X.topLeftCorner(brows,bs), - Y.topLeftCorner(bcols,bs) - ); - } - } -} - -template -UpperBidiagonalization<_MatrixType>& UpperBidiagonalization<_MatrixType>::computeUnblocked(const _MatrixType& matrix) -{ - Index rows = matrix.rows(); - Index cols = matrix.cols(); - EIGEN_ONLY_USED_FOR_DEBUG(cols); - - eigen_assert(rows >= cols && "UpperBidiagonalization is only for Arices satisfying rows>=cols."); - - m_householder = matrix; - - ColVectorType temp(rows); - - upperbidiagonalization_inplace_unblocked(m_householder, - &(m_bidiagonal.template diagonal<0>().coeffRef(0)), - &(m_bidiagonal.template diagonal<1>().coeffRef(0)), - temp.data()); - - m_isInitialized = true; - return *this; -} - -template -UpperBidiagonalization<_MatrixType>& UpperBidiagonalization<_MatrixType>::compute(const _MatrixType& matrix) -{ - Index rows = matrix.rows(); - Index cols = matrix.cols(); - EIGEN_ONLY_USED_FOR_DEBUG(rows); - EIGEN_ONLY_USED_FOR_DEBUG(cols); - - eigen_assert(rows >= cols && "UpperBidiagonalization is only for Arices satisfying rows>=cols."); - - m_householder = matrix; - upperbidiagonalization_inplace_blocked(m_householder, m_bidiagonal); - - m_isInitialized = true; - return *this; -} - -#if 0 -/** \return the Householder QR decomposition of \c *this. - * - * \sa class Bidiagonalization - */ -template -const UpperBidiagonalization::PlainObject> -MatrixBase::bidiagonalization() const -{ - return UpperBidiagonalization(eval()); -} -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BIDIAGONALIZATION_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky.h deleted file mode 100644 index 369e6804afa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky.h +++ /dev/null @@ -1,689 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SIMPLICIAL_CHOLESKY_H -#define EIGEN_SIMPLICIAL_CHOLESKY_H - -namespace Eigen { - -enum SimplicialCholeskyMode { - SimplicialCholeskyLLT, - SimplicialCholeskyLDLT -}; - -namespace internal { - template - struct simplicial_cholesky_grab_input { - typedef CholMatrixType const * ConstCholMatrixPtr; - static void run(const InputMatrixType& input, ConstCholMatrixPtr &pmat, CholMatrixType &tmp) - { - tmp = input; - pmat = &tmp; - } - }; - - template - struct simplicial_cholesky_grab_input { - typedef MatrixType const * ConstMatrixPtr; - static void run(const MatrixType& input, ConstMatrixPtr &pmat, MatrixType &/*tmp*/) - { - pmat = &input; - } - }; -} // end namespace internal - -/** \ingroup SparseCholesky_Module - * \brief A base class for direct sparse Cholesky factorizations - * - * This is a base class for LL^T and LDL^T Cholesky factorizations of sparse matrices that are - * selfadjoint and positive definite. These factorizations allow for solving A.X = B where - * X and B can be either dense or sparse. - * - * In order to reduce the fill-in, a symmetric permutation P is applied prior to the factorization - * such that the factorized matrix is P A P^-1. - * - * \tparam Derived the type of the derived class, that is the actual factorization type. - * - */ -template -class SimplicialCholeskyBase : public SparseSolverBase -{ - typedef SparseSolverBase Base; - using Base::m_isInitialized; - - public: - typedef typename internal::traits::MatrixType MatrixType; - typedef typename internal::traits::OrderingType OrderingType; - enum { UpLo = internal::traits::UpLo }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef CholMatrixType const * ConstCholMatrixPtr; - typedef Matrix VectorType; - typedef Matrix VectorI; - - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - - using Base::derived; - - /** Default constructor */ - SimplicialCholeskyBase() - : m_info(Success), m_shiftOffset(0), m_shiftScale(1) - {} - - explicit SimplicialCholeskyBase(const MatrixType& matrix) - : m_info(Success), m_shiftOffset(0), m_shiftScale(1) - { - derived().compute(matrix); - } - - ~SimplicialCholeskyBase() - { - } - - Derived& derived() { return *static_cast(this); } - const Derived& derived() const { return *static_cast(this); } - - inline Index cols() const { return m_matrix.cols(); } - inline Index rows() const { return m_matrix.rows(); } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the matrix.appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - /** \returns the permutation P - * \sa permutationPinv() */ - const PermutationMatrix& permutationP() const - { return m_P; } - - /** \returns the inverse P^-1 of the permutation P - * \sa permutationP() */ - const PermutationMatrix& permutationPinv() const - { return m_Pinv; } - - /** Sets the shift parameters that will be used to adjust the diagonal coefficients during the numerical factorization. - * - * During the numerical factorization, the diagonal coefficients are transformed by the following linear model:\n - * \c d_ii = \a offset + \a scale * \c d_ii - * - * The default is the identity transformation with \a offset=0, and \a scale=1. - * - * \returns a reference to \c *this. - */ - Derived& setShift(const RealScalar& offset, const RealScalar& scale = 1) - { - m_shiftOffset = offset; - m_shiftScale = scale; - return derived(); - } - -#ifndef EIGEN_PARSED_BY_DOXYGEN - /** \internal */ - template - void dumpMemory(Stream& s) - { - int total = 0; - s << " L: " << ((total+=(m_matrix.cols()+1) * sizeof(int) + m_matrix.nonZeros()*(sizeof(int)+sizeof(Scalar))) >> 20) << "Mb" << "\n"; - s << " diag: " << ((total+=m_diag.size() * sizeof(Scalar)) >> 20) << "Mb" << "\n"; - s << " tree: " << ((total+=m_parent.size() * sizeof(int)) >> 20) << "Mb" << "\n"; - s << " nonzeros: " << ((total+=m_nonZerosPerCol.size() * sizeof(int)) >> 20) << "Mb" << "\n"; - s << " perm: " << ((total+=m_P.size() * sizeof(int)) >> 20) << "Mb" << "\n"; - s << " perm^-1: " << ((total+=m_Pinv.size() * sizeof(int)) >> 20) << "Mb" << "\n"; - s << " TOTAL: " << (total>> 20) << "Mb" << "\n"; - } - - /** \internal */ - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const - { - eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()"); - eigen_assert(m_matrix.rows()==b.rows()); - - if(m_info!=Success) - return; - - if(m_P.size()>0) - dest = m_P * b; - else - dest = b; - - if(m_matrix.nonZeros()>0) // otherwise L==I - derived().matrixL().solveInPlace(dest); - - if(m_diag.size()>0) - dest = m_diag.asDiagonal().inverse() * dest; - - if (m_matrix.nonZeros()>0) // otherwise U==I - derived().matrixU().solveInPlace(dest); - - if(m_P.size()>0) - dest = m_Pinv * dest; - } - - template - void _solve_impl(const SparseMatrixBase &b, SparseMatrixBase &dest) const - { - internal::solve_sparse_through_dense_panels(derived(), b, dest); - } - -#endif // EIGEN_PARSED_BY_DOXYGEN - - protected: - - /** Computes the sparse Cholesky decomposition of \a matrix */ - template - void compute(const MatrixType& matrix) - { - eigen_assert(matrix.rows()==matrix.cols()); - Index size = matrix.cols(); - CholMatrixType tmp(size,size); - ConstCholMatrixPtr pmat; - ordering(matrix, pmat, tmp); - analyzePattern_preordered(*pmat, DoLDLT); - factorize_preordered(*pmat); - } - - template - void factorize(const MatrixType& a) - { - eigen_assert(a.rows()==a.cols()); - Index size = a.cols(); - CholMatrixType tmp(size,size); - ConstCholMatrixPtr pmat; - - if(m_P.size()==0 && (UpLo&Upper)==Upper) - { - // If there is no ordering, try to directly use the input matrix without any copy - internal::simplicial_cholesky_grab_input::run(a, pmat, tmp); - } - else - { - tmp.template selfadjointView() = a.template selfadjointView().twistedBy(m_P); - pmat = &tmp; - } - - factorize_preordered(*pmat); - } - - template - void factorize_preordered(const CholMatrixType& a); - - void analyzePattern(const MatrixType& a, bool doLDLT) - { - eigen_assert(a.rows()==a.cols()); - Index size = a.cols(); - CholMatrixType tmp(size,size); - ConstCholMatrixPtr pmat; - ordering(a, pmat, tmp); - analyzePattern_preordered(*pmat,doLDLT); - } - void analyzePattern_preordered(const CholMatrixType& a, bool doLDLT); - - void ordering(const MatrixType& a, ConstCholMatrixPtr &pmat, CholMatrixType& ap); - - /** keeps off-diagonal entries; drops diagonal entries */ - struct keep_diag { - inline bool operator() (const Index& row, const Index& col, const Scalar&) const - { - return row!=col; - } - }; - - mutable ComputationInfo m_info; - bool m_factorizationIsOk; - bool m_analysisIsOk; - - CholMatrixType m_matrix; - VectorType m_diag; // the diagonal coefficients (LDLT mode) - VectorI m_parent; // elimination tree - VectorI m_nonZerosPerCol; - PermutationMatrix m_P; // the permutation - PermutationMatrix m_Pinv; // the inverse permutation - - RealScalar m_shiftOffset; - RealScalar m_shiftScale; -}; - -template > class SimplicialLLT; -template > class SimplicialLDLT; -template > class SimplicialCholesky; - -namespace internal { - -template struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Ordering OrderingType; - enum { UpLo = _UpLo }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef TriangularView MatrixL; - typedef TriangularView MatrixU; - static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); } - static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); } -}; - -template struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Ordering OrderingType; - enum { UpLo = _UpLo }; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef TriangularView MatrixL; - typedef TriangularView MatrixU; - static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); } - static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); } -}; - -template struct traits > -{ - typedef _MatrixType MatrixType; - typedef _Ordering OrderingType; - enum { UpLo = _UpLo }; -}; - -} - -/** \ingroup SparseCholesky_Module - * \class SimplicialLLT - * \brief A direct sparse LLT Cholesky factorizations - * - * This class provides a LL^T Cholesky factorizations of sparse matrices that are - * selfadjoint and positive definite. The factorization allows for solving A.X = B where - * X and B can be either dense or sparse. - * - * In order to reduce the fill-in, a symmetric permutation P is applied prior to the factorization - * such that the factorized matrix is P A P^-1. - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower - * or Upper. Default is Lower. - * \tparam _Ordering The ordering method to use, either AMDOrdering<> or NaturalOrdering<>. Default is AMDOrdering<> - * - * \implsparsesolverconcept - * - * \sa class SimplicialLDLT, class AMDOrdering, class NaturalOrdering - */ -template - class SimplicialLLT : public SimplicialCholeskyBase > -{ -public: - typedef _MatrixType MatrixType; - enum { UpLo = _UpLo }; - typedef SimplicialCholeskyBase Base; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef Matrix VectorType; - typedef internal::traits Traits; - typedef typename Traits::MatrixL MatrixL; - typedef typename Traits::MatrixU MatrixU; -public: - /** Default constructor */ - SimplicialLLT() : Base() {} - /** Constructs and performs the LLT factorization of \a matrix */ - explicit SimplicialLLT(const MatrixType& matrix) - : Base(matrix) {} - - /** \returns an expression of the factor L */ - inline const MatrixL matrixL() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial LLT not factorized"); - return Traits::getL(Base::m_matrix); - } - - /** \returns an expression of the factor U (= L^*) */ - inline const MatrixU matrixU() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial LLT not factorized"); - return Traits::getU(Base::m_matrix); - } - - /** Computes the sparse Cholesky decomposition of \a matrix */ - SimplicialLLT& compute(const MatrixType& matrix) - { - Base::template compute(matrix); - return *this; - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& a) - { - Base::analyzePattern(a, false); - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - void factorize(const MatrixType& a) - { - Base::template factorize(a); - } - - /** \returns the determinant of the underlying matrix from the current factorization */ - Scalar determinant() const - { - Scalar detL = Base::m_matrix.diagonal().prod(); - return numext::abs2(detL); - } -}; - -/** \ingroup SparseCholesky_Module - * \class SimplicialLDLT - * \brief A direct sparse LDLT Cholesky factorizations without square root. - * - * This class provides a LDL^T Cholesky factorizations without square root of sparse matrices that are - * selfadjoint and positive definite. The factorization allows for solving A.X = B where - * X and B can be either dense or sparse. - * - * In order to reduce the fill-in, a symmetric permutation P is applied prior to the factorization - * such that the factorized matrix is P A P^-1. - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * \tparam _UpLo the triangular part that will be used for the computations. It can be Lower - * or Upper. Default is Lower. - * \tparam _Ordering The ordering method to use, either AMDOrdering<> or NaturalOrdering<>. Default is AMDOrdering<> - * - * \implsparsesolverconcept - * - * \sa class SimplicialLLT, class AMDOrdering, class NaturalOrdering - */ -template - class SimplicialLDLT : public SimplicialCholeskyBase > -{ -public: - typedef _MatrixType MatrixType; - enum { UpLo = _UpLo }; - typedef SimplicialCholeskyBase Base; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef Matrix VectorType; - typedef internal::traits Traits; - typedef typename Traits::MatrixL MatrixL; - typedef typename Traits::MatrixU MatrixU; -public: - /** Default constructor */ - SimplicialLDLT() : Base() {} - - /** Constructs and performs the LLT factorization of \a matrix */ - explicit SimplicialLDLT(const MatrixType& matrix) - : Base(matrix) {} - - /** \returns a vector expression of the diagonal D */ - inline const VectorType vectorD() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial LDLT not factorized"); - return Base::m_diag; - } - /** \returns an expression of the factor L */ - inline const MatrixL matrixL() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial LDLT not factorized"); - return Traits::getL(Base::m_matrix); - } - - /** \returns an expression of the factor U (= L^*) */ - inline const MatrixU matrixU() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial LDLT not factorized"); - return Traits::getU(Base::m_matrix); - } - - /** Computes the sparse Cholesky decomposition of \a matrix */ - SimplicialLDLT& compute(const MatrixType& matrix) - { - Base::template compute(matrix); - return *this; - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& a) - { - Base::analyzePattern(a, true); - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - void factorize(const MatrixType& a) - { - Base::template factorize(a); - } - - /** \returns the determinant of the underlying matrix from the current factorization */ - Scalar determinant() const - { - return Base::m_diag.prod(); - } -}; - -/** \deprecated use SimplicialLDLT or class SimplicialLLT - * \ingroup SparseCholesky_Module - * \class SimplicialCholesky - * - * \sa class SimplicialLDLT, class SimplicialLLT - */ -template - class SimplicialCholesky : public SimplicialCholeskyBase > -{ -public: - typedef _MatrixType MatrixType; - enum { UpLo = _UpLo }; - typedef SimplicialCholeskyBase Base; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix CholMatrixType; - typedef Matrix VectorType; - typedef internal::traits Traits; - typedef internal::traits > LDLTTraits; - typedef internal::traits > LLTTraits; - public: - SimplicialCholesky() : Base(), m_LDLT(true) {} - - explicit SimplicialCholesky(const MatrixType& matrix) - : Base(), m_LDLT(true) - { - compute(matrix); - } - - SimplicialCholesky& setMode(SimplicialCholeskyMode mode) - { - switch(mode) - { - case SimplicialCholeskyLLT: - m_LDLT = false; - break; - case SimplicialCholeskyLDLT: - m_LDLT = true; - break; - default: - break; - } - - return *this; - } - - inline const VectorType vectorD() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial Cholesky not factorized"); - return Base::m_diag; - } - inline const CholMatrixType rawMatrix() const { - eigen_assert(Base::m_factorizationIsOk && "Simplicial Cholesky not factorized"); - return Base::m_matrix; - } - - /** Computes the sparse Cholesky decomposition of \a matrix */ - SimplicialCholesky& compute(const MatrixType& matrix) - { - if(m_LDLT) - Base::template compute(matrix); - else - Base::template compute(matrix); - return *this; - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& a) - { - Base::analyzePattern(a, m_LDLT); - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - void factorize(const MatrixType& a) - { - if(m_LDLT) - Base::template factorize(a); - else - Base::template factorize(a); - } - - /** \internal */ - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const - { - eigen_assert(Base::m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()"); - eigen_assert(Base::m_matrix.rows()==b.rows()); - - if(Base::m_info!=Success) - return; - - if(Base::m_P.size()>0) - dest = Base::m_P * b; - else - dest = b; - - if(Base::m_matrix.nonZeros()>0) // otherwise L==I - { - if(m_LDLT) - LDLTTraits::getL(Base::m_matrix).solveInPlace(dest); - else - LLTTraits::getL(Base::m_matrix).solveInPlace(dest); - } - - if(Base::m_diag.size()>0) - dest = Base::m_diag.real().asDiagonal().inverse() * dest; - - if (Base::m_matrix.nonZeros()>0) // otherwise I==I - { - if(m_LDLT) - LDLTTraits::getU(Base::m_matrix).solveInPlace(dest); - else - LLTTraits::getU(Base::m_matrix).solveInPlace(dest); - } - - if(Base::m_P.size()>0) - dest = Base::m_Pinv * dest; - } - - /** \internal */ - template - void _solve_impl(const SparseMatrixBase &b, SparseMatrixBase &dest) const - { - internal::solve_sparse_through_dense_panels(*this, b, dest); - } - - Scalar determinant() const - { - if(m_LDLT) - { - return Base::m_diag.prod(); - } - else - { - Scalar detL = Diagonal(Base::m_matrix).prod(); - return numext::abs2(detL); - } - } - - protected: - bool m_LDLT; -}; - -template -void SimplicialCholeskyBase::ordering(const MatrixType& a, ConstCholMatrixPtr &pmat, CholMatrixType& ap) -{ - eigen_assert(a.rows()==a.cols()); - const Index size = a.rows(); - pmat = ≈ - // Note that ordering methods compute the inverse permutation - if(!internal::is_same >::value) - { - { - CholMatrixType C; - C = a.template selfadjointView(); - - OrderingType ordering; - ordering(C,m_Pinv); - } - - if(m_Pinv.size()>0) m_P = m_Pinv.inverse(); - else m_P.resize(0); - - ap.resize(size,size); - ap.template selfadjointView() = a.template selfadjointView().twistedBy(m_P); - } - else - { - m_Pinv.resize(0); - m_P.resize(0); - if(int(UpLo)==int(Lower) || MatrixType::IsRowMajor) - { - // we have to transpose the lower part to to the upper one - ap.resize(size,size); - ap.template selfadjointView() = a.template selfadjointView(); - } - else - internal::simplicial_cholesky_grab_input::run(a, pmat, ap); - } -} - -} // end namespace Eigen - -#endif // EIGEN_SIMPLICIAL_CHOLESKY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h deleted file mode 100644 index 7b6183d08b1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h +++ /dev/null @@ -1,199 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2012 Gael Guennebaud - -/* - -NOTE: thes functions vave been adapted from the LDL library: - -LDL Copyright (c) 2005 by Timothy A. Davis. All Rights Reserved. - -LDL License: - - Your use or distribution of LDL or any modified version of - LDL implies that you agree to this License. - - This library is free software; you can redistribute it and/or - modify it under the terms of the GNU Lesser General Public - License as published by the Free Software Foundation; either - version 2.1 of the License, or (at your option) any later version. - - This library is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - Lesser General Public License for more details. - - You should have received a copy of the GNU Lesser General Public - License along with this library; if not, write to the Free Software - Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 - USA - - Permission is hereby granted to use or copy this program under the - terms of the GNU LGPL, provided that the Copyright, this License, - and the Availability of the original version is retained on all copies. - User documentation of any code that uses this code or any modified - version of this code must cite the Copyright, this License, the - Availability note, and "Used by permission." Permission to modify - the code and to distribute modified code is granted, provided the - Copyright, this License, and the Availability note are retained, - and a notice that the code was modified is included. - */ - -#include "../Core/util/NonMPL2.h" - -#ifndef EIGEN_SIMPLICIAL_CHOLESKY_IMPL_H -#define EIGEN_SIMPLICIAL_CHOLESKY_IMPL_H - -namespace Eigen { - -template -void SimplicialCholeskyBase::analyzePattern_preordered(const CholMatrixType& ap, bool doLDLT) -{ - const StorageIndex size = StorageIndex(ap.rows()); - m_matrix.resize(size, size); - m_parent.resize(size); - m_nonZerosPerCol.resize(size); - - ei_declare_aligned_stack_constructed_variable(StorageIndex, tags, size, 0); - - for(StorageIndex k = 0; k < size; ++k) - { - /* L(k,:) pattern: all nodes reachable in etree from nz in A(0:k-1,k) */ - m_parent[k] = -1; /* parent of k is not yet known */ - tags[k] = k; /* mark node k as visited */ - m_nonZerosPerCol[k] = 0; /* count of nonzeros in column k of L */ - for(typename CholMatrixType::InnerIterator it(ap,k); it; ++it) - { - StorageIndex i = it.index(); - if(i < k) - { - /* follow path from i to root of etree, stop at flagged node */ - for(; tags[i] != k; i = m_parent[i]) - { - /* find parent of i if not yet determined */ - if (m_parent[i] == -1) - m_parent[i] = k; - m_nonZerosPerCol[i]++; /* L (k,i) is nonzero */ - tags[i] = k; /* mark i as visited */ - } - } - } - } - - /* construct Lp index array from m_nonZerosPerCol column counts */ - StorageIndex* Lp = m_matrix.outerIndexPtr(); - Lp[0] = 0; - for(StorageIndex k = 0; k < size; ++k) - Lp[k+1] = Lp[k] + m_nonZerosPerCol[k] + (doLDLT ? 0 : 1); - - m_matrix.resizeNonZeros(Lp[size]); - - m_isInitialized = true; - m_info = Success; - m_analysisIsOk = true; - m_factorizationIsOk = false; -} - - -template -template -void SimplicialCholeskyBase::factorize_preordered(const CholMatrixType& ap) -{ - using std::sqrt; - - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - eigen_assert(ap.rows()==ap.cols()); - eigen_assert(m_parent.size()==ap.rows()); - eigen_assert(m_nonZerosPerCol.size()==ap.rows()); - - const StorageIndex size = StorageIndex(ap.rows()); - const StorageIndex* Lp = m_matrix.outerIndexPtr(); - StorageIndex* Li = m_matrix.innerIndexPtr(); - Scalar* Lx = m_matrix.valuePtr(); - - ei_declare_aligned_stack_constructed_variable(Scalar, y, size, 0); - ei_declare_aligned_stack_constructed_variable(StorageIndex, pattern, size, 0); - ei_declare_aligned_stack_constructed_variable(StorageIndex, tags, size, 0); - - bool ok = true; - m_diag.resize(DoLDLT ? size : 0); - - for(StorageIndex k = 0; k < size; ++k) - { - // compute nonzero pattern of kth row of L, in topological order - y[k] = 0.0; // Y(0:k) is now all zero - StorageIndex top = size; // stack for pattern is empty - tags[k] = k; // mark node k as visited - m_nonZerosPerCol[k] = 0; // count of nonzeros in column k of L - for(typename CholMatrixType::InnerIterator it(ap,k); it; ++it) - { - StorageIndex i = it.index(); - if(i <= k) - { - y[i] += numext::conj(it.value()); /* scatter A(i,k) into Y (sum duplicates) */ - Index len; - for(len = 0; tags[i] != k; i = m_parent[i]) - { - pattern[len++] = i; /* L(k,i) is nonzero */ - tags[i] = k; /* mark i as visited */ - } - while(len > 0) - pattern[--top] = pattern[--len]; - } - } - - /* compute numerical values kth row of L (a sparse triangular solve) */ - - RealScalar d = numext::real(y[k]) * m_shiftScale + m_shiftOffset; // get D(k,k), apply the shift function, and clear Y(k) - y[k] = 0.0; - for(; top < size; ++top) - { - Index i = pattern[top]; /* pattern[top:n-1] is pattern of L(:,k) */ - Scalar yi = y[i]; /* get and clear Y(i) */ - y[i] = 0.0; - - /* the nonzero entry L(k,i) */ - Scalar l_ki; - if(DoLDLT) - l_ki = yi / numext::real(m_diag[i]); - else - yi = l_ki = yi / Lx[Lp[i]]; - - Index p2 = Lp[i] + m_nonZerosPerCol[i]; - Index p; - for(p = Lp[i] + (DoLDLT ? 0 : 1); p < p2; ++p) - y[Li[p]] -= numext::conj(Lx[p]) * yi; - d -= numext::real(l_ki * numext::conj(yi)); - Li[p] = k; /* store L(k,i) in column form of L */ - Lx[p] = l_ki; - ++m_nonZerosPerCol[i]; /* increment count of nonzeros in col i */ - } - if(DoLDLT) - { - m_diag[k] = d; - if(d == RealScalar(0)) - { - ok = false; /* failure, D(k,k) is zero */ - break; - } - } - else - { - Index p = Lp[k] + m_nonZerosPerCol[k]++; - Li[p] = k ; /* store L(k,k) = sqrt (d) in column k */ - if(d <= RealScalar(0)) { - ok = false; /* failure, matrix is not positive definite */ - break; - } - Lx[p] = sqrt(d) ; - } - } - - m_info = ok ? Success : NumericalIssue; - m_factorizationIsOk = true; -} - -} // end namespace Eigen - -#endif // EIGEN_SIMPLICIAL_CHOLESKY_IMPL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/AmbiVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/AmbiVector.h deleted file mode 100644 index 2cb7747cc99..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/AmbiVector.h +++ /dev/null @@ -1,378 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_AMBIVECTOR_H -#define EIGEN_AMBIVECTOR_H - -namespace Eigen { - -namespace internal { - -/** \internal - * Hybrid sparse/dense vector class designed for intensive read-write operations. - * - * See BasicSparseLLT and SparseProduct for usage examples. - */ -template -class AmbiVector -{ - public: - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - typedef typename NumTraits::Real RealScalar; - - explicit AmbiVector(Index size) - : m_buffer(0), m_zero(0), m_size(0), m_end(0), m_allocatedSize(0), m_allocatedElements(0), m_mode(-1) - { - resize(size); - } - - void init(double estimatedDensity); - void init(int mode); - - Index nonZeros() const; - - /** Specifies a sub-vector to work on */ - void setBounds(Index start, Index end) { m_start = convert_index(start); m_end = convert_index(end); } - - void setZero(); - - void restart(); - Scalar& coeffRef(Index i); - Scalar& coeff(Index i); - - class Iterator; - - ~AmbiVector() { delete[] m_buffer; } - - void resize(Index size) - { - if (m_allocatedSize < size) - reallocate(size); - m_size = convert_index(size); - } - - StorageIndex size() const { return m_size; } - - protected: - StorageIndex convert_index(Index idx) - { - return internal::convert_index(idx); - } - - void reallocate(Index size) - { - // if the size of the matrix is not too large, let's allocate a bit more than needed such - // that we can handle dense vector even in sparse mode. - delete[] m_buffer; - if (size<1000) - { - Index allocSize = (size * sizeof(ListEl) + sizeof(Scalar) - 1)/sizeof(Scalar); - m_allocatedElements = convert_index((allocSize*sizeof(Scalar))/sizeof(ListEl)); - m_buffer = new Scalar[allocSize]; - } - else - { - m_allocatedElements = convert_index((size*sizeof(Scalar))/sizeof(ListEl)); - m_buffer = new Scalar[size]; - } - m_size = convert_index(size); - m_start = 0; - m_end = m_size; - } - - void reallocateSparse() - { - Index copyElements = m_allocatedElements; - m_allocatedElements = (std::min)(StorageIndex(m_allocatedElements*1.5),m_size); - Index allocSize = m_allocatedElements * sizeof(ListEl); - allocSize = (allocSize + sizeof(Scalar) - 1)/sizeof(Scalar); - Scalar* newBuffer = new Scalar[allocSize]; - std::memcpy(newBuffer, m_buffer, copyElements * sizeof(ListEl)); - delete[] m_buffer; - m_buffer = newBuffer; - } - - protected: - // element type of the linked list - struct ListEl - { - StorageIndex next; - StorageIndex index; - Scalar value; - }; - - // used to store data in both mode - Scalar* m_buffer; - Scalar m_zero; - StorageIndex m_size; - StorageIndex m_start; - StorageIndex m_end; - StorageIndex m_allocatedSize; - StorageIndex m_allocatedElements; - StorageIndex m_mode; - - // linked list mode - StorageIndex m_llStart; - StorageIndex m_llCurrent; - StorageIndex m_llSize; -}; - -/** \returns the number of non zeros in the current sub vector */ -template -Index AmbiVector<_Scalar,_StorageIndex>::nonZeros() const -{ - if (m_mode==IsSparse) - return m_llSize; - else - return m_end - m_start; -} - -template -void AmbiVector<_Scalar,_StorageIndex>::init(double estimatedDensity) -{ - if (estimatedDensity>0.1) - init(IsDense); - else - init(IsSparse); -} - -template -void AmbiVector<_Scalar,_StorageIndex>::init(int mode) -{ - m_mode = mode; - // This is only necessary in sparse mode, but we set these unconditionally to avoid some maybe-uninitialized warnings - // if (m_mode==IsSparse) - { - m_llSize = 0; - m_llStart = -1; - } -} - -/** Must be called whenever we might perform a write access - * with an index smaller than the previous one. - * - * Don't worry, this function is extremely cheap. - */ -template -void AmbiVector<_Scalar,_StorageIndex>::restart() -{ - m_llCurrent = m_llStart; -} - -/** Set all coefficients of current subvector to zero */ -template -void AmbiVector<_Scalar,_StorageIndex>::setZero() -{ - if (m_mode==IsDense) - { - for (Index i=m_start; i -_Scalar& AmbiVector<_Scalar,_StorageIndex>::coeffRef(Index i) -{ - if (m_mode==IsDense) - return m_buffer[i]; - else - { - ListEl* EIGEN_RESTRICT llElements = reinterpret_cast(m_buffer); - // TODO factorize the following code to reduce code generation - eigen_assert(m_mode==IsSparse); - if (m_llSize==0) - { - // this is the first element - m_llStart = 0; - m_llCurrent = 0; - ++m_llSize; - llElements[0].value = Scalar(0); - llElements[0].index = convert_index(i); - llElements[0].next = -1; - return llElements[0].value; - } - else if (i=llElements[m_llCurrent].index && "you must call restart() before inserting an element with lower or equal index"); - while (nextel >= 0 && llElements[nextel].index<=i) - { - m_llCurrent = nextel; - nextel = llElements[nextel].next; - } - - if (llElements[m_llCurrent].index==i) - { - // the coefficient already exists and we found it ! - return llElements[m_llCurrent].value; - } - else - { - if (m_llSize>=m_allocatedElements) - { - reallocateSparse(); - llElements = reinterpret_cast(m_buffer); - } - eigen_internal_assert(m_llSize -_Scalar& AmbiVector<_Scalar,_StorageIndex>::coeff(Index i) -{ - if (m_mode==IsDense) - return m_buffer[i]; - else - { - ListEl* EIGEN_RESTRICT llElements = reinterpret_cast(m_buffer); - eigen_assert(m_mode==IsSparse); - if ((m_llSize==0) || (i= 0 && llElements[elid].index -class AmbiVector<_Scalar,_StorageIndex>::Iterator -{ - public: - typedef _Scalar Scalar; - typedef typename NumTraits::Real RealScalar; - - /** Default constructor - * \param vec the vector on which we iterate - * \param epsilon the minimal value used to prune zero coefficients. - * In practice, all coefficients having a magnitude smaller than \a epsilon - * are skipped. - */ - explicit Iterator(const AmbiVector& vec, const RealScalar& epsilon = 0) - : m_vector(vec) - { - using std::abs; - m_epsilon = epsilon; - m_isDense = m_vector.m_mode==IsDense; - if (m_isDense) - { - m_currentEl = 0; // this is to avoid a compilation warning - m_cachedValue = 0; // this is to avoid a compilation warning - m_cachedIndex = m_vector.m_start-1; - ++(*this); - } - else - { - ListEl* EIGEN_RESTRICT llElements = reinterpret_cast(m_vector.m_buffer); - m_currentEl = m_vector.m_llStart; - while (m_currentEl>=0 && abs(llElements[m_currentEl].value)<=m_epsilon) - m_currentEl = llElements[m_currentEl].next; - if (m_currentEl<0) - { - m_cachedValue = 0; // this is to avoid a compilation warning - m_cachedIndex = -1; - } - else - { - m_cachedIndex = llElements[m_currentEl].index; - m_cachedValue = llElements[m_currentEl].value; - } - } - } - - StorageIndex index() const { return m_cachedIndex; } - Scalar value() const { return m_cachedValue; } - - operator bool() const { return m_cachedIndex>=0; } - - Iterator& operator++() - { - using std::abs; - if (m_isDense) - { - do { - ++m_cachedIndex; - } while (m_cachedIndex(m_vector.m_buffer); - do { - m_currentEl = llElements[m_currentEl].next; - } while (m_currentEl>=0 && abs(llElements[m_currentEl].value)<=m_epsilon); - if (m_currentEl<0) - { - m_cachedIndex = -1; - } - else - { - m_cachedIndex = llElements[m_currentEl].index; - m_cachedValue = llElements[m_currentEl].value; - } - } - return *this; - } - - protected: - const AmbiVector& m_vector; // the target vector - StorageIndex m_currentEl; // the current element in sparse/linked-list mode - RealScalar m_epsilon; // epsilon used to prune zero coefficients - StorageIndex m_cachedIndex; // current coordinate - Scalar m_cachedValue; // current value - bool m_isDense; // mode of the vector -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_AMBIVECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/CompressedStorage.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/CompressedStorage.h deleted file mode 100644 index d89fa0dae47..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/CompressedStorage.h +++ /dev/null @@ -1,258 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPRESSED_STORAGE_H -#define EIGEN_COMPRESSED_STORAGE_H - -namespace Eigen { - -namespace internal { - -/** \internal - * Stores a sparse set of values as a list of values and a list of indices. - * - */ -template -class CompressedStorage -{ - public: - - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - - protected: - - typedef typename NumTraits::Real RealScalar; - - public: - - CompressedStorage() - : m_values(0), m_indices(0), m_size(0), m_allocatedSize(0) - {} - - explicit CompressedStorage(Index size) - : m_values(0), m_indices(0), m_size(0), m_allocatedSize(0) - { - resize(size); - } - - CompressedStorage(const CompressedStorage& other) - : m_values(0), m_indices(0), m_size(0), m_allocatedSize(0) - { - *this = other; - } - - CompressedStorage& operator=(const CompressedStorage& other) - { - resize(other.size()); - if(other.size()>0) - { - internal::smart_copy(other.m_values, other.m_values + m_size, m_values); - internal::smart_copy(other.m_indices, other.m_indices + m_size, m_indices); - } - return *this; - } - - void swap(CompressedStorage& other) - { - std::swap(m_values, other.m_values); - std::swap(m_indices, other.m_indices); - std::swap(m_size, other.m_size); - std::swap(m_allocatedSize, other.m_allocatedSize); - } - - ~CompressedStorage() - { - delete[] m_values; - delete[] m_indices; - } - - void reserve(Index size) - { - Index newAllocatedSize = m_size + size; - if (newAllocatedSize > m_allocatedSize) - reallocate(newAllocatedSize); - } - - void squeeze() - { - if (m_allocatedSize>m_size) - reallocate(m_size); - } - - void resize(Index size, double reserveSizeFactor = 0) - { - if (m_allocatedSize)(NumTraits::highest(), size + Index(reserveSizeFactor*double(size))); - if(realloc_size(i); - } - - inline Index size() const { return m_size; } - inline Index allocatedSize() const { return m_allocatedSize; } - inline void clear() { m_size = 0; } - - const Scalar* valuePtr() const { return m_values; } - Scalar* valuePtr() { return m_values; } - const StorageIndex* indexPtr() const { return m_indices; } - StorageIndex* indexPtr() { return m_indices; } - - inline Scalar& value(Index i) { eigen_internal_assert(m_values!=0); return m_values[i]; } - inline const Scalar& value(Index i) const { eigen_internal_assert(m_values!=0); return m_values[i]; } - - inline StorageIndex& index(Index i) { eigen_internal_assert(m_indices!=0); return m_indices[i]; } - inline const StorageIndex& index(Index i) const { eigen_internal_assert(m_indices!=0); return m_indices[i]; } - - /** \returns the largest \c k such that for all \c j in [0,k) index[\c j]\<\a key */ - inline Index searchLowerIndex(Index key) const - { - return searchLowerIndex(0, m_size, key); - } - - /** \returns the largest \c k in [start,end) such that for all \c j in [start,k) index[\c j]\<\a key */ - inline Index searchLowerIndex(Index start, Index end, Index key) const - { - while(end>start) - { - Index mid = (end+start)>>1; - if (m_indices[mid]=end) - return defaultValue; - else if (end>start && key==m_indices[end-1]) - return m_values[end-1]; - // ^^ optimization: let's first check if it is the last coefficient - // (very common in high level algorithms) - const Index id = searchLowerIndex(start,end-1,key); - return ((id=m_size || m_indices[id]!=key) - { - if (m_allocatedSize newValues(m_allocatedSize); - internal::scoped_array newIndices(m_allocatedSize); - - // copy first chunk - internal::smart_copy(m_values, m_values +id, newValues.ptr()); - internal::smart_copy(m_indices, m_indices+id, newIndices.ptr()); - - // copy the rest - if(m_size>id) - { - internal::smart_copy(m_values +id, m_values +m_size, newValues.ptr() +id+1); - internal::smart_copy(m_indices+id, m_indices+m_size, newIndices.ptr()+id+1); - } - std::swap(m_values,newValues.ptr()); - std::swap(m_indices,newIndices.ptr()); - } - else if(m_size>id) - { - internal::smart_memmove(m_values +id, m_values +m_size, m_values +id+1); - internal::smart_memmove(m_indices+id, m_indices+m_size, m_indices+id+1); - } - m_size++; - m_indices[id] = internal::convert_index(key); - m_values[id] = defaultValue; - } - return m_values[id]; - } - - void prune(const Scalar& reference, const RealScalar& epsilon = NumTraits::dummy_precision()) - { - Index k = 0; - Index n = size(); - for (Index i=0; i newValues(size); - internal::scoped_array newIndices(size); - Index copySize = (std::min)(size, m_size); - if (copySize>0) { - internal::smart_copy(m_values, m_values+copySize, newValues.ptr()); - internal::smart_copy(m_indices, m_indices+copySize, newIndices.ptr()); - } - std::swap(m_values,newValues.ptr()); - std::swap(m_indices,newIndices.ptr()); - m_allocatedSize = size; - } - - protected: - Scalar* m_values; - StorageIndex* m_indices; - Index m_size; - Index m_allocatedSize; - -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPRESSED_STORAGE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/ConservativeSparseSparseProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/ConservativeSparseSparseProduct.h deleted file mode 100644 index 9db119b67f9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/ConservativeSparseSparseProduct.h +++ /dev/null @@ -1,352 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_CONSERVATIVESPARSESPARSEPRODUCT_H -#define EIGEN_CONSERVATIVESPARSESPARSEPRODUCT_H - -namespace Eigen { - -namespace internal { - -template -static void conservative_sparse_sparse_product_impl(const Lhs& lhs, const Rhs& rhs, ResultType& res, bool sortedInsertion = false) -{ - typedef typename remove_all::type::Scalar LhsScalar; - typedef typename remove_all::type::Scalar RhsScalar; - typedef typename remove_all::type::Scalar ResScalar; - - // make sure to call innerSize/outerSize since we fake the storage order. - Index rows = lhs.innerSize(); - Index cols = rhs.outerSize(); - eigen_assert(lhs.outerSize() == rhs.innerSize()); - - ei_declare_aligned_stack_constructed_variable(bool, mask, rows, 0); - ei_declare_aligned_stack_constructed_variable(ResScalar, values, rows, 0); - ei_declare_aligned_stack_constructed_variable(Index, indices, rows, 0); - - std::memset(mask,0,sizeof(bool)*rows); - - evaluator lhsEval(lhs); - evaluator rhsEval(rhs); - - // estimate the number of non zero entries - // given a rhs column containing Y non zeros, we assume that the respective Y columns - // of the lhs differs in average of one non zeros, thus the number of non zeros for - // the product of a rhs column with the lhs is X+Y where X is the average number of non zero - // per column of the lhs. - // Therefore, we have nnz(lhs*rhs) = nnz(lhs) + nnz(rhs) - Index estimated_nnz_prod = lhsEval.nonZerosEstimate() + rhsEval.nonZerosEstimate(); - - res.setZero(); - res.reserve(Index(estimated_nnz_prod)); - // we compute each column of the result, one after the other - for (Index j=0; j::InnerIterator rhsIt(rhsEval, j); rhsIt; ++rhsIt) - { - RhsScalar y = rhsIt.value(); - Index k = rhsIt.index(); - for (typename evaluator::InnerIterator lhsIt(lhsEval, k); lhsIt; ++lhsIt) - { - Index i = lhsIt.index(); - LhsScalar x = lhsIt.value(); - if(!mask[i]) - { - mask[i] = true; - values[i] = x * y; - indices[nnz] = i; - ++nnz; - } - else - values[i] += x * y; - } - } - if(!sortedInsertion) - { - // unordered insertion - for(Index k=0; k use a quick sort - // otherwise => loop through the entire vector - // In order to avoid to perform an expensive log2 when the - // result is clearly very sparse we use a linear bound up to 200. - if((nnz<200 && nnz1) std::sort(indices,indices+nnz); - for(Index k=0; k::Flags&RowMajorBit) ? RowMajor : ColMajor, - int RhsStorageOrder = (traits::Flags&RowMajorBit) ? RowMajor : ColMajor, - int ResStorageOrder = (traits::Flags&RowMajorBit) ? RowMajor : ColMajor> -struct conservative_sparse_sparse_product_selector; - -template -struct conservative_sparse_sparse_product_selector -{ - typedef typename remove_all::type LhsCleaned; - typedef typename LhsCleaned::Scalar Scalar; - - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix RowMajorMatrix; - typedef SparseMatrix ColMajorMatrixAux; - typedef typename sparse_eval::type ColMajorMatrix; - - // If the result is tall and thin (in the extreme case a column vector) - // then it is faster to sort the coefficients inplace instead of transposing twice. - // FIXME, the following heuristic is probably not very good. - if(lhs.rows()>rhs.cols()) - { - ColMajorMatrix resCol(lhs.rows(),rhs.cols()); - // perform sorted insertion - internal::conservative_sparse_sparse_product_impl(lhs, rhs, resCol, true); - res = resCol.markAsRValue(); - } - else - { - ColMajorMatrixAux resCol(lhs.rows(),rhs.cols()); - // ressort to transpose to sort the entries - internal::conservative_sparse_sparse_product_impl(lhs, rhs, resCol, false); - RowMajorMatrix resRow(resCol); - res = resRow.markAsRValue(); - } - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix RowMajorRhs; - typedef SparseMatrix RowMajorRes; - RowMajorRhs rhsRow = rhs; - RowMajorRes resRow(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(rhsRow, lhs, resRow); - res = resRow; - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix RowMajorLhs; - typedef SparseMatrix RowMajorRes; - RowMajorLhs lhsRow = lhs; - RowMajorRes resRow(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(rhs, lhsRow, resRow); - res = resRow; - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix RowMajorMatrix; - RowMajorMatrix resRow(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(rhs, lhs, resRow); - res = resRow; - } -}; - - -template -struct conservative_sparse_sparse_product_selector -{ - typedef typename traits::type>::Scalar Scalar; - - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix ColMajorMatrix; - ColMajorMatrix resCol(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(lhs, rhs, resCol); - res = resCol; - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix ColMajorLhs; - typedef SparseMatrix ColMajorRes; - ColMajorLhs lhsCol = lhs; - ColMajorRes resCol(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(lhsCol, rhs, resCol); - res = resCol; - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix ColMajorRhs; - typedef SparseMatrix ColMajorRes; - ColMajorRhs rhsCol = rhs; - ColMajorRes resCol(lhs.rows(), rhs.cols()); - internal::conservative_sparse_sparse_product_impl(lhs, rhsCol, resCol); - res = resCol; - } -}; - -template -struct conservative_sparse_sparse_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix RowMajorMatrix; - typedef SparseMatrix ColMajorMatrix; - RowMajorMatrix resRow(lhs.rows(),rhs.cols()); - internal::conservative_sparse_sparse_product_impl(rhs, lhs, resRow); - // sort the non zeros: - ColMajorMatrix resCol(resRow); - res = resCol; - } -}; - -} // end namespace internal - - -namespace internal { - -template -static void sparse_sparse_to_dense_product_impl(const Lhs& lhs, const Rhs& rhs, ResultType& res) -{ - typedef typename remove_all::type::Scalar LhsScalar; - typedef typename remove_all::type::Scalar RhsScalar; - Index cols = rhs.outerSize(); - eigen_assert(lhs.outerSize() == rhs.innerSize()); - - evaluator lhsEval(lhs); - evaluator rhsEval(rhs); - - for (Index j=0; j::InnerIterator rhsIt(rhsEval, j); rhsIt; ++rhsIt) - { - RhsScalar y = rhsIt.value(); - Index k = rhsIt.index(); - for (typename evaluator::InnerIterator lhsIt(lhsEval, k); lhsIt; ++lhsIt) - { - Index i = lhsIt.index(); - LhsScalar x = lhsIt.value(); - res.coeffRef(i,j) += x * y; - } - } - } -} - - -} // end namespace internal - -namespace internal { - -template::Flags&RowMajorBit) ? RowMajor : ColMajor, - int RhsStorageOrder = (traits::Flags&RowMajorBit) ? RowMajor : ColMajor> -struct sparse_sparse_to_dense_product_selector; - -template -struct sparse_sparse_to_dense_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - internal::sparse_sparse_to_dense_product_impl(lhs, rhs, res); - } -}; - -template -struct sparse_sparse_to_dense_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix ColMajorLhs; - ColMajorLhs lhsCol(lhs); - internal::sparse_sparse_to_dense_product_impl(lhsCol, rhs, res); - } -}; - -template -struct sparse_sparse_to_dense_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - typedef SparseMatrix ColMajorRhs; - ColMajorRhs rhsCol(rhs); - internal::sparse_sparse_to_dense_product_impl(lhs, rhsCol, res); - } -}; - -template -struct sparse_sparse_to_dense_product_selector -{ - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res) - { - Transpose trRes(res); - internal::sparse_sparse_to_dense_product_impl >(rhs, lhs, trRes); - } -}; - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_CONSERVATIVESPARSESPARSEPRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/MappedSparseMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/MappedSparseMatrix.h deleted file mode 100644 index 67718c85be9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/MappedSparseMatrix.h +++ /dev/null @@ -1,67 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MAPPED_SPARSEMATRIX_H -#define EIGEN_MAPPED_SPARSEMATRIX_H - -namespace Eigen { - -/** \deprecated Use Map > - * \class MappedSparseMatrix - * - * \brief Sparse matrix - * - * \param _Scalar the scalar type, i.e. the type of the coefficients - * - * See http://www.netlib.org/linalg/html_templates/node91.html for details on the storage scheme. - * - */ -namespace internal { -template -struct traits > : traits > -{}; -} // end namespace internal - -template -class MappedSparseMatrix - : public Map > -{ - typedef Map > Base; - - public: - - typedef typename Base::StorageIndex StorageIndex; - typedef typename Base::Scalar Scalar; - - inline MappedSparseMatrix(Index rows, Index cols, Index nnz, StorageIndex* outerIndexPtr, StorageIndex* innerIndexPtr, Scalar* valuePtr, StorageIndex* innerNonZeroPtr = 0) - : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZeroPtr) - {} - - /** Empty destructor */ - inline ~MappedSparseMatrix() {} -}; - -namespace internal { - -template -struct evaluator > - : evaluator > > -{ - typedef MappedSparseMatrix<_Scalar,_Options,_StorageIndex> XprType; - typedef evaluator > Base; - - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_MAPPED_SPARSEMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseAssign.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseAssign.h deleted file mode 100644 index 18352a847b2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseAssign.h +++ /dev/null @@ -1,216 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEASSIGN_H -#define EIGEN_SPARSEASSIGN_H - -namespace Eigen { - -template -template -Derived& SparseMatrixBase::operator=(const EigenBase &other) -{ - internal::call_assignment_no_alias(derived(), other.derived()); - return derived(); -} - -template -template -Derived& SparseMatrixBase::operator=(const ReturnByValue& other) -{ - // TODO use the evaluator mechanism - other.evalTo(derived()); - return derived(); -} - -template -template -inline Derived& SparseMatrixBase::operator=(const SparseMatrixBase& other) -{ - // by default sparse evaluation do not alias, so we can safely bypass the generic call_assignment routine - internal::Assignment > - ::run(derived(), other.derived(), internal::assign_op()); - return derived(); -} - -template -inline Derived& SparseMatrixBase::operator=(const Derived& other) -{ - internal::call_assignment_no_alias(derived(), other.derived()); - return derived(); -} - -namespace internal { - -template<> -struct storage_kind_to_evaluator_kind { - typedef IteratorBased Kind; -}; - -template<> -struct storage_kind_to_shape { - typedef SparseShape Shape; -}; - -struct Sparse2Sparse {}; -struct Sparse2Dense {}; - -template<> struct AssignmentKind { typedef Sparse2Sparse Kind; }; -template<> struct AssignmentKind { typedef Sparse2Sparse Kind; }; -template<> struct AssignmentKind { typedef Sparse2Dense Kind; }; -template<> struct AssignmentKind { typedef Sparse2Dense Kind; }; - - -template -void assign_sparse_to_sparse(DstXprType &dst, const SrcXprType &src) -{ - typedef typename DstXprType::Scalar Scalar; - typedef internal::evaluator DstEvaluatorType; - typedef internal::evaluator SrcEvaluatorType; - - SrcEvaluatorType srcEvaluator(src); - - const bool transpose = (DstEvaluatorType::Flags & RowMajorBit) != (SrcEvaluatorType::Flags & RowMajorBit); - const Index outerEvaluationSize = (SrcEvaluatorType::Flags&RowMajorBit) ? src.rows() : src.cols(); - if ((!transpose) && src.isRValue()) - { - // eval without temporary - dst.resize(src.rows(), src.cols()); - dst.setZero(); - dst.reserve((std::max)(src.rows(),src.cols())*2); - for (Index j=0; j::SupportedAccessPatterns & OuterRandomAccessPattern)==OuterRandomAccessPattern) || - (!((DstEvaluatorType::Flags & RowMajorBit) != (SrcEvaluatorType::Flags & RowMajorBit)))) && - "the transpose operation is supposed to be handled in SparseMatrix::operator="); - - enum { Flip = (DstEvaluatorType::Flags & RowMajorBit) != (SrcEvaluatorType::Flags & RowMajorBit) }; - - - DstXprType temp(src.rows(), src.cols()); - - temp.reserve((std::max)(src.rows(),src.cols())*2); - for (Index j=0; j -struct Assignment -{ - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) - { - assign_sparse_to_sparse(dst.derived(), src.derived()); - } -}; - -// Generic Sparse to Dense assignment -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - if(internal::is_same >::value) - dst.setZero(); - - internal::evaluator srcEval(src); - resize_if_allowed(dst, src, func); - internal::evaluator dstEval(dst); - - const Index outerEvaluationSize = (internal::evaluator::Flags&RowMajorBit) ? src.rows() : src.cols(); - for (Index j=0; j::InnerIterator i(srcEval,j); i; ++i) - func.assignCoeff(dstEval.coeffRef(i.row(),i.col()), i.value()); - } -}; - -// Specialization for "dst = dec.solve(rhs)" -// NOTE we need to specialize it for Sparse2Sparse to avoid ambiguous specialization error -template -struct Assignment, internal::assign_op, Sparse2Sparse> -{ - typedef Solve SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec()._solve_impl(src.rhs(), dst); - } -}; - -struct Diagonal2Sparse {}; - -template<> struct AssignmentKind { typedef Diagonal2Sparse Kind; }; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - typedef typename DstXprType::StorageIndex StorageIndex; - typedef typename DstXprType::Scalar Scalar; - typedef Array ArrayXI; - typedef Array ArrayXS; - template - static void run(SparseMatrix &dst, const SrcXprType &src, const internal::assign_op &/*func*/) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - Index size = src.diagonal().size(); - dst.makeCompressed(); - dst.resizeNonZeros(size); - Map(dst.innerIndexPtr(), size).setLinSpaced(0,StorageIndex(size)-1); - Map(dst.outerIndexPtr(), size+1).setLinSpaced(0,StorageIndex(size)); - Map(dst.valuePtr(), size) = src.diagonal(); - } - - template - static void run(SparseMatrixBase &dst, const SrcXprType &src, const internal::assign_op &/*func*/) - { - dst.diagonal() = src.diagonal(); - } - - static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &/*func*/) - { dst.diagonal() += src.diagonal(); } - - static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &/*func*/) - { dst.diagonal() -= src.diagonal(); } -}; -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSEASSIGN_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseBlock.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseBlock.h deleted file mode 100644 index 511e92b2f9b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseBlock.h +++ /dev/null @@ -1,603 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_BLOCK_H -#define EIGEN_SPARSE_BLOCK_H - -namespace Eigen { - -// Subset of columns or rows -template -class BlockImpl - : public SparseMatrixBase > -{ - typedef typename internal::remove_all::type _MatrixTypeNested; - typedef Block BlockType; -public: - enum { IsRowMajor = internal::traits::IsRowMajor }; -protected: - enum { OuterSize = IsRowMajor ? BlockRows : BlockCols }; - typedef SparseMatrixBase Base; - using Base::convert_index; -public: - EIGEN_SPARSE_PUBLIC_INTERFACE(BlockType) - - inline BlockImpl(XprType& xpr, Index i) - : m_matrix(xpr), m_outerStart(convert_index(i)), m_outerSize(OuterSize) - {} - - inline BlockImpl(XprType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) - : m_matrix(xpr), m_outerStart(convert_index(IsRowMajor ? startRow : startCol)), m_outerSize(convert_index(IsRowMajor ? blockRows : blockCols)) - {} - - EIGEN_STRONG_INLINE Index rows() const { return IsRowMajor ? m_outerSize.value() : m_matrix.rows(); } - EIGEN_STRONG_INLINE Index cols() const { return IsRowMajor ? m_matrix.cols() : m_outerSize.value(); } - - Index nonZeros() const - { - typedef internal::evaluator EvaluatorType; - EvaluatorType matEval(m_matrix); - Index nnz = 0; - Index end = m_outerStart + m_outerSize.value(); - for(Index j=m_outerStart; j::non_const_type m_matrix; - Index m_outerStart; - const internal::variable_if_dynamic m_outerSize; - - protected: - // Disable assignment with clear error message. - // Note that simply removing operator= yields compilation errors with ICC+MSVC - template - BlockImpl& operator=(const T&) - { - EIGEN_STATIC_ASSERT(sizeof(T)==0, THIS_SPARSE_BLOCK_SUBEXPRESSION_IS_READ_ONLY); - return *this; - } -}; - - -/*************************************************************************** -* specialization for SparseMatrix -***************************************************************************/ - -namespace internal { - -template -class sparse_matrix_block_impl - : public SparseCompressedBase > -{ - typedef typename internal::remove_all::type _MatrixTypeNested; - typedef Block BlockType; - typedef SparseCompressedBase > Base; - using Base::convert_index; -public: - enum { IsRowMajor = internal::traits::IsRowMajor }; - EIGEN_SPARSE_PUBLIC_INTERFACE(BlockType) -protected: - typedef typename Base::IndexVector IndexVector; - enum { OuterSize = IsRowMajor ? BlockRows : BlockCols }; -public: - - inline sparse_matrix_block_impl(SparseMatrixType& xpr, Index i) - : m_matrix(xpr), m_outerStart(convert_index(i)), m_outerSize(OuterSize) - {} - - inline sparse_matrix_block_impl(SparseMatrixType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) - : m_matrix(xpr), m_outerStart(convert_index(IsRowMajor ? startRow : startCol)), m_outerSize(convert_index(IsRowMajor ? blockRows : blockCols)) - {} - - template - inline BlockType& operator=(const SparseMatrixBase& other) - { - typedef typename internal::remove_all::type _NestedMatrixType; - _NestedMatrixType& matrix = m_matrix; - // This assignment is slow if this vector set is not empty - // and/or it is not at the end of the nonzeros of the underlying matrix. - - // 1 - eval to a temporary to avoid transposition and/or aliasing issues - Ref > tmp(other.derived()); - eigen_internal_assert(tmp.outerSize()==m_outerSize.value()); - - // 2 - let's check whether there is enough allocated memory - Index nnz = tmp.nonZeros(); - Index start = m_outerStart==0 ? 0 : m_matrix.outerIndexPtr()[m_outerStart]; // starting position of the current block - Index end = m_matrix.outerIndexPtr()[m_outerStart+m_outerSize.value()]; // ending position of the current block - Index block_size = end - start; // available room in the current block - Index tail_size = m_matrix.outerIndexPtr()[m_matrix.outerSize()] - end; - - Index free_size = m_matrix.isCompressed() - ? Index(matrix.data().allocatedSize()) + block_size - : block_size; - - Index tmp_start = tmp.outerIndexPtr()[0]; - - bool update_trailing_pointers = false; - if(nnz>free_size) - { - // realloc manually to reduce copies - typename SparseMatrixType::Storage newdata(m_matrix.data().allocatedSize() - block_size + nnz); - - internal::smart_copy(m_matrix.valuePtr(), m_matrix.valuePtr() + start, newdata.valuePtr()); - internal::smart_copy(m_matrix.innerIndexPtr(), m_matrix.innerIndexPtr() + start, newdata.indexPtr()); - - internal::smart_copy(tmp.valuePtr() + tmp_start, tmp.valuePtr() + tmp_start + nnz, newdata.valuePtr() + start); - internal::smart_copy(tmp.innerIndexPtr() + tmp_start, tmp.innerIndexPtr() + tmp_start + nnz, newdata.indexPtr() + start); - - internal::smart_copy(matrix.valuePtr()+end, matrix.valuePtr()+end + tail_size, newdata.valuePtr()+start+nnz); - internal::smart_copy(matrix.innerIndexPtr()+end, matrix.innerIndexPtr()+end + tail_size, newdata.indexPtr()+start+nnz); - - newdata.resize(m_matrix.outerIndexPtr()[m_matrix.outerSize()] - block_size + nnz); - - matrix.data().swap(newdata); - - update_trailing_pointers = true; - } - else - { - if(m_matrix.isCompressed()) - { - // no need to realloc, simply copy the tail at its respective position and insert tmp - matrix.data().resize(start + nnz + tail_size); - - internal::smart_memmove(matrix.valuePtr()+end, matrix.valuePtr() + end+tail_size, matrix.valuePtr() + start+nnz); - internal::smart_memmove(matrix.innerIndexPtr()+end, matrix.innerIndexPtr() + end+tail_size, matrix.innerIndexPtr() + start+nnz); - - update_trailing_pointers = true; - } - - internal::smart_copy(tmp.valuePtr() + tmp_start, tmp.valuePtr() + tmp_start + nnz, matrix.valuePtr() + start); - internal::smart_copy(tmp.innerIndexPtr() + tmp_start, tmp.innerIndexPtr() + tmp_start + nnz, matrix.innerIndexPtr() + start); - } - - // update outer index pointers and innerNonZeros - if(IsVectorAtCompileTime) - { - if(!m_matrix.isCompressed()) - matrix.innerNonZeroPtr()[m_outerStart] = StorageIndex(nnz); - matrix.outerIndexPtr()[m_outerStart] = StorageIndex(start); - } - else - { - StorageIndex p = StorageIndex(start); - for(Index k=0; k(tmp.innerVector(k).nonZeros()); - if(!m_matrix.isCompressed()) - matrix.innerNonZeroPtr()[m_outerStart+k] = nnz_k; - matrix.outerIndexPtr()[m_outerStart+k] = p; - p += nnz_k; - } - } - - if(update_trailing_pointers) - { - StorageIndex offset = internal::convert_index(nnz - block_size); - for(Index k = m_outerStart + m_outerSize.value(); k<=matrix.outerSize(); ++k) - { - matrix.outerIndexPtr()[k] += offset; - } - } - - return derived(); - } - - inline BlockType& operator=(const BlockType& other) - { - return operator=(other); - } - - inline const Scalar* valuePtr() const - { return m_matrix.valuePtr(); } - inline Scalar* valuePtr() - { return m_matrix.valuePtr(); } - - inline const StorageIndex* innerIndexPtr() const - { return m_matrix.innerIndexPtr(); } - inline StorageIndex* innerIndexPtr() - { return m_matrix.innerIndexPtr(); } - - inline const StorageIndex* outerIndexPtr() const - { return m_matrix.outerIndexPtr() + m_outerStart; } - inline StorageIndex* outerIndexPtr() - { return m_matrix.outerIndexPtr() + m_outerStart; } - - inline const StorageIndex* innerNonZeroPtr() const - { return isCompressed() ? 0 : (m_matrix.innerNonZeroPtr()+m_outerStart); } - inline StorageIndex* innerNonZeroPtr() - { return isCompressed() ? 0 : (m_matrix.innerNonZeroPtr()+m_outerStart); } - - bool isCompressed() const { return m_matrix.innerNonZeroPtr()==0; } - - inline Scalar& coeffRef(Index row, Index col) - { - return m_matrix.coeffRef(row + (IsRowMajor ? m_outerStart : 0), col + (IsRowMajor ? 0 : m_outerStart)); - } - - inline const Scalar coeff(Index row, Index col) const - { - return m_matrix.coeff(row + (IsRowMajor ? m_outerStart : 0), col + (IsRowMajor ? 0 : m_outerStart)); - } - - inline const Scalar coeff(Index index) const - { - return m_matrix.coeff(IsRowMajor ? m_outerStart : index, IsRowMajor ? index : m_outerStart); - } - - const Scalar& lastCoeff() const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(sparse_matrix_block_impl); - eigen_assert(Base::nonZeros()>0); - if(m_matrix.isCompressed()) - return m_matrix.valuePtr()[m_matrix.outerIndexPtr()[m_outerStart+1]-1]; - else - return m_matrix.valuePtr()[m_matrix.outerIndexPtr()[m_outerStart]+m_matrix.innerNonZeroPtr()[m_outerStart]-1]; - } - - EIGEN_STRONG_INLINE Index rows() const { return IsRowMajor ? m_outerSize.value() : m_matrix.rows(); } - EIGEN_STRONG_INLINE Index cols() const { return IsRowMajor ? m_matrix.cols() : m_outerSize.value(); } - - inline const SparseMatrixType& nestedExpression() const { return m_matrix; } - inline SparseMatrixType& nestedExpression() { return m_matrix; } - Index startRow() const { return IsRowMajor ? m_outerStart : 0; } - Index startCol() const { return IsRowMajor ? 0 : m_outerStart; } - Index blockRows() const { return IsRowMajor ? m_outerSize.value() : m_matrix.rows(); } - Index blockCols() const { return IsRowMajor ? m_matrix.cols() : m_outerSize.value(); } - - protected: - - typename internal::ref_selector::non_const_type m_matrix; - Index m_outerStart; - const internal::variable_if_dynamic m_outerSize; - -}; - -} // namespace internal - -template -class BlockImpl,BlockRows,BlockCols,true,Sparse> - : public internal::sparse_matrix_block_impl,BlockRows,BlockCols> -{ -public: - typedef _StorageIndex StorageIndex; - typedef SparseMatrix<_Scalar, _Options, _StorageIndex> SparseMatrixType; - typedef internal::sparse_matrix_block_impl Base; - inline BlockImpl(SparseMatrixType& xpr, Index i) - : Base(xpr, i) - {} - - inline BlockImpl(SparseMatrixType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) - : Base(xpr, startRow, startCol, blockRows, blockCols) - {} - - using Base::operator=; -}; - -template -class BlockImpl,BlockRows,BlockCols,true,Sparse> - : public internal::sparse_matrix_block_impl,BlockRows,BlockCols> -{ -public: - typedef _StorageIndex StorageIndex; - typedef const SparseMatrix<_Scalar, _Options, _StorageIndex> SparseMatrixType; - typedef internal::sparse_matrix_block_impl Base; - inline BlockImpl(SparseMatrixType& xpr, Index i) - : Base(xpr, i) - {} - - inline BlockImpl(SparseMatrixType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) - : Base(xpr, startRow, startCol, blockRows, blockCols) - {} - - using Base::operator=; -private: - template BlockImpl(const SparseMatrixBase& xpr, Index i); - template BlockImpl(const SparseMatrixBase& xpr); -}; - -//---------- - -/** \returns the \a outer -th column (resp. row) of the matrix \c *this if \c *this - * is col-major (resp. row-major). - */ -template -typename SparseMatrixBase::InnerVectorReturnType SparseMatrixBase::innerVector(Index outer) -{ return InnerVectorReturnType(derived(), outer); } - -/** \returns the \a outer -th column (resp. row) of the matrix \c *this if \c *this - * is col-major (resp. row-major). Read-only. - */ -template -const typename SparseMatrixBase::ConstInnerVectorReturnType SparseMatrixBase::innerVector(Index outer) const -{ return ConstInnerVectorReturnType(derived(), outer); } - -/** \returns the \a outer -th column (resp. row) of the matrix \c *this if \c *this - * is col-major (resp. row-major). - */ -template -typename SparseMatrixBase::InnerVectorsReturnType -SparseMatrixBase::innerVectors(Index outerStart, Index outerSize) -{ - return Block(derived(), - IsRowMajor ? outerStart : 0, IsRowMajor ? 0 : outerStart, - IsRowMajor ? outerSize : rows(), IsRowMajor ? cols() : outerSize); - -} - -/** \returns the \a outer -th column (resp. row) of the matrix \c *this if \c *this - * is col-major (resp. row-major). Read-only. - */ -template -const typename SparseMatrixBase::ConstInnerVectorsReturnType -SparseMatrixBase::innerVectors(Index outerStart, Index outerSize) const -{ - return Block(derived(), - IsRowMajor ? outerStart : 0, IsRowMajor ? 0 : outerStart, - IsRowMajor ? outerSize : rows(), IsRowMajor ? cols() : outerSize); - -} - -/** Generic implementation of sparse Block expression. - * Real-only. - */ -template -class BlockImpl - : public SparseMatrixBase >, internal::no_assignment_operator -{ - typedef Block BlockType; - typedef SparseMatrixBase Base; - using Base::convert_index; -public: - enum { IsRowMajor = internal::traits::IsRowMajor }; - EIGEN_SPARSE_PUBLIC_INTERFACE(BlockType) - - typedef typename internal::remove_all::type _MatrixTypeNested; - - /** Column or Row constructor - */ - inline BlockImpl(XprType& xpr, Index i) - : m_matrix(xpr), - m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? convert_index(i) : 0), - m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? convert_index(i) : 0), - m_blockRows(BlockRows==1 ? 1 : xpr.rows()), - m_blockCols(BlockCols==1 ? 1 : xpr.cols()) - {} - - /** Dynamic-size constructor - */ - inline BlockImpl(XprType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) - : m_matrix(xpr), m_startRow(convert_index(startRow)), m_startCol(convert_index(startCol)), m_blockRows(convert_index(blockRows)), m_blockCols(convert_index(blockCols)) - {} - - inline Index rows() const { return m_blockRows.value(); } - inline Index cols() const { return m_blockCols.value(); } - - inline Scalar& coeffRef(Index row, Index col) - { - return m_matrix.coeffRef(row + m_startRow.value(), col + m_startCol.value()); - } - - inline const Scalar coeff(Index row, Index col) const - { - return m_matrix.coeff(row + m_startRow.value(), col + m_startCol.value()); - } - - inline Scalar& coeffRef(Index index) - { - return m_matrix.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), - m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); - } - - inline const Scalar coeff(Index index) const - { - return m_matrix.coeff(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), - m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); - } - - inline const XprType& nestedExpression() const { return m_matrix; } - inline XprType& nestedExpression() { return m_matrix; } - Index startRow() const { return m_startRow.value(); } - Index startCol() const { return m_startCol.value(); } - Index blockRows() const { return m_blockRows.value(); } - Index blockCols() const { return m_blockCols.value(); } - - protected: -// friend class internal::GenericSparseBlockInnerIteratorImpl; - friend struct internal::unary_evaluator, internal::IteratorBased, Scalar >; - - Index nonZeros() const { return Dynamic; } - - typename internal::ref_selector::non_const_type m_matrix; - const internal::variable_if_dynamic m_startRow; - const internal::variable_if_dynamic m_startCol; - const internal::variable_if_dynamic m_blockRows; - const internal::variable_if_dynamic m_blockCols; - - protected: - // Disable assignment with clear error message. - // Note that simply removing operator= yields compilation errors with ICC+MSVC - template - BlockImpl& operator=(const T&) - { - EIGEN_STATIC_ASSERT(sizeof(T)==0, THIS_SPARSE_BLOCK_SUBEXPRESSION_IS_READ_ONLY); - return *this; - } - -}; - -namespace internal { - -template -struct unary_evaluator, IteratorBased > - : public evaluator_base > -{ - class InnerVectorInnerIterator; - class OuterVectorInnerIterator; - public: - typedef Block XprType; - typedef typename XprType::StorageIndex StorageIndex; - typedef typename XprType::Scalar Scalar; - - enum { - IsRowMajor = XprType::IsRowMajor, - - OuterVector = (BlockCols==1 && ArgType::IsRowMajor) - | // FIXME | instead of || to please GCC 4.4.0 stupid warning "suggest parentheses around &&". - // revert to || as soon as not needed anymore. - (BlockRows==1 && !ArgType::IsRowMajor), - - CoeffReadCost = evaluator::CoeffReadCost, - Flags = XprType::Flags - }; - - typedef typename internal::conditional::type InnerIterator; - - explicit unary_evaluator(const XprType& op) - : m_argImpl(op.nestedExpression()), m_block(op) - {} - - inline Index nonZerosEstimate() const { - Index nnz = m_block.nonZeros(); - if(nnz<0) - return m_argImpl.nonZerosEstimate() * m_block.size() / m_block.nestedExpression().size(); - return nnz; - } - - protected: - typedef typename evaluator::InnerIterator EvalIterator; - - evaluator m_argImpl; - const XprType &m_block; -}; - -template -class unary_evaluator, IteratorBased>::InnerVectorInnerIterator - : public EvalIterator -{ - enum { IsRowMajor = unary_evaluator::IsRowMajor }; - const XprType& m_block; - Index m_end; -public: - - EIGEN_STRONG_INLINE InnerVectorInnerIterator(const unary_evaluator& aEval, Index outer) - : EvalIterator(aEval.m_argImpl, outer + (IsRowMajor ? aEval.m_block.startRow() : aEval.m_block.startCol())), - m_block(aEval.m_block), - m_end(IsRowMajor ? aEval.m_block.startCol()+aEval.m_block.blockCols() : aEval.m_block.startRow()+aEval.m_block.blockRows()) - { - while( (EvalIterator::operator bool()) && (EvalIterator::index() < (IsRowMajor ? m_block.startCol() : m_block.startRow())) ) - EvalIterator::operator++(); - } - - inline StorageIndex index() const { return EvalIterator::index() - convert_index(IsRowMajor ? m_block.startCol() : m_block.startRow()); } - inline Index outer() const { return EvalIterator::outer() - (IsRowMajor ? m_block.startRow() : m_block.startCol()); } - inline Index row() const { return EvalIterator::row() - m_block.startRow(); } - inline Index col() const { return EvalIterator::col() - m_block.startCol(); } - - inline operator bool() const { return EvalIterator::operator bool() && EvalIterator::index() < m_end; } -}; - -template -class unary_evaluator, IteratorBased>::OuterVectorInnerIterator -{ - enum { IsRowMajor = unary_evaluator::IsRowMajor }; - const unary_evaluator& m_eval; - Index m_outerPos; - const Index m_innerIndex; - Index m_end; - EvalIterator m_it; -public: - - EIGEN_STRONG_INLINE OuterVectorInnerIterator(const unary_evaluator& aEval, Index outer) - : m_eval(aEval), - m_outerPos( (IsRowMajor ? aEval.m_block.startCol() : aEval.m_block.startRow()) ), - m_innerIndex(IsRowMajor ? aEval.m_block.startRow() : aEval.m_block.startCol()), - m_end(IsRowMajor ? aEval.m_block.startCol()+aEval.m_block.blockCols() : aEval.m_block.startRow()+aEval.m_block.blockRows()), - m_it(m_eval.m_argImpl, m_outerPos) - { - EIGEN_UNUSED_VARIABLE(outer); - eigen_assert(outer==0); - - while(m_it && m_it.index() < m_innerIndex) ++m_it; - if((!m_it) || (m_it.index()!=m_innerIndex)) - ++(*this); - } - - inline StorageIndex index() const { return convert_index(m_outerPos - (IsRowMajor ? m_eval.m_block.startCol() : m_eval.m_block.startRow())); } - inline Index outer() const { return 0; } - inline Index row() const { return IsRowMajor ? 0 : index(); } - inline Index col() const { return IsRowMajor ? index() : 0; } - - inline Scalar value() const { return m_it.value(); } - inline Scalar& valueRef() { return m_it.valueRef(); } - - inline OuterVectorInnerIterator& operator++() - { - // search next non-zero entry - while(++m_outerPos -struct unary_evaluator,BlockRows,BlockCols,true>, IteratorBased> - : evaluator,BlockRows,BlockCols,true> > > -{ - typedef Block,BlockRows,BlockCols,true> XprType; - typedef evaluator > Base; - explicit unary_evaluator(const XprType &xpr) : Base(xpr) {} -}; - -template -struct unary_evaluator,BlockRows,BlockCols,true>, IteratorBased> - : evaluator,BlockRows,BlockCols,true> > > -{ - typedef Block,BlockRows,BlockCols,true> XprType; - typedef evaluator > Base; - explicit unary_evaluator(const XprType &xpr) : Base(xpr) {} -}; - -} // end namespace internal - - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_BLOCK_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseColEtree.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseColEtree.h deleted file mode 100644 index ebe02d1ab02..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseColEtree.h +++ /dev/null @@ -1,206 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -/* - - * NOTE: This file is the modified version of sp_coletree.c file in SuperLU - - * -- SuperLU routine (version 3.1) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * August 1, 2008 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSE_COLETREE_H -#define SPARSE_COLETREE_H - -namespace Eigen { - -namespace internal { - -/** Find the root of the tree/set containing the vertex i : Use Path halving */ -template -Index etree_find (Index i, IndexVector& pp) -{ - Index p = pp(i); // Parent - Index gp = pp(p); // Grand parent - while (gp != p) - { - pp(i) = gp; // Parent pointer on find path is changed to former grand parent - i = gp; - p = pp(i); - gp = pp(p); - } - return p; -} - -/** Compute the column elimination tree of a sparse matrix - * \param mat The matrix in column-major format. - * \param parent The elimination tree - * \param firstRowElt The column index of the first element in each row - * \param perm The permutation to apply to the column of \b mat - */ -template -int coletree(const MatrixType& mat, IndexVector& parent, IndexVector& firstRowElt, typename MatrixType::StorageIndex *perm=0) -{ - typedef typename MatrixType::StorageIndex StorageIndex; - StorageIndex nc = convert_index(mat.cols()); // Number of columns - StorageIndex m = convert_index(mat.rows()); - StorageIndex diagSize = (std::min)(nc,m); - IndexVector root(nc); // root of subtree of etree - root.setZero(); - IndexVector pp(nc); // disjoint sets - pp.setZero(); // Initialize disjoint sets - parent.resize(mat.cols()); - //Compute first nonzero column in each row - firstRowElt.resize(m); - firstRowElt.setConstant(nc); - firstRowElt.segment(0, diagSize).setLinSpaced(diagSize, 0, diagSize-1); - bool found_diag; - for (StorageIndex col = 0; col < nc; col++) - { - StorageIndex pcol = col; - if(perm) pcol = perm[col]; - for (typename MatrixType::InnerIterator it(mat, pcol); it; ++it) - { - Index row = it.row(); - firstRowElt(row) = (std::min)(firstRowElt(row), col); - } - } - /* Compute etree by Liu's algorithm for symmetric matrices, - except use (firstRowElt[r],c) in place of an edge (r,c) of A. - Thus each row clique in A'*A is replaced by a star - centered at its first vertex, which has the same fill. */ - StorageIndex rset, cset, rroot; - for (StorageIndex col = 0; col < nc; col++) - { - found_diag = col>=m; - pp(col) = col; - cset = col; - root(cset) = col; - parent(col) = nc; - /* The diagonal element is treated here even if it does not exist in the matrix - * hence the loop is executed once more */ - StorageIndex pcol = col; - if(perm) pcol = perm[col]; - for (typename MatrixType::InnerIterator it(mat, pcol); it||!found_diag; ++it) - { // A sequence of interleaved find and union is performed - Index i = col; - if(it) i = it.index(); - if (i == col) found_diag = true; - - StorageIndex row = firstRowElt(i); - if (row >= col) continue; - rset = internal::etree_find(row, pp); // Find the name of the set containing row - rroot = root(rset); - if (rroot != col) - { - parent(rroot) = col; - pp(cset) = rset; - cset = rset; - root(cset) = col; - } - } - } - return 0; -} - -/** - * Depth-first search from vertex n. No recursion. - * This routine was contributed by Cédric Doucet, CEDRAT Group, Meylan, France. -*/ -template -void nr_etdfs (typename IndexVector::Scalar n, IndexVector& parent, IndexVector& first_kid, IndexVector& next_kid, IndexVector& post, typename IndexVector::Scalar postnum) -{ - typedef typename IndexVector::Scalar StorageIndex; - StorageIndex current = n, first, next; - while (postnum != n) - { - // No kid for the current node - first = first_kid(current); - - // no kid for the current node - if (first == -1) - { - // Numbering this node because it has no kid - post(current) = postnum++; - - // looking for the next kid - next = next_kid(current); - while (next == -1) - { - // No more kids : back to the parent node - current = parent(current); - // numbering the parent node - post(current) = postnum++; - - // Get the next kid - next = next_kid(current); - } - // stopping criterion - if (postnum == n+1) return; - - // Updating current node - current = next; - } - else - { - current = first; - } - } -} - - -/** - * \brief Post order a tree - * \param n the number of nodes - * \param parent Input tree - * \param post postordered tree - */ -template -void treePostorder(typename IndexVector::Scalar n, IndexVector& parent, IndexVector& post) -{ - typedef typename IndexVector::Scalar StorageIndex; - IndexVector first_kid, next_kid; // Linked list of children - StorageIndex postnum; - // Allocate storage for working arrays and results - first_kid.resize(n+1); - next_kid.setZero(n+1); - post.setZero(n+1); - - // Set up structure describing children - first_kid.setConstant(-1); - for (StorageIndex v = n-1; v >= 0; v--) - { - StorageIndex dad = parent(v); - next_kid(v) = first_kid(dad); - first_kid(dad) = v; - } - - // Depth-first search from dummy root vertex #n - postnum = 0; - internal::nr_etdfs(n, parent, first_kid, next_kid, post, postnum); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // SPARSE_COLETREE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCompressedBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCompressedBase.h deleted file mode 100644 index 5ccb4665614..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCompressedBase.h +++ /dev/null @@ -1,341 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_COMPRESSED_BASE_H -#define EIGEN_SPARSE_COMPRESSED_BASE_H - -namespace Eigen { - -template class SparseCompressedBase; - -namespace internal { - -template -struct traits > : traits -{}; - -} // end namespace internal - -/** \ingroup SparseCore_Module - * \class SparseCompressedBase - * \brief Common base class for sparse [compressed]-{row|column}-storage format. - * - * This class defines the common interface for all derived classes implementing the compressed sparse storage format, such as: - * - SparseMatrix - * - Ref - * - Map - * - */ -template -class SparseCompressedBase - : public SparseMatrixBase -{ - public: - typedef SparseMatrixBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(SparseCompressedBase) - using Base::operator=; - using Base::IsRowMajor; - - class InnerIterator; - class ReverseInnerIterator; - - protected: - typedef typename Base::IndexVector IndexVector; - Eigen::Map innerNonZeros() { return Eigen::Map(innerNonZeroPtr(), isCompressed()?0:derived().outerSize()); } - const Eigen::Map innerNonZeros() const { return Eigen::Map(innerNonZeroPtr(), isCompressed()?0:derived().outerSize()); } - - public: - - /** \returns the number of non zero coefficients */ - inline Index nonZeros() const - { - if(Derived::IsVectorAtCompileTime && outerIndexPtr()==0) - return derived().nonZeros(); - else if(isCompressed()) - return outerIndexPtr()[derived().outerSize()]-outerIndexPtr()[0]; - else if(derived().outerSize()==0) - return 0; - else - return innerNonZeros().sum(); - } - - /** \returns a const pointer to the array of values. - * This function is aimed at interoperability with other libraries. - * \sa innerIndexPtr(), outerIndexPtr() */ - inline const Scalar* valuePtr() const { return derived().valuePtr(); } - /** \returns a non-const pointer to the array of values. - * This function is aimed at interoperability with other libraries. - * \sa innerIndexPtr(), outerIndexPtr() */ - inline Scalar* valuePtr() { return derived().valuePtr(); } - - /** \returns a const pointer to the array of inner indices. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), outerIndexPtr() */ - inline const StorageIndex* innerIndexPtr() const { return derived().innerIndexPtr(); } - /** \returns a non-const pointer to the array of inner indices. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), outerIndexPtr() */ - inline StorageIndex* innerIndexPtr() { return derived().innerIndexPtr(); } - - /** \returns a const pointer to the array of the starting positions of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 for SparseVector - * \sa valuePtr(), innerIndexPtr() */ - inline const StorageIndex* outerIndexPtr() const { return derived().outerIndexPtr(); } - /** \returns a non-const pointer to the array of the starting positions of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 for SparseVector - * \sa valuePtr(), innerIndexPtr() */ - inline StorageIndex* outerIndexPtr() { return derived().outerIndexPtr(); } - - /** \returns a const pointer to the array of the number of non zeros of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 in compressed mode */ - inline const StorageIndex* innerNonZeroPtr() const { return derived().innerNonZeroPtr(); } - /** \returns a non-const pointer to the array of the number of non zeros of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 in compressed mode */ - inline StorageIndex* innerNonZeroPtr() { return derived().innerNonZeroPtr(); } - - /** \returns whether \c *this is in compressed form. */ - inline bool isCompressed() const { return innerNonZeroPtr()==0; } - - /** \returns a read-only view of the stored coefficients as a 1D array expression. - * - * \warning this method is for \b compressed \b storage \b only, and it will trigger an assertion otherwise. - * - * \sa valuePtr(), isCompressed() */ - const Map > coeffs() const { eigen_assert(isCompressed()); return Array::Map(valuePtr(),nonZeros()); } - - /** \returns a read-write view of the stored coefficients as a 1D array expression - * - * \warning this method is for \b compressed \b storage \b only, and it will trigger an assertion otherwise. - * - * Here is an example: - * \include SparseMatrix_coeffs.cpp - * and the output is: - * \include SparseMatrix_coeffs.out - * - * \sa valuePtr(), isCompressed() */ - Map > coeffs() { eigen_assert(isCompressed()); return Array::Map(valuePtr(),nonZeros()); } - - protected: - /** Default constructor. Do nothing. */ - SparseCompressedBase() {} - private: - template explicit SparseCompressedBase(const SparseCompressedBase&); -}; - -template -class SparseCompressedBase::InnerIterator -{ - public: - InnerIterator() - : m_values(0), m_indices(0), m_outer(0), m_id(0), m_end(0) - {} - - InnerIterator(const InnerIterator& other) - : m_values(other.m_values), m_indices(other.m_indices), m_outer(other.m_outer), m_id(other.m_id), m_end(other.m_end) - {} - - InnerIterator& operator=(const InnerIterator& other) - { - m_values = other.m_values; - m_indices = other.m_indices; - const_cast(m_outer).setValue(other.m_outer.value()); - m_id = other.m_id; - m_end = other.m_end; - return *this; - } - - InnerIterator(const SparseCompressedBase& mat, Index outer) - : m_values(mat.valuePtr()), m_indices(mat.innerIndexPtr()), m_outer(outer) - { - if(Derived::IsVectorAtCompileTime && mat.outerIndexPtr()==0) - { - m_id = 0; - m_end = mat.nonZeros(); - } - else - { - m_id = mat.outerIndexPtr()[outer]; - if(mat.isCompressed()) - m_end = mat.outerIndexPtr()[outer+1]; - else - m_end = m_id + mat.innerNonZeroPtr()[outer]; - } - } - - explicit InnerIterator(const SparseCompressedBase& mat) - : m_values(mat.valuePtr()), m_indices(mat.innerIndexPtr()), m_outer(0), m_id(0), m_end(mat.nonZeros()) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - } - - explicit InnerIterator(const internal::CompressedStorage& data) - : m_values(data.valuePtr()), m_indices(data.indexPtr()), m_outer(0), m_id(0), m_end(data.size()) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - } - - inline InnerIterator& operator++() { m_id++; return *this; } - - inline const Scalar& value() const { return m_values[m_id]; } - inline Scalar& valueRef() { return const_cast(m_values[m_id]); } - - inline StorageIndex index() const { return m_indices[m_id]; } - inline Index outer() const { return m_outer.value(); } - inline Index row() const { return IsRowMajor ? m_outer.value() : index(); } - inline Index col() const { return IsRowMajor ? index() : m_outer.value(); } - - inline operator bool() const { return (m_id < m_end); } - - protected: - const Scalar* m_values; - const StorageIndex* m_indices; - typedef internal::variable_if_dynamic OuterType; - const OuterType m_outer; - Index m_id; - Index m_end; - private: - // If you get here, then you're not using the right InnerIterator type, e.g.: - // SparseMatrix A; - // SparseMatrix::InnerIterator it(A,0); - template InnerIterator(const SparseMatrixBase&, Index outer); -}; - -template -class SparseCompressedBase::ReverseInnerIterator -{ - public: - ReverseInnerIterator(const SparseCompressedBase& mat, Index outer) - : m_values(mat.valuePtr()), m_indices(mat.innerIndexPtr()), m_outer(outer) - { - if(Derived::IsVectorAtCompileTime && mat.outerIndexPtr()==0) - { - m_start = 0; - m_id = mat.nonZeros(); - } - else - { - m_start = mat.outerIndexPtr()[outer]; - if(mat.isCompressed()) - m_id = mat.outerIndexPtr()[outer+1]; - else - m_id = m_start + mat.innerNonZeroPtr()[outer]; - } - } - - explicit ReverseInnerIterator(const SparseCompressedBase& mat) - : m_values(mat.valuePtr()), m_indices(mat.innerIndexPtr()), m_outer(0), m_start(0), m_id(mat.nonZeros()) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - } - - explicit ReverseInnerIterator(const internal::CompressedStorage& data) - : m_values(data.valuePtr()), m_indices(data.indexPtr()), m_outer(0), m_start(0), m_id(data.size()) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived); - } - - inline ReverseInnerIterator& operator--() { --m_id; return *this; } - - inline const Scalar& value() const { return m_values[m_id-1]; } - inline Scalar& valueRef() { return const_cast(m_values[m_id-1]); } - - inline StorageIndex index() const { return m_indices[m_id-1]; } - inline Index outer() const { return m_outer.value(); } - inline Index row() const { return IsRowMajor ? m_outer.value() : index(); } - inline Index col() const { return IsRowMajor ? index() : m_outer.value(); } - - inline operator bool() const { return (m_id > m_start); } - - protected: - const Scalar* m_values; - const StorageIndex* m_indices; - typedef internal::variable_if_dynamic OuterType; - const OuterType m_outer; - Index m_start; - Index m_id; -}; - -namespace internal { - -template -struct evaluator > - : evaluator_base -{ - typedef typename Derived::Scalar Scalar; - typedef typename Derived::InnerIterator InnerIterator; - - enum { - CoeffReadCost = NumTraits::ReadCost, - Flags = Derived::Flags - }; - - evaluator() : m_matrix(0), m_zero(0) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - explicit evaluator(const Derived &mat) : m_matrix(&mat), m_zero(0) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_matrix->nonZeros(); - } - - operator Derived&() { return m_matrix->const_cast_derived(); } - operator const Derived&() const { return *m_matrix; } - - typedef typename DenseCoeffsBase::CoeffReturnType CoeffReturnType; - const Scalar& coeff(Index row, Index col) const - { - Index p = find(row,col); - - if(p==Dynamic) - return m_zero; - else - return m_matrix->const_cast_derived().valuePtr()[p]; - } - - Scalar& coeffRef(Index row, Index col) - { - Index p = find(row,col); - eigen_assert(p!=Dynamic && "written coefficient does not exist"); - return m_matrix->const_cast_derived().valuePtr()[p]; - } - -protected: - - Index find(Index row, Index col) const - { - eigen_internal_assert(row>=0 && rowrows() && col>=0 && colcols()); - - const Index outer = Derived::IsRowMajor ? row : col; - const Index inner = Derived::IsRowMajor ? col : row; - - Index start = m_matrix->outerIndexPtr()[outer]; - Index end = m_matrix->isCompressed() ? m_matrix->outerIndexPtr()[outer+1] : m_matrix->outerIndexPtr()[outer] + m_matrix->innerNonZeroPtr()[outer]; - eigen_assert(end>=start && "you are using a non finalized sparse matrix or written coefficient does not exist"); - const Index p = std::lower_bound(m_matrix->innerIndexPtr()+start, m_matrix->innerIndexPtr()+end,inner) - m_matrix->innerIndexPtr(); - - return ((pinnerIndexPtr()[p]==inner)) ? p : Dynamic; - } - - const Derived *m_matrix; - const Scalar m_zero; -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_COMPRESSED_BASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseBinaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseBinaryOp.h deleted file mode 100644 index e315e355060..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseBinaryOp.h +++ /dev/null @@ -1,726 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_CWISE_BINARY_OP_H -#define EIGEN_SPARSE_CWISE_BINARY_OP_H - -namespace Eigen { - -// Here we have to handle 3 cases: -// 1 - sparse op dense -// 2 - dense op sparse -// 3 - sparse op sparse -// We also need to implement a 4th iterator for: -// 4 - dense op dense -// Finally, we also need to distinguish between the product and other operations : -// configuration returned mode -// 1 - sparse op dense product sparse -// generic dense -// 2 - dense op sparse product sparse -// generic dense -// 3 - sparse op sparse product sparse -// generic sparse -// 4 - dense op dense product dense -// generic dense -// -// TODO to ease compiler job, we could specialize product/quotient with a scalar -// and fallback to cwise-unary evaluator using bind1st_op and bind2nd_op. - -template -class CwiseBinaryOpImpl - : public SparseMatrixBase > -{ - public: - typedef CwiseBinaryOp Derived; - typedef SparseMatrixBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Derived) - CwiseBinaryOpImpl() - { - EIGEN_STATIC_ASSERT(( - (!internal::is_same::StorageKind, - typename internal::traits::StorageKind>::value) - || ((internal::evaluator::Flags&RowMajorBit) == (internal::evaluator::Flags&RowMajorBit))), - THE_STORAGE_ORDER_OF_BOTH_SIDES_MUST_MATCH); - } -}; - -namespace internal { - - -// Generic "sparse OP sparse" -template struct binary_sparse_evaluator; - -template -struct binary_evaluator, IteratorBased, IteratorBased> - : evaluator_base > -{ -protected: - typedef typename evaluator::InnerIterator LhsIterator; - typedef typename evaluator::InnerIterator RhsIterator; - typedef CwiseBinaryOp XprType; - typedef typename traits::Scalar Scalar; - typedef typename XprType::StorageIndex StorageIndex; -public: - - class InnerIterator - { - public: - - EIGEN_STRONG_INLINE InnerIterator(const binary_evaluator& aEval, Index outer) - : m_lhsIter(aEval.m_lhsImpl,outer), m_rhsIter(aEval.m_rhsImpl,outer), m_functor(aEval.m_functor) - { - this->operator++(); - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - if (m_lhsIter && m_rhsIter && (m_lhsIter.index() == m_rhsIter.index())) - { - m_id = m_lhsIter.index(); - m_value = m_functor(m_lhsIter.value(), m_rhsIter.value()); - ++m_lhsIter; - ++m_rhsIter; - } - else if (m_lhsIter && (!m_rhsIter || (m_lhsIter.index() < m_rhsIter.index()))) - { - m_id = m_lhsIter.index(); - m_value = m_functor(m_lhsIter.value(), Scalar(0)); - ++m_lhsIter; - } - else if (m_rhsIter && (!m_lhsIter || (m_lhsIter.index() > m_rhsIter.index()))) - { - m_id = m_rhsIter.index(); - m_value = m_functor(Scalar(0), m_rhsIter.value()); - ++m_rhsIter; - } - else - { - m_value = 0; // this is to avoid a compilation warning - m_id = -1; - } - return *this; - } - - EIGEN_STRONG_INLINE Scalar value() const { return m_value; } - - EIGEN_STRONG_INLINE StorageIndex index() const { return m_id; } - EIGEN_STRONG_INLINE Index outer() const { return m_lhsIter.outer(); } - EIGEN_STRONG_INLINE Index row() const { return Lhs::IsRowMajor ? m_lhsIter.row() : index(); } - EIGEN_STRONG_INLINE Index col() const { return Lhs::IsRowMajor ? index() : m_lhsIter.col(); } - - EIGEN_STRONG_INLINE operator bool() const { return m_id>=0; } - - protected: - LhsIterator m_lhsIter; - RhsIterator m_rhsIter; - const BinaryOp& m_functor; - Scalar m_value; - StorageIndex m_id; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - Flags = XprType::Flags - }; - - explicit binary_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_lhsImpl.nonZerosEstimate() + m_rhsImpl.nonZerosEstimate(); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; -}; - -// dense op sparse -template -struct binary_evaluator, IndexBased, IteratorBased> - : evaluator_base > -{ -protected: - typedef typename evaluator::InnerIterator RhsIterator; - typedef CwiseBinaryOp XprType; - typedef typename traits::Scalar Scalar; - typedef typename XprType::StorageIndex StorageIndex; -public: - - class InnerIterator - { - enum { IsRowMajor = (int(Rhs::Flags)&RowMajorBit)==RowMajorBit }; - public: - - EIGEN_STRONG_INLINE InnerIterator(const binary_evaluator& aEval, Index outer) - : m_lhsEval(aEval.m_lhsImpl), m_rhsIter(aEval.m_rhsImpl,outer), m_functor(aEval.m_functor), m_value(0), m_id(-1), m_innerSize(aEval.m_expr.rhs().innerSize()) - { - this->operator++(); - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - ++m_id; - if(m_id &m_lhsEval; - RhsIterator m_rhsIter; - const BinaryOp& m_functor; - Scalar m_value; - StorageIndex m_id; - StorageIndex m_innerSize; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - // Expose storage order of the sparse expression - Flags = (XprType::Flags & ~RowMajorBit) | (int(Rhs::Flags)&RowMajorBit) - }; - - explicit binary_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()), - m_expr(xpr) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_expr.size(); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; - const XprType &m_expr; -}; - -// sparse op dense -template -struct binary_evaluator, IteratorBased, IndexBased> - : evaluator_base > -{ -protected: - typedef typename evaluator::InnerIterator LhsIterator; - typedef CwiseBinaryOp XprType; - typedef typename traits::Scalar Scalar; - typedef typename XprType::StorageIndex StorageIndex; -public: - - class InnerIterator - { - enum { IsRowMajor = (int(Lhs::Flags)&RowMajorBit)==RowMajorBit }; - public: - - EIGEN_STRONG_INLINE InnerIterator(const binary_evaluator& aEval, Index outer) - : m_lhsIter(aEval.m_lhsImpl,outer), m_rhsEval(aEval.m_rhsImpl), m_functor(aEval.m_functor), m_value(0), m_id(-1), m_innerSize(aEval.m_expr.lhs().innerSize()) - { - this->operator++(); - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - ++m_id; - if(m_id &m_rhsEval; - const BinaryOp& m_functor; - Scalar m_value; - StorageIndex m_id; - StorageIndex m_innerSize; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - // Expose storage order of the sparse expression - Flags = (XprType::Flags & ~RowMajorBit) | (int(Lhs::Flags)&RowMajorBit) - }; - - explicit binary_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()), - m_expr(xpr) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_expr.size(); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; - const XprType &m_expr; -}; - -template::Kind, - typename RhsKind = typename evaluator_traits::Kind, - typename LhsScalar = typename traits::Scalar, - typename RhsScalar = typename traits::Scalar> struct sparse_conjunction_evaluator; - -// "sparse .* sparse" -template -struct binary_evaluator, Lhs, Rhs>, IteratorBased, IteratorBased> - : sparse_conjunction_evaluator, Lhs, Rhs> > -{ - typedef CwiseBinaryOp, Lhs, Rhs> XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; -// "dense .* sparse" -template -struct binary_evaluator, Lhs, Rhs>, IndexBased, IteratorBased> - : sparse_conjunction_evaluator, Lhs, Rhs> > -{ - typedef CwiseBinaryOp, Lhs, Rhs> XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; -// "sparse .* dense" -template -struct binary_evaluator, Lhs, Rhs>, IteratorBased, IndexBased> - : sparse_conjunction_evaluator, Lhs, Rhs> > -{ - typedef CwiseBinaryOp, Lhs, Rhs> XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; - -// "sparse ./ dense" -template -struct binary_evaluator, Lhs, Rhs>, IteratorBased, IndexBased> - : sparse_conjunction_evaluator, Lhs, Rhs> > -{ - typedef CwiseBinaryOp, Lhs, Rhs> XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; - -// "sparse && sparse" -template -struct binary_evaluator, IteratorBased, IteratorBased> - : sparse_conjunction_evaluator > -{ - typedef CwiseBinaryOp XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; -// "dense && sparse" -template -struct binary_evaluator, IndexBased, IteratorBased> - : sparse_conjunction_evaluator > -{ - typedef CwiseBinaryOp XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; -// "sparse && dense" -template -struct binary_evaluator, IteratorBased, IndexBased> - : sparse_conjunction_evaluator > -{ - typedef CwiseBinaryOp XprType; - typedef sparse_conjunction_evaluator Base; - explicit binary_evaluator(const XprType& xpr) : Base(xpr) {} -}; - -// "sparse ^ sparse" -template -struct sparse_conjunction_evaluator - : evaluator_base -{ -protected: - typedef typename XprType::Functor BinaryOp; - typedef typename XprType::Lhs LhsArg; - typedef typename XprType::Rhs RhsArg; - typedef typename evaluator::InnerIterator LhsIterator; - typedef typename evaluator::InnerIterator RhsIterator; - typedef typename XprType::StorageIndex StorageIndex; - typedef typename traits::Scalar Scalar; -public: - - class InnerIterator - { - public: - - EIGEN_STRONG_INLINE InnerIterator(const sparse_conjunction_evaluator& aEval, Index outer) - : m_lhsIter(aEval.m_lhsImpl,outer), m_rhsIter(aEval.m_rhsImpl,outer), m_functor(aEval.m_functor) - { - while (m_lhsIter && m_rhsIter && (m_lhsIter.index() != m_rhsIter.index())) - { - if (m_lhsIter.index() < m_rhsIter.index()) - ++m_lhsIter; - else - ++m_rhsIter; - } - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - ++m_lhsIter; - ++m_rhsIter; - while (m_lhsIter && m_rhsIter && (m_lhsIter.index() != m_rhsIter.index())) - { - if (m_lhsIter.index() < m_rhsIter.index()) - ++m_lhsIter; - else - ++m_rhsIter; - } - return *this; - } - - EIGEN_STRONG_INLINE Scalar value() const { return m_functor(m_lhsIter.value(), m_rhsIter.value()); } - - EIGEN_STRONG_INLINE StorageIndex index() const { return m_lhsIter.index(); } - EIGEN_STRONG_INLINE Index outer() const { return m_lhsIter.outer(); } - EIGEN_STRONG_INLINE Index row() const { return m_lhsIter.row(); } - EIGEN_STRONG_INLINE Index col() const { return m_lhsIter.col(); } - - EIGEN_STRONG_INLINE operator bool() const { return (m_lhsIter && m_rhsIter); } - - protected: - LhsIterator m_lhsIter; - RhsIterator m_rhsIter; - const BinaryOp& m_functor; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - Flags = XprType::Flags - }; - - explicit sparse_conjunction_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return (std::min)(m_lhsImpl.nonZerosEstimate(), m_rhsImpl.nonZerosEstimate()); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; -}; - -// "dense ^ sparse" -template -struct sparse_conjunction_evaluator - : evaluator_base -{ -protected: - typedef typename XprType::Functor BinaryOp; - typedef typename XprType::Lhs LhsArg; - typedef typename XprType::Rhs RhsArg; - typedef evaluator LhsEvaluator; - typedef typename evaluator::InnerIterator RhsIterator; - typedef typename XprType::StorageIndex StorageIndex; - typedef typename traits::Scalar Scalar; -public: - - class InnerIterator - { - enum { IsRowMajor = (int(RhsArg::Flags)&RowMajorBit)==RowMajorBit }; - - public: - - EIGEN_STRONG_INLINE InnerIterator(const sparse_conjunction_evaluator& aEval, Index outer) - : m_lhsEval(aEval.m_lhsImpl), m_rhsIter(aEval.m_rhsImpl,outer), m_functor(aEval.m_functor), m_outer(outer) - {} - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - ++m_rhsIter; - return *this; - } - - EIGEN_STRONG_INLINE Scalar value() const - { return m_functor(m_lhsEval.coeff(IsRowMajor?m_outer:m_rhsIter.index(),IsRowMajor?m_rhsIter.index():m_outer), m_rhsIter.value()); } - - EIGEN_STRONG_INLINE StorageIndex index() const { return m_rhsIter.index(); } - EIGEN_STRONG_INLINE Index outer() const { return m_rhsIter.outer(); } - EIGEN_STRONG_INLINE Index row() const { return m_rhsIter.row(); } - EIGEN_STRONG_INLINE Index col() const { return m_rhsIter.col(); } - - EIGEN_STRONG_INLINE operator bool() const { return m_rhsIter; } - - protected: - const LhsEvaluator &m_lhsEval; - RhsIterator m_rhsIter; - const BinaryOp& m_functor; - const Index m_outer; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - // Expose storage order of the sparse expression - Flags = (XprType::Flags & ~RowMajorBit) | (int(RhsArg::Flags)&RowMajorBit) - }; - - explicit sparse_conjunction_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_rhsImpl.nonZerosEstimate(); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; -}; - -// "sparse ^ dense" -template -struct sparse_conjunction_evaluator - : evaluator_base -{ -protected: - typedef typename XprType::Functor BinaryOp; - typedef typename XprType::Lhs LhsArg; - typedef typename XprType::Rhs RhsArg; - typedef typename evaluator::InnerIterator LhsIterator; - typedef evaluator RhsEvaluator; - typedef typename XprType::StorageIndex StorageIndex; - typedef typename traits::Scalar Scalar; -public: - - class InnerIterator - { - enum { IsRowMajor = (int(LhsArg::Flags)&RowMajorBit)==RowMajorBit }; - - public: - - EIGEN_STRONG_INLINE InnerIterator(const sparse_conjunction_evaluator& aEval, Index outer) - : m_lhsIter(aEval.m_lhsImpl,outer), m_rhsEval(aEval.m_rhsImpl), m_functor(aEval.m_functor), m_outer(outer) - {} - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - ++m_lhsIter; - return *this; - } - - EIGEN_STRONG_INLINE Scalar value() const - { return m_functor(m_lhsIter.value(), - m_rhsEval.coeff(IsRowMajor?m_outer:m_lhsIter.index(),IsRowMajor?m_lhsIter.index():m_outer)); } - - EIGEN_STRONG_INLINE StorageIndex index() const { return m_lhsIter.index(); } - EIGEN_STRONG_INLINE Index outer() const { return m_lhsIter.outer(); } - EIGEN_STRONG_INLINE Index row() const { return m_lhsIter.row(); } - EIGEN_STRONG_INLINE Index col() const { return m_lhsIter.col(); } - - EIGEN_STRONG_INLINE operator bool() const { return m_lhsIter; } - - protected: - LhsIterator m_lhsIter; - const evaluator &m_rhsEval; - const BinaryOp& m_functor; - const Index m_outer; - }; - - - enum { - CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, - // Expose storage order of the sparse expression - Flags = (XprType::Flags & ~RowMajorBit) | (int(LhsArg::Flags)&RowMajorBit) - }; - - explicit sparse_conjunction_evaluator(const XprType& xpr) - : m_functor(xpr.functor()), - m_lhsImpl(xpr.lhs()), - m_rhsImpl(xpr.rhs()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_lhsImpl.nonZerosEstimate(); - } - -protected: - const BinaryOp m_functor; - evaluator m_lhsImpl; - evaluator m_rhsImpl; -}; - -} - -/*************************************************************************** -* Implementation of SparseMatrixBase and SparseCwise functions/operators -***************************************************************************/ - -template -template -Derived& SparseMatrixBase::operator+=(const EigenBase &other) -{ - call_assignment(derived(), other.derived(), internal::add_assign_op()); - return derived(); -} - -template -template -Derived& SparseMatrixBase::operator-=(const EigenBase &other) -{ - call_assignment(derived(), other.derived(), internal::assign_op()); - return derived(); -} - -template -template -EIGEN_STRONG_INLINE Derived & -SparseMatrixBase::operator-=(const SparseMatrixBase &other) -{ - return derived() = derived() - other.derived(); -} - -template -template -EIGEN_STRONG_INLINE Derived & -SparseMatrixBase::operator+=(const SparseMatrixBase& other) -{ - return derived() = derived() + other.derived(); -} - -template -template -Derived& SparseMatrixBase::operator+=(const DiagonalBase& other) -{ - call_assignment_no_alias(derived(), other.derived(), internal::add_assign_op()); - return derived(); -} - -template -template -Derived& SparseMatrixBase::operator-=(const DiagonalBase& other) -{ - call_assignment_no_alias(derived(), other.derived(), internal::sub_assign_op()); - return derived(); -} - -template -template -EIGEN_STRONG_INLINE const typename SparseMatrixBase::template CwiseProductDenseReturnType::Type -SparseMatrixBase::cwiseProduct(const MatrixBase &other) const -{ - return typename CwiseProductDenseReturnType::Type(derived(), other.derived()); -} - -template -EIGEN_STRONG_INLINE const CwiseBinaryOp, const DenseDerived, const SparseDerived> -operator+(const MatrixBase &a, const SparseMatrixBase &b) -{ - return CwiseBinaryOp, const DenseDerived, const SparseDerived>(a.derived(), b.derived()); -} - -template -EIGEN_STRONG_INLINE const CwiseBinaryOp, const SparseDerived, const DenseDerived> -operator+(const SparseMatrixBase &a, const MatrixBase &b) -{ - return CwiseBinaryOp, const SparseDerived, const DenseDerived>(a.derived(), b.derived()); -} - -template -EIGEN_STRONG_INLINE const CwiseBinaryOp, const DenseDerived, const SparseDerived> -operator-(const MatrixBase &a, const SparseMatrixBase &b) -{ - return CwiseBinaryOp, const DenseDerived, const SparseDerived>(a.derived(), b.derived()); -} - -template -EIGEN_STRONG_INLINE const CwiseBinaryOp, const SparseDerived, const DenseDerived> -operator-(const SparseMatrixBase &a, const MatrixBase &b) -{ - return CwiseBinaryOp, const SparseDerived, const DenseDerived>(a.derived(), b.derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_CWISE_BINARY_OP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseUnaryOp.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseUnaryOp.h deleted file mode 100644 index df6c28d2b89..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseCwiseUnaryOp.h +++ /dev/null @@ -1,150 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_CWISE_UNARY_OP_H -#define EIGEN_SPARSE_CWISE_UNARY_OP_H - -namespace Eigen { - -namespace internal { - -template -struct unary_evaluator, IteratorBased> - : public evaluator_base > -{ - public: - typedef CwiseUnaryOp XprType; - - class InnerIterator; - - enum { - CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType& op) : m_functor(op.functor()), m_argImpl(op.nestedExpression()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_argImpl.nonZerosEstimate(); - } - - protected: - typedef typename evaluator::InnerIterator EvalIterator; - - const UnaryOp m_functor; - evaluator m_argImpl; -}; - -template -class unary_evaluator, IteratorBased>::InnerIterator - : public unary_evaluator, IteratorBased>::EvalIterator -{ - protected: - typedef typename XprType::Scalar Scalar; - typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; - public: - - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& unaryOp, Index outer) - : Base(unaryOp.m_argImpl,outer), m_functor(unaryOp.m_functor) - {} - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { Base::operator++(); return *this; } - - EIGEN_STRONG_INLINE Scalar value() const { return m_functor(Base::value()); } - - protected: - const UnaryOp m_functor; - private: - Scalar& valueRef(); -}; - -template -struct unary_evaluator, IteratorBased> - : public evaluator_base > -{ - public: - typedef CwiseUnaryView XprType; - - class InnerIterator; - - enum { - CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType& op) : m_functor(op.functor()), m_argImpl(op.nestedExpression()) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - protected: - typedef typename evaluator::InnerIterator EvalIterator; - - const ViewOp m_functor; - evaluator m_argImpl; -}; - -template -class unary_evaluator, IteratorBased>::InnerIterator - : public unary_evaluator, IteratorBased>::EvalIterator -{ - protected: - typedef typename XprType::Scalar Scalar; - typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; - public: - - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& unaryOp, Index outer) - : Base(unaryOp.m_argImpl,outer), m_functor(unaryOp.m_functor) - {} - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { Base::operator++(); return *this; } - - EIGEN_STRONG_INLINE Scalar value() const { return m_functor(Base::value()); } - EIGEN_STRONG_INLINE Scalar& valueRef() { return m_functor(Base::valueRef()); } - - protected: - const ViewOp m_functor; -}; - -} // end namespace internal - -template -EIGEN_STRONG_INLINE Derived& -SparseMatrixBase::operator*=(const Scalar& other) -{ - typedef typename internal::evaluator::InnerIterator EvalIterator; - internal::evaluator thisEval(derived()); - for (Index j=0; j -EIGEN_STRONG_INLINE Derived& -SparseMatrixBase::operator/=(const Scalar& other) -{ - typedef typename internal::evaluator::InnerIterator EvalIterator; - internal::evaluator thisEval(derived()); - for (Index j=0; j -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEDENSEPRODUCT_H -#define EIGEN_SPARSEDENSEPRODUCT_H - -namespace Eigen { - -namespace internal { - -template <> struct product_promote_storage_type { typedef Sparse ret; }; -template <> struct product_promote_storage_type { typedef Sparse ret; }; - -template -struct sparse_time_dense_product_impl; - -template -struct sparse_time_dense_product_impl -{ - typedef typename internal::remove_all::type Lhs; - typedef typename internal::remove_all::type Rhs; - typedef typename internal::remove_all::type Res; - typedef typename evaluator::InnerIterator LhsInnerIterator; - typedef evaluator LhsEval; - static void run(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const typename Res::Scalar& alpha) - { - LhsEval lhsEval(lhs); - - Index n = lhs.outerSize(); -#ifdef EIGEN_HAS_OPENMP - Eigen::initParallel(); - Index threads = Eigen::nbThreads(); -#endif - - for(Index c=0; c1 && lhsEval.nonZerosEstimate() > 20000) - { - #pragma omp parallel for schedule(dynamic,(n+threads*4-1)/(threads*4)) num_threads(threads) - for(Index i=0; i let's disable it for now as it is conflicting with generic scalar*matrix and matrix*scalar operators -// template -// struct ScalarBinaryOpTraits > -// { -// enum { -// Defined = 1 -// }; -// typedef typename CwiseUnaryOp, T2>::PlainObject ReturnType; -// }; - -template -struct sparse_time_dense_product_impl -{ - typedef typename internal::remove_all::type Lhs; - typedef typename internal::remove_all::type Rhs; - typedef typename internal::remove_all::type Res; - typedef typename evaluator::InnerIterator LhsInnerIterator; - static void run(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const AlphaType& alpha) - { - evaluator lhsEval(lhs); - for(Index c=0; c::ReturnType rhs_j(alpha * rhs.coeff(j,c)); - for(LhsInnerIterator it(lhsEval,j); it ;++it) - res.coeffRef(it.index(),c) += it.value() * rhs_j; - } - } - } -}; - -template -struct sparse_time_dense_product_impl -{ - typedef typename internal::remove_all::type Lhs; - typedef typename internal::remove_all::type Rhs; - typedef typename internal::remove_all::type Res; - typedef typename evaluator::InnerIterator LhsInnerIterator; - static void run(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const typename Res::Scalar& alpha) - { - evaluator lhsEval(lhs); - for(Index j=0; j -struct sparse_time_dense_product_impl -{ - typedef typename internal::remove_all::type Lhs; - typedef typename internal::remove_all::type Rhs; - typedef typename internal::remove_all::type Res; - typedef typename evaluator::InnerIterator LhsInnerIterator; - static void run(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const typename Res::Scalar& alpha) - { - evaluator lhsEval(lhs); - for(Index j=0; j -inline void sparse_time_dense_product(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const AlphaType& alpha) -{ - sparse_time_dense_product_impl::run(lhs, rhs, res, alpha); -} - -} // end namespace internal - -namespace internal { - -template -struct generic_product_impl - : generic_product_impl_base > -{ - typedef typename Product::Scalar Scalar; - - template - static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) - { - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhs); - RhsNested rhsNested(rhs); - internal::sparse_time_dense_product(lhsNested, rhsNested, dst, alpha); - } -}; - -template -struct generic_product_impl - : generic_product_impl -{}; - -template -struct generic_product_impl - : generic_product_impl_base > -{ - typedef typename Product::Scalar Scalar; - - template - static void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) - { - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhs); - RhsNested rhsNested(rhs); - - // transpose everything - Transpose dstT(dst); - internal::sparse_time_dense_product(rhsNested.transpose(), lhsNested.transpose(), dstT, alpha); - } -}; - -template -struct generic_product_impl - : generic_product_impl -{}; - -template -struct sparse_dense_outer_product_evaluator -{ -protected: - typedef typename conditional::type Lhs1; - typedef typename conditional::type ActualRhs; - typedef Product ProdXprType; - - // if the actual left-hand side is a dense vector, - // then build a sparse-view so that we can seamlessly iterate over it. - typedef typename conditional::StorageKind,Sparse>::value, - Lhs1, SparseView >::type ActualLhs; - typedef typename conditional::StorageKind,Sparse>::value, - Lhs1 const&, SparseView >::type LhsArg; - - typedef evaluator LhsEval; - typedef evaluator RhsEval; - typedef typename evaluator::InnerIterator LhsIterator; - typedef typename ProdXprType::Scalar Scalar; - -public: - enum { - Flags = NeedToTranspose ? RowMajorBit : 0, - CoeffReadCost = HugeCost - }; - - class InnerIterator : public LhsIterator - { - public: - InnerIterator(const sparse_dense_outer_product_evaluator &xprEval, Index outer) - : LhsIterator(xprEval.m_lhsXprImpl, 0), - m_outer(outer), - m_empty(false), - m_factor(get(xprEval.m_rhsXprImpl, outer, typename internal::traits::StorageKind() )) - {} - - EIGEN_STRONG_INLINE Index outer() const { return m_outer; } - EIGEN_STRONG_INLINE Index row() const { return NeedToTranspose ? m_outer : LhsIterator::index(); } - EIGEN_STRONG_INLINE Index col() const { return NeedToTranspose ? LhsIterator::index() : m_outer; } - - EIGEN_STRONG_INLINE Scalar value() const { return LhsIterator::value() * m_factor; } - EIGEN_STRONG_INLINE operator bool() const { return LhsIterator::operator bool() && (!m_empty); } - - protected: - Scalar get(const RhsEval &rhs, Index outer, Dense = Dense()) const - { - return rhs.coeff(outer); - } - - Scalar get(const RhsEval &rhs, Index outer, Sparse = Sparse()) - { - typename RhsEval::InnerIterator it(rhs, outer); - if (it && it.index()==0 && it.value()!=Scalar(0)) - return it.value(); - m_empty = true; - return Scalar(0); - } - - Index m_outer; - bool m_empty; - Scalar m_factor; - }; - - sparse_dense_outer_product_evaluator(const Lhs1 &lhs, const ActualRhs &rhs) - : m_lhs(lhs), m_lhsXprImpl(m_lhs), m_rhsXprImpl(rhs) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - // transpose case - sparse_dense_outer_product_evaluator(const ActualRhs &rhs, const Lhs1 &lhs) - : m_lhs(lhs), m_lhsXprImpl(m_lhs), m_rhsXprImpl(rhs) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - -protected: - const LhsArg m_lhs; - evaluator m_lhsXprImpl; - evaluator m_rhsXprImpl; -}; - -// sparse * dense outer product -template -struct product_evaluator, OuterProduct, SparseShape, DenseShape> - : sparse_dense_outer_product_evaluator -{ - typedef sparse_dense_outer_product_evaluator Base; - - typedef Product XprType; - typedef typename XprType::PlainObject PlainObject; - - explicit product_evaluator(const XprType& xpr) - : Base(xpr.lhs(), xpr.rhs()) - {} - -}; - -template -struct product_evaluator, OuterProduct, DenseShape, SparseShape> - : sparse_dense_outer_product_evaluator -{ - typedef sparse_dense_outer_product_evaluator Base; - - typedef Product XprType; - typedef typename XprType::PlainObject PlainObject; - - explicit product_evaluator(const XprType& xpr) - : Base(xpr.lhs(), xpr.rhs()) - {} - -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSEDENSEPRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDiagonalProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDiagonalProduct.h deleted file mode 100644 index 941c03be3de..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDiagonalProduct.h +++ /dev/null @@ -1,138 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_DIAGONAL_PRODUCT_H -#define EIGEN_SPARSE_DIAGONAL_PRODUCT_H - -namespace Eigen { - -// The product of a diagonal matrix with a sparse matrix can be easily -// implemented using expression template. -// We have two consider very different cases: -// 1 - diag * row-major sparse -// => each inner vector <=> scalar * sparse vector product -// => so we can reuse CwiseUnaryOp::InnerIterator -// 2 - diag * col-major sparse -// => each inner vector <=> densevector * sparse vector cwise product -// => again, we can reuse specialization of CwiseBinaryOp::InnerIterator -// for that particular case -// The two other cases are symmetric. - -namespace internal { - -enum { - SDP_AsScalarProduct, - SDP_AsCwiseProduct -}; - -template -struct sparse_diagonal_product_evaluator; - -template -struct product_evaluator, ProductTag, DiagonalShape, SparseShape> - : public sparse_diagonal_product_evaluator -{ - typedef Product XprType; - enum { CoeffReadCost = HugeCost, Flags = Rhs::Flags&RowMajorBit, Alignment = 0 }; // FIXME CoeffReadCost & Flags - - typedef sparse_diagonal_product_evaluator Base; - explicit product_evaluator(const XprType& xpr) : Base(xpr.rhs(), xpr.lhs().diagonal()) {} -}; - -template -struct product_evaluator, ProductTag, SparseShape, DiagonalShape> - : public sparse_diagonal_product_evaluator, Lhs::Flags&RowMajorBit?SDP_AsCwiseProduct:SDP_AsScalarProduct> -{ - typedef Product XprType; - enum { CoeffReadCost = HugeCost, Flags = Lhs::Flags&RowMajorBit, Alignment = 0 }; // FIXME CoeffReadCost & Flags - - typedef sparse_diagonal_product_evaluator, Lhs::Flags&RowMajorBit?SDP_AsCwiseProduct:SDP_AsScalarProduct> Base; - explicit product_evaluator(const XprType& xpr) : Base(xpr.lhs(), xpr.rhs().diagonal().transpose()) {} -}; - -template -struct sparse_diagonal_product_evaluator -{ -protected: - typedef typename evaluator::InnerIterator SparseXprInnerIterator; - typedef typename SparseXprType::Scalar Scalar; - -public: - class InnerIterator : public SparseXprInnerIterator - { - public: - InnerIterator(const sparse_diagonal_product_evaluator &xprEval, Index outer) - : SparseXprInnerIterator(xprEval.m_sparseXprImpl, outer), - m_coeff(xprEval.m_diagCoeffImpl.coeff(outer)) - {} - - EIGEN_STRONG_INLINE Scalar value() const { return m_coeff * SparseXprInnerIterator::value(); } - protected: - typename DiagonalCoeffType::Scalar m_coeff; - }; - - sparse_diagonal_product_evaluator(const SparseXprType &sparseXpr, const DiagonalCoeffType &diagCoeff) - : m_sparseXprImpl(sparseXpr), m_diagCoeffImpl(diagCoeff) - {} - - Index nonZerosEstimate() const { return m_sparseXprImpl.nonZerosEstimate(); } - -protected: - evaluator m_sparseXprImpl; - evaluator m_diagCoeffImpl; -}; - - -template -struct sparse_diagonal_product_evaluator -{ - typedef typename SparseXprType::Scalar Scalar; - typedef typename SparseXprType::StorageIndex StorageIndex; - - typedef typename nested_eval::type DiagCoeffNested; - - class InnerIterator - { - typedef typename evaluator::InnerIterator SparseXprIter; - public: - InnerIterator(const sparse_diagonal_product_evaluator &xprEval, Index outer) - : m_sparseIter(xprEval.m_sparseXprEval, outer), m_diagCoeffNested(xprEval.m_diagCoeffNested) - {} - - inline Scalar value() const { return m_sparseIter.value() * m_diagCoeffNested.coeff(index()); } - inline StorageIndex index() const { return m_sparseIter.index(); } - inline Index outer() const { return m_sparseIter.outer(); } - inline Index col() const { return SparseXprType::IsRowMajor ? m_sparseIter.index() : m_sparseIter.outer(); } - inline Index row() const { return SparseXprType::IsRowMajor ? m_sparseIter.outer() : m_sparseIter.index(); } - - EIGEN_STRONG_INLINE InnerIterator& operator++() { ++m_sparseIter; return *this; } - inline operator bool() const { return m_sparseIter; } - - protected: - SparseXprIter m_sparseIter; - DiagCoeffNested m_diagCoeffNested; - }; - - sparse_diagonal_product_evaluator(const SparseXprType &sparseXpr, const DiagCoeffType &diagCoeff) - : m_sparseXprEval(sparseXpr), m_diagCoeffNested(diagCoeff) - {} - - Index nonZerosEstimate() const { return m_sparseXprEval.nonZerosEstimate(); } - -protected: - evaluator m_sparseXprEval; - DiagCoeffNested m_diagCoeffNested; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_DIAGONAL_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDot.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDot.h deleted file mode 100644 index 38bc4aa9ea0..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseDot.h +++ /dev/null @@ -1,98 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_DOT_H -#define EIGEN_SPARSE_DOT_H - -namespace Eigen { - -template -template -typename internal::traits::Scalar -SparseMatrixBase::dot(const MatrixBase& other) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived) - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - - eigen_assert(size() == other.size()); - eigen_assert(other.size()>0 && "you are using a non initialized vector"); - - internal::evaluator thisEval(derived()); - typename internal::evaluator::InnerIterator i(thisEval, 0); - Scalar res(0); - while (i) - { - res += numext::conj(i.value()) * other.coeff(i.index()); - ++i; - } - return res; -} - -template -template -typename internal::traits::Scalar -SparseMatrixBase::dot(const SparseMatrixBase& other) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived) - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - - eigen_assert(size() == other.size()); - - internal::evaluator thisEval(derived()); - typename internal::evaluator::InnerIterator i(thisEval, 0); - - internal::evaluator otherEval(other.derived()); - typename internal::evaluator::InnerIterator j(otherEval, 0); - - Scalar res(0); - while (i && j) - { - if (i.index()==j.index()) - { - res += numext::conj(i.value()) * j.value(); - ++i; ++j; - } - else if (i.index() -inline typename NumTraits::Scalar>::Real -SparseMatrixBase::squaredNorm() const -{ - return numext::real((*this).cwiseAbs2().sum()); -} - -template -inline typename NumTraits::Scalar>::Real -SparseMatrixBase::norm() const -{ - using std::sqrt; - return sqrt(squaredNorm()); -} - -template -inline typename NumTraits::Scalar>::Real -SparseMatrixBase::blueNorm() const -{ - return internal::blueNorm_impl(*this); -} -} // end namespace Eigen - -#endif // EIGEN_SPARSE_DOT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseFuzzy.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseFuzzy.h deleted file mode 100644 index 7d47eb94d2f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseFuzzy.h +++ /dev/null @@ -1,29 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_FUZZY_H -#define EIGEN_SPARSE_FUZZY_H - -namespace Eigen { - -template -template -bool SparseMatrixBase::isApprox(const SparseMatrixBase& other, const RealScalar &prec) const -{ - const typename internal::nested_eval::type actualA(derived()); - typename internal::conditional::type, - const PlainObject>::type actualB(other.derived()); - - return (actualA - actualB).squaredNorm() <= prec * prec * numext::mini(actualA.squaredNorm(), actualB.squaredNorm()); -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_FUZZY_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMap.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMap.h deleted file mode 100644 index f99be3379db..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMap.h +++ /dev/null @@ -1,305 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_MAP_H -#define EIGEN_SPARSE_MAP_H - -namespace Eigen { - -namespace internal { - -template -struct traits, Options, StrideType> > - : public traits > -{ - typedef SparseMatrix PlainObjectType; - typedef traits TraitsBase; - enum { - Flags = TraitsBase::Flags & (~NestByRefBit) - }; -}; - -template -struct traits, Options, StrideType> > - : public traits > -{ - typedef SparseMatrix PlainObjectType; - typedef traits TraitsBase; - enum { - Flags = TraitsBase::Flags & (~ (NestByRefBit | LvalueBit)) - }; -}; - -} // end namespace internal - -template::has_write_access ? WriteAccessors : ReadOnlyAccessors -> class SparseMapBase; - -/** \ingroup SparseCore_Module - * class SparseMapBase - * \brief Common base class for Map and Ref instance of sparse matrix and vector. - */ -template -class SparseMapBase - : public SparseCompressedBase -{ - public: - typedef SparseCompressedBase Base; - typedef typename Base::Scalar Scalar; - typedef typename Base::StorageIndex StorageIndex; - enum { IsRowMajor = Base::IsRowMajor }; - using Base::operator=; - protected: - - typedef typename internal::conditional< - bool(internal::is_lvalue::value), - Scalar *, const Scalar *>::type ScalarPointer; - typedef typename internal::conditional< - bool(internal::is_lvalue::value), - StorageIndex *, const StorageIndex *>::type IndexPointer; - - Index m_outerSize; - Index m_innerSize; - Array m_zero_nnz; - IndexPointer m_outerIndex; - IndexPointer m_innerIndices; - ScalarPointer m_values; - IndexPointer m_innerNonZeros; - - public: - - /** \copydoc SparseMatrixBase::rows() */ - inline Index rows() const { return IsRowMajor ? m_outerSize : m_innerSize; } - /** \copydoc SparseMatrixBase::cols() */ - inline Index cols() const { return IsRowMajor ? m_innerSize : m_outerSize; } - /** \copydoc SparseMatrixBase::innerSize() */ - inline Index innerSize() const { return m_innerSize; } - /** \copydoc SparseMatrixBase::outerSize() */ - inline Index outerSize() const { return m_outerSize; } - /** \copydoc SparseCompressedBase::nonZeros */ - inline Index nonZeros() const { return m_zero_nnz[1]; } - - /** \copydoc SparseCompressedBase::isCompressed */ - bool isCompressed() const { return m_innerNonZeros==0; } - - //---------------------------------------- - // direct access interface - /** \copydoc SparseMatrix::valuePtr */ - inline const Scalar* valuePtr() const { return m_values; } - /** \copydoc SparseMatrix::innerIndexPtr */ - inline const StorageIndex* innerIndexPtr() const { return m_innerIndices; } - /** \copydoc SparseMatrix::outerIndexPtr */ - inline const StorageIndex* outerIndexPtr() const { return m_outerIndex; } - /** \copydoc SparseMatrix::innerNonZeroPtr */ - inline const StorageIndex* innerNonZeroPtr() const { return m_innerNonZeros; } - //---------------------------------------- - - /** \copydoc SparseMatrix::coeff */ - inline Scalar coeff(Index row, Index col) const - { - const Index outer = IsRowMajor ? row : col; - const Index inner = IsRowMajor ? col : row; - - Index start = m_outerIndex[outer]; - Index end = isCompressed() ? m_outerIndex[outer+1] : start + m_innerNonZeros[outer]; - if (start==end) - return Scalar(0); - else if (end>0 && inner==m_innerIndices[end-1]) - return m_values[end-1]; - // ^^ optimization: let's first check if it is the last coefficient - // (very common in high level algorithms) - - const StorageIndex* r = std::lower_bound(&m_innerIndices[start],&m_innerIndices[end-1],inner); - const Index id = r-&m_innerIndices[0]; - return ((*r==inner) && (id(nnz)), m_outerIndex(outerIndexPtr), - m_innerIndices(innerIndexPtr), m_values(valuePtr), m_innerNonZeros(innerNonZerosPtr) - {} - - // for vectors - inline SparseMapBase(Index size, Index nnz, IndexPointer innerIndexPtr, ScalarPointer valuePtr) - : m_outerSize(1), m_innerSize(size), m_zero_nnz(0,internal::convert_index(nnz)), m_outerIndex(m_zero_nnz.data()), - m_innerIndices(innerIndexPtr), m_values(valuePtr), m_innerNonZeros(0) - {} - - /** Empty destructor */ - inline ~SparseMapBase() {} - - protected: - inline SparseMapBase() {} -}; - -/** \ingroup SparseCore_Module - * class SparseMapBase - * \brief Common base class for writable Map and Ref instance of sparse matrix and vector. - */ -template -class SparseMapBase - : public SparseMapBase -{ - typedef MapBase ReadOnlyMapBase; - - public: - typedef SparseMapBase Base; - typedef typename Base::Scalar Scalar; - typedef typename Base::StorageIndex StorageIndex; - enum { IsRowMajor = Base::IsRowMajor }; - - using Base::operator=; - - public: - - //---------------------------------------- - // direct access interface - using Base::valuePtr; - using Base::innerIndexPtr; - using Base::outerIndexPtr; - using Base::innerNonZeroPtr; - /** \copydoc SparseMatrix::valuePtr */ - inline Scalar* valuePtr() { return Base::m_values; } - /** \copydoc SparseMatrix::innerIndexPtr */ - inline StorageIndex* innerIndexPtr() { return Base::m_innerIndices; } - /** \copydoc SparseMatrix::outerIndexPtr */ - inline StorageIndex* outerIndexPtr() { return Base::m_outerIndex; } - /** \copydoc SparseMatrix::innerNonZeroPtr */ - inline StorageIndex* innerNonZeroPtr() { return Base::m_innerNonZeros; } - //---------------------------------------- - - /** \copydoc SparseMatrix::coeffRef */ - inline Scalar& coeffRef(Index row, Index col) - { - const Index outer = IsRowMajor ? row : col; - const Index inner = IsRowMajor ? col : row; - - Index start = Base::m_outerIndex[outer]; - Index end = Base::isCompressed() ? Base::m_outerIndex[outer+1] : start + Base::m_innerNonZeros[outer]; - eigen_assert(end>=start && "you probably called coeffRef on a non finalized matrix"); - eigen_assert(end>start && "coeffRef cannot be called on a zero coefficient"); - StorageIndex* r = std::lower_bound(&Base::m_innerIndices[start],&Base::m_innerIndices[end],inner); - const Index id = r - &Base::m_innerIndices[0]; - eigen_assert((*r==inner) && (id(Base::m_values)[id]; - } - - inline SparseMapBase(Index rows, Index cols, Index nnz, StorageIndex* outerIndexPtr, StorageIndex* innerIndexPtr, - Scalar* valuePtr, StorageIndex* innerNonZerosPtr = 0) - : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZerosPtr) - {} - - // for vectors - inline SparseMapBase(Index size, Index nnz, StorageIndex* innerIndexPtr, Scalar* valuePtr) - : Base(size, nnz, innerIndexPtr, valuePtr) - {} - - /** Empty destructor */ - inline ~SparseMapBase() {} - - protected: - inline SparseMapBase() {} -}; - -/** \ingroup SparseCore_Module - * - * \brief Specialization of class Map for SparseMatrix-like storage. - * - * \tparam SparseMatrixType the equivalent sparse matrix type of the referenced data, it must be a template instance of class SparseMatrix. - * - * \sa class Map, class SparseMatrix, class Ref - */ -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -class Map, Options, StrideType> - : public SparseMapBase, Options, StrideType> > -#else -template -class Map - : public SparseMapBase -#endif -{ - public: - typedef SparseMapBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Map) - enum { IsRowMajor = Base::IsRowMajor }; - - public: - - /** Constructs a read-write Map to a sparse matrix of size \a rows x \a cols, containing \a nnz non-zero coefficients, - * stored as a sparse format as defined by the pointers \a outerIndexPtr, \a innerIndexPtr, and \a valuePtr. - * If the optional parameter \a innerNonZerosPtr is the null pointer, then a standard compressed format is assumed. - * - * This constructor is available only if \c SparseMatrixType is non-const. - * - * More details on the expected storage schemes are given in the \ref TutorialSparse "manual pages". - */ - inline Map(Index rows, Index cols, Index nnz, StorageIndex* outerIndexPtr, - StorageIndex* innerIndexPtr, Scalar* valuePtr, StorageIndex* innerNonZerosPtr = 0) - : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZerosPtr) - {} -#ifndef EIGEN_PARSED_BY_DOXYGEN - /** Empty destructor */ - inline ~Map() {} -}; - -template -class Map, Options, StrideType> - : public SparseMapBase, Options, StrideType> > -{ - public: - typedef SparseMapBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Map) - enum { IsRowMajor = Base::IsRowMajor }; - - public: -#endif - /** This is the const version of the above constructor. - * - * This constructor is available only if \c SparseMatrixType is const, e.g.: - * \code Map > \endcode - */ - inline Map(Index rows, Index cols, Index nnz, const StorageIndex* outerIndexPtr, - const StorageIndex* innerIndexPtr, const Scalar* valuePtr, const StorageIndex* innerNonZerosPtr = 0) - : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZerosPtr) - {} - - /** Empty destructor */ - inline ~Map() {} -}; - -namespace internal { - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Map, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Map, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_MAP_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrix.h deleted file mode 100644 index a5396538ba9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrix.h +++ /dev/null @@ -1,1404 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEMATRIX_H -#define EIGEN_SPARSEMATRIX_H - -namespace Eigen { - -/** \ingroup SparseCore_Module - * - * \class SparseMatrix - * - * \brief A versatible sparse matrix representation - * - * This class implements a more versatile variants of the common \em compressed row/column storage format. - * Each colmun's (resp. row) non zeros are stored as a pair of value with associated row (resp. colmiun) index. - * All the non zeros are stored in a single large buffer. Unlike the \em compressed format, there might be extra - * space inbetween the nonzeros of two successive colmuns (resp. rows) such that insertion of new non-zero - * can be done with limited memory reallocation and copies. - * - * A call to the function makeCompressed() turns the matrix into the standard \em compressed format - * compatible with many library. - * - * More details on this storage sceheme are given in the \ref TutorialSparse "manual pages". - * - * \tparam _Scalar the scalar type, i.e. the type of the coefficients - * \tparam _Options Union of bit flags controlling the storage scheme. Currently the only possibility - * is ColMajor or RowMajor. The default is 0 which means column-major. - * \tparam _StorageIndex the type of the indices. It has to be a \b signed type (e.g., short, int, std::ptrdiff_t). Default is \c int. - * - * \warning In %Eigen 3.2, the undocumented type \c SparseMatrix::Index was improperly defined as the storage index type (e.g., int), - * whereas it is now (starting from %Eigen 3.3) deprecated and always defined as Eigen::Index. - * Codes making use of \c SparseMatrix::Index, might thus likely have to be changed to use \c SparseMatrix::StorageIndex instead. - * - * This class can be extended with the help of the plugin mechanism described on the page - * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_SPARSEMATRIX_PLUGIN. - */ - -namespace internal { -template -struct traits > -{ - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - typedef Sparse StorageKind; - typedef MatrixXpr XprKind; - enum { - RowsAtCompileTime = Dynamic, - ColsAtCompileTime = Dynamic, - MaxRowsAtCompileTime = Dynamic, - MaxColsAtCompileTime = Dynamic, - Flags = _Options | NestByRefBit | LvalueBit | CompressedAccessBit, - SupportedAccessPatterns = InnerRandomAccessPattern - }; -}; - -template -struct traits, DiagIndex> > -{ - typedef SparseMatrix<_Scalar, _Options, _StorageIndex> MatrixType; - typedef typename ref_selector::type MatrixTypeNested; - typedef typename remove_reference::type _MatrixTypeNested; - - typedef _Scalar Scalar; - typedef Dense StorageKind; - typedef _StorageIndex StorageIndex; - typedef MatrixXpr XprKind; - - enum { - RowsAtCompileTime = Dynamic, - ColsAtCompileTime = 1, - MaxRowsAtCompileTime = Dynamic, - MaxColsAtCompileTime = 1, - Flags = LvalueBit - }; -}; - -template -struct traits, DiagIndex> > - : public traits, DiagIndex> > -{ - enum { - Flags = 0 - }; -}; - -} // end namespace internal - -template -class SparseMatrix - : public SparseCompressedBase > -{ - typedef SparseCompressedBase Base; - using Base::convert_index; - friend class SparseVector<_Scalar,0,_StorageIndex>; - public: - using Base::isCompressed; - using Base::nonZeros; - EIGEN_SPARSE_PUBLIC_INTERFACE(SparseMatrix) - using Base::operator+=; - using Base::operator-=; - - typedef MappedSparseMatrix Map; - typedef Diagonal DiagonalReturnType; - typedef Diagonal ConstDiagonalReturnType; - typedef typename Base::InnerIterator InnerIterator; - typedef typename Base::ReverseInnerIterator ReverseInnerIterator; - - - using Base::IsRowMajor; - typedef internal::CompressedStorage Storage; - enum { - Options = _Options - }; - - typedef typename Base::IndexVector IndexVector; - typedef typename Base::ScalarVector ScalarVector; - protected: - typedef SparseMatrix TransposedSparseMatrix; - - Index m_outerSize; - Index m_innerSize; - StorageIndex* m_outerIndex; - StorageIndex* m_innerNonZeros; // optional, if null then the data is compressed - Storage m_data; - - public: - - /** \returns the number of rows of the matrix */ - inline Index rows() const { return IsRowMajor ? m_outerSize : m_innerSize; } - /** \returns the number of columns of the matrix */ - inline Index cols() const { return IsRowMajor ? m_innerSize : m_outerSize; } - - /** \returns the number of rows (resp. columns) of the matrix if the storage order column major (resp. row major) */ - inline Index innerSize() const { return m_innerSize; } - /** \returns the number of columns (resp. rows) of the matrix if the storage order column major (resp. row major) */ - inline Index outerSize() const { return m_outerSize; } - - /** \returns a const pointer to the array of values. - * This function is aimed at interoperability with other libraries. - * \sa innerIndexPtr(), outerIndexPtr() */ - inline const Scalar* valuePtr() const { return m_data.valuePtr(); } - /** \returns a non-const pointer to the array of values. - * This function is aimed at interoperability with other libraries. - * \sa innerIndexPtr(), outerIndexPtr() */ - inline Scalar* valuePtr() { return m_data.valuePtr(); } - - /** \returns a const pointer to the array of inner indices. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), outerIndexPtr() */ - inline const StorageIndex* innerIndexPtr() const { return m_data.indexPtr(); } - /** \returns a non-const pointer to the array of inner indices. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), outerIndexPtr() */ - inline StorageIndex* innerIndexPtr() { return m_data.indexPtr(); } - - /** \returns a const pointer to the array of the starting positions of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), innerIndexPtr() */ - inline const StorageIndex* outerIndexPtr() const { return m_outerIndex; } - /** \returns a non-const pointer to the array of the starting positions of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \sa valuePtr(), innerIndexPtr() */ - inline StorageIndex* outerIndexPtr() { return m_outerIndex; } - - /** \returns a const pointer to the array of the number of non zeros of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 in compressed mode */ - inline const StorageIndex* innerNonZeroPtr() const { return m_innerNonZeros; } - /** \returns a non-const pointer to the array of the number of non zeros of the inner vectors. - * This function is aimed at interoperability with other libraries. - * \warning it returns the null pointer 0 in compressed mode */ - inline StorageIndex* innerNonZeroPtr() { return m_innerNonZeros; } - - /** \internal */ - inline Storage& data() { return m_data; } - /** \internal */ - inline const Storage& data() const { return m_data; } - - /** \returns the value of the matrix at position \a i, \a j - * This function returns Scalar(0) if the element is an explicit \em zero */ - inline Scalar coeff(Index row, Index col) const - { - eigen_assert(row>=0 && row=0 && col=0 && row=0 && col=start && "you probably called coeffRef on a non finalized matrix"); - if(end<=start) - return insert(row,col); - const Index p = m_data.searchLowerIndex(start,end-1,StorageIndex(inner)); - if((pinnerSize() non zeros if reserve(Index) has not been called earlier. - * In this case, the insertion procedure is optimized for a \e sequential insertion mode where elements are assumed to be - * inserted by increasing outer-indices. - * - * If that's not the case, then it is strongly recommended to either use a triplet-list to assemble the matrix, or to first - * call reserve(const SizesType &) to reserve the appropriate number of non-zero elements per inner vector. - * - * Assuming memory has been appropriately reserved, this function performs a sorted insertion in O(1) - * if the elements of each inner vector are inserted in increasing inner index order, and in O(nnz_j) for a random insertion. - * - */ - Scalar& insert(Index row, Index col); - - public: - - /** Removes all non zeros but keep allocated memory - * - * This function does not free the currently allocated memory. To release as much as memory as possible, - * call \code mat.data().squeeze(); \endcode after resizing it. - * - * \sa resize(Index,Index), data() - */ - inline void setZero() - { - m_data.clear(); - memset(m_outerIndex, 0, (m_outerSize+1)*sizeof(StorageIndex)); - if(m_innerNonZeros) - memset(m_innerNonZeros, 0, (m_outerSize)*sizeof(StorageIndex)); - } - - /** Preallocates \a reserveSize non zeros. - * - * Precondition: the matrix must be in compressed mode. */ - inline void reserve(Index reserveSize) - { - eigen_assert(isCompressed() && "This function does not make sense in non compressed mode."); - m_data.reserve(reserveSize); - } - - #ifdef EIGEN_PARSED_BY_DOXYGEN - /** Preallocates \a reserveSize[\c j] non zeros for each column (resp. row) \c j. - * - * This function turns the matrix in non-compressed mode. - * - * The type \c SizesType must expose the following interface: - \code - typedef value_type; - const value_type& operator[](i) const; - \endcode - * for \c i in the [0,this->outerSize()[ range. - * Typical choices include std::vector, Eigen::VectorXi, Eigen::VectorXi::Constant, etc. - */ - template - inline void reserve(const SizesType& reserveSizes); - #else - template - inline void reserve(const SizesType& reserveSizes, const typename SizesType::value_type& enableif = - #if (!EIGEN_COMP_MSVC) || (EIGEN_COMP_MSVC>=1500) // MSVC 2005 fails to compile with this typename - typename - #endif - SizesType::value_type()) - { - EIGEN_UNUSED_VARIABLE(enableif); - reserveInnerVectors(reserveSizes); - } - #endif // EIGEN_PARSED_BY_DOXYGEN - protected: - template - inline void reserveInnerVectors(const SizesType& reserveSizes) - { - if(isCompressed()) - { - Index totalReserveSize = 0; - // turn the matrix into non-compressed mode - m_innerNonZeros = static_cast(std::malloc(m_outerSize * sizeof(StorageIndex))); - if (!m_innerNonZeros) internal::throw_std_bad_alloc(); - - // temporarily use m_innerSizes to hold the new starting points. - StorageIndex* newOuterIndex = m_innerNonZeros; - - StorageIndex count = 0; - for(Index j=0; j=0; --j) - { - StorageIndex innerNNZ = previousOuterIndex - m_outerIndex[j]; - for(Index i=innerNNZ-1; i>=0; --i) - { - m_data.index(newOuterIndex[j]+i) = m_data.index(m_outerIndex[j]+i); - m_data.value(newOuterIndex[j]+i) = m_data.value(m_outerIndex[j]+i); - } - previousOuterIndex = m_outerIndex[j]; - m_outerIndex[j] = newOuterIndex[j]; - m_innerNonZeros[j] = innerNNZ; - } - if(m_outerSize>0) - m_outerIndex[m_outerSize] = m_outerIndex[m_outerSize-1] + m_innerNonZeros[m_outerSize-1] + reserveSizes[m_outerSize-1]; - - m_data.resize(m_outerIndex[m_outerSize]); - } - else - { - StorageIndex* newOuterIndex = static_cast(std::malloc((m_outerSize+1)*sizeof(StorageIndex))); - if (!newOuterIndex) internal::throw_std_bad_alloc(); - - StorageIndex count = 0; - for(Index j=0; j(reserveSizes[j], alreadyReserved); - count += toReserve + m_innerNonZeros[j]; - } - newOuterIndex[m_outerSize] = count; - - m_data.resize(count); - for(Index j=m_outerSize-1; j>=0; --j) - { - Index offset = newOuterIndex[j] - m_outerIndex[j]; - if(offset>0) - { - StorageIndex innerNNZ = m_innerNonZeros[j]; - for(Index i=innerNNZ-1; i>=0; --i) - { - m_data.index(newOuterIndex[j]+i) = m_data.index(m_outerIndex[j]+i); - m_data.value(newOuterIndex[j]+i) = m_data.value(m_outerIndex[j]+i); - } - } - } - - std::swap(m_outerIndex, newOuterIndex); - std::free(newOuterIndex); - } - - } - public: - - //--- low level purely coherent filling --- - - /** \internal - * \returns a reference to the non zero coefficient at position \a row, \a col assuming that: - * - the nonzero does not already exist - * - the new coefficient is the last one according to the storage order - * - * Before filling a given inner vector you must call the statVec(Index) function. - * - * After an insertion session, you should call the finalize() function. - * - * \sa insert, insertBackByOuterInner, startVec */ - inline Scalar& insertBack(Index row, Index col) - { - return insertBackByOuterInner(IsRowMajor?row:col, IsRowMajor?col:row); - } - - /** \internal - * \sa insertBack, startVec */ - inline Scalar& insertBackByOuterInner(Index outer, Index inner) - { - eigen_assert(Index(m_outerIndex[outer+1]) == m_data.size() && "Invalid ordered insertion (invalid outer index)"); - eigen_assert( (m_outerIndex[outer+1]-m_outerIndex[outer]==0 || m_data.index(m_data.size()-1)(m_data.size()); - Index i = m_outerSize; - // find the last filled column - while (i>=0 && m_outerIndex[i]==0) - --i; - ++i; - while (i<=m_outerSize) - { - m_outerIndex[i] = size; - ++i; - } - } - } - - //--- - - template - void setFromTriplets(const InputIterators& begin, const InputIterators& end); - - template - void setFromTriplets(const InputIterators& begin, const InputIterators& end, DupFunctor dup_func); - - void sumupDuplicates() { collapseDuplicates(internal::scalar_sum_op()); } - - template - void collapseDuplicates(DupFunctor dup_func = DupFunctor()); - - //--- - - /** \internal - * same as insert(Index,Index) except that the indices are given relative to the storage order */ - Scalar& insertByOuterInner(Index j, Index i) - { - return insert(IsRowMajor ? j : i, IsRowMajor ? i : j); - } - - /** Turns the matrix into the \em compressed format. - */ - void makeCompressed() - { - if(isCompressed()) - return; - - eigen_internal_assert(m_outerIndex!=0 && m_outerSize>0); - - Index oldStart = m_outerIndex[1]; - m_outerIndex[1] = m_innerNonZeros[0]; - for(Index j=1; j0) - { - for(Index k=0; k(std::malloc(m_outerSize * sizeof(StorageIndex))); - for (Index i = 0; i < m_outerSize; i++) - { - m_innerNonZeros[i] = m_outerIndex[i+1] - m_outerIndex[i]; - } - } - - /** Suppresses all nonzeros which are \b much \b smaller \b than \a reference under the tolerence \a epsilon */ - void prune(const Scalar& reference, const RealScalar& epsilon = NumTraits::dummy_precision()) - { - prune(default_prunning_func(reference,epsilon)); - } - - /** Turns the matrix into compressed format, and suppresses all nonzeros which do not satisfy the predicate \a keep. - * The functor type \a KeepFunc must implement the following function: - * \code - * bool operator() (const Index& row, const Index& col, const Scalar& value) const; - * \endcode - * \sa prune(Scalar,RealScalar) - */ - template - void prune(const KeepFunc& keep = KeepFunc()) - { - // TODO optimize the uncompressed mode to avoid moving and allocating the data twice - makeCompressed(); - - StorageIndex k = 0; - for(Index j=0; jrows() == rows && this->cols() == cols) return; - - // If one dimension is null, then there is nothing to be preserved - if(rows==0 || cols==0) return resize(rows,cols); - - Index innerChange = IsRowMajor ? cols - this->cols() : rows - this->rows(); - Index outerChange = IsRowMajor ? rows - this->rows() : cols - this->cols(); - StorageIndex newInnerSize = convert_index(IsRowMajor ? cols : rows); - - // Deals with inner non zeros - if (m_innerNonZeros) - { - // Resize m_innerNonZeros - StorageIndex *newInnerNonZeros = static_cast(std::realloc(m_innerNonZeros, (m_outerSize + outerChange) * sizeof(StorageIndex))); - if (!newInnerNonZeros) internal::throw_std_bad_alloc(); - m_innerNonZeros = newInnerNonZeros; - - for(Index i=m_outerSize; i(std::malloc((m_outerSize+outerChange+1) * sizeof(StorageIndex))); - if (!m_innerNonZeros) internal::throw_std_bad_alloc(); - for(Index i = 0; i < m_outerSize; i++) - m_innerNonZeros[i] = m_outerIndex[i+1] - m_outerIndex[i]; - } - - // Change the m_innerNonZeros in case of a decrease of inner size - if (m_innerNonZeros && innerChange < 0) - { - for(Index i = 0; i < m_outerSize + (std::min)(outerChange, Index(0)); i++) - { - StorageIndex &n = m_innerNonZeros[i]; - StorageIndex start = m_outerIndex[i]; - while (n > 0 && m_data.index(start+n-1) >= newInnerSize) --n; - } - } - - m_innerSize = newInnerSize; - - // Re-allocate outer index structure if necessary - if (outerChange == 0) - return; - - StorageIndex *newOuterIndex = static_cast(std::realloc(m_outerIndex, (m_outerSize + outerChange + 1) * sizeof(StorageIndex))); - if (!newOuterIndex) internal::throw_std_bad_alloc(); - m_outerIndex = newOuterIndex; - if (outerChange > 0) - { - StorageIndex last = m_outerSize == 0 ? 0 : m_outerIndex[m_outerSize]; - for(Index i=m_outerSize; i(std::malloc((outerSize + 1) * sizeof(StorageIndex))); - if (!m_outerIndex) internal::throw_std_bad_alloc(); - - m_outerSize = outerSize; - } - if(m_innerNonZeros) - { - std::free(m_innerNonZeros); - m_innerNonZeros = 0; - } - memset(m_outerIndex, 0, (m_outerSize+1)*sizeof(StorageIndex)); - } - - /** \internal - * Resize the nonzero vector to \a size */ - void resizeNonZeros(Index size) - { - m_data.resize(size); - } - - /** \returns a const expression of the diagonal coefficients. */ - const ConstDiagonalReturnType diagonal() const { return ConstDiagonalReturnType(*this); } - - /** \returns a read-write expression of the diagonal coefficients. - * \warning If the diagonal entries are written, then all diagonal - * entries \b must already exist, otherwise an assertion will be raised. - */ - DiagonalReturnType diagonal() { return DiagonalReturnType(*this); } - - /** Default constructor yielding an empty \c 0 \c x \c 0 matrix */ - inline SparseMatrix() - : m_outerSize(-1), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - resize(0, 0); - } - - /** Constructs a \a rows \c x \a cols empty matrix */ - inline SparseMatrix(Index rows, Index cols) - : m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - resize(rows, cols); - } - - /** Constructs a sparse matrix from the sparse expression \a other */ - template - inline SparseMatrix(const SparseMatrixBase& other) - : m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - check_template_parameters(); - const bool needToTranspose = (Flags & RowMajorBit) != (internal::evaluator::Flags & RowMajorBit); - if (needToTranspose) - *this = other.derived(); - else - { - #ifdef EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - #endif - internal::call_assignment_no_alias(*this, other.derived()); - } - } - - /** Constructs a sparse matrix from the sparse selfadjoint view \a other */ - template - inline SparseMatrix(const SparseSelfAdjointView& other) - : m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - Base::operator=(other); - } - - /** Copy constructor (it performs a deep copy) */ - inline SparseMatrix(const SparseMatrix& other) - : Base(), m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - *this = other.derived(); - } - - /** \brief Copy constructor with in-place evaluation */ - template - SparseMatrix(const ReturnByValue& other) - : Base(), m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - initAssignment(other); - other.evalTo(*this); - } - - /** \brief Copy constructor with in-place evaluation */ - template - explicit SparseMatrix(const DiagonalBase& other) - : Base(), m_outerSize(0), m_innerSize(0), m_outerIndex(0), m_innerNonZeros(0) - { - check_template_parameters(); - *this = other.derived(); - } - - /** Swaps the content of two sparse matrices of the same type. - * This is a fast operation that simply swaps the underlying pointers and parameters. */ - inline void swap(SparseMatrix& other) - { - //EIGEN_DBG_SPARSE(std::cout << "SparseMatrix:: swap\n"); - std::swap(m_outerIndex, other.m_outerIndex); - std::swap(m_innerSize, other.m_innerSize); - std::swap(m_outerSize, other.m_outerSize); - std::swap(m_innerNonZeros, other.m_innerNonZeros); - m_data.swap(other.m_data); - } - - /** Sets *this to the identity matrix. - * This function also turns the matrix into compressed mode, and drop any reserved memory. */ - inline void setIdentity() - { - eigen_assert(rows() == cols() && "ONLY FOR SQUARED MATRICES"); - this->m_data.resize(rows()); - Eigen::Map(this->m_data.indexPtr(), rows()).setLinSpaced(0, StorageIndex(rows()-1)); - Eigen::Map(this->m_data.valuePtr(), rows()).setOnes(); - Eigen::Map(this->m_outerIndex, rows()+1).setLinSpaced(0, StorageIndex(rows())); - std::free(m_innerNonZeros); - m_innerNonZeros = 0; - } - inline SparseMatrix& operator=(const SparseMatrix& other) - { - if (other.isRValue()) - { - swap(other.const_cast_derived()); - } - else if(this!=&other) - { - #ifdef EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - #endif - initAssignment(other); - if(other.isCompressed()) - { - internal::smart_copy(other.m_outerIndex, other.m_outerIndex + m_outerSize + 1, m_outerIndex); - m_data = other.m_data; - } - else - { - Base::operator=(other); - } - } - return *this; - } - -#ifndef EIGEN_PARSED_BY_DOXYGEN - template - inline SparseMatrix& operator=(const EigenBase& other) - { return Base::operator=(other.derived()); } -#endif // EIGEN_PARSED_BY_DOXYGEN - - template - EIGEN_DONT_INLINE SparseMatrix& operator=(const SparseMatrixBase& other); - - friend std::ostream & operator << (std::ostream & s, const SparseMatrix& m) - { - EIGEN_DBG_SPARSE( - s << "Nonzero entries:\n"; - if(m.isCompressed()) - { - for (Index i=0; i&>(m); - return s; - } - - /** Destructor */ - inline ~SparseMatrix() - { - std::free(m_outerIndex); - std::free(m_innerNonZeros); - } - - /** Overloaded for performance */ - Scalar sum() const; - -# ifdef EIGEN_SPARSEMATRIX_PLUGIN -# include EIGEN_SPARSEMATRIX_PLUGIN -# endif - -protected: - - template - void initAssignment(const Other& other) - { - resize(other.rows(), other.cols()); - if(m_innerNonZeros) - { - std::free(m_innerNonZeros); - m_innerNonZeros = 0; - } - } - - /** \internal - * \sa insert(Index,Index) */ - EIGEN_DONT_INLINE Scalar& insertCompressed(Index row, Index col); - - /** \internal - * A vector object that is equal to 0 everywhere but v at the position i */ - class SingletonVector - { - StorageIndex m_index; - StorageIndex m_value; - public: - typedef StorageIndex value_type; - SingletonVector(Index i, Index v) - : m_index(convert_index(i)), m_value(convert_index(v)) - {} - - StorageIndex operator[](Index i) const { return i==m_index ? m_value : 0; } - }; - - /** \internal - * \sa insert(Index,Index) */ - EIGEN_DONT_INLINE Scalar& insertUncompressed(Index row, Index col); - -public: - /** \internal - * \sa insert(Index,Index) */ - EIGEN_STRONG_INLINE Scalar& insertBackUncompressed(Index row, Index col) - { - const Index outer = IsRowMajor ? row : col; - const Index inner = IsRowMajor ? col : row; - - eigen_assert(!isCompressed()); - eigen_assert(m_innerNonZeros[outer]<=(m_outerIndex[outer+1] - m_outerIndex[outer])); - - Index p = m_outerIndex[outer] + m_innerNonZeros[outer]++; - m_data.index(p) = convert_index(inner); - return (m_data.value(p) = Scalar(0)); - } - -private: - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT(NumTraits::IsSigned,THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE); - EIGEN_STATIC_ASSERT((Options&(ColMajor|RowMajor))==Options,INVALID_MATRIX_TEMPLATE_PARAMETERS); - } - - struct default_prunning_func { - default_prunning_func(const Scalar& ref, const RealScalar& eps) : reference(ref), epsilon(eps) {} - inline bool operator() (const Index&, const Index&, const Scalar& value) const - { - return !internal::isMuchSmallerThan(value, reference, epsilon); - } - Scalar reference; - RealScalar epsilon; - }; -}; - -namespace internal { - -template -void set_from_triplets(const InputIterator& begin, const InputIterator& end, SparseMatrixType& mat, DupFunctor dup_func) -{ - enum { IsRowMajor = SparseMatrixType::IsRowMajor }; - typedef typename SparseMatrixType::Scalar Scalar; - typedef typename SparseMatrixType::StorageIndex StorageIndex; - SparseMatrix trMat(mat.rows(),mat.cols()); - - if(begin!=end) - { - // pass 1: count the nnz per inner-vector - typename SparseMatrixType::IndexVector wi(trMat.outerSize()); - wi.setZero(); - for(InputIterator it(begin); it!=end; ++it) - { - eigen_assert(it->row()>=0 && it->row()col()>=0 && it->col()col() : it->row())++; - } - - // pass 2: insert all the elements into trMat - trMat.reserve(wi); - for(InputIterator it(begin); it!=end; ++it) - trMat.insertBackUncompressed(it->row(),it->col()) = it->value(); - - // pass 3: - trMat.collapseDuplicates(dup_func); - } - - // pass 4: transposed copy -> implicit sorting - mat = trMat; -} - -} - - -/** Fill the matrix \c *this with the list of \em triplets defined by the iterator range \a begin - \a end. - * - * A \em triplet is a tuple (i,j,value) defining a non-zero element. - * The input list of triplets does not have to be sorted, and can contains duplicated elements. - * In any case, the result is a \b sorted and \b compressed sparse matrix where the duplicates have been summed up. - * This is a \em O(n) operation, with \em n the number of triplet elements. - * The initial contents of \c *this is destroyed. - * The matrix \c *this must be properly resized beforehand using the SparseMatrix(Index,Index) constructor, - * or the resize(Index,Index) method. The sizes are not extracted from the triplet list. - * - * The \a InputIterators value_type must provide the following interface: - * \code - * Scalar value() const; // the value - * Scalar row() const; // the row index i - * Scalar col() const; // the column index j - * \endcode - * See for instance the Eigen::Triplet template class. - * - * Here is a typical usage example: - * \code - typedef Triplet T; - std::vector tripletList; - triplets.reserve(estimation_of_entries); - for(...) - { - // ... - tripletList.push_back(T(i,j,v_ij)); - } - SparseMatrixType m(rows,cols); - m.setFromTriplets(tripletList.begin(), tripletList.end()); - // m is ready to go! - * \endcode - * - * \warning The list of triplets is read multiple times (at least twice). Therefore, it is not recommended to define - * an abstract iterator over a complex data-structure that would be expensive to evaluate. The triplets should rather - * be explicitely stored into a std::vector for instance. - */ -template -template -void SparseMatrix::setFromTriplets(const InputIterators& begin, const InputIterators& end) -{ - internal::set_from_triplets >(begin, end, *this, internal::scalar_sum_op()); -} - -/** The same as setFromTriplets but when duplicates are met the functor \a dup_func is applied: - * \code - * value = dup_func(OldValue, NewValue) - * \endcode - * Here is a C++11 example keeping the latest entry only: - * \code - * mat.setFromTriplets(triplets.begin(), triplets.end(), [] (const Scalar&,const Scalar &b) { return b; }); - * \endcode - */ -template -template -void SparseMatrix::setFromTriplets(const InputIterators& begin, const InputIterators& end, DupFunctor dup_func) -{ - internal::set_from_triplets, DupFunctor>(begin, end, *this, dup_func); -} - -/** \internal */ -template -template -void SparseMatrix::collapseDuplicates(DupFunctor dup_func) -{ - eigen_assert(!isCompressed()); - // TODO, in practice we should be able to use m_innerNonZeros for that task - IndexVector wi(innerSize()); - wi.fill(-1); - StorageIndex count = 0; - // for each inner-vector, wi[inner_index] will hold the position of first element into the index/value buffers - for(Index j=0; j=start) - { - // we already meet this entry => accumulate it - m_data.value(wi(i)) = dup_func(m_data.value(wi(i)), m_data.value(k)); - } - else - { - m_data.value(count) = m_data.value(k); - m_data.index(count) = m_data.index(k); - wi(i) = count; - ++count; - } - } - m_outerIndex[j] = start; - } - m_outerIndex[m_outerSize] = count; - - // turn the matrix into compressed form - std::free(m_innerNonZeros); - m_innerNonZeros = 0; - m_data.resize(m_outerIndex[m_outerSize]); -} - -template -template -EIGEN_DONT_INLINE SparseMatrix& SparseMatrix::operator=(const SparseMatrixBase& other) -{ - EIGEN_STATIC_ASSERT((internal::is_same::value), - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - - #ifdef EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - #endif - - const bool needToTranspose = (Flags & RowMajorBit) != (internal::evaluator::Flags & RowMajorBit); - if (needToTranspose) - { - #ifdef EIGEN_SPARSE_TRANSPOSED_COPY_PLUGIN - EIGEN_SPARSE_TRANSPOSED_COPY_PLUGIN - #endif - // two passes algorithm: - // 1 - compute the number of coeffs per dest inner vector - // 2 - do the actual copy/eval - // Since each coeff of the rhs has to be evaluated twice, let's evaluate it if needed - typedef typename internal::nested_eval::type >::type OtherCopy; - typedef typename internal::remove_all::type _OtherCopy; - typedef internal::evaluator<_OtherCopy> OtherCopyEval; - OtherCopy otherCopy(other.derived()); - OtherCopyEval otherCopyEval(otherCopy); - - SparseMatrix dest(other.rows(),other.cols()); - Eigen::Map (dest.m_outerIndex,dest.outerSize()).setZero(); - - // pass 1 - // FIXME the above copy could be merged with that pass - for (Index j=0; jswap(dest); - return *this; - } - else - { - if(other.isRValue()) - { - initAssignment(other.derived()); - } - // there is no special optimization - return Base::operator=(other.derived()); - } -} - -template -typename SparseMatrix<_Scalar,_Options,_StorageIndex>::Scalar& SparseMatrix<_Scalar,_Options,_StorageIndex>::insert(Index row, Index col) -{ - eigen_assert(row>=0 && row=0 && col(std::malloc(m_outerSize * sizeof(StorageIndex))); - if(!m_innerNonZeros) internal::throw_std_bad_alloc(); - - memset(m_innerNonZeros, 0, (m_outerSize)*sizeof(StorageIndex)); - - // pack all inner-vectors to the end of the pre-allocated space - // and allocate the entire free-space to the first inner-vector - StorageIndex end = convert_index(m_data.allocatedSize()); - for(Index j=1; j<=m_outerSize; ++j) - m_outerIndex[j] = end; - } - else - { - // turn the matrix into non-compressed mode - m_innerNonZeros = static_cast(std::malloc(m_outerSize * sizeof(StorageIndex))); - if(!m_innerNonZeros) internal::throw_std_bad_alloc(); - for(Index j=0; j=0 && m_innerNonZeros[j]==0) - m_outerIndex[j--] = p; - - // push back the new element - ++m_innerNonZeros[outer]; - m_data.append(Scalar(0), inner); - - // check for reallocation - if(data_end != m_data.allocatedSize()) - { - // m_data has been reallocated - // -> move remaining inner-vectors back to the end of the free-space - // so that the entire free-space is allocated to the current inner-vector. - eigen_internal_assert(data_end < m_data.allocatedSize()); - StorageIndex new_end = convert_index(m_data.allocatedSize()); - for(Index k=outer+1; k<=m_outerSize; ++k) - if(m_outerIndex[k]==data_end) - m_outerIndex[k] = new_end; - } - return m_data.value(p); - } - - // Second case: the next inner-vector is packed to the end - // and the current inner-vector end match the used-space. - if(m_outerIndex[outer+1]==data_end && m_outerIndex[outer]+m_innerNonZeros[outer]==m_data.size()) - { - eigen_internal_assert(outer+1==m_outerSize || m_innerNonZeros[outer+1]==0); - - // add space for the new element - ++m_innerNonZeros[outer]; - m_data.resize(m_data.size()+1); - - // check for reallocation - if(data_end != m_data.allocatedSize()) - { - // m_data has been reallocated - // -> move remaining inner-vectors back to the end of the free-space - // so that the entire free-space is allocated to the current inner-vector. - eigen_internal_assert(data_end < m_data.allocatedSize()); - StorageIndex new_end = convert_index(m_data.allocatedSize()); - for(Index k=outer+1; k<=m_outerSize; ++k) - if(m_outerIndex[k]==data_end) - m_outerIndex[k] = new_end; - } - - // and insert it at the right position (sorted insertion) - Index startId = m_outerIndex[outer]; - Index p = m_outerIndex[outer]+m_innerNonZeros[outer]-1; - while ( (p > startId) && (m_data.index(p-1) > inner) ) - { - m_data.index(p) = m_data.index(p-1); - m_data.value(p) = m_data.value(p-1); - --p; - } - - m_data.index(p) = convert_index(inner); - return (m_data.value(p) = 0); - } - - if(m_data.size() != m_data.allocatedSize()) - { - // make sure the matrix is compatible to random un-compressed insertion: - m_data.resize(m_data.allocatedSize()); - this->reserveInnerVectors(Array::Constant(m_outerSize, 2)); - } - - return insertUncompressed(row,col); -} - -template -EIGEN_DONT_INLINE typename SparseMatrix<_Scalar,_Options,_StorageIndex>::Scalar& SparseMatrix<_Scalar,_Options,_StorageIndex>::insertUncompressed(Index row, Index col) -{ - eigen_assert(!isCompressed()); - - const Index outer = IsRowMajor ? row : col; - const StorageIndex inner = convert_index(IsRowMajor ? col : row); - - Index room = m_outerIndex[outer+1] - m_outerIndex[outer]; - StorageIndex innerNNZ = m_innerNonZeros[outer]; - if(innerNNZ>=room) - { - // this inner vector is full, we need to reallocate the whole buffer :( - reserve(SingletonVector(outer,std::max(2,innerNNZ))); - } - - Index startId = m_outerIndex[outer]; - Index p = startId + m_innerNonZeros[outer]; - while ( (p > startId) && (m_data.index(p-1) > inner) ) - { - m_data.index(p) = m_data.index(p-1); - m_data.value(p) = m_data.value(p-1); - --p; - } - eigen_assert((p<=startId || m_data.index(p-1)!=inner) && "you cannot insert an element that already exists, you must call coeffRef to this end"); - - m_innerNonZeros[outer]++; - - m_data.index(p) = inner; - return (m_data.value(p) = Scalar(0)); -} - -template -EIGEN_DONT_INLINE typename SparseMatrix<_Scalar,_Options,_StorageIndex>::Scalar& SparseMatrix<_Scalar,_Options,_StorageIndex>::insertCompressed(Index row, Index col) -{ - eigen_assert(isCompressed()); - - const Index outer = IsRowMajor ? row : col; - const Index inner = IsRowMajor ? col : row; - - Index previousOuter = outer; - if (m_outerIndex[outer+1]==0) - { - // we start a new inner vector - while (previousOuter>=0 && m_outerIndex[previousOuter]==0) - { - m_outerIndex[previousOuter] = convert_index(m_data.size()); - --previousOuter; - } - m_outerIndex[outer+1] = m_outerIndex[outer]; - } - - // here we have to handle the tricky case where the outerIndex array - // starts with: [ 0 0 0 0 0 1 ...] and we are inserted in, e.g., - // the 2nd inner vector... - bool isLastVec = (!(previousOuter==-1 && m_data.size()!=0)) - && (std::size_t(m_outerIndex[outer+1]) == m_data.size()); - - std::size_t startId = m_outerIndex[outer]; - // FIXME let's make sure sizeof(long int) == sizeof(std::size_t) - std::size_t p = m_outerIndex[outer+1]; - ++m_outerIndex[outer+1]; - - double reallocRatio = 1; - if (m_data.allocatedSize()<=m_data.size()) - { - // if there is no preallocated memory, let's reserve a minimum of 32 elements - if (m_data.size()==0) - { - m_data.reserve(32); - } - else - { - // we need to reallocate the data, to reduce multiple reallocations - // we use a smart resize algorithm based on the current filling ratio - // in addition, we use double to avoid integers overflows - double nnzEstimate = double(m_outerIndex[outer])*double(m_outerSize)/double(outer+1); - reallocRatio = (nnzEstimate-double(m_data.size()))/double(m_data.size()); - // furthermore we bound the realloc ratio to: - // 1) reduce multiple minor realloc when the matrix is almost filled - // 2) avoid to allocate too much memory when the matrix is almost empty - reallocRatio = (std::min)((std::max)(reallocRatio,1.5),8.); - } - } - m_data.resize(m_data.size()+1,reallocRatio); - - if (!isLastVec) - { - if (previousOuter==-1) - { - // oops wrong guess. - // let's correct the outer offsets - for (Index k=0; k<=(outer+1); ++k) - m_outerIndex[k] = 0; - Index k=outer+1; - while(m_outerIndex[k]==0) - m_outerIndex[k++] = 1; - while (k<=m_outerSize && m_outerIndex[k]!=0) - m_outerIndex[k++]++; - p = 0; - --k; - k = m_outerIndex[k]-1; - while (k>0) - { - m_data.index(k) = m_data.index(k-1); - m_data.value(k) = m_data.value(k-1); - k--; - } - } - else - { - // we are not inserting into the last inner vec - // update outer indices: - Index j = outer+2; - while (j<=m_outerSize && m_outerIndex[j]!=0) - m_outerIndex[j++]++; - --j; - // shift data of last vecs: - Index k = m_outerIndex[j]-1; - while (k>=Index(p)) - { - m_data.index(k) = m_data.index(k-1); - m_data.value(k) = m_data.value(k-1); - k--; - } - } - } - - while ( (p > startId) && (m_data.index(p-1) > inner) ) - { - m_data.index(p) = m_data.index(p-1); - m_data.value(p) = m_data.value(p-1); - --p; - } - - m_data.index(p) = inner; - return (m_data.value(p) = Scalar(0)); -} - -namespace internal { - -template -struct evaluator > - : evaluator > > -{ - typedef evaluator > > Base; - typedef SparseMatrix<_Scalar,_Options,_StorageIndex> SparseMatrixType; - evaluator() : Base() {} - explicit evaluator(const SparseMatrixType &mat) : Base(mat) {} -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSEMATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrixBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrixBase.h deleted file mode 100644 index c6b548f11ad..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseMatrixBase.h +++ /dev/null @@ -1,405 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEMATRIXBASE_H -#define EIGEN_SPARSEMATRIXBASE_H - -namespace Eigen { - -/** \ingroup SparseCore_Module - * - * \class SparseMatrixBase - * - * \brief Base class of any sparse matrices or sparse expressions - * - * \tparam Derived is the derived type, e.g. a sparse matrix type, or an expression, etc. - * - * This class can be extended with the help of the plugin mechanism described on the page - * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_SPARSEMATRIXBASE_PLUGIN. - */ -template class SparseMatrixBase - : public EigenBase -{ - public: - - typedef typename internal::traits::Scalar Scalar; - - /** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex, etc. - * - * It is an alias for the Scalar type */ - typedef Scalar value_type; - - typedef typename internal::packet_traits::type PacketScalar; - typedef typename internal::traits::StorageKind StorageKind; - - /** The integer type used to \b store indices within a SparseMatrix. - * For a \c SparseMatrix it an alias of the third template parameter \c IndexType. */ - typedef typename internal::traits::StorageIndex StorageIndex; - - typedef typename internal::add_const_on_value_type_if_arithmetic< - typename internal::packet_traits::type - >::type PacketReturnType; - - typedef SparseMatrixBase StorageBaseType; - - typedef Matrix IndexVector; - typedef Matrix ScalarVector; - - template - Derived& operator=(const EigenBase &other); - - enum { - - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - /**< The number of rows at compile-time. This is just a copy of the value provided - * by the \a Derived type. If a value is not known at compile-time, - * it is set to the \a Dynamic constant. - * \sa MatrixBase::rows(), MatrixBase::cols(), ColsAtCompileTime, SizeAtCompileTime */ - - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - /**< The number of columns at compile-time. This is just a copy of the value provided - * by the \a Derived type. If a value is not known at compile-time, - * it is set to the \a Dynamic constant. - * \sa MatrixBase::rows(), MatrixBase::cols(), RowsAtCompileTime, SizeAtCompileTime */ - - - SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, - internal::traits::ColsAtCompileTime>::ret), - /**< This is equal to the number of coefficients, i.e. the number of - * rows times the number of columns, or to \a Dynamic if this is not - * known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */ - - MaxRowsAtCompileTime = RowsAtCompileTime, - MaxColsAtCompileTime = ColsAtCompileTime, - - MaxSizeAtCompileTime = (internal::size_at_compile_time::ret), - - IsVectorAtCompileTime = RowsAtCompileTime == 1 || ColsAtCompileTime == 1, - /**< This is set to true if either the number of rows or the number of - * columns is known at compile-time to be equal to 1. Indeed, in that case, - * we are dealing with a column-vector (if there is only one column) or with - * a row-vector (if there is only one row). */ - - Flags = internal::traits::Flags, - /**< This stores expression \ref flags flags which may or may not be inherited by new expressions - * constructed from this one. See the \ref flags "list of flags". - */ - - IsRowMajor = Flags&RowMajorBit ? 1 : 0, - - InnerSizeAtCompileTime = int(IsVectorAtCompileTime) ? int(SizeAtCompileTime) - : int(IsRowMajor) ? int(ColsAtCompileTime) : int(RowsAtCompileTime), - - #ifndef EIGEN_PARSED_BY_DOXYGEN - _HasDirectAccess = (int(Flags)&DirectAccessBit) ? 1 : 0 // workaround sunCC - #endif - }; - - /** \internal the return type of MatrixBase::adjoint() */ - typedef typename internal::conditional::IsComplex, - CwiseUnaryOp, Eigen::Transpose >, - Transpose - >::type AdjointReturnType; - typedef Transpose TransposeReturnType; - typedef typename internal::add_const >::type ConstTransposeReturnType; - - // FIXME storage order do not match evaluator storage order - typedef SparseMatrix PlainObject; - -#ifndef EIGEN_PARSED_BY_DOXYGEN - /** This is the "real scalar" type; if the \a Scalar type is already real numbers - * (e.g. int, float or double) then \a RealScalar is just the same as \a Scalar. If - * \a Scalar is \a std::complex then RealScalar is \a T. - * - * \sa class NumTraits - */ - typedef typename NumTraits::Real RealScalar; - - /** \internal the return type of coeff() - */ - typedef typename internal::conditional<_HasDirectAccess, const Scalar&, Scalar>::type CoeffReturnType; - - /** \internal Represents a matrix with all coefficients equal to one another*/ - typedef CwiseNullaryOp,Matrix > ConstantReturnType; - - /** type of the equivalent dense matrix */ - typedef Matrix DenseMatrixType; - /** type of the equivalent square matrix */ - typedef Matrix SquareMatrixType; - - inline const Derived& derived() const { return *static_cast(this); } - inline Derived& derived() { return *static_cast(this); } - inline Derived& const_cast_derived() const - { return *static_cast(const_cast(this)); } - - typedef EigenBase Base; - -#endif // not EIGEN_PARSED_BY_DOXYGEN - -#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::SparseMatrixBase -#ifdef EIGEN_PARSED_BY_DOXYGEN -#define EIGEN_DOC_UNARY_ADDONS(METHOD,OP) /**

This method does not change the sparsity of \c *this: the OP is applied to explicitly stored coefficients only. \sa SparseCompressedBase::coeffs()

*/ -#define EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL /**

\warning This method returns a read-only expression for any sparse matrices. \sa \ref TutorialSparse_SubMatrices "Sparse block operations"

*/ -#define EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(COND) /**

\warning This method returns a read-write expression for COND sparse matrices only. Otherwise, the returned expression is read-only. \sa \ref TutorialSparse_SubMatrices "Sparse block operations"

*/ -#else -#define EIGEN_DOC_UNARY_ADDONS(X,Y) -#define EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -#define EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(COND) -#endif -# include "../plugins/CommonCwiseUnaryOps.h" -# include "../plugins/CommonCwiseBinaryOps.h" -# include "../plugins/MatrixCwiseUnaryOps.h" -# include "../plugins/MatrixCwiseBinaryOps.h" -# include "../plugins/BlockMethods.h" -# ifdef EIGEN_SPARSEMATRIXBASE_PLUGIN -# include EIGEN_SPARSEMATRIXBASE_PLUGIN -# endif -#undef EIGEN_CURRENT_STORAGE_BASE_CLASS -#undef EIGEN_DOC_UNARY_ADDONS -#undef EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -#undef EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF - - /** \returns the number of rows. \sa cols() */ - inline Index rows() const { return derived().rows(); } - /** \returns the number of columns. \sa rows() */ - inline Index cols() const { return derived().cols(); } - /** \returns the number of coefficients, which is \a rows()*cols(). - * \sa rows(), cols(). */ - inline Index size() const { return rows() * cols(); } - /** \returns true if either the number of rows or the number of columns is equal to 1. - * In other words, this function returns - * \code rows()==1 || cols()==1 \endcode - * \sa rows(), cols(), IsVectorAtCompileTime. */ - inline bool isVector() const { return rows()==1 || cols()==1; } - /** \returns the size of the storage major dimension, - * i.e., the number of columns for a columns major matrix, and the number of rows otherwise */ - Index outerSize() const { return (int(Flags)&RowMajorBit) ? this->rows() : this->cols(); } - /** \returns the size of the inner dimension according to the storage order, - * i.e., the number of rows for a columns major matrix, and the number of cols otherwise */ - Index innerSize() const { return (int(Flags)&RowMajorBit) ? this->cols() : this->rows(); } - - bool isRValue() const { return m_isRValue; } - Derived& markAsRValue() { m_isRValue = true; return derived(); } - - SparseMatrixBase() : m_isRValue(false) { /* TODO check flags */ } - - - template - Derived& operator=(const ReturnByValue& other); - - template - inline Derived& operator=(const SparseMatrixBase& other); - - inline Derived& operator=(const Derived& other); - - protected: - - template - inline Derived& assign(const OtherDerived& other); - - template - inline void assignGeneric(const OtherDerived& other); - - public: - - friend std::ostream & operator << (std::ostream & s, const SparseMatrixBase& m) - { - typedef typename Derived::Nested Nested; - typedef typename internal::remove_all::type NestedCleaned; - - if (Flags&RowMajorBit) - { - Nested nm(m.derived()); - internal::evaluator thisEval(nm); - for (Index row=0; row::InnerIterator it(thisEval, row); it; ++it) - { - for ( ; col thisEval(nm); - if (m.cols() == 1) { - Index row = 0; - for (typename internal::evaluator::InnerIterator it(thisEval, 0); it; ++it) - { - for ( ; row trans = m; - s << static_cast >&>(trans); - } - } - return s; - } - - template - Derived& operator+=(const SparseMatrixBase& other); - template - Derived& operator-=(const SparseMatrixBase& other); - - template - Derived& operator+=(const DiagonalBase& other); - template - Derived& operator-=(const DiagonalBase& other); - - template - Derived& operator+=(const EigenBase &other); - template - Derived& operator-=(const EigenBase &other); - - Derived& operator*=(const Scalar& other); - Derived& operator/=(const Scalar& other); - - template struct CwiseProductDenseReturnType { - typedef CwiseBinaryOp::Scalar, - typename internal::traits::Scalar - >::ReturnType>, - const Derived, - const OtherDerived - > Type; - }; - - template - EIGEN_STRONG_INLINE const typename CwiseProductDenseReturnType::Type - cwiseProduct(const MatrixBase &other) const; - - // sparse * diagonal - template - const Product - operator*(const DiagonalBase &other) const - { return Product(derived(), other.derived()); } - - // diagonal * sparse - template friend - const Product - operator*(const DiagonalBase &lhs, const SparseMatrixBase& rhs) - { return Product(lhs.derived(), rhs.derived()); } - - // sparse * sparse - template - const Product - operator*(const SparseMatrixBase &other) const; - - // sparse * dense - template - const Product - operator*(const MatrixBase &other) const - { return Product(derived(), other.derived()); } - - // dense * sparse - template friend - const Product - operator*(const MatrixBase &lhs, const SparseMatrixBase& rhs) - { return Product(lhs.derived(), rhs.derived()); } - - /** \returns an expression of P H P^-1 where H is the matrix represented by \c *this */ - SparseSymmetricPermutationProduct twistedBy(const PermutationMatrix& perm) const - { - return SparseSymmetricPermutationProduct(derived(), perm); - } - - template - Derived& operator*=(const SparseMatrixBase& other); - - template - inline const TriangularView triangularView() const; - - template struct SelfAdjointViewReturnType { typedef SparseSelfAdjointView Type; }; - template struct ConstSelfAdjointViewReturnType { typedef const SparseSelfAdjointView Type; }; - - template inline - typename ConstSelfAdjointViewReturnType::Type selfadjointView() const; - template inline - typename SelfAdjointViewReturnType::Type selfadjointView(); - - template Scalar dot(const MatrixBase& other) const; - template Scalar dot(const SparseMatrixBase& other) const; - RealScalar squaredNorm() const; - RealScalar norm() const; - RealScalar blueNorm() const; - - TransposeReturnType transpose() { return TransposeReturnType(derived()); } - const ConstTransposeReturnType transpose() const { return ConstTransposeReturnType(derived()); } - const AdjointReturnType adjoint() const { return AdjointReturnType(transpose()); } - - // inner-vector - typedef Block InnerVectorReturnType; - typedef Block ConstInnerVectorReturnType; - InnerVectorReturnType innerVector(Index outer); - const ConstInnerVectorReturnType innerVector(Index outer) const; - - // set of inner-vectors - typedef Block InnerVectorsReturnType; - typedef Block ConstInnerVectorsReturnType; - InnerVectorsReturnType innerVectors(Index outerStart, Index outerSize); - const ConstInnerVectorsReturnType innerVectors(Index outerStart, Index outerSize) const; - - DenseMatrixType toDense() const - { - return DenseMatrixType(derived()); - } - - template - bool isApprox(const SparseMatrixBase& other, - const RealScalar& prec = NumTraits::dummy_precision()) const; - - template - bool isApprox(const MatrixBase& other, - const RealScalar& prec = NumTraits::dummy_precision()) const - { return toDense().isApprox(other,prec); } - - /** \returns the matrix or vector obtained by evaluating this expression. - * - * Notice that in the case of a plain matrix or vector (not an expression) this function just returns - * a const reference, in order to avoid a useless copy. - */ - inline const typename internal::eval::type eval() const - { return typename internal::eval::type(derived()); } - - Scalar sum() const; - - inline const SparseView - pruned(const Scalar& reference = Scalar(0), const RealScalar& epsilon = NumTraits::dummy_precision()) const; - - protected: - - bool m_isRValue; - - static inline StorageIndex convert_index(const Index idx) { - return internal::convert_index(idx); - } - private: - template void evalTo(Dest &) const; -}; - -} // end namespace Eigen - -#endif // EIGEN_SPARSEMATRIXBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparsePermutation.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparsePermutation.h deleted file mode 100644 index ef38357aefd..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparsePermutation.h +++ /dev/null @@ -1,178 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_PERMUTATION_H -#define EIGEN_SPARSE_PERMUTATION_H - -// This file implements sparse * permutation products - -namespace Eigen { - -namespace internal { - -template -struct permutation_matrix_product -{ - typedef typename nested_eval::type MatrixType; - typedef typename remove_all::type MatrixTypeCleaned; - - typedef typename MatrixTypeCleaned::Scalar Scalar; - typedef typename MatrixTypeCleaned::StorageIndex StorageIndex; - - enum { - SrcStorageOrder = MatrixTypeCleaned::Flags&RowMajorBit ? RowMajor : ColMajor, - MoveOuter = SrcStorageOrder==RowMajor ? Side==OnTheLeft : Side==OnTheRight - }; - - typedef typename internal::conditional, - SparseMatrix >::type ReturnType; - - template - static inline void run(Dest& dst, const PermutationType& perm, const ExpressionType& xpr) - { - MatrixType mat(xpr); - if(MoveOuter) - { - SparseMatrix tmp(mat.rows(), mat.cols()); - Matrix sizes(mat.outerSize()); - for(Index j=0; j tmp(mat.rows(), mat.cols()); - Matrix sizes(tmp.outerSize()); - sizes.setZero(); - PermutationMatrix perm_cpy; - if((Side==OnTheLeft) ^ Transposed) - perm_cpy = perm; - else - perm_cpy = perm.transpose(); - - for(Index j=0; j struct product_promote_storage_type { typedef Sparse ret; }; -template struct product_promote_storage_type { typedef Sparse ret; }; - -// TODO, the following two overloads are only needed to define the right temporary type through -// typename traits >::ReturnType -// whereas it should be correctly handled by traits >::PlainObject - -template -struct product_evaluator, ProductTag, PermutationShape, SparseShape> - : public evaluator::ReturnType> -{ - typedef Product XprType; - typedef typename permutation_matrix_product::ReturnType PlainObject; - typedef evaluator Base; - - enum { - Flags = Base::Flags | EvalBeforeNestingBit - }; - - explicit product_evaluator(const XprType& xpr) - : m_result(xpr.rows(), xpr.cols()) - { - ::new (static_cast(this)) Base(m_result); - generic_product_impl::evalTo(m_result, xpr.lhs(), xpr.rhs()); - } - -protected: - PlainObject m_result; -}; - -template -struct product_evaluator, ProductTag, SparseShape, PermutationShape > - : public evaluator::ReturnType> -{ - typedef Product XprType; - typedef typename permutation_matrix_product::ReturnType PlainObject; - typedef evaluator Base; - - enum { - Flags = Base::Flags | EvalBeforeNestingBit - }; - - explicit product_evaluator(const XprType& xpr) - : m_result(xpr.rows(), xpr.cols()) - { - ::new (static_cast(this)) Base(m_result); - generic_product_impl::evalTo(m_result, xpr.lhs(), xpr.rhs()); - } - -protected: - PlainObject m_result; -}; - -} // end namespace internal - -/** \returns the matrix with the permutation applied to the columns - */ -template -inline const Product -operator*(const SparseMatrixBase& matrix, const PermutationBase& perm) -{ return Product(matrix.derived(), perm.derived()); } - -/** \returns the matrix with the permutation applied to the rows - */ -template -inline const Product -operator*( const PermutationBase& perm, const SparseMatrixBase& matrix) -{ return Product(perm.derived(), matrix.derived()); } - - -/** \returns the matrix with the inverse permutation applied to the columns. - */ -template -inline const Product, AliasFreeProduct> -operator*(const SparseMatrixBase& matrix, const InverseImpl& tperm) -{ - return Product, AliasFreeProduct>(matrix.derived(), tperm.derived()); -} - -/** \returns the matrix with the inverse permutation applied to the rows. - */ -template -inline const Product, SparseDerived, AliasFreeProduct> -operator*(const InverseImpl& tperm, const SparseMatrixBase& matrix) -{ - return Product, SparseDerived, AliasFreeProduct>(tperm.derived(), matrix.derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_SELFADJOINTVIEW_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseProduct.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseProduct.h deleted file mode 100644 index 4cbf68781a8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseProduct.h +++ /dev/null @@ -1,169 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEPRODUCT_H -#define EIGEN_SPARSEPRODUCT_H - -namespace Eigen { - -/** \returns an expression of the product of two sparse matrices. - * By default a conservative product preserving the symbolic non zeros is performed. - * The automatic pruning of the small values can be achieved by calling the pruned() function - * in which case a totally different product algorithm is employed: - * \code - * C = (A*B).pruned(); // supress numerical zeros (exact) - * C = (A*B).pruned(ref); - * C = (A*B).pruned(ref,epsilon); - * \endcode - * where \c ref is a meaningful non zero reference value. - * */ -template -template -inline const Product -SparseMatrixBase::operator*(const SparseMatrixBase &other) const -{ - return Product(derived(), other.derived()); -} - -namespace internal { - -// sparse * sparse -template -struct generic_product_impl -{ - template - static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) - { - evalTo(dst, lhs, rhs, typename evaluator_traits::Shape()); - } - - // dense += sparse * sparse - template - static void addTo(Dest& dst, const ActualLhs& lhs, const Rhs& rhs, typename enable_if::Shape,DenseShape>::value,int*>::type* = 0) - { - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhs); - RhsNested rhsNested(rhs); - internal::sparse_sparse_to_dense_product_selector::type, - typename remove_all::type, Dest>::run(lhsNested,rhsNested,dst); - } - - // dense -= sparse * sparse - template - static void subTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, typename enable_if::Shape,DenseShape>::value,int*>::type* = 0) - { - addTo(dst, -lhs, rhs); - } - -protected: - - // sparse = sparse * sparse - template - static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, SparseShape) - { - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhs); - RhsNested rhsNested(rhs); - internal::conservative_sparse_sparse_product_selector::type, - typename remove_all::type, Dest>::run(lhsNested,rhsNested,dst); - } - - // dense = sparse * sparse - template - static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, DenseShape) - { - dst.setZero(); - addTo(dst, lhs, rhs); - } -}; - -// sparse * sparse-triangular -template -struct generic_product_impl - : public generic_product_impl -{}; - -// sparse-triangular * sparse -template -struct generic_product_impl - : public generic_product_impl -{}; - -// dense = sparse-product (can be sparse*sparse, sparse*perm, etc.) -template< typename DstXprType, typename Lhs, typename Rhs> -struct Assignment, internal::assign_op::Scalar>, Sparse2Dense> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - generic_product_impl::evalTo(dst,src.lhs(),src.rhs()); - } -}; - -// dense += sparse-product (can be sparse*sparse, sparse*perm, etc.) -template< typename DstXprType, typename Lhs, typename Rhs> -struct Assignment, internal::add_assign_op::Scalar>, Sparse2Dense> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &) - { - generic_product_impl::addTo(dst,src.lhs(),src.rhs()); - } -}; - -// dense -= sparse-product (can be sparse*sparse, sparse*perm, etc.) -template< typename DstXprType, typename Lhs, typename Rhs> -struct Assignment, internal::sub_assign_op::Scalar>, Sparse2Dense> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &) - { - generic_product_impl::subTo(dst,src.lhs(),src.rhs()); - } -}; - -template -struct unary_evaluator >, IteratorBased> - : public evaluator::PlainObject> -{ - typedef SparseView > XprType; - typedef typename XprType::PlainObject PlainObject; - typedef evaluator Base; - - explicit unary_evaluator(const XprType& xpr) - : m_result(xpr.rows(), xpr.cols()) - { - using std::abs; - ::new (static_cast(this)) Base(m_result); - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(xpr.nestedExpression().lhs()); - RhsNested rhsNested(xpr.nestedExpression().rhs()); - - internal::sparse_sparse_product_with_pruning_selector::type, - typename remove_all::type, PlainObject>::run(lhsNested,rhsNested,m_result, - abs(xpr.reference())*xpr.epsilon()); - } - -protected: - PlainObject m_result; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSEPRODUCT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRedux.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRedux.h deleted file mode 100644 index 45877496270..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRedux.h +++ /dev/null @@ -1,49 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEREDUX_H -#define EIGEN_SPARSEREDUX_H - -namespace Eigen { - -template -typename internal::traits::Scalar -SparseMatrixBase::sum() const -{ - eigen_assert(rows()>0 && cols()>0 && "you are using a non initialized matrix"); - Scalar res(0); - internal::evaluator thisEval(derived()); - for (Index j=0; j::InnerIterator iter(thisEval,j); iter; ++iter) - res += iter.value(); - return res; -} - -template -typename internal::traits >::Scalar -SparseMatrix<_Scalar,_Options,_Index>::sum() const -{ - eigen_assert(rows()>0 && cols()>0 && "you are using a non initialized matrix"); - if(this->isCompressed()) - return Matrix::Map(m_data.valuePtr(), m_data.size()).sum(); - else - return Base::sum(); -} - -template -typename internal::traits >::Scalar -SparseVector<_Scalar,_Options,_Index>::sum() const -{ - eigen_assert(rows()>0 && cols()>0 && "you are using a non initialized matrix"); - return Matrix::Map(m_data.valuePtr(), m_data.size()).sum(); -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSEREDUX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRef.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRef.h deleted file mode 100644 index d91f38f97ce..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseRef.h +++ /dev/null @@ -1,397 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_REF_H -#define EIGEN_SPARSE_REF_H - -namespace Eigen { - -enum { - StandardCompressedFormat = 2 /**< used by Ref to specify whether the input storage must be in standard compressed form */ -}; - -namespace internal { - -template class SparseRefBase; - -template -struct traits, _Options, _StrideType> > - : public traits > -{ - typedef SparseMatrix PlainObjectType; - enum { - Options = _Options, - Flags = traits::Flags | CompressedAccessBit | NestByRefBit - }; - - template struct match { - enum { - StorageOrderMatch = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), - MatchAtCompileTime = (Derived::Flags&CompressedAccessBit) && StorageOrderMatch - }; - typedef typename internal::conditional::type type; - }; - -}; - -template -struct traits, _Options, _StrideType> > - : public traits, _Options, _StrideType> > -{ - enum { - Flags = (traits >::Flags | CompressedAccessBit | NestByRefBit) & ~LvalueBit - }; -}; - -template -struct traits, _Options, _StrideType> > - : public traits > -{ - typedef SparseVector PlainObjectType; - enum { - Options = _Options, - Flags = traits::Flags | CompressedAccessBit | NestByRefBit - }; - - template struct match { - enum { - MatchAtCompileTime = (Derived::Flags&CompressedAccessBit) && Derived::IsVectorAtCompileTime - }; - typedef typename internal::conditional::type type; - }; - -}; - -template -struct traits, _Options, _StrideType> > - : public traits, _Options, _StrideType> > -{ - enum { - Flags = (traits >::Flags | CompressedAccessBit | NestByRefBit) & ~LvalueBit - }; -}; - -template -struct traits > : public traits {}; - -template class SparseRefBase - : public SparseMapBase -{ -public: - - typedef SparseMapBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(SparseRefBase) - - SparseRefBase() - : Base(RowsAtCompileTime==Dynamic?0:RowsAtCompileTime,ColsAtCompileTime==Dynamic?0:ColsAtCompileTime, 0, 0, 0, 0, 0) - {} - -protected: - - template - void construct(Expression& expr) - { - if(expr.outerIndexPtr()==0) - ::new (static_cast(this)) Base(expr.size(), expr.nonZeros(), expr.innerIndexPtr(), expr.valuePtr()); - else - ::new (static_cast(this)) Base(expr.rows(), expr.cols(), expr.nonZeros(), expr.outerIndexPtr(), expr.innerIndexPtr(), expr.valuePtr(), expr.innerNonZeroPtr()); - } -}; - -} // namespace internal - - -/** - * \ingroup SparseCore_Module - * - * \brief A sparse matrix expression referencing an existing sparse expression - * - * \tparam SparseMatrixType the equivalent sparse matrix type of the referenced data, it must be a template instance of class SparseMatrix. - * \tparam Options specifies whether the a standard compressed format is required \c Options is \c #StandardCompressedFormat, or \c 0. - * The default is \c 0. - * - * \sa class Ref - */ -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -class Ref, Options, StrideType > - : public internal::SparseRefBase, Options, StrideType > > -#else -template -class Ref - : public SparseMapBase // yes, that's weird to use Derived here, but that works! -#endif -{ - typedef SparseMatrix PlainObjectType; - typedef internal::traits Traits; - template - inline Ref(const SparseMatrix& expr); - template - inline Ref(const MappedSparseMatrix& expr); - public: - - typedef internal::SparseRefBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Ref) - - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - inline Ref(SparseMatrix& expr) - { - EIGEN_STATIC_ASSERT(bool(Traits::template match >::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); - eigen_assert( ((Options & int(StandardCompressedFormat))==0) || (expr.isCompressed()) ); - Base::construct(expr.derived()); - } - - template - inline Ref(MappedSparseMatrix& expr) - { - EIGEN_STATIC_ASSERT(bool(Traits::template match >::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); - eigen_assert( ((Options & int(StandardCompressedFormat))==0) || (expr.isCompressed()) ); - Base::construct(expr.derived()); - } - - template - inline Ref(const SparseCompressedBase& expr) - #else - /** Implicit constructor from any sparse expression (2D matrix or 1D vector) */ - template - inline Ref(SparseCompressedBase& expr) - #endif - { - EIGEN_STATIC_ASSERT(bool(internal::is_lvalue::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); - EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); - eigen_assert( ((Options & int(StandardCompressedFormat))==0) || (expr.isCompressed()) ); - Base::construct(expr.const_cast_derived()); - } -}; - -// this is the const ref version -template -class Ref, Options, StrideType> - : public internal::SparseRefBase, Options, StrideType> > -{ - typedef SparseMatrix TPlainObjectType; - typedef internal::traits Traits; - public: - - typedef internal::SparseRefBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Ref) - - template - inline Ref(const SparseMatrixBase& expr) : m_hasCopy(false) - { - construct(expr.derived(), typename Traits::template match::type()); - } - - inline Ref(const Ref& other) : Base(other), m_hasCopy(false) { - // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy - } - - template - inline Ref(const RefBase& other) : m_hasCopy(false) { - construct(other.derived(), typename Traits::template match::type()); - } - - ~Ref() { - if(m_hasCopy) { - TPlainObjectType* obj = reinterpret_cast(m_object_bytes); - obj->~TPlainObjectType(); - } - } - - protected: - - template - void construct(const Expression& expr,internal::true_type) - { - if((Options & int(StandardCompressedFormat)) && (!expr.isCompressed())) - { - TPlainObjectType* obj = reinterpret_cast(m_object_bytes); - ::new (obj) TPlainObjectType(expr); - m_hasCopy = true; - Base::construct(*obj); - } - else - { - Base::construct(expr); - } - } - - template - void construct(const Expression& expr, internal::false_type) - { - TPlainObjectType* obj = reinterpret_cast(m_object_bytes); - ::new (obj) TPlainObjectType(expr); - m_hasCopy = true; - Base::construct(*obj); - } - - protected: - char m_object_bytes[sizeof(TPlainObjectType)]; - bool m_hasCopy; -}; - - - -/** - * \ingroup SparseCore_Module - * - * \brief A sparse vector expression referencing an existing sparse vector expression - * - * \tparam SparseVectorType the equivalent sparse vector type of the referenced data, it must be a template instance of class SparseVector. - * - * \sa class Ref - */ -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -class Ref, Options, StrideType > - : public internal::SparseRefBase, Options, StrideType > > -#else -template -class Ref - : public SparseMapBase -#endif -{ - typedef SparseVector PlainObjectType; - typedef internal::traits Traits; - template - inline Ref(const SparseVector& expr); - public: - - typedef internal::SparseRefBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Ref) - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - inline Ref(SparseVector& expr) - { - EIGEN_STATIC_ASSERT(bool(Traits::template match >::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); - Base::construct(expr.derived()); - } - - template - inline Ref(const SparseCompressedBase& expr) - #else - /** Implicit constructor from any 1D sparse vector expression */ - template - inline Ref(SparseCompressedBase& expr) - #endif - { - EIGEN_STATIC_ASSERT(bool(internal::is_lvalue::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); - EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); - Base::construct(expr.const_cast_derived()); - } -}; - -// this is the const ref version -template -class Ref, Options, StrideType> - : public internal::SparseRefBase, Options, StrideType> > -{ - typedef SparseVector TPlainObjectType; - typedef internal::traits Traits; - public: - - typedef internal::SparseRefBase Base; - EIGEN_SPARSE_PUBLIC_INTERFACE(Ref) - - template - inline Ref(const SparseMatrixBase& expr) : m_hasCopy(false) - { - construct(expr.derived(), typename Traits::template match::type()); - } - - inline Ref(const Ref& other) : Base(other), m_hasCopy(false) { - // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy - } - - template - inline Ref(const RefBase& other) : m_hasCopy(false) { - construct(other.derived(), typename Traits::template match::type()); - } - - ~Ref() { - if(m_hasCopy) { - TPlainObjectType* obj = reinterpret_cast(m_object_bytes); - obj->~TPlainObjectType(); - } - } - - protected: - - template - void construct(const Expression& expr,internal::true_type) - { - Base::construct(expr); - } - - template - void construct(const Expression& expr, internal::false_type) - { - TPlainObjectType* obj = reinterpret_cast(m_object_bytes); - ::new (obj) TPlainObjectType(expr); - m_hasCopy = true; - Base::construct(*obj); - } - - protected: - char m_object_bytes[sizeof(TPlainObjectType)]; - bool m_hasCopy; -}; - -namespace internal { - -// FIXME shall we introduce a general evaluatior_ref that we can specialize for any sparse object once, and thus remove this copy-pasta thing... - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Ref, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Ref, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Ref, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -template -struct evaluator, Options, StrideType> > - : evaluator, Options, StrideType> > > -{ - typedef evaluator, Options, StrideType> > > Base; - typedef Ref, Options, StrideType> XprType; - evaluator() : Base() {} - explicit evaluator(const XprType &mat) : Base(mat) {} -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_REF_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSelfAdjointView.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSelfAdjointView.h deleted file mode 100644 index 76117a0107e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSelfAdjointView.h +++ /dev/null @@ -1,656 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_SELFADJOINTVIEW_H -#define EIGEN_SPARSE_SELFADJOINTVIEW_H - -namespace Eigen { - -/** \ingroup SparseCore_Module - * \class SparseSelfAdjointView - * - * \brief Pseudo expression to manipulate a triangular sparse matrix as a selfadjoint matrix. - * - * \param MatrixType the type of the dense matrix storing the coefficients - * \param Mode can be either \c #Lower or \c #Upper - * - * This class is an expression of a sefladjoint matrix from a triangular part of a matrix - * with given dense storage of the coefficients. It is the return type of MatrixBase::selfadjointView() - * and most of the time this is the only way that it is used. - * - * \sa SparseMatrixBase::selfadjointView() - */ -namespace internal { - -template -struct traits > : traits { -}; - -template -void permute_symm_to_symm(const MatrixType& mat, SparseMatrix& _dest, const typename MatrixType::StorageIndex* perm = 0); - -template -void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix& _dest, const typename MatrixType::StorageIndex* perm = 0); - -} - -template class SparseSelfAdjointView - : public EigenBase > -{ - public: - - enum { - Mode = _Mode, - TransposeMode = ((Mode & Upper) ? Lower : 0) | ((Mode & Lower) ? Upper : 0), - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime - }; - - typedef EigenBase Base; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Matrix VectorI; - typedef typename internal::ref_selector::non_const_type MatrixTypeNested; - typedef typename internal::remove_all::type _MatrixTypeNested; - - explicit inline SparseSelfAdjointView(MatrixType& matrix) : m_matrix(matrix) - { - eigen_assert(rows()==cols() && "SelfAdjointView is only for squared matrices"); - } - - inline Index rows() const { return m_matrix.rows(); } - inline Index cols() const { return m_matrix.cols(); } - - /** \internal \returns a reference to the nested matrix */ - const _MatrixTypeNested& matrix() const { return m_matrix; } - typename internal::remove_reference::type& matrix() { return m_matrix; } - - /** \returns an expression of the matrix product between a sparse self-adjoint matrix \c *this and a sparse matrix \a rhs. - * - * Note that there is no algorithmic advantage of performing such a product compared to a general sparse-sparse matrix product. - * Indeed, the SparseSelfadjointView operand is first copied into a temporary SparseMatrix before computing the product. - */ - template - Product - operator*(const SparseMatrixBase& rhs) const - { - return Product(*this, rhs.derived()); - } - - /** \returns an expression of the matrix product between a sparse matrix \a lhs and a sparse self-adjoint matrix \a rhs. - * - * Note that there is no algorithmic advantage of performing such a product compared to a general sparse-sparse matrix product. - * Indeed, the SparseSelfadjointView operand is first copied into a temporary SparseMatrix before computing the product. - */ - template friend - Product - operator*(const SparseMatrixBase& lhs, const SparseSelfAdjointView& rhs) - { - return Product(lhs.derived(), rhs); - } - - /** Efficient sparse self-adjoint matrix times dense vector/matrix product */ - template - Product - operator*(const MatrixBase& rhs) const - { - return Product(*this, rhs.derived()); - } - - /** Efficient dense vector/matrix times sparse self-adjoint matrix product */ - template friend - Product - operator*(const MatrixBase& lhs, const SparseSelfAdjointView& rhs) - { - return Product(lhs.derived(), rhs); - } - - /** Perform a symmetric rank K update of the selfadjoint matrix \c *this: - * \f$ this = this + \alpha ( u u^* ) \f$ where \a u is a vector or matrix. - * - * \returns a reference to \c *this - * - * To perform \f$ this = this + \alpha ( u^* u ) \f$ you can simply - * call this function with u.adjoint(). - */ - template - SparseSelfAdjointView& rankUpdate(const SparseMatrixBase& u, const Scalar& alpha = Scalar(1)); - - /** \returns an expression of P H P^-1 */ - // TODO implement twists in a more evaluator friendly fashion - SparseSymmetricPermutationProduct<_MatrixTypeNested,Mode> twistedBy(const PermutationMatrix& perm) const - { - return SparseSymmetricPermutationProduct<_MatrixTypeNested,Mode>(m_matrix, perm); - } - - template - SparseSelfAdjointView& operator=(const SparseSymmetricPermutationProduct& permutedMatrix) - { - internal::call_assignment_no_alias_no_transpose(*this, permutedMatrix); - return *this; - } - - SparseSelfAdjointView& operator=(const SparseSelfAdjointView& src) - { - PermutationMatrix pnull; - return *this = src.twistedBy(pnull); - } - - template - SparseSelfAdjointView& operator=(const SparseSelfAdjointView& src) - { - PermutationMatrix pnull; - return *this = src.twistedBy(pnull); - } - - void resize(Index rows, Index cols) - { - EIGEN_ONLY_USED_FOR_DEBUG(rows); - EIGEN_ONLY_USED_FOR_DEBUG(cols); - eigen_assert(rows == this->rows() && cols == this->cols() - && "SparseSelfadjointView::resize() does not actually allow to resize."); - } - - protected: - - MatrixTypeNested m_matrix; - //mutable VectorI m_countPerRow; - //mutable VectorI m_countPerCol; - private: - template void evalTo(Dest &) const; -}; - -/*************************************************************************** -* Implementation of SparseMatrixBase methods -***************************************************************************/ - -template -template -typename SparseMatrixBase::template ConstSelfAdjointViewReturnType::Type SparseMatrixBase::selfadjointView() const -{ - return SparseSelfAdjointView(derived()); -} - -template -template -typename SparseMatrixBase::template SelfAdjointViewReturnType::Type SparseMatrixBase::selfadjointView() -{ - return SparseSelfAdjointView(derived()); -} - -/*************************************************************************** -* Implementation of SparseSelfAdjointView methods -***************************************************************************/ - -template -template -SparseSelfAdjointView& -SparseSelfAdjointView::rankUpdate(const SparseMatrixBase& u, const Scalar& alpha) -{ - SparseMatrix tmp = u * u.adjoint(); - if(alpha==Scalar(0)) - m_matrix = tmp.template triangularView(); - else - m_matrix += alpha * tmp.template triangularView(); - - return *this; -} - -namespace internal { - -// TODO currently a selfadjoint expression has the form SelfAdjointView<.,.> -// in the future selfadjoint-ness should be defined by the expression traits -// such that Transpose > is valid. (currently TriangularBase::transpose() is overloaded to make it work) -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef SparseSelfAdjointShape Shape; -}; - -struct SparseSelfAdjoint2Sparse {}; - -template<> struct AssignmentKind { typedef SparseSelfAdjoint2Sparse Kind; }; -template<> struct AssignmentKind { typedef Sparse2Sparse Kind; }; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - typedef typename DstXprType::StorageIndex StorageIndex; - typedef internal::assign_op AssignOpType; - - template - static void run(SparseMatrix &dst, const SrcXprType &src, const AssignOpType&/*func*/) - { - internal::permute_symm_to_fullsymm(src.matrix(), dst); - } - - // FIXME: the handling of += and -= in sparse matrices should be cleanup so that next two overloads could be reduced to: - template - static void run(SparseMatrix &dst, const SrcXprType &src, const AssignFunc& func) - { - SparseMatrix tmp(src.rows(),src.cols()); - run(tmp, src, AssignOpType()); - call_assignment_no_alias_no_transpose(dst, tmp, func); - } - - template - static void run(SparseMatrix &dst, const SrcXprType &src, - const internal::add_assign_op& /* func */) - { - SparseMatrix tmp(src.rows(),src.cols()); - run(tmp, src, AssignOpType()); - dst += tmp; - } - - template - static void run(SparseMatrix &dst, const SrcXprType &src, - const internal::sub_assign_op& /* func */) - { - SparseMatrix tmp(src.rows(),src.cols()); - run(tmp, src, AssignOpType()); - dst -= tmp; - } - - template - static void run(DynamicSparseMatrix& dst, const SrcXprType &src, const AssignOpType&/*func*/) - { - // TODO directly evaluate into dst; - SparseMatrix tmp(dst.rows(),dst.cols()); - internal::permute_symm_to_fullsymm(src.matrix(), tmp); - dst = tmp; - } -}; - -} // end namespace internal - -/*************************************************************************** -* Implementation of sparse self-adjoint time dense matrix -***************************************************************************/ - -namespace internal { - -template -inline void sparse_selfadjoint_time_dense_product(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const AlphaType& alpha) -{ - EIGEN_ONLY_USED_FOR_DEBUG(alpha); - - typedef typename internal::nested_eval::type SparseLhsTypeNested; - typedef typename internal::remove_all::type SparseLhsTypeNestedCleaned; - typedef evaluator LhsEval; - typedef typename LhsEval::InnerIterator LhsIterator; - typedef typename SparseLhsType::Scalar LhsScalar; - - enum { - LhsIsRowMajor = (LhsEval::Flags&RowMajorBit)==RowMajorBit, - ProcessFirstHalf = - ((Mode&(Upper|Lower))==(Upper|Lower)) - || ( (Mode&Upper) && !LhsIsRowMajor) - || ( (Mode&Lower) && LhsIsRowMajor), - ProcessSecondHalf = !ProcessFirstHalf - }; - - SparseLhsTypeNested lhs_nested(lhs); - LhsEval lhsEval(lhs_nested); - - // work on one column at once - for (Index k=0; k::ReturnType rhs_j(alpha*rhs(j,k)); - // accumulator for partial scalar product - typename DenseResType::Scalar res_j(0); - for(; (ProcessFirstHalf ? i && i.index() < j : i) ; ++i) - { - LhsScalar lhs_ij = i.value(); - if(!LhsIsRowMajor) lhs_ij = numext::conj(lhs_ij); - res_j += lhs_ij * rhs.coeff(i.index(),k); - res(i.index(),k) += numext::conj(lhs_ij) * rhs_j; - } - res.coeffRef(j,k) += alpha * res_j; - - // handle diagonal coeff - if (ProcessFirstHalf && i && (i.index()==j)) - res.coeffRef(j,k) += alpha * i.value() * rhs.coeff(j,k); - } - } -} - - -template -struct generic_product_impl -: generic_product_impl_base > -{ - template - static void scaleAndAddTo(Dest& dst, const LhsView& lhsView, const Rhs& rhs, const typename Dest::Scalar& alpha) - { - typedef typename LhsView::_MatrixTypeNested Lhs; - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhsView.matrix()); - RhsNested rhsNested(rhs); - - internal::sparse_selfadjoint_time_dense_product(lhsNested, rhsNested, dst, alpha); - } -}; - -template -struct generic_product_impl -: generic_product_impl_base > -{ - template - static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const RhsView& rhsView, const typename Dest::Scalar& alpha) - { - typedef typename RhsView::_MatrixTypeNested Rhs; - typedef typename nested_eval::type LhsNested; - typedef typename nested_eval::type RhsNested; - LhsNested lhsNested(lhs); - RhsNested rhsNested(rhsView.matrix()); - - // transpose everything - Transpose dstT(dst); - internal::sparse_selfadjoint_time_dense_product(rhsNested.transpose(), lhsNested.transpose(), dstT, alpha); - } -}; - -// NOTE: these two overloads are needed to evaluate the sparse selfadjoint view into a full sparse matrix -// TODO: maybe the copy could be handled by generic_product_impl so that these overloads would not be needed anymore - -template -struct product_evaluator, ProductTag, SparseSelfAdjointShape, SparseShape> - : public evaluator::PlainObject> -{ - typedef Product XprType; - typedef typename XprType::PlainObject PlainObject; - typedef evaluator Base; - - product_evaluator(const XprType& xpr) - : m_lhs(xpr.lhs()), m_result(xpr.rows(), xpr.cols()) - { - ::new (static_cast(this)) Base(m_result); - generic_product_impl::evalTo(m_result, m_lhs, xpr.rhs()); - } - -protected: - typename Rhs::PlainObject m_lhs; - PlainObject m_result; -}; - -template -struct product_evaluator, ProductTag, SparseShape, SparseSelfAdjointShape> - : public evaluator::PlainObject> -{ - typedef Product XprType; - typedef typename XprType::PlainObject PlainObject; - typedef evaluator Base; - - product_evaluator(const XprType& xpr) - : m_rhs(xpr.rhs()), m_result(xpr.rows(), xpr.cols()) - { - ::new (static_cast(this)) Base(m_result); - generic_product_impl::evalTo(m_result, xpr.lhs(), m_rhs); - } - -protected: - typename Lhs::PlainObject m_rhs; - PlainObject m_result; -}; - -} // namespace internal - -/*************************************************************************** -* Implementation of symmetric copies and permutations -***************************************************************************/ -namespace internal { - -template -void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix& _dest, const typename MatrixType::StorageIndex* perm) -{ - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename MatrixType::Scalar Scalar; - typedef SparseMatrix Dest; - typedef Matrix VectorI; - typedef evaluator MatEval; - typedef typename evaluator::InnerIterator MatIterator; - - MatEval matEval(mat); - Dest& dest(_dest.derived()); - enum { - StorageOrderMatch = int(Dest::IsRowMajor) == int(MatrixType::IsRowMajor) - }; - - Index size = mat.rows(); - VectorI count; - count.resize(size); - count.setZero(); - dest.resize(size,size); - for(Index j = 0; jc) || ( Mode==Upper && r(it.index()); - Index r = it.row(); - Index c = it.col(); - - StorageIndex jp = perm ? perm[j] : j; - StorageIndex ip = perm ? perm[i] : i; - - if(Mode==int(Upper|Lower)) - { - Index k = count[StorageOrderMatch ? jp : ip]++; - dest.innerIndexPtr()[k] = StorageOrderMatch ? ip : jp; - dest.valuePtr()[k] = it.value(); - } - else if(r==c) - { - Index k = count[ip]++; - dest.innerIndexPtr()[k] = ip; - dest.valuePtr()[k] = it.value(); - } - else if(( (Mode&Lower)==Lower && r>c) || ( (Mode&Upper)==Upper && r -void permute_symm_to_symm(const MatrixType& mat, SparseMatrix& _dest, const typename MatrixType::StorageIndex* perm) -{ - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename MatrixType::Scalar Scalar; - SparseMatrix& dest(_dest.derived()); - typedef Matrix VectorI; - typedef evaluator MatEval; - typedef typename evaluator::InnerIterator MatIterator; - - enum { - SrcOrder = MatrixType::IsRowMajor ? RowMajor : ColMajor, - StorageOrderMatch = int(SrcOrder) == int(DstOrder), - DstMode = DstOrder==RowMajor ? (_DstMode==Upper ? Lower : Upper) : _DstMode, - SrcMode = SrcOrder==RowMajor ? (_SrcMode==Upper ? Lower : Upper) : _SrcMode - }; - - MatEval matEval(mat); - - Index size = mat.rows(); - VectorI count(size); - count.setZero(); - dest.resize(size,size); - for(StorageIndex j = 0; jj)) - continue; - - StorageIndex ip = perm ? perm[i] : i; - count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++; - } - } - dest.outerIndexPtr()[0] = 0; - for(Index j=0; jj)) - continue; - - StorageIndex jp = perm ? perm[j] : j; - StorageIndex ip = perm? perm[i] : i; - - Index k = count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++; - dest.innerIndexPtr()[k] = int(DstMode)==int(Lower) ? (std::max)(ip,jp) : (std::min)(ip,jp); - - if(!StorageOrderMatch) std::swap(ip,jp); - if( ((int(DstMode)==int(Lower) && ipjp))) - dest.valuePtr()[k] = numext::conj(it.value()); - else - dest.valuePtr()[k] = it.value(); - } - } -} - -} - -// TODO implement twists in a more evaluator friendly fashion - -namespace internal { - -template -struct traits > : traits { -}; - -} - -template -class SparseSymmetricPermutationProduct - : public EigenBase > -{ - public: - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - enum { - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime - }; - protected: - typedef PermutationMatrix Perm; - public: - typedef Matrix VectorI; - typedef typename MatrixType::Nested MatrixTypeNested; - typedef typename internal::remove_all::type NestedExpression; - - SparseSymmetricPermutationProduct(const MatrixType& mat, const Perm& perm) - : m_matrix(mat), m_perm(perm) - {} - - inline Index rows() const { return m_matrix.rows(); } - inline Index cols() const { return m_matrix.cols(); } - - const NestedExpression& matrix() const { return m_matrix; } - const Perm& perm() const { return m_perm; } - - protected: - MatrixTypeNested m_matrix; - const Perm& m_perm; - -}; - -namespace internal { - -template -struct Assignment, internal::assign_op, Sparse2Sparse> -{ - typedef SparseSymmetricPermutationProduct SrcXprType; - typedef typename DstXprType::StorageIndex DstIndex; - template - static void run(SparseMatrix &dst, const SrcXprType &src, const internal::assign_op &) - { - // internal::permute_symm_to_fullsymm(m_matrix,_dest,m_perm.indices().data()); - SparseMatrix tmp; - internal::permute_symm_to_fullsymm(src.matrix(),tmp,src.perm().indices().data()); - dst = tmp; - } - - template - static void run(SparseSelfAdjointView& dst, const SrcXprType &src, const internal::assign_op &) - { - internal::permute_symm_to_symm(src.matrix(),dst.matrix(),src.perm().indices().data()); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_SELFADJOINTVIEW_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSolverBase.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSolverBase.h deleted file mode 100644 index b4c9a422f05..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSolverBase.h +++ /dev/null @@ -1,124 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSESOLVERBASE_H -#define EIGEN_SPARSESOLVERBASE_H - -namespace Eigen { - -namespace internal { - - /** \internal - * Helper functions to solve with a sparse right-hand-side and result. - * The rhs is decomposed into small vertical panels which are solved through dense temporaries. - */ -template -typename enable_if::type -solve_sparse_through_dense_panels(const Decomposition &dec, const Rhs& rhs, Dest &dest) -{ - EIGEN_STATIC_ASSERT((Dest::Flags&RowMajorBit)==0,THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); - typedef typename Dest::Scalar DestScalar; - // we process the sparse rhs per block of NbColsAtOnce columns temporarily stored into a dense matrix. - static const Index NbColsAtOnce = 4; - Index rhsCols = rhs.cols(); - Index size = rhs.rows(); - // the temporary matrices do not need more columns than NbColsAtOnce: - Index tmpCols = (std::min)(rhsCols, NbColsAtOnce); - Eigen::Matrix tmp(size,tmpCols); - Eigen::Matrix tmpX(size,tmpCols); - for(Index k=0; k(rhsCols-k, NbColsAtOnce); - tmp.leftCols(actualCols) = rhs.middleCols(k,actualCols); - tmpX.leftCols(actualCols) = dec.solve(tmp.leftCols(actualCols)); - dest.middleCols(k,actualCols) = tmpX.leftCols(actualCols).sparseView(); - } -} - -// Overload for vector as rhs -template -typename enable_if::type -solve_sparse_through_dense_panels(const Decomposition &dec, const Rhs& rhs, Dest &dest) -{ - typedef typename Dest::Scalar DestScalar; - Index size = rhs.rows(); - Eigen::Matrix rhs_dense(rhs); - Eigen::Matrix dest_dense(size); - dest_dense = dec.solve(rhs_dense); - dest = dest_dense.sparseView(); -} - -} // end namespace internal - -/** \class SparseSolverBase - * \ingroup SparseCore_Module - * \brief A base class for sparse solvers - * - * \tparam Derived the actual type of the solver. - * - */ -template -class SparseSolverBase : internal::noncopyable -{ - public: - - /** Default constructor */ - SparseSolverBase() - : m_isInitialized(false) - {} - - ~SparseSolverBase() - {} - - Derived& derived() { return *static_cast(this); } - const Derived& derived() const { return *static_cast(this); } - - /** \returns an expression of the solution x of \f$ A x = b \f$ using the current decomposition of A. - * - * \sa compute() - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "Solver is not initialized."); - eigen_assert(derived().rows()==b.rows() && "solve(): invalid number of rows of the right hand side matrix b"); - return Solve(derived(), b.derived()); - } - - /** \returns an expression of the solution x of \f$ A x = b \f$ using the current decomposition of A. - * - * \sa compute() - */ - template - inline const Solve - solve(const SparseMatrixBase& b) const - { - eigen_assert(m_isInitialized && "Solver is not initialized."); - eigen_assert(derived().rows()==b.rows() && "solve(): invalid number of rows of the right hand side matrix b"); - return Solve(derived(), b.derived()); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** \internal default implementation of solving with a sparse rhs */ - template - void _solve_impl(const SparseMatrixBase &b, SparseMatrixBase &dest) const - { - internal::solve_sparse_through_dense_panels(derived(), b.derived(), dest.derived()); - } - #endif // EIGEN_PARSED_BY_DOXYGEN - - protected: - - mutable bool m_isInitialized; -}; - -} // end namespace Eigen - -#endif // EIGEN_SPARSESOLVERBASE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSparseProductWithPruning.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSparseProductWithPruning.h deleted file mode 100644 index 88820a48f36..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseSparseProductWithPruning.h +++ /dev/null @@ -1,198 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSESPARSEPRODUCTWITHPRUNING_H -#define EIGEN_SPARSESPARSEPRODUCTWITHPRUNING_H - -namespace Eigen { - -namespace internal { - - -// perform a pseudo in-place sparse * sparse product assuming all matrices are col major -template -static void sparse_sparse_product_with_pruning_impl(const Lhs& lhs, const Rhs& rhs, ResultType& res, const typename ResultType::RealScalar& tolerance) -{ - // return sparse_sparse_product_with_pruning_impl2(lhs,rhs,res); - - typedef typename remove_all::type::Scalar RhsScalar; - typedef typename remove_all::type::Scalar ResScalar; - typedef typename remove_all::type::StorageIndex StorageIndex; - - // make sure to call innerSize/outerSize since we fake the storage order. - Index rows = lhs.innerSize(); - Index cols = rhs.outerSize(); - //Index size = lhs.outerSize(); - eigen_assert(lhs.outerSize() == rhs.innerSize()); - - // allocate a temporary buffer - AmbiVector tempVector(rows); - - // mimics a resizeByInnerOuter: - if(ResultType::IsRowMajor) - res.resize(cols, rows); - else - res.resize(rows, cols); - - evaluator lhsEval(lhs); - evaluator rhsEval(rhs); - - // estimate the number of non zero entries - // given a rhs column containing Y non zeros, we assume that the respective Y columns - // of the lhs differs in average of one non zeros, thus the number of non zeros for - // the product of a rhs column with the lhs is X+Y where X is the average number of non zero - // per column of the lhs. - // Therefore, we have nnz(lhs*rhs) = nnz(lhs) + nnz(rhs) - Index estimated_nnz_prod = lhsEval.nonZerosEstimate() + rhsEval.nonZerosEstimate(); - - res.reserve(estimated_nnz_prod); - double ratioColRes = double(estimated_nnz_prod)/(double(lhs.rows())*double(rhs.cols())); - for (Index j=0; j::InnerIterator rhsIt(rhsEval, j); rhsIt; ++rhsIt) - { - // FIXME should be written like this: tmp += rhsIt.value() * lhs.col(rhsIt.index()) - tempVector.restart(); - RhsScalar x = rhsIt.value(); - for (typename evaluator::InnerIterator lhsIt(lhsEval, rhsIt.index()); lhsIt; ++lhsIt) - { - tempVector.coeffRef(lhsIt.index()) += lhsIt.value() * x; - } - } - res.startVec(j); - for (typename AmbiVector::Iterator it(tempVector,tolerance); it; ++it) - res.insertBackByOuterInner(j,it.index()) = it.value(); - } - res.finalize(); -} - -template::Flags&RowMajorBit, - int RhsStorageOrder = traits::Flags&RowMajorBit, - int ResStorageOrder = traits::Flags&RowMajorBit> -struct sparse_sparse_product_with_pruning_selector; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typename remove_all::type _res(res.rows(), res.cols()); - internal::sparse_sparse_product_with_pruning_impl(lhs, rhs, _res, tolerance); - res.swap(_res); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - // we need a col-major matrix to hold the result - typedef SparseMatrix SparseTemporaryType; - SparseTemporaryType _res(res.rows(), res.cols()); - internal::sparse_sparse_product_with_pruning_impl(lhs, rhs, _res, tolerance); - res = _res; - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - // let's transpose the product to get a column x column product - typename remove_all::type _res(res.rows(), res.cols()); - internal::sparse_sparse_product_with_pruning_impl(rhs, lhs, _res, tolerance); - res.swap(_res); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typedef SparseMatrix ColMajorMatrixLhs; - typedef SparseMatrix ColMajorMatrixRhs; - ColMajorMatrixLhs colLhs(lhs); - ColMajorMatrixRhs colRhs(rhs); - internal::sparse_sparse_product_with_pruning_impl(colLhs, colRhs, res, tolerance); - - // let's transpose the product to get a column x column product -// typedef SparseMatrix SparseTemporaryType; -// SparseTemporaryType _res(res.cols(), res.rows()); -// sparse_sparse_product_with_pruning_impl(rhs, lhs, _res); -// res = _res.transpose(); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typedef SparseMatrix RowMajorMatrixLhs; - RowMajorMatrixLhs rowLhs(lhs); - sparse_sparse_product_with_pruning_selector(rowLhs,rhs,res,tolerance); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typedef SparseMatrix RowMajorMatrixRhs; - RowMajorMatrixRhs rowRhs(rhs); - sparse_sparse_product_with_pruning_selector(lhs,rowRhs,res,tolerance); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typedef SparseMatrix ColMajorMatrixRhs; - ColMajorMatrixRhs colRhs(rhs); - internal::sparse_sparse_product_with_pruning_impl(lhs, colRhs, res, tolerance); - } -}; - -template -struct sparse_sparse_product_with_pruning_selector -{ - typedef typename ResultType::RealScalar RealScalar; - static void run(const Lhs& lhs, const Rhs& rhs, ResultType& res, const RealScalar& tolerance) - { - typedef SparseMatrix ColMajorMatrixLhs; - ColMajorMatrixLhs colLhs(lhs); - internal::sparse_sparse_product_with_pruning_impl(colLhs, rhs, res, tolerance); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSESPARSEPRODUCTWITHPRUNING_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTranspose.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTranspose.h deleted file mode 100644 index 3757d4c6b07..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTranspose.h +++ /dev/null @@ -1,92 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSETRANSPOSE_H -#define EIGEN_SPARSETRANSPOSE_H - -namespace Eigen { - -namespace internal { - template - class SparseTransposeImpl - : public SparseMatrixBase > - {}; - - template - class SparseTransposeImpl - : public SparseCompressedBase > - { - typedef SparseCompressedBase > Base; - public: - using Base::derived; - typedef typename Base::Scalar Scalar; - typedef typename Base::StorageIndex StorageIndex; - - inline Index nonZeros() const { return derived().nestedExpression().nonZeros(); } - - inline const Scalar* valuePtr() const { return derived().nestedExpression().valuePtr(); } - inline const StorageIndex* innerIndexPtr() const { return derived().nestedExpression().innerIndexPtr(); } - inline const StorageIndex* outerIndexPtr() const { return derived().nestedExpression().outerIndexPtr(); } - inline const StorageIndex* innerNonZeroPtr() const { return derived().nestedExpression().innerNonZeroPtr(); } - - inline Scalar* valuePtr() { return derived().nestedExpression().valuePtr(); } - inline StorageIndex* innerIndexPtr() { return derived().nestedExpression().innerIndexPtr(); } - inline StorageIndex* outerIndexPtr() { return derived().nestedExpression().outerIndexPtr(); } - inline StorageIndex* innerNonZeroPtr() { return derived().nestedExpression().innerNonZeroPtr(); } - }; -} - -template class TransposeImpl - : public internal::SparseTransposeImpl -{ - protected: - typedef internal::SparseTransposeImpl Base; -}; - -namespace internal { - -template -struct unary_evaluator, IteratorBased> - : public evaluator_base > -{ - typedef typename evaluator::InnerIterator EvalIterator; - public: - typedef Transpose XprType; - - inline Index nonZerosEstimate() const { - return m_argImpl.nonZerosEstimate(); - } - - class InnerIterator : public EvalIterator - { - public: - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& unaryOp, Index outer) - : EvalIterator(unaryOp.m_argImpl,outer) - {} - - Index row() const { return EvalIterator::col(); } - Index col() const { return EvalIterator::row(); } - }; - - enum { - CoeffReadCost = evaluator::CoeffReadCost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType& op) :m_argImpl(op.nestedExpression()) {} - - protected: - evaluator m_argImpl; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSETRANSPOSE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTriangularView.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTriangularView.h deleted file mode 100644 index 9ac120266a8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseTriangularView.h +++ /dev/null @@ -1,189 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2015 Gael Guennebaud -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_TRIANGULARVIEW_H -#define EIGEN_SPARSE_TRIANGULARVIEW_H - -namespace Eigen { - -/** \ingroup SparseCore_Module - * - * \brief Base class for a triangular part in a \b sparse matrix - * - * This class is an abstract base class of class TriangularView, and objects of type TriangularViewImpl cannot be instantiated. - * It extends class TriangularView with additional methods which are available for sparse expressions only. - * - * \sa class TriangularView, SparseMatrixBase::triangularView() - */ -template class TriangularViewImpl - : public SparseMatrixBase > -{ - enum { SkipFirst = ((Mode&Lower) && !(MatrixType::Flags&RowMajorBit)) - || ((Mode&Upper) && (MatrixType::Flags&RowMajorBit)), - SkipLast = !SkipFirst, - SkipDiag = (Mode&ZeroDiag) ? 1 : 0, - HasUnitDiag = (Mode&UnitDiag) ? 1 : 0 - }; - - typedef TriangularView TriangularViewType; - - protected: - // dummy solve function to make TriangularView happy. - void solve() const; - - typedef SparseMatrixBase Base; - public: - - EIGEN_SPARSE_PUBLIC_INTERFACE(TriangularViewType) - - typedef typename MatrixType::Nested MatrixTypeNested; - typedef typename internal::remove_reference::type MatrixTypeNestedNonRef; - typedef typename internal::remove_all::type MatrixTypeNestedCleaned; - - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE void _solve_impl(const RhsType &rhs, DstType &dst) const { - if(!(internal::is_same::value && internal::extract_data(dst) == internal::extract_data(rhs))) - dst = rhs; - this->solveInPlace(dst); - } - - /** Applies the inverse of \c *this to the dense vector or matrix \a other, "in-place" */ - template void solveInPlace(MatrixBase& other) const; - - /** Applies the inverse of \c *this to the sparse vector or matrix \a other, "in-place" */ - template void solveInPlace(SparseMatrixBase& other) const; - -}; - -namespace internal { - -template -struct unary_evaluator, IteratorBased> - : evaluator_base > -{ - typedef TriangularView XprType; - -protected: - - typedef typename XprType::Scalar Scalar; - typedef typename XprType::StorageIndex StorageIndex; - typedef typename evaluator::InnerIterator EvalIterator; - - enum { SkipFirst = ((Mode&Lower) && !(ArgType::Flags&RowMajorBit)) - || ((Mode&Upper) && (ArgType::Flags&RowMajorBit)), - SkipLast = !SkipFirst, - SkipDiag = (Mode&ZeroDiag) ? 1 : 0, - HasUnitDiag = (Mode&UnitDiag) ? 1 : 0 - }; - -public: - - enum { - CoeffReadCost = evaluator::CoeffReadCost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType &xpr) : m_argImpl(xpr.nestedExpression()), m_arg(xpr.nestedExpression()) {} - - inline Index nonZerosEstimate() const { - return m_argImpl.nonZerosEstimate(); - } - - class InnerIterator : public EvalIterator - { - typedef EvalIterator Base; - public: - - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& xprEval, Index outer) - : Base(xprEval.m_argImpl,outer), m_returnOne(false), m_containsDiag(Base::outer()index()<=outer : this->index()=Base::outer())) - { - if((!SkipFirst) && Base::operator bool()) - Base::operator++(); - m_returnOne = m_containsDiag; - } - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - if(HasUnitDiag && m_returnOne) - m_returnOne = false; - else - { - Base::operator++(); - if(HasUnitDiag && (!SkipFirst) && ((!Base::operator bool()) || Base::index()>=Base::outer())) - { - if((!SkipFirst) && Base::operator bool()) - Base::operator++(); - m_returnOne = m_containsDiag; - } - } - return *this; - } - - EIGEN_STRONG_INLINE operator bool() const - { - if(HasUnitDiag && m_returnOne) - return true; - if(SkipFirst) return Base::operator bool(); - else - { - if (SkipDiag) return (Base::operator bool() && this->index() < this->outer()); - else return (Base::operator bool() && this->index() <= this->outer()); - } - } - -// inline Index row() const { return (ArgType::Flags&RowMajorBit ? Base::outer() : this->index()); } -// inline Index col() const { return (ArgType::Flags&RowMajorBit ? this->index() : Base::outer()); } - inline StorageIndex index() const - { - if(HasUnitDiag && m_returnOne) return internal::convert_index(Base::outer()); - else return Base::index(); - } - inline Scalar value() const - { - if(HasUnitDiag && m_returnOne) return Scalar(1); - else return Base::value(); - } - - protected: - bool m_returnOne; - bool m_containsDiag; - private: - Scalar& valueRef(); - }; - -protected: - evaluator m_argImpl; - const ArgType& m_arg; -}; - -} // end namespace internal - -template -template -inline const TriangularView -SparseMatrixBase::triangularView() const -{ - return TriangularView(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSE_TRIANGULARVIEW_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseUtil.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseUtil.h deleted file mode 100644 index 74df0d49642..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseUtil.h +++ /dev/null @@ -1,178 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEUTIL_H -#define EIGEN_SPARSEUTIL_H - -namespace Eigen { - -#ifdef NDEBUG -#define EIGEN_DBG_SPARSE(X) -#else -#define EIGEN_DBG_SPARSE(X) X -#endif - -#define EIGEN_SPARSE_INHERIT_ASSIGNMENT_OPERATOR(Derived, Op) \ -template \ -EIGEN_STRONG_INLINE Derived& operator Op(const Eigen::SparseMatrixBase& other) \ -{ \ - return Base::operator Op(other.derived()); \ -} \ -EIGEN_STRONG_INLINE Derived& operator Op(const Derived& other) \ -{ \ - return Base::operator Op(other); \ -} - -#define EIGEN_SPARSE_INHERIT_SCALAR_ASSIGNMENT_OPERATOR(Derived, Op) \ -template \ -EIGEN_STRONG_INLINE Derived& operator Op(const Other& scalar) \ -{ \ - return Base::operator Op(scalar); \ -} - -#define EIGEN_SPARSE_INHERIT_ASSIGNMENT_OPERATORS(Derived) \ -EIGEN_SPARSE_INHERIT_ASSIGNMENT_OPERATOR(Derived, =) - - -#define EIGEN_SPARSE_PUBLIC_INTERFACE(Derived) \ - EIGEN_GENERIC_PUBLIC_INTERFACE(Derived) - - -const int CoherentAccessPattern = 0x1; -const int InnerRandomAccessPattern = 0x2 | CoherentAccessPattern; -const int OuterRandomAccessPattern = 0x4 | CoherentAccessPattern; -const int RandomAccessPattern = 0x8 | OuterRandomAccessPattern | InnerRandomAccessPattern; - -template class SparseMatrix; -template class DynamicSparseMatrix; -template class SparseVector; -template class MappedSparseMatrix; - -template class SparseSelfAdjointView; -template class SparseDiagonalProduct; -template class SparseView; - -template class SparseSparseProduct; -template class SparseTimeDenseProduct; -template class DenseTimeSparseProduct; -template class SparseDenseOuterProduct; - -template struct SparseSparseProductReturnType; -template::ColsAtCompileTime,internal::traits::RowsAtCompileTime)> struct DenseSparseProductReturnType; - -template::ColsAtCompileTime,internal::traits::RowsAtCompileTime)> struct SparseDenseProductReturnType; -template class SparseSymmetricPermutationProduct; - -namespace internal { - -template struct sparse_eval; - -template struct eval - : sparse_eval::RowsAtCompileTime,traits::ColsAtCompileTime,traits::Flags> -{}; - -template struct sparse_eval { - typedef typename traits::Scalar _Scalar; - typedef typename traits::StorageIndex _StorageIndex; - public: - typedef SparseVector<_Scalar, RowMajor, _StorageIndex> type; -}; - -template struct sparse_eval { - typedef typename traits::Scalar _Scalar; - typedef typename traits::StorageIndex _StorageIndex; - public: - typedef SparseVector<_Scalar, ColMajor, _StorageIndex> type; -}; - -// TODO this seems almost identical to plain_matrix_type -template struct sparse_eval { - typedef typename traits::Scalar _Scalar; - typedef typename traits::StorageIndex _StorageIndex; - enum { _Options = ((Flags&RowMajorBit)==RowMajorBit) ? RowMajor : ColMajor }; - public: - typedef SparseMatrix<_Scalar, _Options, _StorageIndex> type; -}; - -template struct sparse_eval { - typedef typename traits::Scalar _Scalar; - public: - typedef Matrix<_Scalar, 1, 1> type; -}; - -template struct plain_matrix_type -{ - typedef typename traits::Scalar _Scalar; - typedef typename traits::StorageIndex _StorageIndex; - enum { _Options = ((evaluator::Flags&RowMajorBit)==RowMajorBit) ? RowMajor : ColMajor }; - public: - typedef SparseMatrix<_Scalar, _Options, _StorageIndex> type; -}; - -template -struct plain_object_eval - : sparse_eval::RowsAtCompileTime,traits::ColsAtCompileTime, evaluator::Flags> -{}; - -template -struct solve_traits -{ - typedef typename sparse_eval::Flags>::type PlainObject; -}; - -template -struct generic_xpr_base -{ - typedef SparseMatrixBase type; -}; - -struct SparseTriangularShape { static std::string debugName() { return "SparseTriangularShape"; } }; -struct SparseSelfAdjointShape { static std::string debugName() { return "SparseSelfAdjointShape"; } }; - -template<> struct glue_shapes { typedef SparseSelfAdjointShape type; }; -template<> struct glue_shapes { typedef SparseTriangularShape type; }; - -} // end namespace internal - -/** \ingroup SparseCore_Module - * - * \class Triplet - * - * \brief A small structure to hold a non zero as a triplet (i,j,value). - * - * \sa SparseMatrix::setFromTriplets() - */ -template::StorageIndex > -class Triplet -{ -public: - Triplet() : m_row(0), m_col(0), m_value(0) {} - - Triplet(const StorageIndex& i, const StorageIndex& j, const Scalar& v = Scalar(0)) - : m_row(i), m_col(j), m_value(v) - {} - - /** \returns the row index of the element */ - const StorageIndex& row() const { return m_row; } - - /** \returns the column index of the element */ - const StorageIndex& col() const { return m_col; } - - /** \returns the value of the element */ - const Scalar& value() const { return m_value; } -protected: - StorageIndex m_row, m_col; - Scalar m_value; -}; - -} // end namespace Eigen - -#endif // EIGEN_SPARSEUTIL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseVector.h deleted file mode 100644 index 19b0fbc9d70..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseVector.h +++ /dev/null @@ -1,478 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEVECTOR_H -#define EIGEN_SPARSEVECTOR_H - -namespace Eigen { - -/** \ingroup SparseCore_Module - * \class SparseVector - * - * \brief a sparse vector class - * - * \tparam _Scalar the scalar type, i.e. the type of the coefficients - * - * See http://www.netlib.org/linalg/html_templates/node91.html for details on the storage scheme. - * - * This class can be extended with the help of the plugin mechanism described on the page - * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_SPARSEVECTOR_PLUGIN. - */ - -namespace internal { -template -struct traits > -{ - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - typedef Sparse StorageKind; - typedef MatrixXpr XprKind; - enum { - IsColVector = (_Options & RowMajorBit) ? 0 : 1, - - RowsAtCompileTime = IsColVector ? Dynamic : 1, - ColsAtCompileTime = IsColVector ? 1 : Dynamic, - MaxRowsAtCompileTime = RowsAtCompileTime, - MaxColsAtCompileTime = ColsAtCompileTime, - Flags = _Options | NestByRefBit | LvalueBit | (IsColVector ? 0 : RowMajorBit) | CompressedAccessBit, - SupportedAccessPatterns = InnerRandomAccessPattern - }; -}; - -// Sparse-Vector-Assignment kinds: -enum { - SVA_RuntimeSwitch, - SVA_Inner, - SVA_Outer -}; - -template< typename Dest, typename Src, - int AssignmentKind = !bool(Src::IsVectorAtCompileTime) ? SVA_RuntimeSwitch - : Src::InnerSizeAtCompileTime==1 ? SVA_Outer - : SVA_Inner> -struct sparse_vector_assign_selector; - -} - -template -class SparseVector - : public SparseCompressedBase > -{ - typedef SparseCompressedBase Base; - using Base::convert_index; - public: - EIGEN_SPARSE_PUBLIC_INTERFACE(SparseVector) - EIGEN_SPARSE_INHERIT_ASSIGNMENT_OPERATOR(SparseVector, +=) - EIGEN_SPARSE_INHERIT_ASSIGNMENT_OPERATOR(SparseVector, -=) - - typedef internal::CompressedStorage Storage; - enum { IsColVector = internal::traits::IsColVector }; - - enum { - Options = _Options - }; - - EIGEN_STRONG_INLINE Index rows() const { return IsColVector ? m_size : 1; } - EIGEN_STRONG_INLINE Index cols() const { return IsColVector ? 1 : m_size; } - EIGEN_STRONG_INLINE Index innerSize() const { return m_size; } - EIGEN_STRONG_INLINE Index outerSize() const { return 1; } - - EIGEN_STRONG_INLINE const Scalar* valuePtr() const { return m_data.valuePtr(); } - EIGEN_STRONG_INLINE Scalar* valuePtr() { return m_data.valuePtr(); } - - EIGEN_STRONG_INLINE const StorageIndex* innerIndexPtr() const { return m_data.indexPtr(); } - EIGEN_STRONG_INLINE StorageIndex* innerIndexPtr() { return m_data.indexPtr(); } - - inline const StorageIndex* outerIndexPtr() const { return 0; } - inline StorageIndex* outerIndexPtr() { return 0; } - inline const StorageIndex* innerNonZeroPtr() const { return 0; } - inline StorageIndex* innerNonZeroPtr() { return 0; } - - /** \internal */ - inline Storage& data() { return m_data; } - /** \internal */ - inline const Storage& data() const { return m_data; } - - inline Scalar coeff(Index row, Index col) const - { - eigen_assert(IsColVector ? (col==0 && row>=0 && row=0 && col=0 && i=0 && row=0 && col=0 && i=0 && row=0 && col=0 && i= startId) && (m_data.index(p) > i) ) - { - m_data.index(p+1) = m_data.index(p); - m_data.value(p+1) = m_data.value(p); - --p; - } - m_data.index(p+1) = convert_index(i); - m_data.value(p+1) = 0; - return m_data.value(p+1); - } - - /** - */ - inline void reserve(Index reserveSize) { m_data.reserve(reserveSize); } - - - inline void finalize() {} - - /** \copydoc SparseMatrix::prune(const Scalar&,const RealScalar&) */ - void prune(const Scalar& reference, const RealScalar& epsilon = NumTraits::dummy_precision()) - { - m_data.prune(reference,epsilon); - } - - /** Resizes the sparse vector to \a rows x \a cols - * - * This method is provided for compatibility with matrices. - * For a column vector, \a cols must be equal to 1. - * For a row vector, \a rows must be equal to 1. - * - * \sa resize(Index) - */ - void resize(Index rows, Index cols) - { - eigen_assert((IsColVector ? cols : rows)==1 && "Outer dimension must equal 1"); - resize(IsColVector ? rows : cols); - } - - /** Resizes the sparse vector to \a newSize - * This method deletes all entries, thus leaving an empty sparse vector - * - * \sa conservativeResize(), setZero() */ - void resize(Index newSize) - { - m_size = newSize; - m_data.clear(); - } - - /** Resizes the sparse vector to \a newSize, while leaving old values untouched. - * - * If the size of the vector is decreased, then the storage of the out-of bounds coefficients is kept and reserved. - * Call .data().squeeze() to free extra memory. - * - * \sa reserve(), setZero() - */ - void conservativeResize(Index newSize) - { - if (newSize < m_size) - { - Index i = 0; - while (i - inline SparseVector(const SparseMatrixBase& other) - : m_size(0) - { - #ifdef EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - EIGEN_SPARSE_CREATE_TEMPORARY_PLUGIN - #endif - check_template_parameters(); - *this = other.derived(); - } - - inline SparseVector(const SparseVector& other) - : Base(other), m_size(0) - { - check_template_parameters(); - *this = other.derived(); - } - - /** Swaps the values of \c *this and \a other. - * Overloaded for performance: this version performs a \em shallow swap by swaping pointers and attributes only. - * \sa SparseMatrixBase::swap() - */ - inline void swap(SparseVector& other) - { - std::swap(m_size, other.m_size); - m_data.swap(other.m_data); - } - - template - inline void swap(SparseMatrix& other) - { - eigen_assert(other.outerSize()==1); - std::swap(m_size, other.m_innerSize); - m_data.swap(other.m_data); - } - - inline SparseVector& operator=(const SparseVector& other) - { - if (other.isRValue()) - { - swap(other.const_cast_derived()); - } - else - { - resize(other.size()); - m_data = other.m_data; - } - return *this; - } - - template - inline SparseVector& operator=(const SparseMatrixBase& other) - { - SparseVector tmp(other.size()); - internal::sparse_vector_assign_selector::run(tmp,other.derived()); - this->swap(tmp); - return *this; - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - inline SparseVector& operator=(const SparseSparseProduct& product) - { - return Base::operator=(product); - } - #endif - - friend std::ostream & operator << (std::ostream & s, const SparseVector& m) - { - for (Index i=0; i::IsSigned,THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE); - EIGEN_STATIC_ASSERT((_Options&(ColMajor|RowMajor))==Options,INVALID_MATRIX_TEMPLATE_PARAMETERS); - } - - Storage m_data; - Index m_size; -}; - -namespace internal { - -template -struct evaluator > - : evaluator_base > -{ - typedef SparseVector<_Scalar,_Options,_Index> SparseVectorType; - typedef evaluator_base Base; - typedef typename SparseVectorType::InnerIterator InnerIterator; - typedef typename SparseVectorType::ReverseInnerIterator ReverseInnerIterator; - - enum { - CoeffReadCost = NumTraits<_Scalar>::ReadCost, - Flags = SparseVectorType::Flags - }; - - evaluator() : Base() {} - - explicit evaluator(const SparseVectorType &mat) : m_matrix(&mat) - { - EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); - } - - inline Index nonZerosEstimate() const { - return m_matrix->nonZeros(); - } - - operator SparseVectorType&() { return m_matrix->const_cast_derived(); } - operator const SparseVectorType&() const { return *m_matrix; } - - const SparseVectorType *m_matrix; -}; - -template< typename Dest, typename Src> -struct sparse_vector_assign_selector { - static void run(Dest& dst, const Src& src) { - eigen_internal_assert(src.innerSize()==src.size()); - typedef internal::evaluator SrcEvaluatorType; - SrcEvaluatorType srcEval(src); - for(typename SrcEvaluatorType::InnerIterator it(srcEval, 0); it; ++it) - dst.insert(it.index()) = it.value(); - } -}; - -template< typename Dest, typename Src> -struct sparse_vector_assign_selector { - static void run(Dest& dst, const Src& src) { - eigen_internal_assert(src.outerSize()==src.size()); - typedef internal::evaluator SrcEvaluatorType; - SrcEvaluatorType srcEval(src); - for(Index i=0; i -struct sparse_vector_assign_selector { - static void run(Dest& dst, const Src& src) { - if(src.outerSize()==1) sparse_vector_assign_selector::run(dst, src); - else sparse_vector_assign_selector::run(dst, src); - } -}; - -} - -} // end namespace Eigen - -#endif // EIGEN_SPARSEVECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseView.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseView.h deleted file mode 100644 index 92b3d1f7bad..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/SparseView.h +++ /dev/null @@ -1,254 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2011-2014 Gael Guennebaud -// Copyright (C) 2010 Daniel Lowengrub -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSEVIEW_H -#define EIGEN_SPARSEVIEW_H - -namespace Eigen { - -namespace internal { - -template -struct traits > : traits -{ - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Sparse StorageKind; - enum { - Flags = int(traits::Flags) & (RowMajorBit) - }; -}; - -} // end namespace internal - -/** \ingroup SparseCore_Module - * \class SparseView - * - * \brief Expression of a dense or sparse matrix with zero or too small values removed - * - * \tparam MatrixType the type of the object of which we are removing the small entries - * - * This class represents an expression of a given dense or sparse matrix with - * entries smaller than \c reference * \c epsilon are removed. - * It is the return type of MatrixBase::sparseView() and SparseMatrixBase::pruned() - * and most of the time this is the only way it is used. - * - * \sa MatrixBase::sparseView(), SparseMatrixBase::pruned() - */ -template -class SparseView : public SparseMatrixBase > -{ - typedef typename MatrixType::Nested MatrixTypeNested; - typedef typename internal::remove_all::type _MatrixTypeNested; - typedef SparseMatrixBase Base; -public: - EIGEN_SPARSE_PUBLIC_INTERFACE(SparseView) - typedef typename internal::remove_all::type NestedExpression; - - explicit SparseView(const MatrixType& mat, const Scalar& reference = Scalar(0), - const RealScalar &epsilon = NumTraits::dummy_precision()) - : m_matrix(mat), m_reference(reference), m_epsilon(epsilon) {} - - inline Index rows() const { return m_matrix.rows(); } - inline Index cols() const { return m_matrix.cols(); } - - inline Index innerSize() const { return m_matrix.innerSize(); } - inline Index outerSize() const { return m_matrix.outerSize(); } - - /** \returns the nested expression */ - const typename internal::remove_all::type& - nestedExpression() const { return m_matrix; } - - Scalar reference() const { return m_reference; } - RealScalar epsilon() const { return m_epsilon; } - -protected: - MatrixTypeNested m_matrix; - Scalar m_reference; - RealScalar m_epsilon; -}; - -namespace internal { - -// TODO find a way to unify the two following variants -// This is tricky because implementing an inner iterator on top of an IndexBased evaluator is -// not easy because the evaluators do not expose the sizes of the underlying expression. - -template -struct unary_evaluator, IteratorBased> - : public evaluator_base > -{ - typedef typename evaluator::InnerIterator EvalIterator; - public: - typedef SparseView XprType; - - class InnerIterator : public EvalIterator - { - protected: - typedef typename XprType::Scalar Scalar; - public: - - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& sve, Index outer) - : EvalIterator(sve.m_argImpl,outer), m_view(sve.m_view) - { - incrementToNonZero(); - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - EvalIterator::operator++(); - incrementToNonZero(); - return *this; - } - - using EvalIterator::value; - - protected: - const XprType &m_view; - - private: - void incrementToNonZero() - { - while((bool(*this)) && internal::isMuchSmallerThan(value(), m_view.reference(), m_view.epsilon())) - { - EvalIterator::operator++(); - } - } - }; - - enum { - CoeffReadCost = evaluator::CoeffReadCost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType& xpr) : m_argImpl(xpr.nestedExpression()), m_view(xpr) {} - - protected: - evaluator m_argImpl; - const XprType &m_view; -}; - -template -struct unary_evaluator, IndexBased> - : public evaluator_base > -{ - public: - typedef SparseView XprType; - protected: - enum { IsRowMajor = (XprType::Flags&RowMajorBit)==RowMajorBit }; - typedef typename XprType::Scalar Scalar; - typedef typename XprType::StorageIndex StorageIndex; - public: - - class InnerIterator - { - public: - - EIGEN_STRONG_INLINE InnerIterator(const unary_evaluator& sve, Index outer) - : m_sve(sve), m_inner(0), m_outer(outer), m_end(sve.m_view.innerSize()) - { - incrementToNonZero(); - } - - EIGEN_STRONG_INLINE InnerIterator& operator++() - { - m_inner++; - incrementToNonZero(); - return *this; - } - - EIGEN_STRONG_INLINE Scalar value() const - { - return (IsRowMajor) ? m_sve.m_argImpl.coeff(m_outer, m_inner) - : m_sve.m_argImpl.coeff(m_inner, m_outer); - } - - EIGEN_STRONG_INLINE StorageIndex index() const { return m_inner; } - inline Index row() const { return IsRowMajor ? m_outer : index(); } - inline Index col() const { return IsRowMajor ? index() : m_outer; } - - EIGEN_STRONG_INLINE operator bool() const { return m_inner < m_end && m_inner>=0; } - - protected: - const unary_evaluator &m_sve; - Index m_inner; - const Index m_outer; - const Index m_end; - - private: - void incrementToNonZero() - { - while((bool(*this)) && internal::isMuchSmallerThan(value(), m_sve.m_view.reference(), m_sve.m_view.epsilon())) - { - m_inner++; - } - } - }; - - enum { - CoeffReadCost = evaluator::CoeffReadCost, - Flags = XprType::Flags - }; - - explicit unary_evaluator(const XprType& xpr) : m_argImpl(xpr.nestedExpression()), m_view(xpr) {} - - protected: - evaluator m_argImpl; - const XprType &m_view; -}; - -} // end namespace internal - -/** \ingroup SparseCore_Module - * - * \returns a sparse expression of the dense expression \c *this with values smaller than - * \a reference * \a epsilon removed. - * - * This method is typically used when prototyping to convert a quickly assembled dense Matrix \c D to a SparseMatrix \c S: - * \code - * MatrixXd D(n,m); - * SparseMatrix S; - * S = D.sparseView(); // suppress numerical zeros (exact) - * S = D.sparseView(reference); - * S = D.sparseView(reference,epsilon); - * \endcode - * where \a reference is a meaningful non zero reference value, - * and \a epsilon is a tolerance factor defaulting to NumTraits::dummy_precision(). - * - * \sa SparseMatrixBase::pruned(), class SparseView */ -template -const SparseView MatrixBase::sparseView(const Scalar& reference, - const typename NumTraits::Real& epsilon) const -{ - return SparseView(derived(), reference, epsilon); -} - -/** \returns an expression of \c *this with values smaller than - * \a reference * \a epsilon removed. - * - * This method is typically used in conjunction with the product of two sparse matrices - * to automatically prune the smallest values as follows: - * \code - * C = (A*B).pruned(); // suppress numerical zeros (exact) - * C = (A*B).pruned(ref); - * C = (A*B).pruned(ref,epsilon); - * \endcode - * where \c ref is a meaningful non zero reference value. - * */ -template -const SparseView -SparseMatrixBase::pruned(const Scalar& reference, - const RealScalar& epsilon) const -{ - return SparseView(derived(), reference, epsilon); -} - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/TriangularSolver.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/TriangularSolver.h deleted file mode 100644 index f9c56ba7980..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseCore/TriangularSolver.h +++ /dev/null @@ -1,315 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSETRIANGULARSOLVER_H -#define EIGEN_SPARSETRIANGULARSOLVER_H - -namespace Eigen { - -namespace internal { - -template::Flags) & RowMajorBit> -struct sparse_solve_triangular_selector; - -// forward substitution, row-major -template -struct sparse_solve_triangular_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef evaluator LhsEval; - typedef typename evaluator::InnerIterator LhsIterator; - static void run(const Lhs& lhs, Rhs& other) - { - LhsEval lhsEval(lhs); - for(Index col=0 ; col -struct sparse_solve_triangular_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef evaluator LhsEval; - typedef typename evaluator::InnerIterator LhsIterator; - static void run(const Lhs& lhs, Rhs& other) - { - LhsEval lhsEval(lhs); - for(Index col=0 ; col=0 ; --i) - { - Scalar tmp = other.coeff(i,col); - Scalar l_ii(0); - LhsIterator it(lhsEval, i); - while(it && it.index() -struct sparse_solve_triangular_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef evaluator LhsEval; - typedef typename evaluator::InnerIterator LhsIterator; - static void run(const Lhs& lhs, Rhs& other) - { - LhsEval lhsEval(lhs); - for(Index col=0 ; col -struct sparse_solve_triangular_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef evaluator LhsEval; - typedef typename evaluator::InnerIterator LhsIterator; - static void run(const Lhs& lhs, Rhs& other) - { - LhsEval lhsEval(lhs); - for(Index col=0 ; col=0; --i) - { - Scalar& tmp = other.coeffRef(i,col); - if (tmp!=Scalar(0)) // optimization when other is actually sparse - { - if(!(Mode & UnitDiag)) - { - // TODO replace this by a binary search. make sure the binary search is safe for partially sorted elements - LhsIterator it(lhsEval, i); - while(it && it.index()!=i) - ++it; - eigen_assert(it && it.index()==i); - other.coeffRef(i,col) /= it.value(); - } - LhsIterator it(lhsEval, i); - for(; it && it.index() -template -void TriangularViewImpl::solveInPlace(MatrixBase& other) const -{ - eigen_assert(derived().cols() == derived().rows() && derived().cols() == other.rows()); - eigen_assert((!(Mode & ZeroDiag)) && bool(Mode & (Upper|Lower))); - - enum { copy = internal::traits::Flags & RowMajorBit }; - - typedef typename internal::conditional::type, OtherDerived&>::type OtherCopy; - OtherCopy otherCopy(other.derived()); - - internal::sparse_solve_triangular_selector::type, Mode>::run(derived().nestedExpression(), otherCopy); - - if (copy) - other = otherCopy; -} -#endif - -// pure sparse path - -namespace internal { - -template -struct sparse_solve_triangular_sparse_selector; - -// forward substitution, col-major -template -struct sparse_solve_triangular_sparse_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef typename promote_index_type::StorageIndex, - typename traits::StorageIndex>::type StorageIndex; - static void run(const Lhs& lhs, Rhs& other) - { - const bool IsLower = (UpLo==Lower); - AmbiVector tempVector(other.rows()*2); - tempVector.setBounds(0,other.rows()); - - Rhs res(other.rows(), other.cols()); - res.reserve(other.nonZeros()); - - for(Index col=0 ; col=0; - i+=IsLower?1:-1) - { - tempVector.restart(); - Scalar& ci = tempVector.coeffRef(i); - if (ci!=Scalar(0)) - { - // find - typename Lhs::InnerIterator it(lhs, i); - if(!(Mode & UnitDiag)) - { - if (IsLower) - { - eigen_assert(it.index()==i); - ci /= it.value(); - } - else - ci /= lhs.coeff(i,i); - } - tempVector.restart(); - if (IsLower) - { - if (it.index()==i) - ++it; - for(; it; ++it) - tempVector.coeffRef(it.index()) -= ci * it.value(); - } - else - { - for(; it && it.index()::Iterator it(tempVector/*,1e-12*/); it; ++it) - { - ++ count; -// std::cerr << "fill " << it.index() << ", " << col << "\n"; -// std::cout << it.value() << " "; - // FIXME use insertBack - res.insert(it.index(), col) = it.value(); - } -// std::cout << "tempVector.nonZeros() == " << int(count) << " / " << (other.rows()) << "\n"; - } - res.finalize(); - other = res.markAsRValue(); - } -}; - -} // end namespace internal - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void TriangularViewImpl::solveInPlace(SparseMatrixBase& other) const -{ - eigen_assert(derived().cols() == derived().rows() && derived().cols() == other.rows()); - eigen_assert( (!(Mode & ZeroDiag)) && bool(Mode & (Upper|Lower))); - -// enum { copy = internal::traits::Flags & RowMajorBit }; - -// typedef typename internal::conditional::type, OtherDerived&>::type OtherCopy; -// OtherCopy otherCopy(other.derived()); - - internal::sparse_solve_triangular_sparse_selector::run(derived().nestedExpression(), other.derived()); - -// if (copy) -// other = otherCopy; -} -#endif - -} // end namespace Eigen - -#endif // EIGEN_SPARSETRIANGULARSOLVER_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU.h deleted file mode 100644 index 87f0efe37ed..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU.h +++ /dev/null @@ -1,773 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2012-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -#ifndef EIGEN_SPARSE_LU_H -#define EIGEN_SPARSE_LU_H - -namespace Eigen { - -template > class SparseLU; -template struct SparseLUMatrixLReturnType; -template struct SparseLUMatrixUReturnType; - -/** \ingroup SparseLU_Module - * \class SparseLU - * - * \brief Sparse supernodal LU factorization for general matrices - * - * This class implements the supernodal LU factorization for general matrices. - * It uses the main techniques from the sequential SuperLU package - * (http://crd-legacy.lbl.gov/~xiaoye/SuperLU/). It handles transparently real - * and complex arithmetics with single and double precision, depending on the - * scalar type of your input matrix. - * The code has been optimized to provide BLAS-3 operations during supernode-panel updates. - * It benefits directly from the built-in high-performant Eigen BLAS routines. - * Moreover, when the size of a supernode is very small, the BLAS calls are avoided to - * enable a better optimization from the compiler. For best performance, - * you should compile it with NDEBUG flag to avoid the numerous bounds checking on vectors. - * - * An important parameter of this class is the ordering method. It is used to reorder the columns - * (and eventually the rows) of the matrix to reduce the number of new elements that are created during - * numerical factorization. The cheapest method available is COLAMD. - * See \link OrderingMethods_Module the OrderingMethods module \endlink for the list of - * built-in and external ordering methods. - * - * Simple example with key steps - * \code - * VectorXd x(n), b(n); - * SparseMatrix A; - * SparseLU, COLAMDOrdering > solver; - * // fill A and b; - * // Compute the ordering permutation vector from the structural pattern of A - * solver.analyzePattern(A); - * // Compute the numerical factorization - * solver.factorize(A); - * //Use the factors to solve the linear system - * x = solver.solve(b); - * \endcode - * - * \warning The input matrix A should be in a \b compressed and \b column-major form. - * Otherwise an expensive copy will be made. You can call the inexpensive makeCompressed() to get a compressed matrix. - * - * \note Unlike the initial SuperLU implementation, there is no step to equilibrate the matrix. - * For badly scaled matrices, this step can be useful to reduce the pivoting during factorization. - * If this is the case for your matrices, you can try the basic scaling method at - * "unsupported/Eigen/src/IterativeSolvers/Scaling.h" - * - * \tparam _MatrixType The type of the sparse matrix. It must be a column-major SparseMatrix<> - * \tparam _OrderingType The ordering method to use, either AMD, COLAMD or METIS. Default is COLMAD - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept - * \sa \ref OrderingMethods_Module - */ -template -class SparseLU : public SparseSolverBase >, public internal::SparseLUImpl -{ - protected: - typedef SparseSolverBase > APIBase; - using APIBase::m_isInitialized; - public: - using APIBase::_solve_impl; - - typedef _MatrixType MatrixType; - typedef _OrderingType OrderingType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix NCMatrix; - typedef internal::MappedSuperNodalMatrix SCMatrix; - typedef Matrix ScalarVector; - typedef Matrix IndexVector; - typedef PermutationMatrix PermutationType; - typedef internal::SparseLUImpl Base; - - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - SparseLU():m_lastError(""),m_Ustore(0,0,0,0,0,0),m_symmetricmode(false),m_diagpivotthresh(1.0),m_detPermR(1) - { - initperfvalues(); - } - explicit SparseLU(const MatrixType& matrix) - : m_lastError(""),m_Ustore(0,0,0,0,0,0),m_symmetricmode(false),m_diagpivotthresh(1.0),m_detPermR(1) - { - initperfvalues(); - compute(matrix); - } - - ~SparseLU() - { - // Free all explicit dynamic pointers - } - - void analyzePattern (const MatrixType& matrix); - void factorize (const MatrixType& matrix); - void simplicialfactorize(const MatrixType& matrix); - - /** - * Compute the symbolic and numeric factorization of the input sparse matrix. - * The input matrix should be in column-major storage. - */ - void compute (const MatrixType& matrix) - { - // Analyze - analyzePattern(matrix); - //Factorize - factorize(matrix); - } - - inline Index rows() const { return m_mat.rows(); } - inline Index cols() const { return m_mat.cols(); } - /** Indicate that the pattern of the input matrix is symmetric */ - void isSymmetric(bool sym) - { - m_symmetricmode = sym; - } - - /** \returns an expression of the matrix L, internally stored as supernodes - * The only operation available with this expression is the triangular solve - * \code - * y = b; matrixL().solveInPlace(y); - * \endcode - */ - SparseLUMatrixLReturnType matrixL() const - { - return SparseLUMatrixLReturnType(m_Lstore); - } - /** \returns an expression of the matrix U, - * The only operation available with this expression is the triangular solve - * \code - * y = b; matrixU().solveInPlace(y); - * \endcode - */ - SparseLUMatrixUReturnType > matrixU() const - { - return SparseLUMatrixUReturnType >(m_Lstore, m_Ustore); - } - - /** - * \returns a reference to the row matrix permutation \f$ P_r \f$ such that \f$P_r A P_c^T = L U\f$ - * \sa colsPermutation() - */ - inline const PermutationType& rowsPermutation() const - { - return m_perm_r; - } - /** - * \returns a reference to the column matrix permutation\f$ P_c^T \f$ such that \f$P_r A P_c^T = L U\f$ - * \sa rowsPermutation() - */ - inline const PermutationType& colsPermutation() const - { - return m_perm_c; - } - /** Set the threshold used for a diagonal entry to be an acceptable pivot. */ - void setPivotThreshold(const RealScalar& thresh) - { - m_diagpivotthresh = thresh; - } - -#ifdef EIGEN_PARSED_BY_DOXYGEN - /** \returns the solution X of \f$ A X = B \f$ using the current decomposition of A. - * - * \warning the destination matrix X in X = this->solve(B) must be colmun-major. - * - * \sa compute() - */ - template - inline const Solve solve(const MatrixBase& B) const; -#endif // EIGEN_PARSED_BY_DOXYGEN - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the LU factorization reports a problem, zero diagonal for instance - * \c InvalidInput if the input matrix is invalid - * - * \sa iparm() - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - /** - * \returns A string describing the type of error - */ - std::string lastErrorMessage() const - { - return m_lastError; - } - - template - bool _solve_impl(const MatrixBase &B, MatrixBase &X_base) const - { - Dest& X(X_base.derived()); - eigen_assert(m_factorizationIsOk && "The matrix should be factorized first"); - EIGEN_STATIC_ASSERT((Dest::Flags&RowMajorBit)==0, - THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); - - // Permute the right hand side to form X = Pr*B - // on return, X is overwritten by the computed solution - X.resize(B.rows(),B.cols()); - - // this ugly const_cast_derived() helps to detect aliasing when applying the permutations - for(Index j = 0; j < B.cols(); ++j) - X.col(j) = rowsPermutation() * B.const_cast_derived().col(j); - - //Forward substitution with L - this->matrixL().solveInPlace(X); - this->matrixU().solveInPlace(X); - - // Permute back the solution - for (Index j = 0; j < B.cols(); ++j) - X.col(j) = colsPermutation().inverse() * X.col(j); - - return true; - } - - /** - * \returns the absolute value of the determinant of the matrix of which - * *this is the QR decomposition. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * One way to work around that is to use logAbsDeterminant() instead. - * - * \sa logAbsDeterminant(), signDeterminant() - */ - Scalar absDeterminant() - { - using std::abs; - eigen_assert(m_factorizationIsOk && "The matrix should be factorized first."); - // Initialize with the determinant of the row matrix - Scalar det = Scalar(1.); - // Note that the diagonal blocks of U are stored in supernodes, - // which are available in the L part :) - for (Index j = 0; j < this->cols(); ++j) - { - for (typename SCMatrix::InnerIterator it(m_Lstore, j); it; ++it) - { - if(it.index() == j) - { - det *= abs(it.value()); - break; - } - } - } - return det; - } - - /** \returns the natural log of the absolute value of the determinant of the matrix - * of which **this is the QR decomposition - * - * \note This method is useful to work around the risk of overflow/underflow that's - * inherent to the determinant computation. - * - * \sa absDeterminant(), signDeterminant() - */ - Scalar logAbsDeterminant() const - { - using std::log; - using std::abs; - - eigen_assert(m_factorizationIsOk && "The matrix should be factorized first."); - Scalar det = Scalar(0.); - for (Index j = 0; j < this->cols(); ++j) - { - for (typename SCMatrix::InnerIterator it(m_Lstore, j); it; ++it) - { - if(it.row() < j) continue; - if(it.row() == j) - { - det += log(abs(it.value())); - break; - } - } - } - return det; - } - - /** \returns A number representing the sign of the determinant - * - * \sa absDeterminant(), logAbsDeterminant() - */ - Scalar signDeterminant() - { - eigen_assert(m_factorizationIsOk && "The matrix should be factorized first."); - // Initialize with the determinant of the row matrix - Index det = 1; - // Note that the diagonal blocks of U are stored in supernodes, - // which are available in the L part :) - for (Index j = 0; j < this->cols(); ++j) - { - for (typename SCMatrix::InnerIterator it(m_Lstore, j); it; ++it) - { - if(it.index() == j) - { - if(it.value()<0) - det = -det; - else if(it.value()==0) - return 0; - break; - } - } - } - return det * m_detPermR * m_detPermC; - } - - /** \returns The determinant of the matrix. - * - * \sa absDeterminant(), logAbsDeterminant() - */ - Scalar determinant() - { - eigen_assert(m_factorizationIsOk && "The matrix should be factorized first."); - // Initialize with the determinant of the row matrix - Scalar det = Scalar(1.); - // Note that the diagonal blocks of U are stored in supernodes, - // which are available in the L part :) - for (Index j = 0; j < this->cols(); ++j) - { - for (typename SCMatrix::InnerIterator it(m_Lstore, j); it; ++it) - { - if(it.index() == j) - { - det *= it.value(); - break; - } - } - } - return (m_detPermR * m_detPermC) > 0 ? det : -det; - } - - protected: - // Functions - void initperfvalues() - { - m_perfv.panel_size = 16; - m_perfv.relax = 1; - m_perfv.maxsuper = 128; - m_perfv.rowblk = 16; - m_perfv.colblk = 8; - m_perfv.fillfactor = 20; - } - - // Variables - mutable ComputationInfo m_info; - bool m_factorizationIsOk; - bool m_analysisIsOk; - std::string m_lastError; - NCMatrix m_mat; // The input (permuted ) matrix - SCMatrix m_Lstore; // The lower triangular matrix (supernodal) - MappedSparseMatrix m_Ustore; // The upper triangular matrix - PermutationType m_perm_c; // Column permutation - PermutationType m_perm_r ; // Row permutation - IndexVector m_etree; // Column elimination tree - - typename Base::GlobalLU_t m_glu; - - // SparseLU options - bool m_symmetricmode; - // values for performance - internal::perfvalues m_perfv; - RealScalar m_diagpivotthresh; // Specifies the threshold used for a diagonal entry to be an acceptable pivot - Index m_nnzL, m_nnzU; // Nonzeros in L and U factors - Index m_detPermR, m_detPermC; // Determinants of the permutation matrices - private: - // Disable copy constructor - SparseLU (const SparseLU& ); - -}; // End class SparseLU - - - -// Functions needed by the anaysis phase -/** - * Compute the column permutation to minimize the fill-in - * - * - Apply this permutation to the input matrix - - * - * - Compute the column elimination tree on the permuted matrix - * - * - Postorder the elimination tree and the column permutation - * - */ -template -void SparseLU::analyzePattern(const MatrixType& mat) -{ - - //TODO It is possible as in SuperLU to compute row and columns scaling vectors to equilibrate the matrix mat. - - // Firstly, copy the whole input matrix. - m_mat = mat; - - // Compute fill-in ordering - OrderingType ord; - ord(m_mat,m_perm_c); - - // Apply the permutation to the column of the input matrix - if (m_perm_c.size()) - { - m_mat.uncompress(); //NOTE: The effect of this command is only to create the InnerNonzeros pointers. FIXME : This vector is filled but not subsequently used. - // Then, permute only the column pointers - ei_declare_aligned_stack_constructed_variable(StorageIndex,outerIndexPtr,mat.cols()+1,mat.isCompressed()?const_cast(mat.outerIndexPtr()):0); - - // If the input matrix 'mat' is uncompressed, then the outer-indices do not match the ones of m_mat, and a copy is thus needed. - if(!mat.isCompressed()) - IndexVector::Map(outerIndexPtr, mat.cols()+1) = IndexVector::Map(m_mat.outerIndexPtr(),mat.cols()+1); - - // Apply the permutation and compute the nnz per column. - for (Index i = 0; i < mat.cols(); i++) - { - m_mat.outerIndexPtr()[m_perm_c.indices()(i)] = outerIndexPtr[i]; - m_mat.innerNonZeroPtr()[m_perm_c.indices()(i)] = outerIndexPtr[i+1] - outerIndexPtr[i]; - } - } - - // Compute the column elimination tree of the permuted matrix - IndexVector firstRowElt; - internal::coletree(m_mat, m_etree,firstRowElt); - - // In symmetric mode, do not do postorder here - if (!m_symmetricmode) { - IndexVector post, iwork; - // Post order etree - internal::treePostorder(StorageIndex(m_mat.cols()), m_etree, post); - - - // Renumber etree in postorder - Index m = m_mat.cols(); - iwork.resize(m+1); - for (Index i = 0; i < m; ++i) iwork(post(i)) = post(m_etree(i)); - m_etree = iwork; - - // Postmultiply A*Pc by post, i.e reorder the matrix according to the postorder of the etree - PermutationType post_perm(m); - for (Index i = 0; i < m; i++) - post_perm.indices()(i) = post(i); - - // Combine the two permutations : postorder the permutation for future use - if(m_perm_c.size()) { - m_perm_c = post_perm * m_perm_c; - } - - } // end postordering - - m_analysisIsOk = true; -} - -// Functions needed by the numerical factorization phase - - -/** - * - Numerical factorization - * - Interleaved with the symbolic factorization - * On exit, info is - * - * = 0: successful factorization - * - * > 0: if info = i, and i is - * - * <= A->ncol: U(i,i) is exactly zero. The factorization has - * been completed, but the factor U is exactly singular, - * and division by zero will occur if it is used to solve a - * system of equations. - * - * > A->ncol: number of bytes allocated when memory allocation - * failure occurred, plus A->ncol. If lwork = -1, it is - * the estimated amount of space needed, plus A->ncol. - */ -template -void SparseLU::factorize(const MatrixType& matrix) -{ - using internal::emptyIdxLU; - eigen_assert(m_analysisIsOk && "analyzePattern() should be called first"); - eigen_assert((matrix.rows() == matrix.cols()) && "Only for squared matrices"); - - m_isInitialized = true; - - - // Apply the column permutation computed in analyzepattern() - // m_mat = matrix * m_perm_c.inverse(); - m_mat = matrix; - if (m_perm_c.size()) - { - m_mat.uncompress(); //NOTE: The effect of this command is only to create the InnerNonzeros pointers. - //Then, permute only the column pointers - const StorageIndex * outerIndexPtr; - if (matrix.isCompressed()) outerIndexPtr = matrix.outerIndexPtr(); - else - { - StorageIndex* outerIndexPtr_t = new StorageIndex[matrix.cols()+1]; - for(Index i = 0; i <= matrix.cols(); i++) outerIndexPtr_t[i] = m_mat.outerIndexPtr()[i]; - outerIndexPtr = outerIndexPtr_t; - } - for (Index i = 0; i < matrix.cols(); i++) - { - m_mat.outerIndexPtr()[m_perm_c.indices()(i)] = outerIndexPtr[i]; - m_mat.innerNonZeroPtr()[m_perm_c.indices()(i)] = outerIndexPtr[i+1] - outerIndexPtr[i]; - } - if(!matrix.isCompressed()) delete[] outerIndexPtr; - } - else - { //FIXME This should not be needed if the empty permutation is handled transparently - m_perm_c.resize(matrix.cols()); - for(StorageIndex i = 0; i < matrix.cols(); ++i) m_perm_c.indices()(i) = i; - } - - Index m = m_mat.rows(); - Index n = m_mat.cols(); - Index nnz = m_mat.nonZeros(); - Index maxpanel = m_perfv.panel_size * m; - // Allocate working storage common to the factor routines - Index lwork = 0; - Index info = Base::memInit(m, n, nnz, lwork, m_perfv.fillfactor, m_perfv.panel_size, m_glu); - if (info) - { - m_lastError = "UNABLE TO ALLOCATE WORKING MEMORY\n\n" ; - m_factorizationIsOk = false; - return ; - } - - // Set up pointers for integer working arrays - IndexVector segrep(m); segrep.setZero(); - IndexVector parent(m); parent.setZero(); - IndexVector xplore(m); xplore.setZero(); - IndexVector repfnz(maxpanel); - IndexVector panel_lsub(maxpanel); - IndexVector xprune(n); xprune.setZero(); - IndexVector marker(m*internal::LUNoMarker); marker.setZero(); - - repfnz.setConstant(-1); - panel_lsub.setConstant(-1); - - // Set up pointers for scalar working arrays - ScalarVector dense; - dense.setZero(maxpanel); - ScalarVector tempv; - tempv.setZero(internal::LUnumTempV(m, m_perfv.panel_size, m_perfv.maxsuper, /*m_perfv.rowblk*/m) ); - - // Compute the inverse of perm_c - PermutationType iperm_c(m_perm_c.inverse()); - - // Identify initial relaxed snodes - IndexVector relax_end(n); - if ( m_symmetricmode == true ) - Base::heap_relax_snode(n, m_etree, m_perfv.relax, marker, relax_end); - else - Base::relax_snode(n, m_etree, m_perfv.relax, marker, relax_end); - - - m_perm_r.resize(m); - m_perm_r.indices().setConstant(-1); - marker.setConstant(-1); - m_detPermR = 1; // Record the determinant of the row permutation - - m_glu.supno(0) = emptyIdxLU; m_glu.xsup.setConstant(0); - m_glu.xsup(0) = m_glu.xlsub(0) = m_glu.xusub(0) = m_glu.xlusup(0) = Index(0); - - // Work on one 'panel' at a time. A panel is one of the following : - // (a) a relaxed supernode at the bottom of the etree, or - // (b) panel_size contiguous columns, defined by the user - Index jcol; - IndexVector panel_histo(n); - Index pivrow; // Pivotal row number in the original row matrix - Index nseg1; // Number of segments in U-column above panel row jcol - Index nseg; // Number of segments in each U-column - Index irep; - Index i, k, jj; - for (jcol = 0; jcol < n; ) - { - // Adjust panel size so that a panel won't overlap with the next relaxed snode. - Index panel_size = m_perfv.panel_size; // upper bound on panel width - for (k = jcol + 1; k < (std::min)(jcol+panel_size, n); k++) - { - if (relax_end(k) != emptyIdxLU) - { - panel_size = k - jcol; - break; - } - } - if (k == n) - panel_size = n - jcol; - - // Symbolic outer factorization on a panel of columns - Base::panel_dfs(m, panel_size, jcol, m_mat, m_perm_r.indices(), nseg1, dense, panel_lsub, segrep, repfnz, xprune, marker, parent, xplore, m_glu); - - // Numeric sup-panel updates in topological order - Base::panel_bmod(m, panel_size, jcol, nseg1, dense, tempv, segrep, repfnz, m_glu); - - // Sparse LU within the panel, and below the panel diagonal - for ( jj = jcol; jj< jcol + panel_size; jj++) - { - k = (jj - jcol) * m; // Column index for w-wide arrays - - nseg = nseg1; // begin after all the panel segments - //Depth-first-search for the current column - VectorBlock panel_lsubk(panel_lsub, k, m); - VectorBlock repfnz_k(repfnz, k, m); - info = Base::column_dfs(m, jj, m_perm_r.indices(), m_perfv.maxsuper, nseg, panel_lsubk, segrep, repfnz_k, xprune, marker, parent, xplore, m_glu); - if ( info ) - { - m_lastError = "UNABLE TO EXPAND MEMORY IN COLUMN_DFS() "; - m_info = NumericalIssue; - m_factorizationIsOk = false; - return; - } - // Numeric updates to this column - VectorBlock dense_k(dense, k, m); - VectorBlock segrep_k(segrep, nseg1, m-nseg1); - info = Base::column_bmod(jj, (nseg - nseg1), dense_k, tempv, segrep_k, repfnz_k, jcol, m_glu); - if ( info ) - { - m_lastError = "UNABLE TO EXPAND MEMORY IN COLUMN_BMOD() "; - m_info = NumericalIssue; - m_factorizationIsOk = false; - return; - } - - // Copy the U-segments to ucol(*) - info = Base::copy_to_ucol(jj, nseg, segrep, repfnz_k ,m_perm_r.indices(), dense_k, m_glu); - if ( info ) - { - m_lastError = "UNABLE TO EXPAND MEMORY IN COPY_TO_UCOL() "; - m_info = NumericalIssue; - m_factorizationIsOk = false; - return; - } - - // Form the L-segment - info = Base::pivotL(jj, m_diagpivotthresh, m_perm_r.indices(), iperm_c.indices(), pivrow, m_glu); - if ( info ) - { - m_lastError = "THE MATRIX IS STRUCTURALLY SINGULAR ... ZERO COLUMN AT "; - std::ostringstream returnInfo; - returnInfo << info; - m_lastError += returnInfo.str(); - m_info = NumericalIssue; - m_factorizationIsOk = false; - return; - } - - // Update the determinant of the row permutation matrix - // FIXME: the following test is not correct, we should probably take iperm_c into account and pivrow is not directly the row pivot. - if (pivrow != jj) m_detPermR = -m_detPermR; - - // Prune columns (0:jj-1) using column jj - Base::pruneL(jj, m_perm_r.indices(), pivrow, nseg, segrep, repfnz_k, xprune, m_glu); - - // Reset repfnz for this column - for (i = 0; i < nseg; i++) - { - irep = segrep(i); - repfnz_k(irep) = emptyIdxLU; - } - } // end SparseLU within the panel - jcol += panel_size; // Move to the next panel - } // end for -- end elimination - - m_detPermR = m_perm_r.determinant(); - m_detPermC = m_perm_c.determinant(); - - // Count the number of nonzeros in factors - Base::countnz(n, m_nnzL, m_nnzU, m_glu); - // Apply permutation to the L subscripts - Base::fixupL(n, m_perm_r.indices(), m_glu); - - // Create supernode matrix L - m_Lstore.setInfos(m, n, m_glu.lusup, m_glu.xlusup, m_glu.lsub, m_glu.xlsub, m_glu.supno, m_glu.xsup); - // Create the column major upper sparse matrix U; - new (&m_Ustore) MappedSparseMatrix ( m, n, m_nnzU, m_glu.xusub.data(), m_glu.usub.data(), m_glu.ucol.data() ); - - m_info = Success; - m_factorizationIsOk = true; -} - -template -struct SparseLUMatrixLReturnType : internal::no_assignment_operator -{ - typedef typename MappedSupernodalType::Scalar Scalar; - explicit SparseLUMatrixLReturnType(const MappedSupernodalType& mapL) : m_mapL(mapL) - { } - Index rows() { return m_mapL.rows(); } - Index cols() { return m_mapL.cols(); } - template - void solveInPlace( MatrixBase &X) const - { - m_mapL.solveInPlace(X); - } - const MappedSupernodalType& m_mapL; -}; - -template -struct SparseLUMatrixUReturnType : internal::no_assignment_operator -{ - typedef typename MatrixLType::Scalar Scalar; - SparseLUMatrixUReturnType(const MatrixLType& mapL, const MatrixUType& mapU) - : m_mapL(mapL),m_mapU(mapU) - { } - Index rows() { return m_mapL.rows(); } - Index cols() { return m_mapL.cols(); } - - template void solveInPlace(MatrixBase &X) const - { - Index nrhs = X.cols(); - Index n = X.rows(); - // Backward solve with U - for (Index k = m_mapL.nsuper(); k >= 0; k--) - { - Index fsupc = m_mapL.supToCol()[k]; - Index lda = m_mapL.colIndexPtr()[fsupc+1] - m_mapL.colIndexPtr()[fsupc]; // leading dimension - Index nsupc = m_mapL.supToCol()[k+1] - fsupc; - Index luptr = m_mapL.colIndexPtr()[fsupc]; - - if (nsupc == 1) - { - for (Index j = 0; j < nrhs; j++) - { - X(fsupc, j) /= m_mapL.valuePtr()[luptr]; - } - } - else - { - Map, 0, OuterStride<> > A( &(m_mapL.valuePtr()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); - Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); - U = A.template triangularView().solve(U); - } - - for (Index j = 0; j < nrhs; ++j) - { - for (Index jcol = fsupc; jcol < fsupc + nsupc; jcol++) - { - typename MatrixUType::InnerIterator it(m_mapU, jcol); - for ( ; it; ++it) - { - Index irow = it.index(); - X(irow, j) -= X(jcol, j) * it.value(); - } - } - } - } // End For U-solve - } - const MatrixLType& m_mapL; - const MatrixUType& m_mapU; -}; - -} // End namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLUImpl.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLUImpl.h deleted file mode 100644 index fc0cfc4de1a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLUImpl.h +++ /dev/null @@ -1,66 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -#ifndef SPARSELU_IMPL_H -#define SPARSELU_IMPL_H - -namespace Eigen { -namespace internal { - -/** \ingroup SparseLU_Module - * \class SparseLUImpl - * Base class for sparseLU - */ -template -class SparseLUImpl -{ - public: - typedef Matrix ScalarVector; - typedef Matrix IndexVector; - typedef Matrix ScalarMatrix; - typedef Map > MappedMatrixBlock; - typedef typename ScalarVector::RealScalar RealScalar; - typedef Ref > BlockScalarVector; - typedef Ref > BlockIndexVector; - typedef LU_GlobalLU_t GlobalLU_t; - typedef SparseMatrix MatrixType; - - protected: - template - Index expand(VectorType& vec, Index& length, Index nbElts, Index keep_prev, Index& num_expansions); - Index memInit(Index m, Index n, Index annz, Index lwork, Index fillratio, Index panel_size, GlobalLU_t& glu); - template - Index memXpand(VectorType& vec, Index& maxlen, Index nbElts, MemType memtype, Index& num_expansions); - void heap_relax_snode (const Index n, IndexVector& et, const Index relax_columns, IndexVector& descendants, IndexVector& relax_end); - void relax_snode (const Index n, IndexVector& et, const Index relax_columns, IndexVector& descendants, IndexVector& relax_end); - Index snode_dfs(const Index jcol, const Index kcol,const MatrixType& mat, IndexVector& xprune, IndexVector& marker, GlobalLU_t& glu); - Index snode_bmod (const Index jcol, const Index fsupc, ScalarVector& dense, GlobalLU_t& glu); - Index pivotL(const Index jcol, const RealScalar& diagpivotthresh, IndexVector& perm_r, IndexVector& iperm_c, Index& pivrow, GlobalLU_t& glu); - template - void dfs_kernel(const StorageIndex jj, IndexVector& perm_r, - Index& nseg, IndexVector& panel_lsub, IndexVector& segrep, - Ref repfnz_col, IndexVector& xprune, Ref marker, IndexVector& parent, - IndexVector& xplore, GlobalLU_t& glu, Index& nextl_col, Index krow, Traits& traits); - void panel_dfs(const Index m, const Index w, const Index jcol, MatrixType& A, IndexVector& perm_r, Index& nseg, ScalarVector& dense, IndexVector& panel_lsub, IndexVector& segrep, IndexVector& repfnz, IndexVector& xprune, IndexVector& marker, IndexVector& parent, IndexVector& xplore, GlobalLU_t& glu); - - void panel_bmod(const Index m, const Index w, const Index jcol, const Index nseg, ScalarVector& dense, ScalarVector& tempv, IndexVector& segrep, IndexVector& repfnz, GlobalLU_t& glu); - Index column_dfs(const Index m, const Index jcol, IndexVector& perm_r, Index maxsuper, Index& nseg, BlockIndexVector lsub_col, IndexVector& segrep, BlockIndexVector repfnz, IndexVector& xprune, IndexVector& marker, IndexVector& parent, IndexVector& xplore, GlobalLU_t& glu); - Index column_bmod(const Index jcol, const Index nseg, BlockScalarVector dense, ScalarVector& tempv, BlockIndexVector segrep, BlockIndexVector repfnz, Index fpanelc, GlobalLU_t& glu); - Index copy_to_ucol(const Index jcol, const Index nseg, IndexVector& segrep, BlockIndexVector repfnz ,IndexVector& perm_r, BlockScalarVector dense, GlobalLU_t& glu); - void pruneL(const Index jcol, const IndexVector& perm_r, const Index pivrow, const Index nseg, const IndexVector& segrep, BlockIndexVector repfnz, IndexVector& xprune, GlobalLU_t& glu); - void countnz(const Index n, Index& nnzL, Index& nnzU, GlobalLU_t& glu); - void fixupL(const Index n, const IndexVector& perm_r, GlobalLU_t& glu); - - template - friend struct column_dfs_traits; -}; - -} // end namespace internal -} // namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Memory.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Memory.h deleted file mode 100644 index 4dc42e87ba8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Memory.h +++ /dev/null @@ -1,226 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of [s,d,c,z]memory.c files in SuperLU - - * -- SuperLU routine (version 3.1) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * August 1, 2008 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ - -#ifndef EIGEN_SPARSELU_MEMORY -#define EIGEN_SPARSELU_MEMORY - -namespace Eigen { -namespace internal { - -enum { LUNoMarker = 3 }; -enum {emptyIdxLU = -1}; -inline Index LUnumTempV(Index& m, Index& w, Index& t, Index& b) -{ - return (std::max)(m, (t+b)*w); -} - -template< typename Scalar> -inline Index LUTempSpace(Index&m, Index& w) -{ - return (2*w + 4 + LUNoMarker) * m * sizeof(Index) + (w + 1) * m * sizeof(Scalar); -} - - - - -/** - * Expand the existing storage to accomodate more fill-ins - * \param vec Valid pointer to the vector to allocate or expand - * \param[in,out] length At input, contain the current length of the vector that is to be increased. At output, length of the newly allocated vector - * \param[in] nbElts Current number of elements in the factors - * \param keep_prev 1: use length and do not expand the vector; 0: compute new_len and expand - * \param[in,out] num_expansions Number of times the memory has been expanded - */ -template -template -Index SparseLUImpl::expand(VectorType& vec, Index& length, Index nbElts, Index keep_prev, Index& num_expansions) -{ - - float alpha = 1.5; // Ratio of the memory increase - Index new_len; // New size of the allocated memory - - if(num_expansions == 0 || keep_prev) - new_len = length ; // First time allocate requested - else - new_len = (std::max)(length+1,Index(alpha * length)); - - VectorType old_vec; // Temporary vector to hold the previous values - if (nbElts > 0 ) - old_vec = vec.segment(0,nbElts); - - //Allocate or expand the current vector -#ifdef EIGEN_EXCEPTIONS - try -#endif - { - vec.resize(new_len); - } -#ifdef EIGEN_EXCEPTIONS - catch(std::bad_alloc& ) -#else - if(!vec.size()) -#endif - { - if (!num_expansions) - { - // First time to allocate from LUMemInit() - // Let LUMemInit() deals with it. - return -1; - } - if (keep_prev) - { - // In this case, the memory length should not not be reduced - return new_len; - } - else - { - // Reduce the size and increase again - Index tries = 0; // Number of attempts - do - { - alpha = (alpha + 1)/2; - new_len = (std::max)(length+1,Index(alpha * length)); -#ifdef EIGEN_EXCEPTIONS - try -#endif - { - vec.resize(new_len); - } -#ifdef EIGEN_EXCEPTIONS - catch(std::bad_alloc& ) -#else - if (!vec.size()) -#endif - { - tries += 1; - if ( tries > 10) return new_len; - } - } while (!vec.size()); - } - } - //Copy the previous values to the newly allocated space - if (nbElts > 0) - vec.segment(0, nbElts) = old_vec; - - - length = new_len; - if(num_expansions) ++num_expansions; - return 0; -} - -/** - * \brief Allocate various working space for the numerical factorization phase. - * \param m number of rows of the input matrix - * \param n number of columns - * \param annz number of initial nonzeros in the matrix - * \param lwork if lwork=-1, this routine returns an estimated size of the required memory - * \param glu persistent data to facilitate multiple factors : will be deleted later ?? - * \param fillratio estimated ratio of fill in the factors - * \param panel_size Size of a panel - * \return an estimated size of the required memory if lwork = -1; otherwise, return the size of actually allocated memory when allocation failed, and 0 on success - * \note Unlike SuperLU, this routine does not support successive factorization with the same pattern and the same row permutation - */ -template -Index SparseLUImpl::memInit(Index m, Index n, Index annz, Index lwork, Index fillratio, Index panel_size, GlobalLU_t& glu) -{ - Index& num_expansions = glu.num_expansions; //No memory expansions so far - num_expansions = 0; - glu.nzumax = glu.nzlumax = (std::min)(fillratio * (annz+1) / n, m) * n; // estimated number of nonzeros in U - glu.nzlmax = (std::max)(Index(4), fillratio) * (annz+1) / 4; // estimated nnz in L factor - // Return the estimated size to the user if necessary - Index tempSpace; - tempSpace = (2*panel_size + 4 + LUNoMarker) * m * sizeof(Index) + (panel_size + 1) * m * sizeof(Scalar); - if (lwork == emptyIdxLU) - { - Index estimated_size; - estimated_size = (5 * n + 5) * sizeof(Index) + tempSpace - + (glu.nzlmax + glu.nzumax) * sizeof(Index) + (glu.nzlumax+glu.nzumax) * sizeof(Scalar) + n; - return estimated_size; - } - - // Setup the required space - - // First allocate Integer pointers for L\U factors - glu.xsup.resize(n+1); - glu.supno.resize(n+1); - glu.xlsub.resize(n+1); - glu.xlusup.resize(n+1); - glu.xusub.resize(n+1); - - // Reserve memory for L/U factors - do - { - if( (expand(glu.lusup, glu.nzlumax, 0, 0, num_expansions)<0) - || (expand(glu.ucol, glu.nzumax, 0, 0, num_expansions)<0) - || (expand (glu.lsub, glu.nzlmax, 0, 0, num_expansions)<0) - || (expand (glu.usub, glu.nzumax, 0, 1, num_expansions)<0) ) - { - //Reduce the estimated size and retry - glu.nzlumax /= 2; - glu.nzumax /= 2; - glu.nzlmax /= 2; - if (glu.nzlumax < annz ) return glu.nzlumax; - } - } while (!glu.lusup.size() || !glu.ucol.size() || !glu.lsub.size() || !glu.usub.size()); - - ++num_expansions; - return 0; - -} // end LuMemInit - -/** - * \brief Expand the existing storage - * \param vec vector to expand - * \param[in,out] maxlen On input, previous size of vec (Number of elements to copy ). on output, new size - * \param nbElts current number of elements in the vector. - * \param memtype Type of the element to expand - * \param num_expansions Number of expansions - * \return 0 on success, > 0 size of the memory allocated so far - */ -template -template -Index SparseLUImpl::memXpand(VectorType& vec, Index& maxlen, Index nbElts, MemType memtype, Index& num_expansions) -{ - Index failed_size; - if (memtype == USUB) - failed_size = this->expand(vec, maxlen, nbElts, 1, num_expansions); - else - failed_size = this->expand(vec, maxlen, nbElts, 0, num_expansions); - - if (failed_size) - return failed_size; - - return 0 ; -} - -} // end namespace internal - -} // end namespace Eigen -#endif // EIGEN_SPARSELU_MEMORY diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Structs.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Structs.h deleted file mode 100644 index cf5ec449bec..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Structs.h +++ /dev/null @@ -1,110 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - * NOTE: This file comes from a partly modified version of files slu_[s,d,c,z]defs.h - * -- SuperLU routine (version 4.1) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * November, 2010 - * - * Global data structures used in LU factorization - - * - * nsuper: #supernodes = nsuper + 1, numbered [0, nsuper]. - * (xsup,supno): supno[i] is the supernode no to which i belongs; - * xsup(s) points to the beginning of the s-th supernode. - * e.g. supno 0 1 2 2 3 3 3 4 4 4 4 4 (n=12) - * xsup 0 1 2 4 7 12 - * Note: dfs will be performed on supernode rep. relative to the new - * row pivoting ordering - * - * (xlsub,lsub): lsub[*] contains the compressed subscript of - * rectangular supernodes; xlsub[j] points to the starting - * location of the j-th column in lsub[*]. Note that xlsub - * is indexed by column. - * Storage: original row subscripts - * - * During the course of sparse LU factorization, we also use - * (xlsub,lsub) for the purpose of symmetric pruning. For each - * supernode {s,s+1,...,t=s+r} with first column s and last - * column t, the subscript set - * lsub[j], j=xlsub[s], .., xlsub[s+1]-1 - * is the structure of column s (i.e. structure of this supernode). - * It is used for the storage of numerical values. - * Furthermore, - * lsub[j], j=xlsub[t], .., xlsub[t+1]-1 - * is the structure of the last column t of this supernode. - * It is for the purpose of symmetric pruning. Therefore, the - * structural subscripts can be rearranged without making physical - * interchanges among the numerical values. - * - * However, if the supernode has only one column, then we - * only keep one set of subscripts. For any subscript interchange - * performed, similar interchange must be done on the numerical - * values. - * - * The last column structures (for pruning) will be removed - * after the numercial LU factorization phase. - * - * (xlusup,lusup): lusup[*] contains the numerical values of the - * rectangular supernodes; xlusup[j] points to the starting - * location of the j-th column in storage vector lusup[*] - * Note: xlusup is indexed by column. - * Each rectangular supernode is stored by column-major - * scheme, consistent with Fortran 2-dim array storage. - * - * (xusub,ucol,usub): ucol[*] stores the numerical values of - * U-columns outside the rectangular supernodes. The row - * subscript of nonzero ucol[k] is stored in usub[k]. - * xusub[i] points to the starting location of column i in ucol. - * Storage: new row subscripts; that is subscripts of PA. - */ - -#ifndef EIGEN_LU_STRUCTS -#define EIGEN_LU_STRUCTS -namespace Eigen { -namespace internal { - -typedef enum {LUSUP, UCOL, LSUB, USUB, LLVL, ULVL} MemType; - -template -struct LU_GlobalLU_t { - typedef typename IndexVector::Scalar StorageIndex; - IndexVector xsup; //First supernode column ... xsup(s) points to the beginning of the s-th supernode - IndexVector supno; // Supernode number corresponding to this column (column to supernode mapping) - ScalarVector lusup; // nonzero values of L ordered by columns - IndexVector lsub; // Compressed row indices of L rectangular supernodes. - IndexVector xlusup; // pointers to the beginning of each column in lusup - IndexVector xlsub; // pointers to the beginning of each column in lsub - Index nzlmax; // Current max size of lsub - Index nzlumax; // Current max size of lusup - ScalarVector ucol; // nonzero values of U ordered by columns - IndexVector usub; // row indices of U columns in ucol - IndexVector xusub; // Pointers to the beginning of each column of U in ucol - Index nzumax; // Current max size of ucol - Index n; // Number of columns in the matrix - Index num_expansions; -}; - -// Values to set for performance -struct perfvalues { - Index panel_size; // a panel consists of at most consecutive columns - Index relax; // To control degree of relaxing supernodes. If the number of nodes (columns) - // in a subtree of the elimination tree is less than relax, this subtree is considered - // as one supernode regardless of the row structures of those columns - Index maxsuper; // The maximum size for a supernode in complete LU - Index rowblk; // The minimum row dimension for 2-D blocking to be used; - Index colblk; // The minimum column dimension for 2-D blocking to be used; - Index fillfactor; // The estimated fills factors for L and U, compared with A -}; - -} // end namespace internal - -} // end namespace Eigen -#endif // EIGEN_LU_STRUCTS diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h deleted file mode 100644 index 721e1883ba8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h +++ /dev/null @@ -1,301 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSELU_SUPERNODAL_MATRIX_H -#define EIGEN_SPARSELU_SUPERNODAL_MATRIX_H - -namespace Eigen { -namespace internal { - -/** \ingroup SparseLU_Module - * \brief a class to manipulate the L supernodal factor from the SparseLU factorization - * - * This class contain the data to easily store - * and manipulate the supernodes during the factorization and solution phase of Sparse LU. - * Only the lower triangular matrix has supernodes. - * - * NOTE : This class corresponds to the SCformat structure in SuperLU - * - */ -/* TODO - * InnerIterator as for sparsematrix - * SuperInnerIterator to iterate through all supernodes - * Function for triangular solve - */ -template -class MappedSuperNodalMatrix -{ - public: - typedef _Scalar Scalar; - typedef _StorageIndex StorageIndex; - typedef Matrix IndexVector; - typedef Matrix ScalarVector; - public: - MappedSuperNodalMatrix() - { - - } - MappedSuperNodalMatrix(Index m, Index n, ScalarVector& nzval, IndexVector& nzval_colptr, IndexVector& rowind, - IndexVector& rowind_colptr, IndexVector& col_to_sup, IndexVector& sup_to_col ) - { - setInfos(m, n, nzval, nzval_colptr, rowind, rowind_colptr, col_to_sup, sup_to_col); - } - - ~MappedSuperNodalMatrix() - { - - } - /** - * Set appropriate pointers for the lower triangular supernodal matrix - * These infos are available at the end of the numerical factorization - * FIXME This class will be modified such that it can be use in the course - * of the factorization. - */ - void setInfos(Index m, Index n, ScalarVector& nzval, IndexVector& nzval_colptr, IndexVector& rowind, - IndexVector& rowind_colptr, IndexVector& col_to_sup, IndexVector& sup_to_col ) - { - m_row = m; - m_col = n; - m_nzval = nzval.data(); - m_nzval_colptr = nzval_colptr.data(); - m_rowind = rowind.data(); - m_rowind_colptr = rowind_colptr.data(); - m_nsuper = col_to_sup(n); - m_col_to_sup = col_to_sup.data(); - m_sup_to_col = sup_to_col.data(); - } - - /** - * Number of rows - */ - Index rows() { return m_row; } - - /** - * Number of columns - */ - Index cols() { return m_col; } - - /** - * Return the array of nonzero values packed by column - * - * The size is nnz - */ - Scalar* valuePtr() { return m_nzval; } - - const Scalar* valuePtr() const - { - return m_nzval; - } - /** - * Return the pointers to the beginning of each column in \ref valuePtr() - */ - StorageIndex* colIndexPtr() - { - return m_nzval_colptr; - } - - const StorageIndex* colIndexPtr() const - { - return m_nzval_colptr; - } - - /** - * Return the array of compressed row indices of all supernodes - */ - StorageIndex* rowIndex() { return m_rowind; } - - const StorageIndex* rowIndex() const - { - return m_rowind; - } - - /** - * Return the location in \em rowvaluePtr() which starts each column - */ - StorageIndex* rowIndexPtr() { return m_rowind_colptr; } - - const StorageIndex* rowIndexPtr() const - { - return m_rowind_colptr; - } - - /** - * Return the array of column-to-supernode mapping - */ - StorageIndex* colToSup() { return m_col_to_sup; } - - const StorageIndex* colToSup() const - { - return m_col_to_sup; - } - /** - * Return the array of supernode-to-column mapping - */ - StorageIndex* supToCol() { return m_sup_to_col; } - - const StorageIndex* supToCol() const - { - return m_sup_to_col; - } - - /** - * Return the number of supernodes - */ - Index nsuper() const - { - return m_nsuper; - } - - class InnerIterator; - template - void solveInPlace( MatrixBase&X) const; - - - - - protected: - Index m_row; // Number of rows - Index m_col; // Number of columns - Index m_nsuper; // Number of supernodes - Scalar* m_nzval; //array of nonzero values packed by column - StorageIndex* m_nzval_colptr; //nzval_colptr[j] Stores the location in nzval[] which starts column j - StorageIndex* m_rowind; // Array of compressed row indices of rectangular supernodes - StorageIndex* m_rowind_colptr; //rowind_colptr[j] stores the location in rowind[] which starts column j - StorageIndex* m_col_to_sup; // col_to_sup[j] is the supernode number to which column j belongs - StorageIndex* m_sup_to_col; //sup_to_col[s] points to the starting column of the s-th supernode - - private : -}; - -/** - * \brief InnerIterator class to iterate over nonzero values of the current column in the supernodal matrix L - * - */ -template -class MappedSuperNodalMatrix::InnerIterator -{ - public: - InnerIterator(const MappedSuperNodalMatrix& mat, Index outer) - : m_matrix(mat), - m_outer(outer), - m_supno(mat.colToSup()[outer]), - m_idval(mat.colIndexPtr()[outer]), - m_startidval(m_idval), - m_endidval(mat.colIndexPtr()[outer+1]), - m_idrow(mat.rowIndexPtr()[mat.supToCol()[mat.colToSup()[outer]]]), - m_endidrow(mat.rowIndexPtr()[mat.supToCol()[mat.colToSup()[outer]]+1]) - {} - inline InnerIterator& operator++() - { - m_idval++; - m_idrow++; - return *this; - } - inline Scalar value() const { return m_matrix.valuePtr()[m_idval]; } - - inline Scalar& valueRef() { return const_cast(m_matrix.valuePtr()[m_idval]); } - - inline Index index() const { return m_matrix.rowIndex()[m_idrow]; } - inline Index row() const { return index(); } - inline Index col() const { return m_outer; } - - inline Index supIndex() const { return m_supno; } - - inline operator bool() const - { - return ( (m_idval < m_endidval) && (m_idval >= m_startidval) - && (m_idrow < m_endidrow) ); - } - - protected: - const MappedSuperNodalMatrix& m_matrix; // Supernodal lower triangular matrix - const Index m_outer; // Current column - const Index m_supno; // Current SuperNode number - Index m_idval; // Index to browse the values in the current column - const Index m_startidval; // Start of the column value - const Index m_endidval; // End of the column value - Index m_idrow; // Index to browse the row indices - Index m_endidrow; // End index of row indices of the current column -}; - -/** - * \brief Solve with the supernode triangular matrix - * - */ -template -template -void MappedSuperNodalMatrix::solveInPlace( MatrixBase&X) const -{ - /* Explicit type conversion as the Index type of MatrixBase may be wider than Index */ -// eigen_assert(X.rows() <= NumTraits::highest()); -// eigen_assert(X.cols() <= NumTraits::highest()); - Index n = int(X.rows()); - Index nrhs = Index(X.cols()); - const Scalar * Lval = valuePtr(); // Nonzero values - Matrix work(n, nrhs); // working vector - work.setZero(); - for (Index k = 0; k <= nsuper(); k ++) - { - Index fsupc = supToCol()[k]; // First column of the current supernode - Index istart = rowIndexPtr()[fsupc]; // Pointer index to the subscript of the current column - Index nsupr = rowIndexPtr()[fsupc+1] - istart; // Number of rows in the current supernode - Index nsupc = supToCol()[k+1] - fsupc; // Number of columns in the current supernode - Index nrow = nsupr - nsupc; // Number of rows in the non-diagonal part of the supernode - Index irow; //Current index row - - if (nsupc == 1 ) - { - for (Index j = 0; j < nrhs; j++) - { - InnerIterator it(*this, fsupc); - ++it; // Skip the diagonal element - for (; it; ++it) - { - irow = it.row(); - X(irow, j) -= X(fsupc, j) * it.value(); - } - } - } - else - { - // The supernode has more than one column - Index luptr = colIndexPtr()[fsupc]; - Index lda = colIndexPtr()[fsupc+1] - luptr; - - // Triangular solve - Map, 0, OuterStride<> > A( &(Lval[luptr]), nsupc, nsupc, OuterStride<>(lda) ); - Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); - U = A.template triangularView().solve(U); - - // Matrix-vector product - new (&A) Map, 0, OuterStride<> > ( &(Lval[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); - work.topRows(nrow).noalias() = A * U; - - //Begin Scatter - for (Index j = 0; j < nrhs; j++) - { - Index iptr = istart + nsupc; - for (Index i = 0; i < nrow; i++) - { - irow = rowIndex()[iptr]; - X(irow, j) -= work(i, j); // Scatter operation - work(i, j) = Scalar(0); - iptr++; - } - } - } - } -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SPARSELU_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Utils.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Utils.h deleted file mode 100644 index 9e3dab44d99..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_Utils.h +++ /dev/null @@ -1,80 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -#ifndef EIGEN_SPARSELU_UTILS_H -#define EIGEN_SPARSELU_UTILS_H - -namespace Eigen { -namespace internal { - -/** - * \brief Count Nonzero elements in the factors - */ -template -void SparseLUImpl::countnz(const Index n, Index& nnzL, Index& nnzU, GlobalLU_t& glu) -{ - nnzL = 0; - nnzU = (glu.xusub)(n); - Index nsuper = (glu.supno)(n); - Index jlen; - Index i, j, fsupc; - if (n <= 0 ) return; - // For each supernode - for (i = 0; i <= nsuper; i++) - { - fsupc = glu.xsup(i); - jlen = glu.xlsub(fsupc+1) - glu.xlsub(fsupc); - - for (j = fsupc; j < glu.xsup(i+1); j++) - { - nnzL += jlen; - nnzU += j - fsupc + 1; - jlen--; - } - } -} - -/** - * \brief Fix up the data storage lsub for L-subscripts. - * - * It removes the subscripts sets for structural pruning, - * and applies permutation to the remaining subscripts - * - */ -template -void SparseLUImpl::fixupL(const Index n, const IndexVector& perm_r, GlobalLU_t& glu) -{ - Index fsupc, i, j, k, jstart; - - StorageIndex nextl = 0; - Index nsuper = (glu.supno)(n); - - // For each supernode - for (i = 0; i <= nsuper; i++) - { - fsupc = glu.xsup(i); - jstart = glu.xlsub(fsupc); - glu.xlsub(fsupc) = nextl; - for (j = jstart; j < glu.xlsub(fsupc + 1); j++) - { - glu.lsub(nextl) = perm_r(glu.lsub(j)); // Now indexed into P*A - nextl++; - } - for (k = fsupc+1; k < glu.xsup(i+1); k++) - glu.xlsub(k) = nextl; // other columns in supernode i - } - - glu.xlsub(n) = nextl; -} - -} // end namespace internal - -} // end namespace Eigen -#endif // EIGEN_SPARSELU_UTILS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_bmod.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_bmod.h deleted file mode 100644 index b57f06802e2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_bmod.h +++ /dev/null @@ -1,181 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of xcolumn_bmod.c file in SuperLU - - * -- SuperLU routine (version 3.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * October 15, 2003 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_COLUMN_BMOD_H -#define SPARSELU_COLUMN_BMOD_H - -namespace Eigen { - -namespace internal { -/** - * \brief Performs numeric block updates (sup-col) in topological order - * - * \param jcol current column to update - * \param nseg Number of segments in the U part - * \param dense Store the full representation of the column - * \param tempv working array - * \param segrep segment representative ... - * \param repfnz ??? First nonzero column in each row ??? ... - * \param fpanelc First column in the current panel - * \param glu Global LU data. - * \return 0 - successful return - * > 0 - number of bytes allocated when run out of space - * - */ -template -Index SparseLUImpl::column_bmod(const Index jcol, const Index nseg, BlockScalarVector dense, ScalarVector& tempv, - BlockIndexVector segrep, BlockIndexVector repfnz, Index fpanelc, GlobalLU_t& glu) -{ - Index jsupno, k, ksub, krep, ksupno; - Index lptr, nrow, isub, irow, nextlu, new_next, ufirst; - Index fsupc, nsupc, nsupr, luptr, kfnz, no_zeros; - /* krep = representative of current k-th supernode - * fsupc = first supernodal column - * nsupc = number of columns in a supernode - * nsupr = number of rows in a supernode - * luptr = location of supernodal LU-block in storage - * kfnz = first nonz in the k-th supernodal segment - * no_zeros = no lf leading zeros in a supernodal U-segment - */ - - jsupno = glu.supno(jcol); - // For each nonzero supernode segment of U[*,j] in topological order - k = nseg - 1; - Index d_fsupc; // distance between the first column of the current panel and the - // first column of the current snode - Index fst_col; // First column within small LU update - Index segsize; - for (ksub = 0; ksub < nseg; ksub++) - { - krep = segrep(k); k--; - ksupno = glu.supno(krep); - if (jsupno != ksupno ) - { - // outside the rectangular supernode - fsupc = glu.xsup(ksupno); - fst_col = (std::max)(fsupc, fpanelc); - - // Distance from the current supernode to the current panel; - // d_fsupc = 0 if fsupc > fpanelc - d_fsupc = fst_col - fsupc; - - luptr = glu.xlusup(fst_col) + d_fsupc; - lptr = glu.xlsub(fsupc) + d_fsupc; - - kfnz = repfnz(krep); - kfnz = (std::max)(kfnz, fpanelc); - - segsize = krep - kfnz + 1; - nsupc = krep - fst_col + 1; - nsupr = glu.xlsub(fsupc+1) - glu.xlsub(fsupc); - nrow = nsupr - d_fsupc - nsupc; - Index lda = glu.xlusup(fst_col+1) - glu.xlusup(fst_col); - - - // Perform a triangular solver and block update, - // then scatter the result of sup-col update to dense - no_zeros = kfnz - fst_col; - if(segsize==1) - LU_kernel_bmod<1>::run(segsize, dense, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - else - LU_kernel_bmod::run(segsize, dense, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - } // end if jsupno - } // end for each segment - - // Process the supernodal portion of L\U[*,j] - nextlu = glu.xlusup(jcol); - fsupc = glu.xsup(jsupno); - - // copy the SPA dense into L\U[*,j] - Index mem; - new_next = nextlu + glu.xlsub(fsupc + 1) - glu.xlsub(fsupc); - Index offset = internal::first_multiple(new_next, internal::packet_traits::size) - new_next; - if(offset) - new_next += offset; - while (new_next > glu.nzlumax ) - { - mem = memXpand(glu.lusup, glu.nzlumax, nextlu, LUSUP, glu.num_expansions); - if (mem) return mem; - } - - for (isub = glu.xlsub(fsupc); isub < glu.xlsub(fsupc+1); isub++) - { - irow = glu.lsub(isub); - glu.lusup(nextlu) = dense(irow); - dense(irow) = Scalar(0.0); - ++nextlu; - } - - if(offset) - { - glu.lusup.segment(nextlu,offset).setZero(); - nextlu += offset; - } - glu.xlusup(jcol + 1) = StorageIndex(nextlu); // close L\U(*,jcol); - - /* For more updates within the panel (also within the current supernode), - * should start from the first column of the panel, or the first column - * of the supernode, whichever is bigger. There are two cases: - * 1) fsupc < fpanelc, then fst_col <-- fpanelc - * 2) fsupc >= fpanelc, then fst_col <-- fsupc - */ - fst_col = (std::max)(fsupc, fpanelc); - - if (fst_col < jcol) - { - // Distance between the current supernode and the current panel - // d_fsupc = 0 if fsupc >= fpanelc - d_fsupc = fst_col - fsupc; - - lptr = glu.xlsub(fsupc) + d_fsupc; - luptr = glu.xlusup(fst_col) + d_fsupc; - nsupr = glu.xlsub(fsupc+1) - glu.xlsub(fsupc); // leading dimension - nsupc = jcol - fst_col; // excluding jcol - nrow = nsupr - d_fsupc - nsupc; - - // points to the beginning of jcol in snode L\U(jsupno) - ufirst = glu.xlusup(jcol) + d_fsupc; - Index lda = glu.xlusup(jcol+1) - glu.xlusup(jcol); - MappedMatrixBlock A( &(glu.lusup.data()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); - VectorBlock u(glu.lusup, ufirst, nsupc); - u = A.template triangularView().solve(u); - - new (&A) MappedMatrixBlock ( &(glu.lusup.data()[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); - VectorBlock l(glu.lusup, ufirst+nsupc, nrow); - l.noalias() -= A * u; - - } // End if fst_col - return 0; -} - -} // end namespace internal -} // end namespace Eigen - -#endif // SPARSELU_COLUMN_BMOD_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_dfs.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_dfs.h deleted file mode 100644 index c98b30e323f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_column_dfs.h +++ /dev/null @@ -1,179 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of [s,d,c,z]column_dfs.c file in SuperLU - - * -- SuperLU routine (version 2.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * November 15, 1997 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_COLUMN_DFS_H -#define SPARSELU_COLUMN_DFS_H - -template class SparseLUImpl; -namespace Eigen { - -namespace internal { - -template -struct column_dfs_traits : no_assignment_operator -{ - typedef typename ScalarVector::Scalar Scalar; - typedef typename IndexVector::Scalar StorageIndex; - column_dfs_traits(Index jcol, Index& jsuper, typename SparseLUImpl::GlobalLU_t& glu, SparseLUImpl& luImpl) - : m_jcol(jcol), m_jsuper_ref(jsuper), m_glu(glu), m_luImpl(luImpl) - {} - bool update_segrep(Index /*krep*/, Index /*jj*/) - { - return true; - } - void mem_expand(IndexVector& lsub, Index& nextl, Index chmark) - { - if (nextl >= m_glu.nzlmax) - m_luImpl.memXpand(lsub, m_glu.nzlmax, nextl, LSUB, m_glu.num_expansions); - if (chmark != (m_jcol-1)) m_jsuper_ref = emptyIdxLU; - } - enum { ExpandMem = true }; - - Index m_jcol; - Index& m_jsuper_ref; - typename SparseLUImpl::GlobalLU_t& m_glu; - SparseLUImpl& m_luImpl; -}; - - -/** - * \brief Performs a symbolic factorization on column jcol and decide the supernode boundary - * - * A supernode representative is the last column of a supernode. - * The nonzeros in U[*,j] are segments that end at supernodes representatives. - * The routine returns a list of the supernodal representatives - * in topological order of the dfs that generates them. - * The location of the first nonzero in each supernodal segment - * (supernodal entry location) is also returned. - * - * \param m number of rows in the matrix - * \param jcol Current column - * \param perm_r Row permutation - * \param maxsuper Maximum number of column allowed in a supernode - * \param [in,out] nseg Number of segments in current U[*,j] - new segments appended - * \param lsub_col defines the rhs vector to start the dfs - * \param [in,out] segrep Segment representatives - new segments appended - * \param repfnz First nonzero location in each row - * \param xprune - * \param marker marker[i] == jj, if i was visited during dfs of current column jj; - * \param parent - * \param xplore working array - * \param glu global LU data - * \return 0 success - * > 0 number of bytes allocated when run out of space - * - */ -template -Index SparseLUImpl::column_dfs(const Index m, const Index jcol, IndexVector& perm_r, Index maxsuper, Index& nseg, - BlockIndexVector lsub_col, IndexVector& segrep, BlockIndexVector repfnz, IndexVector& xprune, - IndexVector& marker, IndexVector& parent, IndexVector& xplore, GlobalLU_t& glu) -{ - - Index jsuper = glu.supno(jcol); - Index nextl = glu.xlsub(jcol); - VectorBlock marker2(marker, 2*m, m); - - - column_dfs_traits traits(jcol, jsuper, glu, *this); - - // For each nonzero in A(*,jcol) do dfs - for (Index k = 0; ((k < m) ? lsub_col[k] != emptyIdxLU : false) ; k++) - { - Index krow = lsub_col(k); - lsub_col(k) = emptyIdxLU; - Index kmark = marker2(krow); - - // krow was visited before, go to the next nonz; - if (kmark == jcol) continue; - - dfs_kernel(StorageIndex(jcol), perm_r, nseg, glu.lsub, segrep, repfnz, xprune, marker2, parent, - xplore, glu, nextl, krow, traits); - } // for each nonzero ... - - Index fsupc; - StorageIndex nsuper = glu.supno(jcol); - StorageIndex jcolp1 = StorageIndex(jcol) + 1; - Index jcolm1 = jcol - 1; - - // check to see if j belongs in the same supernode as j-1 - if ( jcol == 0 ) - { // Do nothing for column 0 - nsuper = glu.supno(0) = 0 ; - } - else - { - fsupc = glu.xsup(nsuper); - StorageIndex jptr = glu.xlsub(jcol); // Not yet compressed - StorageIndex jm1ptr = glu.xlsub(jcolm1); - - // Use supernodes of type T2 : see SuperLU paper - if ( (nextl-jptr != jptr-jm1ptr-1) ) jsuper = emptyIdxLU; - - // Make sure the number of columns in a supernode doesn't - // exceed threshold - if ( (jcol - fsupc) >= maxsuper) jsuper = emptyIdxLU; - - /* If jcol starts a new supernode, reclaim storage space in - * glu.lsub from previous supernode. Note we only store - * the subscript set of the first and last columns of - * a supernode. (first for num values, last for pruning) - */ - if (jsuper == emptyIdxLU) - { // starts a new supernode - if ( (fsupc < jcolm1-1) ) - { // >= 3 columns in nsuper - StorageIndex ito = glu.xlsub(fsupc+1); - glu.xlsub(jcolm1) = ito; - StorageIndex istop = ito + jptr - jm1ptr; - xprune(jcolm1) = istop; // intialize xprune(jcol-1) - glu.xlsub(jcol) = istop; - - for (StorageIndex ifrom = jm1ptr; ifrom < nextl; ++ifrom, ++ito) - glu.lsub(ito) = glu.lsub(ifrom); - nextl = ito; // = istop + length(jcol) - } - nsuper++; - glu.supno(jcol) = nsuper; - } // if a new supernode - } // end else: jcol > 0 - - // Tidy up the pointers before exit - glu.xsup(nsuper+1) = jcolp1; - glu.supno(jcolp1) = nsuper; - xprune(jcol) = StorageIndex(nextl); // Intialize upper bound for pruning - glu.xlsub(jcolp1) = StorageIndex(nextl); - - return 0; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_copy_to_ucol.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_copy_to_ucol.h deleted file mode 100644 index c32d8d8b14b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_copy_to_ucol.h +++ /dev/null @@ -1,107 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -/* - - * NOTE: This file is the modified version of [s,d,c,z]copy_to_ucol.c file in SuperLU - - * -- SuperLU routine (version 2.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * November 15, 1997 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_COPY_TO_UCOL_H -#define SPARSELU_COPY_TO_UCOL_H - -namespace Eigen { -namespace internal { - -/** - * \brief Performs numeric block updates (sup-col) in topological order - * - * \param jcol current column to update - * \param nseg Number of segments in the U part - * \param segrep segment representative ... - * \param repfnz First nonzero column in each row ... - * \param perm_r Row permutation - * \param dense Store the full representation of the column - * \param glu Global LU data. - * \return 0 - successful return - * > 0 - number of bytes allocated when run out of space - * - */ -template -Index SparseLUImpl::copy_to_ucol(const Index jcol, const Index nseg, IndexVector& segrep, - BlockIndexVector repfnz ,IndexVector& perm_r, BlockScalarVector dense, GlobalLU_t& glu) -{ - Index ksub, krep, ksupno; - - Index jsupno = glu.supno(jcol); - - // For each nonzero supernode segment of U[*,j] in topological order - Index k = nseg - 1, i; - StorageIndex nextu = glu.xusub(jcol); - Index kfnz, isub, segsize; - Index new_next,irow; - Index fsupc, mem; - for (ksub = 0; ksub < nseg; ksub++) - { - krep = segrep(k); k--; - ksupno = glu.supno(krep); - if (jsupno != ksupno ) // should go into ucol(); - { - kfnz = repfnz(krep); - if (kfnz != emptyIdxLU) - { // Nonzero U-segment - fsupc = glu.xsup(ksupno); - isub = glu.xlsub(fsupc) + kfnz - fsupc; - segsize = krep - kfnz + 1; - new_next = nextu + segsize; - while (new_next > glu.nzumax) - { - mem = memXpand(glu.ucol, glu.nzumax, nextu, UCOL, glu.num_expansions); - if (mem) return mem; - mem = memXpand(glu.usub, glu.nzumax, nextu, USUB, glu.num_expansions); - if (mem) return mem; - - } - - for (i = 0; i < segsize; i++) - { - irow = glu.lsub(isub); - glu.usub(nextu) = perm_r(irow); // Unlike the L part, the U part is stored in its final order - glu.ucol(nextu) = dense(irow); - dense(irow) = Scalar(0.0); - nextu++; - isub++; - } - - } // end nonzero U-segment - - } // end if jsupno - - } // end for each segment - glu.xusub(jcol + 1) = nextu; // close U(*,jcol) - return 0; -} - -} // namespace internal -} // end namespace Eigen - -#endif // SPARSELU_COPY_TO_UCOL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_gemm_kernel.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_gemm_kernel.h deleted file mode 100644 index 95ba7413f29..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_gemm_kernel.h +++ /dev/null @@ -1,280 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSELU_GEMM_KERNEL_H -#define EIGEN_SPARSELU_GEMM_KERNEL_H - -namespace Eigen { - -namespace internal { - - -/** \internal - * A general matrix-matrix product kernel optimized for the SparseLU factorization. - * - A, B, and C must be column major - * - lda and ldc must be multiples of the respective packet size - * - C must have the same alignment as A - */ -template -EIGEN_DONT_INLINE -void sparselu_gemm(Index m, Index n, Index d, const Scalar* A, Index lda, const Scalar* B, Index ldb, Scalar* C, Index ldc) -{ - using namespace Eigen::internal; - - typedef typename packet_traits::type Packet; - enum { - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - PacketSize = packet_traits::size, - PM = 8, // peeling in M - RN = 2, // register blocking - RK = NumberOfRegisters>=16 ? 4 : 2, // register blocking - BM = 4096/sizeof(Scalar), // number of rows of A-C per chunk - SM = PM*PacketSize // step along M - }; - Index d_end = (d/RK)*RK; // number of columns of A (rows of B) suitable for full register blocking - Index n_end = (n/RN)*RN; // number of columns of B-C suitable for processing RN columns at once - Index i0 = internal::first_default_aligned(A,m); - - eigen_internal_assert(((lda%PacketSize)==0) && ((ldc%PacketSize)==0) && (i0==internal::first_default_aligned(C,m))); - - // handle the non aligned rows of A and C without any optimization: - for(Index i=0; i(BM, m-ib); // actual number of rows - Index actual_b_end1 = (actual_b/SM)*SM; // actual number of rows suitable for peeling - Index actual_b_end2 = (actual_b/PacketSize)*PacketSize; // actual number of rows suitable for vectorization - - // Let's process two columns of B-C at once - for(Index j=0; j(Bc0[0]); } - { b10 = pset1(Bc0[1]); } - if(RK==4) { b20 = pset1(Bc0[2]); } - if(RK==4) { b30 = pset1(Bc0[3]); } - { b01 = pset1(Bc1[0]); } - { b11 = pset1(Bc1[1]); } - if(RK==4) { b21 = pset1(Bc1[2]); } - if(RK==4) { b31 = pset1(Bc1[3]); } - - Packet a0, a1, a2, a3, c0, c1, t0, t1; - - const Scalar* A0 = A+ib+(k+0)*lda; - const Scalar* A1 = A+ib+(k+1)*lda; - const Scalar* A2 = A+ib+(k+2)*lda; - const Scalar* A3 = A+ib+(k+3)*lda; - - Scalar* C0 = C+ib+(j+0)*ldc; - Scalar* C1 = C+ib+(j+1)*ldc; - - a0 = pload(A0); - a1 = pload(A1); - if(RK==4) - { - a2 = pload(A2); - a3 = pload(A3); - } - else - { - // workaround "may be used uninitialized in this function" warning - a2 = a3 = a0; - } - -#define KMADD(c, a, b, tmp) {tmp = b; tmp = pmul(a,tmp); c = padd(c,tmp);} -#define WORK(I) \ - c0 = pload(C0+i+(I)*PacketSize); \ - c1 = pload(C1+i+(I)*PacketSize); \ - KMADD(c0, a0, b00, t0) \ - KMADD(c1, a0, b01, t1) \ - a0 = pload(A0+i+(I+1)*PacketSize); \ - KMADD(c0, a1, b10, t0) \ - KMADD(c1, a1, b11, t1) \ - a1 = pload(A1+i+(I+1)*PacketSize); \ - if(RK==4){ KMADD(c0, a2, b20, t0) }\ - if(RK==4){ KMADD(c1, a2, b21, t1) }\ - if(RK==4){ a2 = pload(A2+i+(I+1)*PacketSize); }\ - if(RK==4){ KMADD(c0, a3, b30, t0) }\ - if(RK==4){ KMADD(c1, a3, b31, t1) }\ - if(RK==4){ a3 = pload(A3+i+(I+1)*PacketSize); }\ - pstore(C0+i+(I)*PacketSize, c0); \ - pstore(C1+i+(I)*PacketSize, c1) - - // process rows of A' - C' with aggressive vectorization and peeling - for(Index i=0; i0) - { - const Scalar* Bc0 = B+(n-1)*ldb; - - for(Index k=0; k(Bc0[0]); - b10 = pset1(Bc0[1]); - if(RK==4) b20 = pset1(Bc0[2]); - if(RK==4) b30 = pset1(Bc0[3]); - - Packet a0, a1, a2, a3, c0, t0/*, t1*/; - - const Scalar* A0 = A+ib+(k+0)*lda; - const Scalar* A1 = A+ib+(k+1)*lda; - const Scalar* A2 = A+ib+(k+2)*lda; - const Scalar* A3 = A+ib+(k+3)*lda; - - Scalar* C0 = C+ib+(n_end)*ldc; - - a0 = pload(A0); - a1 = pload(A1); - if(RK==4) - { - a2 = pload(A2); - a3 = pload(A3); - } - else - { - // workaround "may be used uninitialized in this function" warning - a2 = a3 = a0; - } - -#define WORK(I) \ - c0 = pload(C0+i+(I)*PacketSize); \ - KMADD(c0, a0, b00, t0) \ - a0 = pload(A0+i+(I+1)*PacketSize); \ - KMADD(c0, a1, b10, t0) \ - a1 = pload(A1+i+(I+1)*PacketSize); \ - if(RK==4){ KMADD(c0, a2, b20, t0) }\ - if(RK==4){ a2 = pload(A2+i+(I+1)*PacketSize); }\ - if(RK==4){ KMADD(c0, a3, b30, t0) }\ - if(RK==4){ a3 = pload(A3+i+(I+1)*PacketSize); }\ - pstore(C0+i+(I)*PacketSize, c0); - - // agressive vectorization and peeling - for(Index i=0; i0) - { - for(Index j=0; j1 ? Aligned : 0 - }; - typedef Map, Alignment > MapVector; - typedef Map, Alignment > ConstMapVector; - if(rd==1) MapVector(C+j*ldc+ib,actual_b) += B[0+d_end+j*ldb] * ConstMapVector(A+(d_end+0)*lda+ib, actual_b); - - else if(rd==2) MapVector(C+j*ldc+ib,actual_b) += B[0+d_end+j*ldb] * ConstMapVector(A+(d_end+0)*lda+ib, actual_b) - + B[1+d_end+j*ldb] * ConstMapVector(A+(d_end+1)*lda+ib, actual_b); - - else MapVector(C+j*ldc+ib,actual_b) += B[0+d_end+j*ldb] * ConstMapVector(A+(d_end+0)*lda+ib, actual_b) - + B[1+d_end+j*ldb] * ConstMapVector(A+(d_end+1)*lda+ib, actual_b) - + B[2+d_end+j*ldb] * ConstMapVector(A+(d_end+2)*lda+ib, actual_b); - } - } - - } // blocking on the rows of A and C -} -#undef KMADD - -} // namespace internal - -} // namespace Eigen - -#endif // EIGEN_SPARSELU_GEMM_KERNEL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_heap_relax_snode.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_heap_relax_snode.h deleted file mode 100644 index 6f75d500e5f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_heap_relax_snode.h +++ /dev/null @@ -1,126 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* This file is a modified version of heap_relax_snode.c file in SuperLU - * -- SuperLU routine (version 3.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * October 15, 2003 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ - -#ifndef SPARSELU_HEAP_RELAX_SNODE_H -#define SPARSELU_HEAP_RELAX_SNODE_H - -namespace Eigen { -namespace internal { - -/** - * \brief Identify the initial relaxed supernodes - * - * This routine applied to a symmetric elimination tree. - * It assumes that the matrix has been reordered according to the postorder of the etree - * \param n The number of columns - * \param et elimination tree - * \param relax_columns Maximum number of columns allowed in a relaxed snode - * \param descendants Number of descendants of each node in the etree - * \param relax_end last column in a supernode - */ -template -void SparseLUImpl::heap_relax_snode (const Index n, IndexVector& et, const Index relax_columns, IndexVector& descendants, IndexVector& relax_end) -{ - - // The etree may not be postordered, but its heap ordered - IndexVector post; - internal::treePostorder(StorageIndex(n), et, post); // Post order etree - IndexVector inv_post(n+1); - for (StorageIndex i = 0; i < n+1; ++i) inv_post(post(i)) = i; // inv_post = post.inverse()??? - - // Renumber etree in postorder - IndexVector iwork(n); - IndexVector et_save(n+1); - for (Index i = 0; i < n; ++i) - { - iwork(post(i)) = post(et(i)); - } - et_save = et; // Save the original etree - et = iwork; - - // compute the number of descendants of each node in the etree - relax_end.setConstant(emptyIdxLU); - Index j, parent; - descendants.setZero(); - for (j = 0; j < n; j++) - { - parent = et(j); - if (parent != n) // not the dummy root - descendants(parent) += descendants(j) + 1; - } - // Identify the relaxed supernodes by postorder traversal of the etree - Index snode_start; // beginning of a snode - StorageIndex k; - Index nsuper_et_post = 0; // Number of relaxed snodes in postordered etree - Index nsuper_et = 0; // Number of relaxed snodes in the original etree - StorageIndex l; - for (j = 0; j < n; ) - { - parent = et(j); - snode_start = j; - while ( parent != n && descendants(parent) < relax_columns ) - { - j = parent; - parent = et(j); - } - // Found a supernode in postordered etree, j is the last column - ++nsuper_et_post; - k = StorageIndex(n); - for (Index i = snode_start; i <= j; ++i) - k = (std::min)(k, inv_post(i)); - l = inv_post(j); - if ( (l - k) == (j - snode_start) ) // Same number of columns in the snode - { - // This is also a supernode in the original etree - relax_end(k) = l; // Record last column - ++nsuper_et; - } - else - { - for (Index i = snode_start; i <= j; ++i) - { - l = inv_post(i); - if (descendants(i) == 0) - { - relax_end(l) = l; - ++nsuper_et; - } - } - } - j++; - // Search for a new leaf - while (descendants(j) != 0 && j < n) j++; - } // End postorder traversal of the etree - - // Recover the original etree - et = et_save; -} - -} // end namespace internal - -} // end namespace Eigen -#endif // SPARSELU_HEAP_RELAX_SNODE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_kernel_bmod.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_kernel_bmod.h deleted file mode 100644 index 8c1b3e8bc67..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_kernel_bmod.h +++ /dev/null @@ -1,130 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef SPARSELU_KERNEL_BMOD_H -#define SPARSELU_KERNEL_BMOD_H - -namespace Eigen { -namespace internal { - -template struct LU_kernel_bmod -{ - /** \internal - * \brief Performs numeric block updates from a given supernode to a single column - * - * \param segsize Size of the segment (and blocks ) to use for updates - * \param[in,out] dense Packed values of the original matrix - * \param tempv temporary vector to use for updates - * \param lusup array containing the supernodes - * \param lda Leading dimension in the supernode - * \param nrow Number of rows in the rectangular part of the supernode - * \param lsub compressed row subscripts of supernodes - * \param lptr pointer to the first column of the current supernode in lsub - * \param no_zeros Number of nonzeros elements before the diagonal part of the supernode - */ - template - static EIGEN_DONT_INLINE void run(const Index segsize, BlockScalarVector& dense, ScalarVector& tempv, ScalarVector& lusup, Index& luptr, const Index lda, - const Index nrow, IndexVector& lsub, const Index lptr, const Index no_zeros); -}; - -template -template -EIGEN_DONT_INLINE void LU_kernel_bmod::run(const Index segsize, BlockScalarVector& dense, ScalarVector& tempv, ScalarVector& lusup, Index& luptr, const Index lda, - const Index nrow, IndexVector& lsub, const Index lptr, const Index no_zeros) -{ - typedef typename ScalarVector::Scalar Scalar; - // First, copy U[*,j] segment from dense(*) to tempv(*) - // The result of triangular solve is in tempv[*]; - // The result of matric-vector update is in dense[*] - Index isub = lptr + no_zeros; - Index i; - Index irow; - for (i = 0; i < ((SegSizeAtCompileTime==Dynamic)?segsize:SegSizeAtCompileTime); i++) - { - irow = lsub(isub); - tempv(i) = dense(irow); - ++isub; - } - // Dense triangular solve -- start effective triangle - luptr += lda * no_zeros + no_zeros; - // Form Eigen matrix and vector - Map, 0, OuterStride<> > A( &(lusup.data()[luptr]), segsize, segsize, OuterStride<>(lda) ); - Map > u(tempv.data(), segsize); - - u = A.template triangularView().solve(u); - - // Dense matrix-vector product y <-- B*x - luptr += segsize; - const Index PacketSize = internal::packet_traits::size; - Index ldl = internal::first_multiple(nrow, PacketSize); - Map, 0, OuterStride<> > B( &(lusup.data()[luptr]), nrow, segsize, OuterStride<>(lda) ); - Index aligned_offset = internal::first_default_aligned(tempv.data()+segsize, PacketSize); - Index aligned_with_B_offset = (PacketSize-internal::first_default_aligned(B.data(), PacketSize))%PacketSize; - Map, 0, OuterStride<> > l(tempv.data()+segsize+aligned_offset+aligned_with_B_offset, nrow, OuterStride<>(ldl) ); - - l.setZero(); - internal::sparselu_gemm(l.rows(), l.cols(), B.cols(), B.data(), B.outerStride(), u.data(), u.outerStride(), l.data(), l.outerStride()); - - // Scatter tempv[] into SPA dense[] as a temporary storage - isub = lptr + no_zeros; - for (i = 0; i < ((SegSizeAtCompileTime==Dynamic)?segsize:SegSizeAtCompileTime); i++) - { - irow = lsub(isub++); - dense(irow) = tempv(i); - } - - // Scatter l into SPA dense[] - for (i = 0; i < nrow; i++) - { - irow = lsub(isub++); - dense(irow) -= l(i); - } -} - -template <> struct LU_kernel_bmod<1> -{ - template - static EIGEN_DONT_INLINE void run(const Index /*segsize*/, BlockScalarVector& dense, ScalarVector& /*tempv*/, ScalarVector& lusup, Index& luptr, - const Index lda, const Index nrow, IndexVector& lsub, const Index lptr, const Index no_zeros); -}; - - -template -EIGEN_DONT_INLINE void LU_kernel_bmod<1>::run(const Index /*segsize*/, BlockScalarVector& dense, ScalarVector& /*tempv*/, ScalarVector& lusup, Index& luptr, - const Index lda, const Index nrow, IndexVector& lsub, const Index lptr, const Index no_zeros) -{ - typedef typename ScalarVector::Scalar Scalar; - typedef typename IndexVector::Scalar StorageIndex; - Scalar f = dense(lsub(lptr + no_zeros)); - luptr += lda * no_zeros + no_zeros + 1; - const Scalar* a(lusup.data() + luptr); - const StorageIndex* irow(lsub.data()+lptr + no_zeros + 1); - Index i = 0; - for (; i+1 < nrow; i+=2) - { - Index i0 = *(irow++); - Index i1 = *(irow++); - Scalar a0 = *(a++); - Scalar a1 = *(a++); - Scalar d0 = dense.coeff(i0); - Scalar d1 = dense.coeff(i1); - d0 -= f*a0; - d1 -= f*a1; - dense.coeffRef(i0) = d0; - dense.coeffRef(i1) = d1; - } - if(i -// Copyright (C) 2012 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of [s,d,c,z]panel_bmod.c file in SuperLU - - * -- SuperLU routine (version 3.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * October 15, 2003 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_PANEL_BMOD_H -#define SPARSELU_PANEL_BMOD_H - -namespace Eigen { -namespace internal { - -/** - * \brief Performs numeric block updates (sup-panel) in topological order. - * - * Before entering this routine, the original nonzeros in the panel - * were already copied i nto the spa[m,w] - * - * \param m number of rows in the matrix - * \param w Panel size - * \param jcol Starting column of the panel - * \param nseg Number of segments in the U part - * \param dense Store the full representation of the panel - * \param tempv working array - * \param segrep segment representative... first row in the segment - * \param repfnz First nonzero rows - * \param glu Global LU data. - * - * - */ -template -void SparseLUImpl::panel_bmod(const Index m, const Index w, const Index jcol, - const Index nseg, ScalarVector& dense, ScalarVector& tempv, - IndexVector& segrep, IndexVector& repfnz, GlobalLU_t& glu) -{ - - Index ksub,jj,nextl_col; - Index fsupc, nsupc, nsupr, nrow; - Index krep, kfnz; - Index lptr; // points to the row subscripts of a supernode - Index luptr; // ... - Index segsize,no_zeros ; - // For each nonz supernode segment of U[*,j] in topological order - Index k = nseg - 1; - const Index PacketSize = internal::packet_traits::size; - - for (ksub = 0; ksub < nseg; ksub++) - { // For each updating supernode - /* krep = representative of current k-th supernode - * fsupc = first supernodal column - * nsupc = number of columns in a supernode - * nsupr = number of rows in a supernode - */ - krep = segrep(k); k--; - fsupc = glu.xsup(glu.supno(krep)); - nsupc = krep - fsupc + 1; - nsupr = glu.xlsub(fsupc+1) - glu.xlsub(fsupc); - nrow = nsupr - nsupc; - lptr = glu.xlsub(fsupc); - - // loop over the panel columns to detect the actual number of columns and rows - Index u_rows = 0; - Index u_cols = 0; - for (jj = jcol; jj < jcol + w; jj++) - { - nextl_col = (jj-jcol) * m; - VectorBlock repfnz_col(repfnz, nextl_col, m); // First nonzero column index for each row - - kfnz = repfnz_col(krep); - if ( kfnz == emptyIdxLU ) - continue; // skip any zero segment - - segsize = krep - kfnz + 1; - u_cols++; - u_rows = (std::max)(segsize,u_rows); - } - - if(nsupc >= 2) - { - Index ldu = internal::first_multiple(u_rows, PacketSize); - Map > U(tempv.data(), u_rows, u_cols, OuterStride<>(ldu)); - - // gather U - Index u_col = 0; - for (jj = jcol; jj < jcol + w; jj++) - { - nextl_col = (jj-jcol) * m; - VectorBlock repfnz_col(repfnz, nextl_col, m); // First nonzero column index for each row - VectorBlock dense_col(dense, nextl_col, m); // Scatter/gather entire matrix column from/to here - - kfnz = repfnz_col(krep); - if ( kfnz == emptyIdxLU ) - continue; // skip any zero segment - - segsize = krep - kfnz + 1; - luptr = glu.xlusup(fsupc); - no_zeros = kfnz - fsupc; - - Index isub = lptr + no_zeros; - Index off = u_rows-segsize; - for (Index i = 0; i < off; i++) U(i,u_col) = 0; - for (Index i = 0; i < segsize; i++) - { - Index irow = glu.lsub(isub); - U(i+off,u_col) = dense_col(irow); - ++isub; - } - u_col++; - } - // solve U = A^-1 U - luptr = glu.xlusup(fsupc); - Index lda = glu.xlusup(fsupc+1) - glu.xlusup(fsupc); - no_zeros = (krep - u_rows + 1) - fsupc; - luptr += lda * no_zeros + no_zeros; - MappedMatrixBlock A(glu.lusup.data()+luptr, u_rows, u_rows, OuterStride<>(lda) ); - U = A.template triangularView().solve(U); - - // update - luptr += u_rows; - MappedMatrixBlock B(glu.lusup.data()+luptr, nrow, u_rows, OuterStride<>(lda) ); - eigen_assert(tempv.size()>w*ldu + nrow*w + 1); - - Index ldl = internal::first_multiple(nrow, PacketSize); - Index offset = (PacketSize-internal::first_default_aligned(B.data(), PacketSize)) % PacketSize; - MappedMatrixBlock L(tempv.data()+w*ldu+offset, nrow, u_cols, OuterStride<>(ldl)); - - L.setZero(); - internal::sparselu_gemm(L.rows(), L.cols(), B.cols(), B.data(), B.outerStride(), U.data(), U.outerStride(), L.data(), L.outerStride()); - - // scatter U and L - u_col = 0; - for (jj = jcol; jj < jcol + w; jj++) - { - nextl_col = (jj-jcol) * m; - VectorBlock repfnz_col(repfnz, nextl_col, m); // First nonzero column index for each row - VectorBlock dense_col(dense, nextl_col, m); // Scatter/gather entire matrix column from/to here - - kfnz = repfnz_col(krep); - if ( kfnz == emptyIdxLU ) - continue; // skip any zero segment - - segsize = krep - kfnz + 1; - no_zeros = kfnz - fsupc; - Index isub = lptr + no_zeros; - - Index off = u_rows-segsize; - for (Index i = 0; i < segsize; i++) - { - Index irow = glu.lsub(isub++); - dense_col(irow) = U.coeff(i+off,u_col); - U.coeffRef(i+off,u_col) = 0; - } - - // Scatter l into SPA dense[] - for (Index i = 0; i < nrow; i++) - { - Index irow = glu.lsub(isub++); - dense_col(irow) -= L.coeff(i,u_col); - L.coeffRef(i,u_col) = 0; - } - u_col++; - } - } - else // level 2 only - { - // Sequence through each column in the panel - for (jj = jcol; jj < jcol + w; jj++) - { - nextl_col = (jj-jcol) * m; - VectorBlock repfnz_col(repfnz, nextl_col, m); // First nonzero column index for each row - VectorBlock dense_col(dense, nextl_col, m); // Scatter/gather entire matrix column from/to here - - kfnz = repfnz_col(krep); - if ( kfnz == emptyIdxLU ) - continue; // skip any zero segment - - segsize = krep - kfnz + 1; - luptr = glu.xlusup(fsupc); - - Index lda = glu.xlusup(fsupc+1)-glu.xlusup(fsupc);// nsupr - - // Perform a trianglar solve and block update, - // then scatter the result of sup-col update to dense[] - no_zeros = kfnz - fsupc; - if(segsize==1) LU_kernel_bmod<1>::run(segsize, dense_col, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - else if(segsize==2) LU_kernel_bmod<2>::run(segsize, dense_col, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - else if(segsize==3) LU_kernel_bmod<3>::run(segsize, dense_col, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - else LU_kernel_bmod::run(segsize, dense_col, tempv, glu.lusup, luptr, lda, nrow, glu.lsub, lptr, no_zeros); - } // End for each column in the panel - } - - } // End for each updating supernode -} // end panel bmod - -} // end namespace internal - -} // end namespace Eigen - -#endif // SPARSELU_PANEL_BMOD_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_panel_dfs.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_panel_dfs.h deleted file mode 100644 index 155df733687..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_panel_dfs.h +++ /dev/null @@ -1,258 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of [s,d,c,z]panel_dfs.c file in SuperLU - - * -- SuperLU routine (version 2.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * November 15, 1997 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_PANEL_DFS_H -#define SPARSELU_PANEL_DFS_H - -namespace Eigen { - -namespace internal { - -template -struct panel_dfs_traits -{ - typedef typename IndexVector::Scalar StorageIndex; - panel_dfs_traits(Index jcol, StorageIndex* marker) - : m_jcol(jcol), m_marker(marker) - {} - bool update_segrep(Index krep, StorageIndex jj) - { - if(m_marker[krep] -template -void SparseLUImpl::dfs_kernel(const StorageIndex jj, IndexVector& perm_r, - Index& nseg, IndexVector& panel_lsub, IndexVector& segrep, - Ref repfnz_col, IndexVector& xprune, Ref marker, IndexVector& parent, - IndexVector& xplore, GlobalLU_t& glu, - Index& nextl_col, Index krow, Traits& traits - ) -{ - - StorageIndex kmark = marker(krow); - - // For each unmarked krow of jj - marker(krow) = jj; - StorageIndex kperm = perm_r(krow); - if (kperm == emptyIdxLU ) { - // krow is in L : place it in structure of L(*, jj) - panel_lsub(nextl_col++) = StorageIndex(krow); // krow is indexed into A - - traits.mem_expand(panel_lsub, nextl_col, kmark); - } - else - { - // krow is in U : if its supernode-representative krep - // has been explored, update repfnz(*) - // krep = supernode representative of the current row - StorageIndex krep = glu.xsup(glu.supno(kperm)+1) - 1; - // First nonzero element in the current column: - StorageIndex myfnz = repfnz_col(krep); - - if (myfnz != emptyIdxLU ) - { - // Representative visited before - if (myfnz > kperm ) repfnz_col(krep) = kperm; - - } - else - { - // Otherwise, perform dfs starting at krep - StorageIndex oldrep = emptyIdxLU; - parent(krep) = oldrep; - repfnz_col(krep) = kperm; - StorageIndex xdfs = glu.xlsub(krep); - Index maxdfs = xprune(krep); - - StorageIndex kpar; - do - { - // For each unmarked kchild of krep - while (xdfs < maxdfs) - { - StorageIndex kchild = glu.lsub(xdfs); - xdfs++; - StorageIndex chmark = marker(kchild); - - if (chmark != jj ) - { - marker(kchild) = jj; - StorageIndex chperm = perm_r(kchild); - - if (chperm == emptyIdxLU) - { - // case kchild is in L: place it in L(*, j) - panel_lsub(nextl_col++) = kchild; - traits.mem_expand(panel_lsub, nextl_col, chmark); - } - else - { - // case kchild is in U : - // chrep = its supernode-rep. If its rep has been explored, - // update its repfnz(*) - StorageIndex chrep = glu.xsup(glu.supno(chperm)+1) - 1; - myfnz = repfnz_col(chrep); - - if (myfnz != emptyIdxLU) - { // Visited before - if (myfnz > chperm) - repfnz_col(chrep) = chperm; - } - else - { // Cont. dfs at snode-rep of kchild - xplore(krep) = xdfs; - oldrep = krep; - krep = chrep; // Go deeper down G(L) - parent(krep) = oldrep; - repfnz_col(krep) = chperm; - xdfs = glu.xlsub(krep); - maxdfs = xprune(krep); - - } // end if myfnz != -1 - } // end if chperm == -1 - - } // end if chmark !=jj - } // end while xdfs < maxdfs - - // krow has no more unexplored nbrs : - // Place snode-rep krep in postorder DFS, if this - // segment is seen for the first time. (Note that - // "repfnz(krep)" may change later.) - // Baktrack dfs to its parent - if(traits.update_segrep(krep,jj)) - //if (marker1(krep) < jcol ) - { - segrep(nseg) = krep; - ++nseg; - //marker1(krep) = jj; - } - - kpar = parent(krep); // Pop recursion, mimic recursion - if (kpar == emptyIdxLU) - break; // dfs done - krep = kpar; - xdfs = xplore(krep); - maxdfs = xprune(krep); - - } while (kpar != emptyIdxLU); // Do until empty stack - - } // end if (myfnz = -1) - - } // end if (kperm == -1) -} - -/** - * \brief Performs a symbolic factorization on a panel of columns [jcol, jcol+w) - * - * A supernode representative is the last column of a supernode. - * The nonzeros in U[*,j] are segments that end at supernodes representatives - * - * The routine returns a list of the supernodal representatives - * in topological order of the dfs that generates them. This list is - * a superset of the topological order of each individual column within - * the panel. - * The location of the first nonzero in each supernodal segment - * (supernodal entry location) is also returned. Each column has - * a separate list for this purpose. - * - * Two markers arrays are used for dfs : - * marker[i] == jj, if i was visited during dfs of current column jj; - * marker1[i] >= jcol, if i was visited by earlier columns in this panel; - * - * \param[in] m number of rows in the matrix - * \param[in] w Panel size - * \param[in] jcol Starting column of the panel - * \param[in] A Input matrix in column-major storage - * \param[in] perm_r Row permutation - * \param[out] nseg Number of U segments - * \param[out] dense Accumulate the column vectors of the panel - * \param[out] panel_lsub Subscripts of the row in the panel - * \param[out] segrep Segment representative i.e first nonzero row of each segment - * \param[out] repfnz First nonzero location in each row - * \param[out] xprune The pruned elimination tree - * \param[out] marker work vector - * \param parent The elimination tree - * \param xplore work vector - * \param glu The global data structure - * - */ - -template -void SparseLUImpl::panel_dfs(const Index m, const Index w, const Index jcol, MatrixType& A, IndexVector& perm_r, Index& nseg, ScalarVector& dense, IndexVector& panel_lsub, IndexVector& segrep, IndexVector& repfnz, IndexVector& xprune, IndexVector& marker, IndexVector& parent, IndexVector& xplore, GlobalLU_t& glu) -{ - Index nextl_col; // Next available position in panel_lsub[*,jj] - - // Initialize pointers - VectorBlock marker1(marker, m, m); - nseg = 0; - - panel_dfs_traits traits(jcol, marker1.data()); - - // For each column in the panel - for (StorageIndex jj = StorageIndex(jcol); jj < jcol + w; jj++) - { - nextl_col = (jj - jcol) * m; - - VectorBlock repfnz_col(repfnz, nextl_col, m); // First nonzero location in each row - VectorBlock dense_col(dense,nextl_col, m); // Accumulate a column vector here - - - // For each nnz in A[*, jj] do depth first search - for (typename MatrixType::InnerIterator it(A, jj); it; ++it) - { - Index krow = it.row(); - dense_col(krow) = it.value(); - - StorageIndex kmark = marker(krow); - if (kmark == jj) - continue; // krow visited before, go to the next nonzero - - dfs_kernel(jj, perm_r, nseg, panel_lsub, segrep, repfnz_col, xprune, marker, parent, - xplore, glu, nextl_col, krow, traits); - }// end for nonzeros in column jj - - } // end for column jj -} - -} // end namespace internal -} // end namespace Eigen - -#endif // SPARSELU_PANEL_DFS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pivotL.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pivotL.h deleted file mode 100644 index a86dac93fa9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pivotL.h +++ /dev/null @@ -1,137 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of xpivotL.c file in SuperLU - - * -- SuperLU routine (version 3.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * October 15, 2003 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_PIVOTL_H -#define SPARSELU_PIVOTL_H - -namespace Eigen { -namespace internal { - -/** - * \brief Performs the numerical pivotin on the current column of L, and the CDIV operation. - * - * Pivot policy : - * (1) Compute thresh = u * max_(i>=j) abs(A_ij); - * (2) IF user specifies pivot row k and abs(A_kj) >= thresh THEN - * pivot row = k; - * ELSE IF abs(A_jj) >= thresh THEN - * pivot row = j; - * ELSE - * pivot row = m; - * - * Note: If you absolutely want to use a given pivot order, then set u=0.0. - * - * \param jcol The current column of L - * \param diagpivotthresh diagonal pivoting threshold - * \param[in,out] perm_r Row permutation (threshold pivoting) - * \param[in] iperm_c column permutation - used to finf diagonal of Pc*A*Pc' - * \param[out] pivrow The pivot row - * \param glu Global LU data - * \return 0 if success, i > 0 if U(i,i) is exactly zero - * - */ -template -Index SparseLUImpl::pivotL(const Index jcol, const RealScalar& diagpivotthresh, IndexVector& perm_r, IndexVector& iperm_c, Index& pivrow, GlobalLU_t& glu) -{ - - Index fsupc = (glu.xsup)((glu.supno)(jcol)); // First column in the supernode containing the column jcol - Index nsupc = jcol - fsupc; // Number of columns in the supernode portion, excluding jcol; nsupc >=0 - Index lptr = glu.xlsub(fsupc); // pointer to the starting location of the row subscripts for this supernode portion - Index nsupr = glu.xlsub(fsupc+1) - lptr; // Number of rows in the supernode - Index lda = glu.xlusup(fsupc+1) - glu.xlusup(fsupc); // leading dimension - Scalar* lu_sup_ptr = &(glu.lusup.data()[glu.xlusup(fsupc)]); // Start of the current supernode - Scalar* lu_col_ptr = &(glu.lusup.data()[glu.xlusup(jcol)]); // Start of jcol in the supernode - StorageIndex* lsub_ptr = &(glu.lsub.data()[lptr]); // Start of row indices of the supernode - - // Determine the largest abs numerical value for partial pivoting - Index diagind = iperm_c(jcol); // diagonal index - RealScalar pivmax(-1.0); - Index pivptr = nsupc; - Index diag = emptyIdxLU; - RealScalar rtemp; - Index isub, icol, itemp, k; - for (isub = nsupc; isub < nsupr; ++isub) { - using std::abs; - rtemp = abs(lu_col_ptr[isub]); - if (rtemp > pivmax) { - pivmax = rtemp; - pivptr = isub; - } - if (lsub_ptr[isub] == diagind) diag = isub; - } - - // Test for singularity - if ( pivmax <= RealScalar(0.0) ) { - // if pivmax == -1, the column is structurally empty, otherwise it is only numerically zero - pivrow = pivmax < RealScalar(0.0) ? diagind : lsub_ptr[pivptr]; - perm_r(pivrow) = StorageIndex(jcol); - return (jcol+1); - } - - RealScalar thresh = diagpivotthresh * pivmax; - - // Choose appropriate pivotal element - - { - // Test if the diagonal element can be used as a pivot (given the threshold value) - if (diag >= 0 ) - { - // Diagonal element exists - using std::abs; - rtemp = abs(lu_col_ptr[diag]); - if (rtemp != RealScalar(0.0) && rtemp >= thresh) pivptr = diag; - } - pivrow = lsub_ptr[pivptr]; - } - - // Record pivot row - perm_r(pivrow) = StorageIndex(jcol); - // Interchange row subscripts - if (pivptr != nsupc ) - { - std::swap( lsub_ptr[pivptr], lsub_ptr[nsupc] ); - // Interchange numerical values as well, for the two rows in the whole snode - // such that L is indexed the same way as A - for (icol = 0; icol <= nsupc; icol++) - { - itemp = pivptr + icol * lda; - std::swap(lu_sup_ptr[itemp], lu_sup_ptr[nsupc + icol * lda]); - } - } - // cdiv operations - Scalar temp = Scalar(1.0) / lu_col_ptr[nsupc]; - for (k = nsupc+1; k < nsupr; k++) - lu_col_ptr[k] *= temp; - return 0; -} - -} // end namespace internal -} // end namespace Eigen - -#endif // SPARSELU_PIVOTL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pruneL.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pruneL.h deleted file mode 100644 index ad32fed5e6b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_pruneL.h +++ /dev/null @@ -1,136 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* - - * NOTE: This file is the modified version of [s,d,c,z]pruneL.c file in SuperLU - - * -- SuperLU routine (version 2.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * November 15, 1997 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ -#ifndef SPARSELU_PRUNEL_H -#define SPARSELU_PRUNEL_H - -namespace Eigen { -namespace internal { - -/** - * \brief Prunes the L-structure. - * - * It prunes the L-structure of supernodes whose L-structure contains the current pivot row "pivrow" - * - * - * \param jcol The current column of L - * \param[in] perm_r Row permutation - * \param[out] pivrow The pivot row - * \param nseg Number of segments - * \param segrep - * \param repfnz - * \param[out] xprune - * \param glu Global LU data - * - */ -template -void SparseLUImpl::pruneL(const Index jcol, const IndexVector& perm_r, const Index pivrow, const Index nseg, - const IndexVector& segrep, BlockIndexVector repfnz, IndexVector& xprune, GlobalLU_t& glu) -{ - // For each supernode-rep irep in U(*,j] - Index jsupno = glu.supno(jcol); - Index i,irep,irep1; - bool movnum, do_prune = false; - Index kmin = 0, kmax = 0, minloc, maxloc,krow; - for (i = 0; i < nseg; i++) - { - irep = segrep(i); - irep1 = irep + 1; - do_prune = false; - - // Don't prune with a zero U-segment - if (repfnz(irep) == emptyIdxLU) continue; - - // If a snode overlaps with the next panel, then the U-segment - // is fragmented into two parts -- irep and irep1. We should let - // pruning occur at the rep-column in irep1s snode. - if (glu.supno(irep) == glu.supno(irep1) ) continue; // don't prune - - // If it has not been pruned & it has a nonz in row L(pivrow,i) - if (glu.supno(irep) != jsupno ) - { - if ( xprune (irep) >= glu.xlsub(irep1) ) - { - kmin = glu.xlsub(irep); - kmax = glu.xlsub(irep1) - 1; - for (krow = kmin; krow <= kmax; krow++) - { - if (glu.lsub(krow) == pivrow) - { - do_prune = true; - break; - } - } - } - - if (do_prune) - { - // do a quicksort-type partition - // movnum=true means that the num values have to be exchanged - movnum = false; - if (irep == glu.xsup(glu.supno(irep)) ) // Snode of size 1 - movnum = true; - - while (kmin <= kmax) - { - if (perm_r(glu.lsub(kmax)) == emptyIdxLU) - kmax--; - else if ( perm_r(glu.lsub(kmin)) != emptyIdxLU) - kmin++; - else - { - // kmin below pivrow (not yet pivoted), and kmax - // above pivrow: interchange the two suscripts - std::swap(glu.lsub(kmin), glu.lsub(kmax)); - - // If the supernode has only one column, then we - // only keep one set of subscripts. For any subscript - // intercnahge performed, similar interchange must be - // done on the numerical values. - if (movnum) - { - minloc = glu.xlusup(irep) + ( kmin - glu.xlsub(irep) ); - maxloc = glu.xlusup(irep) + ( kmax - glu.xlsub(irep) ); - std::swap(glu.lusup(minloc), glu.lusup(maxloc)); - } - kmin++; - kmax--; - } - } // end while - - xprune(irep) = StorageIndex(kmin); //Pruning - } // end if do_prune - } // end pruning - } // End for each U-segment -} - -} // end namespace internal -} // end namespace Eigen - -#endif // SPARSELU_PRUNEL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_relax_snode.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_relax_snode.h deleted file mode 100644 index c408d01b406..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseLU/SparseLU_relax_snode.h +++ /dev/null @@ -1,83 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012 Désiré Nuentsa-Wakam -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* This file is a modified version of heap_relax_snode.c file in SuperLU - * -- SuperLU routine (version 3.0) -- - * Univ. of California Berkeley, Xerox Palo Alto Research Center, - * and Lawrence Berkeley National Lab. - * October 15, 2003 - * - * Copyright (c) 1994 by Xerox Corporation. All rights reserved. - * - * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY - * EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. - * - * Permission is hereby granted to use or copy this program for any - * purpose, provided the above notices are retained on all copies. - * Permission to modify the code and to distribute modified code is - * granted, provided the above notices are retained, and a notice that - * the code was modified is included with the above copyright notice. - */ - -#ifndef SPARSELU_RELAX_SNODE_H -#define SPARSELU_RELAX_SNODE_H - -namespace Eigen { - -namespace internal { - -/** - * \brief Identify the initial relaxed supernodes - * - * This routine is applied to a column elimination tree. - * It assumes that the matrix has been reordered according to the postorder of the etree - * \param n the number of columns - * \param et elimination tree - * \param relax_columns Maximum number of columns allowed in a relaxed snode - * \param descendants Number of descendants of each node in the etree - * \param relax_end last column in a supernode - */ -template -void SparseLUImpl::relax_snode (const Index n, IndexVector& et, const Index relax_columns, IndexVector& descendants, IndexVector& relax_end) -{ - - // compute the number of descendants of each node in the etree - Index parent; - relax_end.setConstant(emptyIdxLU); - descendants.setZero(); - for (Index j = 0; j < n; j++) - { - parent = et(j); - if (parent != n) // not the dummy root - descendants(parent) += descendants(j) + 1; - } - // Identify the relaxed supernodes by postorder traversal of the etree - Index snode_start; // beginning of a snode - for (Index j = 0; j < n; ) - { - parent = et(j); - snode_start = j; - while ( parent != n && descendants(parent) < relax_columns ) - { - j = parent; - parent = et(j); - } - // Found a supernode in postordered etree, j is the last column - relax_end(snode_start) = StorageIndex(j); // Record last column - j++; - // Search for a new leaf - while (descendants(j) != 0 && j < n) j++; - } // End postorder traversal of the etree - -} - -} // end namespace internal - -} // end namespace Eigen -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseQR/SparseQR.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseQR/SparseQR.h deleted file mode 100644 index 7409fcae940..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SparseQR/SparseQR.h +++ /dev/null @@ -1,745 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2012-2013 Desire Nuentsa -// Copyright (C) 2012-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SPARSE_QR_H -#define EIGEN_SPARSE_QR_H - -namespace Eigen { - -template class SparseQR; -template struct SparseQRMatrixQReturnType; -template struct SparseQRMatrixQTransposeReturnType; -template struct SparseQR_QProduct; -namespace internal { - template struct traits > - { - typedef typename SparseQRType::MatrixType ReturnType; - typedef typename ReturnType::StorageIndex StorageIndex; - typedef typename ReturnType::StorageKind StorageKind; - enum { - RowsAtCompileTime = Dynamic, - ColsAtCompileTime = Dynamic - }; - }; - template struct traits > - { - typedef typename SparseQRType::MatrixType ReturnType; - }; - template struct traits > - { - typedef typename Derived::PlainObject ReturnType; - }; -} // End namespace internal - -/** - * \ingroup SparseQR_Module - * \class SparseQR - * \brief Sparse left-looking rank-revealing QR factorization - * - * This class implements a left-looking rank-revealing QR decomposition - * of sparse matrices. When a column has a norm less than a given tolerance - * it is implicitly permuted to the end. The QR factorization thus obtained is - * given by A*P = Q*R where R is upper triangular or trapezoidal. - * - * P is the column permutation which is the product of the fill-reducing and the - * rank-revealing permutations. Use colsPermutation() to get it. - * - * Q is the orthogonal matrix represented as products of Householder reflectors. - * Use matrixQ() to get an expression and matrixQ().adjoint() to get the adjoint. - * You can then apply it to a vector. - * - * R is the sparse triangular or trapezoidal matrix. The later occurs when A is rank-deficient. - * matrixR().topLeftCorner(rank(), rank()) always returns a triangular factor of full rank. - * - * \tparam _MatrixType The type of the sparse matrix A, must be a column-major SparseMatrix<> - * \tparam _OrderingType The fill-reducing ordering method. See the \link OrderingMethods_Module - * OrderingMethods \endlink module for the list of built-in and external ordering methods. - * - * \implsparsesolverconcept - * - * \warning The input sparse matrix A must be in compressed mode (see SparseMatrix::makeCompressed()). - * \warning For complex matrices matrixQ().transpose() will actually return the adjoint matrix. - * - */ -template -class SparseQR : public SparseSolverBase > -{ - protected: - typedef SparseSolverBase > Base; - using Base::m_isInitialized; - public: - using Base::_solve_impl; - typedef _MatrixType MatrixType; - typedef _OrderingType OrderingType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef SparseMatrix QRMatrixType; - typedef Matrix IndexVector; - typedef Matrix ScalarVector; - typedef PermutationMatrix PermutationType; - - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - SparseQR () : m_analysisIsok(false), m_lastError(""), m_useDefaultThreshold(true),m_isQSorted(false),m_isEtreeOk(false) - { } - - /** Construct a QR factorization of the matrix \a mat. - * - * \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()). - * - * \sa compute() - */ - explicit SparseQR(const MatrixType& mat) : m_analysisIsok(false), m_lastError(""), m_useDefaultThreshold(true),m_isQSorted(false),m_isEtreeOk(false) - { - compute(mat); - } - - /** Computes the QR factorization of the sparse matrix \a mat. - * - * \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()). - * - * \sa analyzePattern(), factorize() - */ - void compute(const MatrixType& mat) - { - analyzePattern(mat); - factorize(mat); - } - void analyzePattern(const MatrixType& mat); - void factorize(const MatrixType& mat); - - /** \returns the number of rows of the represented matrix. - */ - inline Index rows() const { return m_pmat.rows(); } - - /** \returns the number of columns of the represented matrix. - */ - inline Index cols() const { return m_pmat.cols();} - - /** \returns a const reference to the \b sparse upper triangular matrix R of the QR factorization. - * \warning The entries of the returned matrix are not sorted. This means that using it in algorithms - * expecting sorted entries will fail. This include random coefficient accesses (SpaseMatrix::coeff()), - * and coefficient-wise operations. Matrix products and triangular solves are fine though. - * - * To sort the entries, you can assign it to a row-major matrix, and if a column-major matrix - * is required, you can copy it again: - * \code - * SparseMatrix R = qr.matrixR(); // column-major, not sorted! - * SparseMatrix Rr = qr.matrixR(); // row-major, sorted - * SparseMatrix Rc = Rr; // column-major, sorted - * \endcode - */ - const QRMatrixType& matrixR() const { return m_R; } - - /** \returns the number of non linearly dependent columns as determined by the pivoting threshold. - * - * \sa setPivotThreshold() - */ - Index rank() const - { - eigen_assert(m_isInitialized && "The factorization should be called first, use compute()"); - return m_nonzeropivots; - } - - /** \returns an expression of the matrix Q as products of sparse Householder reflectors. - * The common usage of this function is to apply it to a dense matrix or vector - * \code - * VectorXd B1, B2; - * // Initialize B1 - * B2 = matrixQ() * B1; - * \endcode - * - * To get a plain SparseMatrix representation of Q: - * \code - * SparseMatrix Q; - * Q = SparseQR >(A).matrixQ(); - * \endcode - * Internally, this call simply performs a sparse product between the matrix Q - * and a sparse identity matrix. However, due to the fact that the sparse - * reflectors are stored unsorted, two transpositions are needed to sort - * them before performing the product. - */ - SparseQRMatrixQReturnType matrixQ() const - { return SparseQRMatrixQReturnType(*this); } - - /** \returns a const reference to the column permutation P that was applied to A such that A*P = Q*R - * It is the combination of the fill-in reducing permutation and numerical column pivoting. - */ - const PermutationType& colsPermutation() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_outputPerm_c; - } - - /** \returns A string describing the type of error. - * This method is provided to ease debugging, not to handle errors. - */ - std::string lastErrorMessage() const { return m_lastError; } - - /** \internal */ - template - bool _solve_impl(const MatrixBase &B, MatrixBase &dest) const - { - eigen_assert(m_isInitialized && "The factorization should be called first, use compute()"); - eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix"); - - Index rank = this->rank(); - - // Compute Q^* * b; - typename Dest::PlainObject y, b; - y = this->matrixQ().adjoint() * B; - b = y; - - // Solve with the triangular matrix R - y.resize((std::max)(cols(),y.rows()),y.cols()); - y.topRows(rank) = this->matrixR().topLeftCorner(rank, rank).template triangularView().solve(b.topRows(rank)); - y.bottomRows(y.rows()-rank).setZero(); - - // Apply the column permutation - if (m_perm_c.size()) dest = colsPermutation() * y.topRows(cols()); - else dest = y.topRows(cols()); - - m_info = Success; - return true; - } - - /** Sets the threshold that is used to determine linearly dependent columns during the factorization. - * - * In practice, if during the factorization the norm of the column that has to be eliminated is below - * this threshold, then the entire column is treated as zero, and it is moved at the end. - */ - void setPivotThreshold(const RealScalar& threshold) - { - m_useDefaultThreshold = false; - m_threshold = threshold; - } - - /** \returns the solution X of \f$ A X = B \f$ using the current decomposition of A. - * - * \sa compute() - */ - template - inline const Solve solve(const MatrixBase& B) const - { - eigen_assert(m_isInitialized && "The factorization should be called first, use compute()"); - eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix"); - return Solve(*this, B.derived()); - } - template - inline const Solve solve(const SparseMatrixBase& B) const - { - eigen_assert(m_isInitialized && "The factorization should be called first, use compute()"); - eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix"); - return Solve(*this, B.derived()); - } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was successful, - * \c NumericalIssue if the QR factorization reports a numerical problem - * \c InvalidInput if the input matrix is invalid - * - * \sa iparm() - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - - /** \internal */ - inline void _sort_matrix_Q() - { - if(this->m_isQSorted) return; - // The matrix Q is sorted during the transposition - SparseMatrix mQrm(this->m_Q); - this->m_Q = mQrm; - this->m_isQSorted = true; - } - - - protected: - bool m_analysisIsok; - bool m_factorizationIsok; - mutable ComputationInfo m_info; - std::string m_lastError; - QRMatrixType m_pmat; // Temporary matrix - QRMatrixType m_R; // The triangular factor matrix - QRMatrixType m_Q; // The orthogonal reflectors - ScalarVector m_hcoeffs; // The Householder coefficients - PermutationType m_perm_c; // Fill-reducing Column permutation - PermutationType m_pivotperm; // The permutation for rank revealing - PermutationType m_outputPerm_c; // The final column permutation - RealScalar m_threshold; // Threshold to determine null Householder reflections - bool m_useDefaultThreshold; // Use default threshold - Index m_nonzeropivots; // Number of non zero pivots found - IndexVector m_etree; // Column elimination tree - IndexVector m_firstRowElt; // First element in each row - bool m_isQSorted; // whether Q is sorted or not - bool m_isEtreeOk; // whether the elimination tree match the initial input matrix - - template friend struct SparseQR_QProduct; - -}; - -/** \brief Preprocessing step of a QR factorization - * - * \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()). - * - * In this step, the fill-reducing permutation is computed and applied to the columns of A - * and the column elimination tree is computed as well. Only the sparsity pattern of \a mat is exploited. - * - * \note In this step it is assumed that there is no empty row in the matrix \a mat. - */ -template -void SparseQR::analyzePattern(const MatrixType& mat) -{ - eigen_assert(mat.isCompressed() && "SparseQR requires a sparse matrix in compressed mode. Call .makeCompressed() before passing it to SparseQR"); - // Copy to a column major matrix if the input is rowmajor - typename internal::conditional::type matCpy(mat); - // Compute the column fill reducing ordering - OrderingType ord; - ord(matCpy, m_perm_c); - Index n = mat.cols(); - Index m = mat.rows(); - Index diagSize = (std::min)(m,n); - - if (!m_perm_c.size()) - { - m_perm_c.resize(n); - m_perm_c.indices().setLinSpaced(n, 0,StorageIndex(n-1)); - } - - // Compute the column elimination tree of the permuted matrix - m_outputPerm_c = m_perm_c.inverse(); - internal::coletree(matCpy, m_etree, m_firstRowElt, m_outputPerm_c.indices().data()); - m_isEtreeOk = true; - - m_R.resize(m, n); - m_Q.resize(m, diagSize); - - // Allocate space for nonzero elements : rough estimation - m_R.reserve(2*mat.nonZeros()); //FIXME Get a more accurate estimation through symbolic factorization with the etree - m_Q.reserve(2*mat.nonZeros()); - m_hcoeffs.resize(diagSize); - m_analysisIsok = true; -} - -/** \brief Performs the numerical QR factorization of the input matrix - * - * The function SparseQR::analyzePattern(const MatrixType&) must have been called beforehand with - * a matrix having the same sparsity pattern than \a mat. - * - * \param mat The sparse column-major matrix - */ -template -void SparseQR::factorize(const MatrixType& mat) -{ - using std::abs; - - eigen_assert(m_analysisIsok && "analyzePattern() should be called before this step"); - StorageIndex m = StorageIndex(mat.rows()); - StorageIndex n = StorageIndex(mat.cols()); - StorageIndex diagSize = (std::min)(m,n); - IndexVector mark((std::max)(m,n)); mark.setConstant(-1); // Record the visited nodes - IndexVector Ridx(n), Qidx(m); // Store temporarily the row indexes for the current column of R and Q - Index nzcolR, nzcolQ; // Number of nonzero for the current column of R and Q - ScalarVector tval(m); // The dense vector used to compute the current column - RealScalar pivotThreshold = m_threshold; - - m_R.setZero(); - m_Q.setZero(); - m_pmat = mat; - if(!m_isEtreeOk) - { - m_outputPerm_c = m_perm_c.inverse(); - internal::coletree(m_pmat, m_etree, m_firstRowElt, m_outputPerm_c.indices().data()); - m_isEtreeOk = true; - } - - m_pmat.uncompress(); // To have the innerNonZeroPtr allocated - - // Apply the fill-in reducing permutation lazily: - { - // If the input is row major, copy the original column indices, - // otherwise directly use the input matrix - // - IndexVector originalOuterIndicesCpy; - const StorageIndex *originalOuterIndices = mat.outerIndexPtr(); - if(MatrixType::IsRowMajor) - { - originalOuterIndicesCpy = IndexVector::Map(m_pmat.outerIndexPtr(),n+1); - originalOuterIndices = originalOuterIndicesCpy.data(); - } - - for (int i = 0; i < n; i++) - { - Index p = m_perm_c.size() ? m_perm_c.indices()(i) : i; - m_pmat.outerIndexPtr()[p] = originalOuterIndices[i]; - m_pmat.innerNonZeroPtr()[p] = originalOuterIndices[i+1] - originalOuterIndices[i]; - } - } - - /* Compute the default threshold as in MatLab, see: - * Tim Davis, "Algorithm 915, SuiteSparseQR: Multifrontal Multithreaded Rank-Revealing - * Sparse QR Factorization, ACM Trans. on Math. Soft. 38(1), 2011, Page 8:3 - */ - if(m_useDefaultThreshold) - { - RealScalar max2Norm = 0.0; - for (int j = 0; j < n; j++) max2Norm = numext::maxi(max2Norm, m_pmat.col(j).norm()); - if(max2Norm==RealScalar(0)) - max2Norm = RealScalar(1); - pivotThreshold = 20 * (m + n) * max2Norm * NumTraits::epsilon(); - } - - // Initialize the numerical permutation - m_pivotperm.setIdentity(n); - - StorageIndex nonzeroCol = 0; // Record the number of valid pivots - m_Q.startVec(0); - - // Left looking rank-revealing QR factorization: compute a column of R and Q at a time - for (StorageIndex col = 0; col < n; ++col) - { - mark.setConstant(-1); - m_R.startVec(col); - mark(nonzeroCol) = col; - Qidx(0) = nonzeroCol; - nzcolR = 0; nzcolQ = 1; - bool found_diag = nonzeroCol>=m; - tval.setZero(); - - // Symbolic factorization: find the nonzero locations of the column k of the factors R and Q, i.e., - // all the nodes (with indexes lower than rank) reachable through the column elimination tree (etree) rooted at node k. - // Note: if the diagonal entry does not exist, then its contribution must be explicitly added, - // thus the trick with found_diag that permits to do one more iteration on the diagonal element if this one has not been found. - for (typename QRMatrixType::InnerIterator itp(m_pmat, col); itp || !found_diag; ++itp) - { - StorageIndex curIdx = nonzeroCol; - if(itp) curIdx = StorageIndex(itp.row()); - if(curIdx == nonzeroCol) found_diag = true; - - // Get the nonzeros indexes of the current column of R - StorageIndex st = m_firstRowElt(curIdx); // The traversal of the etree starts here - if (st < 0 ) - { - m_lastError = "Empty row found during numerical factorization"; - m_info = InvalidInput; - return; - } - - // Traverse the etree - Index bi = nzcolR; - for (; mark(st) != col; st = m_etree(st)) - { - Ridx(nzcolR) = st; // Add this row to the list, - mark(st) = col; // and mark this row as visited - nzcolR++; - } - - // Reverse the list to get the topological ordering - Index nt = nzcolR-bi; - for(Index i = 0; i < nt/2; i++) std::swap(Ridx(bi+i), Ridx(nzcolR-i-1)); - - // Copy the current (curIdx,pcol) value of the input matrix - if(itp) tval(curIdx) = itp.value(); - else tval(curIdx) = Scalar(0); - - // Compute the pattern of Q(:,k) - if(curIdx > nonzeroCol && mark(curIdx) != col ) - { - Qidx(nzcolQ) = curIdx; // Add this row to the pattern of Q, - mark(curIdx) = col; // and mark it as visited - nzcolQ++; - } - } - - // Browse all the indexes of R(:,col) in reverse order - for (Index i = nzcolR-1; i >= 0; i--) - { - Index curIdx = Ridx(i); - - // Apply the curIdx-th householder vector to the current column (temporarily stored into tval) - Scalar tdot(0); - - // First compute q' * tval - tdot = m_Q.col(curIdx).dot(tval); - - tdot *= m_hcoeffs(curIdx); - - // Then update tval = tval - q * tau - // FIXME: tval -= tdot * m_Q.col(curIdx) should amount to the same (need to check/add support for efficient "dense ?= sparse") - for (typename QRMatrixType::InnerIterator itq(m_Q, curIdx); itq; ++itq) - tval(itq.row()) -= itq.value() * tdot; - - // Detect fill-in for the current column of Q - if(m_etree(Ridx(i)) == nonzeroCol) - { - for (typename QRMatrixType::InnerIterator itq(m_Q, curIdx); itq; ++itq) - { - StorageIndex iQ = StorageIndex(itq.row()); - if (mark(iQ) != col) - { - Qidx(nzcolQ++) = iQ; // Add this row to the pattern of Q, - mark(iQ) = col; // and mark it as visited - } - } - } - } // End update current column - - Scalar tau = RealScalar(0); - RealScalar beta = 0; - - if(nonzeroCol < diagSize) - { - // Compute the Householder reflection that eliminate the current column - // FIXME this step should call the Householder module. - Scalar c0 = nzcolQ ? tval(Qidx(0)) : Scalar(0); - - // First, the squared norm of Q((col+1):m, col) - RealScalar sqrNorm = 0.; - for (Index itq = 1; itq < nzcolQ; ++itq) sqrNorm += numext::abs2(tval(Qidx(itq))); - if(sqrNorm == RealScalar(0) && numext::imag(c0) == RealScalar(0)) - { - beta = numext::real(c0); - tval(Qidx(0)) = 1; - } - else - { - using std::sqrt; - beta = sqrt(numext::abs2(c0) + sqrNorm); - if(numext::real(c0) >= RealScalar(0)) - beta = -beta; - tval(Qidx(0)) = 1; - for (Index itq = 1; itq < nzcolQ; ++itq) - tval(Qidx(itq)) /= (c0 - beta); - tau = numext::conj((beta-c0) / beta); - - } - } - - // Insert values in R - for (Index i = nzcolR-1; i >= 0; i--) - { - Index curIdx = Ridx(i); - if(curIdx < nonzeroCol) - { - m_R.insertBackByOuterInnerUnordered(col, curIdx) = tval(curIdx); - tval(curIdx) = Scalar(0.); - } - } - - if(nonzeroCol < diagSize && abs(beta) >= pivotThreshold) - { - m_R.insertBackByOuterInner(col, nonzeroCol) = beta; - // The householder coefficient - m_hcoeffs(nonzeroCol) = tau; - // Record the householder reflections - for (Index itq = 0; itq < nzcolQ; ++itq) - { - Index iQ = Qidx(itq); - m_Q.insertBackByOuterInnerUnordered(nonzeroCol,iQ) = tval(iQ); - tval(iQ) = Scalar(0.); - } - nonzeroCol++; - if(nonzeroCol -struct SparseQR_QProduct : ReturnByValue > -{ - typedef typename SparseQRType::QRMatrixType MatrixType; - typedef typename SparseQRType::Scalar Scalar; - // Get the references - SparseQR_QProduct(const SparseQRType& qr, const Derived& other, bool transpose) : - m_qr(qr),m_other(other),m_transpose(transpose) {} - inline Index rows() const { return m_qr.matrixQ().rows(); } - inline Index cols() const { return m_other.cols(); } - - // Assign to a vector - template - void evalTo(DesType& res) const - { - Index m = m_qr.rows(); - Index n = m_qr.cols(); - Index diagSize = (std::min)(m,n); - res = m_other; - if (m_transpose) - { - eigen_assert(m_qr.m_Q.rows() == m_other.rows() && "Non conforming object sizes"); - //Compute res = Q' * other column by column - for(Index j = 0; j < res.cols(); j++){ - for (Index k = 0; k < diagSize; k++) - { - Scalar tau = Scalar(0); - tau = m_qr.m_Q.col(k).dot(res.col(j)); - if(tau==Scalar(0)) continue; - tau = tau * m_qr.m_hcoeffs(k); - res.col(j) -= tau * m_qr.m_Q.col(k); - } - } - } - else - { - eigen_assert(m_qr.matrixQ().cols() == m_other.rows() && "Non conforming object sizes"); - - res.conservativeResize(rows(), cols()); - - // Compute res = Q * other column by column - for(Index j = 0; j < res.cols(); j++) - { - for (Index k = diagSize-1; k >=0; k--) - { - Scalar tau = Scalar(0); - tau = m_qr.m_Q.col(k).dot(res.col(j)); - if(tau==Scalar(0)) continue; - tau = tau * numext::conj(m_qr.m_hcoeffs(k)); - res.col(j) -= tau * m_qr.m_Q.col(k); - } - } - } - } - - const SparseQRType& m_qr; - const Derived& m_other; - bool m_transpose; // TODO this actually means adjoint -}; - -template -struct SparseQRMatrixQReturnType : public EigenBase > -{ - typedef typename SparseQRType::Scalar Scalar; - typedef Matrix DenseMatrix; - enum { - RowsAtCompileTime = Dynamic, - ColsAtCompileTime = Dynamic - }; - explicit SparseQRMatrixQReturnType(const SparseQRType& qr) : m_qr(qr) {} - template - SparseQR_QProduct operator*(const MatrixBase& other) - { - return SparseQR_QProduct(m_qr,other.derived(),false); - } - // To use for operations with the adjoint of Q - SparseQRMatrixQTransposeReturnType adjoint() const - { - return SparseQRMatrixQTransposeReturnType(m_qr); - } - inline Index rows() const { return m_qr.rows(); } - inline Index cols() const { return m_qr.rows(); } - // To use for operations with the transpose of Q FIXME this is the same as adjoint at the moment - SparseQRMatrixQTransposeReturnType transpose() const - { - return SparseQRMatrixQTransposeReturnType(m_qr); - } - const SparseQRType& m_qr; -}; - -// TODO this actually represents the adjoint of Q -template -struct SparseQRMatrixQTransposeReturnType -{ - explicit SparseQRMatrixQTransposeReturnType(const SparseQRType& qr) : m_qr(qr) {} - template - SparseQR_QProduct operator*(const MatrixBase& other) - { - return SparseQR_QProduct(m_qr,other.derived(), true); - } - const SparseQRType& m_qr; -}; - -namespace internal { - -template -struct evaluator_traits > -{ - typedef typename SparseQRType::MatrixType MatrixType; - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef SparseShape Shape; -}; - -template< typename DstXprType, typename SparseQRType> -struct Assignment, internal::assign_op, Sparse2Sparse> -{ - typedef SparseQRMatrixQReturnType SrcXprType; - typedef typename DstXprType::Scalar Scalar; - typedef typename DstXprType::StorageIndex StorageIndex; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) - { - typename DstXprType::PlainObject idMat(src.rows(), src.cols()); - idMat.setIdentity(); - // Sort the sparse householder reflectors if needed - const_cast(&src.m_qr)->_sort_matrix_Q(); - dst = SparseQR_QProduct(src.m_qr, idMat, false); - } -}; - -template< typename DstXprType, typename SparseQRType> -struct Assignment, internal::assign_op, Sparse2Dense> -{ - typedef SparseQRMatrixQReturnType SrcXprType; - typedef typename DstXprType::Scalar Scalar; - typedef typename DstXprType::StorageIndex StorageIndex; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) - { - dst = src.m_qr.matrixQ() * DstXprType::Identity(src.m_qr.rows(), src.m_qr.rows()); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdDeque.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdDeque.h deleted file mode 100644 index af158f425d4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdDeque.h +++ /dev/null @@ -1,128 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2009 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STDDEQUE_H -#define EIGEN_STDDEQUE_H - -#include "details.h" - -/** - * This section contains a convenience MACRO which allows an easy specialization of - * std::deque such that for data types with alignment issues the correct allocator - * is used automatically. - */ -#define EIGEN_DEFINE_STL_DEQUE_SPECIALIZATION(...) \ -namespace std \ -{ \ - template<> \ - class deque<__VA_ARGS__, std::allocator<__VA_ARGS__> > \ - : public deque<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > \ - { \ - typedef deque<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > deque_base; \ - public: \ - typedef __VA_ARGS__ value_type; \ - typedef deque_base::allocator_type allocator_type; \ - typedef deque_base::size_type size_type; \ - typedef deque_base::iterator iterator; \ - explicit deque(const allocator_type& a = allocator_type()) : deque_base(a) {} \ - template \ - deque(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) : deque_base(first, last, a) {} \ - deque(const deque& c) : deque_base(c) {} \ - explicit deque(size_type num, const value_type& val = value_type()) : deque_base(num, val) {} \ - deque(iterator start, iterator end) : deque_base(start, end) {} \ - deque& operator=(const deque& x) { \ - deque_base::operator=(x); \ - return *this; \ - } \ - }; \ -} - -// check whether we really need the std::deque specialization -#if !EIGEN_HAS_CXX11_CONTAINERS && !(defined(_GLIBCXX_DEQUE) && (!EIGEN_GNUC_AT_LEAST(4,1))) /* Note that before gcc-4.1 we already have: std::deque::resize(size_type,const T&). */ - -namespace std { - -#define EIGEN_STD_DEQUE_SPECIALIZATION_BODY \ - public: \ - typedef T value_type; \ - typedef typename deque_base::allocator_type allocator_type; \ - typedef typename deque_base::size_type size_type; \ - typedef typename deque_base::iterator iterator; \ - typedef typename deque_base::const_iterator const_iterator; \ - explicit deque(const allocator_type& a = allocator_type()) : deque_base(a) {} \ - template \ - deque(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) \ - : deque_base(first, last, a) {} \ - deque(const deque& c) : deque_base(c) {} \ - explicit deque(size_type num, const value_type& val = value_type()) : deque_base(num, val) {} \ - deque(iterator start, iterator end) : deque_base(start, end) {} \ - deque& operator=(const deque& x) { \ - deque_base::operator=(x); \ - return *this; \ - } - - template - class deque > - : public deque > -{ - typedef deque > deque_base; - EIGEN_STD_DEQUE_SPECIALIZATION_BODY - - void resize(size_type new_size) - { resize(new_size, T()); } - -#if defined(_DEQUE_) - // workaround MSVC std::deque implementation - void resize(size_type new_size, const value_type& x) - { - if (deque_base::size() < new_size) - deque_base::_Insert_n(deque_base::end(), new_size - deque_base::size(), x); - else if (new_size < deque_base::size()) - deque_base::erase(deque_base::begin() + new_size, deque_base::end()); - } - void push_back(const value_type& x) - { deque_base::push_back(x); } - void push_front(const value_type& x) - { deque_base::push_front(x); } - using deque_base::insert; - iterator insert(const_iterator position, const value_type& x) - { return deque_base::insert(position,x); } - void insert(const_iterator position, size_type new_size, const value_type& x) - { deque_base::insert(position, new_size, x); } -#elif defined(_GLIBCXX_DEQUE) && EIGEN_GNUC_AT_LEAST(4,2) && !EIGEN_GNUC_AT_LEAST(10, 1) - // workaround GCC std::deque implementation - // GCC 10.1 doesn't let us access _Deque_impl _M_impl anymore and we have to - // fall-back to the default case - void resize(size_type new_size, const value_type& x) - { - if (new_size < deque_base::size()) - deque_base::_M_erase_at_end(this->_M_impl._M_start + new_size); - else - deque_base::insert(deque_base::end(), new_size - deque_base::size(), x); - } -#else - // either non-GCC or GCC between 4.1 and 10.1 - // default implementation which should always work. - void resize(size_type new_size, const value_type& x) - { - if (new_size < deque_base::size()) - deque_base::erase(deque_base::begin() + new_size, deque_base::end()); - else if (new_size > deque_base::size()) - deque_base::insert(deque_base::end(), new_size - deque_base::size(), x); - } -#endif - }; -} - -#endif // check whether specialization is actually required - -#endif // EIGEN_STDDEQUE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdList.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdList.h deleted file mode 100644 index e1eba498596..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdList.h +++ /dev/null @@ -1,106 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STDLIST_H -#define EIGEN_STDLIST_H - -#include "details.h" - -/** - * This section contains a convenience MACRO which allows an easy specialization of - * std::list such that for data types with alignment issues the correct allocator - * is used automatically. - */ -#define EIGEN_DEFINE_STL_LIST_SPECIALIZATION(...) \ -namespace std \ -{ \ - template<> \ - class list<__VA_ARGS__, std::allocator<__VA_ARGS__> > \ - : public list<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > \ - { \ - typedef list<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > list_base; \ - public: \ - typedef __VA_ARGS__ value_type; \ - typedef list_base::allocator_type allocator_type; \ - typedef list_base::size_type size_type; \ - typedef list_base::iterator iterator; \ - explicit list(const allocator_type& a = allocator_type()) : list_base(a) {} \ - template \ - list(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) : list_base(first, last, a) {} \ - list(const list& c) : list_base(c) {} \ - explicit list(size_type num, const value_type& val = value_type()) : list_base(num, val) {} \ - list(iterator start, iterator end) : list_base(start, end) {} \ - list& operator=(const list& x) { \ - list_base::operator=(x); \ - return *this; \ - } \ - }; \ -} - -// check whether we really need the std::list specialization -#if !EIGEN_HAS_CXX11_CONTAINERS && !(defined(_GLIBCXX_LIST) && (!EIGEN_GNUC_AT_LEAST(4,1))) /* Note that before gcc-4.1 we already have: std::list::resize(size_type,const T&). */ - -namespace std -{ - -#define EIGEN_STD_LIST_SPECIALIZATION_BODY \ - public: \ - typedef T value_type; \ - typedef typename list_base::allocator_type allocator_type; \ - typedef typename list_base::size_type size_type; \ - typedef typename list_base::iterator iterator; \ - typedef typename list_base::const_iterator const_iterator; \ - explicit list(const allocator_type& a = allocator_type()) : list_base(a) {} \ - template \ - list(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) \ - : list_base(first, last, a) {} \ - list(const list& c) : list_base(c) {} \ - explicit list(size_type num, const value_type& val = value_type()) : list_base(num, val) {} \ - list(iterator start, iterator end) : list_base(start, end) {} \ - list& operator=(const list& x) { \ - list_base::operator=(x); \ - return *this; \ - } - - template - class list > - : public list > - { - typedef list > list_base; - EIGEN_STD_LIST_SPECIALIZATION_BODY - - void resize(size_type new_size) - { resize(new_size, T()); } - - void resize(size_type new_size, const value_type& x) - { - if (list_base::size() < new_size) - list_base::insert(list_base::end(), new_size - list_base::size(), x); - else - while (new_size < list_base::size()) list_base::pop_back(); - } - -#if defined(_LIST_) - // workaround MSVC std::list implementation - void push_back(const value_type& x) - { list_base::push_back(x); } - using list_base::insert; - iterator insert(const_iterator position, const value_type& x) - { return list_base::insert(position,x); } - void insert(const_iterator position, size_type new_size, const value_type& x) - { list_base::insert(position, new_size, x); } -#endif - }; -} - -#endif // check whether specialization is actually required - -#endif // EIGEN_STDLIST_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdVector.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdVector.h deleted file mode 100644 index ec22821d261..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/StdVector.h +++ /dev/null @@ -1,131 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2009 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STDVECTOR_H -#define EIGEN_STDVECTOR_H - -#include "details.h" - -/** - * This section contains a convenience MACRO which allows an easy specialization of - * std::vector such that for data types with alignment issues the correct allocator - * is used automatically. - */ -#define EIGEN_DEFINE_STL_VECTOR_SPECIALIZATION(...) \ -namespace std \ -{ \ - template<> \ - class vector<__VA_ARGS__, std::allocator<__VA_ARGS__> > \ - : public vector<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > \ - { \ - typedef vector<__VA_ARGS__, EIGEN_ALIGNED_ALLOCATOR<__VA_ARGS__> > vector_base; \ - public: \ - typedef __VA_ARGS__ value_type; \ - typedef vector_base::allocator_type allocator_type; \ - typedef vector_base::size_type size_type; \ - typedef vector_base::iterator iterator; \ - explicit vector(const allocator_type& a = allocator_type()) : vector_base(a) {} \ - template \ - vector(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) : vector_base(first, last, a) {} \ - vector(const vector& c) : vector_base(c) {} \ - explicit vector(size_type num, const value_type& val = value_type()) : vector_base(num, val) {} \ - vector(iterator start, iterator end) : vector_base(start, end) {} \ - vector& operator=(const vector& x) { \ - vector_base::operator=(x); \ - return *this; \ - } \ - }; \ -} - -// Don't specialize if containers are implemented according to C++11 -#if !EIGEN_HAS_CXX11_CONTAINERS - -namespace std { - -#define EIGEN_STD_VECTOR_SPECIALIZATION_BODY \ - public: \ - typedef T value_type; \ - typedef typename vector_base::allocator_type allocator_type; \ - typedef typename vector_base::size_type size_type; \ - typedef typename vector_base::iterator iterator; \ - typedef typename vector_base::const_iterator const_iterator; \ - explicit vector(const allocator_type& a = allocator_type()) : vector_base(a) {} \ - template \ - vector(InputIterator first, InputIterator last, const allocator_type& a = allocator_type()) \ - : vector_base(first, last, a) {} \ - vector(const vector& c) : vector_base(c) {} \ - explicit vector(size_type num, const value_type& val = value_type()) : vector_base(num, val) {} \ - vector(iterator start, iterator end) : vector_base(start, end) {} \ - vector& operator=(const vector& x) { \ - vector_base::operator=(x); \ - return *this; \ - } - - template - class vector > - : public vector > -{ - typedef vector > vector_base; - EIGEN_STD_VECTOR_SPECIALIZATION_BODY - - void resize(size_type new_size) - { resize(new_size, T()); } - -#if defined(_VECTOR_) - // workaround MSVC std::vector implementation - void resize(size_type new_size, const value_type& x) - { - if (vector_base::size() < new_size) - vector_base::_Insert_n(vector_base::end(), new_size - vector_base::size(), x); - else if (new_size < vector_base::size()) - vector_base::erase(vector_base::begin() + new_size, vector_base::end()); - } - void push_back(const value_type& x) - { vector_base::push_back(x); } - using vector_base::insert; - iterator insert(const_iterator position, const value_type& x) - { return vector_base::insert(position,x); } - void insert(const_iterator position, size_type new_size, const value_type& x) - { vector_base::insert(position, new_size, x); } -#elif defined(_GLIBCXX_VECTOR) && (!(EIGEN_GNUC_AT_LEAST(4,1))) - /* Note that before gcc-4.1 we already have: std::vector::resize(size_type,const T&). - * However, this specialization is still needed to make the above EIGEN_DEFINE_STL_VECTOR_SPECIALIZATION trick to work. */ - void resize(size_type new_size, const value_type& x) - { - vector_base::resize(new_size,x); - } -#elif defined(_GLIBCXX_VECTOR) && EIGEN_GNUC_AT_LEAST(4,2) - // workaround GCC std::vector implementation - void resize(size_type new_size, const value_type& x) - { - if (new_size < vector_base::size()) - vector_base::_M_erase_at_end(this->_M_impl._M_start + new_size); - else - vector_base::insert(vector_base::end(), new_size - vector_base::size(), x); - } -#else - // either GCC 4.1 or non-GCC - // default implementation which should always work. - void resize(size_type new_size, const value_type& x) - { - if (new_size < vector_base::size()) - vector_base::erase(vector_base::begin() + new_size, vector_base::end()); - else if (new_size > vector_base::size()) - vector_base::insert(vector_base::end(), new_size - vector_base::size(), x); - } -#endif - }; -} -#endif // !EIGEN_HAS_CXX11_CONTAINERS - - -#endif // EIGEN_STDVECTOR_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/details.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/details.h deleted file mode 100644 index 2cfd13e03aa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/StlSupport/details.h +++ /dev/null @@ -1,84 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2009 Hauke Heibel -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STL_DETAILS_H -#define EIGEN_STL_DETAILS_H - -#ifndef EIGEN_ALIGNED_ALLOCATOR - #define EIGEN_ALIGNED_ALLOCATOR Eigen::aligned_allocator -#endif - -namespace Eigen { - - // This one is needed to prevent reimplementing the whole std::vector. - template - class aligned_allocator_indirection : public EIGEN_ALIGNED_ALLOCATOR - { - public: - typedef std::size_t size_type; - typedef std::ptrdiff_t difference_type; - typedef T* pointer; - typedef const T* const_pointer; - typedef T& reference; - typedef const T& const_reference; - typedef T value_type; - - template - struct rebind - { - typedef aligned_allocator_indirection other; - }; - - aligned_allocator_indirection() {} - aligned_allocator_indirection(const aligned_allocator_indirection& ) : EIGEN_ALIGNED_ALLOCATOR() {} - aligned_allocator_indirection(const EIGEN_ALIGNED_ALLOCATOR& ) {} - template - aligned_allocator_indirection(const aligned_allocator_indirection& ) {} - template - aligned_allocator_indirection(const EIGEN_ALIGNED_ALLOCATOR& ) {} - ~aligned_allocator_indirection() {} - }; - -#if EIGEN_COMP_MSVC - - // sometimes, MSVC detects, at compile time, that the argument x - // in std::vector::resize(size_t s,T x) won't be aligned and generate an error - // even if this function is never called. Whence this little wrapper. -#define EIGEN_WORKAROUND_MSVC_STL_SUPPORT(T) \ - typename Eigen::internal::conditional< \ - Eigen::internal::is_arithmetic::value, \ - T, \ - Eigen::internal::workaround_msvc_stl_support \ - >::type - - namespace internal { - template struct workaround_msvc_stl_support : public T - { - inline workaround_msvc_stl_support() : T() {} - inline workaround_msvc_stl_support(const T& other) : T(other) {} - inline operator T& () { return *static_cast(this); } - inline operator const T& () const { return *static_cast(this); } - template - inline T& operator=(const OtherT& other) - { T::operator=(other); return *this; } - inline workaround_msvc_stl_support& operator=(const workaround_msvc_stl_support& other) - { T::operator=(other); return *this; } - }; - } - -#else - -#define EIGEN_WORKAROUND_MSVC_STL_SUPPORT(T) T - -#endif - -} - -#endif // EIGEN_STL_DETAILS_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SuperLUSupport/SuperLUSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SuperLUSupport/SuperLUSupport.h deleted file mode 100644 index 7261c7d0fc8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/SuperLUSupport/SuperLUSupport.h +++ /dev/null @@ -1,1027 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SUPERLUSUPPORT_H -#define EIGEN_SUPERLUSUPPORT_H - -namespace Eigen { - -#if defined(SUPERLU_MAJOR_VERSION) && (SUPERLU_MAJOR_VERSION >= 5) -#define DECL_GSSVX(PREFIX,FLOATTYPE,KEYTYPE) \ - extern "C" { \ - extern void PREFIX##gssvx(superlu_options_t *, SuperMatrix *, int *, int *, int *, \ - char *, FLOATTYPE *, FLOATTYPE *, SuperMatrix *, SuperMatrix *, \ - void *, int, SuperMatrix *, SuperMatrix *, \ - FLOATTYPE *, FLOATTYPE *, FLOATTYPE *, FLOATTYPE *, \ - GlobalLU_t *, mem_usage_t *, SuperLUStat_t *, int *); \ - } \ - inline float SuperLU_gssvx(superlu_options_t *options, SuperMatrix *A, \ - int *perm_c, int *perm_r, int *etree, char *equed, \ - FLOATTYPE *R, FLOATTYPE *C, SuperMatrix *L, \ - SuperMatrix *U, void *work, int lwork, \ - SuperMatrix *B, SuperMatrix *X, \ - FLOATTYPE *recip_pivot_growth, \ - FLOATTYPE *rcond, FLOATTYPE *ferr, FLOATTYPE *berr, \ - SuperLUStat_t *stats, int *info, KEYTYPE) { \ - mem_usage_t mem_usage; \ - GlobalLU_t gLU; \ - PREFIX##gssvx(options, A, perm_c, perm_r, etree, equed, R, C, L, \ - U, work, lwork, B, X, recip_pivot_growth, rcond, \ - ferr, berr, &gLU, &mem_usage, stats, info); \ - return mem_usage.for_lu; /* bytes used by the factor storage */ \ - } -#else // version < 5.0 -#define DECL_GSSVX(PREFIX,FLOATTYPE,KEYTYPE) \ - extern "C" { \ - extern void PREFIX##gssvx(superlu_options_t *, SuperMatrix *, int *, int *, int *, \ - char *, FLOATTYPE *, FLOATTYPE *, SuperMatrix *, SuperMatrix *, \ - void *, int, SuperMatrix *, SuperMatrix *, \ - FLOATTYPE *, FLOATTYPE *, FLOATTYPE *, FLOATTYPE *, \ - mem_usage_t *, SuperLUStat_t *, int *); \ - } \ - inline float SuperLU_gssvx(superlu_options_t *options, SuperMatrix *A, \ - int *perm_c, int *perm_r, int *etree, char *equed, \ - FLOATTYPE *R, FLOATTYPE *C, SuperMatrix *L, \ - SuperMatrix *U, void *work, int lwork, \ - SuperMatrix *B, SuperMatrix *X, \ - FLOATTYPE *recip_pivot_growth, \ - FLOATTYPE *rcond, FLOATTYPE *ferr, FLOATTYPE *berr, \ - SuperLUStat_t *stats, int *info, KEYTYPE) { \ - mem_usage_t mem_usage; \ - PREFIX##gssvx(options, A, perm_c, perm_r, etree, equed, R, C, L, \ - U, work, lwork, B, X, recip_pivot_growth, rcond, \ - ferr, berr, &mem_usage, stats, info); \ - return mem_usage.for_lu; /* bytes used by the factor storage */ \ - } -#endif - -DECL_GSSVX(s,float,float) -DECL_GSSVX(c,float,std::complex) -DECL_GSSVX(d,double,double) -DECL_GSSVX(z,double,std::complex) - -#ifdef MILU_ALPHA -#define EIGEN_SUPERLU_HAS_ILU -#endif - -#ifdef EIGEN_SUPERLU_HAS_ILU - -// similarly for the incomplete factorization using gsisx -#define DECL_GSISX(PREFIX,FLOATTYPE,KEYTYPE) \ - extern "C" { \ - extern void PREFIX##gsisx(superlu_options_t *, SuperMatrix *, int *, int *, int *, \ - char *, FLOATTYPE *, FLOATTYPE *, SuperMatrix *, SuperMatrix *, \ - void *, int, SuperMatrix *, SuperMatrix *, FLOATTYPE *, FLOATTYPE *, \ - mem_usage_t *, SuperLUStat_t *, int *); \ - } \ - inline float SuperLU_gsisx(superlu_options_t *options, SuperMatrix *A, \ - int *perm_c, int *perm_r, int *etree, char *equed, \ - FLOATTYPE *R, FLOATTYPE *C, SuperMatrix *L, \ - SuperMatrix *U, void *work, int lwork, \ - SuperMatrix *B, SuperMatrix *X, \ - FLOATTYPE *recip_pivot_growth, \ - FLOATTYPE *rcond, \ - SuperLUStat_t *stats, int *info, KEYTYPE) { \ - mem_usage_t mem_usage; \ - PREFIX##gsisx(options, A, perm_c, perm_r, etree, equed, R, C, L, \ - U, work, lwork, B, X, recip_pivot_growth, rcond, \ - &mem_usage, stats, info); \ - return mem_usage.for_lu; /* bytes used by the factor storage */ \ - } - -DECL_GSISX(s,float,float) -DECL_GSISX(c,float,std::complex) -DECL_GSISX(d,double,double) -DECL_GSISX(z,double,std::complex) - -#endif - -template -struct SluMatrixMapHelper; - -/** \internal - * - * A wrapper class for SuperLU matrices. It supports only compressed sparse matrices - * and dense matrices. Supernodal and other fancy format are not supported by this wrapper. - * - * This wrapper class mainly aims to avoids the need of dynamic allocation of the storage structure. - */ -struct SluMatrix : SuperMatrix -{ - SluMatrix() - { - Store = &storage; - } - - SluMatrix(const SluMatrix& other) - : SuperMatrix(other) - { - Store = &storage; - storage = other.storage; - } - - SluMatrix& operator=(const SluMatrix& other) - { - SuperMatrix::operator=(static_cast(other)); - Store = &storage; - storage = other.storage; - return *this; - } - - struct - { - union {int nnz;int lda;}; - void *values; - int *innerInd; - int *outerInd; - } storage; - - void setStorageType(Stype_t t) - { - Stype = t; - if (t==SLU_NC || t==SLU_NR || t==SLU_DN) - Store = &storage; - else - { - eigen_assert(false && "storage type not supported"); - Store = 0; - } - } - - template - void setScalarType() - { - if (internal::is_same::value) - Dtype = SLU_S; - else if (internal::is_same::value) - Dtype = SLU_D; - else if (internal::is_same >::value) - Dtype = SLU_C; - else if (internal::is_same >::value) - Dtype = SLU_Z; - else - { - eigen_assert(false && "Scalar type not supported by SuperLU"); - } - } - - template - static SluMatrix Map(MatrixBase& _mat) - { - MatrixType& mat(_mat.derived()); - eigen_assert( ((MatrixType::Flags&RowMajorBit)!=RowMajorBit) && "row-major dense matrices are not supported by SuperLU"); - SluMatrix res; - res.setStorageType(SLU_DN); - res.setScalarType(); - res.Mtype = SLU_GE; - - res.nrow = internal::convert_index(mat.rows()); - res.ncol = internal::convert_index(mat.cols()); - - res.storage.lda = internal::convert_index(MatrixType::IsVectorAtCompileTime ? mat.size() : mat.outerStride()); - res.storage.values = (void*)(mat.data()); - return res; - } - - template - static SluMatrix Map(SparseMatrixBase& a_mat) - { - MatrixType &mat(a_mat.derived()); - SluMatrix res; - if ((MatrixType::Flags&RowMajorBit)==RowMajorBit) - { - res.setStorageType(SLU_NR); - res.nrow = internal::convert_index(mat.cols()); - res.ncol = internal::convert_index(mat.rows()); - } - else - { - res.setStorageType(SLU_NC); - res.nrow = internal::convert_index(mat.rows()); - res.ncol = internal::convert_index(mat.cols()); - } - - res.Mtype = SLU_GE; - - res.storage.nnz = internal::convert_index(mat.nonZeros()); - res.storage.values = mat.valuePtr(); - res.storage.innerInd = mat.innerIndexPtr(); - res.storage.outerInd = mat.outerIndexPtr(); - - res.setScalarType(); - - // FIXME the following is not very accurate - if (MatrixType::Flags & Upper) - res.Mtype = SLU_TRU; - if (MatrixType::Flags & Lower) - res.Mtype = SLU_TRL; - - eigen_assert(((MatrixType::Flags & SelfAdjoint)==0) && "SelfAdjoint matrix shape not supported by SuperLU"); - - return res; - } -}; - -template -struct SluMatrixMapHelper > -{ - typedef Matrix MatrixType; - static void run(MatrixType& mat, SluMatrix& res) - { - eigen_assert( ((Options&RowMajor)!=RowMajor) && "row-major dense matrices is not supported by SuperLU"); - res.setStorageType(SLU_DN); - res.setScalarType(); - res.Mtype = SLU_GE; - - res.nrow = mat.rows(); - res.ncol = mat.cols(); - - res.storage.lda = mat.outerStride(); - res.storage.values = mat.data(); - } -}; - -template -struct SluMatrixMapHelper > -{ - typedef Derived MatrixType; - static void run(MatrixType& mat, SluMatrix& res) - { - if ((MatrixType::Flags&RowMajorBit)==RowMajorBit) - { - res.setStorageType(SLU_NR); - res.nrow = mat.cols(); - res.ncol = mat.rows(); - } - else - { - res.setStorageType(SLU_NC); - res.nrow = mat.rows(); - res.ncol = mat.cols(); - } - - res.Mtype = SLU_GE; - - res.storage.nnz = mat.nonZeros(); - res.storage.values = mat.valuePtr(); - res.storage.innerInd = mat.innerIndexPtr(); - res.storage.outerInd = mat.outerIndexPtr(); - - res.setScalarType(); - - // FIXME the following is not very accurate - if (MatrixType::Flags & Upper) - res.Mtype = SLU_TRU; - if (MatrixType::Flags & Lower) - res.Mtype = SLU_TRL; - - eigen_assert(((MatrixType::Flags & SelfAdjoint)==0) && "SelfAdjoint matrix shape not supported by SuperLU"); - } -}; - -namespace internal { - -template -SluMatrix asSluMatrix(MatrixType& mat) -{ - return SluMatrix::Map(mat); -} - -/** View a Super LU matrix as an Eigen expression */ -template -MappedSparseMatrix map_superlu(SluMatrix& sluMat) -{ - eigen_assert(((Flags&RowMajor)==RowMajor && sluMat.Stype == SLU_NR) - || ((Flags&ColMajor)==ColMajor && sluMat.Stype == SLU_NC)); - - Index outerSize = (Flags&RowMajor)==RowMajor ? sluMat.ncol : sluMat.nrow; - - return MappedSparseMatrix( - sluMat.nrow, sluMat.ncol, sluMat.storage.outerInd[outerSize], - sluMat.storage.outerInd, sluMat.storage.innerInd, reinterpret_cast(sluMat.storage.values) ); -} - -} // end namespace internal - -/** \ingroup SuperLUSupport_Module - * \class SuperLUBase - * \brief The base class for the direct and incomplete LU factorization of SuperLU - */ -template -class SuperLUBase : public SparseSolverBase -{ - protected: - typedef SparseSolverBase Base; - using Base::derived; - using Base::m_isInitialized; - public: - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Matrix Vector; - typedef Matrix IntRowVectorType; - typedef Matrix IntColVectorType; - typedef Map > PermutationMap; - typedef SparseMatrix LUMatrixType; - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - - SuperLUBase() {} - - ~SuperLUBase() - { - clearFactors(); - } - - inline Index rows() const { return m_matrix.rows(); } - inline Index cols() const { return m_matrix.cols(); } - - /** \returns a reference to the Super LU option object to configure the Super LU algorithms. */ - inline superlu_options_t& options() { return m_sluOptions; } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the matrix.appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - /** Computes the sparse Cholesky decomposition of \a matrix */ - void compute(const MatrixType& matrix) - { - derived().analyzePattern(matrix); - derived().factorize(matrix); - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& /*matrix*/) - { - m_isInitialized = true; - m_info = Success; - m_analysisIsOk = true; - m_factorizationIsOk = false; - } - - template - void dumpMemory(Stream& /*s*/) - {} - - protected: - - void initFactorization(const MatrixType& a) - { - set_default_options(&this->m_sluOptions); - - const Index size = a.rows(); - m_matrix = a; - - m_sluA = internal::asSluMatrix(m_matrix); - clearFactors(); - - m_p.resize(size); - m_q.resize(size); - m_sluRscale.resize(size); - m_sluCscale.resize(size); - m_sluEtree.resize(size); - - // set empty B and X - m_sluB.setStorageType(SLU_DN); - m_sluB.setScalarType(); - m_sluB.Mtype = SLU_GE; - m_sluB.storage.values = 0; - m_sluB.nrow = 0; - m_sluB.ncol = 0; - m_sluB.storage.lda = internal::convert_index(size); - m_sluX = m_sluB; - - m_extractedDataAreDirty = true; - } - - void init() - { - m_info = InvalidInput; - m_isInitialized = false; - m_sluL.Store = 0; - m_sluU.Store = 0; - } - - void extractData() const; - - void clearFactors() - { - if(m_sluL.Store) - Destroy_SuperNode_Matrix(&m_sluL); - if(m_sluU.Store) - Destroy_CompCol_Matrix(&m_sluU); - - m_sluL.Store = 0; - m_sluU.Store = 0; - - memset(&m_sluL,0,sizeof m_sluL); - memset(&m_sluU,0,sizeof m_sluU); - } - - // cached data to reduce reallocation, etc. - mutable LUMatrixType m_l; - mutable LUMatrixType m_u; - mutable IntColVectorType m_p; - mutable IntRowVectorType m_q; - - mutable LUMatrixType m_matrix; // copy of the factorized matrix - mutable SluMatrix m_sluA; - mutable SuperMatrix m_sluL, m_sluU; - mutable SluMatrix m_sluB, m_sluX; - mutable SuperLUStat_t m_sluStat; - mutable superlu_options_t m_sluOptions; - mutable std::vector m_sluEtree; - mutable Matrix m_sluRscale, m_sluCscale; - mutable Matrix m_sluFerr, m_sluBerr; - mutable char m_sluEqued; - - mutable ComputationInfo m_info; - int m_factorizationIsOk; - int m_analysisIsOk; - mutable bool m_extractedDataAreDirty; - - private: - SuperLUBase(SuperLUBase& ) { } -}; - - -/** \ingroup SuperLUSupport_Module - * \class SuperLU - * \brief A sparse direct LU factorization and solver based on the SuperLU library - * - * This class allows to solve for A.X = B sparse linear problems via a direct LU factorization - * using the SuperLU library. The sparse matrix A must be squared and invertible. The vectors or matrices - * X and B can be either dense or sparse. - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * - * \warning This class is only for the 4.x versions of SuperLU. The 3.x and 5.x versions are not supported. - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SparseLU - */ -template -class SuperLU : public SuperLUBase<_MatrixType,SuperLU<_MatrixType> > -{ - public: - typedef SuperLUBase<_MatrixType,SuperLU> Base; - typedef _MatrixType MatrixType; - typedef typename Base::Scalar Scalar; - typedef typename Base::RealScalar RealScalar; - typedef typename Base::StorageIndex StorageIndex; - typedef typename Base::IntRowVectorType IntRowVectorType; - typedef typename Base::IntColVectorType IntColVectorType; - typedef typename Base::PermutationMap PermutationMap; - typedef typename Base::LUMatrixType LUMatrixType; - typedef TriangularView LMatrixType; - typedef TriangularView UMatrixType; - - public: - using Base::_solve_impl; - - SuperLU() : Base() { init(); } - - explicit SuperLU(const MatrixType& matrix) : Base() - { - init(); - Base::compute(matrix); - } - - ~SuperLU() - { - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& matrix) - { - m_info = InvalidInput; - m_isInitialized = false; - Base::analyzePattern(matrix); - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - void factorize(const MatrixType& matrix); - - /** \internal */ - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const; - - inline const LMatrixType& matrixL() const - { - if (m_extractedDataAreDirty) this->extractData(); - return m_l; - } - - inline const UMatrixType& matrixU() const - { - if (m_extractedDataAreDirty) this->extractData(); - return m_u; - } - - inline const IntColVectorType& permutationP() const - { - if (m_extractedDataAreDirty) this->extractData(); - return m_p; - } - - inline const IntRowVectorType& permutationQ() const - { - if (m_extractedDataAreDirty) this->extractData(); - return m_q; - } - - Scalar determinant() const; - - protected: - - using Base::m_matrix; - using Base::m_sluOptions; - using Base::m_sluA; - using Base::m_sluB; - using Base::m_sluX; - using Base::m_p; - using Base::m_q; - using Base::m_sluEtree; - using Base::m_sluEqued; - using Base::m_sluRscale; - using Base::m_sluCscale; - using Base::m_sluL; - using Base::m_sluU; - using Base::m_sluStat; - using Base::m_sluFerr; - using Base::m_sluBerr; - using Base::m_l; - using Base::m_u; - - using Base::m_analysisIsOk; - using Base::m_factorizationIsOk; - using Base::m_extractedDataAreDirty; - using Base::m_isInitialized; - using Base::m_info; - - void init() - { - Base::init(); - - set_default_options(&this->m_sluOptions); - m_sluOptions.PrintStat = NO; - m_sluOptions.ConditionNumber = NO; - m_sluOptions.Trans = NOTRANS; - m_sluOptions.ColPerm = COLAMD; - } - - - private: - SuperLU(SuperLU& ) { } -}; - -template -void SuperLU::factorize(const MatrixType& a) -{ - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - if(!m_analysisIsOk) - { - m_info = InvalidInput; - return; - } - - this->initFactorization(a); - - m_sluOptions.ColPerm = COLAMD; - int info = 0; - RealScalar recip_pivot_growth, rcond; - RealScalar ferr, berr; - - StatInit(&m_sluStat); - SuperLU_gssvx(&m_sluOptions, &m_sluA, m_q.data(), m_p.data(), &m_sluEtree[0], - &m_sluEqued, &m_sluRscale[0], &m_sluCscale[0], - &m_sluL, &m_sluU, - NULL, 0, - &m_sluB, &m_sluX, - &recip_pivot_growth, &rcond, - &ferr, &berr, - &m_sluStat, &info, Scalar()); - StatFree(&m_sluStat); - - m_extractedDataAreDirty = true; - - // FIXME how to better check for errors ??? - m_info = info == 0 ? Success : NumericalIssue; - m_factorizationIsOk = true; -} - -template -template -void SuperLU::_solve_impl(const MatrixBase &b, MatrixBase& x) const -{ - eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or analyzePattern()/factorize()"); - - const Index size = m_matrix.rows(); - const Index rhsCols = b.cols(); - eigen_assert(size==b.rows()); - - m_sluOptions.Trans = NOTRANS; - m_sluOptions.Fact = FACTORED; - m_sluOptions.IterRefine = NOREFINE; - - - m_sluFerr.resize(rhsCols); - m_sluBerr.resize(rhsCols); - - Ref > b_ref(b); - Ref > x_ref(x); - - m_sluB = SluMatrix::Map(b_ref.const_cast_derived()); - m_sluX = SluMatrix::Map(x_ref.const_cast_derived()); - - typename Rhs::PlainObject b_cpy; - if(m_sluEqued!='N') - { - b_cpy = b; - m_sluB = SluMatrix::Map(b_cpy.const_cast_derived()); - } - - StatInit(&m_sluStat); - int info = 0; - RealScalar recip_pivot_growth, rcond; - SuperLU_gssvx(&m_sluOptions, &m_sluA, - m_q.data(), m_p.data(), - &m_sluEtree[0], &m_sluEqued, - &m_sluRscale[0], &m_sluCscale[0], - &m_sluL, &m_sluU, - NULL, 0, - &m_sluB, &m_sluX, - &recip_pivot_growth, &rcond, - &m_sluFerr[0], &m_sluBerr[0], - &m_sluStat, &info, Scalar()); - StatFree(&m_sluStat); - - if(x.derived().data() != x_ref.data()) - x = x_ref; - - m_info = info==0 ? Success : NumericalIssue; -} - -// the code of this extractData() function has been adapted from the SuperLU's Matlab support code, -// -// Copyright (c) 1994 by Xerox Corporation. All rights reserved. -// -// THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY -// EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. -// -template -void SuperLUBase::extractData() const -{ - eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for extracting factors, you must first call either compute() or analyzePattern()/factorize()"); - if (m_extractedDataAreDirty) - { - int upper; - int fsupc, istart, nsupr; - int lastl = 0, lastu = 0; - SCformat *Lstore = static_cast(m_sluL.Store); - NCformat *Ustore = static_cast(m_sluU.Store); - Scalar *SNptr; - - const Index size = m_matrix.rows(); - m_l.resize(size,size); - m_l.resizeNonZeros(Lstore->nnz); - m_u.resize(size,size); - m_u.resizeNonZeros(Ustore->nnz); - - int* Lcol = m_l.outerIndexPtr(); - int* Lrow = m_l.innerIndexPtr(); - Scalar* Lval = m_l.valuePtr(); - - int* Ucol = m_u.outerIndexPtr(); - int* Urow = m_u.innerIndexPtr(); - Scalar* Uval = m_u.valuePtr(); - - Ucol[0] = 0; - Ucol[0] = 0; - - /* for each supernode */ - for (int k = 0; k <= Lstore->nsuper; ++k) - { - fsupc = L_FST_SUPC(k); - istart = L_SUB_START(fsupc); - nsupr = L_SUB_START(fsupc+1) - istart; - upper = 1; - - /* for each column in the supernode */ - for (int j = fsupc; j < L_FST_SUPC(k+1); ++j) - { - SNptr = &((Scalar*)Lstore->nzval)[L_NZ_START(j)]; - - /* Extract U */ - for (int i = U_NZ_START(j); i < U_NZ_START(j+1); ++i) - { - Uval[lastu] = ((Scalar*)Ustore->nzval)[i]; - /* Matlab doesn't like explicit zero. */ - if (Uval[lastu] != 0.0) - Urow[lastu++] = U_SUB(i); - } - for (int i = 0; i < upper; ++i) - { - /* upper triangle in the supernode */ - Uval[lastu] = SNptr[i]; - /* Matlab doesn't like explicit zero. */ - if (Uval[lastu] != 0.0) - Urow[lastu++] = L_SUB(istart+i); - } - Ucol[j+1] = lastu; - - /* Extract L */ - Lval[lastl] = 1.0; /* unit diagonal */ - Lrow[lastl++] = L_SUB(istart + upper - 1); - for (int i = upper; i < nsupr; ++i) - { - Lval[lastl] = SNptr[i]; - /* Matlab doesn't like explicit zero. */ - if (Lval[lastl] != 0.0) - Lrow[lastl++] = L_SUB(istart+i); - } - Lcol[j+1] = lastl; - - ++upper; - } /* for j ... */ - - } /* for k ... */ - - // squeeze the matrices : - m_l.resizeNonZeros(lastl); - m_u.resizeNonZeros(lastu); - - m_extractedDataAreDirty = false; - } -} - -template -typename SuperLU::Scalar SuperLU::determinant() const -{ - eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for computing the determinant, you must first call either compute() or analyzePattern()/factorize()"); - - if (m_extractedDataAreDirty) - this->extractData(); - - Scalar det = Scalar(1); - for (int j=0; j 0) - { - int lastId = m_u.outerIndexPtr()[j+1]-1; - eigen_assert(m_u.innerIndexPtr()[lastId]<=j); - if (m_u.innerIndexPtr()[lastId]==j) - det *= m_u.valuePtr()[lastId]; - } - } - if(PermutationMap(m_p.data(),m_p.size()).determinant()*PermutationMap(m_q.data(),m_q.size()).determinant()<0) - det = -det; - if(m_sluEqued!='N') - return det/m_sluRscale.prod()/m_sluCscale.prod(); - else - return det; -} - -#ifdef EIGEN_PARSED_BY_DOXYGEN -#define EIGEN_SUPERLU_HAS_ILU -#endif - -#ifdef EIGEN_SUPERLU_HAS_ILU - -/** \ingroup SuperLUSupport_Module - * \class SuperILU - * \brief A sparse direct \b incomplete LU factorization and solver based on the SuperLU library - * - * This class allows to solve for an approximate solution of A.X = B sparse linear problems via an incomplete LU factorization - * using the SuperLU library. This class is aimed to be used as a preconditioner of the iterative linear solvers. - * - * \warning This class is only for the 4.x versions of SuperLU. The 3.x and 5.x versions are not supported. - * - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class IncompleteLUT, class ConjugateGradient, class BiCGSTAB - */ - -template -class SuperILU : public SuperLUBase<_MatrixType,SuperILU<_MatrixType> > -{ - public: - typedef SuperLUBase<_MatrixType,SuperILU> Base; - typedef _MatrixType MatrixType; - typedef typename Base::Scalar Scalar; - typedef typename Base::RealScalar RealScalar; - - public: - using Base::_solve_impl; - - SuperILU() : Base() { init(); } - - SuperILU(const MatrixType& matrix) : Base() - { - init(); - Base::compute(matrix); - } - - ~SuperILU() - { - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize() - */ - void analyzePattern(const MatrixType& matrix) - { - Base::analyzePattern(matrix); - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the symbolic decomposition has been performed. - * - * \sa analyzePattern() - */ - void factorize(const MatrixType& matrix); - - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** \internal */ - template - void _solve_impl(const MatrixBase &b, MatrixBase &dest) const; - #endif // EIGEN_PARSED_BY_DOXYGEN - - protected: - - using Base::m_matrix; - using Base::m_sluOptions; - using Base::m_sluA; - using Base::m_sluB; - using Base::m_sluX; - using Base::m_p; - using Base::m_q; - using Base::m_sluEtree; - using Base::m_sluEqued; - using Base::m_sluRscale; - using Base::m_sluCscale; - using Base::m_sluL; - using Base::m_sluU; - using Base::m_sluStat; - using Base::m_sluFerr; - using Base::m_sluBerr; - using Base::m_l; - using Base::m_u; - - using Base::m_analysisIsOk; - using Base::m_factorizationIsOk; - using Base::m_extractedDataAreDirty; - using Base::m_isInitialized; - using Base::m_info; - - void init() - { - Base::init(); - - ilu_set_default_options(&m_sluOptions); - m_sluOptions.PrintStat = NO; - m_sluOptions.ConditionNumber = NO; - m_sluOptions.Trans = NOTRANS; - m_sluOptions.ColPerm = MMD_AT_PLUS_A; - - // no attempt to preserve column sum - m_sluOptions.ILU_MILU = SILU; - // only basic ILU(k) support -- no direct control over memory consumption - // better to use ILU_DropRule = DROP_BASIC | DROP_AREA - // and set ILU_FillFactor to max memory growth - m_sluOptions.ILU_DropRule = DROP_BASIC; - m_sluOptions.ILU_DropTol = NumTraits::dummy_precision()*10; - } - - private: - SuperILU(SuperILU& ) { } -}; - -template -void SuperILU::factorize(const MatrixType& a) -{ - eigen_assert(m_analysisIsOk && "You must first call analyzePattern()"); - if(!m_analysisIsOk) - { - m_info = InvalidInput; - return; - } - - this->initFactorization(a); - - int info = 0; - RealScalar recip_pivot_growth, rcond; - - StatInit(&m_sluStat); - SuperLU_gsisx(&m_sluOptions, &m_sluA, m_q.data(), m_p.data(), &m_sluEtree[0], - &m_sluEqued, &m_sluRscale[0], &m_sluCscale[0], - &m_sluL, &m_sluU, - NULL, 0, - &m_sluB, &m_sluX, - &recip_pivot_growth, &rcond, - &m_sluStat, &info, Scalar()); - StatFree(&m_sluStat); - - // FIXME how to better check for errors ??? - m_info = info == 0 ? Success : NumericalIssue; - m_factorizationIsOk = true; -} - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void SuperILU::_solve_impl(const MatrixBase &b, MatrixBase& x) const -{ - eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or analyzePattern()/factorize()"); - - const int size = m_matrix.rows(); - const int rhsCols = b.cols(); - eigen_assert(size==b.rows()); - - m_sluOptions.Trans = NOTRANS; - m_sluOptions.Fact = FACTORED; - m_sluOptions.IterRefine = NOREFINE; - - m_sluFerr.resize(rhsCols); - m_sluBerr.resize(rhsCols); - - Ref > b_ref(b); - Ref > x_ref(x); - - m_sluB = SluMatrix::Map(b_ref.const_cast_derived()); - m_sluX = SluMatrix::Map(x_ref.const_cast_derived()); - - typename Rhs::PlainObject b_cpy; - if(m_sluEqued!='N') - { - b_cpy = b; - m_sluB = SluMatrix::Map(b_cpy.const_cast_derived()); - } - - int info = 0; - RealScalar recip_pivot_growth, rcond; - - StatInit(&m_sluStat); - SuperLU_gsisx(&m_sluOptions, &m_sluA, - m_q.data(), m_p.data(), - &m_sluEtree[0], &m_sluEqued, - &m_sluRscale[0], &m_sluCscale[0], - &m_sluL, &m_sluU, - NULL, 0, - &m_sluB, &m_sluX, - &recip_pivot_growth, &rcond, - &m_sluStat, &info, Scalar()); - StatFree(&m_sluStat); - - if(x.derived().data() != x_ref.data()) - x = x_ref; - - m_info = info==0 ? Success : NumericalIssue; -} -#endif - -#endif - -} // end namespace Eigen - -#endif // EIGEN_SUPERLUSUPPORT_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/UmfPackSupport/UmfPackSupport.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/UmfPackSupport/UmfPackSupport.h deleted file mode 100644 index 91c09ab1338..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/UmfPackSupport/UmfPackSupport.h +++ /dev/null @@ -1,506 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2011 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_UMFPACKSUPPORT_H -#define EIGEN_UMFPACKSUPPORT_H - -namespace Eigen { - -/* TODO extract L, extract U, compute det, etc... */ - -// generic double/complex wrapper functions: - - -inline void umfpack_defaults(double control[UMFPACK_CONTROL], double) -{ umfpack_di_defaults(control); } - -inline void umfpack_defaults(double control[UMFPACK_CONTROL], std::complex) -{ umfpack_zi_defaults(control); } - -inline void umfpack_report_info(double control[UMFPACK_CONTROL], double info[UMFPACK_INFO], double) -{ umfpack_di_report_info(control, info);} - -inline void umfpack_report_info(double control[UMFPACK_CONTROL], double info[UMFPACK_INFO], std::complex) -{ umfpack_zi_report_info(control, info);} - -inline void umfpack_report_status(double control[UMFPACK_CONTROL], int status, double) -{ umfpack_di_report_status(control, status);} - -inline void umfpack_report_status(double control[UMFPACK_CONTROL], int status, std::complex) -{ umfpack_zi_report_status(control, status);} - -inline void umfpack_report_control(double control[UMFPACK_CONTROL], double) -{ umfpack_di_report_control(control);} - -inline void umfpack_report_control(double control[UMFPACK_CONTROL], std::complex) -{ umfpack_zi_report_control(control);} - -inline void umfpack_free_numeric(void **Numeric, double) -{ umfpack_di_free_numeric(Numeric); *Numeric = 0; } - -inline void umfpack_free_numeric(void **Numeric, std::complex) -{ umfpack_zi_free_numeric(Numeric); *Numeric = 0; } - -inline void umfpack_free_symbolic(void **Symbolic, double) -{ umfpack_di_free_symbolic(Symbolic); *Symbolic = 0; } - -inline void umfpack_free_symbolic(void **Symbolic, std::complex) -{ umfpack_zi_free_symbolic(Symbolic); *Symbolic = 0; } - -inline int umfpack_symbolic(int n_row,int n_col, - const int Ap[], const int Ai[], const double Ax[], void **Symbolic, - const double Control [UMFPACK_CONTROL], double Info [UMFPACK_INFO]) -{ - return umfpack_di_symbolic(n_row,n_col,Ap,Ai,Ax,Symbolic,Control,Info); -} - -inline int umfpack_symbolic(int n_row,int n_col, - const int Ap[], const int Ai[], const std::complex Ax[], void **Symbolic, - const double Control [UMFPACK_CONTROL], double Info [UMFPACK_INFO]) -{ - return umfpack_zi_symbolic(n_row,n_col,Ap,Ai,&numext::real_ref(Ax[0]),0,Symbolic,Control,Info); -} - -inline int umfpack_numeric( const int Ap[], const int Ai[], const double Ax[], - void *Symbolic, void **Numeric, - const double Control[UMFPACK_CONTROL],double Info [UMFPACK_INFO]) -{ - return umfpack_di_numeric(Ap,Ai,Ax,Symbolic,Numeric,Control,Info); -} - -inline int umfpack_numeric( const int Ap[], const int Ai[], const std::complex Ax[], - void *Symbolic, void **Numeric, - const double Control[UMFPACK_CONTROL],double Info [UMFPACK_INFO]) -{ - return umfpack_zi_numeric(Ap,Ai,&numext::real_ref(Ax[0]),0,Symbolic,Numeric,Control,Info); -} - -inline int umfpack_solve( int sys, const int Ap[], const int Ai[], const double Ax[], - double X[], const double B[], void *Numeric, - const double Control[UMFPACK_CONTROL], double Info[UMFPACK_INFO]) -{ - return umfpack_di_solve(sys,Ap,Ai,Ax,X,B,Numeric,Control,Info); -} - -inline int umfpack_solve( int sys, const int Ap[], const int Ai[], const std::complex Ax[], - std::complex X[], const std::complex B[], void *Numeric, - const double Control[UMFPACK_CONTROL], double Info[UMFPACK_INFO]) -{ - return umfpack_zi_solve(sys,Ap,Ai,&numext::real_ref(Ax[0]),0,&numext::real_ref(X[0]),0,&numext::real_ref(B[0]),0,Numeric,Control,Info); -} - -inline int umfpack_get_lunz(int *lnz, int *unz, int *n_row, int *n_col, int *nz_udiag, void *Numeric, double) -{ - return umfpack_di_get_lunz(lnz,unz,n_row,n_col,nz_udiag,Numeric); -} - -inline int umfpack_get_lunz(int *lnz, int *unz, int *n_row, int *n_col, int *nz_udiag, void *Numeric, std::complex) -{ - return umfpack_zi_get_lunz(lnz,unz,n_row,n_col,nz_udiag,Numeric); -} - -inline int umfpack_get_numeric(int Lp[], int Lj[], double Lx[], int Up[], int Ui[], double Ux[], - int P[], int Q[], double Dx[], int *do_recip, double Rs[], void *Numeric) -{ - return umfpack_di_get_numeric(Lp,Lj,Lx,Up,Ui,Ux,P,Q,Dx,do_recip,Rs,Numeric); -} - -inline int umfpack_get_numeric(int Lp[], int Lj[], std::complex Lx[], int Up[], int Ui[], std::complex Ux[], - int P[], int Q[], std::complex Dx[], int *do_recip, double Rs[], void *Numeric) -{ - double& lx0_real = numext::real_ref(Lx[0]); - double& ux0_real = numext::real_ref(Ux[0]); - double& dx0_real = numext::real_ref(Dx[0]); - return umfpack_zi_get_numeric(Lp,Lj,Lx?&lx0_real:0,0,Up,Ui,Ux?&ux0_real:0,0,P,Q, - Dx?&dx0_real:0,0,do_recip,Rs,Numeric); -} - -inline int umfpack_get_determinant(double *Mx, double *Ex, void *NumericHandle, double User_Info [UMFPACK_INFO]) -{ - return umfpack_di_get_determinant(Mx,Ex,NumericHandle,User_Info); -} - -inline int umfpack_get_determinant(std::complex *Mx, double *Ex, void *NumericHandle, double User_Info [UMFPACK_INFO]) -{ - double& mx_real = numext::real_ref(*Mx); - return umfpack_zi_get_determinant(&mx_real,0,Ex,NumericHandle,User_Info); -} - - -/** \ingroup UmfPackSupport_Module - * \brief A sparse LU factorization and solver based on UmfPack - * - * This class allows to solve for A.X = B sparse linear problems via a LU factorization - * using the UmfPack library. The sparse matrix A must be squared and full rank. - * The vectors or matrices X and B can be either dense or sparse. - * - * \warning The input matrix A should be in a \b compressed and \b column-major form. - * Otherwise an expensive copy will be made. You can call the inexpensive makeCompressed() to get a compressed matrix. - * \tparam _MatrixType the type of the sparse matrix A, it must be a SparseMatrix<> - * - * \implsparsesolverconcept - * - * \sa \ref TutorialSparseSolverConcept, class SparseLU - */ -template -class UmfPackLU : public SparseSolverBase > -{ - protected: - typedef SparseSolverBase > Base; - using Base::m_isInitialized; - public: - using Base::_solve_impl; - typedef _MatrixType MatrixType; - typedef typename MatrixType::Scalar Scalar; - typedef typename MatrixType::RealScalar RealScalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef Matrix Vector; - typedef Matrix IntRowVectorType; - typedef Matrix IntColVectorType; - typedef SparseMatrix LUMatrixType; - typedef SparseMatrix UmfpackMatrixType; - typedef Ref UmfpackMatrixRef; - enum { - ColsAtCompileTime = MatrixType::ColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - - public: - - typedef Array UmfpackControl; - typedef Array UmfpackInfo; - - UmfPackLU() - : m_dummy(0,0), mp_matrix(m_dummy) - { - init(); - } - - template - explicit UmfPackLU(const InputMatrixType& matrix) - : mp_matrix(matrix) - { - init(); - compute(matrix); - } - - ~UmfPackLU() - { - if(m_symbolic) umfpack_free_symbolic(&m_symbolic,Scalar()); - if(m_numeric) umfpack_free_numeric(&m_numeric,Scalar()); - } - - inline Index rows() const { return mp_matrix.rows(); } - inline Index cols() const { return mp_matrix.cols(); } - - /** \brief Reports whether previous computation was successful. - * - * \returns \c Success if computation was succesful, - * \c NumericalIssue if the matrix.appears to be negative. - */ - ComputationInfo info() const - { - eigen_assert(m_isInitialized && "Decomposition is not initialized."); - return m_info; - } - - inline const LUMatrixType& matrixL() const - { - if (m_extractedDataAreDirty) extractData(); - return m_l; - } - - inline const LUMatrixType& matrixU() const - { - if (m_extractedDataAreDirty) extractData(); - return m_u; - } - - inline const IntColVectorType& permutationP() const - { - if (m_extractedDataAreDirty) extractData(); - return m_p; - } - - inline const IntRowVectorType& permutationQ() const - { - if (m_extractedDataAreDirty) extractData(); - return m_q; - } - - /** Computes the sparse Cholesky decomposition of \a matrix - * Note that the matrix should be column-major, and in compressed format for best performance. - * \sa SparseMatrix::makeCompressed(). - */ - template - void compute(const InputMatrixType& matrix) - { - if(m_symbolic) umfpack_free_symbolic(&m_symbolic,Scalar()); - if(m_numeric) umfpack_free_numeric(&m_numeric,Scalar()); - grab(matrix.derived()); - analyzePattern_impl(); - factorize_impl(); - } - - /** Performs a symbolic decomposition on the sparcity of \a matrix. - * - * This function is particularly useful when solving for several problems having the same structure. - * - * \sa factorize(), compute() - */ - template - void analyzePattern(const InputMatrixType& matrix) - { - if(m_symbolic) umfpack_free_symbolic(&m_symbolic,Scalar()); - if(m_numeric) umfpack_free_numeric(&m_numeric,Scalar()); - - grab(matrix.derived()); - - analyzePattern_impl(); - } - - /** Provides the return status code returned by UmfPack during the numeric - * factorization. - * - * \sa factorize(), compute() - */ - inline int umfpackFactorizeReturncode() const - { - eigen_assert(m_numeric && "UmfPackLU: you must first call factorize()"); - return m_fact_errorCode; - } - - /** Provides access to the control settings array used by UmfPack. - * - * If this array contains NaN's, the default values are used. - * - * See UMFPACK documentation for details. - */ - inline const UmfpackControl& umfpackControl() const - { - return m_control; - } - - /** Provides access to the control settings array used by UmfPack. - * - * If this array contains NaN's, the default values are used. - * - * See UMFPACK documentation for details. - */ - inline UmfpackControl& umfpackControl() - { - return m_control; - } - - /** Performs a numeric decomposition of \a matrix - * - * The given matrix must has the same sparcity than the matrix on which the pattern anylysis has been performed. - * - * \sa analyzePattern(), compute() - */ - template - void factorize(const InputMatrixType& matrix) - { - eigen_assert(m_analysisIsOk && "UmfPackLU: you must first call analyzePattern()"); - if(m_numeric) - umfpack_free_numeric(&m_numeric,Scalar()); - - grab(matrix.derived()); - - factorize_impl(); - } - - /** Prints the current UmfPack control settings. - * - * \sa umfpackControl() - */ - void umfpackReportControl() - { - umfpack_report_control(m_control.data(), Scalar()); - } - - /** Prints statistics collected by UmfPack. - * - * \sa analyzePattern(), compute() - */ - void umfpackReportInfo() - { - eigen_assert(m_analysisIsOk && "UmfPackLU: you must first call analyzePattern()"); - umfpack_report_info(m_control.data(), m_umfpackInfo.data(), Scalar()); - } - - /** Prints the status of the previous factorization operation performed by UmfPack (symbolic or numerical factorization). - * - * \sa analyzePattern(), compute() - */ - void umfpackReportStatus() { - eigen_assert(m_analysisIsOk && "UmfPackLU: you must first call analyzePattern()"); - umfpack_report_status(m_control.data(), m_fact_errorCode, Scalar()); - } - - /** \internal */ - template - bool _solve_impl(const MatrixBase &b, MatrixBase &x) const; - - Scalar determinant() const; - - void extractData() const; - - protected: - - void init() - { - m_info = InvalidInput; - m_isInitialized = false; - m_numeric = 0; - m_symbolic = 0; - m_extractedDataAreDirty = true; - - umfpack_defaults(m_control.data(), Scalar()); - } - - void analyzePattern_impl() - { - m_fact_errorCode = umfpack_symbolic(internal::convert_index(mp_matrix.rows()), - internal::convert_index(mp_matrix.cols()), - mp_matrix.outerIndexPtr(), mp_matrix.innerIndexPtr(), mp_matrix.valuePtr(), - &m_symbolic, m_control.data(), m_umfpackInfo.data()); - - m_isInitialized = true; - m_info = m_fact_errorCode ? InvalidInput : Success; - m_analysisIsOk = true; - m_factorizationIsOk = false; - m_extractedDataAreDirty = true; - } - - void factorize_impl() - { - - m_fact_errorCode = umfpack_numeric(mp_matrix.outerIndexPtr(), mp_matrix.innerIndexPtr(), mp_matrix.valuePtr(), - m_symbolic, &m_numeric, m_control.data(), m_umfpackInfo.data()); - - m_info = m_fact_errorCode == UMFPACK_OK ? Success : NumericalIssue; - m_factorizationIsOk = true; - m_extractedDataAreDirty = true; - } - - template - void grab(const EigenBase &A) - { - mp_matrix.~UmfpackMatrixRef(); - ::new (&mp_matrix) UmfpackMatrixRef(A.derived()); - } - - void grab(const UmfpackMatrixRef &A) - { - if(&(A.derived()) != &mp_matrix) - { - mp_matrix.~UmfpackMatrixRef(); - ::new (&mp_matrix) UmfpackMatrixRef(A); - } - } - - // cached data to reduce reallocation, etc. - mutable LUMatrixType m_l; - int m_fact_errorCode; - UmfpackControl m_control; - mutable UmfpackInfo m_umfpackInfo; - - mutable LUMatrixType m_u; - mutable IntColVectorType m_p; - mutable IntRowVectorType m_q; - - UmfpackMatrixType m_dummy; - UmfpackMatrixRef mp_matrix; - - void* m_numeric; - void* m_symbolic; - - mutable ComputationInfo m_info; - int m_factorizationIsOk; - int m_analysisIsOk; - mutable bool m_extractedDataAreDirty; - - private: - UmfPackLU(const UmfPackLU& ) { } -}; - - -template -void UmfPackLU::extractData() const -{ - if (m_extractedDataAreDirty) - { - // get size of the data - int lnz, unz, rows, cols, nz_udiag; - umfpack_get_lunz(&lnz, &unz, &rows, &cols, &nz_udiag, m_numeric, Scalar()); - - // allocate data - m_l.resize(rows,(std::min)(rows,cols)); - m_l.resizeNonZeros(lnz); - - m_u.resize((std::min)(rows,cols),cols); - m_u.resizeNonZeros(unz); - - m_p.resize(rows); - m_q.resize(cols); - - // extract - umfpack_get_numeric(m_l.outerIndexPtr(), m_l.innerIndexPtr(), m_l.valuePtr(), - m_u.outerIndexPtr(), m_u.innerIndexPtr(), m_u.valuePtr(), - m_p.data(), m_q.data(), 0, 0, 0, m_numeric); - - m_extractedDataAreDirty = false; - } -} - -template -typename UmfPackLU::Scalar UmfPackLU::determinant() const -{ - Scalar det; - umfpack_get_determinant(&det, 0, m_numeric, 0); - return det; -} - -template -template -bool UmfPackLU::_solve_impl(const MatrixBase &b, MatrixBase &x) const -{ - Index rhsCols = b.cols(); - eigen_assert((BDerived::Flags&RowMajorBit)==0 && "UmfPackLU backend does not support non col-major rhs yet"); - eigen_assert((XDerived::Flags&RowMajorBit)==0 && "UmfPackLU backend does not support non col-major result yet"); - eigen_assert(b.derived().data() != x.derived().data() && " Umfpack does not support inplace solve"); - - int errorCode; - Scalar* x_ptr = 0; - Matrix x_tmp; - if(x.innerStride()!=1) - { - x_tmp.resize(x.rows()); - x_ptr = x_tmp.data(); - } - for (int j=0; j -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MISC_IMAGE_H -#define EIGEN_MISC_IMAGE_H - -namespace Eigen { - -namespace internal { - -/** \class image_retval_base - * - */ -template -struct traits > -{ - typedef typename DecompositionType::MatrixType MatrixType; - typedef Matrix< - typename MatrixType::Scalar, - MatrixType::RowsAtCompileTime, // the image is a subspace of the destination space, whose - // dimension is the number of rows of the original matrix - Dynamic, // we don't know at compile time the dimension of the image (the rank) - MatrixType::Options, - MatrixType::MaxRowsAtCompileTime, // the image matrix will consist of columns from the original matrix, - MatrixType::MaxColsAtCompileTime // so it has the same number of rows and at most as many columns. - > ReturnType; -}; - -template struct image_retval_base - : public ReturnByValue > -{ - typedef _DecompositionType DecompositionType; - typedef typename DecompositionType::MatrixType MatrixType; - typedef ReturnByValue Base; - - image_retval_base(const DecompositionType& dec, const MatrixType& originalMatrix) - : m_dec(dec), m_rank(dec.rank()), - m_cols(m_rank == 0 ? 1 : m_rank), - m_originalMatrix(originalMatrix) - {} - - inline Index rows() const { return m_dec.rows(); } - inline Index cols() const { return m_cols; } - inline Index rank() const { return m_rank; } - inline const DecompositionType& dec() const { return m_dec; } - inline const MatrixType& originalMatrix() const { return m_originalMatrix; } - - template inline void evalTo(Dest& dst) const - { - static_cast*>(this)->evalTo(dst); - } - - protected: - const DecompositionType& m_dec; - Index m_rank, m_cols; - const MatrixType& m_originalMatrix; -}; - -} // end namespace internal - -#define EIGEN_MAKE_IMAGE_HELPERS(DecompositionType) \ - typedef typename DecompositionType::MatrixType MatrixType; \ - typedef typename MatrixType::Scalar Scalar; \ - typedef typename MatrixType::RealScalar RealScalar; \ - typedef Eigen::internal::image_retval_base Base; \ - using Base::dec; \ - using Base::originalMatrix; \ - using Base::rank; \ - using Base::rows; \ - using Base::cols; \ - image_retval(const DecompositionType& dec, const MatrixType& originalMatrix) \ - : Base(dec, originalMatrix) {} - -} // end namespace Eigen - -#endif // EIGEN_MISC_IMAGE_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h deleted file mode 100644 index bef5d6ff583..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h +++ /dev/null @@ -1,79 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MISC_KERNEL_H -#define EIGEN_MISC_KERNEL_H - -namespace Eigen { - -namespace internal { - -/** \class kernel_retval_base - * - */ -template -struct traits > -{ - typedef typename DecompositionType::MatrixType MatrixType; - typedef Matrix< - typename MatrixType::Scalar, - MatrixType::ColsAtCompileTime, // the number of rows in the "kernel matrix" - // is the number of cols of the original matrix - // so that the product "matrix * kernel = zero" makes sense - Dynamic, // we don't know at compile-time the dimension of the kernel - MatrixType::Options, - MatrixType::MaxColsAtCompileTime, // see explanation for 2nd template parameter - MatrixType::MaxColsAtCompileTime // the kernel is a subspace of the domain space, - // whose dimension is the number of columns of the original matrix - > ReturnType; -}; - -template struct kernel_retval_base - : public ReturnByValue > -{ - typedef _DecompositionType DecompositionType; - typedef ReturnByValue Base; - - explicit kernel_retval_base(const DecompositionType& dec) - : m_dec(dec), - m_rank(dec.rank()), - m_cols(m_rank==dec.cols() ? 1 : dec.cols() - m_rank) - {} - - inline Index rows() const { return m_dec.cols(); } - inline Index cols() const { return m_cols; } - inline Index rank() const { return m_rank; } - inline const DecompositionType& dec() const { return m_dec; } - - template inline void evalTo(Dest& dst) const - { - static_cast*>(this)->evalTo(dst); - } - - protected: - const DecompositionType& m_dec; - Index m_rank, m_cols; -}; - -} // end namespace internal - -#define EIGEN_MAKE_KERNEL_HELPERS(DecompositionType) \ - typedef typename DecompositionType::MatrixType MatrixType; \ - typedef typename MatrixType::Scalar Scalar; \ - typedef typename MatrixType::RealScalar RealScalar; \ - typedef Eigen::internal::kernel_retval_base Base; \ - using Base::dec; \ - using Base::rank; \ - using Base::rows; \ - using Base::cols; \ - kernel_retval(const DecompositionType& dec) : Base(dec) {} - -} // end namespace Eigen - -#endif // EIGEN_MISC_KERNEL_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/RealSvd2x2.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/RealSvd2x2.h deleted file mode 100644 index abb4d3c2fc7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/RealSvd2x2.h +++ /dev/null @@ -1,55 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Benoit Jacob -// Copyright (C) 2013-2016 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_REALSVD2X2_H -#define EIGEN_REALSVD2X2_H - -namespace Eigen { - -namespace internal { - -template -void real_2x2_jacobi_svd(const MatrixType& matrix, Index p, Index q, - JacobiRotation *j_left, - JacobiRotation *j_right) -{ - using std::sqrt; - using std::abs; - Matrix m; - m << numext::real(matrix.coeff(p,p)), numext::real(matrix.coeff(p,q)), - numext::real(matrix.coeff(q,p)), numext::real(matrix.coeff(q,q)); - JacobiRotation rot1; - RealScalar t = m.coeff(0,0) + m.coeff(1,1); - RealScalar d = m.coeff(1,0) - m.coeff(0,1); - - if(abs(d) < (std::numeric_limits::min)()) - { - rot1.s() = RealScalar(0); - rot1.c() = RealScalar(1); - } - else - { - // If d!=0, then t/d cannot overflow because the magnitude of the - // entries forming d are not too small compared to the ones forming t. - RealScalar u = t / d; - RealScalar tmp = sqrt(RealScalar(1) + numext::abs2(u)); - rot1.s() = RealScalar(1) / tmp; - rot1.c() = u / tmp; - } - m.applyOnTheLeft(0,1,rot1); - j_right->makeJacobi(m,0,1); - *j_left = rot1 * j_right->transpose(); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_REALSVD2X2_H diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/blas.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/blas.h deleted file mode 100644 index 25215b15e8a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/blas.h +++ /dev/null @@ -1,440 +0,0 @@ -#ifndef BLAS_H -#define BLAS_H - -#ifdef __cplusplus -extern "C" -{ -#endif - -#define BLASFUNC(FUNC) FUNC##_ - -#ifdef __WIN64__ -typedef long long BLASLONG; -typedef unsigned long long BLASULONG; -#else -typedef long BLASLONG; -typedef unsigned long BLASULONG; -#endif - -int BLASFUNC(xerbla)(const char *, int *info, int); - -float BLASFUNC(sdot) (int *, float *, int *, float *, int *); -float BLASFUNC(sdsdot)(int *, float *, float *, int *, float *, int *); - -double BLASFUNC(dsdot) (int *, float *, int *, float *, int *); -double BLASFUNC(ddot) (int *, double *, int *, double *, int *); -double BLASFUNC(qdot) (int *, double *, int *, double *, int *); - -int BLASFUNC(cdotuw) (int *, float *, int *, float *, int *, float*); -int BLASFUNC(cdotcw) (int *, float *, int *, float *, int *, float*); -int BLASFUNC(zdotuw) (int *, double *, int *, double *, int *, double*); -int BLASFUNC(zdotcw) (int *, double *, int *, double *, int *, double*); - -int BLASFUNC(saxpy) (const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(daxpy) (const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(qaxpy) (const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(caxpy) (const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(zaxpy) (const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(xaxpy) (const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(caxpyc)(const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(zaxpyc)(const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(xaxpyc)(const int *, const double *, const double *, const int *, double *, const int *); - -int BLASFUNC(scopy) (int *, float *, int *, float *, int *); -int BLASFUNC(dcopy) (int *, double *, int *, double *, int *); -int BLASFUNC(qcopy) (int *, double *, int *, double *, int *); -int BLASFUNC(ccopy) (int *, float *, int *, float *, int *); -int BLASFUNC(zcopy) (int *, double *, int *, double *, int *); -int BLASFUNC(xcopy) (int *, double *, int *, double *, int *); - -int BLASFUNC(sswap) (int *, float *, int *, float *, int *); -int BLASFUNC(dswap) (int *, double *, int *, double *, int *); -int BLASFUNC(qswap) (int *, double *, int *, double *, int *); -int BLASFUNC(cswap) (int *, float *, int *, float *, int *); -int BLASFUNC(zswap) (int *, double *, int *, double *, int *); -int BLASFUNC(xswap) (int *, double *, int *, double *, int *); - -float BLASFUNC(sasum) (int *, float *, int *); -float BLASFUNC(scasum)(int *, float *, int *); -double BLASFUNC(dasum) (int *, double *, int *); -double BLASFUNC(qasum) (int *, double *, int *); -double BLASFUNC(dzasum)(int *, double *, int *); -double BLASFUNC(qxasum)(int *, double *, int *); - -int BLASFUNC(isamax)(int *, float *, int *); -int BLASFUNC(idamax)(int *, double *, int *); -int BLASFUNC(iqamax)(int *, double *, int *); -int BLASFUNC(icamax)(int *, float *, int *); -int BLASFUNC(izamax)(int *, double *, int *); -int BLASFUNC(ixamax)(int *, double *, int *); - -int BLASFUNC(ismax) (int *, float *, int *); -int BLASFUNC(idmax) (int *, double *, int *); -int BLASFUNC(iqmax) (int *, double *, int *); -int BLASFUNC(icmax) (int *, float *, int *); -int BLASFUNC(izmax) (int *, double *, int *); -int BLASFUNC(ixmax) (int *, double *, int *); - -int BLASFUNC(isamin)(int *, float *, int *); -int BLASFUNC(idamin)(int *, double *, int *); -int BLASFUNC(iqamin)(int *, double *, int *); -int BLASFUNC(icamin)(int *, float *, int *); -int BLASFUNC(izamin)(int *, double *, int *); -int BLASFUNC(ixamin)(int *, double *, int *); - -int BLASFUNC(ismin)(int *, float *, int *); -int BLASFUNC(idmin)(int *, double *, int *); -int BLASFUNC(iqmin)(int *, double *, int *); -int BLASFUNC(icmin)(int *, float *, int *); -int BLASFUNC(izmin)(int *, double *, int *); -int BLASFUNC(ixmin)(int *, double *, int *); - -float BLASFUNC(samax) (int *, float *, int *); -double BLASFUNC(damax) (int *, double *, int *); -double BLASFUNC(qamax) (int *, double *, int *); -float BLASFUNC(scamax)(int *, float *, int *); -double BLASFUNC(dzamax)(int *, double *, int *); -double BLASFUNC(qxamax)(int *, double *, int *); - -float BLASFUNC(samin) (int *, float *, int *); -double BLASFUNC(damin) (int *, double *, int *); -double BLASFUNC(qamin) (int *, double *, int *); -float BLASFUNC(scamin)(int *, float *, int *); -double BLASFUNC(dzamin)(int *, double *, int *); -double BLASFUNC(qxamin)(int *, double *, int *); - -float BLASFUNC(smax) (int *, float *, int *); -double BLASFUNC(dmax) (int *, double *, int *); -double BLASFUNC(qmax) (int *, double *, int *); -float BLASFUNC(scmax) (int *, float *, int *); -double BLASFUNC(dzmax) (int *, double *, int *); -double BLASFUNC(qxmax) (int *, double *, int *); - -float BLASFUNC(smin) (int *, float *, int *); -double BLASFUNC(dmin) (int *, double *, int *); -double BLASFUNC(qmin) (int *, double *, int *); -float BLASFUNC(scmin) (int *, float *, int *); -double BLASFUNC(dzmin) (int *, double *, int *); -double BLASFUNC(qxmin) (int *, double *, int *); - -int BLASFUNC(sscal) (int *, float *, float *, int *); -int BLASFUNC(dscal) (int *, double *, double *, int *); -int BLASFUNC(qscal) (int *, double *, double *, int *); -int BLASFUNC(cscal) (int *, float *, float *, int *); -int BLASFUNC(zscal) (int *, double *, double *, int *); -int BLASFUNC(xscal) (int *, double *, double *, int *); -int BLASFUNC(csscal)(int *, float *, float *, int *); -int BLASFUNC(zdscal)(int *, double *, double *, int *); -int BLASFUNC(xqscal)(int *, double *, double *, int *); - -float BLASFUNC(snrm2) (int *, float *, int *); -float BLASFUNC(scnrm2)(int *, float *, int *); - -double BLASFUNC(dnrm2) (int *, double *, int *); -double BLASFUNC(qnrm2) (int *, double *, int *); -double BLASFUNC(dznrm2)(int *, double *, int *); -double BLASFUNC(qxnrm2)(int *, double *, int *); - -int BLASFUNC(srot) (int *, float *, int *, float *, int *, float *, float *); -int BLASFUNC(drot) (int *, double *, int *, double *, int *, double *, double *); -int BLASFUNC(qrot) (int *, double *, int *, double *, int *, double *, double *); -int BLASFUNC(csrot) (int *, float *, int *, float *, int *, float *, float *); -int BLASFUNC(zdrot) (int *, double *, int *, double *, int *, double *, double *); -int BLASFUNC(xqrot) (int *, double *, int *, double *, int *, double *, double *); - -int BLASFUNC(srotg) (float *, float *, float *, float *); -int BLASFUNC(drotg) (double *, double *, double *, double *); -int BLASFUNC(qrotg) (double *, double *, double *, double *); -int BLASFUNC(crotg) (float *, float *, float *, float *); -int BLASFUNC(zrotg) (double *, double *, double *, double *); -int BLASFUNC(xrotg) (double *, double *, double *, double *); - -int BLASFUNC(srotmg)(float *, float *, float *, float *, float *); -int BLASFUNC(drotmg)(double *, double *, double *, double *, double *); - -int BLASFUNC(srotm) (int *, float *, int *, float *, int *, float *); -int BLASFUNC(drotm) (int *, double *, int *, double *, int *, double *); -int BLASFUNC(qrotm) (int *, double *, int *, double *, int *, double *); - -/* Level 2 routines */ - -int BLASFUNC(sger)(int *, int *, float *, float *, int *, - float *, int *, float *, int *); -int BLASFUNC(dger)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); -int BLASFUNC(qger)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); -int BLASFUNC(cgeru)(int *, int *, float *, float *, int *, - float *, int *, float *, int *); -int BLASFUNC(cgerc)(int *, int *, float *, float *, int *, - float *, int *, float *, int *); -int BLASFUNC(zgeru)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); -int BLASFUNC(zgerc)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); -int BLASFUNC(xgeru)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); -int BLASFUNC(xgerc)(int *, int *, double *, double *, int *, - double *, int *, double *, int *); - -int BLASFUNC(sgemv)(const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dgemv)(const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(qgemv)(const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(cgemv)(const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zgemv)(const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xgemv)(const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(strsv) (const char *, const char *, const char *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(dtrsv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(qtrsv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(ctrsv) (const char *, const char *, const char *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(ztrsv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(xtrsv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); - -int BLASFUNC(stpsv) (char *, char *, char *, int *, float *, float *, int *); -int BLASFUNC(dtpsv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(qtpsv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(ctpsv) (char *, char *, char *, int *, float *, float *, int *); -int BLASFUNC(ztpsv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(xtpsv) (char *, char *, char *, int *, double *, double *, int *); - -int BLASFUNC(strmv) (const char *, const char *, const char *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(dtrmv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(qtrmv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(ctrmv) (const char *, const char *, const char *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(ztrmv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(xtrmv) (const char *, const char *, const char *, const int *, const double *, const int *, double *, const int *); - -int BLASFUNC(stpmv) (char *, char *, char *, int *, float *, float *, int *); -int BLASFUNC(dtpmv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(qtpmv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(ctpmv) (char *, char *, char *, int *, float *, float *, int *); -int BLASFUNC(ztpmv) (char *, char *, char *, int *, double *, double *, int *); -int BLASFUNC(xtpmv) (char *, char *, char *, int *, double *, double *, int *); - -int BLASFUNC(stbmv) (char *, char *, char *, int *, int *, float *, int *, float *, int *); -int BLASFUNC(dtbmv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(qtbmv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(ctbmv) (char *, char *, char *, int *, int *, float *, int *, float *, int *); -int BLASFUNC(ztbmv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(xtbmv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); - -int BLASFUNC(stbsv) (char *, char *, char *, int *, int *, float *, int *, float *, int *); -int BLASFUNC(dtbsv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(qtbsv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(ctbsv) (char *, char *, char *, int *, int *, float *, int *, float *, int *); -int BLASFUNC(ztbsv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); -int BLASFUNC(xtbsv) (char *, char *, char *, int *, int *, double *, int *, double *, int *); - -int BLASFUNC(ssymv) (const char *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dsymv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(qsymv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(sspmv) (char *, int *, float *, float *, - float *, int *, float *, float *, int *); -int BLASFUNC(dspmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); -int BLASFUNC(qspmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); - -int BLASFUNC(ssyr) (const char *, const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(dsyr) (const char *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(qsyr) (const char *, const int *, const double *, const double *, const int *, double *, const int *); - -int BLASFUNC(ssyr2) (const char *, const int *, const float *, const float *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(dsyr2) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(qsyr2) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(csyr2) (const char *, const int *, const float *, const float *, const int *, const float *, const int *, float *, const int *); -int BLASFUNC(zsyr2) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, double *, const int *); -int BLASFUNC(xsyr2) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, double *, const int *); - -int BLASFUNC(sspr) (char *, int *, float *, float *, int *, - float *); -int BLASFUNC(dspr) (char *, int *, double *, double *, int *, - double *); -int BLASFUNC(qspr) (char *, int *, double *, double *, int *, - double *); - -int BLASFUNC(sspr2) (char *, int *, float *, - float *, int *, float *, int *, float *); -int BLASFUNC(dspr2) (char *, int *, double *, - double *, int *, double *, int *, double *); -int BLASFUNC(qspr2) (char *, int *, double *, - double *, int *, double *, int *, double *); -int BLASFUNC(cspr2) (char *, int *, float *, - float *, int *, float *, int *, float *); -int BLASFUNC(zspr2) (char *, int *, double *, - double *, int *, double *, int *, double *); -int BLASFUNC(xspr2) (char *, int *, double *, - double *, int *, double *, int *, double *); - -int BLASFUNC(cher) (char *, int *, float *, float *, int *, - float *, int *); -int BLASFUNC(zher) (char *, int *, double *, double *, int *, - double *, int *); -int BLASFUNC(xher) (char *, int *, double *, double *, int *, - double *, int *); - -int BLASFUNC(chpr) (char *, int *, float *, float *, int *, float *); -int BLASFUNC(zhpr) (char *, int *, double *, double *, int *, double *); -int BLASFUNC(xhpr) (char *, int *, double *, double *, int *, double *); - -int BLASFUNC(cher2) (char *, int *, float *, - float *, int *, float *, int *, float *, int *); -int BLASFUNC(zher2) (char *, int *, double *, - double *, int *, double *, int *, double *, int *); -int BLASFUNC(xher2) (char *, int *, double *, - double *, int *, double *, int *, double *, int *); - -int BLASFUNC(chpr2) (char *, int *, float *, - float *, int *, float *, int *, float *); -int BLASFUNC(zhpr2) (char *, int *, double *, - double *, int *, double *, int *, double *); -int BLASFUNC(xhpr2) (char *, int *, double *, - double *, int *, double *, int *, double *); - -int BLASFUNC(chemv) (const char *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zhemv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xhemv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(chpmv) (char *, int *, float *, float *, - float *, int *, float *, float *, int *); -int BLASFUNC(zhpmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); -int BLASFUNC(xhpmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); - -int BLASFUNC(snorm)(char *, int *, int *, float *, int *); -int BLASFUNC(dnorm)(char *, int *, int *, double *, int *); -int BLASFUNC(cnorm)(char *, int *, int *, float *, int *); -int BLASFUNC(znorm)(char *, int *, int *, double *, int *); - -int BLASFUNC(sgbmv)(char *, int *, int *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(dgbmv)(char *, int *, int *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(qgbmv)(char *, int *, int *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(cgbmv)(char *, int *, int *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(zgbmv)(char *, int *, int *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(xgbmv)(char *, int *, int *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); - -int BLASFUNC(ssbmv)(char *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(dsbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(qsbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(csbmv)(char *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(zsbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(xsbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); - -int BLASFUNC(chbmv)(char *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(zhbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(xhbmv)(char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); - -/* Level 3 routines */ - -int BLASFUNC(sgemm)(const char *, const char *, const int *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dgemm)(const char *, const char *, const int *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(qgemm)(const char *, const char *, const int *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(cgemm)(const char *, const char *, const int *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zgemm)(const char *, const char *, const int *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xgemm)(const char *, const char *, const int *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(cgemm3m)(char *, char *, int *, int *, int *, float *, - float *, int *, float *, int *, float *, float *, int *); -int BLASFUNC(zgemm3m)(char *, char *, int *, int *, int *, double *, - double *, int *, double *, int *, double *, double *, int *); -int BLASFUNC(xgemm3m)(char *, char *, int *, int *, int *, double *, - double *, int *, double *, int *, double *, double *, int *); - -int BLASFUNC(sge2mm)(char *, char *, char *, int *, int *, - float *, float *, int *, float *, int *, - float *, float *, int *); -int BLASFUNC(dge2mm)(char *, char *, char *, int *, int *, - double *, double *, int *, double *, int *, - double *, double *, int *); -int BLASFUNC(cge2mm)(char *, char *, char *, int *, int *, - float *, float *, int *, float *, int *, - float *, float *, int *); -int BLASFUNC(zge2mm)(char *, char *, char *, int *, int *, - double *, double *, int *, double *, int *, - double *, double *, int *); - -int BLASFUNC(strsm)(const char *, const char *, const char *, const char *, const int *, const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(dtrsm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(qtrsm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(ctrsm)(const char *, const char *, const char *, const char *, const int *, const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(ztrsm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(xtrsm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); - -int BLASFUNC(strmm)(const char *, const char *, const char *, const char *, const int *, const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(dtrmm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(qtrmm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(ctrmm)(const char *, const char *, const char *, const char *, const int *, const int *, const float *, const float *, const int *, float *, const int *); -int BLASFUNC(ztrmm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); -int BLASFUNC(xtrmm)(const char *, const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, double *, const int *); - -int BLASFUNC(ssymm)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dsymm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(qsymm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(csymm)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zsymm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xsymm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(csymm3m)(char *, char *, int *, int *, float *, float *, int *, float *, int *, float *, float *, int *); -int BLASFUNC(zsymm3m)(char *, char *, int *, int *, double *, double *, int *, double *, int *, double *, double *, int *); -int BLASFUNC(xsymm3m)(char *, char *, int *, int *, double *, double *, int *, double *, int *, double *, double *, int *); - -int BLASFUNC(ssyrk)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dsyrk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(qsyrk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(csyrk)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zsyrk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xsyrk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(ssyr2k)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(dsyr2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); -int BLASFUNC(qsyr2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); -int BLASFUNC(csyr2k)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zsyr2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); -int BLASFUNC(xsyr2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); - -int BLASFUNC(chemm)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zhemm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xhemm)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(chemm3m)(char *, char *, int *, int *, float *, float *, int *, - float *, int *, float *, float *, int *); -int BLASFUNC(zhemm3m)(char *, char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); -int BLASFUNC(xhemm3m)(char *, char *, int *, int *, double *, double *, int *, - double *, int *, double *, double *, int *); - -int BLASFUNC(cherk)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zherk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xherk)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, double *, const int *); - -int BLASFUNC(cher2k)(const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zher2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xher2k)(const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(cher2m)(const char *, const char *, const char *, const int *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zher2m)(const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); -int BLASFUNC(xher2m)(const char *, const char *, const char *, const int *, const int *, const double *, const double *, const int *, const double*, const int *, const double *, double *, const int *); - - -#ifdef __cplusplus -} -#endif - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapack.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapack.h deleted file mode 100644 index 249f3575c6d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapack.h +++ /dev/null @@ -1,152 +0,0 @@ -#ifndef LAPACK_H -#define LAPACK_H - -#include "blas.h" - -#ifdef __cplusplus -extern "C" -{ -#endif - -int BLASFUNC(csymv) (const char *, const int *, const float *, const float *, const int *, const float *, const int *, const float *, float *, const int *); -int BLASFUNC(zsymv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); -int BLASFUNC(xsymv) (const char *, const int *, const double *, const double *, const int *, const double *, const int *, const double *, double *, const int *); - - -int BLASFUNC(cspmv) (char *, int *, float *, float *, - float *, int *, float *, float *, int *); -int BLASFUNC(zspmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); -int BLASFUNC(xspmv) (char *, int *, double *, double *, - double *, int *, double *, double *, int *); - -int BLASFUNC(csyr) (char *, int *, float *, float *, int *, - float *, int *); -int BLASFUNC(zsyr) (char *, int *, double *, double *, int *, - double *, int *); -int BLASFUNC(xsyr) (char *, int *, double *, double *, int *, - double *, int *); - -int BLASFUNC(cspr) (char *, int *, float *, float *, int *, - float *); -int BLASFUNC(zspr) (char *, int *, double *, double *, int *, - double *); -int BLASFUNC(xspr) (char *, int *, double *, double *, int *, - double *); - -int BLASFUNC(sgemt)(char *, int *, int *, float *, float *, int *, - float *, int *); -int BLASFUNC(dgemt)(char *, int *, int *, double *, double *, int *, - double *, int *); -int BLASFUNC(cgemt)(char *, int *, int *, float *, float *, int *, - float *, int *); -int BLASFUNC(zgemt)(char *, int *, int *, double *, double *, int *, - double *, int *); - -int BLASFUNC(sgema)(char *, char *, int *, int *, float *, - float *, int *, float *, float *, int *, float *, int *); -int BLASFUNC(dgema)(char *, char *, int *, int *, double *, - double *, int *, double*, double *, int *, double*, int *); -int BLASFUNC(cgema)(char *, char *, int *, int *, float *, - float *, int *, float *, float *, int *, float *, int *); -int BLASFUNC(zgema)(char *, char *, int *, int *, double *, - double *, int *, double*, double *, int *, double*, int *); - -int BLASFUNC(sgems)(char *, char *, int *, int *, float *, - float *, int *, float *, float *, int *, float *, int *); -int BLASFUNC(dgems)(char *, char *, int *, int *, double *, - double *, int *, double*, double *, int *, double*, int *); -int BLASFUNC(cgems)(char *, char *, int *, int *, float *, - float *, int *, float *, float *, int *, float *, int *); -int BLASFUNC(zgems)(char *, char *, int *, int *, double *, - double *, int *, double*, double *, int *, double*, int *); - -int BLASFUNC(sgetf2)(int *, int *, float *, int *, int *, int *); -int BLASFUNC(dgetf2)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(qgetf2)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(cgetf2)(int *, int *, float *, int *, int *, int *); -int BLASFUNC(zgetf2)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(xgetf2)(int *, int *, double *, int *, int *, int *); - -int BLASFUNC(sgetrf)(int *, int *, float *, int *, int *, int *); -int BLASFUNC(dgetrf)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(qgetrf)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(cgetrf)(int *, int *, float *, int *, int *, int *); -int BLASFUNC(zgetrf)(int *, int *, double *, int *, int *, int *); -int BLASFUNC(xgetrf)(int *, int *, double *, int *, int *, int *); - -int BLASFUNC(slaswp)(int *, float *, int *, int *, int *, int *, int *); -int BLASFUNC(dlaswp)(int *, double *, int *, int *, int *, int *, int *); -int BLASFUNC(qlaswp)(int *, double *, int *, int *, int *, int *, int *); -int BLASFUNC(claswp)(int *, float *, int *, int *, int *, int *, int *); -int BLASFUNC(zlaswp)(int *, double *, int *, int *, int *, int *, int *); -int BLASFUNC(xlaswp)(int *, double *, int *, int *, int *, int *, int *); - -int BLASFUNC(sgetrs)(char *, int *, int *, float *, int *, int *, float *, int *, int *); -int BLASFUNC(dgetrs)(char *, int *, int *, double *, int *, int *, double *, int *, int *); -int BLASFUNC(qgetrs)(char *, int *, int *, double *, int *, int *, double *, int *, int *); -int BLASFUNC(cgetrs)(char *, int *, int *, float *, int *, int *, float *, int *, int *); -int BLASFUNC(zgetrs)(char *, int *, int *, double *, int *, int *, double *, int *, int *); -int BLASFUNC(xgetrs)(char *, int *, int *, double *, int *, int *, double *, int *, int *); - -int BLASFUNC(sgesv)(int *, int *, float *, int *, int *, float *, int *, int *); -int BLASFUNC(dgesv)(int *, int *, double *, int *, int *, double*, int *, int *); -int BLASFUNC(qgesv)(int *, int *, double *, int *, int *, double*, int *, int *); -int BLASFUNC(cgesv)(int *, int *, float *, int *, int *, float *, int *, int *); -int BLASFUNC(zgesv)(int *, int *, double *, int *, int *, double*, int *, int *); -int BLASFUNC(xgesv)(int *, int *, double *, int *, int *, double*, int *, int *); - -int BLASFUNC(spotf2)(char *, int *, float *, int *, int *); -int BLASFUNC(dpotf2)(char *, int *, double *, int *, int *); -int BLASFUNC(qpotf2)(char *, int *, double *, int *, int *); -int BLASFUNC(cpotf2)(char *, int *, float *, int *, int *); -int BLASFUNC(zpotf2)(char *, int *, double *, int *, int *); -int BLASFUNC(xpotf2)(char *, int *, double *, int *, int *); - -int BLASFUNC(spotrf)(char *, int *, float *, int *, int *); -int BLASFUNC(dpotrf)(char *, int *, double *, int *, int *); -int BLASFUNC(qpotrf)(char *, int *, double *, int *, int *); -int BLASFUNC(cpotrf)(char *, int *, float *, int *, int *); -int BLASFUNC(zpotrf)(char *, int *, double *, int *, int *); -int BLASFUNC(xpotrf)(char *, int *, double *, int *, int *); - -int BLASFUNC(slauu2)(char *, int *, float *, int *, int *); -int BLASFUNC(dlauu2)(char *, int *, double *, int *, int *); -int BLASFUNC(qlauu2)(char *, int *, double *, int *, int *); -int BLASFUNC(clauu2)(char *, int *, float *, int *, int *); -int BLASFUNC(zlauu2)(char *, int *, double *, int *, int *); -int BLASFUNC(xlauu2)(char *, int *, double *, int *, int *); - -int BLASFUNC(slauum)(char *, int *, float *, int *, int *); -int BLASFUNC(dlauum)(char *, int *, double *, int *, int *); -int BLASFUNC(qlauum)(char *, int *, double *, int *, int *); -int BLASFUNC(clauum)(char *, int *, float *, int *, int *); -int BLASFUNC(zlauum)(char *, int *, double *, int *, int *); -int BLASFUNC(xlauum)(char *, int *, double *, int *, int *); - -int BLASFUNC(strti2)(char *, char *, int *, float *, int *, int *); -int BLASFUNC(dtrti2)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(qtrti2)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(ctrti2)(char *, char *, int *, float *, int *, int *); -int BLASFUNC(ztrti2)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(xtrti2)(char *, char *, int *, double *, int *, int *); - -int BLASFUNC(strtri)(char *, char *, int *, float *, int *, int *); -int BLASFUNC(dtrtri)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(qtrtri)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(ctrtri)(char *, char *, int *, float *, int *, int *); -int BLASFUNC(ztrtri)(char *, char *, int *, double *, int *, int *); -int BLASFUNC(xtrtri)(char *, char *, int *, double *, int *, int *); - -int BLASFUNC(spotri)(char *, int *, float *, int *, int *); -int BLASFUNC(dpotri)(char *, int *, double *, int *, int *); -int BLASFUNC(qpotri)(char *, int *, double *, int *, int *); -int BLASFUNC(cpotri)(char *, int *, float *, int *, int *); -int BLASFUNC(zpotri)(char *, int *, double *, int *, int *); -int BLASFUNC(xpotri)(char *, int *, double *, int *, int *); - -#ifdef __cplusplus -} -#endif - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke.h deleted file mode 100644 index 8c7e79b034c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke.h +++ /dev/null @@ -1,16291 +0,0 @@ -/***************************************************************************** - Copyright (c) 2010, Intel Corp. - All rights reserved. - - Redistribution and use in source and binary forms, with or without - modification, are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, - this list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright - notice, this list of conditions and the following disclaimer in the - documentation and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors - may be used to endorse or promote products derived from this software - without specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" - AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE - IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE - ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE - LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR - CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF - SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS - INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN - CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) - ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF - THE POSSIBILITY OF SUCH DAMAGE. -****************************************************************************** -* Contents: Native C interface to LAPACK -* Author: Intel Corporation -* Generated November, 2011 -*****************************************************************************/ - -#ifndef _MKL_LAPACKE_H_ - -#ifndef _LAPACKE_H_ -#define _LAPACKE_H_ - -/* -* Turn on HAVE_LAPACK_CONFIG_H to redefine C-LAPACK datatypes -*/ -#ifdef HAVE_LAPACK_CONFIG_H -#include "lapacke_config.h" -#endif - -#ifdef __cplusplus -extern "C" { -#endif /* __cplusplus */ - -#include - -#ifndef lapack_int -#define lapack_int int -#endif - -#ifndef lapack_logical -#define lapack_logical lapack_int -#endif - -/* Complex types are structures equivalent to the -* Fortran complex types COMPLEX(4) and COMPLEX(8). -* -* One can also redefine the types with his own types -* for example by including in the code definitions like -* -* #define lapack_complex_float std::complex -* #define lapack_complex_double std::complex -* -* or define these types in the command line: -* -* -Dlapack_complex_float="std::complex" -* -Dlapack_complex_double="std::complex" -*/ - -#ifndef LAPACK_COMPLEX_CUSTOM - -/* Complex type (single precision) */ -#ifndef lapack_complex_float -#include -#define lapack_complex_float float _Complex -#endif - -#ifndef lapack_complex_float_real -#define lapack_complex_float_real(z) (creal(z)) -#endif - -#ifndef lapack_complex_float_imag -#define lapack_complex_float_imag(z) (cimag(z)) -#endif - -lapack_complex_float lapack_make_complex_float( float re, float im ); - -/* Complex type (double precision) */ -#ifndef lapack_complex_double -#include -#define lapack_complex_double double _Complex -#endif - -#ifndef lapack_complex_double_real -#define lapack_complex_double_real(z) (creal(z)) -#endif - -#ifndef lapack_complex_double_imag -#define lapack_complex_double_imag(z) (cimag(z)) -#endif - -lapack_complex_double lapack_make_complex_double( double re, double im ); - -#endif - -#ifndef LAPACKE_malloc -#define LAPACKE_malloc( size ) malloc( size ) -#endif -#ifndef LAPACKE_free -#define LAPACKE_free( p ) free( p ) -#endif - -#define LAPACK_C2INT( x ) (lapack_int)(*((float*)&x )) -#define LAPACK_Z2INT( x ) (lapack_int)(*((double*)&x )) - -#define LAPACK_ROW_MAJOR 101 -#define LAPACK_COL_MAJOR 102 - -#define LAPACK_WORK_MEMORY_ERROR -1010 -#define LAPACK_TRANSPOSE_MEMORY_ERROR -1011 - -/* Callback logical functions of one, two, or three arguments are used -* to select eigenvalues to sort to the top left of the Schur form. -* The value is selected if function returns TRUE (non-zero). */ - -typedef lapack_logical (*LAPACK_S_SELECT2) ( const float*, const float* ); -typedef lapack_logical (*LAPACK_S_SELECT3) - ( const float*, const float*, const float* ); -typedef lapack_logical (*LAPACK_D_SELECT2) ( const double*, const double* ); -typedef lapack_logical (*LAPACK_D_SELECT3) - ( const double*, const double*, const double* ); - -typedef lapack_logical (*LAPACK_C_SELECT1) ( const lapack_complex_float* ); -typedef lapack_logical (*LAPACK_C_SELECT2) - ( const lapack_complex_float*, const lapack_complex_float* ); -typedef lapack_logical (*LAPACK_Z_SELECT1) ( const lapack_complex_double* ); -typedef lapack_logical (*LAPACK_Z_SELECT2) - ( const lapack_complex_double*, const lapack_complex_double* ); - -#include "lapacke_mangling.h" - -#define LAPACK_lsame LAPACK_GLOBAL(lsame,LSAME) -lapack_logical LAPACK_lsame( char* ca, char* cb, - lapack_int lca, lapack_int lcb ); - -/* C-LAPACK function prototypes */ - -lapack_int LAPACKE_sbdsdc( int matrix_order, char uplo, char compq, - lapack_int n, float* d, float* e, float* u, - lapack_int ldu, float* vt, lapack_int ldvt, float* q, - lapack_int* iq ); -lapack_int LAPACKE_dbdsdc( int matrix_order, char uplo, char compq, - lapack_int n, double* d, double* e, double* u, - lapack_int ldu, double* vt, lapack_int ldvt, - double* q, lapack_int* iq ); - -lapack_int LAPACKE_sbdsqr( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - float* d, float* e, float* vt, lapack_int ldvt, - float* u, lapack_int ldu, float* c, lapack_int ldc ); -lapack_int LAPACKE_dbdsqr( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - double* d, double* e, double* vt, lapack_int ldvt, - double* u, lapack_int ldu, double* c, - lapack_int ldc ); -lapack_int LAPACKE_cbdsqr( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - float* d, float* e, lapack_complex_float* vt, - lapack_int ldvt, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* c, - lapack_int ldc ); -lapack_int LAPACKE_zbdsqr( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - double* d, double* e, lapack_complex_double* vt, - lapack_int ldvt, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* c, - lapack_int ldc ); - -lapack_int LAPACKE_sdisna( char job, lapack_int m, lapack_int n, const float* d, - float* sep ); -lapack_int LAPACKE_ddisna( char job, lapack_int m, lapack_int n, - const double* d, double* sep ); - -lapack_int LAPACKE_sgbbrd( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, float* ab, lapack_int ldab, float* d, - float* e, float* q, lapack_int ldq, float* pt, - lapack_int ldpt, float* c, lapack_int ldc ); -lapack_int LAPACKE_dgbbrd( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, double* ab, lapack_int ldab, - double* d, double* e, double* q, lapack_int ldq, - double* pt, lapack_int ldpt, double* c, - lapack_int ldc ); -lapack_int LAPACKE_cgbbrd( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, lapack_complex_float* ab, - lapack_int ldab, float* d, float* e, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* pt, lapack_int ldpt, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zgbbrd( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, lapack_complex_double* ab, - lapack_int ldab, double* d, double* e, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* pt, lapack_int ldpt, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_sgbcon( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, const lapack_int* ipiv, float anorm, - float* rcond ); -lapack_int LAPACKE_dgbcon( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, const lapack_int* ipiv, - double anorm, double* rcond ); -lapack_int LAPACKE_cgbcon( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zgbcon( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, lapack_int ldab, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_sgbequ( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, float* r, float* c, float* rowcnd, - float* colcnd, float* amax ); -lapack_int LAPACKE_dgbequ( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); -lapack_int LAPACKE_cgbequ( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, lapack_int ldab, - float* r, float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_zgbequ( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, lapack_int ldab, - double* r, double* c, double* rowcnd, double* colcnd, - double* amax ); - -lapack_int LAPACKE_sgbequb( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, float* r, float* c, float* rowcnd, - float* colcnd, float* amax ); -lapack_int LAPACKE_dgbequb( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); -lapack_int LAPACKE_cgbequb( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, lapack_int ldab, - float* r, float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_zgbequb( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, lapack_int ldab, - double* r, double* c, double* rowcnd, - double* colcnd, double* amax ); - -lapack_int LAPACKE_sgbrfs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const float* ab, lapack_int ldab, const float* afb, - lapack_int ldafb, const lapack_int* ipiv, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_dgbrfs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const double* ab, lapack_int ldab, const double* afb, - lapack_int ldafb, const lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr ); -lapack_int LAPACKE_cgbrfs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* afb, lapack_int ldafb, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zgbrfs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - const lapack_complex_double* afb, lapack_int ldafb, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sgbrfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const float* ab, lapack_int ldab, - const float* afb, lapack_int ldafb, - const lapack_int* ipiv, const float* r, - const float* c, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dgbrfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const double* ab, lapack_int ldab, - const double* afb, lapack_int ldafb, - const lapack_int* ipiv, const double* r, - const double* c, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); -lapack_int LAPACKE_cgbrfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const lapack_complex_float* ab, - lapack_int ldab, const lapack_complex_float* afb, - lapack_int ldafb, const lapack_int* ipiv, - const float* r, const float* c, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_zgbrfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const lapack_complex_double* ab, - lapack_int ldab, const lapack_complex_double* afb, - lapack_int ldafb, const lapack_int* ipiv, - const double* r, const double* c, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); - -lapack_int LAPACKE_sgbsv( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, float* ab, - lapack_int ldab, lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgbsv( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, double* ab, - lapack_int ldab, lapack_int* ipiv, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cgbsv( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgbsv( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sgbsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, float* ab, lapack_int ldab, - float* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, float* r, float* c, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* rpivot ); -lapack_int LAPACKE_dgbsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, double* ab, lapack_int ldab, - double* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, double* r, double* c, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* rpivot ); -lapack_int LAPACKE_cgbsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - float* r, float* c, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, float* rpivot ); -lapack_int LAPACKE_zgbsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - double* r, double* c, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, double* rpivot ); - -lapack_int LAPACKE_sgbsvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, float* ab, lapack_int ldab, - float* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, float* r, float* c, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dgbsvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, double* ab, lapack_int ldab, - double* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, double* r, double* c, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); -lapack_int LAPACKE_cgbsvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - float* r, float* c, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_zgbsvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - double* r, double* c, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); - -lapack_int LAPACKE_sgbtrf( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, float* ab, - lapack_int ldab, lapack_int* ipiv ); -lapack_int LAPACKE_dgbtrf( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, double* ab, - lapack_int ldab, lapack_int* ipiv ); -lapack_int LAPACKE_cgbtrf( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - lapack_complex_float* ab, lapack_int ldab, - lapack_int* ipiv ); -lapack_int LAPACKE_zgbtrf( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - lapack_complex_double* ab, lapack_int ldab, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgbtrs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const float* ab, lapack_int ldab, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dgbtrs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const double* ab, lapack_int ldab, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_cgbtrs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgbtrs( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sgebak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const float* scale, - lapack_int m, float* v, lapack_int ldv ); -lapack_int LAPACKE_dgebak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const double* scale, - lapack_int m, double* v, lapack_int ldv ); -lapack_int LAPACKE_cgebak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const float* scale, - lapack_int m, lapack_complex_float* v, - lapack_int ldv ); -lapack_int LAPACKE_zgebak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const double* scale, - lapack_int m, lapack_complex_double* v, - lapack_int ldv ); - -lapack_int LAPACKE_sgebal( int matrix_order, char job, lapack_int n, float* a, - lapack_int lda, lapack_int* ilo, lapack_int* ihi, - float* scale ); -lapack_int LAPACKE_dgebal( int matrix_order, char job, lapack_int n, double* a, - lapack_int lda, lapack_int* ilo, lapack_int* ihi, - double* scale ); -lapack_int LAPACKE_cgebal( int matrix_order, char job, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ilo, lapack_int* ihi, float* scale ); -lapack_int LAPACKE_zgebal( int matrix_order, char job, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ilo, lapack_int* ihi, double* scale ); - -lapack_int LAPACKE_sgebrd( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* d, float* e, - float* tauq, float* taup ); -lapack_int LAPACKE_dgebrd( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* d, double* e, - double* tauq, double* taup ); -lapack_int LAPACKE_cgebrd( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, float* d, - float* e, lapack_complex_float* tauq, - lapack_complex_float* taup ); -lapack_int LAPACKE_zgebrd( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, double* d, - double* e, lapack_complex_double* tauq, - lapack_complex_double* taup ); - -lapack_int LAPACKE_sgecon( int matrix_order, char norm, lapack_int n, - const float* a, lapack_int lda, float anorm, - float* rcond ); -lapack_int LAPACKE_dgecon( int matrix_order, char norm, lapack_int n, - const double* a, lapack_int lda, double anorm, - double* rcond ); -lapack_int LAPACKE_cgecon( int matrix_order, char norm, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float anorm, float* rcond ); -lapack_int LAPACKE_zgecon( int matrix_order, char norm, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double anorm, double* rcond ); - -lapack_int LAPACKE_sgeequ( int matrix_order, lapack_int m, lapack_int n, - const float* a, lapack_int lda, float* r, float* c, - float* rowcnd, float* colcnd, float* amax ); -lapack_int LAPACKE_dgeequ( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, double* r, - double* c, double* rowcnd, double* colcnd, - double* amax ); -lapack_int LAPACKE_cgeequ( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* r, float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_zgeequ( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* r, double* c, double* rowcnd, double* colcnd, - double* amax ); - -lapack_int LAPACKE_sgeequb( int matrix_order, lapack_int m, lapack_int n, - const float* a, lapack_int lda, float* r, float* c, - float* rowcnd, float* colcnd, float* amax ); -lapack_int LAPACKE_dgeequb( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, double* r, - double* c, double* rowcnd, double* colcnd, - double* amax ); -lapack_int LAPACKE_cgeequb( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* r, float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_zgeequb( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* r, double* c, double* rowcnd, - double* colcnd, double* amax ); - -lapack_int LAPACKE_sgees( int matrix_order, char jobvs, char sort, - LAPACK_S_SELECT2 select, lapack_int n, float* a, - lapack_int lda, lapack_int* sdim, float* wr, - float* wi, float* vs, lapack_int ldvs ); -lapack_int LAPACKE_dgees( int matrix_order, char jobvs, char sort, - LAPACK_D_SELECT2 select, lapack_int n, double* a, - lapack_int lda, lapack_int* sdim, double* wr, - double* wi, double* vs, lapack_int ldvs ); -lapack_int LAPACKE_cgees( int matrix_order, char jobvs, char sort, - LAPACK_C_SELECT1 select, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* sdim, lapack_complex_float* w, - lapack_complex_float* vs, lapack_int ldvs ); -lapack_int LAPACKE_zgees( int matrix_order, char jobvs, char sort, - LAPACK_Z_SELECT1 select, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* sdim, lapack_complex_double* w, - lapack_complex_double* vs, lapack_int ldvs ); - -lapack_int LAPACKE_sgeesx( int matrix_order, char jobvs, char sort, - LAPACK_S_SELECT2 select, char sense, lapack_int n, - float* a, lapack_int lda, lapack_int* sdim, - float* wr, float* wi, float* vs, lapack_int ldvs, - float* rconde, float* rcondv ); -lapack_int LAPACKE_dgeesx( int matrix_order, char jobvs, char sort, - LAPACK_D_SELECT2 select, char sense, lapack_int n, - double* a, lapack_int lda, lapack_int* sdim, - double* wr, double* wi, double* vs, lapack_int ldvs, - double* rconde, double* rcondv ); -lapack_int LAPACKE_cgeesx( int matrix_order, char jobvs, char sort, - LAPACK_C_SELECT1 select, char sense, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* sdim, lapack_complex_float* w, - lapack_complex_float* vs, lapack_int ldvs, - float* rconde, float* rcondv ); -lapack_int LAPACKE_zgeesx( int matrix_order, char jobvs, char sort, - LAPACK_Z_SELECT1 select, char sense, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* sdim, lapack_complex_double* w, - lapack_complex_double* vs, lapack_int ldvs, - double* rconde, double* rcondv ); - -lapack_int LAPACKE_sgeev( int matrix_order, char jobvl, char jobvr, - lapack_int n, float* a, lapack_int lda, float* wr, - float* wi, float* vl, lapack_int ldvl, float* vr, - lapack_int ldvr ); -lapack_int LAPACKE_dgeev( int matrix_order, char jobvl, char jobvr, - lapack_int n, double* a, lapack_int lda, double* wr, - double* wi, double* vl, lapack_int ldvl, double* vr, - lapack_int ldvr ); -lapack_int LAPACKE_cgeev( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_float* a, lapack_int lda, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int ldvl, lapack_complex_float* vr, - lapack_int ldvr ); -lapack_int LAPACKE_zgeev( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* w, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr ); - -lapack_int LAPACKE_sgeevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, float* a, - lapack_int lda, float* wr, float* wi, float* vl, - lapack_int ldvl, float* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, float* scale, - float* abnrm, float* rconde, float* rcondv ); -lapack_int LAPACKE_dgeevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, double* a, - lapack_int lda, double* wr, double* wi, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, double* scale, - double* abnrm, double* rconde, double* rcondv ); -lapack_int LAPACKE_cgeevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int ldvl, lapack_complex_float* vr, - lapack_int ldvr, lapack_int* ilo, lapack_int* ihi, - float* scale, float* abnrm, float* rconde, - float* rcondv ); -lapack_int LAPACKE_zgeevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* w, lapack_complex_double* vl, - lapack_int ldvl, lapack_complex_double* vr, - lapack_int ldvr, lapack_int* ilo, lapack_int* ihi, - double* scale, double* abnrm, double* rconde, - double* rcondv ); - -lapack_int LAPACKE_sgehrd( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, float* a, lapack_int lda, - float* tau ); -lapack_int LAPACKE_dgehrd( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, double* a, lapack_int lda, - double* tau ); -lapack_int LAPACKE_cgehrd( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* tau ); -lapack_int LAPACKE_zgehrd( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* tau ); - -lapack_int LAPACKE_sgejsv( int matrix_order, char joba, char jobu, char jobv, - char jobr, char jobt, char jobp, lapack_int m, - lapack_int n, float* a, lapack_int lda, float* sva, - float* u, lapack_int ldu, float* v, lapack_int ldv, - float* stat, lapack_int* istat ); -lapack_int LAPACKE_dgejsv( int matrix_order, char joba, char jobu, char jobv, - char jobr, char jobt, char jobp, lapack_int m, - lapack_int n, double* a, lapack_int lda, double* sva, - double* u, lapack_int ldu, double* v, lapack_int ldv, - double* stat, lapack_int* istat ); - -lapack_int LAPACKE_sgelq2( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgelq2( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgelq2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgelq2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgelqf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgelqf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgelqf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgelqf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgels( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* b, lapack_int ldb ); -lapack_int LAPACKE_dgels( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* b, lapack_int ldb ); -lapack_int LAPACKE_cgels( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zgels( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sgelsd( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_dgelsd( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* s, double rcond, - lapack_int* rank ); -lapack_int LAPACKE_cgelsd( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_zgelsd( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* s, double rcond, - lapack_int* rank ); - -lapack_int LAPACKE_sgelss( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_dgelss( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* s, double rcond, - lapack_int* rank ); -lapack_int LAPACKE_cgelss( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_zgelss( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* s, double rcond, - lapack_int* rank ); - -lapack_int LAPACKE_sgelsy( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, float* b, - lapack_int ldb, lapack_int* jpvt, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_dgelsy( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, lapack_int* jpvt, - double rcond, lapack_int* rank ); -lapack_int LAPACKE_cgelsy( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_int* jpvt, float rcond, - lapack_int* rank ); -lapack_int LAPACKE_zgelsy( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_int* jpvt, double rcond, - lapack_int* rank ); - -lapack_int LAPACKE_sgeqlf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgeqlf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgeqlf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqlf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgeqp3( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* jpvt, - float* tau ); -lapack_int LAPACKE_dgeqp3( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* jpvt, - double* tau ); -lapack_int LAPACKE_cgeqp3( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqp3( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_double* tau ); - -lapack_int LAPACKE_sgeqpf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* jpvt, - float* tau ); -lapack_int LAPACKE_dgeqpf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* jpvt, - double* tau ); -lapack_int LAPACKE_cgeqpf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqpf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_double* tau ); - -lapack_int LAPACKE_sgeqr2( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgeqr2( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgeqr2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqr2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgeqrf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgeqrf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgeqrf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqrf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgeqrfp( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgeqrfp( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgeqrfp( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgeqrfp( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgerfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dgerfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr ); -lapack_int LAPACKE_cgerfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zgerfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sgerfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* r, - const float* c, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dgerfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* r, - const double* c, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); -lapack_int LAPACKE_cgerfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* r, - const float* c, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_zgerfsx( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* r, - const double* c, const lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); - -lapack_int LAPACKE_sgerqf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dgerqf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_cgerqf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_zgerqf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_sgesdd( int matrix_order, char jobz, lapack_int m, - lapack_int n, float* a, lapack_int lda, float* s, - float* u, lapack_int ldu, float* vt, - lapack_int ldvt ); -lapack_int LAPACKE_dgesdd( int matrix_order, char jobz, lapack_int m, - lapack_int n, double* a, lapack_int lda, double* s, - double* u, lapack_int ldu, double* vt, - lapack_int ldvt ); -lapack_int LAPACKE_cgesdd( int matrix_order, char jobz, lapack_int m, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float* s, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* vt, - lapack_int ldvt ); -lapack_int LAPACKE_zgesdd( int matrix_order, char jobz, lapack_int m, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double* s, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* vt, - lapack_int ldvt ); - -lapack_int LAPACKE_sgesv( int matrix_order, lapack_int n, lapack_int nrhs, - float* a, lapack_int lda, lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgesv( int matrix_order, lapack_int n, lapack_int nrhs, - double* a, lapack_int lda, lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cgesv( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgesv( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); -lapack_int LAPACKE_dsgesv( int matrix_order, lapack_int n, lapack_int nrhs, - double* a, lapack_int lda, lapack_int* ipiv, - double* b, lapack_int ldb, double* x, lapack_int ldx, - lapack_int* iter ); -lapack_int LAPACKE_zcgesv( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, lapack_int* iter ); - -lapack_int LAPACKE_sgesvd( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, float* a, lapack_int lda, - float* s, float* u, lapack_int ldu, float* vt, - lapack_int ldvt, float* superb ); -lapack_int LAPACKE_dgesvd( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, double* a, - lapack_int lda, double* s, double* u, lapack_int ldu, - double* vt, lapack_int ldvt, double* superb ); -lapack_int LAPACKE_cgesvd( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, lapack_complex_float* a, - lapack_int lda, float* s, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* vt, - lapack_int ldvt, float* superb ); -lapack_int LAPACKE_zgesvd( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, lapack_complex_double* a, - lapack_int lda, double* s, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* vt, - lapack_int ldvt, double* superb ); - -lapack_int LAPACKE_sgesvj( int matrix_order, char joba, char jobu, char jobv, - lapack_int m, lapack_int n, float* a, lapack_int lda, - float* sva, lapack_int mv, float* v, lapack_int ldv, - float* stat ); -lapack_int LAPACKE_dgesvj( int matrix_order, char joba, char jobu, char jobv, - lapack_int m, lapack_int n, double* a, - lapack_int lda, double* sva, lapack_int mv, - double* v, lapack_int ldv, double* stat ); - -lapack_int LAPACKE_sgesvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - float* b, lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* rpivot ); -lapack_int LAPACKE_dgesvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* rpivot ); -lapack_int LAPACKE_cgesvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* rpivot ); -lapack_int LAPACKE_zgesvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* rpivot ); - -lapack_int LAPACKE_sgesvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - float* b, lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dgesvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); -lapack_int LAPACKE_cgesvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_zgesvxx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); - -lapack_int LAPACKE_sgetf2( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_dgetf2( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_cgetf2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zgetf2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgetrf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_dgetrf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_cgetrf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zgetrf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgetri( int matrix_order, lapack_int n, float* a, - lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_dgetri( int matrix_order, lapack_int n, double* a, - lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_cgetri( int matrix_order, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zgetri( int matrix_order, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); - -lapack_int LAPACKE_sgetrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dgetrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_cgetrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zgetrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sggbak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const float* lscale, - const float* rscale, lapack_int m, float* v, - lapack_int ldv ); -lapack_int LAPACKE_dggbak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const double* lscale, - const double* rscale, lapack_int m, double* v, - lapack_int ldv ); -lapack_int LAPACKE_cggbak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const float* lscale, - const float* rscale, lapack_int m, - lapack_complex_float* v, lapack_int ldv ); -lapack_int LAPACKE_zggbak( int matrix_order, char job, char side, lapack_int n, - lapack_int ilo, lapack_int ihi, const double* lscale, - const double* rscale, lapack_int m, - lapack_complex_double* v, lapack_int ldv ); - -lapack_int LAPACKE_sggbal( int matrix_order, char job, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale ); -lapack_int LAPACKE_dggbal( int matrix_order, char job, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, double* lscale, - double* rscale ); -lapack_int LAPACKE_cggbal( int matrix_order, char job, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale ); -lapack_int LAPACKE_zggbal( int matrix_order, char job, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, double* lscale, - double* rscale ); - -lapack_int LAPACKE_sgges( int matrix_order, char jobvsl, char jobvsr, char sort, - LAPACK_S_SELECT3 selctg, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - lapack_int* sdim, float* alphar, float* alphai, - float* beta, float* vsl, lapack_int ldvsl, float* vsr, - lapack_int ldvsr ); -lapack_int LAPACKE_dgges( int matrix_order, char jobvsl, char jobvsr, char sort, - LAPACK_D_SELECT3 selctg, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - lapack_int* sdim, double* alphar, double* alphai, - double* beta, double* vsl, lapack_int ldvsl, - double* vsr, lapack_int ldvsr ); -lapack_int LAPACKE_cgges( int matrix_order, char jobvsl, char jobvsr, char sort, - LAPACK_C_SELECT2 selctg, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_int* sdim, lapack_complex_float* alpha, - lapack_complex_float* beta, lapack_complex_float* vsl, - lapack_int ldvsl, lapack_complex_float* vsr, - lapack_int ldvsr ); -lapack_int LAPACKE_zgges( int matrix_order, char jobvsl, char jobvsr, char sort, - LAPACK_Z_SELECT2 selctg, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_int* sdim, lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int ldvsl, - lapack_complex_double* vsr, lapack_int ldvsr ); - -lapack_int LAPACKE_sggesx( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_S_SELECT3 selctg, char sense, - lapack_int n, float* a, lapack_int lda, float* b, - lapack_int ldb, lapack_int* sdim, float* alphar, - float* alphai, float* beta, float* vsl, - lapack_int ldvsl, float* vsr, lapack_int ldvsr, - float* rconde, float* rcondv ); -lapack_int LAPACKE_dggesx( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_D_SELECT3 selctg, char sense, - lapack_int n, double* a, lapack_int lda, double* b, - lapack_int ldb, lapack_int* sdim, double* alphar, - double* alphai, double* beta, double* vsl, - lapack_int ldvsl, double* vsr, lapack_int ldvsr, - double* rconde, double* rcondv ); -lapack_int LAPACKE_cggesx( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_C_SELECT2 selctg, char sense, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_int* sdim, - lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* vsl, lapack_int ldvsl, - lapack_complex_float* vsr, lapack_int ldvsr, - float* rconde, float* rcondv ); -lapack_int LAPACKE_zggesx( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_Z_SELECT2 selctg, char sense, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_int* sdim, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int ldvsl, - lapack_complex_double* vsr, lapack_int ldvsr, - double* rconde, double* rcondv ); - -lapack_int LAPACKE_sggev( int matrix_order, char jobvl, char jobvr, - lapack_int n, float* a, lapack_int lda, float* b, - lapack_int ldb, float* alphar, float* alphai, - float* beta, float* vl, lapack_int ldvl, float* vr, - lapack_int ldvr ); -lapack_int LAPACKE_dggev( int matrix_order, char jobvl, char jobvr, - lapack_int n, double* a, lapack_int lda, double* b, - lapack_int ldb, double* alphar, double* alphai, - double* beta, double* vl, lapack_int ldvl, double* vr, - lapack_int ldvr ); -lapack_int LAPACKE_cggev( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* alpha, - lapack_complex_float* beta, lapack_complex_float* vl, - lapack_int ldvl, lapack_complex_float* vr, - lapack_int ldvr ); -lapack_int LAPACKE_zggev( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr ); - -lapack_int LAPACKE_sggevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* alphar, float* alphai, float* beta, float* vl, - lapack_int ldvl, float* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale, float* abnrm, float* bbnrm, - float* rconde, float* rcondv ); -lapack_int LAPACKE_dggevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* alphar, double* alphai, double* beta, - double* vl, lapack_int ldvl, double* vr, - lapack_int ldvr, lapack_int* ilo, lapack_int* ihi, - double* lscale, double* rscale, double* abnrm, - double* bbnrm, double* rconde, double* rcondv ); -lapack_int LAPACKE_cggevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* alpha, - lapack_complex_float* beta, lapack_complex_float* vl, - lapack_int ldvl, lapack_complex_float* vr, - lapack_int ldvr, lapack_int* ilo, lapack_int* ihi, - float* lscale, float* rscale, float* abnrm, - float* bbnrm, float* rconde, float* rcondv ); -lapack_int LAPACKE_zggevx( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, double* lscale, - double* rscale, double* abnrm, double* bbnrm, - double* rconde, double* rcondv ); - -lapack_int LAPACKE_sggglm( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, float* a, lapack_int lda, float* b, - lapack_int ldb, float* d, float* x, float* y ); -lapack_int LAPACKE_dggglm( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, double* a, lapack_int lda, double* b, - lapack_int ldb, double* d, double* x, double* y ); -lapack_int LAPACKE_cggglm( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* d, - lapack_complex_float* x, lapack_complex_float* y ); -lapack_int LAPACKE_zggglm( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* d, - lapack_complex_double* x, lapack_complex_double* y ); - -lapack_int LAPACKE_sgghrd( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - float* a, lapack_int lda, float* b, lapack_int ldb, - float* q, lapack_int ldq, float* z, lapack_int ldz ); -lapack_int LAPACKE_dgghrd( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - double* a, lapack_int lda, double* b, lapack_int ldb, - double* q, lapack_int ldq, double* z, - lapack_int ldz ); -lapack_int LAPACKE_cgghrd( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zgghrd( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_sgglse( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, float* a, lapack_int lda, float* b, - lapack_int ldb, float* c, float* d, float* x ); -lapack_int LAPACKE_dgglse( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, double* a, lapack_int lda, double* b, - lapack_int ldb, double* c, double* d, double* x ); -lapack_int LAPACKE_cgglse( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* c, - lapack_complex_float* d, lapack_complex_float* x ); -lapack_int LAPACKE_zgglse( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* c, - lapack_complex_double* d, lapack_complex_double* x ); - -lapack_int LAPACKE_sggqrf( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, float* a, lapack_int lda, float* taua, - float* b, lapack_int ldb, float* taub ); -lapack_int LAPACKE_dggqrf( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, double* a, lapack_int lda, - double* taua, double* b, lapack_int ldb, - double* taub ); -lapack_int LAPACKE_cggqrf( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* taua, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* taub ); -lapack_int LAPACKE_zggqrf( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* taua, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* taub ); - -lapack_int LAPACKE_sggrqf( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, float* a, lapack_int lda, float* taua, - float* b, lapack_int ldb, float* taub ); -lapack_int LAPACKE_dggrqf( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, double* a, lapack_int lda, - double* taua, double* b, lapack_int ldb, - double* taub ); -lapack_int LAPACKE_cggrqf( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* taua, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* taub ); -lapack_int LAPACKE_zggrqf( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* taua, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* taub ); - -lapack_int LAPACKE_sggsvd( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int n, lapack_int p, - lapack_int* k, lapack_int* l, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* alpha, float* beta, float* u, lapack_int ldu, - float* v, lapack_int ldv, float* q, lapack_int ldq, - lapack_int* iwork ); -lapack_int LAPACKE_dggsvd( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int n, lapack_int p, - lapack_int* k, lapack_int* l, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* alpha, double* beta, double* u, - lapack_int ldu, double* v, lapack_int ldv, double* q, - lapack_int ldq, lapack_int* iwork ); -lapack_int LAPACKE_cggsvd( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int n, lapack_int p, - lapack_int* k, lapack_int* l, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - float* alpha, float* beta, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* v, - lapack_int ldv, lapack_complex_float* q, - lapack_int ldq, lapack_int* iwork ); -lapack_int LAPACKE_zggsvd( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int n, lapack_int p, - lapack_int* k, lapack_int* l, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double* alpha, double* beta, - lapack_complex_double* u, lapack_int ldu, - lapack_complex_double* v, lapack_int ldv, - lapack_complex_double* q, lapack_int ldq, - lapack_int* iwork ); - -lapack_int LAPACKE_sggsvp( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, float tola, - float tolb, lapack_int* k, lapack_int* l, float* u, - lapack_int ldu, float* v, lapack_int ldv, float* q, - lapack_int ldq ); -lapack_int LAPACKE_dggsvp( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double tola, double tolb, lapack_int* k, - lapack_int* l, double* u, lapack_int ldu, double* v, - lapack_int ldv, double* q, lapack_int ldq ); -lapack_int LAPACKE_cggsvp( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, float tola, - float tolb, lapack_int* k, lapack_int* l, - lapack_complex_float* u, lapack_int ldu, - lapack_complex_float* v, lapack_int ldv, - lapack_complex_float* q, lapack_int ldq ); -lapack_int LAPACKE_zggsvp( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double tola, double tolb, lapack_int* k, - lapack_int* l, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* v, - lapack_int ldv, lapack_complex_double* q, - lapack_int ldq ); - -lapack_int LAPACKE_sgtcon( char norm, lapack_int n, const float* dl, - const float* d, const float* du, const float* du2, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_dgtcon( char norm, lapack_int n, const double* dl, - const double* d, const double* du, const double* du2, - const lapack_int* ipiv, double anorm, - double* rcond ); -lapack_int LAPACKE_cgtcon( char norm, lapack_int n, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zgtcon( char norm, lapack_int n, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_sgtrfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* dl, const float* d, - const float* du, const float* dlf, const float* df, - const float* duf, const float* du2, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dgtrfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* dl, const double* d, - const double* du, const double* dlf, - const double* df, const double* duf, - const double* du2, const lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr ); -lapack_int LAPACKE_cgtrfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* dlf, - const lapack_complex_float* df, - const lapack_complex_float* duf, - const lapack_complex_float* du2, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zgtrfs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* dlf, - const lapack_complex_double* df, - const lapack_complex_double* duf, - const lapack_complex_double* du2, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sgtsv( int matrix_order, lapack_int n, lapack_int nrhs, - float* dl, float* d, float* du, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgtsv( int matrix_order, lapack_int n, lapack_int nrhs, - double* dl, double* d, double* du, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cgtsv( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_float* dl, lapack_complex_float* d, - lapack_complex_float* du, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgtsv( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* dl, lapack_complex_double* d, - lapack_complex_double* du, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sgtsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, const float* dl, - const float* d, const float* du, float* dlf, - float* df, float* duf, float* du2, lapack_int* ipiv, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dgtsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, const double* dl, - const double* d, const double* du, double* dlf, - double* df, double* duf, double* du2, - lapack_int* ipiv, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* ferr, double* berr ); -lapack_int LAPACKE_cgtsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - lapack_complex_float* dlf, lapack_complex_float* df, - lapack_complex_float* duf, lapack_complex_float* du2, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_zgtsvx( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - lapack_complex_double* dlf, - lapack_complex_double* df, - lapack_complex_double* duf, - lapack_complex_double* du2, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_sgttrf( lapack_int n, float* dl, float* d, float* du, - float* du2, lapack_int* ipiv ); -lapack_int LAPACKE_dgttrf( lapack_int n, double* dl, double* d, double* du, - double* du2, lapack_int* ipiv ); -lapack_int LAPACKE_cgttrf( lapack_int n, lapack_complex_float* dl, - lapack_complex_float* d, lapack_complex_float* du, - lapack_complex_float* du2, lapack_int* ipiv ); -lapack_int LAPACKE_zgttrf( lapack_int n, lapack_complex_double* dl, - lapack_complex_double* d, lapack_complex_double* du, - lapack_complex_double* du2, lapack_int* ipiv ); - -lapack_int LAPACKE_sgttrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* dl, const float* d, - const float* du, const float* du2, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dgttrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* dl, const double* d, - const double* du, const double* du2, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_cgttrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgttrs( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_chbev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, lapack_complex_float* ab, - lapack_int ldab, float* w, lapack_complex_float* z, - lapack_int ldz ); -lapack_int LAPACKE_zhbev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, lapack_complex_double* ab, - lapack_int ldab, double* w, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_chbevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, lapack_complex_float* ab, - lapack_int ldab, float* w, lapack_complex_float* z, - lapack_int ldz ); -lapack_int LAPACKE_zhbevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, lapack_complex_double* ab, - lapack_int ldab, double* w, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_chbevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* q, lapack_int ldq, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* ifail ); -lapack_int LAPACKE_zhbevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* q, lapack_int ldq, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_chbgst( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* bb, lapack_int ldbb, - lapack_complex_float* x, lapack_int ldx ); -lapack_int LAPACKE_zhbgst( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - const lapack_complex_double* bb, lapack_int ldbb, - lapack_complex_double* x, lapack_int ldx ); - -lapack_int LAPACKE_chbgv( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* bb, lapack_int ldbb, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhbgv( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* bb, lapack_int ldbb, double* w, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chbgvd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* bb, lapack_int ldbb, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhbgvd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* bb, lapack_int ldbb, - double* w, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_chbgvx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* bb, lapack_int ldbb, - lapack_complex_float* q, lapack_int ldq, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* ifail ); -lapack_int LAPACKE_zhbgvx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* bb, lapack_int ldbb, - lapack_complex_double* q, lapack_int ldq, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_chbtrd( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int kd, lapack_complex_float* ab, - lapack_int ldab, float* d, float* e, - lapack_complex_float* q, lapack_int ldq ); -lapack_int LAPACKE_zhbtrd( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int kd, lapack_complex_double* ab, - lapack_int ldab, double* d, double* e, - lapack_complex_double* q, lapack_int ldq ); - -lapack_int LAPACKE_checon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zhecon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_cheequb( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zheequb( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_cheev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, float* w ); -lapack_int LAPACKE_zheev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, double* w ); - -lapack_int LAPACKE_cheevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, float* w ); -lapack_int LAPACKE_zheevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double* w ); - -lapack_int LAPACKE_cheevr( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_int* isuppz ); -lapack_int LAPACKE_zheevr( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, lapack_int ldz, - lapack_int* isuppz ); - -lapack_int LAPACKE_cheevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_zheevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_chegst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhegst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_chegv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* w ); -lapack_int LAPACKE_zhegv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* w ); - -lapack_int LAPACKE_chegvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* w ); -lapack_int LAPACKE_zhegvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* w ); - -lapack_int LAPACKE_chegvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* ifail ); -lapack_int LAPACKE_zhegvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_cherfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zherfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_cherfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* s, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_zherfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* s, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); - -lapack_int LAPACKE_chesv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zhesv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_chesvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* af, - lapack_int ldaf, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zhesvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* af, - lapack_int ldaf, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_chesvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_zhesvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); - -lapack_int LAPACKE_chetrd( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, float* d, - float* e, lapack_complex_float* tau ); -lapack_int LAPACKE_zhetrd( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, double* d, - double* e, lapack_complex_double* tau ); - -lapack_int LAPACKE_chetrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zhetrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_chetri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zhetri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); - -lapack_int LAPACKE_chetrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zhetrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_chfrk( int matrix_order, char transr, char uplo, char trans, - lapack_int n, lapack_int k, float alpha, - const lapack_complex_float* a, lapack_int lda, - float beta, lapack_complex_float* c ); -lapack_int LAPACKE_zhfrk( int matrix_order, char transr, char uplo, char trans, - lapack_int n, lapack_int k, double alpha, - const lapack_complex_double* a, lapack_int lda, - double beta, lapack_complex_double* c ); - -lapack_int LAPACKE_shgeqz( int matrix_order, char job, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - float* h, lapack_int ldh, float* t, lapack_int ldt, - float* alphar, float* alphai, float* beta, float* q, - lapack_int ldq, float* z, lapack_int ldz ); -lapack_int LAPACKE_dhgeqz( int matrix_order, char job, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - double* h, lapack_int ldh, double* t, lapack_int ldt, - double* alphar, double* alphai, double* beta, - double* q, lapack_int ldq, double* z, - lapack_int ldz ); -lapack_int LAPACKE_chgeqz( int matrix_order, char job, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_float* h, lapack_int ldh, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* alpha, - lapack_complex_float* beta, lapack_complex_float* q, - lapack_int ldq, lapack_complex_float* z, - lapack_int ldz ); -lapack_int LAPACKE_zhgeqz( int matrix_order, char job, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_double* h, lapack_int ldh, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chpcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zhpcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_chpev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_float* ap, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhpev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_double* ap, double* w, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chpevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_float* ap, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhpevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_complex_double* ap, double* w, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chpevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_float* ap, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* ifail ); -lapack_int LAPACKE_zhpevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_complex_double* ap, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_chpgst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_float* ap, - const lapack_complex_float* bp ); -lapack_int LAPACKE_zhpgst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_double* ap, - const lapack_complex_double* bp ); - -lapack_int LAPACKE_chpgv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_float* ap, - lapack_complex_float* bp, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhpgv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_double* ap, - lapack_complex_double* bp, double* w, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chpgvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_float* ap, - lapack_complex_float* bp, float* w, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zhpgvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_double* ap, - lapack_complex_double* bp, double* w, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_chpgvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_float* ap, lapack_complex_float* bp, - float vl, float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_zhpgvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_double* ap, lapack_complex_double* bp, - double vl, double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_chprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zhprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_chpsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhpsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_chpsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - lapack_complex_float* afp, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zhpsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - lapack_complex_double* afp, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_chptrd( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, float* d, float* e, - lapack_complex_float* tau ); -lapack_int LAPACKE_zhptrd( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, double* d, double* e, - lapack_complex_double* tau ); - -lapack_int LAPACKE_chptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, lapack_int* ipiv ); -lapack_int LAPACKE_zhptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, lapack_int* ipiv ); - -lapack_int LAPACKE_chptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, const lapack_int* ipiv ); -lapack_int LAPACKE_zhptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, const lapack_int* ipiv ); - -lapack_int LAPACKE_chptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_shsein( int matrix_order, char job, char eigsrc, char initv, - lapack_logical* select, lapack_int n, const float* h, - lapack_int ldh, float* wr, const float* wi, - float* vl, lapack_int ldvl, float* vr, - lapack_int ldvr, lapack_int mm, lapack_int* m, - lapack_int* ifaill, lapack_int* ifailr ); -lapack_int LAPACKE_dhsein( int matrix_order, char job, char eigsrc, char initv, - lapack_logical* select, lapack_int n, - const double* h, lapack_int ldh, double* wr, - const double* wi, double* vl, lapack_int ldvl, - double* vr, lapack_int ldvr, lapack_int mm, - lapack_int* m, lapack_int* ifaill, - lapack_int* ifailr ); -lapack_int LAPACKE_chsein( int matrix_order, char job, char eigsrc, char initv, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* h, lapack_int ldh, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int ldvl, lapack_complex_float* vr, - lapack_int ldvr, lapack_int mm, lapack_int* m, - lapack_int* ifaill, lapack_int* ifailr ); -lapack_int LAPACKE_zhsein( int matrix_order, char job, char eigsrc, char initv, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* h, lapack_int ldh, - lapack_complex_double* w, lapack_complex_double* vl, - lapack_int ldvl, lapack_complex_double* vr, - lapack_int ldvr, lapack_int mm, lapack_int* m, - lapack_int* ifaill, lapack_int* ifailr ); - -lapack_int LAPACKE_shseqr( int matrix_order, char job, char compz, lapack_int n, - lapack_int ilo, lapack_int ihi, float* h, - lapack_int ldh, float* wr, float* wi, float* z, - lapack_int ldz ); -lapack_int LAPACKE_dhseqr( int matrix_order, char job, char compz, lapack_int n, - lapack_int ilo, lapack_int ihi, double* h, - lapack_int ldh, double* wr, double* wi, double* z, - lapack_int ldz ); -lapack_int LAPACKE_chseqr( int matrix_order, char job, char compz, lapack_int n, - lapack_int ilo, lapack_int ihi, - lapack_complex_float* h, lapack_int ldh, - lapack_complex_float* w, lapack_complex_float* z, - lapack_int ldz ); -lapack_int LAPACKE_zhseqr( int matrix_order, char job, char compz, lapack_int n, - lapack_int ilo, lapack_int ihi, - lapack_complex_double* h, lapack_int ldh, - lapack_complex_double* w, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_clacgv( lapack_int n, lapack_complex_float* x, - lapack_int incx ); -lapack_int LAPACKE_zlacgv( lapack_int n, lapack_complex_double* x, - lapack_int incx ); - -lapack_int LAPACKE_slacpy( int matrix_order, char uplo, lapack_int m, - lapack_int n, const float* a, lapack_int lda, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dlacpy( int matrix_order, char uplo, lapack_int m, - lapack_int n, const double* a, lapack_int lda, double* b, - lapack_int ldb ); -lapack_int LAPACKE_clacpy( int matrix_order, char uplo, lapack_int m, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zlacpy( int matrix_order, char uplo, lapack_int m, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_zlag2c( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_float* sa, lapack_int ldsa ); - -lapack_int LAPACKE_slag2d( int matrix_order, lapack_int m, lapack_int n, - const float* sa, lapack_int ldsa, double* a, - lapack_int lda ); - -lapack_int LAPACKE_dlag2s( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, float* sa, - lapack_int ldsa ); - -lapack_int LAPACKE_clag2z( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* sa, lapack_int ldsa, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_slagge( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* d, - float* a, lapack_int lda, lapack_int* iseed ); -lapack_int LAPACKE_dlagge( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* d, - double* a, lapack_int lda, lapack_int* iseed ); -lapack_int LAPACKE_clagge( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* d, - lapack_complex_float* a, lapack_int lda, - lapack_int* iseed ); -lapack_int LAPACKE_zlagge( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* d, - lapack_complex_double* a, lapack_int lda, - lapack_int* iseed ); - -float LAPACKE_slamch( char cmach ); -double LAPACKE_dlamch( char cmach ); - -float LAPACKE_slange( int matrix_order, char norm, lapack_int m, - lapack_int n, const float* a, lapack_int lda ); -double LAPACKE_dlange( int matrix_order, char norm, lapack_int m, - lapack_int n, const double* a, lapack_int lda ); -float LAPACKE_clange( int matrix_order, char norm, lapack_int m, - lapack_int n, const lapack_complex_float* a, - lapack_int lda ); -double LAPACKE_zlange( int matrix_order, char norm, lapack_int m, - lapack_int n, const lapack_complex_double* a, - lapack_int lda ); - -float LAPACKE_clanhe( int matrix_order, char norm, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda ); -double LAPACKE_zlanhe( int matrix_order, char norm, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda ); - -float LAPACKE_slansy( int matrix_order, char norm, char uplo, lapack_int n, - const float* a, lapack_int lda ); -double LAPACKE_dlansy( int matrix_order, char norm, char uplo, lapack_int n, - const double* a, lapack_int lda ); -float LAPACKE_clansy( int matrix_order, char norm, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda ); -double LAPACKE_zlansy( int matrix_order, char norm, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda ); - -float LAPACKE_slantr( int matrix_order, char norm, char uplo, char diag, - lapack_int m, lapack_int n, const float* a, - lapack_int lda ); -double LAPACKE_dlantr( int matrix_order, char norm, char uplo, char diag, - lapack_int m, lapack_int n, const double* a, - lapack_int lda ); -float LAPACKE_clantr( int matrix_order, char norm, char uplo, char diag, - lapack_int m, lapack_int n, const lapack_complex_float* a, - lapack_int lda ); -double LAPACKE_zlantr( int matrix_order, char norm, char uplo, char diag, - lapack_int m, lapack_int n, const lapack_complex_double* a, - lapack_int lda ); - - -lapack_int LAPACKE_slarfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, const float* v, lapack_int ldv, - const float* t, lapack_int ldt, float* c, - lapack_int ldc ); -lapack_int LAPACKE_dlarfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, const double* v, lapack_int ldv, - const double* t, lapack_int ldt, double* c, - lapack_int ldc ); -lapack_int LAPACKE_clarfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, const lapack_complex_float* v, - lapack_int ldv, const lapack_complex_float* t, - lapack_int ldt, lapack_complex_float* c, - lapack_int ldc ); -lapack_int LAPACKE_zlarfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, const lapack_complex_double* v, - lapack_int ldv, const lapack_complex_double* t, - lapack_int ldt, lapack_complex_double* c, - lapack_int ldc ); - -lapack_int LAPACKE_slarfg( lapack_int n, float* alpha, float* x, - lapack_int incx, float* tau ); -lapack_int LAPACKE_dlarfg( lapack_int n, double* alpha, double* x, - lapack_int incx, double* tau ); -lapack_int LAPACKE_clarfg( lapack_int n, lapack_complex_float* alpha, - lapack_complex_float* x, lapack_int incx, - lapack_complex_float* tau ); -lapack_int LAPACKE_zlarfg( lapack_int n, lapack_complex_double* alpha, - lapack_complex_double* x, lapack_int incx, - lapack_complex_double* tau ); - -lapack_int LAPACKE_slarft( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, const float* v, - lapack_int ldv, const float* tau, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dlarft( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, const double* v, - lapack_int ldv, const double* tau, double* t, - lapack_int ldt ); -lapack_int LAPACKE_clarft( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* tau, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zlarft( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* tau, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_slarfx( int matrix_order, char side, lapack_int m, - lapack_int n, const float* v, float tau, float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_dlarfx( int matrix_order, char side, lapack_int m, - lapack_int n, const double* v, double tau, double* c, - lapack_int ldc, double* work ); -lapack_int LAPACKE_clarfx( int matrix_order, char side, lapack_int m, - lapack_int n, const lapack_complex_float* v, - lapack_complex_float tau, lapack_complex_float* c, - lapack_int ldc, lapack_complex_float* work ); -lapack_int LAPACKE_zlarfx( int matrix_order, char side, lapack_int m, - lapack_int n, const lapack_complex_double* v, - lapack_complex_double tau, lapack_complex_double* c, - lapack_int ldc, lapack_complex_double* work ); - -lapack_int LAPACKE_slarnv( lapack_int idist, lapack_int* iseed, lapack_int n, - float* x ); -lapack_int LAPACKE_dlarnv( lapack_int idist, lapack_int* iseed, lapack_int n, - double* x ); -lapack_int LAPACKE_clarnv( lapack_int idist, lapack_int* iseed, lapack_int n, - lapack_complex_float* x ); -lapack_int LAPACKE_zlarnv( lapack_int idist, lapack_int* iseed, lapack_int n, - lapack_complex_double* x ); - -lapack_int LAPACKE_slaset( int matrix_order, char uplo, lapack_int m, - lapack_int n, float alpha, float beta, float* a, - lapack_int lda ); -lapack_int LAPACKE_dlaset( int matrix_order, char uplo, lapack_int m, - lapack_int n, double alpha, double beta, double* a, - lapack_int lda ); -lapack_int LAPACKE_claset( int matrix_order, char uplo, lapack_int m, - lapack_int n, lapack_complex_float alpha, - lapack_complex_float beta, lapack_complex_float* a, - lapack_int lda ); -lapack_int LAPACKE_zlaset( int matrix_order, char uplo, lapack_int m, - lapack_int n, lapack_complex_double alpha, - lapack_complex_double beta, lapack_complex_double* a, - lapack_int lda ); - -lapack_int LAPACKE_slasrt( char id, lapack_int n, float* d ); -lapack_int LAPACKE_dlasrt( char id, lapack_int n, double* d ); - -lapack_int LAPACKE_slaswp( int matrix_order, lapack_int n, float* a, - lapack_int lda, lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); -lapack_int LAPACKE_dlaswp( int matrix_order, lapack_int n, double* a, - lapack_int lda, lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); -lapack_int LAPACKE_claswp( int matrix_order, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int k1, lapack_int k2, const lapack_int* ipiv, - lapack_int incx ); -lapack_int LAPACKE_zlaswp( int matrix_order, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int k1, lapack_int k2, const lapack_int* ipiv, - lapack_int incx ); - -lapack_int LAPACKE_slatms( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, float* d, - lapack_int mode, float cond, float dmax, - lapack_int kl, lapack_int ku, char pack, float* a, - lapack_int lda ); -lapack_int LAPACKE_dlatms( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, double* d, - lapack_int mode, double cond, double dmax, - lapack_int kl, lapack_int ku, char pack, double* a, - lapack_int lda ); -lapack_int LAPACKE_clatms( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, float* d, - lapack_int mode, float cond, float dmax, - lapack_int kl, lapack_int ku, char pack, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zlatms( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, double* d, - lapack_int mode, double cond, double dmax, - lapack_int kl, lapack_int ku, char pack, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_slauum( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda ); -lapack_int LAPACKE_dlauum( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda ); -lapack_int LAPACKE_clauum( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zlauum( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_sopgtr( int matrix_order, char uplo, lapack_int n, - const float* ap, const float* tau, float* q, - lapack_int ldq ); -lapack_int LAPACKE_dopgtr( int matrix_order, char uplo, lapack_int n, - const double* ap, const double* tau, double* q, - lapack_int ldq ); - -lapack_int LAPACKE_sopmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, const float* ap, - const float* tau, float* c, lapack_int ldc ); -lapack_int LAPACKE_dopmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, const double* ap, - const double* tau, double* c, lapack_int ldc ); - -lapack_int LAPACKE_sorgbr( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorgbr( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, double* a, - lapack_int lda, const double* tau ); - -lapack_int LAPACKE_sorghr( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorghr( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, double* a, lapack_int lda, - const double* tau ); - -lapack_int LAPACKE_sorglq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorglq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau ); - -lapack_int LAPACKE_sorgql( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorgql( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau ); - -lapack_int LAPACKE_sorgqr( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorgqr( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau ); - -lapack_int LAPACKE_sorgrq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau ); -lapack_int LAPACKE_dorgrq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau ); - -lapack_int LAPACKE_sorgtr( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, const float* tau ); -lapack_int LAPACKE_dorgtr( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda, const double* tau ); - -lapack_int LAPACKE_sormbr( int matrix_order, char vect, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, const float* tau, - float* c, lapack_int ldc ); -lapack_int LAPACKE_dormbr( int matrix_order, char vect, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, const double* tau, - double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormhr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc ); -lapack_int LAPACKE_dormhr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormlq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, const float* tau, - float* c, lapack_int ldc ); -lapack_int LAPACKE_dormlq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, const double* tau, - double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormql( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, const float* tau, - float* c, lapack_int ldc ); -lapack_int LAPACKE_dormql( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, const double* tau, - double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormqr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, const float* tau, - float* c, lapack_int ldc ); -lapack_int LAPACKE_dormqr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, const double* tau, - double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormrq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, const float* tau, - float* c, lapack_int ldc ); -lapack_int LAPACKE_dormrq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, const double* tau, - double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormrz( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc ); -lapack_int LAPACKE_dormrz( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc ); - -lapack_int LAPACKE_sormtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, const float* a, - lapack_int lda, const float* tau, float* c, - lapack_int ldc ); -lapack_int LAPACKE_dormtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, const double* a, - lapack_int lda, const double* tau, double* c, - lapack_int ldc ); - -lapack_int LAPACKE_spbcon( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const float* ab, lapack_int ldab, - float anorm, float* rcond ); -lapack_int LAPACKE_dpbcon( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const double* ab, lapack_int ldab, - double anorm, double* rcond ); -lapack_int LAPACKE_cpbcon( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_float* ab, - lapack_int ldab, float anorm, float* rcond ); -lapack_int LAPACKE_zpbcon( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_double* ab, - lapack_int ldab, double anorm, double* rcond ); - -lapack_int LAPACKE_spbequ( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const float* ab, lapack_int ldab, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_dpbequ( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const double* ab, lapack_int ldab, - double* s, double* scond, double* amax ); -lapack_int LAPACKE_cpbequ( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_float* ab, - lapack_int ldab, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_zpbequ( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_double* ab, - lapack_int ldab, double* s, double* scond, - double* amax ); - -lapack_int LAPACKE_spbrfs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const float* ab, - lapack_int ldab, const float* afb, lapack_int ldafb, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_dpbrfs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const double* ab, - lapack_int ldab, const double* afb, lapack_int ldafb, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr ); -lapack_int LAPACKE_cpbrfs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* afb, lapack_int ldafb, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zpbrfs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - const lapack_complex_double* afb, lapack_int ldafb, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_spbstf( int matrix_order, char uplo, lapack_int n, - lapack_int kb, float* bb, lapack_int ldbb ); -lapack_int LAPACKE_dpbstf( int matrix_order, char uplo, lapack_int n, - lapack_int kb, double* bb, lapack_int ldbb ); -lapack_int LAPACKE_cpbstf( int matrix_order, char uplo, lapack_int n, - lapack_int kb, lapack_complex_float* bb, - lapack_int ldbb ); -lapack_int LAPACKE_zpbstf( int matrix_order, char uplo, lapack_int n, - lapack_int kb, lapack_complex_double* bb, - lapack_int ldbb ); - -lapack_int LAPACKE_spbsv( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, float* ab, - lapack_int ldab, float* b, lapack_int ldb ); -lapack_int LAPACKE_dpbsv( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, double* ab, - lapack_int ldab, double* b, lapack_int ldb ); -lapack_int LAPACKE_cpbsv( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpbsv( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spbsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, float* ab, - lapack_int ldab, float* afb, lapack_int ldafb, - char* equed, float* s, float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dpbsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, double* ab, - lapack_int ldab, double* afb, lapack_int ldafb, - char* equed, double* s, double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* ferr, double* berr ); -lapack_int LAPACKE_cpbsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* afb, lapack_int ldafb, - char* equed, float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_zpbsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* afb, lapack_int ldafb, - char* equed, double* s, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr ); - -lapack_int LAPACKE_spbtrf( int matrix_order, char uplo, lapack_int n, - lapack_int kd, float* ab, lapack_int ldab ); -lapack_int LAPACKE_dpbtrf( int matrix_order, char uplo, lapack_int n, - lapack_int kd, double* ab, lapack_int ldab ); -lapack_int LAPACKE_cpbtrf( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_complex_float* ab, - lapack_int ldab ); -lapack_int LAPACKE_zpbtrf( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_complex_double* ab, - lapack_int ldab ); - -lapack_int LAPACKE_spbtrs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const float* ab, - lapack_int ldab, float* b, lapack_int ldb ); -lapack_int LAPACKE_dpbtrs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const double* ab, - lapack_int ldab, double* b, lapack_int ldb ); -lapack_int LAPACKE_cpbtrs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpbtrs( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spftrf( int matrix_order, char transr, char uplo, - lapack_int n, float* a ); -lapack_int LAPACKE_dpftrf( int matrix_order, char transr, char uplo, - lapack_int n, double* a ); -lapack_int LAPACKE_cpftrf( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_float* a ); -lapack_int LAPACKE_zpftrf( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_double* a ); - -lapack_int LAPACKE_spftri( int matrix_order, char transr, char uplo, - lapack_int n, float* a ); -lapack_int LAPACKE_dpftri( int matrix_order, char transr, char uplo, - lapack_int n, double* a ); -lapack_int LAPACKE_cpftri( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_float* a ); -lapack_int LAPACKE_zpftri( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_double* a ); - -lapack_int LAPACKE_spftrs( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, const float* a, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dpftrs( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, const double* a, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cpftrs( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpftrs( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spocon( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float anorm, - float* rcond ); -lapack_int LAPACKE_dpocon( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double anorm, - double* rcond ); -lapack_int LAPACKE_cpocon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float anorm, float* rcond ); -lapack_int LAPACKE_zpocon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double anorm, double* rcond ); - -lapack_int LAPACKE_spoequ( int matrix_order, lapack_int n, const float* a, - lapack_int lda, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dpoequ( int matrix_order, lapack_int n, const double* a, - lapack_int lda, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cpoequ( int matrix_order, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zpoequ( int matrix_order, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_spoequb( int matrix_order, lapack_int n, const float* a, - lapack_int lda, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dpoequb( int matrix_order, lapack_int n, const double* a, - lapack_int lda, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cpoequb( int matrix_order, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zpoequb( int matrix_order, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_sporfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dporfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const double* af, lapack_int ldaf, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr ); -lapack_int LAPACKE_cporfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_zporfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* ferr, double* berr ); - -lapack_int LAPACKE_sporfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, lapack_int ldaf, - const float* s, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dporfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, lapack_int ldaf, - const double* s, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); -lapack_int LAPACKE_cporfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, lapack_int ldaf, - const float* s, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_zporfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, lapack_int ldaf, - const double* s, const lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); - -lapack_int LAPACKE_sposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); -lapack_int LAPACKE_dsposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* x, lapack_int ldx, - lapack_int* iter ); -lapack_int LAPACKE_zcposv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, lapack_int* iter ); - -lapack_int LAPACKE_sposvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, float* af, - lapack_int ldaf, char* equed, float* s, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_dposvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* af, lapack_int ldaf, char* equed, double* s, - double* b, lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); -lapack_int LAPACKE_cposvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* af, - lapack_int ldaf, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zposvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* af, - lapack_int ldaf, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_sposvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - char* equed, float* s, float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_dposvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - char* equed, double* s, double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); -lapack_int LAPACKE_cposvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - char* equed, float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_zposvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - char* equed, double* s, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); - -lapack_int LAPACKE_spotrf( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda ); -lapack_int LAPACKE_dpotrf( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda ); -lapack_int LAPACKE_cpotrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zpotrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_spotri( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda ); -lapack_int LAPACKE_dpotri( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda ); -lapack_int LAPACKE_cpotri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zpotri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_spotrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dpotrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cpotrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zpotrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sppcon( int matrix_order, char uplo, lapack_int n, - const float* ap, float anorm, float* rcond ); -lapack_int LAPACKE_dppcon( int matrix_order, char uplo, lapack_int n, - const double* ap, double anorm, double* rcond ); -lapack_int LAPACKE_cppcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, float anorm, - float* rcond ); -lapack_int LAPACKE_zppcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, double anorm, - double* rcond ); - -lapack_int LAPACKE_sppequ( int matrix_order, char uplo, lapack_int n, - const float* ap, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dppequ( int matrix_order, char uplo, lapack_int n, - const double* ap, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cppequ( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, float* s, - float* scond, float* amax ); -lapack_int LAPACKE_zppequ( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, double* s, - double* scond, double* amax ); - -lapack_int LAPACKE_spprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, const float* afp, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_dpprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, const double* afp, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr ); -lapack_int LAPACKE_cpprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zpprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sppsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* ap, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dppsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* ap, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cppsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zppsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sppsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, float* ap, float* afp, char* equed, - float* s, float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dppsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, double* ap, double* afp, - char* equed, double* s, double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* ferr, double* berr ); -lapack_int LAPACKE_cppsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_complex_float* afp, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zppsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_complex_double* afp, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_spptrf( int matrix_order, char uplo, lapack_int n, - float* ap ); -lapack_int LAPACKE_dpptrf( int matrix_order, char uplo, lapack_int n, - double* ap ); -lapack_int LAPACKE_cpptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap ); -lapack_int LAPACKE_zpptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap ); - -lapack_int LAPACKE_spptri( int matrix_order, char uplo, lapack_int n, - float* ap ); -lapack_int LAPACKE_dpptri( int matrix_order, char uplo, lapack_int n, - double* ap ); -lapack_int LAPACKE_cpptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap ); -lapack_int LAPACKE_zpptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap ); - -lapack_int LAPACKE_spptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dpptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cpptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spstrf( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, lapack_int* piv, lapack_int* rank, - float tol ); -lapack_int LAPACKE_dpstrf( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda, lapack_int* piv, lapack_int* rank, - double tol ); -lapack_int LAPACKE_cpstrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* piv, lapack_int* rank, float tol ); -lapack_int LAPACKE_zpstrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* piv, lapack_int* rank, double tol ); - -lapack_int LAPACKE_sptcon( lapack_int n, const float* d, const float* e, - float anorm, float* rcond ); -lapack_int LAPACKE_dptcon( lapack_int n, const double* d, const double* e, - double anorm, double* rcond ); -lapack_int LAPACKE_cptcon( lapack_int n, const float* d, - const lapack_complex_float* e, float anorm, - float* rcond ); -lapack_int LAPACKE_zptcon( lapack_int n, const double* d, - const lapack_complex_double* e, double anorm, - double* rcond ); - -lapack_int LAPACKE_spteqr( int matrix_order, char compz, lapack_int n, float* d, - float* e, float* z, lapack_int ldz ); -lapack_int LAPACKE_dpteqr( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz ); -lapack_int LAPACKE_cpteqr( int matrix_order, char compz, lapack_int n, float* d, - float* e, lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zpteqr( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_sptrfs( int matrix_order, lapack_int n, lapack_int nrhs, - const float* d, const float* e, const float* df, - const float* ef, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_dptrfs( int matrix_order, lapack_int n, lapack_int nrhs, - const double* d, const double* e, const double* df, - const double* ef, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* ferr, - double* berr ); -lapack_int LAPACKE_cptrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, const float* df, - const lapack_complex_float* ef, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zptrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, const double* df, - const lapack_complex_double* ef, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sptsv( int matrix_order, lapack_int n, lapack_int nrhs, - float* d, float* e, float* b, lapack_int ldb ); -lapack_int LAPACKE_dptsv( int matrix_order, lapack_int n, lapack_int nrhs, - double* d, double* e, double* b, lapack_int ldb ); -lapack_int LAPACKE_cptsv( int matrix_order, lapack_int n, lapack_int nrhs, - float* d, lapack_complex_float* e, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zptsv( int matrix_order, lapack_int n, lapack_int nrhs, - double* d, lapack_complex_double* e, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sptsvx( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const float* d, const float* e, - float* df, float* ef, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dptsvx( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const double* d, const double* e, - double* df, double* ef, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); -lapack_int LAPACKE_cptsvx( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, float* df, - lapack_complex_float* ef, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zptsvx( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, double* df, - lapack_complex_double* ef, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_spttrf( lapack_int n, float* d, float* e ); -lapack_int LAPACKE_dpttrf( lapack_int n, double* d, double* e ); -lapack_int LAPACKE_cpttrf( lapack_int n, float* d, lapack_complex_float* e ); -lapack_int LAPACKE_zpttrf( lapack_int n, double* d, lapack_complex_double* e ); - -lapack_int LAPACKE_spttrs( int matrix_order, lapack_int n, lapack_int nrhs, - const float* d, const float* e, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dpttrs( int matrix_order, lapack_int n, lapack_int nrhs, - const double* d, const double* e, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cpttrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpttrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_ssbev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, float* ab, lapack_int ldab, float* w, - float* z, lapack_int ldz ); -lapack_int LAPACKE_dsbev( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, double* ab, lapack_int ldab, double* w, - double* z, lapack_int ldz ); - -lapack_int LAPACKE_ssbevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, float* ab, lapack_int ldab, float* w, - float* z, lapack_int ldz ); -lapack_int LAPACKE_dsbevd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int kd, double* ab, lapack_int ldab, - double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_ssbevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int kd, float* ab, - lapack_int ldab, float* q, lapack_int ldq, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dsbevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int kd, double* ab, - lapack_int ldab, double* q, lapack_int ldq, - double vl, double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* ifail ); - -lapack_int LAPACKE_ssbgst( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, float* ab, - lapack_int ldab, const float* bb, lapack_int ldbb, - float* x, lapack_int ldx ); -lapack_int LAPACKE_dsbgst( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, double* ab, - lapack_int ldab, const double* bb, lapack_int ldbb, - double* x, lapack_int ldx ); - -lapack_int LAPACKE_ssbgv( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, float* ab, - lapack_int ldab, float* bb, lapack_int ldbb, float* w, - float* z, lapack_int ldz ); -lapack_int LAPACKE_dsbgv( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, double* ab, - lapack_int ldab, double* bb, lapack_int ldbb, - double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_ssbgvd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, float* ab, - lapack_int ldab, float* bb, lapack_int ldbb, - float* w, float* z, lapack_int ldz ); -lapack_int LAPACKE_dsbgvd( int matrix_order, char jobz, char uplo, lapack_int n, - lapack_int ka, lapack_int kb, double* ab, - lapack_int ldab, double* bb, lapack_int ldbb, - double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_ssbgvx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - float* ab, lapack_int ldab, float* bb, - lapack_int ldbb, float* q, lapack_int ldq, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dsbgvx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - double* ab, lapack_int ldab, double* bb, - lapack_int ldbb, double* q, lapack_int ldq, - double vl, double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* ifail ); - -lapack_int LAPACKE_ssbtrd( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int kd, float* ab, lapack_int ldab, float* d, - float* e, float* q, lapack_int ldq ); -lapack_int LAPACKE_dsbtrd( int matrix_order, char vect, char uplo, lapack_int n, - lapack_int kd, double* ab, lapack_int ldab, - double* d, double* e, double* q, lapack_int ldq ); - -lapack_int LAPACKE_ssfrk( int matrix_order, char transr, char uplo, char trans, - lapack_int n, lapack_int k, float alpha, - const float* a, lapack_int lda, float beta, - float* c ); -lapack_int LAPACKE_dsfrk( int matrix_order, char transr, char uplo, char trans, - lapack_int n, lapack_int k, double alpha, - const double* a, lapack_int lda, double beta, - double* c ); - -lapack_int LAPACKE_sspcon( int matrix_order, char uplo, lapack_int n, - const float* ap, const lapack_int* ipiv, float anorm, - float* rcond ); -lapack_int LAPACKE_dspcon( int matrix_order, char uplo, lapack_int n, - const double* ap, const lapack_int* ipiv, - double anorm, double* rcond ); -lapack_int LAPACKE_cspcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zspcon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_sspev( int matrix_order, char jobz, char uplo, lapack_int n, - float* ap, float* w, float* z, lapack_int ldz ); -lapack_int LAPACKE_dspev( int matrix_order, char jobz, char uplo, lapack_int n, - double* ap, double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sspevd( int matrix_order, char jobz, char uplo, lapack_int n, - float* ap, float* w, float* z, lapack_int ldz ); -lapack_int LAPACKE_dspevd( int matrix_order, char jobz, char uplo, lapack_int n, - double* ap, double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sspevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, float* ap, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dspevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, double* ap, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_sspgst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, float* ap, const float* bp ); -lapack_int LAPACKE_dspgst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, double* ap, const double* bp ); - -lapack_int LAPACKE_sspgv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* ap, float* bp, - float* w, float* z, lapack_int ldz ); -lapack_int LAPACKE_dspgv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* ap, double* bp, - double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sspgvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* ap, float* bp, - float* w, float* z, lapack_int ldz ); -lapack_int LAPACKE_dspgvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* ap, double* bp, - double* w, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sspgvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, float* ap, - float* bp, float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, float* w, - float* z, lapack_int ldz, lapack_int* ifail ); -lapack_int LAPACKE_dspgvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, double* ap, - double* bp, double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, double* z, lapack_int ldz, - lapack_int* ifail ); - -lapack_int LAPACKE_ssprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, const float* afp, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dsprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, const double* afp, - const lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr ); -lapack_int LAPACKE_csprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zsprfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_sspsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* ap, lapack_int* ipiv, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dspsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* ap, lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cspsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zspsv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sspsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, float* afp, - lapack_int* ipiv, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dspsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, double* afp, - lapack_int* ipiv, const double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* ferr, double* berr ); -lapack_int LAPACKE_cspsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - lapack_complex_float* afp, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zspsvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - lapack_complex_double* afp, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_ssptrd( int matrix_order, char uplo, lapack_int n, float* ap, - float* d, float* e, float* tau ); -lapack_int LAPACKE_dsptrd( int matrix_order, char uplo, lapack_int n, - double* ap, double* d, double* e, double* tau ); - -lapack_int LAPACKE_ssptrf( int matrix_order, char uplo, lapack_int n, float* ap, - lapack_int* ipiv ); -lapack_int LAPACKE_dsptrf( int matrix_order, char uplo, lapack_int n, - double* ap, lapack_int* ipiv ); -lapack_int LAPACKE_csptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, lapack_int* ipiv ); -lapack_int LAPACKE_zsptrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, lapack_int* ipiv ); - -lapack_int LAPACKE_ssptri( int matrix_order, char uplo, lapack_int n, float* ap, - const lapack_int* ipiv ); -lapack_int LAPACKE_dsptri( int matrix_order, char uplo, lapack_int n, - double* ap, const lapack_int* ipiv ); -lapack_int LAPACKE_csptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, const lapack_int* ipiv ); -lapack_int LAPACKE_zsptri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, const lapack_int* ipiv ); - -lapack_int LAPACKE_ssptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dsptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_csptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zsptrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* ap, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sstebz( char range, char order, lapack_int n, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - const float* d, const float* e, lapack_int* m, - lapack_int* nsplit, float* w, lapack_int* iblock, - lapack_int* isplit ); -lapack_int LAPACKE_dstebz( char range, char order, lapack_int n, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, const double* d, const double* e, - lapack_int* m, lapack_int* nsplit, double* w, - lapack_int* iblock, lapack_int* isplit ); - -lapack_int LAPACKE_sstedc( int matrix_order, char compz, lapack_int n, float* d, - float* e, float* z, lapack_int ldz ); -lapack_int LAPACKE_dstedc( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz ); -lapack_int LAPACKE_cstedc( int matrix_order, char compz, lapack_int n, float* d, - float* e, lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zstedc( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_sstegr( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* isuppz ); -lapack_int LAPACKE_dstegr( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* isuppz ); -lapack_int LAPACKE_cstegr( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* isuppz ); -lapack_int LAPACKE_zstegr( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* isuppz ); - -lapack_int LAPACKE_sstein( int matrix_order, lapack_int n, const float* d, - const float* e, lapack_int m, const float* w, - const lapack_int* iblock, const lapack_int* isplit, - float* z, lapack_int ldz, lapack_int* ifailv ); -lapack_int LAPACKE_dstein( int matrix_order, lapack_int n, const double* d, - const double* e, lapack_int m, const double* w, - const lapack_int* iblock, const lapack_int* isplit, - double* z, lapack_int ldz, lapack_int* ifailv ); -lapack_int LAPACKE_cstein( int matrix_order, lapack_int n, const float* d, - const float* e, lapack_int m, const float* w, - const lapack_int* iblock, const lapack_int* isplit, - lapack_complex_float* z, lapack_int ldz, - lapack_int* ifailv ); -lapack_int LAPACKE_zstein( int matrix_order, lapack_int n, const double* d, - const double* e, lapack_int m, const double* w, - const lapack_int* iblock, const lapack_int* isplit, - lapack_complex_double* z, lapack_int ldz, - lapack_int* ifailv ); - -lapack_int LAPACKE_sstemr( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, lapack_int* m, - float* w, float* z, lapack_int ldz, lapack_int nzc, - lapack_int* isuppz, lapack_logical* tryrac ); -lapack_int LAPACKE_dstemr( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - lapack_int* m, double* w, double* z, lapack_int ldz, - lapack_int nzc, lapack_int* isuppz, - lapack_logical* tryrac ); -lapack_int LAPACKE_cstemr( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, lapack_int* m, - float* w, lapack_complex_float* z, lapack_int ldz, - lapack_int nzc, lapack_int* isuppz, - lapack_logical* tryrac ); -lapack_int LAPACKE_zstemr( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int ldz, lapack_int nzc, lapack_int* isuppz, - lapack_logical* tryrac ); - -lapack_int LAPACKE_ssteqr( int matrix_order, char compz, lapack_int n, float* d, - float* e, float* z, lapack_int ldz ); -lapack_int LAPACKE_dsteqr( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz ); -lapack_int LAPACKE_csteqr( int matrix_order, char compz, lapack_int n, float* d, - float* e, lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zsteqr( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz ); - -lapack_int LAPACKE_ssterf( lapack_int n, float* d, float* e ); -lapack_int LAPACKE_dsterf( lapack_int n, double* d, double* e ); - -lapack_int LAPACKE_sstev( int matrix_order, char jobz, lapack_int n, float* d, - float* e, float* z, lapack_int ldz ); -lapack_int LAPACKE_dstev( int matrix_order, char jobz, lapack_int n, double* d, - double* e, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sstevd( int matrix_order, char jobz, lapack_int n, float* d, - float* e, float* z, lapack_int ldz ); -lapack_int LAPACKE_dstevd( int matrix_order, char jobz, lapack_int n, double* d, - double* e, double* z, lapack_int ldz ); - -lapack_int LAPACKE_sstevr( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* isuppz ); -lapack_int LAPACKE_dstevr( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* isuppz ); - -lapack_int LAPACKE_sstevx( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dstevx( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* ifail ); - -lapack_int LAPACKE_ssycon( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_dsycon( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond ); -lapack_int LAPACKE_csycon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, float* rcond ); -lapack_int LAPACKE_zsycon( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond ); - -lapack_int LAPACKE_ssyequb( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float* s, - float* scond, float* amax ); -lapack_int LAPACKE_dsyequb( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double* s, - double* scond, double* amax ); -lapack_int LAPACKE_csyequb( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zsyequb( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_ssyev( int matrix_order, char jobz, char uplo, lapack_int n, - float* a, lapack_int lda, float* w ); -lapack_int LAPACKE_dsyev( int matrix_order, char jobz, char uplo, lapack_int n, - double* a, lapack_int lda, double* w ); - -lapack_int LAPACKE_ssyevd( int matrix_order, char jobz, char uplo, lapack_int n, - float* a, lapack_int lda, float* w ); -lapack_int LAPACKE_dsyevd( int matrix_order, char jobz, char uplo, lapack_int n, - double* a, lapack_int lda, double* w ); - -lapack_int LAPACKE_ssyevr( int matrix_order, char jobz, char range, char uplo, - lapack_int n, float* a, lapack_int lda, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* isuppz ); -lapack_int LAPACKE_dsyevr( int matrix_order, char jobz, char range, char uplo, - lapack_int n, double* a, lapack_int lda, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* isuppz ); - -lapack_int LAPACKE_ssyevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, float* a, lapack_int lda, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dsyevx( int matrix_order, char jobz, char range, char uplo, - lapack_int n, double* a, lapack_int lda, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* ifail ); - -lapack_int LAPACKE_ssygst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, float* a, lapack_int lda, - const float* b, lapack_int ldb ); -lapack_int LAPACKE_dsygst( int matrix_order, lapack_int itype, char uplo, - lapack_int n, double* a, lapack_int lda, - const double* b, lapack_int ldb ); - -lapack_int LAPACKE_ssygv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* a, lapack_int lda, - float* b, lapack_int ldb, float* w ); -lapack_int LAPACKE_dsygv( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* a, lapack_int lda, - double* b, lapack_int ldb, double* w ); - -lapack_int LAPACKE_ssygvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* a, lapack_int lda, - float* b, lapack_int ldb, float* w ); -lapack_int LAPACKE_dsygvd( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* a, lapack_int lda, - double* b, lapack_int ldb, double* w ); - -lapack_int LAPACKE_ssygvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, float vl, - float vu, lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, lapack_int ldz, - lapack_int* ifail ); -lapack_int LAPACKE_dsygvx( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* ifail ); - -lapack_int LAPACKE_ssyrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dsyrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr ); -lapack_int LAPACKE_csyrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_zsyrfs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_ssyrfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* s, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dsyrfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* s, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); -lapack_int LAPACKE_csyrfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* s, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params ); -lapack_int LAPACKE_zsyrfsx( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, lapack_int ldaf, - const lapack_int* ipiv, const double* s, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params ); - -lapack_int LAPACKE_ssysv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dsysv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_csysv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zsysv( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_ssysvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - float* af, lapack_int ldaf, lapack_int* ipiv, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr ); -lapack_int LAPACKE_dsysvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - double* af, lapack_int ldaf, lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr ); -lapack_int LAPACKE_csysvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* af, - lapack_int ldaf, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr ); -lapack_int LAPACKE_zsysvx( int matrix_order, char fact, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* af, - lapack_int ldaf, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr ); - -lapack_int LAPACKE_ssysvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_dsysvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); -lapack_int LAPACKE_csysvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params ); -lapack_int LAPACKE_zsysvxx( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params ); - -lapack_int LAPACKE_ssytrd( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, float* d, float* e, float* tau ); -lapack_int LAPACKE_dsytrd( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda, double* d, double* e, double* tau ); - -lapack_int LAPACKE_ssytrf( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_dsytrf( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_csytrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zsytrf( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_ssytri( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_dsytri( int matrix_order, char uplo, lapack_int n, double* a, - lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_csytri( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zsytri( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); - -lapack_int LAPACKE_ssytrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_dsytrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_csytrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zsytrs( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stbcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, lapack_int kd, const float* ab, - lapack_int ldab, float* rcond ); -lapack_int LAPACKE_dtbcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, lapack_int kd, const double* ab, - lapack_int ldab, double* rcond ); -lapack_int LAPACKE_ctbcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, lapack_int kd, - const lapack_complex_float* ab, lapack_int ldab, - float* rcond ); -lapack_int LAPACKE_ztbcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, lapack_int kd, - const lapack_complex_double* ab, lapack_int ldab, - double* rcond ); - -lapack_int LAPACKE_stbrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const float* ab, lapack_int ldab, const float* b, - lapack_int ldb, const float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_dtbrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const double* ab, lapack_int ldab, const double* b, - lapack_int ldb, const double* x, lapack_int ldx, - double* ferr, double* berr ); -lapack_int LAPACKE_ctbrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_ztbrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_stbtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const float* ab, lapack_int ldab, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dtbtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const double* ab, lapack_int ldab, double* b, - lapack_int ldb ); -lapack_int LAPACKE_ctbtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztbtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stfsm( int matrix_order, char transr, char side, char uplo, - char trans, char diag, lapack_int m, lapack_int n, - float alpha, const float* a, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dtfsm( int matrix_order, char transr, char side, char uplo, - char trans, char diag, lapack_int m, lapack_int n, - double alpha, const double* a, double* b, - lapack_int ldb ); -lapack_int LAPACKE_ctfsm( int matrix_order, char transr, char side, char uplo, - char trans, char diag, lapack_int m, lapack_int n, - lapack_complex_float alpha, - const lapack_complex_float* a, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztfsm( int matrix_order, char transr, char side, char uplo, - char trans, char diag, lapack_int m, lapack_int n, - lapack_complex_double alpha, - const lapack_complex_double* a, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stftri( int matrix_order, char transr, char uplo, char diag, - lapack_int n, float* a ); -lapack_int LAPACKE_dtftri( int matrix_order, char transr, char uplo, char diag, - lapack_int n, double* a ); -lapack_int LAPACKE_ctftri( int matrix_order, char transr, char uplo, char diag, - lapack_int n, lapack_complex_float* a ); -lapack_int LAPACKE_ztftri( int matrix_order, char transr, char uplo, char diag, - lapack_int n, lapack_complex_double* a ); - -lapack_int LAPACKE_stfttp( int matrix_order, char transr, char uplo, - lapack_int n, const float* arf, float* ap ); -lapack_int LAPACKE_dtfttp( int matrix_order, char transr, char uplo, - lapack_int n, const double* arf, double* ap ); -lapack_int LAPACKE_ctfttp( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* arf, - lapack_complex_float* ap ); -lapack_int LAPACKE_ztfttp( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* arf, - lapack_complex_double* ap ); - -lapack_int LAPACKE_stfttr( int matrix_order, char transr, char uplo, - lapack_int n, const float* arf, float* a, - lapack_int lda ); -lapack_int LAPACKE_dtfttr( int matrix_order, char transr, char uplo, - lapack_int n, const double* arf, double* a, - lapack_int lda ); -lapack_int LAPACKE_ctfttr( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* arf, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_ztfttr( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* arf, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_stgevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const float* s, lapack_int lds, const float* p, - lapack_int ldp, float* vl, lapack_int ldvl, - float* vr, lapack_int ldvr, lapack_int mm, - lapack_int* m ); -lapack_int LAPACKE_dtgevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const double* s, lapack_int lds, const double* p, - lapack_int ldp, double* vl, lapack_int ldvl, - double* vr, lapack_int ldvr, lapack_int mm, - lapack_int* m ); -lapack_int LAPACKE_ctgevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* s, lapack_int lds, - const lapack_complex_float* p, lapack_int ldp, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ztgevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* s, lapack_int lds, - const lapack_complex_double* p, lapack_int ldp, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m ); - -lapack_int LAPACKE_stgexc( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, float* q, - lapack_int ldq, float* z, lapack_int ldz, - lapack_int* ifst, lapack_int* ilst ); -lapack_int LAPACKE_dtgexc( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, double* q, - lapack_int ldq, double* z, lapack_int ldz, - lapack_int* ifst, lapack_int* ilst ); -lapack_int LAPACKE_ctgexc( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz, - lapack_int ifst, lapack_int ilst ); -lapack_int LAPACKE_ztgexc( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz, - lapack_int ifst, lapack_int ilst ); - -lapack_int LAPACKE_stgsen( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* alphar, float* alphai, float* beta, float* q, - lapack_int ldq, float* z, lapack_int ldz, - lapack_int* m, float* pl, float* pr, float* dif ); -lapack_int LAPACKE_dtgsen( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - double* a, lapack_int lda, double* b, lapack_int ldb, - double* alphar, double* alphai, double* beta, - double* q, lapack_int ldq, double* z, lapack_int ldz, - lapack_int* m, double* pl, double* pr, double* dif ); -lapack_int LAPACKE_ctgsen( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* alpha, - lapack_complex_float* beta, lapack_complex_float* q, - lapack_int ldq, lapack_complex_float* z, - lapack_int ldz, lapack_int* m, float* pl, float* pr, - float* dif ); -lapack_int LAPACKE_ztgsen( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz, - lapack_int* m, double* pl, double* pr, double* dif ); - -lapack_int LAPACKE_stgsja( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_int k, lapack_int l, float* a, lapack_int lda, - float* b, lapack_int ldb, float tola, float tolb, - float* alpha, float* beta, float* u, lapack_int ldu, - float* v, lapack_int ldv, float* q, lapack_int ldq, - lapack_int* ncycle ); -lapack_int LAPACKE_dtgsja( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_int k, lapack_int l, double* a, - lapack_int lda, double* b, lapack_int ldb, - double tola, double tolb, double* alpha, - double* beta, double* u, lapack_int ldu, double* v, - lapack_int ldv, double* q, lapack_int ldq, - lapack_int* ncycle ); -lapack_int LAPACKE_ctgsja( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_int k, lapack_int l, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float tola, float tolb, float* alpha, - float* beta, lapack_complex_float* u, lapack_int ldu, - lapack_complex_float* v, lapack_int ldv, - lapack_complex_float* q, lapack_int ldq, - lapack_int* ncycle ); -lapack_int LAPACKE_ztgsja( int matrix_order, char jobu, char jobv, char jobq, - lapack_int m, lapack_int p, lapack_int n, - lapack_int k, lapack_int l, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double tola, double tolb, - double* alpha, double* beta, - lapack_complex_double* u, lapack_int ldu, - lapack_complex_double* v, lapack_int ldv, - lapack_complex_double* q, lapack_int ldq, - lapack_int* ncycle ); - -lapack_int LAPACKE_stgsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const float* a, lapack_int lda, const float* b, - lapack_int ldb, const float* vl, lapack_int ldvl, - const float* vr, lapack_int ldvr, float* s, - float* dif, lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_dtgsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const double* a, lapack_int lda, const double* b, - lapack_int ldb, const double* vl, lapack_int ldvl, - const double* vr, lapack_int ldvr, double* s, - double* dif, lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ctgsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* vl, lapack_int ldvl, - const lapack_complex_float* vr, lapack_int ldvr, - float* s, float* dif, lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ztgsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* vl, lapack_int ldvl, - const lapack_complex_double* vr, lapack_int ldvr, - double* s, double* dif, lapack_int mm, - lapack_int* m ); - -lapack_int LAPACKE_stgsyl( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, const float* a, - lapack_int lda, const float* b, lapack_int ldb, - float* c, lapack_int ldc, const float* d, - lapack_int ldd, const float* e, lapack_int lde, - float* f, lapack_int ldf, float* scale, float* dif ); -lapack_int LAPACKE_dtgsyl( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, const double* a, - lapack_int lda, const double* b, lapack_int ldb, - double* c, lapack_int ldc, const double* d, - lapack_int ldd, const double* e, lapack_int lde, - double* f, lapack_int ldf, double* scale, - double* dif ); -lapack_int LAPACKE_ctgsyl( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* c, lapack_int ldc, - const lapack_complex_float* d, lapack_int ldd, - const lapack_complex_float* e, lapack_int lde, - lapack_complex_float* f, lapack_int ldf, - float* scale, float* dif ); -lapack_int LAPACKE_ztgsyl( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* c, lapack_int ldc, - const lapack_complex_double* d, lapack_int ldd, - const lapack_complex_double* e, lapack_int lde, - lapack_complex_double* f, lapack_int ldf, - double* scale, double* dif ); - -lapack_int LAPACKE_stpcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const float* ap, float* rcond ); -lapack_int LAPACKE_dtpcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const double* ap, double* rcond ); -lapack_int LAPACKE_ctpcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const lapack_complex_float* ap, - float* rcond ); -lapack_int LAPACKE_ztpcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const lapack_complex_double* ap, - double* rcond ); - -lapack_int LAPACKE_stprfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const float* ap, - const float* b, lapack_int ldb, const float* x, - lapack_int ldx, float* ferr, float* berr ); -lapack_int LAPACKE_dtprfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const double* ap, - const double* b, lapack_int ldb, const double* x, - lapack_int ldx, double* ferr, double* berr ); -lapack_int LAPACKE_ctprfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_ztprfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_stptri( int matrix_order, char uplo, char diag, lapack_int n, - float* ap ); -lapack_int LAPACKE_dtptri( int matrix_order, char uplo, char diag, lapack_int n, - double* ap ); -lapack_int LAPACKE_ctptri( int matrix_order, char uplo, char diag, lapack_int n, - lapack_complex_float* ap ); -lapack_int LAPACKE_ztptri( int matrix_order, char uplo, char diag, lapack_int n, - lapack_complex_double* ap ); - -lapack_int LAPACKE_stptrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const float* ap, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dtptrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const double* ap, - double* b, lapack_int ldb ); -lapack_int LAPACKE_ctptrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztptrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stpttf( int matrix_order, char transr, char uplo, - lapack_int n, const float* ap, float* arf ); -lapack_int LAPACKE_dtpttf( int matrix_order, char transr, char uplo, - lapack_int n, const double* ap, double* arf ); -lapack_int LAPACKE_ctpttf( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* ap, - lapack_complex_float* arf ); -lapack_int LAPACKE_ztpttf( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* ap, - lapack_complex_double* arf ); - -lapack_int LAPACKE_stpttr( int matrix_order, char uplo, lapack_int n, - const float* ap, float* a, lapack_int lda ); -lapack_int LAPACKE_dtpttr( int matrix_order, char uplo, lapack_int n, - const double* ap, double* a, lapack_int lda ); -lapack_int LAPACKE_ctpttr( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_ztpttr( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_strcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const float* a, lapack_int lda, - float* rcond ); -lapack_int LAPACKE_dtrcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const double* a, lapack_int lda, - double* rcond ); -lapack_int LAPACKE_ctrcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, float* rcond ); -lapack_int LAPACKE_ztrcon( int matrix_order, char norm, char uplo, char diag, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, double* rcond ); - -lapack_int LAPACKE_strevc( int matrix_order, char side, char howmny, - lapack_logical* select, lapack_int n, const float* t, - lapack_int ldt, float* vl, lapack_int ldvl, - float* vr, lapack_int ldvr, lapack_int mm, - lapack_int* m ); -lapack_int LAPACKE_dtrevc( int matrix_order, char side, char howmny, - lapack_logical* select, lapack_int n, - const double* t, lapack_int ldt, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ctrevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ztrevc( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m ); - -lapack_int LAPACKE_strexc( int matrix_order, char compq, lapack_int n, float* t, - lapack_int ldt, float* q, lapack_int ldq, - lapack_int* ifst, lapack_int* ilst ); -lapack_int LAPACKE_dtrexc( int matrix_order, char compq, lapack_int n, - double* t, lapack_int ldt, double* q, lapack_int ldq, - lapack_int* ifst, lapack_int* ilst ); -lapack_int LAPACKE_ctrexc( int matrix_order, char compq, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* q, lapack_int ldq, - lapack_int ifst, lapack_int ilst ); -lapack_int LAPACKE_ztrexc( int matrix_order, char compq, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* q, lapack_int ldq, - lapack_int ifst, lapack_int ilst ); - -lapack_int LAPACKE_strrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* b, lapack_int ldb, - const float* x, lapack_int ldx, float* ferr, - float* berr ); -lapack_int LAPACKE_dtrrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* b, lapack_int ldb, - const double* x, lapack_int ldx, double* ferr, - double* berr ); -lapack_int LAPACKE_ctrrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr ); -lapack_int LAPACKE_ztrrfs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr ); - -lapack_int LAPACKE_strsen( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, float* t, - lapack_int ldt, float* q, lapack_int ldq, float* wr, - float* wi, lapack_int* m, float* s, float* sep ); -lapack_int LAPACKE_dtrsen( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - double* t, lapack_int ldt, double* q, lapack_int ldq, - double* wr, double* wi, lapack_int* m, double* s, - double* sep ); -lapack_int LAPACKE_ctrsen( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* w, lapack_int* m, float* s, - float* sep ); -lapack_int LAPACKE_ztrsen( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* w, lapack_int* m, double* s, - double* sep ); - -lapack_int LAPACKE_strsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const float* t, lapack_int ldt, const float* vl, - lapack_int ldvl, const float* vr, lapack_int ldvr, - float* s, float* sep, lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_dtrsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const double* t, lapack_int ldt, const double* vl, - lapack_int ldvl, const double* vr, lapack_int ldvr, - double* s, double* sep, lapack_int mm, - lapack_int* m ); -lapack_int LAPACKE_ctrsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* t, lapack_int ldt, - const lapack_complex_float* vl, lapack_int ldvl, - const lapack_complex_float* vr, lapack_int ldvr, - float* s, float* sep, lapack_int mm, lapack_int* m ); -lapack_int LAPACKE_ztrsna( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* t, lapack_int ldt, - const lapack_complex_double* vl, lapack_int ldvl, - const lapack_complex_double* vr, lapack_int ldvr, - double* s, double* sep, lapack_int mm, - lapack_int* m ); - -lapack_int LAPACKE_strsyl( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const float* a, lapack_int lda, const float* b, - lapack_int ldb, float* c, lapack_int ldc, - float* scale ); -lapack_int LAPACKE_dtrsyl( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const double* a, lapack_int lda, const double* b, - lapack_int ldb, double* c, lapack_int ldc, - double* scale ); -lapack_int LAPACKE_ctrsyl( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* c, lapack_int ldc, - float* scale ); -lapack_int LAPACKE_ztrsyl( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* c, lapack_int ldc, - double* scale ); - -lapack_int LAPACKE_strtri( int matrix_order, char uplo, char diag, lapack_int n, - float* a, lapack_int lda ); -lapack_int LAPACKE_dtrtri( int matrix_order, char uplo, char diag, lapack_int n, - double* a, lapack_int lda ); -lapack_int LAPACKE_ctrtri( int matrix_order, char uplo, char diag, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_ztrtri( int matrix_order, char uplo, char diag, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_strtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, float* b, lapack_int ldb ); -lapack_int LAPACKE_dtrtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, double* b, lapack_int ldb ); -lapack_int LAPACKE_ctrtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztrtrs( int matrix_order, char uplo, char trans, char diag, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_strttf( int matrix_order, char transr, char uplo, - lapack_int n, const float* a, lapack_int lda, - float* arf ); -lapack_int LAPACKE_dtrttf( int matrix_order, char transr, char uplo, - lapack_int n, const double* a, lapack_int lda, - double* arf ); -lapack_int LAPACKE_ctrttf( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* arf ); -lapack_int LAPACKE_ztrttf( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* arf ); - -lapack_int LAPACKE_strttp( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float* ap ); -lapack_int LAPACKE_dtrttp( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double* ap ); -lapack_int LAPACKE_ctrttp( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* ap ); -lapack_int LAPACKE_ztrttp( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* ap ); - -lapack_int LAPACKE_stzrzf( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau ); -lapack_int LAPACKE_dtzrzf( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau ); -lapack_int LAPACKE_ctzrzf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau ); -lapack_int LAPACKE_ztzrzf( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau ); - -lapack_int LAPACKE_cungbr( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zungbr( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cunghr( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zunghr( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cunglq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zunglq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cungql( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zungql( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cungqr( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zungqr( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cungrq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau ); -lapack_int LAPACKE_zungrq( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau ); - -lapack_int LAPACKE_cungtr( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau ); -lapack_int LAPACKE_zungtr( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau ); - -lapack_int LAPACKE_cunmbr( int matrix_order, char vect, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmbr( int matrix_order, char vect, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmhr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmhr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmlq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmlq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmql( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmql( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmqr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmqr( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmrq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmrq( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmrz( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmrz( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cunmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zunmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_cupgtr( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_complex_float* tau, - lapack_complex_float* q, lapack_int ldq ); -lapack_int LAPACKE_zupgtr( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_complex_double* tau, - lapack_complex_double* q, lapack_int ldq ); - -lapack_int LAPACKE_cupmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, - const lapack_complex_float* ap, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc ); -lapack_int LAPACKE_zupmtr( int matrix_order, char side, char uplo, char trans, - lapack_int m, lapack_int n, - const lapack_complex_double* ap, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc ); - -lapack_int LAPACKE_sbdsdc_work( int matrix_order, char uplo, char compq, - lapack_int n, float* d, float* e, float* u, - lapack_int ldu, float* vt, lapack_int ldvt, - float* q, lapack_int* iq, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dbdsdc_work( int matrix_order, char uplo, char compq, - lapack_int n, double* d, double* e, double* u, - lapack_int ldu, double* vt, lapack_int ldvt, - double* q, lapack_int* iq, double* work, - lapack_int* iwork ); - -lapack_int LAPACKE_sbdsqr_work( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - float* d, float* e, float* vt, lapack_int ldvt, - float* u, lapack_int ldu, float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_dbdsqr_work( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - double* d, double* e, double* vt, - lapack_int ldvt, double* u, lapack_int ldu, - double* c, lapack_int ldc, double* work ); -lapack_int LAPACKE_cbdsqr_work( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - float* d, float* e, lapack_complex_float* vt, - lapack_int ldvt, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_zbdsqr_work( int matrix_order, char uplo, lapack_int n, - lapack_int ncvt, lapack_int nru, lapack_int ncc, - double* d, double* e, lapack_complex_double* vt, - lapack_int ldvt, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* c, - lapack_int ldc, double* work ); - -lapack_int LAPACKE_sdisna_work( char job, lapack_int m, lapack_int n, - const float* d, float* sep ); -lapack_int LAPACKE_ddisna_work( char job, lapack_int m, lapack_int n, - const double* d, double* sep ); - -lapack_int LAPACKE_sgbbrd_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, float* ab, lapack_int ldab, - float* d, float* e, float* q, lapack_int ldq, - float* pt, lapack_int ldpt, float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_dgbbrd_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, double* ab, lapack_int ldab, - double* d, double* e, double* q, lapack_int ldq, - double* pt, lapack_int ldpt, double* c, - lapack_int ldc, double* work ); -lapack_int LAPACKE_cgbbrd_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, lapack_complex_float* ab, - lapack_int ldab, float* d, float* e, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* pt, lapack_int ldpt, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgbbrd_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int ncc, lapack_int kl, - lapack_int ku, lapack_complex_double* ab, - lapack_int ldab, double* d, double* e, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* pt, lapack_int ldpt, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgbcon_work( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, const lapack_int* ipiv, - float anorm, float* rcond, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgbcon_work( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, const lapack_int* ipiv, - double anorm, double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgbcon_work( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgbcon_work( int matrix_order, char norm, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, - lapack_int ldab, const lapack_int* ipiv, - double anorm, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgbequ_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, float* r, float* c, - float* rowcnd, float* colcnd, float* amax ); -lapack_int LAPACKE_dgbequ_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); -lapack_int LAPACKE_cgbequ_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, lapack_int ldab, - float* r, float* c, float* rowcnd, - float* colcnd, float* amax ); -lapack_int LAPACKE_zgbequ_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); - -lapack_int LAPACKE_sgbequb_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* ab, - lapack_int ldab, float* r, float* c, - float* rowcnd, float* colcnd, float* amax ); -lapack_int LAPACKE_dgbequb_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); -lapack_int LAPACKE_cgbequb_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_float* ab, - lapack_int ldab, float* r, float* c, - float* rowcnd, float* colcnd, float* amax ); -lapack_int LAPACKE_zgbequb_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - const lapack_complex_double* ab, - lapack_int ldab, double* r, double* c, - double* rowcnd, double* colcnd, double* amax ); - -lapack_int LAPACKE_sgbrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const float* ab, lapack_int ldab, - const float* afb, lapack_int ldafb, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgbrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const double* ab, lapack_int ldab, - const double* afb, lapack_int ldafb, - const lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgbrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* afb, - lapack_int ldafb, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgbrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, - const lapack_complex_double* afb, - lapack_int ldafb, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgbrfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const float* ab, - lapack_int ldab, const float* afb, - lapack_int ldafb, const lapack_int* ipiv, - const float* r, const float* c, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgbrfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, const double* ab, - lapack_int ldab, const double* afb, - lapack_int ldafb, const lapack_int* ipiv, - const double* r, const double* c, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgbrfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, - const lapack_complex_float* ab, - lapack_int ldab, - const lapack_complex_float* afb, - lapack_int ldafb, const lapack_int* ipiv, - const float* r, const float* c, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgbrfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, - const lapack_complex_double* afb, - lapack_int ldafb, const lapack_int* ipiv, - const double* r, const double* c, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sgbsv_work( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, float* ab, - lapack_int ldab, lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgbsv_work( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, double* ab, - lapack_int ldab, lapack_int* ipiv, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cgbsv_work( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgbsv_work( int matrix_order, lapack_int n, lapack_int kl, - lapack_int ku, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sgbsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, float* ab, lapack_int ldab, - float* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, float* r, float* c, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dgbsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, double* ab, lapack_int ldab, - double* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, double* r, double* c, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cgbsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - float* r, float* c, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgbsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* afb, - lapack_int ldafb, lapack_int* ipiv, char* equed, - double* r, double* c, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sgbsvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, float* ab, lapack_int ldab, - float* afb, lapack_int ldafb, lapack_int* ipiv, - char* equed, float* r, float* c, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgbsvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, double* ab, lapack_int ldab, - double* afb, lapack_int ldafb, - lapack_int* ipiv, char* equed, double* r, - double* c, double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgbsvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* afb, - lapack_int ldafb, lapack_int* ipiv, - char* equed, float* r, float* c, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgbsvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int kl, lapack_int ku, - lapack_int nrhs, lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* afb, - lapack_int ldafb, lapack_int* ipiv, - char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sgbtrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, float* ab, - lapack_int ldab, lapack_int* ipiv ); -lapack_int LAPACKE_dgbtrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, double* ab, - lapack_int ldab, lapack_int* ipiv ); -lapack_int LAPACKE_cgbtrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - lapack_complex_float* ab, lapack_int ldab, - lapack_int* ipiv ); -lapack_int LAPACKE_zgbtrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, - lapack_complex_double* ab, lapack_int ldab, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgbtrs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const float* ab, lapack_int ldab, - const lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgbtrs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const double* ab, lapack_int ldab, - const lapack_int* ipiv, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cgbtrs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgbtrs_work( int matrix_order, char trans, lapack_int n, - lapack_int kl, lapack_int ku, lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sgebak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const float* scale, lapack_int m, float* v, - lapack_int ldv ); -lapack_int LAPACKE_dgebak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const double* scale, lapack_int m, double* v, - lapack_int ldv ); -lapack_int LAPACKE_cgebak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const float* scale, lapack_int m, - lapack_complex_float* v, lapack_int ldv ); -lapack_int LAPACKE_zgebak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const double* scale, lapack_int m, - lapack_complex_double* v, lapack_int ldv ); - -lapack_int LAPACKE_sgebal_work( int matrix_order, char job, lapack_int n, - float* a, lapack_int lda, lapack_int* ilo, - lapack_int* ihi, float* scale ); -lapack_int LAPACKE_dgebal_work( int matrix_order, char job, lapack_int n, - double* a, lapack_int lda, lapack_int* ilo, - lapack_int* ihi, double* scale ); -lapack_int LAPACKE_cgebal_work( int matrix_order, char job, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ilo, lapack_int* ihi, - float* scale ); -lapack_int LAPACKE_zgebal_work( int matrix_order, char job, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ilo, lapack_int* ihi, - double* scale ); - -lapack_int LAPACKE_sgebrd_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* d, float* e, - float* tauq, float* taup, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dgebrd_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* d, double* e, - double* tauq, double* taup, double* work, - lapack_int lwork ); -lapack_int LAPACKE_cgebrd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - float* d, float* e, lapack_complex_float* tauq, - lapack_complex_float* taup, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgebrd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double* d, double* e, - lapack_complex_double* tauq, - lapack_complex_double* taup, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgecon_work( int matrix_order, char norm, lapack_int n, - const float* a, lapack_int lda, float anorm, - float* rcond, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dgecon_work( int matrix_order, char norm, lapack_int n, - const double* a, lapack_int lda, double anorm, - double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgecon_work( int matrix_order, char norm, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float anorm, float* rcond, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgecon_work( int matrix_order, char norm, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double anorm, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgeequ_work( int matrix_order, lapack_int m, lapack_int n, - const float* a, lapack_int lda, float* r, - float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_dgeequ_work( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, double* r, - double* c, double* rowcnd, double* colcnd, - double* amax ); -lapack_int LAPACKE_cgeequ_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* r, float* c, float* rowcnd, - float* colcnd, float* amax ); -lapack_int LAPACKE_zgeequ_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* r, double* c, double* rowcnd, - double* colcnd, double* amax ); - -lapack_int LAPACKE_sgeequb_work( int matrix_order, lapack_int m, lapack_int n, - const float* a, lapack_int lda, float* r, - float* c, float* rowcnd, float* colcnd, - float* amax ); -lapack_int LAPACKE_dgeequb_work( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, double* r, - double* c, double* rowcnd, double* colcnd, - double* amax ); -lapack_int LAPACKE_cgeequb_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* r, float* c, float* rowcnd, - float* colcnd, float* amax ); -lapack_int LAPACKE_zgeequb_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* r, double* c, double* rowcnd, - double* colcnd, double* amax ); - -lapack_int LAPACKE_sgees_work( int matrix_order, char jobvs, char sort, - LAPACK_S_SELECT2 select, lapack_int n, float* a, - lapack_int lda, lapack_int* sdim, float* wr, - float* wi, float* vs, lapack_int ldvs, - float* work, lapack_int lwork, - lapack_logical* bwork ); -lapack_int LAPACKE_dgees_work( int matrix_order, char jobvs, char sort, - LAPACK_D_SELECT2 select, lapack_int n, double* a, - lapack_int lda, lapack_int* sdim, double* wr, - double* wi, double* vs, lapack_int ldvs, - double* work, lapack_int lwork, - lapack_logical* bwork ); -lapack_int LAPACKE_cgees_work( int matrix_order, char jobvs, char sort, - LAPACK_C_SELECT1 select, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* sdim, lapack_complex_float* w, - lapack_complex_float* vs, lapack_int ldvs, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_logical* bwork ); -lapack_int LAPACKE_zgees_work( int matrix_order, char jobvs, char sort, - LAPACK_Z_SELECT1 select, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* sdim, lapack_complex_double* w, - lapack_complex_double* vs, lapack_int ldvs, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_logical* bwork ); - -lapack_int LAPACKE_sgeesx_work( int matrix_order, char jobvs, char sort, - LAPACK_S_SELECT2 select, char sense, - lapack_int n, float* a, lapack_int lda, - lapack_int* sdim, float* wr, float* wi, - float* vs, lapack_int ldvs, float* rconde, - float* rcondv, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork, - lapack_logical* bwork ); -lapack_int LAPACKE_dgeesx_work( int matrix_order, char jobvs, char sort, - LAPACK_D_SELECT2 select, char sense, - lapack_int n, double* a, lapack_int lda, - lapack_int* sdim, double* wr, double* wi, - double* vs, lapack_int ldvs, double* rconde, - double* rcondv, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork, - lapack_logical* bwork ); -lapack_int LAPACKE_cgeesx_work( int matrix_order, char jobvs, char sort, - LAPACK_C_SELECT1 select, char sense, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_int* sdim, - lapack_complex_float* w, - lapack_complex_float* vs, lapack_int ldvs, - float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_logical* bwork ); -lapack_int LAPACKE_zgeesx_work( int matrix_order, char jobvs, char sort, - LAPACK_Z_SELECT1 select, char sense, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_int* sdim, - lapack_complex_double* w, - lapack_complex_double* vs, lapack_int ldvs, - double* rconde, double* rcondv, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_logical* bwork ); - -lapack_int LAPACKE_sgeev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, float* a, lapack_int lda, - float* wr, float* wi, float* vl, lapack_int ldvl, - float* vr, lapack_int ldvr, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dgeev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, double* a, lapack_int lda, - double* wr, double* wi, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgeev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* w, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zgeev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* w, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_sgeevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, float* a, - lapack_int lda, float* wr, float* wi, float* vl, - lapack_int ldvl, float* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, float* scale, - float* abnrm, float* rconde, float* rcondv, - float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dgeevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, double* a, - lapack_int lda, double* wr, double* wi, - double* vl, lapack_int ldvl, double* vr, - lapack_int ldvr, lapack_int* ilo, - lapack_int* ihi, double* scale, double* abnrm, - double* rconde, double* rcondv, double* work, - lapack_int lwork, lapack_int* iwork ); -lapack_int LAPACKE_cgeevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* w, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, float* scale, - float* abnrm, float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zgeevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* w, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, double* scale, - double* abnrm, double* rconde, double* rcondv, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_sgehrd_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, float* a, lapack_int lda, - float* tau, float* work, lapack_int lwork ); -lapack_int LAPACKE_dgehrd_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, double* a, lapack_int lda, - double* tau, double* work, lapack_int lwork ); -lapack_int LAPACKE_cgehrd_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgehrd_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgejsv_work( int matrix_order, char joba, char jobu, - char jobv, char jobr, char jobt, char jobp, - lapack_int m, lapack_int n, float* a, - lapack_int lda, float* sva, float* u, - lapack_int ldu, float* v, lapack_int ldv, - float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dgejsv_work( int matrix_order, char joba, char jobu, - char jobv, char jobr, char jobt, char jobp, - lapack_int m, lapack_int n, double* a, - lapack_int lda, double* sva, double* u, - lapack_int ldu, double* v, lapack_int ldv, - double* work, lapack_int lwork, - lapack_int* iwork ); - -lapack_int LAPACKE_sgelq2_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work ); -lapack_int LAPACKE_dgelq2_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work ); -lapack_int LAPACKE_cgelq2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work ); -lapack_int LAPACKE_zgelq2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work ); - -lapack_int LAPACKE_sgelqf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgelqf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgelqf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgelqf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgels_work( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgels_work( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgels_work( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgels_work( int matrix_order, char trans, lapack_int m, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgelsd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - float* b, lapack_int ldb, float* s, float rcond, - lapack_int* rank, float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dgelsd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* s, - double rcond, lapack_int* rank, double* work, - lapack_int lwork, lapack_int* iwork ); -lapack_int LAPACKE_cgelsd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int* iwork ); -lapack_int LAPACKE_zgelsd_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* s, double rcond, - lapack_int* rank, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int* iwork ); - -lapack_int LAPACKE_sgelss_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - float* b, lapack_int ldb, float* s, float rcond, - lapack_int* rank, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dgelss_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* s, - double rcond, lapack_int* rank, double* work, - lapack_int lwork ); -lapack_int LAPACKE_cgelss_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* s, float rcond, - lapack_int* rank, lapack_complex_float* work, - lapack_int lwork, float* rwork ); -lapack_int LAPACKE_zgelss_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double* s, double rcond, - lapack_int* rank, lapack_complex_double* work, - lapack_int lwork, double* rwork ); - -lapack_int LAPACKE_sgelsy_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - float* b, lapack_int ldb, lapack_int* jpvt, - float rcond, lapack_int* rank, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dgelsy_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, lapack_int* jpvt, - double rcond, lapack_int* rank, double* work, - lapack_int lwork ); -lapack_int LAPACKE_cgelsy_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_int* jpvt, float rcond, - lapack_int* rank, lapack_complex_float* work, - lapack_int lwork, float* rwork ); -lapack_int LAPACKE_zgelsy_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_int* jpvt, double rcond, - lapack_int* rank, lapack_complex_double* work, - lapack_int lwork, double* rwork ); - -lapack_int LAPACKE_sgeqlf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgeqlf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgeqlf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgeqlf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgeqp3_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* jpvt, - float* tau, float* work, lapack_int lwork ); -lapack_int LAPACKE_dgeqp3_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* jpvt, - double* tau, double* work, lapack_int lwork ); -lapack_int LAPACKE_cgeqp3_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zgeqp3_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_sgeqpf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* jpvt, - float* tau, float* work ); -lapack_int LAPACKE_dgeqpf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* jpvt, - double* tau, double* work ); -lapack_int LAPACKE_cgeqpf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_float* tau, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgeqpf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* jpvt, lapack_complex_double* tau, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgeqr2_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work ); -lapack_int LAPACKE_dgeqr2_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work ); -lapack_int LAPACKE_cgeqr2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work ); -lapack_int LAPACKE_zgeqr2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work ); - -lapack_int LAPACKE_sgeqrf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgeqrf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgeqrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgeqrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgeqrfp_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgeqrfp_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgeqrfp_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgeqrfp_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sgerfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgerfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cgerfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgerfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgerfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, - lapack_int ldaf, const lapack_int* ipiv, - const float* r, const float* c, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgerfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* r, const double* c, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgerfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const float* r, const float* c, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgerfsx_work( int matrix_order, char trans, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* r, const double* c, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sgerqf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgerqf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgerqf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgerqf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgesdd_work( int matrix_order, char jobz, lapack_int m, - lapack_int n, float* a, lapack_int lda, - float* s, float* u, lapack_int ldu, float* vt, - lapack_int ldvt, float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dgesdd_work( int matrix_order, char jobz, lapack_int m, - lapack_int n, double* a, lapack_int lda, - double* s, double* u, lapack_int ldu, - double* vt, lapack_int ldvt, double* work, - lapack_int lwork, lapack_int* iwork ); -lapack_int LAPACKE_cgesdd_work( int matrix_order, char jobz, lapack_int m, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float* s, - lapack_complex_float* u, lapack_int ldu, - lapack_complex_float* vt, lapack_int ldvt, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int* iwork ); -lapack_int LAPACKE_zgesdd_work( int matrix_order, char jobz, lapack_int m, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double* s, - lapack_complex_double* u, lapack_int ldu, - lapack_complex_double* vt, lapack_int ldvt, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int* iwork ); - -lapack_int LAPACKE_sgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - float* a, lapack_int lda, lapack_int* ipiv, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - double* a, lapack_int lda, lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); -lapack_int LAPACKE_dsgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - double* a, lapack_int lda, lapack_int* ipiv, - double* b, lapack_int ldb, double* x, - lapack_int ldx, double* work, float* swork, - lapack_int* iter ); -lapack_int LAPACKE_zcgesv_work( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, lapack_complex_double* work, - lapack_complex_float* swork, double* rwork, - lapack_int* iter ); - -lapack_int LAPACKE_sgesvd_work( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, float* a, - lapack_int lda, float* s, float* u, - lapack_int ldu, float* vt, lapack_int ldvt, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgesvd_work( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, double* a, - lapack_int lda, double* s, double* u, - lapack_int ldu, double* vt, lapack_int ldvt, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgesvd_work( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - float* s, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* vt, - lapack_int ldvt, lapack_complex_float* work, - lapack_int lwork, float* rwork ); -lapack_int LAPACKE_zgesvd_work( int matrix_order, char jobu, char jobvt, - lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double* s, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* vt, - lapack_int ldvt, lapack_complex_double* work, - lapack_int lwork, double* rwork ); - -lapack_int LAPACKE_sgesvj_work( int matrix_order, char joba, char jobu, - char jobv, lapack_int m, lapack_int n, float* a, - lapack_int lda, float* sva, lapack_int mv, - float* v, lapack_int ldv, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dgesvj_work( int matrix_order, char joba, char jobu, - char jobv, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* sva, - lapack_int mv, double* v, lapack_int ldv, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgesvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, - float* c, float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dgesvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, - double* c, double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, double* work, lapack_int* iwork ); -lapack_int LAPACKE_cgesvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, - float* c, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zgesvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, - double* c, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sgesvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, - float* c, float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgesvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, - double* c, double* b, lapack_int ldb, - double* x, lapack_int ldx, double* rcond, - double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgesvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* r, - float* c, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgesvxx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* r, - double* c, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgetf2_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_dgetf2_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_cgetf2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zgetf2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgetrf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_dgetrf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, lapack_int* ipiv ); -lapack_int LAPACKE_cgetrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv ); -lapack_int LAPACKE_zgetrf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv ); - -lapack_int LAPACKE_sgetri_work( int matrix_order, lapack_int n, float* a, - lapack_int lda, const lapack_int* ipiv, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dgetri_work( int matrix_order, lapack_int n, double* a, - lapack_int lda, const lapack_int* ipiv, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cgetri_work( int matrix_order, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgetri_work( int matrix_order, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgetrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgetrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cgetrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zgetrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sggbak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const float* lscale, const float* rscale, - lapack_int m, float* v, lapack_int ldv ); -lapack_int LAPACKE_dggbak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const double* lscale, const double* rscale, - lapack_int m, double* v, lapack_int ldv ); -lapack_int LAPACKE_cggbak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const float* lscale, const float* rscale, - lapack_int m, lapack_complex_float* v, - lapack_int ldv ); -lapack_int LAPACKE_zggbak_work( int matrix_order, char job, char side, - lapack_int n, lapack_int ilo, lapack_int ihi, - const double* lscale, const double* rscale, - lapack_int m, lapack_complex_double* v, - lapack_int ldv ); - -lapack_int LAPACKE_sggbal_work( int matrix_order, char job, lapack_int n, - float* a, lapack_int lda, float* b, - lapack_int ldb, lapack_int* ilo, - lapack_int* ihi, float* lscale, float* rscale, - float* work ); -lapack_int LAPACKE_dggbal_work( int matrix_order, char job, lapack_int n, - double* a, lapack_int lda, double* b, - lapack_int ldb, lapack_int* ilo, - lapack_int* ihi, double* lscale, double* rscale, - double* work ); -lapack_int LAPACKE_cggbal_work( int matrix_order, char job, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale, float* work ); -lapack_int LAPACKE_zggbal_work( int matrix_order, char job, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_int* ilo, lapack_int* ihi, - double* lscale, double* rscale, double* work ); - -lapack_int LAPACKE_sgges_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_S_SELECT3 selctg, lapack_int n, - float* a, lapack_int lda, float* b, - lapack_int ldb, lapack_int* sdim, float* alphar, - float* alphai, float* beta, float* vsl, - lapack_int ldvsl, float* vsr, lapack_int ldvsr, - float* work, lapack_int lwork, - lapack_logical* bwork ); -lapack_int LAPACKE_dgges_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_D_SELECT3 selctg, lapack_int n, - double* a, lapack_int lda, double* b, - lapack_int ldb, lapack_int* sdim, double* alphar, - double* alphai, double* beta, double* vsl, - lapack_int ldvsl, double* vsr, lapack_int ldvsr, - double* work, lapack_int lwork, - lapack_logical* bwork ); -lapack_int LAPACKE_cgges_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_C_SELECT2 selctg, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_int* sdim, lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* vsl, lapack_int ldvsl, - lapack_complex_float* vsr, lapack_int ldvsr, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_logical* bwork ); -lapack_int LAPACKE_zgges_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_Z_SELECT2 selctg, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_int* sdim, lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int ldvsl, - lapack_complex_double* vsr, lapack_int ldvsr, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_logical* bwork ); - -lapack_int LAPACKE_sggesx_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_S_SELECT3 selctg, char sense, - lapack_int n, float* a, lapack_int lda, - float* b, lapack_int ldb, lapack_int* sdim, - float* alphar, float* alphai, float* beta, - float* vsl, lapack_int ldvsl, float* vsr, - lapack_int ldvsr, float* rconde, float* rcondv, - float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork, - lapack_logical* bwork ); -lapack_int LAPACKE_dggesx_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_D_SELECT3 selctg, char sense, - lapack_int n, double* a, lapack_int lda, - double* b, lapack_int ldb, lapack_int* sdim, - double* alphar, double* alphai, double* beta, - double* vsl, lapack_int ldvsl, double* vsr, - lapack_int ldvsr, double* rconde, - double* rcondv, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork, - lapack_logical* bwork ); -lapack_int LAPACKE_cggesx_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_C_SELECT2 selctg, char sense, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_int* sdim, - lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* vsl, lapack_int ldvsl, - lapack_complex_float* vsr, lapack_int ldvsr, - float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int* iwork, - lapack_int liwork, lapack_logical* bwork ); -lapack_int LAPACKE_zggesx_work( int matrix_order, char jobvsl, char jobvsr, - char sort, LAPACK_Z_SELECT2 selctg, char sense, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_int* sdim, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int ldvsl, - lapack_complex_double* vsr, lapack_int ldvsr, - double* rconde, double* rcondv, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int* iwork, - lapack_int liwork, lapack_logical* bwork ); - -lapack_int LAPACKE_sggev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, float* a, lapack_int lda, float* b, - lapack_int ldb, float* alphar, float* alphai, - float* beta, float* vl, lapack_int ldvl, - float* vr, lapack_int ldvr, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dggev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, double* a, lapack_int lda, - double* b, lapack_int ldb, double* alphar, - double* alphai, double* beta, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - double* work, lapack_int lwork ); -lapack_int LAPACKE_cggev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zggev_work( int matrix_order, char jobvl, char jobvr, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_sggevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* alphar, float* alphai, float* beta, - float* vl, lapack_int ldvl, float* vr, - lapack_int ldvr, lapack_int* ilo, - lapack_int* ihi, float* lscale, float* rscale, - float* abnrm, float* bbnrm, float* rconde, - float* rcondv, float* work, lapack_int lwork, - lapack_int* iwork, lapack_logical* bwork ); -lapack_int LAPACKE_dggevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* alphar, double* alphai, double* beta, - double* vl, lapack_int ldvl, double* vr, - lapack_int ldvr, lapack_int* ilo, - lapack_int* ihi, double* lscale, double* rscale, - double* abnrm, double* bbnrm, double* rconde, - double* rcondv, double* work, lapack_int lwork, - lapack_int* iwork, lapack_logical* bwork ); -lapack_int LAPACKE_cggevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale, float* abnrm, float* bbnrm, - float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int* iwork, - lapack_logical* bwork ); -lapack_int LAPACKE_zggevx_work( int matrix_order, char balanc, char jobvl, - char jobvr, char sense, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int* ilo, lapack_int* ihi, - double* lscale, double* rscale, double* abnrm, - double* bbnrm, double* rconde, double* rcondv, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int* iwork, - lapack_logical* bwork ); - -lapack_int LAPACKE_sggglm_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, float* a, lapack_int lda, - float* b, lapack_int ldb, float* d, float* x, - float* y, float* work, lapack_int lwork ); -lapack_int LAPACKE_dggglm_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, double* a, lapack_int lda, - double* b, lapack_int ldb, double* d, double* x, - double* y, double* work, lapack_int lwork ); -lapack_int LAPACKE_cggglm_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* d, - lapack_complex_float* x, - lapack_complex_float* y, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zggglm_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* d, - lapack_complex_double* x, - lapack_complex_double* y, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sgghrd_work( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - float* a, lapack_int lda, float* b, - lapack_int ldb, float* q, lapack_int ldq, - float* z, lapack_int ldz ); -lapack_int LAPACKE_dgghrd_work( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - double* a, lapack_int lda, double* b, - lapack_int ldb, double* q, lapack_int ldq, - double* z, lapack_int ldz ); -lapack_int LAPACKE_cgghrd_work( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz ); -lapack_int LAPACKE_zgghrd_work( int matrix_order, char compq, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz ); - -lapack_int LAPACKE_sgglse_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, float* a, lapack_int lda, - float* b, lapack_int ldb, float* c, float* d, - float* x, float* work, lapack_int lwork ); -lapack_int LAPACKE_dgglse_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, double* a, lapack_int lda, - double* b, lapack_int ldb, double* c, double* d, - double* x, double* work, lapack_int lwork ); -lapack_int LAPACKE_cgglse_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* c, - lapack_complex_float* d, - lapack_complex_float* x, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zgglse_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* c, - lapack_complex_double* d, - lapack_complex_double* x, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sggqrf_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, float* a, lapack_int lda, - float* taua, float* b, lapack_int ldb, - float* taub, float* work, lapack_int lwork ); -lapack_int LAPACKE_dggqrf_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, double* a, lapack_int lda, - double* taua, double* b, lapack_int ldb, - double* taub, double* work, lapack_int lwork ); -lapack_int LAPACKE_cggqrf_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* taua, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* taub, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zggqrf_work( int matrix_order, lapack_int n, lapack_int m, - lapack_int p, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* taua, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* taub, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sggrqf_work( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, float* a, lapack_int lda, - float* taua, float* b, lapack_int ldb, - float* taub, float* work, lapack_int lwork ); -lapack_int LAPACKE_dggrqf_work( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, double* a, lapack_int lda, - double* taua, double* b, lapack_int ldb, - double* taub, double* work, lapack_int lwork ); -lapack_int LAPACKE_cggrqf_work( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* taua, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* taub, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zggrqf_work( int matrix_order, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* taua, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* taub, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_sggsvd_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int n, - lapack_int p, lapack_int* k, lapack_int* l, - float* a, lapack_int lda, float* b, - lapack_int ldb, float* alpha, float* beta, - float* u, lapack_int ldu, float* v, - lapack_int ldv, float* q, lapack_int ldq, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dggsvd_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int n, - lapack_int p, lapack_int* k, lapack_int* l, - double* a, lapack_int lda, double* b, - lapack_int ldb, double* alpha, double* beta, - double* u, lapack_int ldu, double* v, - lapack_int ldv, double* q, lapack_int ldq, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cggsvd_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int n, - lapack_int p, lapack_int* k, lapack_int* l, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - float* alpha, float* beta, - lapack_complex_float* u, lapack_int ldu, - lapack_complex_float* v, lapack_int ldv, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* work, float* rwork, - lapack_int* iwork ); -lapack_int LAPACKE_zggsvd_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int n, - lapack_int p, lapack_int* k, lapack_int* l, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double* alpha, double* beta, - lapack_complex_double* u, lapack_int ldu, - lapack_complex_double* v, lapack_int ldv, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* work, double* rwork, - lapack_int* iwork ); - -lapack_int LAPACKE_sggsvp_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, float* a, lapack_int lda, - float* b, lapack_int ldb, float tola, - float tolb, lapack_int* k, lapack_int* l, - float* u, lapack_int ldu, float* v, - lapack_int ldv, float* q, lapack_int ldq, - lapack_int* iwork, float* tau, float* work ); -lapack_int LAPACKE_dggsvp_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, double* a, lapack_int lda, - double* b, lapack_int ldb, double tola, - double tolb, lapack_int* k, lapack_int* l, - double* u, lapack_int ldu, double* v, - lapack_int ldv, double* q, lapack_int ldq, - lapack_int* iwork, double* tau, double* work ); -lapack_int LAPACKE_cggsvp_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float tola, float tolb, - lapack_int* k, lapack_int* l, - lapack_complex_float* u, lapack_int ldu, - lapack_complex_float* v, lapack_int ldv, - lapack_complex_float* q, lapack_int ldq, - lapack_int* iwork, float* rwork, - lapack_complex_float* tau, - lapack_complex_float* work ); -lapack_int LAPACKE_zggsvp_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, double tola, double tolb, - lapack_int* k, lapack_int* l, - lapack_complex_double* u, lapack_int ldu, - lapack_complex_double* v, lapack_int ldv, - lapack_complex_double* q, lapack_int ldq, - lapack_int* iwork, double* rwork, - lapack_complex_double* tau, - lapack_complex_double* work ); - -lapack_int LAPACKE_sgtcon_work( char norm, lapack_int n, const float* dl, - const float* d, const float* du, - const float* du2, const lapack_int* ipiv, - float anorm, float* rcond, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgtcon_work( char norm, lapack_int n, const double* dl, - const double* d, const double* du, - const double* du2, const lapack_int* ipiv, - double anorm, double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgtcon_work( char norm, lapack_int n, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work ); -lapack_int LAPACKE_zgtcon_work( char norm, lapack_int n, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, - const lapack_int* ipiv, double anorm, - double* rcond, lapack_complex_double* work ); - -lapack_int LAPACKE_sgtrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* dl, - const float* d, const float* du, - const float* dlf, const float* df, - const float* duf, const float* du2, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dgtrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* dl, - const double* d, const double* du, - const double* dlf, const double* df, - const double* duf, const double* du2, - const lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cgtrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* dlf, - const lapack_complex_float* df, - const lapack_complex_float* duf, - const lapack_complex_float* du2, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgtrfs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* dlf, - const lapack_complex_double* df, - const lapack_complex_double* duf, - const lapack_complex_double* du2, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgtsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - float* dl, float* d, float* du, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dgtsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - double* dl, double* d, double* du, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cgtsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_float* dl, - lapack_complex_float* d, - lapack_complex_float* du, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zgtsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - lapack_complex_double* dl, - lapack_complex_double* d, - lapack_complex_double* du, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sgtsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, const float* dl, - const float* d, const float* du, float* dlf, - float* df, float* duf, float* du2, - lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dgtsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, const double* dl, - const double* d, const double* du, double* dlf, - double* df, double* duf, double* du2, - lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cgtsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - lapack_complex_float* dlf, - lapack_complex_float* df, - lapack_complex_float* duf, - lapack_complex_float* du2, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zgtsvx_work( int matrix_order, char fact, char trans, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - lapack_complex_double* dlf, - lapack_complex_double* df, - lapack_complex_double* duf, - lapack_complex_double* du2, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sgttrf_work( lapack_int n, float* dl, float* d, float* du, - float* du2, lapack_int* ipiv ); -lapack_int LAPACKE_dgttrf_work( lapack_int n, double* dl, double* d, double* du, - double* du2, lapack_int* ipiv ); -lapack_int LAPACKE_cgttrf_work( lapack_int n, lapack_complex_float* dl, - lapack_complex_float* d, - lapack_complex_float* du, - lapack_complex_float* du2, lapack_int* ipiv ); -lapack_int LAPACKE_zgttrf_work( lapack_int n, lapack_complex_double* dl, - lapack_complex_double* d, - lapack_complex_double* du, - lapack_complex_double* du2, lapack_int* ipiv ); - -lapack_int LAPACKE_sgttrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const float* dl, - const float* d, const float* du, - const float* du2, const lapack_int* ipiv, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dgttrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const double* dl, - const double* d, const double* du, - const double* du2, const lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cgttrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zgttrs_work( int matrix_order, char trans, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, - const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_chbev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_float* ab, lapack_int ldab, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zhbev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_double* ab, lapack_int ldab, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chbevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_float* ab, lapack_int ldab, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zhbevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_double* ab, lapack_int ldab, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_chbevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int kd, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* q, lapack_int ldq, - float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - float* rwork, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_zhbevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int kd, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* q, lapack_int ldq, - double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - double* rwork, lapack_int* iwork, - lapack_int* ifail ); - -lapack_int LAPACKE_chbgst_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* bb, lapack_int ldbb, - lapack_complex_float* x, lapack_int ldx, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zhbgst_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - const lapack_complex_double* bb, - lapack_int ldbb, lapack_complex_double* x, - lapack_int ldx, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chbgv_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* bb, lapack_int ldbb, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zhbgv_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* bb, lapack_int ldbb, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chbgvd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* bb, lapack_int ldbb, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zhbgvd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* bb, lapack_int ldbb, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_chbgvx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int ka, - lapack_int kb, lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* bb, - lapack_int ldbb, lapack_complex_float* q, - lapack_int ldq, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_zhbgvx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int ka, - lapack_int kb, lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* bb, - lapack_int ldbb, lapack_complex_double* q, - lapack_int ldq, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_chbtrd_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_float* ab, lapack_int ldab, - float* d, float* e, lapack_complex_float* q, - lapack_int ldq, lapack_complex_float* work ); -lapack_int LAPACKE_zhbtrd_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int kd, - lapack_complex_double* ab, lapack_int ldab, - double* d, double* e, lapack_complex_double* q, - lapack_int ldq, lapack_complex_double* work ); - -lapack_int LAPACKE_checon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work ); -lapack_int LAPACKE_zhecon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond, lapack_complex_double* work ); - -lapack_int LAPACKE_cheequb_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax, - lapack_complex_float* work ); -lapack_int LAPACKE_zheequb_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax, - lapack_complex_double* work ); - -lapack_int LAPACKE_cheev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float* w, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zheev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double* w, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_cheevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, float* w, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_zheevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, double* w, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_cheevr_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_int* isuppz, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_zheevr_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_int* isuppz, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_cheevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_zheevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_chegst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhegst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_chegv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb, float* w, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zhegv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double* w, lapack_complex_double* work, - lapack_int lwork, double* rwork ); - -lapack_int LAPACKE_chegvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - float* w, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zhegvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double* w, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_chegvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_zhegvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_cherfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zherfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_cherfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const float* s, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zherfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* s, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chesv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zhesv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_chesvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - lapack_int lwork, float* rwork ); -lapack_int LAPACKE_zhesvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_chesvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zhesvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chetrd_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - float* d, float* e, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zhetrd_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - double* d, double* e, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_chetrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_float* work, - lapack_int lwork ); -lapack_int LAPACKE_zhetrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* work, - lapack_int lwork ); - -lapack_int LAPACKE_chetri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work ); -lapack_int LAPACKE_zhetri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work ); - -lapack_int LAPACKE_chetrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zhetrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_chfrk_work( int matrix_order, char transr, char uplo, - char trans, lapack_int n, lapack_int k, - float alpha, const lapack_complex_float* a, - lapack_int lda, float beta, - lapack_complex_float* c ); -lapack_int LAPACKE_zhfrk_work( int matrix_order, char transr, char uplo, - char trans, lapack_int n, lapack_int k, - double alpha, const lapack_complex_double* a, - lapack_int lda, double beta, - lapack_complex_double* c ); - -lapack_int LAPACKE_shgeqz_work( int matrix_order, char job, char compq, - char compz, lapack_int n, lapack_int ilo, - lapack_int ihi, float* h, lapack_int ldh, - float* t, lapack_int ldt, float* alphar, - float* alphai, float* beta, float* q, - lapack_int ldq, float* z, lapack_int ldz, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dhgeqz_work( int matrix_order, char job, char compq, - char compz, lapack_int n, lapack_int ilo, - lapack_int ihi, double* h, lapack_int ldh, - double* t, lapack_int ldt, double* alphar, - double* alphai, double* beta, double* q, - lapack_int ldq, double* z, lapack_int ldz, - double* work, lapack_int lwork ); -lapack_int LAPACKE_chgeqz_work( int matrix_order, char job, char compq, - char compz, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_float* h, - lapack_int ldh, lapack_complex_float* t, - lapack_int ldt, lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, lapack_int lwork, - float* rwork ); -lapack_int LAPACKE_zhgeqz_work( int matrix_order, char job, char compq, - char compz, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_double* h, - lapack_int ldh, lapack_complex_double* t, - lapack_int ldt, lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_chpcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work ); -lapack_int LAPACKE_zhpcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_int* ipiv, double anorm, - double* rcond, lapack_complex_double* work ); - -lapack_int LAPACKE_chpev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_float* ap, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zhpev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_double* ap, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_chpevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_float* ap, - float* w, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zhpevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_complex_double* ap, - double* w, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_chpevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_float* ap, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_zhpevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, - lapack_complex_double* ap, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_chpgst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_float* ap, - const lapack_complex_float* bp ); -lapack_int LAPACKE_zhpgst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, lapack_complex_double* ap, - const lapack_complex_double* bp ); - -lapack_int LAPACKE_chpgv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_float* ap, - lapack_complex_float* bp, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zhpgv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_double* ap, - lapack_complex_double* bp, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_chpgvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_float* ap, - lapack_complex_float* bp, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_zhpgvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, - lapack_complex_double* ap, - lapack_complex_double* bp, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int lrwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_chpgvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_float* ap, - lapack_complex_float* bp, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_zhpgvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, - lapack_complex_double* ap, - lapack_complex_double* bp, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_chprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zhprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_chpsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhpsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_chpsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - lapack_complex_float* afp, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zhpsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - lapack_complex_double* afp, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_chptrd_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, float* d, float* e, - lapack_complex_float* tau ); -lapack_int LAPACKE_zhptrd_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, double* d, double* e, - lapack_complex_double* tau ); - -lapack_int LAPACKE_chptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, lapack_int* ipiv ); -lapack_int LAPACKE_zhptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, lapack_int* ipiv ); - -lapack_int LAPACKE_chptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, - const lapack_int* ipiv, - lapack_complex_float* work ); -lapack_int LAPACKE_zhptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, - const lapack_int* ipiv, - lapack_complex_double* work ); - -lapack_int LAPACKE_chptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zhptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_shsein_work( int matrix_order, char job, char eigsrc, - char initv, lapack_logical* select, - lapack_int n, const float* h, lapack_int ldh, - float* wr, const float* wi, float* vl, - lapack_int ldvl, float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, float* work, - lapack_int* ifaill, lapack_int* ifailr ); -lapack_int LAPACKE_dhsein_work( int matrix_order, char job, char eigsrc, - char initv, lapack_logical* select, - lapack_int n, const double* h, lapack_int ldh, - double* wr, const double* wi, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, double* work, - lapack_int* ifaill, lapack_int* ifailr ); -lapack_int LAPACKE_chsein_work( int matrix_order, char job, char eigsrc, - char initv, const lapack_logical* select, - lapack_int n, const lapack_complex_float* h, - lapack_int ldh, lapack_complex_float* w, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_float* work, float* rwork, - lapack_int* ifaill, lapack_int* ifailr ); -lapack_int LAPACKE_zhsein_work( int matrix_order, char job, char eigsrc, - char initv, const lapack_logical* select, - lapack_int n, const lapack_complex_double* h, - lapack_int ldh, lapack_complex_double* w, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_double* work, double* rwork, - lapack_int* ifaill, lapack_int* ifailr ); - -lapack_int LAPACKE_shseqr_work( int matrix_order, char job, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - float* h, lapack_int ldh, float* wr, float* wi, - float* z, lapack_int ldz, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dhseqr_work( int matrix_order, char job, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - double* h, lapack_int ldh, double* wr, - double* wi, double* z, lapack_int ldz, - double* work, lapack_int lwork ); -lapack_int LAPACKE_chseqr_work( int matrix_order, char job, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_float* h, lapack_int ldh, - lapack_complex_float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zhseqr_work( int matrix_order, char job, char compz, - lapack_int n, lapack_int ilo, lapack_int ihi, - lapack_complex_double* h, lapack_int ldh, - lapack_complex_double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_clacgv_work( lapack_int n, lapack_complex_float* x, - lapack_int incx ); -lapack_int LAPACKE_zlacgv_work( lapack_int n, lapack_complex_double* x, - lapack_int incx ); - -lapack_int LAPACKE_slacpy_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, const float* a, lapack_int lda, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dlacpy_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, const double* a, lapack_int lda, - double* b, lapack_int ldb ); -lapack_int LAPACKE_clacpy_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zlacpy_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_zlag2c_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_float* sa, lapack_int ldsa ); - -lapack_int LAPACKE_slag2d_work( int matrix_order, lapack_int m, lapack_int n, - const float* sa, lapack_int ldsa, double* a, - lapack_int lda ); - -lapack_int LAPACKE_dlag2s_work( int matrix_order, lapack_int m, lapack_int n, - const double* a, lapack_int lda, float* sa, - lapack_int ldsa ); - -lapack_int LAPACKE_clag2z_work( int matrix_order, lapack_int m, lapack_int n, - const lapack_complex_float* sa, lapack_int ldsa, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_slagge_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* d, - float* a, lapack_int lda, lapack_int* iseed, - float* work ); -lapack_int LAPACKE_dlagge_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* d, - double* a, lapack_int lda, lapack_int* iseed, - double* work ); -lapack_int LAPACKE_clagge_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const float* d, - lapack_complex_float* a, lapack_int lda, - lapack_int* iseed, lapack_complex_float* work ); -lapack_int LAPACKE_zlagge_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int kl, lapack_int ku, const double* d, - lapack_complex_double* a, lapack_int lda, - lapack_int* iseed, - lapack_complex_double* work ); - -lapack_int LAPACKE_claghe_work( int matrix_order, lapack_int n, lapack_int k, - const float* d, lapack_complex_float* a, - lapack_int lda, lapack_int* iseed, - lapack_complex_float* work ); -lapack_int LAPACKE_zlaghe_work( int matrix_order, lapack_int n, lapack_int k, - const double* d, lapack_complex_double* a, - lapack_int lda, lapack_int* iseed, - lapack_complex_double* work ); - -lapack_int LAPACKE_slagsy_work( int matrix_order, lapack_int n, lapack_int k, - const float* d, float* a, lapack_int lda, - lapack_int* iseed, float* work ); -lapack_int LAPACKE_dlagsy_work( int matrix_order, lapack_int n, lapack_int k, - const double* d, double* a, lapack_int lda, - lapack_int* iseed, double* work ); -lapack_int LAPACKE_clagsy_work( int matrix_order, lapack_int n, lapack_int k, - const float* d, lapack_complex_float* a, - lapack_int lda, lapack_int* iseed, - lapack_complex_float* work ); -lapack_int LAPACKE_zlagsy_work( int matrix_order, lapack_int n, lapack_int k, - const double* d, lapack_complex_double* a, - lapack_int lda, lapack_int* iseed, - lapack_complex_double* work ); - -lapack_int LAPACKE_slapmr_work( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, float* x, - lapack_int ldx, lapack_int* k ); -lapack_int LAPACKE_dlapmr_work( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, double* x, - lapack_int ldx, lapack_int* k ); -lapack_int LAPACKE_clapmr_work( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, - lapack_complex_float* x, lapack_int ldx, - lapack_int* k ); -lapack_int LAPACKE_zlapmr_work( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, - lapack_complex_double* x, lapack_int ldx, - lapack_int* k ); - -lapack_int LAPACKE_slartgp_work( float f, float g, float* cs, float* sn, - float* r ); -lapack_int LAPACKE_dlartgp_work( double f, double g, double* cs, double* sn, - double* r ); - -lapack_int LAPACKE_slartgs_work( float x, float y, float sigma, float* cs, - float* sn ); -lapack_int LAPACKE_dlartgs_work( double x, double y, double sigma, double* cs, - double* sn ); - -float LAPACKE_slapy2_work( float x, float y ); -double LAPACKE_dlapy2_work( double x, double y ); - -float LAPACKE_slapy3_work( float x, float y, float z ); -double LAPACKE_dlapy3_work( double x, double y, double z ); - -float LAPACKE_slamch_work( char cmach ); -double LAPACKE_dlamch_work( char cmach ); - -float LAPACKE_slange_work( int matrix_order, char norm, lapack_int m, - lapack_int n, const float* a, lapack_int lda, - float* work ); -double LAPACKE_dlange_work( int matrix_order, char norm, lapack_int m, - lapack_int n, const double* a, lapack_int lda, - double* work ); -float LAPACKE_clange_work( int matrix_order, char norm, lapack_int m, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, float* work ); -double LAPACKE_zlange_work( int matrix_order, char norm, lapack_int m, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, double* work ); - -float LAPACKE_clanhe_work( int matrix_order, char norm, char uplo, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, float* work ); -double LAPACKE_zlanhe_work( int matrix_order, char norm, char uplo, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, double* work ); - -float LAPACKE_slansy_work( int matrix_order, char norm, char uplo, - lapack_int n, const float* a, lapack_int lda, - float* work ); -double LAPACKE_dlansy_work( int matrix_order, char norm, char uplo, - lapack_int n, const double* a, lapack_int lda, - double* work ); -float LAPACKE_clansy_work( int matrix_order, char norm, char uplo, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, float* work ); -double LAPACKE_zlansy_work( int matrix_order, char norm, char uplo, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, double* work ); - -float LAPACKE_slantr_work( int matrix_order, char norm, char uplo, - char diag, lapack_int m, lapack_int n, const float* a, - lapack_int lda, float* work ); -double LAPACKE_dlantr_work( int matrix_order, char norm, char uplo, - char diag, lapack_int m, lapack_int n, - const double* a, lapack_int lda, double* work ); -float LAPACKE_clantr_work( int matrix_order, char norm, char uplo, - char diag, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* work ); -double LAPACKE_zlantr_work( int matrix_order, char norm, char uplo, - char diag, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* work ); - -lapack_int LAPACKE_slarfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, const float* v, - lapack_int ldv, const float* t, lapack_int ldt, - float* c, lapack_int ldc, float* work, - lapack_int ldwork ); -lapack_int LAPACKE_dlarfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, const double* v, - lapack_int ldv, const double* t, lapack_int ldt, - double* c, lapack_int ldc, double* work, - lapack_int ldwork ); -lapack_int LAPACKE_clarfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int ldwork ); -lapack_int LAPACKE_zlarfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, - lapack_int ldwork ); - -lapack_int LAPACKE_slarfg_work( lapack_int n, float* alpha, float* x, - lapack_int incx, float* tau ); -lapack_int LAPACKE_dlarfg_work( lapack_int n, double* alpha, double* x, - lapack_int incx, double* tau ); -lapack_int LAPACKE_clarfg_work( lapack_int n, lapack_complex_float* alpha, - lapack_complex_float* x, lapack_int incx, - lapack_complex_float* tau ); -lapack_int LAPACKE_zlarfg_work( lapack_int n, lapack_complex_double* alpha, - lapack_complex_double* x, lapack_int incx, - lapack_complex_double* tau ); - -lapack_int LAPACKE_slarft_work( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, const float* v, - lapack_int ldv, const float* tau, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dlarft_work( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, const double* v, - lapack_int ldv, const double* tau, double* t, - lapack_int ldt ); -lapack_int LAPACKE_clarft_work( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* tau, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zlarft_work( int matrix_order, char direct, char storev, - lapack_int n, lapack_int k, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* tau, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_slarfx_work( int matrix_order, char side, lapack_int m, - lapack_int n, const float* v, float tau, - float* c, lapack_int ldc, float* work ); -lapack_int LAPACKE_dlarfx_work( int matrix_order, char side, lapack_int m, - lapack_int n, const double* v, double tau, - double* c, lapack_int ldc, double* work ); -lapack_int LAPACKE_clarfx_work( int matrix_order, char side, lapack_int m, - lapack_int n, const lapack_complex_float* v, - lapack_complex_float tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work ); -lapack_int LAPACKE_zlarfx_work( int matrix_order, char side, lapack_int m, - lapack_int n, const lapack_complex_double* v, - lapack_complex_double tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work ); - -lapack_int LAPACKE_slarnv_work( lapack_int idist, lapack_int* iseed, - lapack_int n, float* x ); -lapack_int LAPACKE_dlarnv_work( lapack_int idist, lapack_int* iseed, - lapack_int n, double* x ); -lapack_int LAPACKE_clarnv_work( lapack_int idist, lapack_int* iseed, - lapack_int n, lapack_complex_float* x ); -lapack_int LAPACKE_zlarnv_work( lapack_int idist, lapack_int* iseed, - lapack_int n, lapack_complex_double* x ); - -lapack_int LAPACKE_slaset_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, float alpha, float beta, float* a, - lapack_int lda ); -lapack_int LAPACKE_dlaset_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, double alpha, double beta, - double* a, lapack_int lda ); -lapack_int LAPACKE_claset_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, lapack_complex_float alpha, - lapack_complex_float beta, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zlaset_work( int matrix_order, char uplo, lapack_int m, - lapack_int n, lapack_complex_double alpha, - lapack_complex_double beta, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_slasrt_work( char id, lapack_int n, float* d ); -lapack_int LAPACKE_dlasrt_work( char id, lapack_int n, double* d ); - -lapack_int LAPACKE_slaswp_work( int matrix_order, lapack_int n, float* a, - lapack_int lda, lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); -lapack_int LAPACKE_dlaswp_work( int matrix_order, lapack_int n, double* a, - lapack_int lda, lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); -lapack_int LAPACKE_claswp_work( int matrix_order, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); -lapack_int LAPACKE_zlaswp_work( int matrix_order, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int k1, lapack_int k2, - const lapack_int* ipiv, lapack_int incx ); - -lapack_int LAPACKE_slatms_work( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, - float* d, lapack_int mode, float cond, - float dmax, lapack_int kl, lapack_int ku, - char pack, float* a, lapack_int lda, - float* work ); -lapack_int LAPACKE_dlatms_work( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, - double* d, lapack_int mode, double cond, - double dmax, lapack_int kl, lapack_int ku, - char pack, double* a, lapack_int lda, - double* work ); -lapack_int LAPACKE_clatms_work( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, - float* d, lapack_int mode, float cond, - float dmax, lapack_int kl, lapack_int ku, - char pack, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* work ); -lapack_int LAPACKE_zlatms_work( int matrix_order, lapack_int m, lapack_int n, - char dist, lapack_int* iseed, char sym, - double* d, lapack_int mode, double cond, - double dmax, lapack_int kl, lapack_int ku, - char pack, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* work ); - -lapack_int LAPACKE_slauum_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda ); -lapack_int LAPACKE_dlauum_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda ); -lapack_int LAPACKE_clauum_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zlauum_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_sopgtr_work( int matrix_order, char uplo, lapack_int n, - const float* ap, const float* tau, float* q, - lapack_int ldq, float* work ); -lapack_int LAPACKE_dopgtr_work( int matrix_order, char uplo, lapack_int n, - const double* ap, const double* tau, double* q, - lapack_int ldq, double* work ); - -lapack_int LAPACKE_sopmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const float* ap, const float* tau, float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_dopmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const double* ap, const double* tau, double* c, - lapack_int ldc, double* work ); - -lapack_int LAPACKE_sorgbr_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, float* a, - lapack_int lda, const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorgbr_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, double* a, - lapack_int lda, const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorghr_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, float* a, lapack_int lda, - const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorghr_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, double* a, lapack_int lda, - const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorglq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorglq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorgql_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorgql_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorgqr_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorgqr_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorgrq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, float* a, lapack_int lda, - const float* tau, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dorgrq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, double* a, lapack_int lda, - const double* tau, double* work, - lapack_int lwork ); - -lapack_int LAPACKE_sorgtr_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, const float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dorgtr_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, const double* tau, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormbr_work( int matrix_order, char vect, char side, - char trans, lapack_int m, lapack_int n, - lapack_int k, const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormbr_work( int matrix_order, char vect, char side, - char trans, lapack_int m, lapack_int n, - lapack_int k, const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormhr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormhr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormlq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormlq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormql_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormql_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormqr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormqr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormrq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormrq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormrz_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormrz_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_sormtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const float* a, lapack_int lda, - const float* tau, float* c, lapack_int ldc, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dormtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const double* a, lapack_int lda, - const double* tau, double* c, lapack_int ldc, - double* work, lapack_int lwork ); - -lapack_int LAPACKE_spbcon_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const float* ab, lapack_int ldab, - float anorm, float* rcond, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dpbcon_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const double* ab, - lapack_int ldab, double anorm, double* rcond, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cpbcon_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_float* ab, - lapack_int ldab, float anorm, float* rcond, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zpbcon_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_double* ab, - lapack_int ldab, double anorm, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_spbequ_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const float* ab, lapack_int ldab, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_dpbequ_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const double* ab, - lapack_int ldab, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cpbequ_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_float* ab, - lapack_int ldab, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_zpbequ_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, const lapack_complex_double* ab, - lapack_int ldab, double* s, double* scond, - double* amax ); - -lapack_int LAPACKE_spbrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const float* ab, - lapack_int ldab, const float* afb, - lapack_int ldafb, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dpbrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const double* ab, lapack_int ldab, - const double* afb, lapack_int ldafb, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cpbrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - const lapack_complex_float* afb, - lapack_int ldafb, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zpbrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, - const lapack_complex_double* afb, - lapack_int ldafb, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_spbstf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kb, float* bb, lapack_int ldbb ); -lapack_int LAPACKE_dpbstf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kb, double* bb, lapack_int ldbb ); -lapack_int LAPACKE_cpbstf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kb, lapack_complex_float* bb, - lapack_int ldbb ); -lapack_int LAPACKE_zpbstf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kb, lapack_complex_double* bb, - lapack_int ldbb ); - -lapack_int LAPACKE_spbsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, float* ab, - lapack_int ldab, float* b, lapack_int ldb ); -lapack_int LAPACKE_dpbsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, double* ab, - lapack_int ldab, double* b, lapack_int ldb ); -lapack_int LAPACKE_cpbsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpbsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spbsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int kd, lapack_int nrhs, - float* ab, lapack_int ldab, float* afb, - lapack_int ldafb, char* equed, float* s, - float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dpbsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int kd, lapack_int nrhs, - double* ab, lapack_int ldab, double* afb, - lapack_int ldafb, char* equed, double* s, - double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, double* work, lapack_int* iwork ); -lapack_int LAPACKE_cpbsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int kd, lapack_int nrhs, - lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* afb, lapack_int ldafb, - char* equed, float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zpbsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int kd, lapack_int nrhs, - lapack_complex_double* ab, lapack_int ldab, - lapack_complex_double* afb, lapack_int ldafb, - char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_spbtrf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, float* ab, lapack_int ldab ); -lapack_int LAPACKE_dpbtrf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, double* ab, lapack_int ldab ); -lapack_int LAPACKE_cpbtrf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_complex_float* ab, - lapack_int ldab ); -lapack_int LAPACKE_zpbtrf_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_complex_double* ab, - lapack_int ldab ); - -lapack_int LAPACKE_spbtrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, const float* ab, - lapack_int ldab, float* b, lapack_int ldb ); -lapack_int LAPACKE_dpbtrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const double* ab, lapack_int ldab, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cpbtrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_float* ab, lapack_int ldab, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpbtrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int kd, lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_spftrf_work( int matrix_order, char transr, char uplo, - lapack_int n, float* a ); -lapack_int LAPACKE_dpftrf_work( int matrix_order, char transr, char uplo, - lapack_int n, double* a ); -lapack_int LAPACKE_cpftrf_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_float* a ); -lapack_int LAPACKE_zpftrf_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_double* a ); - -lapack_int LAPACKE_spftri_work( int matrix_order, char transr, char uplo, - lapack_int n, float* a ); -lapack_int LAPACKE_dpftri_work( int matrix_order, char transr, char uplo, - lapack_int n, double* a ); -lapack_int LAPACKE_cpftri_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_float* a ); -lapack_int LAPACKE_zpftri_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_complex_double* a ); - -lapack_int LAPACKE_spftrs_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, const float* a, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dpftrs_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, const double* a, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cpftrs_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpftrs_work( int matrix_order, char transr, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spocon_work( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float anorm, - float* rcond, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dpocon_work( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double anorm, - double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cpocon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float anorm, float* rcond, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zpocon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double anorm, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_spoequ_work( int matrix_order, lapack_int n, const float* a, - lapack_int lda, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dpoequ_work( int matrix_order, lapack_int n, const double* a, - lapack_int lda, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cpoequ_work( int matrix_order, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zpoequ_work( int matrix_order, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_spoequb_work( int matrix_order, lapack_int n, const float* a, - lapack_int lda, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dpoequb_work( int matrix_order, lapack_int n, - const double* a, lapack_int lda, double* s, - double* scond, double* amax ); -lapack_int LAPACKE_cpoequb_work( int matrix_order, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax ); -lapack_int LAPACKE_zpoequb_work( int matrix_order, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax ); - -lapack_int LAPACKE_sporfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dporfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cporfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zporfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sporfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, - lapack_int ldaf, const float* s, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dporfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const double* s, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cporfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, - lapack_int ldaf, const float* s, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zporfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, - lapack_int ldaf, const double* s, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); -lapack_int LAPACKE_dsposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - double* b, lapack_int ldb, double* x, - lapack_int ldx, double* work, float* swork, - lapack_int* iter ); -lapack_int LAPACKE_zcposv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, lapack_complex_double* work, - lapack_complex_float* swork, double* rwork, - lapack_int* iter ); - -lapack_int LAPACKE_sposvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - char* equed, float* s, float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dposvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - char* equed, double* s, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cposvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - char* equed, float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zposvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sposvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - char* equed, float* s, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dposvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - char* equed, double* s, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cposvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - char* equed, float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zposvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_spotrf_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda ); -lapack_int LAPACKE_dpotrf_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda ); -lapack_int LAPACKE_cpotrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zpotrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_spotri_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda ); -lapack_int LAPACKE_dpotri_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda ); -lapack_int LAPACKE_cpotri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zpotri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_spotrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dpotrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, double* b, lapack_int ldb ); -lapack_int LAPACKE_cpotrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zpotrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sppcon_work( int matrix_order, char uplo, lapack_int n, - const float* ap, float anorm, float* rcond, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dppcon_work( int matrix_order, char uplo, lapack_int n, - const double* ap, double anorm, double* rcond, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cppcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, float anorm, - float* rcond, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zppcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, double anorm, - double* rcond, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_sppequ_work( int matrix_order, char uplo, lapack_int n, - const float* ap, float* s, float* scond, - float* amax ); -lapack_int LAPACKE_dppequ_work( int matrix_order, char uplo, lapack_int n, - const double* ap, double* s, double* scond, - double* amax ); -lapack_int LAPACKE_cppequ_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, float* s, - float* scond, float* amax ); -lapack_int LAPACKE_zppequ_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, double* s, - double* scond, double* amax ); - -lapack_int LAPACKE_spprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, - const float* afp, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dpprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, - const double* afp, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cpprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zpprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sppsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* ap, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dppsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* ap, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cppsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zppsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sppsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* ap, - float* afp, char* equed, float* s, float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dppsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* ap, - double* afp, char* equed, double* s, double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cppsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* ap, - lapack_complex_float* afp, char* equed, - float* s, lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zppsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* ap, - lapack_complex_double* afp, char* equed, - double* s, lapack_complex_double* b, - lapack_int ldb, lapack_complex_double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_spptrf_work( int matrix_order, char uplo, lapack_int n, - float* ap ); -lapack_int LAPACKE_dpptrf_work( int matrix_order, char uplo, lapack_int n, - double* ap ); -lapack_int LAPACKE_cpptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap ); -lapack_int LAPACKE_zpptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap ); - -lapack_int LAPACKE_spptri_work( int matrix_order, char uplo, lapack_int n, - float* ap ); -lapack_int LAPACKE_dpptri_work( int matrix_order, char uplo, lapack_int n, - double* ap ); -lapack_int LAPACKE_cpptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap ); -lapack_int LAPACKE_zpptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap ); - -lapack_int LAPACKE_spptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dpptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cpptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_spstrf_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, lapack_int* piv, - lapack_int* rank, float tol, float* work ); -lapack_int LAPACKE_dpstrf_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, lapack_int* piv, - lapack_int* rank, double tol, double* work ); -lapack_int LAPACKE_cpstrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* piv, lapack_int* rank, float tol, - float* work ); -lapack_int LAPACKE_zpstrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* piv, lapack_int* rank, double tol, - double* work ); - -lapack_int LAPACKE_sptcon_work( lapack_int n, const float* d, const float* e, - float anorm, float* rcond, float* work ); -lapack_int LAPACKE_dptcon_work( lapack_int n, const double* d, const double* e, - double anorm, double* rcond, double* work ); -lapack_int LAPACKE_cptcon_work( lapack_int n, const float* d, - const lapack_complex_float* e, float anorm, - float* rcond, float* work ); -lapack_int LAPACKE_zptcon_work( lapack_int n, const double* d, - const lapack_complex_double* e, double anorm, - double* rcond, double* work ); - -lapack_int LAPACKE_spteqr_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, float* z, lapack_int ldz, - float* work ); -lapack_int LAPACKE_dpteqr_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz, - double* work ); -lapack_int LAPACKE_cpteqr_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, lapack_complex_float* z, - lapack_int ldz, float* work ); -lapack_int LAPACKE_zpteqr_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz, double* work ); - -lapack_int LAPACKE_sptrfs_work( int matrix_order, lapack_int n, lapack_int nrhs, - const float* d, const float* e, const float* df, - const float* ef, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* ferr, - float* berr, float* work ); -lapack_int LAPACKE_dptrfs_work( int matrix_order, lapack_int n, lapack_int nrhs, - const double* d, const double* e, - const double* df, const double* ef, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr, - double* work ); -lapack_int LAPACKE_cptrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, const float* df, - const lapack_complex_float* ef, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zptrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, - const double* df, - const lapack_complex_double* ef, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sptsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - float* d, float* e, float* b, lapack_int ldb ); -lapack_int LAPACKE_dptsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - double* d, double* e, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cptsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - float* d, lapack_complex_float* e, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zptsv_work( int matrix_order, lapack_int n, lapack_int nrhs, - double* d, lapack_complex_double* e, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sptsvx_work( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const float* d, const float* e, - float* df, float* ef, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work ); -lapack_int LAPACKE_dptsvx_work( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const double* d, - const double* e, double* df, double* ef, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* ferr, - double* berr, double* work ); -lapack_int LAPACKE_cptsvx_work( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, float* df, - lapack_complex_float* ef, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zptsvx_work( int matrix_order, char fact, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, double* df, - lapack_complex_double* ef, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_spttrf_work( lapack_int n, float* d, float* e ); -lapack_int LAPACKE_dpttrf_work( lapack_int n, double* d, double* e ); -lapack_int LAPACKE_cpttrf_work( lapack_int n, float* d, - lapack_complex_float* e ); -lapack_int LAPACKE_zpttrf_work( lapack_int n, double* d, - lapack_complex_double* e ); - -lapack_int LAPACKE_spttrs_work( int matrix_order, lapack_int n, lapack_int nrhs, - const float* d, const float* e, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dpttrs_work( int matrix_order, lapack_int n, lapack_int nrhs, - const double* d, const double* e, double* b, - lapack_int ldb ); -lapack_int LAPACKE_cpttrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* d, - const lapack_complex_float* e, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zpttrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* d, - const lapack_complex_double* e, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_ssbev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, float* ab, - lapack_int ldab, float* w, float* z, - lapack_int ldz, float* work ); -lapack_int LAPACKE_dsbev_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, double* ab, - lapack_int ldab, double* w, double* z, - lapack_int ldz, double* work ); - -lapack_int LAPACKE_ssbevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, float* ab, - lapack_int ldab, float* w, float* z, - lapack_int ldz, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dsbevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int kd, double* ab, - lapack_int ldab, double* w, double* z, - lapack_int ldz, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_ssbevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int kd, - float* ab, lapack_int ldab, float* q, - lapack_int ldq, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, - lapack_int ldz, float* work, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_dsbevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int kd, - double* ab, lapack_int ldab, double* q, - lapack_int ldq, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, double* z, - lapack_int ldz, double* work, lapack_int* iwork, - lapack_int* ifail ); - -lapack_int LAPACKE_ssbgst_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - float* ab, lapack_int ldab, const float* bb, - lapack_int ldbb, float* x, lapack_int ldx, - float* work ); -lapack_int LAPACKE_dsbgst_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - double* ab, lapack_int ldab, const double* bb, - lapack_int ldbb, double* x, lapack_int ldx, - double* work ); - -lapack_int LAPACKE_ssbgv_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - float* ab, lapack_int ldab, float* bb, - lapack_int ldbb, float* w, float* z, - lapack_int ldz, float* work ); -lapack_int LAPACKE_dsbgv_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - double* ab, lapack_int ldab, double* bb, - lapack_int ldbb, double* w, double* z, - lapack_int ldz, double* work ); - -lapack_int LAPACKE_ssbgvd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - float* ab, lapack_int ldab, float* bb, - lapack_int ldbb, float* w, float* z, - lapack_int ldz, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dsbgvd_work( int matrix_order, char jobz, char uplo, - lapack_int n, lapack_int ka, lapack_int kb, - double* ab, lapack_int ldab, double* bb, - lapack_int ldbb, double* w, double* z, - lapack_int ldz, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_ssbgvx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int ka, - lapack_int kb, float* ab, lapack_int ldab, - float* bb, lapack_int ldbb, float* q, - lapack_int ldq, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, - lapack_int ldz, float* work, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_dsbgvx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, lapack_int ka, - lapack_int kb, double* ab, lapack_int ldab, - double* bb, lapack_int ldbb, double* q, - lapack_int ldq, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, double* z, - lapack_int ldz, double* work, lapack_int* iwork, - lapack_int* ifail ); - -lapack_int LAPACKE_ssbtrd_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int kd, float* ab, - lapack_int ldab, float* d, float* e, float* q, - lapack_int ldq, float* work ); -lapack_int LAPACKE_dsbtrd_work( int matrix_order, char vect, char uplo, - lapack_int n, lapack_int kd, double* ab, - lapack_int ldab, double* d, double* e, - double* q, lapack_int ldq, double* work ); - -lapack_int LAPACKE_ssfrk_work( int matrix_order, char transr, char uplo, - char trans, lapack_int n, lapack_int k, - float alpha, const float* a, lapack_int lda, - float beta, float* c ); -lapack_int LAPACKE_dsfrk_work( int matrix_order, char transr, char uplo, - char trans, lapack_int n, lapack_int k, - double alpha, const double* a, lapack_int lda, - double beta, double* c ); - -lapack_int LAPACKE_sspcon_work( int matrix_order, char uplo, lapack_int n, - const float* ap, const lapack_int* ipiv, - float anorm, float* rcond, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dspcon_work( int matrix_order, char uplo, lapack_int n, - const double* ap, const lapack_int* ipiv, - double anorm, double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_cspcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work ); -lapack_int LAPACKE_zspcon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_int* ipiv, double anorm, - double* rcond, lapack_complex_double* work ); - -lapack_int LAPACKE_sspev_work( int matrix_order, char jobz, char uplo, - lapack_int n, float* ap, float* w, float* z, - lapack_int ldz, float* work ); -lapack_int LAPACKE_dspev_work( int matrix_order, char jobz, char uplo, - lapack_int n, double* ap, double* w, double* z, - lapack_int ldz, double* work ); - -lapack_int LAPACKE_sspevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, float* ap, float* w, float* z, - lapack_int ldz, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dspevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, double* ap, double* w, double* z, - lapack_int ldz, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_sspevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, float* ap, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, float* z, - lapack_int ldz, float* work, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_dspevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, double* ap, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - double* z, lapack_int ldz, double* work, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_sspgst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, float* ap, const float* bp ); -lapack_int LAPACKE_dspgst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, double* ap, const double* bp ); - -lapack_int LAPACKE_sspgv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* ap, float* bp, - float* w, float* z, lapack_int ldz, - float* work ); -lapack_int LAPACKE_dspgv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* ap, double* bp, - double* w, double* z, lapack_int ldz, - double* work ); - -lapack_int LAPACKE_sspgvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* ap, float* bp, - float* w, float* z, lapack_int ldz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_dspgvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* ap, double* bp, - double* w, double* z, lapack_int ldz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_sspgvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, float* ap, - float* bp, float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, float* z, lapack_int ldz, float* work, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_dspgvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, double* ap, - double* bp, double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, double* z, lapack_int ldz, - double* work, lapack_int* iwork, - lapack_int* ifail ); - -lapack_int LAPACKE_ssprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, - const float* afp, const lapack_int* ipiv, - const float* b, lapack_int ldb, float* x, - lapack_int ldx, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dsprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, - const double* afp, const lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_csprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zsprfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_sspsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* ap, lapack_int* ipiv, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dspsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* ap, lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_cspsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* ap, - lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zspsv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* ap, - lapack_int* ipiv, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_sspsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, const float* ap, - float* afp, lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dspsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, const double* ap, - double* afp, lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_cspsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - lapack_complex_float* afp, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zspsvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - lapack_complex_double* afp, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_ssptrd_work( int matrix_order, char uplo, lapack_int n, - float* ap, float* d, float* e, float* tau ); -lapack_int LAPACKE_dsptrd_work( int matrix_order, char uplo, lapack_int n, - double* ap, double* d, double* e, double* tau ); - -lapack_int LAPACKE_ssptrf_work( int matrix_order, char uplo, lapack_int n, - float* ap, lapack_int* ipiv ); -lapack_int LAPACKE_dsptrf_work( int matrix_order, char uplo, lapack_int n, - double* ap, lapack_int* ipiv ); -lapack_int LAPACKE_csptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, lapack_int* ipiv ); -lapack_int LAPACKE_zsptrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, lapack_int* ipiv ); - -lapack_int LAPACKE_ssptri_work( int matrix_order, char uplo, lapack_int n, - float* ap, const lapack_int* ipiv, - float* work ); -lapack_int LAPACKE_dsptri_work( int matrix_order, char uplo, lapack_int n, - double* ap, const lapack_int* ipiv, - double* work ); -lapack_int LAPACKE_csptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* ap, - const lapack_int* ipiv, - lapack_complex_float* work ); -lapack_int LAPACKE_zsptri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* ap, - const lapack_int* ipiv, - lapack_complex_double* work ); - -lapack_int LAPACKE_ssptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* ap, - const lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dsptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* ap, - const lapack_int* ipiv, double* b, - lapack_int ldb ); -lapack_int LAPACKE_csptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_zsptrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_sstebz_work( char range, char order, lapack_int n, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, const float* d, const float* e, - lapack_int* m, lapack_int* nsplit, float* w, - lapack_int* iblock, lapack_int* isplit, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dstebz_work( char range, char order, lapack_int n, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, const double* d, const double* e, - lapack_int* m, lapack_int* nsplit, double* w, - lapack_int* iblock, lapack_int* isplit, - double* work, lapack_int* iwork ); - -lapack_int LAPACKE_sstedc_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, float* z, lapack_int ldz, - float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dstedc_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_cstedc_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, lapack_complex_float* z, - lapack_int ldz, lapack_complex_float* work, - lapack_int lwork, float* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zstedc_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz, lapack_complex_double* work, - lapack_int lwork, double* rwork, - lapack_int lrwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_sstegr_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, float* z, - lapack_int ldz, lapack_int* isuppz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_dstegr_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - double* z, lapack_int ldz, lapack_int* isuppz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_cstegr_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_int* isuppz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zstegr_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int* isuppz, double* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_sstein_work( int matrix_order, lapack_int n, const float* d, - const float* e, lapack_int m, const float* w, - const lapack_int* iblock, - const lapack_int* isplit, float* z, - lapack_int ldz, float* work, lapack_int* iwork, - lapack_int* ifailv ); -lapack_int LAPACKE_dstein_work( int matrix_order, lapack_int n, const double* d, - const double* e, lapack_int m, const double* w, - const lapack_int* iblock, - const lapack_int* isplit, double* z, - lapack_int ldz, double* work, lapack_int* iwork, - lapack_int* ifailv ); -lapack_int LAPACKE_cstein_work( int matrix_order, lapack_int n, const float* d, - const float* e, lapack_int m, const float* w, - const lapack_int* iblock, - const lapack_int* isplit, - lapack_complex_float* z, lapack_int ldz, - float* work, lapack_int* iwork, - lapack_int* ifailv ); -lapack_int LAPACKE_zstein_work( int matrix_order, lapack_int n, const double* d, - const double* e, lapack_int m, const double* w, - const lapack_int* iblock, - const lapack_int* isplit, - lapack_complex_double* z, lapack_int ldz, - double* work, lapack_int* iwork, - lapack_int* ifailv ); - -lapack_int LAPACKE_sstemr_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - lapack_int* m, float* w, float* z, - lapack_int ldz, lapack_int nzc, - lapack_int* isuppz, lapack_logical* tryrac, - float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dstemr_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int nzc, - lapack_int* isuppz, lapack_logical* tryrac, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_cstemr_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - lapack_int* m, float* w, - lapack_complex_float* z, lapack_int ldz, - lapack_int nzc, lapack_int* isuppz, - lapack_logical* tryrac, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_zstemr_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - lapack_int* m, double* w, - lapack_complex_double* z, lapack_int ldz, - lapack_int nzc, lapack_int* isuppz, - lapack_logical* tryrac, double* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_ssteqr_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, float* z, lapack_int ldz, - float* work ); -lapack_int LAPACKE_dsteqr_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz, - double* work ); -lapack_int LAPACKE_csteqr_work( int matrix_order, char compz, lapack_int n, - float* d, float* e, lapack_complex_float* z, - lapack_int ldz, float* work ); -lapack_int LAPACKE_zsteqr_work( int matrix_order, char compz, lapack_int n, - double* d, double* e, lapack_complex_double* z, - lapack_int ldz, double* work ); - -lapack_int LAPACKE_ssterf_work( lapack_int n, float* d, float* e ); -lapack_int LAPACKE_dsterf_work( lapack_int n, double* d, double* e ); - -lapack_int LAPACKE_sstev_work( int matrix_order, char jobz, lapack_int n, - float* d, float* e, float* z, lapack_int ldz, - float* work ); -lapack_int LAPACKE_dstev_work( int matrix_order, char jobz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz, - double* work ); - -lapack_int LAPACKE_sstevd_work( int matrix_order, char jobz, lapack_int n, - float* d, float* e, float* z, lapack_int ldz, - float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dstevd_work( int matrix_order, char jobz, lapack_int n, - double* d, double* e, double* z, lapack_int ldz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_sstevr_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, float* z, - lapack_int ldz, lapack_int* isuppz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_dstevr_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - double* z, lapack_int ldz, lapack_int* isuppz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_sstevx_work( int matrix_order, char jobz, char range, - lapack_int n, float* d, float* e, float vl, - float vu, lapack_int il, lapack_int iu, - float abstol, lapack_int* m, float* w, float* z, - lapack_int ldz, float* work, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_dstevx_work( int matrix_order, char jobz, char range, - lapack_int n, double* d, double* e, double vl, - double vu, lapack_int il, lapack_int iu, - double abstol, lapack_int* m, double* w, - double* z, lapack_int ldz, double* work, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_ssycon_work( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, - float* rcond, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dsycon_work( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_csycon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, float anorm, - float* rcond, lapack_complex_float* work ); -lapack_int LAPACKE_zsycon_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, double anorm, - double* rcond, lapack_complex_double* work ); - -lapack_int LAPACKE_ssyequb_work( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float* s, - float* scond, float* amax, float* work ); -lapack_int LAPACKE_dsyequb_work( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double* s, - double* scond, double* amax, double* work ); -lapack_int LAPACKE_csyequb_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* s, float* scond, float* amax, - lapack_complex_float* work ); -lapack_int LAPACKE_zsyequb_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* s, double* scond, double* amax, - lapack_complex_double* work ); - -lapack_int LAPACKE_ssyev_work( int matrix_order, char jobz, char uplo, - lapack_int n, float* a, lapack_int lda, float* w, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dsyev_work( int matrix_order, char jobz, char uplo, - lapack_int n, double* a, lapack_int lda, - double* w, double* work, lapack_int lwork ); - -lapack_int LAPACKE_ssyevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, float* a, lapack_int lda, - float* w, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dsyevd_work( int matrix_order, char jobz, char uplo, - lapack_int n, double* a, lapack_int lda, - double* w, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_ssyevr_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, float* a, - lapack_int lda, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, - lapack_int ldz, lapack_int* isuppz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_dsyevr_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, double* a, - lapack_int lda, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, double* z, - lapack_int ldz, lapack_int* isuppz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_ssyevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, float* a, - lapack_int lda, float vl, float vu, - lapack_int il, lapack_int iu, float abstol, - lapack_int* m, float* w, float* z, - lapack_int ldz, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int* ifail ); -lapack_int LAPACKE_dsyevx_work( int matrix_order, char jobz, char range, - char uplo, lapack_int n, double* a, - lapack_int lda, double vl, double vu, - lapack_int il, lapack_int iu, double abstol, - lapack_int* m, double* w, double* z, - lapack_int ldz, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_ssygst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, float* a, lapack_int lda, - const float* b, lapack_int ldb ); -lapack_int LAPACKE_dsygst_work( int matrix_order, lapack_int itype, char uplo, - lapack_int n, double* a, lapack_int lda, - const double* b, lapack_int ldb ); - -lapack_int LAPACKE_ssygv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* w, float* work, lapack_int lwork ); -lapack_int LAPACKE_dsygv_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* w, double* work, lapack_int lwork ); - -lapack_int LAPACKE_ssygvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* w, float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dsygvd_work( int matrix_order, lapack_int itype, char jobz, - char uplo, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* w, double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); - -lapack_int LAPACKE_ssygvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float vl, float vu, lapack_int il, - lapack_int iu, float abstol, lapack_int* m, - float* w, float* z, lapack_int ldz, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int* ifail ); -lapack_int LAPACKE_dsygvx_work( int matrix_order, lapack_int itype, char jobz, - char range, char uplo, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double vl, double vu, lapack_int il, - lapack_int iu, double abstol, lapack_int* m, - double* w, double* z, lapack_int ldz, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int* ifail ); - -lapack_int LAPACKE_ssyrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const float* af, lapack_int ldaf, - const lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dsyrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* b, lapack_int ldb, double* x, - lapack_int ldx, double* ferr, double* berr, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_csyrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_zsyrfs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_ssyrfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, const float* af, - lapack_int ldaf, const lapack_int* ipiv, - const float* s, const float* b, lapack_int ldb, - float* x, lapack_int ldx, float* rcond, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dsyrfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, const double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* s, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_csyrfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* af, - lapack_int ldaf, const lapack_int* ipiv, - const float* s, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zsyrfsx_work( int matrix_order, char uplo, char equed, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* af, - lapack_int ldaf, const lapack_int* ipiv, - const double* s, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_ssysv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, float* a, lapack_int lda, - lapack_int* ipiv, float* b, lapack_int ldb, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dsysv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, double* a, lapack_int lda, - lapack_int* ipiv, double* b, lapack_int ldb, - double* work, lapack_int lwork ); -lapack_int LAPACKE_csysv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_float* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zsysv_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, lapack_complex_double* a, - lapack_int lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_ssysvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, const float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, const float* b, - lapack_int ldb, float* x, lapack_int ldx, - float* rcond, float* ferr, float* berr, - float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dsysvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, const double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, const double* b, - lapack_int ldb, double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - double* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_csysvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int ldb, lapack_complex_float* x, - lapack_int ldx, float* rcond, float* ferr, - float* berr, lapack_complex_float* work, - lapack_int lwork, float* rwork ); -lapack_int LAPACKE_zsysvx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, lapack_int lwork, - double* rwork ); - -lapack_int LAPACKE_ssysvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, float* a, - lapack_int lda, float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, - float* b, lapack_int ldb, float* x, - lapack_int ldx, float* rcond, float* rpvgrw, - float* berr, lapack_int n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int nparams, float* params, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dsysvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, double* a, - lapack_int lda, double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, - double* b, lapack_int ldb, double* x, - lapack_int ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int nparams, double* params, - double* work, lapack_int* iwork ); -lapack_int LAPACKE_csysvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* x, lapack_int ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int nparams, - float* params, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_zsysvxx_work( int matrix_order, char fact, char uplo, - lapack_int n, lapack_int nrhs, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* af, lapack_int ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* x, lapack_int ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int nparams, - double* params, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_ssytrd_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, float* d, float* e, - float* tau, float* work, lapack_int lwork ); -lapack_int LAPACKE_dsytrd_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, double* d, double* e, - double* tau, double* work, lapack_int lwork ); - -lapack_int LAPACKE_ssytrf_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, lapack_int* ipiv, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dsytrf_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, lapack_int* ipiv, - double* work, lapack_int lwork ); -lapack_int LAPACKE_csytrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_float* work, - lapack_int lwork ); -lapack_int LAPACKE_zsytrf_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_int* ipiv, lapack_complex_double* work, - lapack_int lwork ); - -lapack_int LAPACKE_ssytri_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, - const lapack_int* ipiv, float* work ); -lapack_int LAPACKE_dsytri_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, - const lapack_int* ipiv, double* work ); -lapack_int LAPACKE_csytri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work ); -lapack_int LAPACKE_zsytri_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work ); - -lapack_int LAPACKE_ssytrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const lapack_int* ipiv, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dsytrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const lapack_int* ipiv, - double* b, lapack_int ldb ); -lapack_int LAPACKE_csytrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_zsytrs_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stbcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, lapack_int kd, - const float* ab, lapack_int ldab, float* rcond, - float* work, lapack_int* iwork ); -lapack_int LAPACKE_dtbcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, lapack_int kd, - const double* ab, lapack_int ldab, - double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_ctbcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, lapack_int kd, - const lapack_complex_float* ab, lapack_int ldab, - float* rcond, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_ztbcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, lapack_int kd, - const lapack_complex_double* ab, - lapack_int ldab, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_stbrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const float* ab, - lapack_int ldab, const float* b, lapack_int ldb, - const float* x, lapack_int ldx, float* ferr, - float* berr, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dtbrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const double* ab, - lapack_int ldab, const double* b, - lapack_int ldb, const double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_ctbrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const lapack_complex_float* ab, - lapack_int ldab, const lapack_complex_float* b, - lapack_int ldb, const lapack_complex_float* x, - lapack_int ldx, float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztbrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, const lapack_complex_double* b, - lapack_int ldb, const lapack_complex_double* x, - lapack_int ldx, double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_stbtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const float* ab, - lapack_int ldab, float* b, lapack_int ldb ); -lapack_int LAPACKE_dtbtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const double* ab, - lapack_int ldab, double* b, lapack_int ldb ); -lapack_int LAPACKE_ctbtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, const lapack_complex_float* ab, - lapack_int ldab, lapack_complex_float* b, - lapack_int ldb ); -lapack_int LAPACKE_ztbtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int kd, - lapack_int nrhs, - const lapack_complex_double* ab, - lapack_int ldab, lapack_complex_double* b, - lapack_int ldb ); - -lapack_int LAPACKE_stfsm_work( int matrix_order, char transr, char side, - char uplo, char trans, char diag, lapack_int m, - lapack_int n, float alpha, const float* a, - float* b, lapack_int ldb ); -lapack_int LAPACKE_dtfsm_work( int matrix_order, char transr, char side, - char uplo, char trans, char diag, lapack_int m, - lapack_int n, double alpha, const double* a, - double* b, lapack_int ldb ); -lapack_int LAPACKE_ctfsm_work( int matrix_order, char transr, char side, - char uplo, char trans, char diag, lapack_int m, - lapack_int n, lapack_complex_float alpha, - const lapack_complex_float* a, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztfsm_work( int matrix_order, char transr, char side, - char uplo, char trans, char diag, lapack_int m, - lapack_int n, lapack_complex_double alpha, - const lapack_complex_double* a, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stftri_work( int matrix_order, char transr, char uplo, - char diag, lapack_int n, float* a ); -lapack_int LAPACKE_dtftri_work( int matrix_order, char transr, char uplo, - char diag, lapack_int n, double* a ); -lapack_int LAPACKE_ctftri_work( int matrix_order, char transr, char uplo, - char diag, lapack_int n, - lapack_complex_float* a ); -lapack_int LAPACKE_ztftri_work( int matrix_order, char transr, char uplo, - char diag, lapack_int n, - lapack_complex_double* a ); - -lapack_int LAPACKE_stfttp_work( int matrix_order, char transr, char uplo, - lapack_int n, const float* arf, float* ap ); -lapack_int LAPACKE_dtfttp_work( int matrix_order, char transr, char uplo, - lapack_int n, const double* arf, double* ap ); -lapack_int LAPACKE_ctfttp_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* arf, - lapack_complex_float* ap ); -lapack_int LAPACKE_ztfttp_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* arf, - lapack_complex_double* ap ); - -lapack_int LAPACKE_stfttr_work( int matrix_order, char transr, char uplo, - lapack_int n, const float* arf, float* a, - lapack_int lda ); -lapack_int LAPACKE_dtfttr_work( int matrix_order, char transr, char uplo, - lapack_int n, const double* arf, double* a, - lapack_int lda ); -lapack_int LAPACKE_ctfttr_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* arf, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_ztfttr_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* arf, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_stgevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const float* s, lapack_int lds, const float* p, - lapack_int ldp, float* vl, lapack_int ldvl, - float* vr, lapack_int ldvr, lapack_int mm, - lapack_int* m, float* work ); -lapack_int LAPACKE_dtgevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const double* s, lapack_int lds, - const double* p, lapack_int ldp, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, double* work ); -lapack_int LAPACKE_ctgevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* s, lapack_int lds, - const lapack_complex_float* p, lapack_int ldp, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztgevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* s, lapack_int lds, - const lapack_complex_double* p, lapack_int ldp, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_stgexc_work( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, float* a, - lapack_int lda, float* b, lapack_int ldb, - float* q, lapack_int ldq, float* z, - lapack_int ldz, lapack_int* ifst, - lapack_int* ilst, float* work, - lapack_int lwork ); -lapack_int LAPACKE_dtgexc_work( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* q, lapack_int ldq, double* z, - lapack_int ldz, lapack_int* ifst, - lapack_int* ilst, double* work, - lapack_int lwork ); -lapack_int LAPACKE_ctgexc_work( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz, - lapack_int ifst, lapack_int ilst ); -lapack_int LAPACKE_ztgexc_work( int matrix_order, lapack_logical wantq, - lapack_logical wantz, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz, - lapack_int ifst, lapack_int ilst ); - -lapack_int LAPACKE_stgsen_work( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - float* a, lapack_int lda, float* b, - lapack_int ldb, float* alphar, float* alphai, - float* beta, float* q, lapack_int ldq, float* z, - lapack_int ldz, lapack_int* m, float* pl, - float* pr, float* dif, float* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); -lapack_int LAPACKE_dtgsen_work( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - double* a, lapack_int lda, double* b, - lapack_int ldb, double* alphar, double* alphai, - double* beta, double* q, lapack_int ldq, - double* z, lapack_int ldz, lapack_int* m, - double* pl, double* pr, double* dif, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_ctgsen_work( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* alpha, - lapack_complex_float* beta, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* z, lapack_int ldz, - lapack_int* m, float* pl, float* pr, float* dif, - lapack_complex_float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_ztgsen_work( int matrix_order, lapack_int ijob, - lapack_logical wantq, lapack_logical wantz, - const lapack_logical* select, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* alpha, - lapack_complex_double* beta, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* z, lapack_int ldz, - lapack_int* m, double* pl, double* pr, - double* dif, lapack_complex_double* work, - lapack_int lwork, lapack_int* iwork, - lapack_int liwork ); - -lapack_int LAPACKE_stgsja_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_int k, lapack_int l, - float* a, lapack_int lda, float* b, - lapack_int ldb, float tola, float tolb, - float* alpha, float* beta, float* u, - lapack_int ldu, float* v, lapack_int ldv, - float* q, lapack_int ldq, float* work, - lapack_int* ncycle ); -lapack_int LAPACKE_dtgsja_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_int k, lapack_int l, - double* a, lapack_int lda, double* b, - lapack_int ldb, double tola, double tolb, - double* alpha, double* beta, double* u, - lapack_int ldu, double* v, lapack_int ldv, - double* q, lapack_int ldq, double* work, - lapack_int* ncycle ); -lapack_int LAPACKE_ctgsja_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_int k, lapack_int l, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - float tola, float tolb, float* alpha, - float* beta, lapack_complex_float* u, - lapack_int ldu, lapack_complex_float* v, - lapack_int ldv, lapack_complex_float* q, - lapack_int ldq, lapack_complex_float* work, - lapack_int* ncycle ); -lapack_int LAPACKE_ztgsja_work( int matrix_order, char jobu, char jobv, - char jobq, lapack_int m, lapack_int p, - lapack_int n, lapack_int k, lapack_int l, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - double tola, double tolb, double* alpha, - double* beta, lapack_complex_double* u, - lapack_int ldu, lapack_complex_double* v, - lapack_int ldv, lapack_complex_double* q, - lapack_int ldq, lapack_complex_double* work, - lapack_int* ncycle ); - -lapack_int LAPACKE_stgsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const float* a, lapack_int lda, const float* b, - lapack_int ldb, const float* vl, - lapack_int ldvl, const float* vr, - lapack_int ldvr, float* s, float* dif, - lapack_int mm, lapack_int* m, float* work, - lapack_int lwork, lapack_int* iwork ); -lapack_int LAPACKE_dtgsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const double* a, lapack_int lda, - const double* b, lapack_int ldb, - const double* vl, lapack_int ldvl, - const double* vr, lapack_int ldvr, double* s, - double* dif, lapack_int mm, lapack_int* m, - double* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_ctgsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* vl, lapack_int ldvl, - const lapack_complex_float* vr, lapack_int ldvr, - float* s, float* dif, lapack_int mm, - lapack_int* m, lapack_complex_float* work, - lapack_int lwork, lapack_int* iwork ); -lapack_int LAPACKE_ztgsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* vl, - lapack_int ldvl, - const lapack_complex_double* vr, - lapack_int ldvr, double* s, double* dif, - lapack_int mm, lapack_int* m, - lapack_complex_double* work, lapack_int lwork, - lapack_int* iwork ); - -lapack_int LAPACKE_stgsyl_work( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, const float* a, - lapack_int lda, const float* b, lapack_int ldb, - float* c, lapack_int ldc, const float* d, - lapack_int ldd, const float* e, lapack_int lde, - float* f, lapack_int ldf, float* scale, - float* dif, float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dtgsyl_work( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, const double* a, - lapack_int lda, const double* b, lapack_int ldb, - double* c, lapack_int ldc, const double* d, - lapack_int ldd, const double* e, lapack_int lde, - double* f, lapack_int ldf, double* scale, - double* dif, double* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_ctgsyl_work( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* c, lapack_int ldc, - const lapack_complex_float* d, lapack_int ldd, - const lapack_complex_float* e, lapack_int lde, - lapack_complex_float* f, lapack_int ldf, - float* scale, float* dif, - lapack_complex_float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_ztgsyl_work( int matrix_order, char trans, lapack_int ijob, - lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* c, lapack_int ldc, - const lapack_complex_double* d, lapack_int ldd, - const lapack_complex_double* e, lapack_int lde, - lapack_complex_double* f, lapack_int ldf, - double* scale, double* dif, - lapack_complex_double* work, lapack_int lwork, - lapack_int* iwork ); - -lapack_int LAPACKE_stpcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, const float* ap, - float* rcond, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dtpcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, const double* ap, - double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_ctpcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, - const lapack_complex_float* ap, float* rcond, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztpcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, - const lapack_complex_double* ap, double* rcond, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_stprfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const float* ap, const float* b, lapack_int ldb, - const float* x, lapack_int ldx, float* ferr, - float* berr, float* work, lapack_int* iwork ); -lapack_int LAPACKE_dtprfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const double* ap, const double* b, - lapack_int ldb, const double* x, lapack_int ldx, - double* ferr, double* berr, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_ctprfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztprfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_stptri_work( int matrix_order, char uplo, char diag, - lapack_int n, float* ap ); -lapack_int LAPACKE_dtptri_work( int matrix_order, char uplo, char diag, - lapack_int n, double* ap ); -lapack_int LAPACKE_ctptri_work( int matrix_order, char uplo, char diag, - lapack_int n, lapack_complex_float* ap ); -lapack_int LAPACKE_ztptri_work( int matrix_order, char uplo, char diag, - lapack_int n, lapack_complex_double* ap ); - -lapack_int LAPACKE_stptrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const float* ap, float* b, lapack_int ldb ); -lapack_int LAPACKE_dtptrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const double* ap, double* b, lapack_int ldb ); -lapack_int LAPACKE_ctptrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_float* ap, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztptrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_double* ap, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_stpttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const float* ap, float* arf ); -lapack_int LAPACKE_dtpttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const double* ap, double* arf ); -lapack_int LAPACKE_ctpttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* ap, - lapack_complex_float* arf ); -lapack_int LAPACKE_ztpttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* ap, - lapack_complex_double* arf ); - -lapack_int LAPACKE_stpttr_work( int matrix_order, char uplo, lapack_int n, - const float* ap, float* a, lapack_int lda ); -lapack_int LAPACKE_dtpttr_work( int matrix_order, char uplo, lapack_int n, - const double* ap, double* a, lapack_int lda ); -lapack_int LAPACKE_ctpttr_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_ztpttr_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_strcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, const float* a, - lapack_int lda, float* rcond, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dtrcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, const double* a, - lapack_int lda, double* rcond, double* work, - lapack_int* iwork ); -lapack_int LAPACKE_ctrcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - float* rcond, lapack_complex_float* work, - float* rwork ); -lapack_int LAPACKE_ztrcon_work( int matrix_order, char norm, char uplo, - char diag, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - double* rcond, lapack_complex_double* work, - double* rwork ); - -lapack_int LAPACKE_strevc_work( int matrix_order, char side, char howmny, - lapack_logical* select, lapack_int n, - const float* t, lapack_int ldt, float* vl, - lapack_int ldvl, float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, float* work ); -lapack_int LAPACKE_dtrevc_work( int matrix_order, char side, char howmny, - lapack_logical* select, lapack_int n, - const double* t, lapack_int ldt, double* vl, - lapack_int ldvl, double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, double* work ); -lapack_int LAPACKE_ctrevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* vl, lapack_int ldvl, - lapack_complex_float* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztrevc_work( int matrix_order, char side, char howmny, - const lapack_logical* select, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* vl, lapack_int ldvl, - lapack_complex_double* vr, lapack_int ldvr, - lapack_int mm, lapack_int* m, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_strexc_work( int matrix_order, char compq, lapack_int n, - float* t, lapack_int ldt, float* q, - lapack_int ldq, lapack_int* ifst, - lapack_int* ilst, float* work ); -lapack_int LAPACKE_dtrexc_work( int matrix_order, char compq, lapack_int n, - double* t, lapack_int ldt, double* q, - lapack_int ldq, lapack_int* ifst, - lapack_int* ilst, double* work ); -lapack_int LAPACKE_ctrexc_work( int matrix_order, char compq, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* q, lapack_int ldq, - lapack_int ifst, lapack_int ilst ); -lapack_int LAPACKE_ztrexc_work( int matrix_order, char compq, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* q, lapack_int ldq, - lapack_int ifst, lapack_int ilst ); - -lapack_int LAPACKE_strrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const float* a, lapack_int lda, const float* b, - lapack_int ldb, const float* x, lapack_int ldx, - float* ferr, float* berr, float* work, - lapack_int* iwork ); -lapack_int LAPACKE_dtrrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const double* a, lapack_int lda, - const double* b, lapack_int ldb, - const double* x, lapack_int ldx, double* ferr, - double* berr, double* work, lapack_int* iwork ); -lapack_int LAPACKE_ctrrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - const lapack_complex_float* x, lapack_int ldx, - float* ferr, float* berr, - lapack_complex_float* work, float* rwork ); -lapack_int LAPACKE_ztrrfs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - const lapack_complex_double* x, lapack_int ldx, - double* ferr, double* berr, - lapack_complex_double* work, double* rwork ); - -lapack_int LAPACKE_strsen_work( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - float* t, lapack_int ldt, float* q, - lapack_int ldq, float* wr, float* wi, - lapack_int* m, float* s, float* sep, - float* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_dtrsen_work( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - double* t, lapack_int ldt, double* q, - lapack_int ldq, double* wr, double* wi, - lapack_int* m, double* s, double* sep, - double* work, lapack_int lwork, - lapack_int* iwork, lapack_int liwork ); -lapack_int LAPACKE_ctrsen_work( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* w, lapack_int* m, - float* s, float* sep, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_ztrsen_work( int matrix_order, char job, char compq, - const lapack_logical* select, lapack_int n, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* w, lapack_int* m, - double* s, double* sep, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_strsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const float* t, lapack_int ldt, const float* vl, - lapack_int ldvl, const float* vr, - lapack_int ldvr, float* s, float* sep, - lapack_int mm, lapack_int* m, float* work, - lapack_int ldwork, lapack_int* iwork ); -lapack_int LAPACKE_dtrsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const double* t, lapack_int ldt, - const double* vl, lapack_int ldvl, - const double* vr, lapack_int ldvr, double* s, - double* sep, lapack_int mm, lapack_int* m, - double* work, lapack_int ldwork, - lapack_int* iwork ); -lapack_int LAPACKE_ctrsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_float* t, lapack_int ldt, - const lapack_complex_float* vl, lapack_int ldvl, - const lapack_complex_float* vr, lapack_int ldvr, - float* s, float* sep, lapack_int mm, - lapack_int* m, lapack_complex_float* work, - lapack_int ldwork, float* rwork ); -lapack_int LAPACKE_ztrsna_work( int matrix_order, char job, char howmny, - const lapack_logical* select, lapack_int n, - const lapack_complex_double* t, lapack_int ldt, - const lapack_complex_double* vl, - lapack_int ldvl, - const lapack_complex_double* vr, - lapack_int ldvr, double* s, double* sep, - lapack_int mm, lapack_int* m, - lapack_complex_double* work, lapack_int ldwork, - double* rwork ); - -lapack_int LAPACKE_strsyl_work( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const float* a, lapack_int lda, const float* b, - lapack_int ldb, float* c, lapack_int ldc, - float* scale ); -lapack_int LAPACKE_dtrsyl_work( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const double* a, lapack_int lda, - const double* b, lapack_int ldb, double* c, - lapack_int ldc, double* scale ); -lapack_int LAPACKE_ctrsyl_work( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* c, lapack_int ldc, - float* scale ); -lapack_int LAPACKE_ztrsyl_work( int matrix_order, char trana, char tranb, - lapack_int isgn, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* c, lapack_int ldc, - double* scale ); - -lapack_int LAPACKE_strtri_work( int matrix_order, char uplo, char diag, - lapack_int n, float* a, lapack_int lda ); -lapack_int LAPACKE_dtrtri_work( int matrix_order, char uplo, char diag, - lapack_int n, double* a, lapack_int lda ); -lapack_int LAPACKE_ctrtri_work( int matrix_order, char uplo, char diag, - lapack_int n, lapack_complex_float* a, - lapack_int lda ); -lapack_int LAPACKE_ztrtri_work( int matrix_order, char uplo, char diag, - lapack_int n, lapack_complex_double* a, - lapack_int lda ); - -lapack_int LAPACKE_strtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const float* a, lapack_int lda, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dtrtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const double* a, lapack_int lda, double* b, - lapack_int ldb ); -lapack_int LAPACKE_ctrtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztrtrs_work( int matrix_order, char uplo, char trans, - char diag, lapack_int n, lapack_int nrhs, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_strttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const float* a, lapack_int lda, - float* arf ); -lapack_int LAPACKE_dtrttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const double* a, lapack_int lda, - double* arf ); -lapack_int LAPACKE_ctrttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_float* a, - lapack_int lda, lapack_complex_float* arf ); -lapack_int LAPACKE_ztrttf_work( int matrix_order, char transr, char uplo, - lapack_int n, const lapack_complex_double* a, - lapack_int lda, lapack_complex_double* arf ); - -lapack_int LAPACKE_strttp_work( int matrix_order, char uplo, lapack_int n, - const float* a, lapack_int lda, float* ap ); -lapack_int LAPACKE_dtrttp_work( int matrix_order, char uplo, lapack_int n, - const double* a, lapack_int lda, double* ap ); -lapack_int LAPACKE_ctrttp_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - lapack_complex_float* ap ); -lapack_int LAPACKE_ztrttp_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - lapack_complex_double* ap ); - -lapack_int LAPACKE_stzrzf_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* tau, - float* work, lapack_int lwork ); -lapack_int LAPACKE_dtzrzf_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* tau, - double* work, lapack_int lwork ); -lapack_int LAPACKE_ctzrzf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_ztzrzf_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cungbr_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, - lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zungbr_work( int matrix_order, char vect, lapack_int m, - lapack_int n, lapack_int k, - lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunghr_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunghr_work( int matrix_order, lapack_int n, lapack_int ilo, - lapack_int ihi, lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunglq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunglq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cungql_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zungql_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cungqr_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zungqr_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cungrq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zungrq_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int k, lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cungtr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zungtr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmbr_work( int matrix_order, char vect, char side, - char trans, lapack_int m, lapack_int n, - lapack_int k, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmbr_work( int matrix_order, char vect, char side, - char trans, lapack_int m, lapack_int n, - lapack_int k, const lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmhr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmhr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int ilo, - lapack_int ihi, const lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmlq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmlq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmql_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmql_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmqr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmqr_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmrq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmrq_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmrz_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const lapack_complex_float* a, - lapack_int lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmrz_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, const lapack_complex_double* a, - lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cunmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const lapack_complex_float* a, lapack_int lda, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_zunmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const lapack_complex_double* a, lapack_int lda, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work, lapack_int lwork ); - -lapack_int LAPACKE_cupgtr_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_float* ap, - const lapack_complex_float* tau, - lapack_complex_float* q, lapack_int ldq, - lapack_complex_float* work ); -lapack_int LAPACKE_zupgtr_work( int matrix_order, char uplo, lapack_int n, - const lapack_complex_double* ap, - const lapack_complex_double* tau, - lapack_complex_double* q, lapack_int ldq, - lapack_complex_double* work ); - -lapack_int LAPACKE_cupmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const lapack_complex_float* ap, - const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int ldc, - lapack_complex_float* work ); -lapack_int LAPACKE_zupmtr_work( int matrix_order, char side, char uplo, - char trans, lapack_int m, lapack_int n, - const lapack_complex_double* ap, - const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int ldc, - lapack_complex_double* work ); - -lapack_int LAPACKE_claghe( int matrix_order, lapack_int n, lapack_int k, - const float* d, lapack_complex_float* a, - lapack_int lda, lapack_int* iseed ); -lapack_int LAPACKE_zlaghe( int matrix_order, lapack_int n, lapack_int k, - const double* d, lapack_complex_double* a, - lapack_int lda, lapack_int* iseed ); - -lapack_int LAPACKE_slagsy( int matrix_order, lapack_int n, lapack_int k, - const float* d, float* a, lapack_int lda, - lapack_int* iseed ); -lapack_int LAPACKE_dlagsy( int matrix_order, lapack_int n, lapack_int k, - const double* d, double* a, lapack_int lda, - lapack_int* iseed ); -lapack_int LAPACKE_clagsy( int matrix_order, lapack_int n, lapack_int k, - const float* d, lapack_complex_float* a, - lapack_int lda, lapack_int* iseed ); -lapack_int LAPACKE_zlagsy( int matrix_order, lapack_int n, lapack_int k, - const double* d, lapack_complex_double* a, - lapack_int lda, lapack_int* iseed ); - -lapack_int LAPACKE_slapmr( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, float* x, lapack_int ldx, - lapack_int* k ); -lapack_int LAPACKE_dlapmr( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, double* x, - lapack_int ldx, lapack_int* k ); -lapack_int LAPACKE_clapmr( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, lapack_complex_float* x, - lapack_int ldx, lapack_int* k ); -lapack_int LAPACKE_zlapmr( int matrix_order, lapack_logical forwrd, - lapack_int m, lapack_int n, lapack_complex_double* x, - lapack_int ldx, lapack_int* k ); - - -float LAPACKE_slapy2( float x, float y ); -double LAPACKE_dlapy2( double x, double y ); - -float LAPACKE_slapy3( float x, float y, float z ); -double LAPACKE_dlapy3( double x, double y, double z ); - -lapack_int LAPACKE_slartgp( float f, float g, float* cs, float* sn, float* r ); -lapack_int LAPACKE_dlartgp( double f, double g, double* cs, double* sn, - double* r ); - -lapack_int LAPACKE_slartgs( float x, float y, float sigma, float* cs, - float* sn ); -lapack_int LAPACKE_dlartgs( double x, double y, double sigma, double* cs, - double* sn ); - - -//LAPACK 3.3.0 -lapack_int LAPACKE_cbbcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, lapack_int m, - lapack_int p, lapack_int q, float* theta, float* phi, - lapack_complex_float* u1, lapack_int ldu1, - lapack_complex_float* u2, lapack_int ldu2, - lapack_complex_float* v1t, lapack_int ldv1t, - lapack_complex_float* v2t, lapack_int ldv2t, - float* b11d, float* b11e, float* b12d, float* b12e, - float* b21d, float* b21e, float* b22d, float* b22e ); -lapack_int LAPACKE_cbbcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - lapack_int m, lapack_int p, lapack_int q, - float* theta, float* phi, - lapack_complex_float* u1, lapack_int ldu1, - lapack_complex_float* u2, lapack_int ldu2, - lapack_complex_float* v1t, lapack_int ldv1t, - lapack_complex_float* v2t, lapack_int ldv2t, - float* b11d, float* b11e, float* b12d, - float* b12e, float* b21d, float* b21e, - float* b22d, float* b22e, float* rwork, - lapack_int lrwork ); -lapack_int LAPACKE_cheswapr( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_cheswapr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_chetri2( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_chetri2_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_chetri2x( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, lapack_int nb ); -lapack_int LAPACKE_chetri2x_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int nb ); -lapack_int LAPACKE_chetrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_chetrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work ); -lapack_int LAPACKE_csyconv( int matrix_order, char uplo, char way, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_csyconv_work( int matrix_order, char uplo, char way, - lapack_int n, lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* work ); -lapack_int LAPACKE_csyswapr( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_csyswapr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_csytri2( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_csytri2_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_csytri2x( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, lapack_int nb ); -lapack_int LAPACKE_csytri2x_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int nb ); -lapack_int LAPACKE_csytrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_csytrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_float* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work ); -lapack_int LAPACKE_cunbdb( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_float* x11, lapack_int ldx11, - lapack_complex_float* x12, lapack_int ldx12, - lapack_complex_float* x21, lapack_int ldx21, - lapack_complex_float* x22, lapack_int ldx22, - float* theta, float* phi, - lapack_complex_float* taup1, - lapack_complex_float* taup2, - lapack_complex_float* tauq1, - lapack_complex_float* tauq2 ); -lapack_int LAPACKE_cunbdb_work( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_float* x11, lapack_int ldx11, - lapack_complex_float* x12, lapack_int ldx12, - lapack_complex_float* x21, lapack_int ldx21, - lapack_complex_float* x22, lapack_int ldx22, - float* theta, float* phi, - lapack_complex_float* taup1, - lapack_complex_float* taup2, - lapack_complex_float* tauq1, - lapack_complex_float* tauq2, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_cuncsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_float* x11, lapack_int ldx11, - lapack_complex_float* x12, lapack_int ldx12, - lapack_complex_float* x21, lapack_int ldx21, - lapack_complex_float* x22, lapack_int ldx22, - float* theta, lapack_complex_float* u1, - lapack_int ldu1, lapack_complex_float* u2, - lapack_int ldu2, lapack_complex_float* v1t, - lapack_int ldv1t, lapack_complex_float* v2t, - lapack_int ldv2t ); -lapack_int LAPACKE_cuncsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - char signs, lapack_int m, lapack_int p, - lapack_int q, lapack_complex_float* x11, - lapack_int ldx11, lapack_complex_float* x12, - lapack_int ldx12, lapack_complex_float* x21, - lapack_int ldx21, lapack_complex_float* x22, - lapack_int ldx22, float* theta, - lapack_complex_float* u1, lapack_int ldu1, - lapack_complex_float* u2, lapack_int ldu2, - lapack_complex_float* v1t, lapack_int ldv1t, - lapack_complex_float* v2t, lapack_int ldv2t, - lapack_complex_float* work, lapack_int lwork, - float* rwork, lapack_int lrwork, - lapack_int* iwork ); -lapack_int LAPACKE_dbbcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, lapack_int m, - lapack_int p, lapack_int q, double* theta, - double* phi, double* u1, lapack_int ldu1, double* u2, - lapack_int ldu2, double* v1t, lapack_int ldv1t, - double* v2t, lapack_int ldv2t, double* b11d, - double* b11e, double* b12d, double* b12e, - double* b21d, double* b21e, double* b22d, - double* b22e ); -lapack_int LAPACKE_dbbcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - lapack_int m, lapack_int p, lapack_int q, - double* theta, double* phi, double* u1, - lapack_int ldu1, double* u2, lapack_int ldu2, - double* v1t, lapack_int ldv1t, double* v2t, - lapack_int ldv2t, double* b11d, double* b11e, - double* b12d, double* b12e, double* b21d, - double* b21e, double* b22d, double* b22e, - double* work, lapack_int lwork ); -lapack_int LAPACKE_dorbdb( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - double* x11, lapack_int ldx11, double* x12, - lapack_int ldx12, double* x21, lapack_int ldx21, - double* x22, lapack_int ldx22, double* theta, - double* phi, double* taup1, double* taup2, - double* tauq1, double* tauq2 ); -lapack_int LAPACKE_dorbdb_work( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - double* x11, lapack_int ldx11, double* x12, - lapack_int ldx12, double* x21, lapack_int ldx21, - double* x22, lapack_int ldx22, double* theta, - double* phi, double* taup1, double* taup2, - double* tauq1, double* tauq2, double* work, - lapack_int lwork ); -lapack_int LAPACKE_dorcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - double* x11, lapack_int ldx11, double* x12, - lapack_int ldx12, double* x21, lapack_int ldx21, - double* x22, lapack_int ldx22, double* theta, - double* u1, lapack_int ldu1, double* u2, - lapack_int ldu2, double* v1t, lapack_int ldv1t, - double* v2t, lapack_int ldv2t ); -lapack_int LAPACKE_dorcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - char signs, lapack_int m, lapack_int p, - lapack_int q, double* x11, lapack_int ldx11, - double* x12, lapack_int ldx12, double* x21, - lapack_int ldx21, double* x22, lapack_int ldx22, - double* theta, double* u1, lapack_int ldu1, - double* u2, lapack_int ldu2, double* v1t, - lapack_int ldv1t, double* v2t, lapack_int ldv2t, - double* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_dsyconv( int matrix_order, char uplo, char way, lapack_int n, - double* a, lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_dsyconv_work( int matrix_order, char uplo, char way, - lapack_int n, double* a, lapack_int lda, - const lapack_int* ipiv, double* work ); -lapack_int LAPACKE_dsyswapr( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int i1, lapack_int i2 ); -lapack_int LAPACKE_dsyswapr_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int i1, lapack_int i2 ); -lapack_int LAPACKE_dsytri2( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_dsytri2_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int lwork ); -lapack_int LAPACKE_dsytri2x( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, const lapack_int* ipiv, - lapack_int nb ); -lapack_int LAPACKE_dsytri2x_work( int matrix_order, char uplo, lapack_int n, - double* a, lapack_int lda, - const lapack_int* ipiv, double* work, - lapack_int nb ); -lapack_int LAPACKE_dsytrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, lapack_int lda, - const lapack_int* ipiv, double* b, lapack_int ldb ); -lapack_int LAPACKE_dsytrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const double* a, - lapack_int lda, const lapack_int* ipiv, - double* b, lapack_int ldb, double* work ); -lapack_int LAPACKE_sbbcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, lapack_int m, - lapack_int p, lapack_int q, float* theta, float* phi, - float* u1, lapack_int ldu1, float* u2, - lapack_int ldu2, float* v1t, lapack_int ldv1t, - float* v2t, lapack_int ldv2t, float* b11d, - float* b11e, float* b12d, float* b12e, float* b21d, - float* b21e, float* b22d, float* b22e ); -lapack_int LAPACKE_sbbcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - lapack_int m, lapack_int p, lapack_int q, - float* theta, float* phi, float* u1, - lapack_int ldu1, float* u2, lapack_int ldu2, - float* v1t, lapack_int ldv1t, float* v2t, - lapack_int ldv2t, float* b11d, float* b11e, - float* b12d, float* b12e, float* b21d, - float* b21e, float* b22d, float* b22e, - float* work, lapack_int lwork ); -lapack_int LAPACKE_sorbdb( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, float* x11, - lapack_int ldx11, float* x12, lapack_int ldx12, - float* x21, lapack_int ldx21, float* x22, - lapack_int ldx22, float* theta, float* phi, - float* taup1, float* taup2, float* tauq1, - float* tauq2 ); -lapack_int LAPACKE_sorbdb_work( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - float* x11, lapack_int ldx11, float* x12, - lapack_int ldx12, float* x21, lapack_int ldx21, - float* x22, lapack_int ldx22, float* theta, - float* phi, float* taup1, float* taup2, - float* tauq1, float* tauq2, float* work, - lapack_int lwork ); -lapack_int LAPACKE_sorcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, float* x11, - lapack_int ldx11, float* x12, lapack_int ldx12, - float* x21, lapack_int ldx21, float* x22, - lapack_int ldx22, float* theta, float* u1, - lapack_int ldu1, float* u2, lapack_int ldu2, - float* v1t, lapack_int ldv1t, float* v2t, - lapack_int ldv2t ); -lapack_int LAPACKE_sorcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - char signs, lapack_int m, lapack_int p, - lapack_int q, float* x11, lapack_int ldx11, - float* x12, lapack_int ldx12, float* x21, - lapack_int ldx21, float* x22, lapack_int ldx22, - float* theta, float* u1, lapack_int ldu1, - float* u2, lapack_int ldu2, float* v1t, - lapack_int ldv1t, float* v2t, lapack_int ldv2t, - float* work, lapack_int lwork, - lapack_int* iwork ); -lapack_int LAPACKE_ssyconv( int matrix_order, char uplo, char way, lapack_int n, - float* a, lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_ssyconv_work( int matrix_order, char uplo, char way, - lapack_int n, float* a, lapack_int lda, - const lapack_int* ipiv, float* work ); -lapack_int LAPACKE_ssyswapr( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int i1, lapack_int i2 ); -lapack_int LAPACKE_ssyswapr_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int i1, lapack_int i2 ); -lapack_int LAPACKE_ssytri2( int matrix_order, char uplo, lapack_int n, float* a, - lapack_int lda, const lapack_int* ipiv ); -lapack_int LAPACKE_ssytri2_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int lwork ); -lapack_int LAPACKE_ssytri2x( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, const lapack_int* ipiv, - lapack_int nb ); -lapack_int LAPACKE_ssytri2x_work( int matrix_order, char uplo, lapack_int n, - float* a, lapack_int lda, - const lapack_int* ipiv, float* work, - lapack_int nb ); -lapack_int LAPACKE_ssytrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, lapack_int lda, - const lapack_int* ipiv, float* b, lapack_int ldb ); -lapack_int LAPACKE_ssytrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const float* a, - lapack_int lda, const lapack_int* ipiv, - float* b, lapack_int ldb, float* work ); -lapack_int LAPACKE_zbbcsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, lapack_int m, - lapack_int p, lapack_int q, double* theta, - double* phi, lapack_complex_double* u1, - lapack_int ldu1, lapack_complex_double* u2, - lapack_int ldu2, lapack_complex_double* v1t, - lapack_int ldv1t, lapack_complex_double* v2t, - lapack_int ldv2t, double* b11d, double* b11e, - double* b12d, double* b12e, double* b21d, - double* b21e, double* b22d, double* b22e ); -lapack_int LAPACKE_zbbcsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - lapack_int m, lapack_int p, lapack_int q, - double* theta, double* phi, - lapack_complex_double* u1, lapack_int ldu1, - lapack_complex_double* u2, lapack_int ldu2, - lapack_complex_double* v1t, lapack_int ldv1t, - lapack_complex_double* v2t, lapack_int ldv2t, - double* b11d, double* b11e, double* b12d, - double* b12e, double* b21d, double* b21e, - double* b22d, double* b22e, double* rwork, - lapack_int lrwork ); -lapack_int LAPACKE_zheswapr( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_zheswapr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_zhetri2( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zhetri2_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int lwork ); -lapack_int LAPACKE_zhetri2x( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, lapack_int nb ); -lapack_int LAPACKE_zhetri2x_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int nb ); -lapack_int LAPACKE_zhetrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); -lapack_int LAPACKE_zhetrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work ); -lapack_int LAPACKE_zsyconv( int matrix_order, char uplo, char way, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zsyconv_work( int matrix_order, char uplo, char way, - lapack_int n, lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* work ); -lapack_int LAPACKE_zsyswapr( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_zsyswapr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int i1, - lapack_int i2 ); -lapack_int LAPACKE_zsytri2( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv ); -lapack_int LAPACKE_zsytri2_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int lwork ); -lapack_int LAPACKE_zsytri2x( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, lapack_int nb ); -lapack_int LAPACKE_zsytri2x_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double* a, lapack_int lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int nb ); -lapack_int LAPACKE_zsytrs2( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb ); -lapack_int LAPACKE_zsytrs2_work( int matrix_order, char uplo, lapack_int n, - lapack_int nrhs, const lapack_complex_double* a, - lapack_int lda, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work ); -lapack_int LAPACKE_zunbdb( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_double* x11, lapack_int ldx11, - lapack_complex_double* x12, lapack_int ldx12, - lapack_complex_double* x21, lapack_int ldx21, - lapack_complex_double* x22, lapack_int ldx22, - double* theta, double* phi, - lapack_complex_double* taup1, - lapack_complex_double* taup2, - lapack_complex_double* tauq1, - lapack_complex_double* tauq2 ); -lapack_int LAPACKE_zunbdb_work( int matrix_order, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_double* x11, lapack_int ldx11, - lapack_complex_double* x12, lapack_int ldx12, - lapack_complex_double* x21, lapack_int ldx21, - lapack_complex_double* x22, lapack_int ldx22, - double* theta, double* phi, - lapack_complex_double* taup1, - lapack_complex_double* taup2, - lapack_complex_double* tauq1, - lapack_complex_double* tauq2, - lapack_complex_double* work, lapack_int lwork ); -lapack_int LAPACKE_zuncsd( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, char signs, - lapack_int m, lapack_int p, lapack_int q, - lapack_complex_double* x11, lapack_int ldx11, - lapack_complex_double* x12, lapack_int ldx12, - lapack_complex_double* x21, lapack_int ldx21, - lapack_complex_double* x22, lapack_int ldx22, - double* theta, lapack_complex_double* u1, - lapack_int ldu1, lapack_complex_double* u2, - lapack_int ldu2, lapack_complex_double* v1t, - lapack_int ldv1t, lapack_complex_double* v2t, - lapack_int ldv2t ); -lapack_int LAPACKE_zuncsd_work( int matrix_order, char jobu1, char jobu2, - char jobv1t, char jobv2t, char trans, - char signs, lapack_int m, lapack_int p, - lapack_int q, lapack_complex_double* x11, - lapack_int ldx11, lapack_complex_double* x12, - lapack_int ldx12, lapack_complex_double* x21, - lapack_int ldx21, lapack_complex_double* x22, - lapack_int ldx22, double* theta, - lapack_complex_double* u1, lapack_int ldu1, - lapack_complex_double* u2, lapack_int ldu2, - lapack_complex_double* v1t, lapack_int ldv1t, - lapack_complex_double* v2t, lapack_int ldv2t, - lapack_complex_double* work, lapack_int lwork, - double* rwork, lapack_int lrwork, - lapack_int* iwork ); -//LAPACK 3.4.0 -lapack_int LAPACKE_sgemqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const float* v, lapack_int ldv, - const float* t, lapack_int ldt, float* c, - lapack_int ldc ); -lapack_int LAPACKE_dgemqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const double* v, lapack_int ldv, - const double* t, lapack_int ldt, double* c, - lapack_int ldc ); -lapack_int LAPACKE_cgemqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const lapack_complex_float* v, - lapack_int ldv, const lapack_complex_float* t, - lapack_int ldt, lapack_complex_float* c, - lapack_int ldc ); -lapack_int LAPACKE_zgemqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const lapack_complex_double* v, - lapack_int ldv, const lapack_complex_double* t, - lapack_int ldt, lapack_complex_double* c, - lapack_int ldc ); - -lapack_int LAPACKE_sgeqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, float* a, lapack_int lda, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dgeqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, double* a, lapack_int lda, double* t, - lapack_int ldt ); -lapack_int LAPACKE_cgeqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* t, - lapack_int ldt ); -lapack_int LAPACKE_zgeqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* t, - lapack_int ldt ); - -lapack_int LAPACKE_sgeqrt2( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dgeqrt2( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* t, - lapack_int ldt ); -lapack_int LAPACKE_cgeqrt2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zgeqrt2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_sgeqrt3( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dgeqrt3( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* t, - lapack_int ldt ); -lapack_int LAPACKE_cgeqrt3( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zgeqrt3( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_stpmqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, const float* v, - lapack_int ldv, const float* t, lapack_int ldt, - float* a, lapack_int lda, float* b, - lapack_int ldb ); -lapack_int LAPACKE_dtpmqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, const double* v, - lapack_int ldv, const double* t, lapack_int ldt, - double* a, lapack_int lda, double* b, - lapack_int ldb ); -lapack_int LAPACKE_ctpmqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb ); -lapack_int LAPACKE_ztpmqrt( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb ); - -lapack_int LAPACKE_dtpqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, double* a, - lapack_int lda, double* b, lapack_int ldb, double* t, - lapack_int ldt ); -lapack_int LAPACKE_ctpqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* t, - lapack_complex_float* b, lapack_int ldb, - lapack_int ldt ); -lapack_int LAPACKE_ztpqrt( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_stpqrt2( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* b, lapack_int ldb, - float* t, lapack_int ldt ); -lapack_int LAPACKE_dtpqrt2( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* b, - lapack_int ldb, double* t, lapack_int ldt ); -lapack_int LAPACKE_ctpqrt2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_ztpqrt2( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_stprfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, lapack_int l, const float* v, - lapack_int ldv, const float* t, lapack_int ldt, - float* a, lapack_int lda, float* b, lapack_int ldb, - lapack_int myldwork ); -lapack_int LAPACKE_dtprfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, lapack_int l, const double* v, - lapack_int ldv, const double* t, lapack_int ldt, - double* a, lapack_int lda, double* b, lapack_int ldb, - lapack_int myldwork ); -lapack_int LAPACKE_ctprfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, lapack_int l, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_int myldwork ); -lapack_int LAPACKE_ztprfb( int matrix_order, char side, char trans, char direct, - char storev, lapack_int m, lapack_int n, - lapack_int k, lapack_int l, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_int myldwork ); - -lapack_int LAPACKE_sgemqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const float* v, lapack_int ldv, - const float* t, lapack_int ldt, float* c, - lapack_int ldc, float* work ); -lapack_int LAPACKE_dgemqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const double* v, lapack_int ldv, - const double* t, lapack_int ldt, double* c, - lapack_int ldc, double* work ); -lapack_int LAPACKE_cgemqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const lapack_complex_float* v, - lapack_int ldv, const lapack_complex_float* t, - lapack_int ldt, lapack_complex_float* c, - lapack_int ldc, lapack_complex_float* work ); -lapack_int LAPACKE_zgemqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int nb, const lapack_complex_double* v, - lapack_int ldv, const lapack_complex_double* t, - lapack_int ldt, lapack_complex_double* c, - lapack_int ldc, lapack_complex_double* work ); - -lapack_int LAPACKE_sgeqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, float* a, lapack_int lda, - float* t, lapack_int ldt, float* work ); -lapack_int LAPACKE_dgeqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, double* a, lapack_int lda, - double* t, lapack_int ldt, double* work ); -lapack_int LAPACKE_cgeqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, lapack_complex_float* a, - lapack_int lda, lapack_complex_float* t, - lapack_int ldt, lapack_complex_float* work ); -lapack_int LAPACKE_zgeqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int nb, lapack_complex_double* a, - lapack_int lda, lapack_complex_double* t, - lapack_int ldt, lapack_complex_double* work ); - -lapack_int LAPACKE_sgeqrt2_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dgeqrt2_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* t, - lapack_int ldt ); -lapack_int LAPACKE_cgeqrt2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zgeqrt2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_sgeqrt3_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* t, - lapack_int ldt ); -lapack_int LAPACKE_dgeqrt3_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* t, - lapack_int ldt ); -lapack_int LAPACKE_cgeqrt3_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_zgeqrt3_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_stpmqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, const float* v, - lapack_int ldv, const float* t, lapack_int ldt, - float* a, lapack_int lda, float* b, - lapack_int ldb, float* work ); -lapack_int LAPACKE_dtpmqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, const double* v, - lapack_int ldv, const double* t, - lapack_int ldt, double* a, lapack_int lda, - double* b, lapack_int ldb, double* work ); -lapack_int LAPACKE_ctpmqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* work ); -lapack_int LAPACKE_ztpmqrt_work( int matrix_order, char side, char trans, - lapack_int m, lapack_int n, lapack_int k, - lapack_int l, lapack_int nb, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* work ); - -lapack_int LAPACKE_dtpqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, double* a, - lapack_int lda, double* b, lapack_int ldb, - double* t, lapack_int ldt, double* work ); -lapack_int LAPACKE_ctpqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* t, - lapack_complex_float* b, lapack_int ldb, - lapack_int ldt, lapack_complex_float* work ); -lapack_int LAPACKE_ztpqrt_work( int matrix_order, lapack_int m, lapack_int n, - lapack_int l, lapack_int nb, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* work ); - -lapack_int LAPACKE_stpqrt2_work( int matrix_order, lapack_int m, lapack_int n, - float* a, lapack_int lda, float* b, - lapack_int ldb, float* t, lapack_int ldt ); -lapack_int LAPACKE_dtpqrt2_work( int matrix_order, lapack_int m, lapack_int n, - double* a, lapack_int lda, double* b, - lapack_int ldb, double* t, lapack_int ldt ); -lapack_int LAPACKE_ctpqrt2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - lapack_complex_float* t, lapack_int ldt ); -lapack_int LAPACKE_ztpqrt2_work( int matrix_order, lapack_int m, lapack_int n, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - lapack_complex_double* t, lapack_int ldt ); - -lapack_int LAPACKE_stprfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, lapack_int l, - const float* v, lapack_int ldv, const float* t, - lapack_int ldt, float* a, lapack_int lda, - float* b, lapack_int ldb, const float* mywork, - lapack_int myldwork ); -lapack_int LAPACKE_dtprfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, lapack_int l, - const double* v, lapack_int ldv, - const double* t, lapack_int ldt, double* a, - lapack_int lda, double* b, lapack_int ldb, - const double* mywork, lapack_int myldwork ); -lapack_int LAPACKE_ctprfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, lapack_int l, - const lapack_complex_float* v, lapack_int ldv, - const lapack_complex_float* t, lapack_int ldt, - lapack_complex_float* a, lapack_int lda, - lapack_complex_float* b, lapack_int ldb, - const float* mywork, lapack_int myldwork ); -lapack_int LAPACKE_ztprfb_work( int matrix_order, char side, char trans, - char direct, char storev, lapack_int m, - lapack_int n, lapack_int k, lapack_int l, - const lapack_complex_double* v, lapack_int ldv, - const lapack_complex_double* t, lapack_int ldt, - lapack_complex_double* a, lapack_int lda, - lapack_complex_double* b, lapack_int ldb, - const double* mywork, lapack_int myldwork ); -//LAPACK 3.X.X -lapack_int LAPACKE_csyr( int matrix_order, char uplo, lapack_int n, - lapack_complex_float alpha, - const lapack_complex_float* x, lapack_int incx, - lapack_complex_float* a, lapack_int lda ); -lapack_int LAPACKE_zsyr( int matrix_order, char uplo, lapack_int n, - lapack_complex_double alpha, - const lapack_complex_double* x, lapack_int incx, - lapack_complex_double* a, lapack_int lda ); - -lapack_int LAPACKE_csyr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_float alpha, - const lapack_complex_float* x, - lapack_int incx, lapack_complex_float* a, - lapack_int lda ); -lapack_int LAPACKE_zsyr_work( int matrix_order, char uplo, lapack_int n, - lapack_complex_double alpha, - const lapack_complex_double* x, - lapack_int incx, lapack_complex_double* a, - lapack_int lda ); - - - -#define LAPACK_sgetrf LAPACK_GLOBAL(sgetrf,SGETRF) -#define LAPACK_dgetrf LAPACK_GLOBAL(dgetrf,DGETRF) -#define LAPACK_cgetrf LAPACK_GLOBAL(cgetrf,CGETRF) -#define LAPACK_zgetrf LAPACK_GLOBAL(zgetrf,ZGETRF) -#define LAPACK_sgbtrf LAPACK_GLOBAL(sgbtrf,SGBTRF) -#define LAPACK_dgbtrf LAPACK_GLOBAL(dgbtrf,DGBTRF) -#define LAPACK_cgbtrf LAPACK_GLOBAL(cgbtrf,CGBTRF) -#define LAPACK_zgbtrf LAPACK_GLOBAL(zgbtrf,ZGBTRF) -#define LAPACK_sgttrf LAPACK_GLOBAL(sgttrf,SGTTRF) -#define LAPACK_dgttrf LAPACK_GLOBAL(dgttrf,DGTTRF) -#define LAPACK_cgttrf LAPACK_GLOBAL(cgttrf,CGTTRF) -#define LAPACK_zgttrf LAPACK_GLOBAL(zgttrf,ZGTTRF) -#define LAPACK_spotrf LAPACK_GLOBAL(spotrf,SPOTRF) -#define LAPACK_dpotrf LAPACK_GLOBAL(dpotrf,DPOTRF) -#define LAPACK_cpotrf LAPACK_GLOBAL(cpotrf,CPOTRF) -#define LAPACK_zpotrf LAPACK_GLOBAL(zpotrf,ZPOTRF) -#define LAPACK_dpstrf LAPACK_GLOBAL(dpstrf,DPSTRF) -#define LAPACK_spstrf LAPACK_GLOBAL(spstrf,SPSTRF) -#define LAPACK_zpstrf LAPACK_GLOBAL(zpstrf,ZPSTRF) -#define LAPACK_cpstrf LAPACK_GLOBAL(cpstrf,CPSTRF) -#define LAPACK_dpftrf LAPACK_GLOBAL(dpftrf,DPFTRF) -#define LAPACK_spftrf LAPACK_GLOBAL(spftrf,SPFTRF) -#define LAPACK_zpftrf LAPACK_GLOBAL(zpftrf,ZPFTRF) -#define LAPACK_cpftrf LAPACK_GLOBAL(cpftrf,CPFTRF) -#define LAPACK_spptrf LAPACK_GLOBAL(spptrf,SPPTRF) -#define LAPACK_dpptrf LAPACK_GLOBAL(dpptrf,DPPTRF) -#define LAPACK_cpptrf LAPACK_GLOBAL(cpptrf,CPPTRF) -#define LAPACK_zpptrf LAPACK_GLOBAL(zpptrf,ZPPTRF) -#define LAPACK_spbtrf LAPACK_GLOBAL(spbtrf,SPBTRF) -#define LAPACK_dpbtrf LAPACK_GLOBAL(dpbtrf,DPBTRF) -#define LAPACK_cpbtrf LAPACK_GLOBAL(cpbtrf,CPBTRF) -#define LAPACK_zpbtrf LAPACK_GLOBAL(zpbtrf,ZPBTRF) -#define LAPACK_spttrf LAPACK_GLOBAL(spttrf,SPTTRF) -#define LAPACK_dpttrf LAPACK_GLOBAL(dpttrf,DPTTRF) -#define LAPACK_cpttrf LAPACK_GLOBAL(cpttrf,CPTTRF) -#define LAPACK_zpttrf LAPACK_GLOBAL(zpttrf,ZPTTRF) -#define LAPACK_ssytrf LAPACK_GLOBAL(ssytrf,SSYTRF) -#define LAPACK_dsytrf LAPACK_GLOBAL(dsytrf,DSYTRF) -#define LAPACK_csytrf LAPACK_GLOBAL(csytrf,CSYTRF) -#define LAPACK_zsytrf LAPACK_GLOBAL(zsytrf,ZSYTRF) -#define LAPACK_chetrf LAPACK_GLOBAL(chetrf,CHETRF) -#define LAPACK_zhetrf LAPACK_GLOBAL(zhetrf,ZHETRF) -#define LAPACK_ssptrf LAPACK_GLOBAL(ssptrf,SSPTRF) -#define LAPACK_dsptrf LAPACK_GLOBAL(dsptrf,DSPTRF) -#define LAPACK_csptrf LAPACK_GLOBAL(csptrf,CSPTRF) -#define LAPACK_zsptrf LAPACK_GLOBAL(zsptrf,ZSPTRF) -#define LAPACK_chptrf LAPACK_GLOBAL(chptrf,CHPTRF) -#define LAPACK_zhptrf LAPACK_GLOBAL(zhptrf,ZHPTRF) -#define LAPACK_sgetrs LAPACK_GLOBAL(sgetrs,SGETRS) -#define LAPACK_dgetrs LAPACK_GLOBAL(dgetrs,DGETRS) -#define LAPACK_cgetrs LAPACK_GLOBAL(cgetrs,CGETRS) -#define LAPACK_zgetrs LAPACK_GLOBAL(zgetrs,ZGETRS) -#define LAPACK_sgbtrs LAPACK_GLOBAL(sgbtrs,SGBTRS) -#define LAPACK_dgbtrs LAPACK_GLOBAL(dgbtrs,DGBTRS) -#define LAPACK_cgbtrs LAPACK_GLOBAL(cgbtrs,CGBTRS) -#define LAPACK_zgbtrs LAPACK_GLOBAL(zgbtrs,ZGBTRS) -#define LAPACK_sgttrs LAPACK_GLOBAL(sgttrs,SGTTRS) -#define LAPACK_dgttrs LAPACK_GLOBAL(dgttrs,DGTTRS) -#define LAPACK_cgttrs LAPACK_GLOBAL(cgttrs,CGTTRS) -#define LAPACK_zgttrs LAPACK_GLOBAL(zgttrs,ZGTTRS) -#define LAPACK_spotrs LAPACK_GLOBAL(spotrs,SPOTRS) -#define LAPACK_dpotrs LAPACK_GLOBAL(dpotrs,DPOTRS) -#define LAPACK_cpotrs LAPACK_GLOBAL(cpotrs,CPOTRS) -#define LAPACK_zpotrs LAPACK_GLOBAL(zpotrs,ZPOTRS) -#define LAPACK_dpftrs LAPACK_GLOBAL(dpftrs,DPFTRS) -#define LAPACK_spftrs LAPACK_GLOBAL(spftrs,SPFTRS) -#define LAPACK_zpftrs LAPACK_GLOBAL(zpftrs,ZPFTRS) -#define LAPACK_cpftrs LAPACK_GLOBAL(cpftrs,CPFTRS) -#define LAPACK_spptrs LAPACK_GLOBAL(spptrs,SPPTRS) -#define LAPACK_dpptrs LAPACK_GLOBAL(dpptrs,DPPTRS) -#define LAPACK_cpptrs LAPACK_GLOBAL(cpptrs,CPPTRS) -#define LAPACK_zpptrs LAPACK_GLOBAL(zpptrs,ZPPTRS) -#define LAPACK_spbtrs LAPACK_GLOBAL(spbtrs,SPBTRS) -#define LAPACK_dpbtrs LAPACK_GLOBAL(dpbtrs,DPBTRS) -#define LAPACK_cpbtrs LAPACK_GLOBAL(cpbtrs,CPBTRS) -#define LAPACK_zpbtrs LAPACK_GLOBAL(zpbtrs,ZPBTRS) -#define LAPACK_spttrs LAPACK_GLOBAL(spttrs,SPTTRS) -#define LAPACK_dpttrs LAPACK_GLOBAL(dpttrs,DPTTRS) -#define LAPACK_cpttrs LAPACK_GLOBAL(cpttrs,CPTTRS) -#define LAPACK_zpttrs LAPACK_GLOBAL(zpttrs,ZPTTRS) -#define LAPACK_ssytrs LAPACK_GLOBAL(ssytrs,SSYTRS) -#define LAPACK_dsytrs LAPACK_GLOBAL(dsytrs,DSYTRS) -#define LAPACK_csytrs LAPACK_GLOBAL(csytrs,CSYTRS) -#define LAPACK_zsytrs LAPACK_GLOBAL(zsytrs,ZSYTRS) -#define LAPACK_chetrs LAPACK_GLOBAL(chetrs,CHETRS) -#define LAPACK_zhetrs LAPACK_GLOBAL(zhetrs,ZHETRS) -#define LAPACK_ssptrs LAPACK_GLOBAL(ssptrs,SSPTRS) -#define LAPACK_dsptrs LAPACK_GLOBAL(dsptrs,DSPTRS) -#define LAPACK_csptrs LAPACK_GLOBAL(csptrs,CSPTRS) -#define LAPACK_zsptrs LAPACK_GLOBAL(zsptrs,ZSPTRS) -#define LAPACK_chptrs LAPACK_GLOBAL(chptrs,CHPTRS) -#define LAPACK_zhptrs LAPACK_GLOBAL(zhptrs,ZHPTRS) -#define LAPACK_strtrs LAPACK_GLOBAL(strtrs,STRTRS) -#define LAPACK_dtrtrs LAPACK_GLOBAL(dtrtrs,DTRTRS) -#define LAPACK_ctrtrs LAPACK_GLOBAL(ctrtrs,CTRTRS) -#define LAPACK_ztrtrs LAPACK_GLOBAL(ztrtrs,ZTRTRS) -#define LAPACK_stptrs LAPACK_GLOBAL(stptrs,STPTRS) -#define LAPACK_dtptrs LAPACK_GLOBAL(dtptrs,DTPTRS) -#define LAPACK_ctptrs LAPACK_GLOBAL(ctptrs,CTPTRS) -#define LAPACK_ztptrs LAPACK_GLOBAL(ztptrs,ZTPTRS) -#define LAPACK_stbtrs LAPACK_GLOBAL(stbtrs,STBTRS) -#define LAPACK_dtbtrs LAPACK_GLOBAL(dtbtrs,DTBTRS) -#define LAPACK_ctbtrs LAPACK_GLOBAL(ctbtrs,CTBTRS) -#define LAPACK_ztbtrs LAPACK_GLOBAL(ztbtrs,ZTBTRS) -#define LAPACK_sgecon LAPACK_GLOBAL(sgecon,SGECON) -#define LAPACK_dgecon LAPACK_GLOBAL(dgecon,DGECON) -#define LAPACK_cgecon LAPACK_GLOBAL(cgecon,CGECON) -#define LAPACK_zgecon LAPACK_GLOBAL(zgecon,ZGECON) -#define LAPACK_sgbcon LAPACK_GLOBAL(sgbcon,SGBCON) -#define LAPACK_dgbcon LAPACK_GLOBAL(dgbcon,DGBCON) -#define LAPACK_cgbcon LAPACK_GLOBAL(cgbcon,CGBCON) -#define LAPACK_zgbcon LAPACK_GLOBAL(zgbcon,ZGBCON) -#define LAPACK_sgtcon LAPACK_GLOBAL(sgtcon,SGTCON) -#define LAPACK_dgtcon LAPACK_GLOBAL(dgtcon,DGTCON) -#define LAPACK_cgtcon LAPACK_GLOBAL(cgtcon,CGTCON) -#define LAPACK_zgtcon LAPACK_GLOBAL(zgtcon,ZGTCON) -#define LAPACK_spocon LAPACK_GLOBAL(spocon,SPOCON) -#define LAPACK_dpocon LAPACK_GLOBAL(dpocon,DPOCON) -#define LAPACK_cpocon LAPACK_GLOBAL(cpocon,CPOCON) -#define LAPACK_zpocon LAPACK_GLOBAL(zpocon,ZPOCON) -#define LAPACK_sppcon LAPACK_GLOBAL(sppcon,SPPCON) -#define LAPACK_dppcon LAPACK_GLOBAL(dppcon,DPPCON) -#define LAPACK_cppcon LAPACK_GLOBAL(cppcon,CPPCON) -#define LAPACK_zppcon LAPACK_GLOBAL(zppcon,ZPPCON) -#define LAPACK_spbcon LAPACK_GLOBAL(spbcon,SPBCON) -#define LAPACK_dpbcon LAPACK_GLOBAL(dpbcon,DPBCON) -#define LAPACK_cpbcon LAPACK_GLOBAL(cpbcon,CPBCON) -#define LAPACK_zpbcon LAPACK_GLOBAL(zpbcon,ZPBCON) -#define LAPACK_sptcon LAPACK_GLOBAL(sptcon,SPTCON) -#define LAPACK_dptcon LAPACK_GLOBAL(dptcon,DPTCON) -#define LAPACK_cptcon LAPACK_GLOBAL(cptcon,CPTCON) -#define LAPACK_zptcon LAPACK_GLOBAL(zptcon,ZPTCON) -#define LAPACK_ssycon LAPACK_GLOBAL(ssycon,SSYCON) -#define LAPACK_dsycon LAPACK_GLOBAL(dsycon,DSYCON) -#define LAPACK_csycon LAPACK_GLOBAL(csycon,CSYCON) -#define LAPACK_zsycon LAPACK_GLOBAL(zsycon,ZSYCON) -#define LAPACK_checon LAPACK_GLOBAL(checon,CHECON) -#define LAPACK_zhecon LAPACK_GLOBAL(zhecon,ZHECON) -#define LAPACK_sspcon LAPACK_GLOBAL(sspcon,SSPCON) -#define LAPACK_dspcon LAPACK_GLOBAL(dspcon,DSPCON) -#define LAPACK_cspcon LAPACK_GLOBAL(cspcon,CSPCON) -#define LAPACK_zspcon LAPACK_GLOBAL(zspcon,ZSPCON) -#define LAPACK_chpcon LAPACK_GLOBAL(chpcon,CHPCON) -#define LAPACK_zhpcon LAPACK_GLOBAL(zhpcon,ZHPCON) -#define LAPACK_strcon LAPACK_GLOBAL(strcon,STRCON) -#define LAPACK_dtrcon LAPACK_GLOBAL(dtrcon,DTRCON) -#define LAPACK_ctrcon LAPACK_GLOBAL(ctrcon,CTRCON) -#define LAPACK_ztrcon LAPACK_GLOBAL(ztrcon,ZTRCON) -#define LAPACK_stpcon LAPACK_GLOBAL(stpcon,STPCON) -#define LAPACK_dtpcon LAPACK_GLOBAL(dtpcon,DTPCON) -#define LAPACK_ctpcon LAPACK_GLOBAL(ctpcon,CTPCON) -#define LAPACK_ztpcon LAPACK_GLOBAL(ztpcon,ZTPCON) -#define LAPACK_stbcon LAPACK_GLOBAL(stbcon,STBCON) -#define LAPACK_dtbcon LAPACK_GLOBAL(dtbcon,DTBCON) -#define LAPACK_ctbcon LAPACK_GLOBAL(ctbcon,CTBCON) -#define LAPACK_ztbcon LAPACK_GLOBAL(ztbcon,ZTBCON) -#define LAPACK_sgerfs LAPACK_GLOBAL(sgerfs,SGERFS) -#define LAPACK_dgerfs LAPACK_GLOBAL(dgerfs,DGERFS) -#define LAPACK_cgerfs LAPACK_GLOBAL(cgerfs,CGERFS) -#define LAPACK_zgerfs LAPACK_GLOBAL(zgerfs,ZGERFS) -#define LAPACK_dgerfsx LAPACK_GLOBAL(dgerfsx,DGERFSX) -#define LAPACK_sgerfsx LAPACK_GLOBAL(sgerfsx,SGERFSX) -#define LAPACK_zgerfsx LAPACK_GLOBAL(zgerfsx,ZGERFSX) -#define LAPACK_cgerfsx LAPACK_GLOBAL(cgerfsx,CGERFSX) -#define LAPACK_sgbrfs LAPACK_GLOBAL(sgbrfs,SGBRFS) -#define LAPACK_dgbrfs LAPACK_GLOBAL(dgbrfs,DGBRFS) -#define LAPACK_cgbrfs LAPACK_GLOBAL(cgbrfs,CGBRFS) -#define LAPACK_zgbrfs LAPACK_GLOBAL(zgbrfs,ZGBRFS) -#define LAPACK_dgbrfsx LAPACK_GLOBAL(dgbrfsx,DGBRFSX) -#define LAPACK_sgbrfsx LAPACK_GLOBAL(sgbrfsx,SGBRFSX) -#define LAPACK_zgbrfsx LAPACK_GLOBAL(zgbrfsx,ZGBRFSX) -#define LAPACK_cgbrfsx LAPACK_GLOBAL(cgbrfsx,CGBRFSX) -#define LAPACK_sgtrfs LAPACK_GLOBAL(sgtrfs,SGTRFS) -#define LAPACK_dgtrfs LAPACK_GLOBAL(dgtrfs,DGTRFS) -#define LAPACK_cgtrfs LAPACK_GLOBAL(cgtrfs,CGTRFS) -#define LAPACK_zgtrfs LAPACK_GLOBAL(zgtrfs,ZGTRFS) -#define LAPACK_sporfs LAPACK_GLOBAL(sporfs,SPORFS) -#define LAPACK_dporfs LAPACK_GLOBAL(dporfs,DPORFS) -#define LAPACK_cporfs LAPACK_GLOBAL(cporfs,CPORFS) -#define LAPACK_zporfs LAPACK_GLOBAL(zporfs,ZPORFS) -#define LAPACK_dporfsx LAPACK_GLOBAL(dporfsx,DPORFSX) -#define LAPACK_sporfsx LAPACK_GLOBAL(sporfsx,SPORFSX) -#define LAPACK_zporfsx LAPACK_GLOBAL(zporfsx,ZPORFSX) -#define LAPACK_cporfsx LAPACK_GLOBAL(cporfsx,CPORFSX) -#define LAPACK_spprfs LAPACK_GLOBAL(spprfs,SPPRFS) -#define LAPACK_dpprfs LAPACK_GLOBAL(dpprfs,DPPRFS) -#define LAPACK_cpprfs LAPACK_GLOBAL(cpprfs,CPPRFS) -#define LAPACK_zpprfs LAPACK_GLOBAL(zpprfs,ZPPRFS) -#define LAPACK_spbrfs LAPACK_GLOBAL(spbrfs,SPBRFS) -#define LAPACK_dpbrfs LAPACK_GLOBAL(dpbrfs,DPBRFS) -#define LAPACK_cpbrfs LAPACK_GLOBAL(cpbrfs,CPBRFS) -#define LAPACK_zpbrfs LAPACK_GLOBAL(zpbrfs,ZPBRFS) -#define LAPACK_sptrfs LAPACK_GLOBAL(sptrfs,SPTRFS) -#define LAPACK_dptrfs LAPACK_GLOBAL(dptrfs,DPTRFS) -#define LAPACK_cptrfs LAPACK_GLOBAL(cptrfs,CPTRFS) -#define LAPACK_zptrfs LAPACK_GLOBAL(zptrfs,ZPTRFS) -#define LAPACK_ssyrfs LAPACK_GLOBAL(ssyrfs,SSYRFS) -#define LAPACK_dsyrfs LAPACK_GLOBAL(dsyrfs,DSYRFS) -#define LAPACK_csyrfs LAPACK_GLOBAL(csyrfs,CSYRFS) -#define LAPACK_zsyrfs LAPACK_GLOBAL(zsyrfs,ZSYRFS) -#define LAPACK_dsyrfsx LAPACK_GLOBAL(dsyrfsx,DSYRFSX) -#define LAPACK_ssyrfsx LAPACK_GLOBAL(ssyrfsx,SSYRFSX) -#define LAPACK_zsyrfsx LAPACK_GLOBAL(zsyrfsx,ZSYRFSX) -#define LAPACK_csyrfsx LAPACK_GLOBAL(csyrfsx,CSYRFSX) -#define LAPACK_cherfs LAPACK_GLOBAL(cherfs,CHERFS) -#define LAPACK_zherfs LAPACK_GLOBAL(zherfs,ZHERFS) -#define LAPACK_zherfsx LAPACK_GLOBAL(zherfsx,ZHERFSX) -#define LAPACK_cherfsx LAPACK_GLOBAL(cherfsx,CHERFSX) -#define LAPACK_ssprfs LAPACK_GLOBAL(ssprfs,SSPRFS) -#define LAPACK_dsprfs LAPACK_GLOBAL(dsprfs,DSPRFS) -#define LAPACK_csprfs LAPACK_GLOBAL(csprfs,CSPRFS) -#define LAPACK_zsprfs LAPACK_GLOBAL(zsprfs,ZSPRFS) -#define LAPACK_chprfs LAPACK_GLOBAL(chprfs,CHPRFS) -#define LAPACK_zhprfs LAPACK_GLOBAL(zhprfs,ZHPRFS) -#define LAPACK_strrfs LAPACK_GLOBAL(strrfs,STRRFS) -#define LAPACK_dtrrfs LAPACK_GLOBAL(dtrrfs,DTRRFS) -#define LAPACK_ctrrfs LAPACK_GLOBAL(ctrrfs,CTRRFS) -#define LAPACK_ztrrfs LAPACK_GLOBAL(ztrrfs,ZTRRFS) -#define LAPACK_stprfs LAPACK_GLOBAL(stprfs,STPRFS) -#define LAPACK_dtprfs LAPACK_GLOBAL(dtprfs,DTPRFS) -#define LAPACK_ctprfs LAPACK_GLOBAL(ctprfs,CTPRFS) -#define LAPACK_ztprfs LAPACK_GLOBAL(ztprfs,ZTPRFS) -#define LAPACK_stbrfs LAPACK_GLOBAL(stbrfs,STBRFS) -#define LAPACK_dtbrfs LAPACK_GLOBAL(dtbrfs,DTBRFS) -#define LAPACK_ctbrfs LAPACK_GLOBAL(ctbrfs,CTBRFS) -#define LAPACK_ztbrfs LAPACK_GLOBAL(ztbrfs,ZTBRFS) -#define LAPACK_sgetri LAPACK_GLOBAL(sgetri,SGETRI) -#define LAPACK_dgetri LAPACK_GLOBAL(dgetri,DGETRI) -#define LAPACK_cgetri LAPACK_GLOBAL(cgetri,CGETRI) -#define LAPACK_zgetri LAPACK_GLOBAL(zgetri,ZGETRI) -#define LAPACK_spotri LAPACK_GLOBAL(spotri,SPOTRI) -#define LAPACK_dpotri LAPACK_GLOBAL(dpotri,DPOTRI) -#define LAPACK_cpotri LAPACK_GLOBAL(cpotri,CPOTRI) -#define LAPACK_zpotri LAPACK_GLOBAL(zpotri,ZPOTRI) -#define LAPACK_dpftri LAPACK_GLOBAL(dpftri,DPFTRI) -#define LAPACK_spftri LAPACK_GLOBAL(spftri,SPFTRI) -#define LAPACK_zpftri LAPACK_GLOBAL(zpftri,ZPFTRI) -#define LAPACK_cpftri LAPACK_GLOBAL(cpftri,CPFTRI) -#define LAPACK_spptri LAPACK_GLOBAL(spptri,SPPTRI) -#define LAPACK_dpptri LAPACK_GLOBAL(dpptri,DPPTRI) -#define LAPACK_cpptri LAPACK_GLOBAL(cpptri,CPPTRI) -#define LAPACK_zpptri LAPACK_GLOBAL(zpptri,ZPPTRI) -#define LAPACK_ssytri LAPACK_GLOBAL(ssytri,SSYTRI) -#define LAPACK_dsytri LAPACK_GLOBAL(dsytri,DSYTRI) -#define LAPACK_csytri LAPACK_GLOBAL(csytri,CSYTRI) -#define LAPACK_zsytri LAPACK_GLOBAL(zsytri,ZSYTRI) -#define LAPACK_chetri LAPACK_GLOBAL(chetri,CHETRI) -#define LAPACK_zhetri LAPACK_GLOBAL(zhetri,ZHETRI) -#define LAPACK_ssptri LAPACK_GLOBAL(ssptri,SSPTRI) -#define LAPACK_dsptri LAPACK_GLOBAL(dsptri,DSPTRI) -#define LAPACK_csptri LAPACK_GLOBAL(csptri,CSPTRI) -#define LAPACK_zsptri LAPACK_GLOBAL(zsptri,ZSPTRI) -#define LAPACK_chptri LAPACK_GLOBAL(chptri,CHPTRI) -#define LAPACK_zhptri LAPACK_GLOBAL(zhptri,ZHPTRI) -#define LAPACK_strtri LAPACK_GLOBAL(strtri,STRTRI) -#define LAPACK_dtrtri LAPACK_GLOBAL(dtrtri,DTRTRI) -#define LAPACK_ctrtri LAPACK_GLOBAL(ctrtri,CTRTRI) -#define LAPACK_ztrtri LAPACK_GLOBAL(ztrtri,ZTRTRI) -#define LAPACK_dtftri LAPACK_GLOBAL(dtftri,DTFTRI) -#define LAPACK_stftri LAPACK_GLOBAL(stftri,STFTRI) -#define LAPACK_ztftri LAPACK_GLOBAL(ztftri,ZTFTRI) -#define LAPACK_ctftri LAPACK_GLOBAL(ctftri,CTFTRI) -#define LAPACK_stptri LAPACK_GLOBAL(stptri,STPTRI) -#define LAPACK_dtptri LAPACK_GLOBAL(dtptri,DTPTRI) -#define LAPACK_ctptri LAPACK_GLOBAL(ctptri,CTPTRI) -#define LAPACK_ztptri LAPACK_GLOBAL(ztptri,ZTPTRI) -#define LAPACK_sgeequ LAPACK_GLOBAL(sgeequ,SGEEQU) -#define LAPACK_dgeequ LAPACK_GLOBAL(dgeequ,DGEEQU) -#define LAPACK_cgeequ LAPACK_GLOBAL(cgeequ,CGEEQU) -#define LAPACK_zgeequ LAPACK_GLOBAL(zgeequ,ZGEEQU) -#define LAPACK_dgeequb LAPACK_GLOBAL(dgeequb,DGEEQUB) -#define LAPACK_sgeequb LAPACK_GLOBAL(sgeequb,SGEEQUB) -#define LAPACK_zgeequb LAPACK_GLOBAL(zgeequb,ZGEEQUB) -#define LAPACK_cgeequb LAPACK_GLOBAL(cgeequb,CGEEQUB) -#define LAPACK_sgbequ LAPACK_GLOBAL(sgbequ,SGBEQU) -#define LAPACK_dgbequ LAPACK_GLOBAL(dgbequ,DGBEQU) -#define LAPACK_cgbequ LAPACK_GLOBAL(cgbequ,CGBEQU) -#define LAPACK_zgbequ LAPACK_GLOBAL(zgbequ,ZGBEQU) -#define LAPACK_dgbequb LAPACK_GLOBAL(dgbequb,DGBEQUB) -#define LAPACK_sgbequb LAPACK_GLOBAL(sgbequb,SGBEQUB) -#define LAPACK_zgbequb LAPACK_GLOBAL(zgbequb,ZGBEQUB) -#define LAPACK_cgbequb LAPACK_GLOBAL(cgbequb,CGBEQUB) -#define LAPACK_spoequ LAPACK_GLOBAL(spoequ,SPOEQU) -#define LAPACK_dpoequ LAPACK_GLOBAL(dpoequ,DPOEQU) -#define LAPACK_cpoequ LAPACK_GLOBAL(cpoequ,CPOEQU) -#define LAPACK_zpoequ LAPACK_GLOBAL(zpoequ,ZPOEQU) -#define LAPACK_dpoequb LAPACK_GLOBAL(dpoequb,DPOEQUB) -#define LAPACK_spoequb LAPACK_GLOBAL(spoequb,SPOEQUB) -#define LAPACK_zpoequb LAPACK_GLOBAL(zpoequb,ZPOEQUB) -#define LAPACK_cpoequb LAPACK_GLOBAL(cpoequb,CPOEQUB) -#define LAPACK_sppequ LAPACK_GLOBAL(sppequ,SPPEQU) -#define LAPACK_dppequ LAPACK_GLOBAL(dppequ,DPPEQU) -#define LAPACK_cppequ LAPACK_GLOBAL(cppequ,CPPEQU) -#define LAPACK_zppequ LAPACK_GLOBAL(zppequ,ZPPEQU) -#define LAPACK_spbequ LAPACK_GLOBAL(spbequ,SPBEQU) -#define LAPACK_dpbequ LAPACK_GLOBAL(dpbequ,DPBEQU) -#define LAPACK_cpbequ LAPACK_GLOBAL(cpbequ,CPBEQU) -#define LAPACK_zpbequ LAPACK_GLOBAL(zpbequ,ZPBEQU) -#define LAPACK_dsyequb LAPACK_GLOBAL(dsyequb,DSYEQUB) -#define LAPACK_ssyequb LAPACK_GLOBAL(ssyequb,SSYEQUB) -#define LAPACK_zsyequb LAPACK_GLOBAL(zsyequb,ZSYEQUB) -#define LAPACK_csyequb LAPACK_GLOBAL(csyequb,CSYEQUB) -#define LAPACK_zheequb LAPACK_GLOBAL(zheequb,ZHEEQUB) -#define LAPACK_cheequb LAPACK_GLOBAL(cheequb,CHEEQUB) -#define LAPACK_sgesv LAPACK_GLOBAL(sgesv,SGESV) -#define LAPACK_dgesv LAPACK_GLOBAL(dgesv,DGESV) -#define LAPACK_cgesv LAPACK_GLOBAL(cgesv,CGESV) -#define LAPACK_zgesv LAPACK_GLOBAL(zgesv,ZGESV) -#define LAPACK_dsgesv LAPACK_GLOBAL(dsgesv,DSGESV) -#define LAPACK_zcgesv LAPACK_GLOBAL(zcgesv,ZCGESV) -#define LAPACK_sgesvx LAPACK_GLOBAL(sgesvx,SGESVX) -#define LAPACK_dgesvx LAPACK_GLOBAL(dgesvx,DGESVX) -#define LAPACK_cgesvx LAPACK_GLOBAL(cgesvx,CGESVX) -#define LAPACK_zgesvx LAPACK_GLOBAL(zgesvx,ZGESVX) -#define LAPACK_dgesvxx LAPACK_GLOBAL(dgesvxx,DGESVXX) -#define LAPACK_sgesvxx LAPACK_GLOBAL(sgesvxx,SGESVXX) -#define LAPACK_zgesvxx LAPACK_GLOBAL(zgesvxx,ZGESVXX) -#define LAPACK_cgesvxx LAPACK_GLOBAL(cgesvxx,CGESVXX) -#define LAPACK_sgbsv LAPACK_GLOBAL(sgbsv,SGBSV) -#define LAPACK_dgbsv LAPACK_GLOBAL(dgbsv,DGBSV) -#define LAPACK_cgbsv LAPACK_GLOBAL(cgbsv,CGBSV) -#define LAPACK_zgbsv LAPACK_GLOBAL(zgbsv,ZGBSV) -#define LAPACK_sgbsvx LAPACK_GLOBAL(sgbsvx,SGBSVX) -#define LAPACK_dgbsvx LAPACK_GLOBAL(dgbsvx,DGBSVX) -#define LAPACK_cgbsvx LAPACK_GLOBAL(cgbsvx,CGBSVX) -#define LAPACK_zgbsvx LAPACK_GLOBAL(zgbsvx,ZGBSVX) -#define LAPACK_dgbsvxx LAPACK_GLOBAL(dgbsvxx,DGBSVXX) -#define LAPACK_sgbsvxx LAPACK_GLOBAL(sgbsvxx,SGBSVXX) -#define LAPACK_zgbsvxx LAPACK_GLOBAL(zgbsvxx,ZGBSVXX) -#define LAPACK_cgbsvxx LAPACK_GLOBAL(cgbsvxx,CGBSVXX) -#define LAPACK_sgtsv LAPACK_GLOBAL(sgtsv,SGTSV) -#define LAPACK_dgtsv LAPACK_GLOBAL(dgtsv,DGTSV) -#define LAPACK_cgtsv LAPACK_GLOBAL(cgtsv,CGTSV) -#define LAPACK_zgtsv LAPACK_GLOBAL(zgtsv,ZGTSV) -#define LAPACK_sgtsvx LAPACK_GLOBAL(sgtsvx,SGTSVX) -#define LAPACK_dgtsvx LAPACK_GLOBAL(dgtsvx,DGTSVX) -#define LAPACK_cgtsvx LAPACK_GLOBAL(cgtsvx,CGTSVX) -#define LAPACK_zgtsvx LAPACK_GLOBAL(zgtsvx,ZGTSVX) -#define LAPACK_sposv LAPACK_GLOBAL(sposv,SPOSV) -#define LAPACK_dposv LAPACK_GLOBAL(dposv,DPOSV) -#define LAPACK_cposv LAPACK_GLOBAL(cposv,CPOSV) -#define LAPACK_zposv LAPACK_GLOBAL(zposv,ZPOSV) -#define LAPACK_dsposv LAPACK_GLOBAL(dsposv,DSPOSV) -#define LAPACK_zcposv LAPACK_GLOBAL(zcposv,ZCPOSV) -#define LAPACK_sposvx LAPACK_GLOBAL(sposvx,SPOSVX) -#define LAPACK_dposvx LAPACK_GLOBAL(dposvx,DPOSVX) -#define LAPACK_cposvx LAPACK_GLOBAL(cposvx,CPOSVX) -#define LAPACK_zposvx LAPACK_GLOBAL(zposvx,ZPOSVX) -#define LAPACK_dposvxx LAPACK_GLOBAL(dposvxx,DPOSVXX) -#define LAPACK_sposvxx LAPACK_GLOBAL(sposvxx,SPOSVXX) -#define LAPACK_zposvxx LAPACK_GLOBAL(zposvxx,ZPOSVXX) -#define LAPACK_cposvxx LAPACK_GLOBAL(cposvxx,CPOSVXX) -#define LAPACK_sppsv LAPACK_GLOBAL(sppsv,SPPSV) -#define LAPACK_dppsv LAPACK_GLOBAL(dppsv,DPPSV) -#define LAPACK_cppsv LAPACK_GLOBAL(cppsv,CPPSV) -#define LAPACK_zppsv LAPACK_GLOBAL(zppsv,ZPPSV) -#define LAPACK_sppsvx LAPACK_GLOBAL(sppsvx,SPPSVX) -#define LAPACK_dppsvx LAPACK_GLOBAL(dppsvx,DPPSVX) -#define LAPACK_cppsvx LAPACK_GLOBAL(cppsvx,CPPSVX) -#define LAPACK_zppsvx LAPACK_GLOBAL(zppsvx,ZPPSVX) -#define LAPACK_spbsv LAPACK_GLOBAL(spbsv,SPBSV) -#define LAPACK_dpbsv LAPACK_GLOBAL(dpbsv,DPBSV) -#define LAPACK_cpbsv LAPACK_GLOBAL(cpbsv,CPBSV) -#define LAPACK_zpbsv LAPACK_GLOBAL(zpbsv,ZPBSV) -#define LAPACK_spbsvx LAPACK_GLOBAL(spbsvx,SPBSVX) -#define LAPACK_dpbsvx LAPACK_GLOBAL(dpbsvx,DPBSVX) -#define LAPACK_cpbsvx LAPACK_GLOBAL(cpbsvx,CPBSVX) -#define LAPACK_zpbsvx LAPACK_GLOBAL(zpbsvx,ZPBSVX) -#define LAPACK_sptsv LAPACK_GLOBAL(sptsv,SPTSV) -#define LAPACK_dptsv LAPACK_GLOBAL(dptsv,DPTSV) -#define LAPACK_cptsv LAPACK_GLOBAL(cptsv,CPTSV) -#define LAPACK_zptsv LAPACK_GLOBAL(zptsv,ZPTSV) -#define LAPACK_sptsvx LAPACK_GLOBAL(sptsvx,SPTSVX) -#define LAPACK_dptsvx LAPACK_GLOBAL(dptsvx,DPTSVX) -#define LAPACK_cptsvx LAPACK_GLOBAL(cptsvx,CPTSVX) -#define LAPACK_zptsvx LAPACK_GLOBAL(zptsvx,ZPTSVX) -#define LAPACK_ssysv LAPACK_GLOBAL(ssysv,SSYSV) -#define LAPACK_dsysv LAPACK_GLOBAL(dsysv,DSYSV) -#define LAPACK_csysv LAPACK_GLOBAL(csysv,CSYSV) -#define LAPACK_zsysv LAPACK_GLOBAL(zsysv,ZSYSV) -#define LAPACK_ssysvx LAPACK_GLOBAL(ssysvx,SSYSVX) -#define LAPACK_dsysvx LAPACK_GLOBAL(dsysvx,DSYSVX) -#define LAPACK_csysvx LAPACK_GLOBAL(csysvx,CSYSVX) -#define LAPACK_zsysvx LAPACK_GLOBAL(zsysvx,ZSYSVX) -#define LAPACK_dsysvxx LAPACK_GLOBAL(dsysvxx,DSYSVXX) -#define LAPACK_ssysvxx LAPACK_GLOBAL(ssysvxx,SSYSVXX) -#define LAPACK_zsysvxx LAPACK_GLOBAL(zsysvxx,ZSYSVXX) -#define LAPACK_csysvxx LAPACK_GLOBAL(csysvxx,CSYSVXX) -#define LAPACK_chesv LAPACK_GLOBAL(chesv,CHESV) -#define LAPACK_zhesv LAPACK_GLOBAL(zhesv,ZHESV) -#define LAPACK_chesvx LAPACK_GLOBAL(chesvx,CHESVX) -#define LAPACK_zhesvx LAPACK_GLOBAL(zhesvx,ZHESVX) -#define LAPACK_zhesvxx LAPACK_GLOBAL(zhesvxx,ZHESVXX) -#define LAPACK_chesvxx LAPACK_GLOBAL(chesvxx,CHESVXX) -#define LAPACK_sspsv LAPACK_GLOBAL(sspsv,SSPSV) -#define LAPACK_dspsv LAPACK_GLOBAL(dspsv,DSPSV) -#define LAPACK_cspsv LAPACK_GLOBAL(cspsv,CSPSV) -#define LAPACK_zspsv LAPACK_GLOBAL(zspsv,ZSPSV) -#define LAPACK_sspsvx LAPACK_GLOBAL(sspsvx,SSPSVX) -#define LAPACK_dspsvx LAPACK_GLOBAL(dspsvx,DSPSVX) -#define LAPACK_cspsvx LAPACK_GLOBAL(cspsvx,CSPSVX) -#define LAPACK_zspsvx LAPACK_GLOBAL(zspsvx,ZSPSVX) -#define LAPACK_chpsv LAPACK_GLOBAL(chpsv,CHPSV) -#define LAPACK_zhpsv LAPACK_GLOBAL(zhpsv,ZHPSV) -#define LAPACK_chpsvx LAPACK_GLOBAL(chpsvx,CHPSVX) -#define LAPACK_zhpsvx LAPACK_GLOBAL(zhpsvx,ZHPSVX) -#define LAPACK_sgeqrf LAPACK_GLOBAL(sgeqrf,SGEQRF) -#define LAPACK_dgeqrf LAPACK_GLOBAL(dgeqrf,DGEQRF) -#define LAPACK_cgeqrf LAPACK_GLOBAL(cgeqrf,CGEQRF) -#define LAPACK_zgeqrf LAPACK_GLOBAL(zgeqrf,ZGEQRF) -#define LAPACK_sgeqpf LAPACK_GLOBAL(sgeqpf,SGEQPF) -#define LAPACK_dgeqpf LAPACK_GLOBAL(dgeqpf,DGEQPF) -#define LAPACK_cgeqpf LAPACK_GLOBAL(cgeqpf,CGEQPF) -#define LAPACK_zgeqpf LAPACK_GLOBAL(zgeqpf,ZGEQPF) -#define LAPACK_sgeqp3 LAPACK_GLOBAL(sgeqp3,SGEQP3) -#define LAPACK_dgeqp3 LAPACK_GLOBAL(dgeqp3,DGEQP3) -#define LAPACK_cgeqp3 LAPACK_GLOBAL(cgeqp3,CGEQP3) -#define LAPACK_zgeqp3 LAPACK_GLOBAL(zgeqp3,ZGEQP3) -#define LAPACK_sorgqr LAPACK_GLOBAL(sorgqr,SORGQR) -#define LAPACK_dorgqr LAPACK_GLOBAL(dorgqr,DORGQR) -#define LAPACK_sormqr LAPACK_GLOBAL(sormqr,SORMQR) -#define LAPACK_dormqr LAPACK_GLOBAL(dormqr,DORMQR) -#define LAPACK_cungqr LAPACK_GLOBAL(cungqr,CUNGQR) -#define LAPACK_zungqr LAPACK_GLOBAL(zungqr,ZUNGQR) -#define LAPACK_cunmqr LAPACK_GLOBAL(cunmqr,CUNMQR) -#define LAPACK_zunmqr LAPACK_GLOBAL(zunmqr,ZUNMQR) -#define LAPACK_sgelqf LAPACK_GLOBAL(sgelqf,SGELQF) -#define LAPACK_dgelqf LAPACK_GLOBAL(dgelqf,DGELQF) -#define LAPACK_cgelqf LAPACK_GLOBAL(cgelqf,CGELQF) -#define LAPACK_zgelqf LAPACK_GLOBAL(zgelqf,ZGELQF) -#define LAPACK_sorglq LAPACK_GLOBAL(sorglq,SORGLQ) -#define LAPACK_dorglq LAPACK_GLOBAL(dorglq,DORGLQ) -#define LAPACK_sormlq LAPACK_GLOBAL(sormlq,SORMLQ) -#define LAPACK_dormlq LAPACK_GLOBAL(dormlq,DORMLQ) -#define LAPACK_cunglq LAPACK_GLOBAL(cunglq,CUNGLQ) -#define LAPACK_zunglq LAPACK_GLOBAL(zunglq,ZUNGLQ) -#define LAPACK_cunmlq LAPACK_GLOBAL(cunmlq,CUNMLQ) -#define LAPACK_zunmlq LAPACK_GLOBAL(zunmlq,ZUNMLQ) -#define LAPACK_sgeqlf LAPACK_GLOBAL(sgeqlf,SGEQLF) -#define LAPACK_dgeqlf LAPACK_GLOBAL(dgeqlf,DGEQLF) -#define LAPACK_cgeqlf LAPACK_GLOBAL(cgeqlf,CGEQLF) -#define LAPACK_zgeqlf LAPACK_GLOBAL(zgeqlf,ZGEQLF) -#define LAPACK_sorgql LAPACK_GLOBAL(sorgql,SORGQL) -#define LAPACK_dorgql LAPACK_GLOBAL(dorgql,DORGQL) -#define LAPACK_cungql LAPACK_GLOBAL(cungql,CUNGQL) -#define LAPACK_zungql LAPACK_GLOBAL(zungql,ZUNGQL) -#define LAPACK_sormql LAPACK_GLOBAL(sormql,SORMQL) -#define LAPACK_dormql LAPACK_GLOBAL(dormql,DORMQL) -#define LAPACK_cunmql LAPACK_GLOBAL(cunmql,CUNMQL) -#define LAPACK_zunmql LAPACK_GLOBAL(zunmql,ZUNMQL) -#define LAPACK_sgerqf LAPACK_GLOBAL(sgerqf,SGERQF) -#define LAPACK_dgerqf LAPACK_GLOBAL(dgerqf,DGERQF) -#define LAPACK_cgerqf LAPACK_GLOBAL(cgerqf,CGERQF) -#define LAPACK_zgerqf LAPACK_GLOBAL(zgerqf,ZGERQF) -#define LAPACK_sorgrq LAPACK_GLOBAL(sorgrq,SORGRQ) -#define LAPACK_dorgrq LAPACK_GLOBAL(dorgrq,DORGRQ) -#define LAPACK_cungrq LAPACK_GLOBAL(cungrq,CUNGRQ) -#define LAPACK_zungrq LAPACK_GLOBAL(zungrq,ZUNGRQ) -#define LAPACK_sormrq LAPACK_GLOBAL(sormrq,SORMRQ) -#define LAPACK_dormrq LAPACK_GLOBAL(dormrq,DORMRQ) -#define LAPACK_cunmrq LAPACK_GLOBAL(cunmrq,CUNMRQ) -#define LAPACK_zunmrq LAPACK_GLOBAL(zunmrq,ZUNMRQ) -#define LAPACK_stzrzf LAPACK_GLOBAL(stzrzf,STZRZF) -#define LAPACK_dtzrzf LAPACK_GLOBAL(dtzrzf,DTZRZF) -#define LAPACK_ctzrzf LAPACK_GLOBAL(ctzrzf,CTZRZF) -#define LAPACK_ztzrzf LAPACK_GLOBAL(ztzrzf,ZTZRZF) -#define LAPACK_sormrz LAPACK_GLOBAL(sormrz,SORMRZ) -#define LAPACK_dormrz LAPACK_GLOBAL(dormrz,DORMRZ) -#define LAPACK_cunmrz LAPACK_GLOBAL(cunmrz,CUNMRZ) -#define LAPACK_zunmrz LAPACK_GLOBAL(zunmrz,ZUNMRZ) -#define LAPACK_sggqrf LAPACK_GLOBAL(sggqrf,SGGQRF) -#define LAPACK_dggqrf LAPACK_GLOBAL(dggqrf,DGGQRF) -#define LAPACK_cggqrf LAPACK_GLOBAL(cggqrf,CGGQRF) -#define LAPACK_zggqrf LAPACK_GLOBAL(zggqrf,ZGGQRF) -#define LAPACK_sggrqf LAPACK_GLOBAL(sggrqf,SGGRQF) -#define LAPACK_dggrqf LAPACK_GLOBAL(dggrqf,DGGRQF) -#define LAPACK_cggrqf LAPACK_GLOBAL(cggrqf,CGGRQF) -#define LAPACK_zggrqf LAPACK_GLOBAL(zggrqf,ZGGRQF) -#define LAPACK_sgebrd LAPACK_GLOBAL(sgebrd,SGEBRD) -#define LAPACK_dgebrd LAPACK_GLOBAL(dgebrd,DGEBRD) -#define LAPACK_cgebrd LAPACK_GLOBAL(cgebrd,CGEBRD) -#define LAPACK_zgebrd LAPACK_GLOBAL(zgebrd,ZGEBRD) -#define LAPACK_sgbbrd LAPACK_GLOBAL(sgbbrd,SGBBRD) -#define LAPACK_dgbbrd LAPACK_GLOBAL(dgbbrd,DGBBRD) -#define LAPACK_cgbbrd LAPACK_GLOBAL(cgbbrd,CGBBRD) -#define LAPACK_zgbbrd LAPACK_GLOBAL(zgbbrd,ZGBBRD) -#define LAPACK_sorgbr LAPACK_GLOBAL(sorgbr,SORGBR) -#define LAPACK_dorgbr LAPACK_GLOBAL(dorgbr,DORGBR) -#define LAPACK_sormbr LAPACK_GLOBAL(sormbr,SORMBR) -#define LAPACK_dormbr LAPACK_GLOBAL(dormbr,DORMBR) -#define LAPACK_cungbr LAPACK_GLOBAL(cungbr,CUNGBR) -#define LAPACK_zungbr LAPACK_GLOBAL(zungbr,ZUNGBR) -#define LAPACK_cunmbr LAPACK_GLOBAL(cunmbr,CUNMBR) -#define LAPACK_zunmbr LAPACK_GLOBAL(zunmbr,ZUNMBR) -#define LAPACK_sbdsqr LAPACK_GLOBAL(sbdsqr,SBDSQR) -#define LAPACK_dbdsqr LAPACK_GLOBAL(dbdsqr,DBDSQR) -#define LAPACK_cbdsqr LAPACK_GLOBAL(cbdsqr,CBDSQR) -#define LAPACK_zbdsqr LAPACK_GLOBAL(zbdsqr,ZBDSQR) -#define LAPACK_sbdsdc LAPACK_GLOBAL(sbdsdc,SBDSDC) -#define LAPACK_dbdsdc LAPACK_GLOBAL(dbdsdc,DBDSDC) -#define LAPACK_ssytrd LAPACK_GLOBAL(ssytrd,SSYTRD) -#define LAPACK_dsytrd LAPACK_GLOBAL(dsytrd,DSYTRD) -#define LAPACK_sorgtr LAPACK_GLOBAL(sorgtr,SORGTR) -#define LAPACK_dorgtr LAPACK_GLOBAL(dorgtr,DORGTR) -#define LAPACK_sormtr LAPACK_GLOBAL(sormtr,SORMTR) -#define LAPACK_dormtr LAPACK_GLOBAL(dormtr,DORMTR) -#define LAPACK_chetrd LAPACK_GLOBAL(chetrd,CHETRD) -#define LAPACK_zhetrd LAPACK_GLOBAL(zhetrd,ZHETRD) -#define LAPACK_cungtr LAPACK_GLOBAL(cungtr,CUNGTR) -#define LAPACK_zungtr LAPACK_GLOBAL(zungtr,ZUNGTR) -#define LAPACK_cunmtr LAPACK_GLOBAL(cunmtr,CUNMTR) -#define LAPACK_zunmtr LAPACK_GLOBAL(zunmtr,ZUNMTR) -#define LAPACK_ssptrd LAPACK_GLOBAL(ssptrd,SSPTRD) -#define LAPACK_dsptrd LAPACK_GLOBAL(dsptrd,DSPTRD) -#define LAPACK_sopgtr LAPACK_GLOBAL(sopgtr,SOPGTR) -#define LAPACK_dopgtr LAPACK_GLOBAL(dopgtr,DOPGTR) -#define LAPACK_sopmtr LAPACK_GLOBAL(sopmtr,SOPMTR) -#define LAPACK_dopmtr LAPACK_GLOBAL(dopmtr,DOPMTR) -#define LAPACK_chptrd LAPACK_GLOBAL(chptrd,CHPTRD) -#define LAPACK_zhptrd LAPACK_GLOBAL(zhptrd,ZHPTRD) -#define LAPACK_cupgtr LAPACK_GLOBAL(cupgtr,CUPGTR) -#define LAPACK_zupgtr LAPACK_GLOBAL(zupgtr,ZUPGTR) -#define LAPACK_cupmtr LAPACK_GLOBAL(cupmtr,CUPMTR) -#define LAPACK_zupmtr LAPACK_GLOBAL(zupmtr,ZUPMTR) -#define LAPACK_ssbtrd LAPACK_GLOBAL(ssbtrd,SSBTRD) -#define LAPACK_dsbtrd LAPACK_GLOBAL(dsbtrd,DSBTRD) -#define LAPACK_chbtrd LAPACK_GLOBAL(chbtrd,CHBTRD) -#define LAPACK_zhbtrd LAPACK_GLOBAL(zhbtrd,ZHBTRD) -#define LAPACK_ssterf LAPACK_GLOBAL(ssterf,SSTERF) -#define LAPACK_dsterf LAPACK_GLOBAL(dsterf,DSTERF) -#define LAPACK_ssteqr LAPACK_GLOBAL(ssteqr,SSTEQR) -#define LAPACK_dsteqr LAPACK_GLOBAL(dsteqr,DSTEQR) -#define LAPACK_csteqr LAPACK_GLOBAL(csteqr,CSTEQR) -#define LAPACK_zsteqr LAPACK_GLOBAL(zsteqr,ZSTEQR) -#define LAPACK_sstemr LAPACK_GLOBAL(sstemr,SSTEMR) -#define LAPACK_dstemr LAPACK_GLOBAL(dstemr,DSTEMR) -#define LAPACK_cstemr LAPACK_GLOBAL(cstemr,CSTEMR) -#define LAPACK_zstemr LAPACK_GLOBAL(zstemr,ZSTEMR) -#define LAPACK_sstedc LAPACK_GLOBAL(sstedc,SSTEDC) -#define LAPACK_dstedc LAPACK_GLOBAL(dstedc,DSTEDC) -#define LAPACK_cstedc LAPACK_GLOBAL(cstedc,CSTEDC) -#define LAPACK_zstedc LAPACK_GLOBAL(zstedc,ZSTEDC) -#define LAPACK_sstegr LAPACK_GLOBAL(sstegr,SSTEGR) -#define LAPACK_dstegr LAPACK_GLOBAL(dstegr,DSTEGR) -#define LAPACK_cstegr LAPACK_GLOBAL(cstegr,CSTEGR) -#define LAPACK_zstegr LAPACK_GLOBAL(zstegr,ZSTEGR) -#define LAPACK_spteqr LAPACK_GLOBAL(spteqr,SPTEQR) -#define LAPACK_dpteqr LAPACK_GLOBAL(dpteqr,DPTEQR) -#define LAPACK_cpteqr LAPACK_GLOBAL(cpteqr,CPTEQR) -#define LAPACK_zpteqr LAPACK_GLOBAL(zpteqr,ZPTEQR) -#define LAPACK_sstebz LAPACK_GLOBAL(sstebz,SSTEBZ) -#define LAPACK_dstebz LAPACK_GLOBAL(dstebz,DSTEBZ) -#define LAPACK_sstein LAPACK_GLOBAL(sstein,SSTEIN) -#define LAPACK_dstein LAPACK_GLOBAL(dstein,DSTEIN) -#define LAPACK_cstein LAPACK_GLOBAL(cstein,CSTEIN) -#define LAPACK_zstein LAPACK_GLOBAL(zstein,ZSTEIN) -#define LAPACK_sdisna LAPACK_GLOBAL(sdisna,SDISNA) -#define LAPACK_ddisna LAPACK_GLOBAL(ddisna,DDISNA) -#define LAPACK_ssygst LAPACK_GLOBAL(ssygst,SSYGST) -#define LAPACK_dsygst LAPACK_GLOBAL(dsygst,DSYGST) -#define LAPACK_chegst LAPACK_GLOBAL(chegst,CHEGST) -#define LAPACK_zhegst LAPACK_GLOBAL(zhegst,ZHEGST) -#define LAPACK_sspgst LAPACK_GLOBAL(sspgst,SSPGST) -#define LAPACK_dspgst LAPACK_GLOBAL(dspgst,DSPGST) -#define LAPACK_chpgst LAPACK_GLOBAL(chpgst,CHPGST) -#define LAPACK_zhpgst LAPACK_GLOBAL(zhpgst,ZHPGST) -#define LAPACK_ssbgst LAPACK_GLOBAL(ssbgst,SSBGST) -#define LAPACK_dsbgst LAPACK_GLOBAL(dsbgst,DSBGST) -#define LAPACK_chbgst LAPACK_GLOBAL(chbgst,CHBGST) -#define LAPACK_zhbgst LAPACK_GLOBAL(zhbgst,ZHBGST) -#define LAPACK_spbstf LAPACK_GLOBAL(spbstf,SPBSTF) -#define LAPACK_dpbstf LAPACK_GLOBAL(dpbstf,DPBSTF) -#define LAPACK_cpbstf LAPACK_GLOBAL(cpbstf,CPBSTF) -#define LAPACK_zpbstf LAPACK_GLOBAL(zpbstf,ZPBSTF) -#define LAPACK_sgehrd LAPACK_GLOBAL(sgehrd,SGEHRD) -#define LAPACK_dgehrd LAPACK_GLOBAL(dgehrd,DGEHRD) -#define LAPACK_cgehrd LAPACK_GLOBAL(cgehrd,CGEHRD) -#define LAPACK_zgehrd LAPACK_GLOBAL(zgehrd,ZGEHRD) -#define LAPACK_sorghr LAPACK_GLOBAL(sorghr,SORGHR) -#define LAPACK_dorghr LAPACK_GLOBAL(dorghr,DORGHR) -#define LAPACK_sormhr LAPACK_GLOBAL(sormhr,SORMHR) -#define LAPACK_dormhr LAPACK_GLOBAL(dormhr,DORMHR) -#define LAPACK_cunghr LAPACK_GLOBAL(cunghr,CUNGHR) -#define LAPACK_zunghr LAPACK_GLOBAL(zunghr,ZUNGHR) -#define LAPACK_cunmhr LAPACK_GLOBAL(cunmhr,CUNMHR) -#define LAPACK_zunmhr LAPACK_GLOBAL(zunmhr,ZUNMHR) -#define LAPACK_sgebal LAPACK_GLOBAL(sgebal,SGEBAL) -#define LAPACK_dgebal LAPACK_GLOBAL(dgebal,DGEBAL) -#define LAPACK_cgebal LAPACK_GLOBAL(cgebal,CGEBAL) -#define LAPACK_zgebal LAPACK_GLOBAL(zgebal,ZGEBAL) -#define LAPACK_sgebak LAPACK_GLOBAL(sgebak,SGEBAK) -#define LAPACK_dgebak LAPACK_GLOBAL(dgebak,DGEBAK) -#define LAPACK_cgebak LAPACK_GLOBAL(cgebak,CGEBAK) -#define LAPACK_zgebak LAPACK_GLOBAL(zgebak,ZGEBAK) -#define LAPACK_shseqr LAPACK_GLOBAL(shseqr,SHSEQR) -#define LAPACK_dhseqr LAPACK_GLOBAL(dhseqr,DHSEQR) -#define LAPACK_chseqr LAPACK_GLOBAL(chseqr,CHSEQR) -#define LAPACK_zhseqr LAPACK_GLOBAL(zhseqr,ZHSEQR) -#define LAPACK_shsein LAPACK_GLOBAL(shsein,SHSEIN) -#define LAPACK_dhsein LAPACK_GLOBAL(dhsein,DHSEIN) -#define LAPACK_chsein LAPACK_GLOBAL(chsein,CHSEIN) -#define LAPACK_zhsein LAPACK_GLOBAL(zhsein,ZHSEIN) -#define LAPACK_strevc LAPACK_GLOBAL(strevc,STREVC) -#define LAPACK_dtrevc LAPACK_GLOBAL(dtrevc,DTREVC) -#define LAPACK_ctrevc LAPACK_GLOBAL(ctrevc,CTREVC) -#define LAPACK_ztrevc LAPACK_GLOBAL(ztrevc,ZTREVC) -#define LAPACK_strsna LAPACK_GLOBAL(strsna,STRSNA) -#define LAPACK_dtrsna LAPACK_GLOBAL(dtrsna,DTRSNA) -#define LAPACK_ctrsna LAPACK_GLOBAL(ctrsna,CTRSNA) -#define LAPACK_ztrsna LAPACK_GLOBAL(ztrsna,ZTRSNA) -#define LAPACK_strexc LAPACK_GLOBAL(strexc,STREXC) -#define LAPACK_dtrexc LAPACK_GLOBAL(dtrexc,DTREXC) -#define LAPACK_ctrexc LAPACK_GLOBAL(ctrexc,CTREXC) -#define LAPACK_ztrexc LAPACK_GLOBAL(ztrexc,ZTREXC) -#define LAPACK_strsen LAPACK_GLOBAL(strsen,STRSEN) -#define LAPACK_dtrsen LAPACK_GLOBAL(dtrsen,DTRSEN) -#define LAPACK_ctrsen LAPACK_GLOBAL(ctrsen,CTRSEN) -#define LAPACK_ztrsen LAPACK_GLOBAL(ztrsen,ZTRSEN) -#define LAPACK_strsyl LAPACK_GLOBAL(strsyl,STRSYL) -#define LAPACK_dtrsyl LAPACK_GLOBAL(dtrsyl,DTRSYL) -#define LAPACK_ctrsyl LAPACK_GLOBAL(ctrsyl,CTRSYL) -#define LAPACK_ztrsyl LAPACK_GLOBAL(ztrsyl,ZTRSYL) -#define LAPACK_sgghrd LAPACK_GLOBAL(sgghrd,SGGHRD) -#define LAPACK_dgghrd LAPACK_GLOBAL(dgghrd,DGGHRD) -#define LAPACK_cgghrd LAPACK_GLOBAL(cgghrd,CGGHRD) -#define LAPACK_zgghrd LAPACK_GLOBAL(zgghrd,ZGGHRD) -#define LAPACK_sggbal LAPACK_GLOBAL(sggbal,SGGBAL) -#define LAPACK_dggbal LAPACK_GLOBAL(dggbal,DGGBAL) -#define LAPACK_cggbal LAPACK_GLOBAL(cggbal,CGGBAL) -#define LAPACK_zggbal LAPACK_GLOBAL(zggbal,ZGGBAL) -#define LAPACK_sggbak LAPACK_GLOBAL(sggbak,SGGBAK) -#define LAPACK_dggbak LAPACK_GLOBAL(dggbak,DGGBAK) -#define LAPACK_cggbak LAPACK_GLOBAL(cggbak,CGGBAK) -#define LAPACK_zggbak LAPACK_GLOBAL(zggbak,ZGGBAK) -#define LAPACK_shgeqz LAPACK_GLOBAL(shgeqz,SHGEQZ) -#define LAPACK_dhgeqz LAPACK_GLOBAL(dhgeqz,DHGEQZ) -#define LAPACK_chgeqz LAPACK_GLOBAL(chgeqz,CHGEQZ) -#define LAPACK_zhgeqz LAPACK_GLOBAL(zhgeqz,ZHGEQZ) -#define LAPACK_stgevc LAPACK_GLOBAL(stgevc,STGEVC) -#define LAPACK_dtgevc LAPACK_GLOBAL(dtgevc,DTGEVC) -#define LAPACK_ctgevc LAPACK_GLOBAL(ctgevc,CTGEVC) -#define LAPACK_ztgevc LAPACK_GLOBAL(ztgevc,ZTGEVC) -#define LAPACK_stgexc LAPACK_GLOBAL(stgexc,STGEXC) -#define LAPACK_dtgexc LAPACK_GLOBAL(dtgexc,DTGEXC) -#define LAPACK_ctgexc LAPACK_GLOBAL(ctgexc,CTGEXC) -#define LAPACK_ztgexc LAPACK_GLOBAL(ztgexc,ZTGEXC) -#define LAPACK_stgsen LAPACK_GLOBAL(stgsen,STGSEN) -#define LAPACK_dtgsen LAPACK_GLOBAL(dtgsen,DTGSEN) -#define LAPACK_ctgsen LAPACK_GLOBAL(ctgsen,CTGSEN) -#define LAPACK_ztgsen LAPACK_GLOBAL(ztgsen,ZTGSEN) -#define LAPACK_stgsyl LAPACK_GLOBAL(stgsyl,STGSYL) -#define LAPACK_dtgsyl LAPACK_GLOBAL(dtgsyl,DTGSYL) -#define LAPACK_ctgsyl LAPACK_GLOBAL(ctgsyl,CTGSYL) -#define LAPACK_ztgsyl LAPACK_GLOBAL(ztgsyl,ZTGSYL) -#define LAPACK_stgsna LAPACK_GLOBAL(stgsna,STGSNA) -#define LAPACK_dtgsna LAPACK_GLOBAL(dtgsna,DTGSNA) -#define LAPACK_ctgsna LAPACK_GLOBAL(ctgsna,CTGSNA) -#define LAPACK_ztgsna LAPACK_GLOBAL(ztgsna,ZTGSNA) -#define LAPACK_sggsvp LAPACK_GLOBAL(sggsvp,SGGSVP) -#define LAPACK_dggsvp LAPACK_GLOBAL(dggsvp,DGGSVP) -#define LAPACK_cggsvp LAPACK_GLOBAL(cggsvp,CGGSVP) -#define LAPACK_zggsvp LAPACK_GLOBAL(zggsvp,ZGGSVP) -#define LAPACK_stgsja LAPACK_GLOBAL(stgsja,STGSJA) -#define LAPACK_dtgsja LAPACK_GLOBAL(dtgsja,DTGSJA) -#define LAPACK_ctgsja LAPACK_GLOBAL(ctgsja,CTGSJA) -#define LAPACK_ztgsja LAPACK_GLOBAL(ztgsja,ZTGSJA) -#define LAPACK_sgels LAPACK_GLOBAL(sgels,SGELS) -#define LAPACK_dgels LAPACK_GLOBAL(dgels,DGELS) -#define LAPACK_cgels LAPACK_GLOBAL(cgels,CGELS) -#define LAPACK_zgels LAPACK_GLOBAL(zgels,ZGELS) -#define LAPACK_sgelsy LAPACK_GLOBAL(sgelsy,SGELSY) -#define LAPACK_dgelsy LAPACK_GLOBAL(dgelsy,DGELSY) -#define LAPACK_cgelsy LAPACK_GLOBAL(cgelsy,CGELSY) -#define LAPACK_zgelsy LAPACK_GLOBAL(zgelsy,ZGELSY) -#define LAPACK_sgelss LAPACK_GLOBAL(sgelss,SGELSS) -#define LAPACK_dgelss LAPACK_GLOBAL(dgelss,DGELSS) -#define LAPACK_cgelss LAPACK_GLOBAL(cgelss,CGELSS) -#define LAPACK_zgelss LAPACK_GLOBAL(zgelss,ZGELSS) -#define LAPACK_sgelsd LAPACK_GLOBAL(sgelsd,SGELSD) -#define LAPACK_dgelsd LAPACK_GLOBAL(dgelsd,DGELSD) -#define LAPACK_cgelsd LAPACK_GLOBAL(cgelsd,CGELSD) -#define LAPACK_zgelsd LAPACK_GLOBAL(zgelsd,ZGELSD) -#define LAPACK_sgglse LAPACK_GLOBAL(sgglse,SGGLSE) -#define LAPACK_dgglse LAPACK_GLOBAL(dgglse,DGGLSE) -#define LAPACK_cgglse LAPACK_GLOBAL(cgglse,CGGLSE) -#define LAPACK_zgglse LAPACK_GLOBAL(zgglse,ZGGLSE) -#define LAPACK_sggglm LAPACK_GLOBAL(sggglm,SGGGLM) -#define LAPACK_dggglm LAPACK_GLOBAL(dggglm,DGGGLM) -#define LAPACK_cggglm LAPACK_GLOBAL(cggglm,CGGGLM) -#define LAPACK_zggglm LAPACK_GLOBAL(zggglm,ZGGGLM) -#define LAPACK_ssyev LAPACK_GLOBAL(ssyev,SSYEV) -#define LAPACK_dsyev LAPACK_GLOBAL(dsyev,DSYEV) -#define LAPACK_cheev LAPACK_GLOBAL(cheev,CHEEV) -#define LAPACK_zheev LAPACK_GLOBAL(zheev,ZHEEV) -#define LAPACK_ssyevd LAPACK_GLOBAL(ssyevd,SSYEVD) -#define LAPACK_dsyevd LAPACK_GLOBAL(dsyevd,DSYEVD) -#define LAPACK_cheevd LAPACK_GLOBAL(cheevd,CHEEVD) -#define LAPACK_zheevd LAPACK_GLOBAL(zheevd,ZHEEVD) -#define LAPACK_ssyevx LAPACK_GLOBAL(ssyevx,SSYEVX) -#define LAPACK_dsyevx LAPACK_GLOBAL(dsyevx,DSYEVX) -#define LAPACK_cheevx LAPACK_GLOBAL(cheevx,CHEEVX) -#define LAPACK_zheevx LAPACK_GLOBAL(zheevx,ZHEEVX) -#define LAPACK_ssyevr LAPACK_GLOBAL(ssyevr,SSYEVR) -#define LAPACK_dsyevr LAPACK_GLOBAL(dsyevr,DSYEVR) -#define LAPACK_cheevr LAPACK_GLOBAL(cheevr,CHEEVR) -#define LAPACK_zheevr LAPACK_GLOBAL(zheevr,ZHEEVR) -#define LAPACK_sspev LAPACK_GLOBAL(sspev,SSPEV) -#define LAPACK_dspev LAPACK_GLOBAL(dspev,DSPEV) -#define LAPACK_chpev LAPACK_GLOBAL(chpev,CHPEV) -#define LAPACK_zhpev LAPACK_GLOBAL(zhpev,ZHPEV) -#define LAPACK_sspevd LAPACK_GLOBAL(sspevd,SSPEVD) -#define LAPACK_dspevd LAPACK_GLOBAL(dspevd,DSPEVD) -#define LAPACK_chpevd LAPACK_GLOBAL(chpevd,CHPEVD) -#define LAPACK_zhpevd LAPACK_GLOBAL(zhpevd,ZHPEVD) -#define LAPACK_sspevx LAPACK_GLOBAL(sspevx,SSPEVX) -#define LAPACK_dspevx LAPACK_GLOBAL(dspevx,DSPEVX) -#define LAPACK_chpevx LAPACK_GLOBAL(chpevx,CHPEVX) -#define LAPACK_zhpevx LAPACK_GLOBAL(zhpevx,ZHPEVX) -#define LAPACK_ssbev LAPACK_GLOBAL(ssbev,SSBEV) -#define LAPACK_dsbev LAPACK_GLOBAL(dsbev,DSBEV) -#define LAPACK_chbev LAPACK_GLOBAL(chbev,CHBEV) -#define LAPACK_zhbev LAPACK_GLOBAL(zhbev,ZHBEV) -#define LAPACK_ssbevd LAPACK_GLOBAL(ssbevd,SSBEVD) -#define LAPACK_dsbevd LAPACK_GLOBAL(dsbevd,DSBEVD) -#define LAPACK_chbevd LAPACK_GLOBAL(chbevd,CHBEVD) -#define LAPACK_zhbevd LAPACK_GLOBAL(zhbevd,ZHBEVD) -#define LAPACK_ssbevx LAPACK_GLOBAL(ssbevx,SSBEVX) -#define LAPACK_dsbevx LAPACK_GLOBAL(dsbevx,DSBEVX) -#define LAPACK_chbevx LAPACK_GLOBAL(chbevx,CHBEVX) -#define LAPACK_zhbevx LAPACK_GLOBAL(zhbevx,ZHBEVX) -#define LAPACK_sstev LAPACK_GLOBAL(sstev,SSTEV) -#define LAPACK_dstev LAPACK_GLOBAL(dstev,DSTEV) -#define LAPACK_sstevd LAPACK_GLOBAL(sstevd,SSTEVD) -#define LAPACK_dstevd LAPACK_GLOBAL(dstevd,DSTEVD) -#define LAPACK_sstevx LAPACK_GLOBAL(sstevx,SSTEVX) -#define LAPACK_dstevx LAPACK_GLOBAL(dstevx,DSTEVX) -#define LAPACK_sstevr LAPACK_GLOBAL(sstevr,SSTEVR) -#define LAPACK_dstevr LAPACK_GLOBAL(dstevr,DSTEVR) -#define LAPACK_sgees LAPACK_GLOBAL(sgees,SGEES) -#define LAPACK_dgees LAPACK_GLOBAL(dgees,DGEES) -#define LAPACK_cgees LAPACK_GLOBAL(cgees,CGEES) -#define LAPACK_zgees LAPACK_GLOBAL(zgees,ZGEES) -#define LAPACK_sgeesx LAPACK_GLOBAL(sgeesx,SGEESX) -#define LAPACK_dgeesx LAPACK_GLOBAL(dgeesx,DGEESX) -#define LAPACK_cgeesx LAPACK_GLOBAL(cgeesx,CGEESX) -#define LAPACK_zgeesx LAPACK_GLOBAL(zgeesx,ZGEESX) -#define LAPACK_sgeev LAPACK_GLOBAL(sgeev,SGEEV) -#define LAPACK_dgeev LAPACK_GLOBAL(dgeev,DGEEV) -#define LAPACK_cgeev LAPACK_GLOBAL(cgeev,CGEEV) -#define LAPACK_zgeev LAPACK_GLOBAL(zgeev,ZGEEV) -#define LAPACK_sgeevx LAPACK_GLOBAL(sgeevx,SGEEVX) -#define LAPACK_dgeevx LAPACK_GLOBAL(dgeevx,DGEEVX) -#define LAPACK_cgeevx LAPACK_GLOBAL(cgeevx,CGEEVX) -#define LAPACK_zgeevx LAPACK_GLOBAL(zgeevx,ZGEEVX) -#define LAPACK_sgesvd LAPACK_GLOBAL(sgesvd,SGESVD) -#define LAPACK_dgesvd LAPACK_GLOBAL(dgesvd,DGESVD) -#define LAPACK_cgesvd LAPACK_GLOBAL(cgesvd,CGESVD) -#define LAPACK_zgesvd LAPACK_GLOBAL(zgesvd,ZGESVD) -#define LAPACK_sgesdd LAPACK_GLOBAL(sgesdd,SGESDD) -#define LAPACK_dgesdd LAPACK_GLOBAL(dgesdd,DGESDD) -#define LAPACK_cgesdd LAPACK_GLOBAL(cgesdd,CGESDD) -#define LAPACK_zgesdd LAPACK_GLOBAL(zgesdd,ZGESDD) -#define LAPACK_dgejsv LAPACK_GLOBAL(dgejsv,DGEJSV) -#define LAPACK_sgejsv LAPACK_GLOBAL(sgejsv,SGEJSV) -#define LAPACK_dgesvj LAPACK_GLOBAL(dgesvj,DGESVJ) -#define LAPACK_sgesvj LAPACK_GLOBAL(sgesvj,SGESVJ) -#define LAPACK_sggsvd LAPACK_GLOBAL(sggsvd,SGGSVD) -#define LAPACK_dggsvd LAPACK_GLOBAL(dggsvd,DGGSVD) -#define LAPACK_cggsvd LAPACK_GLOBAL(cggsvd,CGGSVD) -#define LAPACK_zggsvd LAPACK_GLOBAL(zggsvd,ZGGSVD) -#define LAPACK_ssygv LAPACK_GLOBAL(ssygv,SSYGV) -#define LAPACK_dsygv LAPACK_GLOBAL(dsygv,DSYGV) -#define LAPACK_chegv LAPACK_GLOBAL(chegv,CHEGV) -#define LAPACK_zhegv LAPACK_GLOBAL(zhegv,ZHEGV) -#define LAPACK_ssygvd LAPACK_GLOBAL(ssygvd,SSYGVD) -#define LAPACK_dsygvd LAPACK_GLOBAL(dsygvd,DSYGVD) -#define LAPACK_chegvd LAPACK_GLOBAL(chegvd,CHEGVD) -#define LAPACK_zhegvd LAPACK_GLOBAL(zhegvd,ZHEGVD) -#define LAPACK_ssygvx LAPACK_GLOBAL(ssygvx,SSYGVX) -#define LAPACK_dsygvx LAPACK_GLOBAL(dsygvx,DSYGVX) -#define LAPACK_chegvx LAPACK_GLOBAL(chegvx,CHEGVX) -#define LAPACK_zhegvx LAPACK_GLOBAL(zhegvx,ZHEGVX) -#define LAPACK_sspgv LAPACK_GLOBAL(sspgv,SSPGV) -#define LAPACK_dspgv LAPACK_GLOBAL(dspgv,DSPGV) -#define LAPACK_chpgv LAPACK_GLOBAL(chpgv,CHPGV) -#define LAPACK_zhpgv LAPACK_GLOBAL(zhpgv,ZHPGV) -#define LAPACK_sspgvd LAPACK_GLOBAL(sspgvd,SSPGVD) -#define LAPACK_dspgvd LAPACK_GLOBAL(dspgvd,DSPGVD) -#define LAPACK_chpgvd LAPACK_GLOBAL(chpgvd,CHPGVD) -#define LAPACK_zhpgvd LAPACK_GLOBAL(zhpgvd,ZHPGVD) -#define LAPACK_sspgvx LAPACK_GLOBAL(sspgvx,SSPGVX) -#define LAPACK_dspgvx LAPACK_GLOBAL(dspgvx,DSPGVX) -#define LAPACK_chpgvx LAPACK_GLOBAL(chpgvx,CHPGVX) -#define LAPACK_zhpgvx LAPACK_GLOBAL(zhpgvx,ZHPGVX) -#define LAPACK_ssbgv LAPACK_GLOBAL(ssbgv,SSBGV) -#define LAPACK_dsbgv LAPACK_GLOBAL(dsbgv,DSBGV) -#define LAPACK_chbgv LAPACK_GLOBAL(chbgv,CHBGV) -#define LAPACK_zhbgv LAPACK_GLOBAL(zhbgv,ZHBGV) -#define LAPACK_ssbgvd LAPACK_GLOBAL(ssbgvd,SSBGVD) -#define LAPACK_dsbgvd LAPACK_GLOBAL(dsbgvd,DSBGVD) -#define LAPACK_chbgvd LAPACK_GLOBAL(chbgvd,CHBGVD) -#define LAPACK_zhbgvd LAPACK_GLOBAL(zhbgvd,ZHBGVD) -#define LAPACK_ssbgvx LAPACK_GLOBAL(ssbgvx,SSBGVX) -#define LAPACK_dsbgvx LAPACK_GLOBAL(dsbgvx,DSBGVX) -#define LAPACK_chbgvx LAPACK_GLOBAL(chbgvx,CHBGVX) -#define LAPACK_zhbgvx LAPACK_GLOBAL(zhbgvx,ZHBGVX) -#define LAPACK_sgges LAPACK_GLOBAL(sgges,SGGES) -#define LAPACK_dgges LAPACK_GLOBAL(dgges,DGGES) -#define LAPACK_cgges LAPACK_GLOBAL(cgges,CGGES) -#define LAPACK_zgges LAPACK_GLOBAL(zgges,ZGGES) -#define LAPACK_sggesx LAPACK_GLOBAL(sggesx,SGGESX) -#define LAPACK_dggesx LAPACK_GLOBAL(dggesx,DGGESX) -#define LAPACK_cggesx LAPACK_GLOBAL(cggesx,CGGESX) -#define LAPACK_zggesx LAPACK_GLOBAL(zggesx,ZGGESX) -#define LAPACK_sggev LAPACK_GLOBAL(sggev,SGGEV) -#define LAPACK_dggev LAPACK_GLOBAL(dggev,DGGEV) -#define LAPACK_cggev LAPACK_GLOBAL(cggev,CGGEV) -#define LAPACK_zggev LAPACK_GLOBAL(zggev,ZGGEV) -#define LAPACK_sggevx LAPACK_GLOBAL(sggevx,SGGEVX) -#define LAPACK_dggevx LAPACK_GLOBAL(dggevx,DGGEVX) -#define LAPACK_cggevx LAPACK_GLOBAL(cggevx,CGGEVX) -#define LAPACK_zggevx LAPACK_GLOBAL(zggevx,ZGGEVX) -#define LAPACK_dsfrk LAPACK_GLOBAL(dsfrk,DSFRK) -#define LAPACK_ssfrk LAPACK_GLOBAL(ssfrk,SSFRK) -#define LAPACK_zhfrk LAPACK_GLOBAL(zhfrk,ZHFRK) -#define LAPACK_chfrk LAPACK_GLOBAL(chfrk,CHFRK) -#define LAPACK_dtfsm LAPACK_GLOBAL(dtfsm,DTFSM) -#define LAPACK_stfsm LAPACK_GLOBAL(stfsm,STFSM) -#define LAPACK_ztfsm LAPACK_GLOBAL(ztfsm,ZTFSM) -#define LAPACK_ctfsm LAPACK_GLOBAL(ctfsm,CTFSM) -#define LAPACK_dtfttp LAPACK_GLOBAL(dtfttp,DTFTTP) -#define LAPACK_stfttp LAPACK_GLOBAL(stfttp,STFTTP) -#define LAPACK_ztfttp LAPACK_GLOBAL(ztfttp,ZTFTTP) -#define LAPACK_ctfttp LAPACK_GLOBAL(ctfttp,CTFTTP) -#define LAPACK_dtfttr LAPACK_GLOBAL(dtfttr,DTFTTR) -#define LAPACK_stfttr LAPACK_GLOBAL(stfttr,STFTTR) -#define LAPACK_ztfttr LAPACK_GLOBAL(ztfttr,ZTFTTR) -#define LAPACK_ctfttr LAPACK_GLOBAL(ctfttr,CTFTTR) -#define LAPACK_dtpttf LAPACK_GLOBAL(dtpttf,DTPTTF) -#define LAPACK_stpttf LAPACK_GLOBAL(stpttf,STPTTF) -#define LAPACK_ztpttf LAPACK_GLOBAL(ztpttf,ZTPTTF) -#define LAPACK_ctpttf LAPACK_GLOBAL(ctpttf,CTPTTF) -#define LAPACK_dtpttr LAPACK_GLOBAL(dtpttr,DTPTTR) -#define LAPACK_stpttr LAPACK_GLOBAL(stpttr,STPTTR) -#define LAPACK_ztpttr LAPACK_GLOBAL(ztpttr,ZTPTTR) -#define LAPACK_ctpttr LAPACK_GLOBAL(ctpttr,CTPTTR) -#define LAPACK_dtrttf LAPACK_GLOBAL(dtrttf,DTRTTF) -#define LAPACK_strttf LAPACK_GLOBAL(strttf,STRTTF) -#define LAPACK_ztrttf LAPACK_GLOBAL(ztrttf,ZTRTTF) -#define LAPACK_ctrttf LAPACK_GLOBAL(ctrttf,CTRTTF) -#define LAPACK_dtrttp LAPACK_GLOBAL(dtrttp,DTRTTP) -#define LAPACK_strttp LAPACK_GLOBAL(strttp,STRTTP) -#define LAPACK_ztrttp LAPACK_GLOBAL(ztrttp,ZTRTTP) -#define LAPACK_ctrttp LAPACK_GLOBAL(ctrttp,CTRTTP) -#define LAPACK_sgeqrfp LAPACK_GLOBAL(sgeqrfp,SGEQRFP) -#define LAPACK_dgeqrfp LAPACK_GLOBAL(dgeqrfp,DGEQRFP) -#define LAPACK_cgeqrfp LAPACK_GLOBAL(cgeqrfp,CGEQRFP) -#define LAPACK_zgeqrfp LAPACK_GLOBAL(zgeqrfp,ZGEQRFP) -#define LAPACK_clacgv LAPACK_GLOBAL(clacgv,CLACGV) -#define LAPACK_zlacgv LAPACK_GLOBAL(zlacgv,ZLACGV) -#define LAPACK_slarnv LAPACK_GLOBAL(slarnv,SLARNV) -#define LAPACK_dlarnv LAPACK_GLOBAL(dlarnv,DLARNV) -#define LAPACK_clarnv LAPACK_GLOBAL(clarnv,CLARNV) -#define LAPACK_zlarnv LAPACK_GLOBAL(zlarnv,ZLARNV) -#define LAPACK_sgeqr2 LAPACK_GLOBAL(sgeqr2,SGEQR2) -#define LAPACK_dgeqr2 LAPACK_GLOBAL(dgeqr2,DGEQR2) -#define LAPACK_cgeqr2 LAPACK_GLOBAL(cgeqr2,CGEQR2) -#define LAPACK_zgeqr2 LAPACK_GLOBAL(zgeqr2,ZGEQR2) -#define LAPACK_slacpy LAPACK_GLOBAL(slacpy,SLACPY) -#define LAPACK_dlacpy LAPACK_GLOBAL(dlacpy,DLACPY) -#define LAPACK_clacpy LAPACK_GLOBAL(clacpy,CLACPY) -#define LAPACK_zlacpy LAPACK_GLOBAL(zlacpy,ZLACPY) -#define LAPACK_sgetf2 LAPACK_GLOBAL(sgetf2,SGETF2) -#define LAPACK_dgetf2 LAPACK_GLOBAL(dgetf2,DGETF2) -#define LAPACK_cgetf2 LAPACK_GLOBAL(cgetf2,CGETF2) -#define LAPACK_zgetf2 LAPACK_GLOBAL(zgetf2,ZGETF2) -#define LAPACK_slaswp LAPACK_GLOBAL(slaswp,SLASWP) -#define LAPACK_dlaswp LAPACK_GLOBAL(dlaswp,DLASWP) -#define LAPACK_claswp LAPACK_GLOBAL(claswp,CLASWP) -#define LAPACK_zlaswp LAPACK_GLOBAL(zlaswp,ZLASWP) -#define LAPACK_slange LAPACK_GLOBAL(slange,SLANGE) -#define LAPACK_dlange LAPACK_GLOBAL(dlange,DLANGE) -#define LAPACK_clange LAPACK_GLOBAL(clange,CLANGE) -#define LAPACK_zlange LAPACK_GLOBAL(zlange,ZLANGE) -#define LAPACK_clanhe LAPACK_GLOBAL(clanhe,CLANHE) -#define LAPACK_zlanhe LAPACK_GLOBAL(zlanhe,ZLANHE) -#define LAPACK_slansy LAPACK_GLOBAL(slansy,SLANSY) -#define LAPACK_dlansy LAPACK_GLOBAL(dlansy,DLANSY) -#define LAPACK_clansy LAPACK_GLOBAL(clansy,CLANSY) -#define LAPACK_zlansy LAPACK_GLOBAL(zlansy,ZLANSY) -#define LAPACK_slantr LAPACK_GLOBAL(slantr,SLANTR) -#define LAPACK_dlantr LAPACK_GLOBAL(dlantr,DLANTR) -#define LAPACK_clantr LAPACK_GLOBAL(clantr,CLANTR) -#define LAPACK_zlantr LAPACK_GLOBAL(zlantr,ZLANTR) -#define LAPACK_slamch LAPACK_GLOBAL(slamch,SLAMCH) -#define LAPACK_dlamch LAPACK_GLOBAL(dlamch,DLAMCH) -#define LAPACK_sgelq2 LAPACK_GLOBAL(sgelq2,SGELQ2) -#define LAPACK_dgelq2 LAPACK_GLOBAL(dgelq2,DGELQ2) -#define LAPACK_cgelq2 LAPACK_GLOBAL(cgelq2,CGELQ2) -#define LAPACK_zgelq2 LAPACK_GLOBAL(zgelq2,ZGELQ2) -#define LAPACK_slarfb LAPACK_GLOBAL(slarfb,SLARFB) -#define LAPACK_dlarfb LAPACK_GLOBAL(dlarfb,DLARFB) -#define LAPACK_clarfb LAPACK_GLOBAL(clarfb,CLARFB) -#define LAPACK_zlarfb LAPACK_GLOBAL(zlarfb,ZLARFB) -#define LAPACK_slarfg LAPACK_GLOBAL(slarfg,SLARFG) -#define LAPACK_dlarfg LAPACK_GLOBAL(dlarfg,DLARFG) -#define LAPACK_clarfg LAPACK_GLOBAL(clarfg,CLARFG) -#define LAPACK_zlarfg LAPACK_GLOBAL(zlarfg,ZLARFG) -#define LAPACK_slarft LAPACK_GLOBAL(slarft,SLARFT) -#define LAPACK_dlarft LAPACK_GLOBAL(dlarft,DLARFT) -#define LAPACK_clarft LAPACK_GLOBAL(clarft,CLARFT) -#define LAPACK_zlarft LAPACK_GLOBAL(zlarft,ZLARFT) -#define LAPACK_slarfx LAPACK_GLOBAL(slarfx,SLARFX) -#define LAPACK_dlarfx LAPACK_GLOBAL(dlarfx,DLARFX) -#define LAPACK_clarfx LAPACK_GLOBAL(clarfx,CLARFX) -#define LAPACK_zlarfx LAPACK_GLOBAL(zlarfx,ZLARFX) -#define LAPACK_slatms LAPACK_GLOBAL(slatms,SLATMS) -#define LAPACK_dlatms LAPACK_GLOBAL(dlatms,DLATMS) -#define LAPACK_clatms LAPACK_GLOBAL(clatms,CLATMS) -#define LAPACK_zlatms LAPACK_GLOBAL(zlatms,ZLATMS) -#define LAPACK_slag2d LAPACK_GLOBAL(slag2d,SLAG2D) -#define LAPACK_dlag2s LAPACK_GLOBAL(dlag2s,DLAG2S) -#define LAPACK_clag2z LAPACK_GLOBAL(clag2z,CLAG2Z) -#define LAPACK_zlag2c LAPACK_GLOBAL(zlag2c,ZLAG2C) -#define LAPACK_slauum LAPACK_GLOBAL(slauum,SLAUUM) -#define LAPACK_dlauum LAPACK_GLOBAL(dlauum,DLAUUM) -#define LAPACK_clauum LAPACK_GLOBAL(clauum,CLAUUM) -#define LAPACK_zlauum LAPACK_GLOBAL(zlauum,ZLAUUM) -#define LAPACK_slagge LAPACK_GLOBAL(slagge,SLAGGE) -#define LAPACK_dlagge LAPACK_GLOBAL(dlagge,DLAGGE) -#define LAPACK_clagge LAPACK_GLOBAL(clagge,CLAGGE) -#define LAPACK_zlagge LAPACK_GLOBAL(zlagge,ZLAGGE) -#define LAPACK_slaset LAPACK_GLOBAL(slaset,SLASET) -#define LAPACK_dlaset LAPACK_GLOBAL(dlaset,DLASET) -#define LAPACK_claset LAPACK_GLOBAL(claset,CLASET) -#define LAPACK_zlaset LAPACK_GLOBAL(zlaset,ZLASET) -#define LAPACK_slasrt LAPACK_GLOBAL(slasrt,SLASRT) -#define LAPACK_dlasrt LAPACK_GLOBAL(dlasrt,DLASRT) -#define LAPACK_slagsy LAPACK_GLOBAL(slagsy,SLAGSY) -#define LAPACK_dlagsy LAPACK_GLOBAL(dlagsy,DLAGSY) -#define LAPACK_clagsy LAPACK_GLOBAL(clagsy,CLAGSY) -#define LAPACK_zlagsy LAPACK_GLOBAL(zlagsy,ZLAGSY) -#define LAPACK_claghe LAPACK_GLOBAL(claghe,CLAGHE) -#define LAPACK_zlaghe LAPACK_GLOBAL(zlaghe,ZLAGHE) -#define LAPACK_slapmr LAPACK_GLOBAL(slapmr,SLAPMR) -#define LAPACK_dlapmr LAPACK_GLOBAL(dlapmr,DLAPMR) -#define LAPACK_clapmr LAPACK_GLOBAL(clapmr,CLAPMR) -#define LAPACK_zlapmr LAPACK_GLOBAL(zlapmr,ZLAPMR) -#define LAPACK_slapy2 LAPACK_GLOBAL(slapy2,SLAPY2) -#define LAPACK_dlapy2 LAPACK_GLOBAL(dlapy2,DLAPY2) -#define LAPACK_slapy3 LAPACK_GLOBAL(slapy3,SLAPY3) -#define LAPACK_dlapy3 LAPACK_GLOBAL(dlapy3,DLAPY3) -#define LAPACK_slartgp LAPACK_GLOBAL(slartgp,SLARTGP) -#define LAPACK_dlartgp LAPACK_GLOBAL(dlartgp,DLARTGP) -#define LAPACK_slartgs LAPACK_GLOBAL(slartgs,SLARTGS) -#define LAPACK_dlartgs LAPACK_GLOBAL(dlartgs,DLARTGS) -// LAPACK 3.3.0 -#define LAPACK_cbbcsd LAPACK_GLOBAL(cbbcsd,CBBCSD) -#define LAPACK_cheswapr LAPACK_GLOBAL(cheswapr,CHESWAPR) -#define LAPACK_chetri2 LAPACK_GLOBAL(chetri2,CHETRI2) -#define LAPACK_chetri2x LAPACK_GLOBAL(chetri2x,CHETRI2X) -#define LAPACK_chetrs2 LAPACK_GLOBAL(chetrs2,CHETRS2) -#define LAPACK_csyconv LAPACK_GLOBAL(csyconv,CSYCONV) -#define LAPACK_csyswapr LAPACK_GLOBAL(csyswapr,CSYSWAPR) -#define LAPACK_csytri2 LAPACK_GLOBAL(csytri2,CSYTRI2) -#define LAPACK_csytri2x LAPACK_GLOBAL(csytri2x,CSYTRI2X) -#define LAPACK_csytrs2 LAPACK_GLOBAL(csytrs2,CSYTRS2) -#define LAPACK_cunbdb LAPACK_GLOBAL(cunbdb,CUNBDB) -#define LAPACK_cuncsd LAPACK_GLOBAL(cuncsd,CUNCSD) -#define LAPACK_dbbcsd LAPACK_GLOBAL(dbbcsd,DBBCSD) -#define LAPACK_dorbdb LAPACK_GLOBAL(dorbdb,DORBDB) -#define LAPACK_dorcsd LAPACK_GLOBAL(dorcsd,DORCSD) -#define LAPACK_dsyconv LAPACK_GLOBAL(dsyconv,DSYCONV) -#define LAPACK_dsyswapr LAPACK_GLOBAL(dsyswapr,DSYSWAPR) -#define LAPACK_dsytri2 LAPACK_GLOBAL(dsytri2,DSYTRI2) -#define LAPACK_dsytri2x LAPACK_GLOBAL(dsytri2x,DSYTRI2X) -#define LAPACK_dsytrs2 LAPACK_GLOBAL(dsytrs2,DSYTRS2) -#define LAPACK_sbbcsd LAPACK_GLOBAL(sbbcsd,SBBCSD) -#define LAPACK_sorbdb LAPACK_GLOBAL(sorbdb,SORBDB) -#define LAPACK_sorcsd LAPACK_GLOBAL(sorcsd,SORCSD) -#define LAPACK_ssyconv LAPACK_GLOBAL(ssyconv,SSYCONV) -#define LAPACK_ssyswapr LAPACK_GLOBAL(ssyswapr,SSYSWAPR) -#define LAPACK_ssytri2 LAPACK_GLOBAL(ssytri2,SSYTRI2) -#define LAPACK_ssytri2x LAPACK_GLOBAL(ssytri2x,SSYTRI2X) -#define LAPACK_ssytrs2 LAPACK_GLOBAL(ssytrs2,SSYTRS2) -#define LAPACK_zbbcsd LAPACK_GLOBAL(zbbcsd,ZBBCSD) -#define LAPACK_zheswapr LAPACK_GLOBAL(zheswapr,ZHESWAPR) -#define LAPACK_zhetri2 LAPACK_GLOBAL(zhetri2,ZHETRI2) -#define LAPACK_zhetri2x LAPACK_GLOBAL(zhetri2x,ZHETRI2X) -#define LAPACK_zhetrs2 LAPACK_GLOBAL(zhetrs2,ZHETRS2) -#define LAPACK_zsyconv LAPACK_GLOBAL(zsyconv,ZSYCONV) -#define LAPACK_zsyswapr LAPACK_GLOBAL(zsyswapr,ZSYSWAPR) -#define LAPACK_zsytri2 LAPACK_GLOBAL(zsytri2,ZSYTRI2) -#define LAPACK_zsytri2x LAPACK_GLOBAL(zsytri2x,ZSYTRI2X) -#define LAPACK_zsytrs2 LAPACK_GLOBAL(zsytrs2,ZSYTRS2) -#define LAPACK_zunbdb LAPACK_GLOBAL(zunbdb,ZUNBDB) -#define LAPACK_zuncsd LAPACK_GLOBAL(zuncsd,ZUNCSD) -// LAPACK 3.4.0 -#define LAPACK_sgemqrt LAPACK_GLOBAL(sgemqrt,SGEMQRT) -#define LAPACK_dgemqrt LAPACK_GLOBAL(dgemqrt,DGEMQRT) -#define LAPACK_cgemqrt LAPACK_GLOBAL(cgemqrt,CGEMQRT) -#define LAPACK_zgemqrt LAPACK_GLOBAL(zgemqrt,ZGEMQRT) -#define LAPACK_sgeqrt LAPACK_GLOBAL(sgeqrt,SGEQRT) -#define LAPACK_dgeqrt LAPACK_GLOBAL(dgeqrt,DGEQRT) -#define LAPACK_cgeqrt LAPACK_GLOBAL(cgeqrt,CGEQRT) -#define LAPACK_zgeqrt LAPACK_GLOBAL(zgeqrt,ZGEQRT) -#define LAPACK_sgeqrt2 LAPACK_GLOBAL(sgeqrt2,SGEQRT2) -#define LAPACK_dgeqrt2 LAPACK_GLOBAL(dgeqrt2,DGEQRT2) -#define LAPACK_cgeqrt2 LAPACK_GLOBAL(cgeqrt2,CGEQRT2) -#define LAPACK_zgeqrt2 LAPACK_GLOBAL(zgeqrt2,ZGEQRT2) -#define LAPACK_sgeqrt3 LAPACK_GLOBAL(sgeqrt3,SGEQRT3) -#define LAPACK_dgeqrt3 LAPACK_GLOBAL(dgeqrt3,DGEQRT3) -#define LAPACK_cgeqrt3 LAPACK_GLOBAL(cgeqrt3,CGEQRT3) -#define LAPACK_zgeqrt3 LAPACK_GLOBAL(zgeqrt3,ZGEQRT3) -#define LAPACK_stpmqrt LAPACK_GLOBAL(stpmqrt,STPMQRT) -#define LAPACK_dtpmqrt LAPACK_GLOBAL(dtpmqrt,DTPMQRT) -#define LAPACK_ctpmqrt LAPACK_GLOBAL(ctpmqrt,CTPMQRT) -#define LAPACK_ztpmqrt LAPACK_GLOBAL(ztpmqrt,ZTPMQRT) -#define LAPACK_dtpqrt LAPACK_GLOBAL(dtpqrt,DTPQRT) -#define LAPACK_ctpqrt LAPACK_GLOBAL(ctpqrt,CTPQRT) -#define LAPACK_ztpqrt LAPACK_GLOBAL(ztpqrt,ZTPQRT) -#define LAPACK_stpqrt2 LAPACK_GLOBAL(stpqrt2,STPQRT2) -#define LAPACK_dtpqrt2 LAPACK_GLOBAL(dtpqrt2,DTPQRT2) -#define LAPACK_ctpqrt2 LAPACK_GLOBAL(ctpqrt2,CTPQRT2) -#define LAPACK_ztpqrt2 LAPACK_GLOBAL(ztpqrt2,ZTPQRT2) -#define LAPACK_stprfb LAPACK_GLOBAL(stprfb,STPRFB) -#define LAPACK_dtprfb LAPACK_GLOBAL(dtprfb,DTPRFB) -#define LAPACK_ctprfb LAPACK_GLOBAL(ctprfb,CTPRFB) -#define LAPACK_ztprfb LAPACK_GLOBAL(ztprfb,ZTPRFB) -// LAPACK 3.X.X -#define LAPACK_csyr LAPACK_GLOBAL(csyr,CSYR) -#define LAPACK_zsyr LAPACK_GLOBAL(zsyr,ZSYR) - - -void LAPACK_sgetrf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_dgetrf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_cgetrf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* ipiv, lapack_int *info ); -void LAPACK_zgetrf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, lapack_int *info ); -void LAPACK_sgbtrf( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, float* ab, lapack_int* ldab, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_dgbtrf( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, double* ab, lapack_int* ldab, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_cgbtrf( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_complex_float* ab, lapack_int* ldab, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_zgbtrf( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_complex_double* ab, lapack_int* ldab, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_sgttrf( lapack_int* n, float* dl, float* d, float* du, float* du2, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_dgttrf( lapack_int* n, double* dl, double* d, double* du, - double* du2, lapack_int* ipiv, lapack_int *info ); -void LAPACK_cgttrf( lapack_int* n, lapack_complex_float* dl, - lapack_complex_float* d, lapack_complex_float* du, - lapack_complex_float* du2, lapack_int* ipiv, - lapack_int *info ); -void LAPACK_zgttrf( lapack_int* n, lapack_complex_double* dl, - lapack_complex_double* d, lapack_complex_double* du, - lapack_complex_double* du2, lapack_int* ipiv, - lapack_int *info ); -void LAPACK_spotrf( char* uplo, lapack_int* n, float* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dpotrf( char* uplo, lapack_int* n, double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_cpotrf( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_zpotrf( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_dpstrf( char* uplo, lapack_int* n, double* a, lapack_int* lda, - lapack_int* piv, lapack_int* rank, double* tol, - double* work, lapack_int *info ); -void LAPACK_spstrf( char* uplo, lapack_int* n, float* a, lapack_int* lda, - lapack_int* piv, lapack_int* rank, float* tol, float* work, - lapack_int *info ); -void LAPACK_zpstrf( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* piv, lapack_int* rank, - double* tol, double* work, lapack_int *info ); -void LAPACK_cpstrf( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* piv, lapack_int* rank, - float* tol, float* work, lapack_int *info ); -void LAPACK_dpftrf( char* transr, char* uplo, lapack_int* n, double* a, - lapack_int *info ); -void LAPACK_spftrf( char* transr, char* uplo, lapack_int* n, float* a, - lapack_int *info ); -void LAPACK_zpftrf( char* transr, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int *info ); -void LAPACK_cpftrf( char* transr, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int *info ); -void LAPACK_spptrf( char* uplo, lapack_int* n, float* ap, lapack_int *info ); -void LAPACK_dpptrf( char* uplo, lapack_int* n, double* ap, lapack_int *info ); -void LAPACK_cpptrf( char* uplo, lapack_int* n, lapack_complex_float* ap, - lapack_int *info ); -void LAPACK_zpptrf( char* uplo, lapack_int* n, lapack_complex_double* ap, - lapack_int *info ); -void LAPACK_spbtrf( char* uplo, lapack_int* n, lapack_int* kd, float* ab, - lapack_int* ldab, lapack_int *info ); -void LAPACK_dpbtrf( char* uplo, lapack_int* n, lapack_int* kd, double* ab, - lapack_int* ldab, lapack_int *info ); -void LAPACK_cpbtrf( char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_float* ab, lapack_int* ldab, - lapack_int *info ); -void LAPACK_zpbtrf( char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_double* ab, lapack_int* ldab, - lapack_int *info ); -void LAPACK_spttrf( lapack_int* n, float* d, float* e, lapack_int *info ); -void LAPACK_dpttrf( lapack_int* n, double* d, double* e, lapack_int *info ); -void LAPACK_cpttrf( lapack_int* n, float* d, lapack_complex_float* e, - lapack_int *info ); -void LAPACK_zpttrf( lapack_int* n, double* d, lapack_complex_double* e, - lapack_int *info ); -void LAPACK_ssytrf( char* uplo, lapack_int* n, float* a, lapack_int* lda, - lapack_int* ipiv, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dsytrf( char* uplo, lapack_int* n, double* a, lapack_int* lda, - lapack_int* ipiv, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_csytrf( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* ipiv, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zsytrf( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_chetrf( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* ipiv, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zhetrf( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ssptrf( char* uplo, lapack_int* n, float* ap, lapack_int* ipiv, - lapack_int *info ); -void LAPACK_dsptrf( char* uplo, lapack_int* n, double* ap, lapack_int* ipiv, - lapack_int *info ); -void LAPACK_csptrf( char* uplo, lapack_int* n, lapack_complex_float* ap, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_zsptrf( char* uplo, lapack_int* n, lapack_complex_double* ap, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_chptrf( char* uplo, lapack_int* n, lapack_complex_float* ap, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_zhptrf( char* uplo, lapack_int* n, lapack_complex_double* ap, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_sgetrs( char* trans, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, const lapack_int* ipiv, - float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_dgetrs( char* trans, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const lapack_int* ipiv, - double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_cgetrs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zgetrs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_sgbtrs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const float* ab, lapack_int* ldab, - const lapack_int* ipiv, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dgbtrs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const double* ab, lapack_int* ldab, - const lapack_int* ipiv, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cgbtrs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const lapack_complex_float* ab, - lapack_int* ldab, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zgbtrs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const lapack_complex_double* ab, - lapack_int* ldab, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_sgttrs( char* trans, lapack_int* n, lapack_int* nrhs, - const float* dl, const float* d, const float* du, - const float* du2, const lapack_int* ipiv, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dgttrs( char* trans, lapack_int* n, lapack_int* nrhs, - const double* dl, const double* d, const double* du, - const double* du2, const lapack_int* ipiv, double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_cgttrs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zgttrs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_spotrs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* a, - lapack_int* lda, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dpotrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_cpotrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zpotrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dpftrs( char* transr, char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_spftrs( char* transr, char* uplo, lapack_int* n, lapack_int* nrhs, - const float* a, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zpftrs( char* transr, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_cpftrs( char* transr, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_spptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const float* ap, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dpptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* ap, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cpptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zpptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_spbtrs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const float* ab, lapack_int* ldab, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dpbtrs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const double* ab, lapack_int* ldab, double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_cpbtrs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zpbtrs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_spttrs( lapack_int* n, lapack_int* nrhs, const float* d, - const float* e, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dpttrs( lapack_int* n, lapack_int* nrhs, const double* d, - const double* e, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cpttrs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* d, - const lapack_complex_float* e, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zpttrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* d, const lapack_complex_double* e, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_ssytrs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* a, - lapack_int* lda, const lapack_int* ipiv, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dsytrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const lapack_int* ipiv, - double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_csytrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zsytrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_chetrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zhetrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_ssptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const float* ap, const lapack_int* ipiv, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dsptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* ap, const lapack_int* ipiv, double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_csptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zsptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_chptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zhptrs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_strtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const float* a, lapack_int* lda, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dtrtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const double* a, lapack_int* lda, - double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_ctrtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_ztrtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_stptrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const float* ap, float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dtptrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const double* ap, double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_ctptrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* ap, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_ztptrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_double* ap, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_stbtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, const float* ab, - lapack_int* ldab, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dtbtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, const double* ab, - lapack_int* ldab, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_ctbtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, - const lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_ztbtrs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, - const lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_sgecon( char* norm, lapack_int* n, const float* a, lapack_int* lda, - float* anorm, float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgecon( char* norm, lapack_int* n, const double* a, lapack_int* lda, - double* anorm, double* rcond, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cgecon( char* norm, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* anorm, float* rcond, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zgecon( char* norm, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* anorm, double* rcond, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sgbcon( char* norm, lapack_int* n, lapack_int* kl, lapack_int* ku, - const float* ab, lapack_int* ldab, const lapack_int* ipiv, - float* anorm, float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgbcon( char* norm, lapack_int* n, lapack_int* kl, lapack_int* ku, - const double* ab, lapack_int* ldab, const lapack_int* ipiv, - double* anorm, double* rcond, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cgbcon( char* norm, lapack_int* n, lapack_int* kl, lapack_int* ku, - const lapack_complex_float* ab, lapack_int* ldab, - const lapack_int* ipiv, float* anorm, float* rcond, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zgbcon( char* norm, lapack_int* n, lapack_int* kl, lapack_int* ku, - const lapack_complex_double* ab, lapack_int* ldab, - const lapack_int* ipiv, double* anorm, double* rcond, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sgtcon( char* norm, lapack_int* n, const float* dl, const float* d, - const float* du, const float* du2, const lapack_int* ipiv, - float* anorm, float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgtcon( char* norm, lapack_int* n, const double* dl, - const double* d, const double* du, const double* du2, - const lapack_int* ipiv, double* anorm, double* rcond, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cgtcon( char* norm, lapack_int* n, const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* du2, const lapack_int* ipiv, - float* anorm, float* rcond, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zgtcon( char* norm, lapack_int* n, const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* du2, const lapack_int* ipiv, - double* anorm, double* rcond, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_spocon( char* uplo, lapack_int* n, const float* a, lapack_int* lda, - float* anorm, float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dpocon( char* uplo, lapack_int* n, const double* a, lapack_int* lda, - double* anorm, double* rcond, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cpocon( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* anorm, float* rcond, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zpocon( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* anorm, double* rcond, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sppcon( char* uplo, lapack_int* n, const float* ap, float* anorm, - float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dppcon( char* uplo, lapack_int* n, const double* ap, double* anorm, - double* rcond, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cppcon( char* uplo, lapack_int* n, const lapack_complex_float* ap, - float* anorm, float* rcond, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zppcon( char* uplo, lapack_int* n, const lapack_complex_double* ap, - double* anorm, double* rcond, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_spbcon( char* uplo, lapack_int* n, lapack_int* kd, const float* ab, - lapack_int* ldab, float* anorm, float* rcond, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dpbcon( char* uplo, lapack_int* n, lapack_int* kd, const double* ab, - lapack_int* ldab, double* anorm, double* rcond, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cpbcon( char* uplo, lapack_int* n, lapack_int* kd, - const lapack_complex_float* ab, lapack_int* ldab, - float* anorm, float* rcond, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zpbcon( char* uplo, lapack_int* n, lapack_int* kd, - const lapack_complex_double* ab, lapack_int* ldab, - double* anorm, double* rcond, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_sptcon( lapack_int* n, const float* d, const float* e, float* anorm, - float* rcond, float* work, lapack_int *info ); -void LAPACK_dptcon( lapack_int* n, const double* d, const double* e, - double* anorm, double* rcond, double* work, - lapack_int *info ); -void LAPACK_cptcon( lapack_int* n, const float* d, - const lapack_complex_float* e, float* anorm, float* rcond, - float* work, lapack_int *info ); -void LAPACK_zptcon( lapack_int* n, const double* d, - const lapack_complex_double* e, double* anorm, - double* rcond, double* work, lapack_int *info ); -void LAPACK_ssycon( char* uplo, lapack_int* n, const float* a, lapack_int* lda, - const lapack_int* ipiv, float* anorm, float* rcond, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dsycon( char* uplo, lapack_int* n, const double* a, lapack_int* lda, - const lapack_int* ipiv, double* anorm, double* rcond, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_csycon( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, float* anorm, - float* rcond, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zsycon( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, const lapack_int* ipiv, double* anorm, - double* rcond, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_checon( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, float* anorm, - float* rcond, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zhecon( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, const lapack_int* ipiv, double* anorm, - double* rcond, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_sspcon( char* uplo, lapack_int* n, const float* ap, - const lapack_int* ipiv, float* anorm, float* rcond, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dspcon( char* uplo, lapack_int* n, const double* ap, - const lapack_int* ipiv, double* anorm, double* rcond, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cspcon( char* uplo, lapack_int* n, const lapack_complex_float* ap, - const lapack_int* ipiv, float* anorm, float* rcond, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zspcon( char* uplo, lapack_int* n, const lapack_complex_double* ap, - const lapack_int* ipiv, double* anorm, double* rcond, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_chpcon( char* uplo, lapack_int* n, const lapack_complex_float* ap, - const lapack_int* ipiv, float* anorm, float* rcond, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zhpcon( char* uplo, lapack_int* n, const lapack_complex_double* ap, - const lapack_int* ipiv, double* anorm, double* rcond, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_strcon( char* norm, char* uplo, char* diag, lapack_int* n, - const float* a, lapack_int* lda, float* rcond, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dtrcon( char* norm, char* uplo, char* diag, lapack_int* n, - const double* a, lapack_int* lda, double* rcond, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_ctrcon( char* norm, char* uplo, char* diag, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, - float* rcond, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztrcon( char* norm, char* uplo, char* diag, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, - double* rcond, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_stpcon( char* norm, char* uplo, char* diag, lapack_int* n, - const float* ap, float* rcond, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dtpcon( char* norm, char* uplo, char* diag, lapack_int* n, - const double* ap, double* rcond, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ctpcon( char* norm, char* uplo, char* diag, lapack_int* n, - const lapack_complex_float* ap, float* rcond, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztpcon( char* norm, char* uplo, char* diag, lapack_int* n, - const lapack_complex_double* ap, double* rcond, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_stbcon( char* norm, char* uplo, char* diag, lapack_int* n, - lapack_int* kd, const float* ab, lapack_int* ldab, - float* rcond, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dtbcon( char* norm, char* uplo, char* diag, lapack_int* n, - lapack_int* kd, const double* ab, lapack_int* ldab, - double* rcond, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_ctbcon( char* norm, char* uplo, char* diag, lapack_int* n, - lapack_int* kd, const lapack_complex_float* ab, - lapack_int* ldab, float* rcond, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_ztbcon( char* norm, char* uplo, char* diag, lapack_int* n, - lapack_int* kd, const lapack_complex_double* ab, - lapack_int* ldab, double* rcond, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sgerfs( char* trans, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, const float* af, - lapack_int* ldaf, const lapack_int* ipiv, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* ferr, - float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgerfs( char* trans, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const lapack_int* ipiv, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cgerfs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zgerfs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_dgerfsx( char* trans, char* equed, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const lapack_int* ipiv, const double* r, - const double* c, const double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* rcond, double* berr, - lapack_int* n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int* nparams, double* params, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_sgerfsx( char* trans, char* equed, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, const float* af, - lapack_int* ldaf, const lapack_int* ipiv, const float* r, - const float* c, const float* b, lapack_int* ldb, float* x, - lapack_int* ldx, float* rcond, float* berr, - lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_zgerfsx( char* trans, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const double* r, const double* c, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cgerfsx( char* trans, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const float* r, const float* c, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* berr, lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sgbrfs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const float* ab, lapack_int* ldab, - const float* afb, lapack_int* ldafb, const lapack_int* ipiv, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgbrfs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const double* ab, lapack_int* ldab, - const double* afb, lapack_int* ldafb, - const lapack_int* ipiv, const double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* ferr, double* berr, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cgbrfs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const lapack_complex_float* ab, - lapack_int* ldab, const lapack_complex_float* afb, - lapack_int* ldafb, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zgbrfs( char* trans, lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, const lapack_complex_double* ab, - lapack_int* ldab, const lapack_complex_double* afb, - lapack_int* ldafb, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_dgbrfsx( char* trans, char* equed, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, const double* ab, - lapack_int* ldab, const double* afb, lapack_int* ldafb, - const lapack_int* ipiv, const double* r, const double* c, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* rcond, double* berr, - lapack_int* n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int* nparams, double* params, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_sgbrfsx( char* trans, char* equed, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, const float* ab, - lapack_int* ldab, const float* afb, lapack_int* ldafb, - const lapack_int* ipiv, const float* r, const float* c, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* rcond, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_zgbrfsx( char* trans, char* equed, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, - const lapack_complex_double* ab, lapack_int* ldab, - const lapack_complex_double* afb, lapack_int* ldafb, - const lapack_int* ipiv, const double* r, const double* c, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cgbrfsx( char* trans, char* equed, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, - const lapack_complex_float* ab, lapack_int* ldab, - const lapack_complex_float* afb, lapack_int* ldafb, - const lapack_int* ipiv, const float* r, const float* c, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* berr, lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sgtrfs( char* trans, lapack_int* n, lapack_int* nrhs, - const float* dl, const float* d, const float* du, - const float* dlf, const float* df, const float* duf, - const float* du2, const lapack_int* ipiv, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* ferr, - float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgtrfs( char* trans, lapack_int* n, lapack_int* nrhs, - const double* dl, const double* d, const double* du, - const double* dlf, const double* df, const double* duf, - const double* du2, const lapack_int* ipiv, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cgtrfs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, - const lapack_complex_float* dlf, - const lapack_complex_float* df, - const lapack_complex_float* duf, - const lapack_complex_float* du2, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zgtrfs( char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, - const lapack_complex_double* dlf, - const lapack_complex_double* df, - const lapack_complex_double* duf, - const lapack_complex_double* du2, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sporfs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* a, - lapack_int* lda, const float* af, lapack_int* ldaf, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dporfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* ferr, double* berr, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cporfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zporfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_dporfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const double* s, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* rcond, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_sporfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, const float* af, - lapack_int* ldaf, const float* s, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* berr, lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_zporfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const double* s, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cporfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const float* s, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_spprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const float* ap, const float* afp, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* ferr, - float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dpprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* ap, const double* afp, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cpprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, - const lapack_complex_float* afp, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zpprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_spbrfs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const float* ab, lapack_int* ldab, const float* afb, - lapack_int* ldafb, const float* b, lapack_int* ldb, - float* x, lapack_int* ldx, float* ferr, float* berr, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dpbrfs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const double* ab, lapack_int* ldab, const double* afb, - lapack_int* ldafb, const double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* ferr, double* berr, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cpbrfs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const lapack_complex_float* ab, lapack_int* ldab, - const lapack_complex_float* afb, lapack_int* ldafb, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zpbrfs( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - const lapack_complex_double* ab, lapack_int* ldab, - const lapack_complex_double* afb, lapack_int* ldafb, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sptrfs( lapack_int* n, lapack_int* nrhs, const float* d, - const float* e, const float* df, const float* ef, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* ferr, float* berr, float* work, lapack_int *info ); -void LAPACK_dptrfs( lapack_int* n, lapack_int* nrhs, const double* d, - const double* e, const double* df, const double* ef, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* ferr, double* berr, double* work, - lapack_int *info ); -void LAPACK_cptrfs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* d, - const lapack_complex_float* e, const float* df, - const lapack_complex_float* ef, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zptrfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* d, const lapack_complex_double* e, - const double* df, const lapack_complex_double* ef, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_ssyrfs( char* uplo, lapack_int* n, lapack_int* nrhs, const float* a, - lapack_int* lda, const float* af, lapack_int* ldaf, - const lapack_int* ipiv, const float* b, lapack_int* ldb, - float* x, lapack_int* ldx, float* ferr, float* berr, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dsyrfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const lapack_int* ipiv, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_csyrfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zsyrfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_dsyrfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, const double* af, - lapack_int* ldaf, const lapack_int* ipiv, const double* s, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* rcond, double* berr, - lapack_int* n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int* nparams, double* params, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_ssyrfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, const float* af, - lapack_int* ldaf, const lapack_int* ipiv, const float* s, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* rcond, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_zsyrfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const double* s, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_csyrfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const float* s, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* berr, lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_cherfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zherfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_zherfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* af, lapack_int* ldaf, - const lapack_int* ipiv, const double* s, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cherfsx( char* uplo, char* equed, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* af, lapack_int* ldaf, - const lapack_int* ipiv, const float* s, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* berr, lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ssprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const float* ap, const float* afp, const lapack_int* ipiv, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dsprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const double* ap, const double* afp, const lapack_int* ipiv, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* ferr, double* berr, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_csprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, - const lapack_complex_float* afp, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zsprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_chprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, - const lapack_complex_float* afp, const lapack_int* ipiv, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zhprfs( char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, - const lapack_complex_double* afp, const lapack_int* ipiv, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* ferr, - double* berr, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_strrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const float* a, lapack_int* lda, - const float* b, lapack_int* ldb, const float* x, - lapack_int* ldx, float* ferr, float* berr, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dtrrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const double* a, lapack_int* lda, - const double* b, lapack_int* ldb, const double* x, - lapack_int* ldx, double* ferr, double* berr, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ctrrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* b, - lapack_int* ldb, const lapack_complex_float* x, - lapack_int* ldx, float* ferr, float* berr, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztrrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* b, - lapack_int* ldb, const lapack_complex_double* x, - lapack_int* ldx, double* ferr, double* berr, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_stprfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const float* ap, const float* b, - lapack_int* ldb, const float* x, lapack_int* ldx, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dtprfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const double* ap, const double* b, - lapack_int* ldb, const double* x, lapack_int* ldx, - double* ferr, double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_ctprfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* ap, - const lapack_complex_float* b, lapack_int* ldb, - const lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztprfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* nrhs, const lapack_complex_double* ap, - const lapack_complex_double* b, lapack_int* ldb, - const lapack_complex_double* x, lapack_int* ldx, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_stbrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, const float* ab, - lapack_int* ldab, const float* b, lapack_int* ldb, - const float* x, lapack_int* ldx, float* ferr, float* berr, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dtbrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, const double* ab, - lapack_int* ldab, const double* b, lapack_int* ldb, - const double* x, lapack_int* ldx, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_ctbrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, - const lapack_complex_float* ab, lapack_int* ldab, - const lapack_complex_float* b, lapack_int* ldb, - const lapack_complex_float* x, lapack_int* ldx, float* ferr, - float* berr, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztbrfs( char* uplo, char* trans, char* diag, lapack_int* n, - lapack_int* kd, lapack_int* nrhs, - const lapack_complex_double* ab, lapack_int* ldab, - const lapack_complex_double* b, lapack_int* ldb, - const lapack_complex_double* x, lapack_int* ldx, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_sgetri( lapack_int* n, float* a, lapack_int* lda, - const lapack_int* ipiv, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgetri( lapack_int* n, double* a, lapack_int* lda, - const lapack_int* ipiv, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgetri( lapack_int* n, lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zgetri( lapack_int* n, lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_spotri( char* uplo, lapack_int* n, float* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dpotri( char* uplo, lapack_int* n, double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_cpotri( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_zpotri( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_dpftri( char* transr, char* uplo, lapack_int* n, double* a, - lapack_int *info ); -void LAPACK_spftri( char* transr, char* uplo, lapack_int* n, float* a, - lapack_int *info ); -void LAPACK_zpftri( char* transr, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int *info ); -void LAPACK_cpftri( char* transr, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int *info ); -void LAPACK_spptri( char* uplo, lapack_int* n, float* ap, lapack_int *info ); -void LAPACK_dpptri( char* uplo, lapack_int* n, double* ap, lapack_int *info ); -void LAPACK_cpptri( char* uplo, lapack_int* n, lapack_complex_float* ap, - lapack_int *info ); -void LAPACK_zpptri( char* uplo, lapack_int* n, lapack_complex_double* ap, - lapack_int *info ); -void LAPACK_ssytri( char* uplo, lapack_int* n, float* a, lapack_int* lda, - const lapack_int* ipiv, float* work, lapack_int *info ); -void LAPACK_dsytri( char* uplo, lapack_int* n, double* a, lapack_int* lda, - const lapack_int* ipiv, double* work, lapack_int *info ); -void LAPACK_csytri( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zsytri( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_chetri( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zhetri( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_ssptri( char* uplo, lapack_int* n, float* ap, - const lapack_int* ipiv, float* work, lapack_int *info ); -void LAPACK_dsptri( char* uplo, lapack_int* n, double* ap, - const lapack_int* ipiv, double* work, lapack_int *info ); -void LAPACK_csptri( char* uplo, lapack_int* n, lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zsptri( char* uplo, lapack_int* n, lapack_complex_double* ap, - const lapack_int* ipiv, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_chptri( char* uplo, lapack_int* n, lapack_complex_float* ap, - const lapack_int* ipiv, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zhptri( char* uplo, lapack_int* n, lapack_complex_double* ap, - const lapack_int* ipiv, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_strtri( char* uplo, char* diag, lapack_int* n, float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_dtrtri( char* uplo, char* diag, lapack_int* n, double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_ctrtri( char* uplo, char* diag, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_ztrtri( char* uplo, char* diag, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dtftri( char* transr, char* uplo, char* diag, lapack_int* n, - double* a, lapack_int *info ); -void LAPACK_stftri( char* transr, char* uplo, char* diag, lapack_int* n, - float* a, lapack_int *info ); -void LAPACK_ztftri( char* transr, char* uplo, char* diag, lapack_int* n, - lapack_complex_double* a, lapack_int *info ); -void LAPACK_ctftri( char* transr, char* uplo, char* diag, lapack_int* n, - lapack_complex_float* a, lapack_int *info ); -void LAPACK_stptri( char* uplo, char* diag, lapack_int* n, float* ap, - lapack_int *info ); -void LAPACK_dtptri( char* uplo, char* diag, lapack_int* n, double* ap, - lapack_int *info ); -void LAPACK_ctptri( char* uplo, char* diag, lapack_int* n, - lapack_complex_float* ap, lapack_int *info ); -void LAPACK_ztptri( char* uplo, char* diag, lapack_int* n, - lapack_complex_double* ap, lapack_int *info ); -void LAPACK_sgeequ( lapack_int* m, lapack_int* n, const float* a, - lapack_int* lda, float* r, float* c, float* rowcnd, - float* colcnd, float* amax, lapack_int *info ); -void LAPACK_dgeequ( lapack_int* m, lapack_int* n, const double* a, - lapack_int* lda, double* r, double* c, double* rowcnd, - double* colcnd, double* amax, lapack_int *info ); -void LAPACK_cgeequ( lapack_int* m, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* r, float* c, float* rowcnd, - float* colcnd, float* amax, lapack_int *info ); -void LAPACK_zgeequ( lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, double* r, - double* c, double* rowcnd, double* colcnd, double* amax, - lapack_int *info ); -void LAPACK_dgeequb( lapack_int* m, lapack_int* n, const double* a, - lapack_int* lda, double* r, double* c, double* rowcnd, - double* colcnd, double* amax, lapack_int *info ); -void LAPACK_sgeequb( lapack_int* m, lapack_int* n, const float* a, - lapack_int* lda, float* r, float* c, float* rowcnd, - float* colcnd, float* amax, lapack_int *info ); -void LAPACK_zgeequb( lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, double* r, - double* c, double* rowcnd, double* colcnd, double* amax, - lapack_int *info ); -void LAPACK_cgeequb( lapack_int* m, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, float* r, - float* c, float* rowcnd, float* colcnd, float* amax, - lapack_int *info ); -void LAPACK_sgbequ( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const float* ab, lapack_int* ldab, float* r, - float* c, float* rowcnd, float* colcnd, float* amax, - lapack_int *info ); -void LAPACK_dgbequ( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const double* ab, lapack_int* ldab, - double* r, double* c, double* rowcnd, double* colcnd, - double* amax, lapack_int *info ); -void LAPACK_cgbequ( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const lapack_complex_float* ab, - lapack_int* ldab, float* r, float* c, float* rowcnd, - float* colcnd, float* amax, lapack_int *info ); -void LAPACK_zgbequ( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const lapack_complex_double* ab, - lapack_int* ldab, double* r, double* c, double* rowcnd, - double* colcnd, double* amax, lapack_int *info ); -void LAPACK_dgbequb( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const double* ab, lapack_int* ldab, - double* r, double* c, double* rowcnd, double* colcnd, - double* amax, lapack_int *info ); -void LAPACK_sgbequb( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const float* ab, lapack_int* ldab, - float* r, float* c, float* rowcnd, float* colcnd, - float* amax, lapack_int *info ); -void LAPACK_zgbequb( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const lapack_complex_double* ab, - lapack_int* ldab, double* r, double* c, double* rowcnd, - double* colcnd, double* amax, lapack_int *info ); -void LAPACK_cgbequb( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const lapack_complex_float* ab, - lapack_int* ldab, float* r, float* c, float* rowcnd, - float* colcnd, float* amax, lapack_int *info ); -void LAPACK_spoequ( lapack_int* n, const float* a, lapack_int* lda, float* s, - float* scond, float* amax, lapack_int *info ); -void LAPACK_dpoequ( lapack_int* n, const double* a, lapack_int* lda, double* s, - double* scond, double* amax, lapack_int *info ); -void LAPACK_cpoequ( lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* s, float* scond, float* amax, - lapack_int *info ); -void LAPACK_zpoequ( lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* s, double* scond, double* amax, - lapack_int *info ); -void LAPACK_dpoequb( lapack_int* n, const double* a, lapack_int* lda, double* s, - double* scond, double* amax, lapack_int *info ); -void LAPACK_spoequb( lapack_int* n, const float* a, lapack_int* lda, float* s, - float* scond, float* amax, lapack_int *info ); -void LAPACK_zpoequb( lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* s, double* scond, double* amax, - lapack_int *info ); -void LAPACK_cpoequb( lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* s, float* scond, float* amax, - lapack_int *info ); -void LAPACK_sppequ( char* uplo, lapack_int* n, const float* ap, float* s, - float* scond, float* amax, lapack_int *info ); -void LAPACK_dppequ( char* uplo, lapack_int* n, const double* ap, double* s, - double* scond, double* amax, lapack_int *info ); -void LAPACK_cppequ( char* uplo, lapack_int* n, const lapack_complex_float* ap, - float* s, float* scond, float* amax, lapack_int *info ); -void LAPACK_zppequ( char* uplo, lapack_int* n, const lapack_complex_double* ap, - double* s, double* scond, double* amax, lapack_int *info ); -void LAPACK_spbequ( char* uplo, lapack_int* n, lapack_int* kd, const float* ab, - lapack_int* ldab, float* s, float* scond, float* amax, - lapack_int *info ); -void LAPACK_dpbequ( char* uplo, lapack_int* n, lapack_int* kd, const double* ab, - lapack_int* ldab, double* s, double* scond, double* amax, - lapack_int *info ); -void LAPACK_cpbequ( char* uplo, lapack_int* n, lapack_int* kd, - const lapack_complex_float* ab, lapack_int* ldab, float* s, - float* scond, float* amax, lapack_int *info ); -void LAPACK_zpbequ( char* uplo, lapack_int* n, lapack_int* kd, - const lapack_complex_double* ab, lapack_int* ldab, - double* s, double* scond, double* amax, lapack_int *info ); -void LAPACK_dsyequb( char* uplo, lapack_int* n, const double* a, - lapack_int* lda, double* s, double* scond, double* amax, - double* work, lapack_int *info ); -void LAPACK_ssyequb( char* uplo, lapack_int* n, const float* a, lapack_int* lda, - float* s, float* scond, float* amax, float* work, - lapack_int *info ); -void LAPACK_zsyequb( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* s, double* scond, double* amax, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_csyequb( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* s, float* scond, float* amax, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zheequb( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, double* s, double* scond, double* amax, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_cheequb( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, float* s, float* scond, float* amax, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_sgesv( lapack_int* n, lapack_int* nrhs, float* a, lapack_int* lda, - lapack_int* ipiv, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dgesv( lapack_int* n, lapack_int* nrhs, double* a, lapack_int* lda, - lapack_int* ipiv, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cgesv( lapack_int* n, lapack_int* nrhs, lapack_complex_float* a, - lapack_int* lda, lapack_int* ipiv, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zgesv( lapack_int* n, lapack_int* nrhs, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_dsgesv( lapack_int* n, lapack_int* nrhs, double* a, lapack_int* lda, - lapack_int* ipiv, double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* work, float* swork, - lapack_int* iter, lapack_int *info ); -void LAPACK_zcgesv( lapack_int* n, lapack_int* nrhs, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - lapack_complex_double* work, lapack_complex_float* swork, - double* rwork, lapack_int* iter, lapack_int *info ); -void LAPACK_sgesvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgesvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int* ldb, double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cgesvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zgesvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_dgesvxx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int* ldb, double* x, lapack_int* ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int* n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int* nparams, double* params, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_sgesvxx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_zgesvxx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cgesvxx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* r, float* c, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sgbsv( lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, float* ab, lapack_int* ldab, - lapack_int* ipiv, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dgbsv( lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, double* ab, lapack_int* ldab, - lapack_int* ipiv, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cgbsv( lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, lapack_complex_float* ab, lapack_int* ldab, - lapack_int* ipiv, lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zgbsv( lapack_int* n, lapack_int* kl, lapack_int* ku, - lapack_int* nrhs, lapack_complex_double* ab, - lapack_int* ldab, lapack_int* ipiv, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_sgbsvx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, float* ab, - lapack_int* ldab, float* afb, lapack_int* ldafb, - lapack_int* ipiv, char* equed, float* r, float* c, float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgbsvx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, double* ab, - lapack_int* ldab, double* afb, lapack_int* ldafb, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int* ldb, double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cgbsvx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, lapack_complex_float* ab, - lapack_int* ldab, lapack_complex_float* afb, - lapack_int* ldafb, lapack_int* ipiv, char* equed, float* r, - float* c, lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zgbsvx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, lapack_complex_double* ab, - lapack_int* ldab, lapack_complex_double* afb, - lapack_int* ldafb, lapack_int* ipiv, char* equed, double* r, - double* c, lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_dgbsvxx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, double* ab, - lapack_int* ldab, double* afb, lapack_int* ldafb, - lapack_int* ipiv, char* equed, double* r, double* c, - double* b, lapack_int* ldb, double* x, lapack_int* ldx, - double* rcond, double* rpvgrw, double* berr, - lapack_int* n_err_bnds, double* err_bnds_norm, - double* err_bnds_comp, lapack_int* nparams, double* params, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_sgbsvxx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, float* ab, - lapack_int* ldab, float* afb, lapack_int* ldafb, - lapack_int* ipiv, char* equed, float* r, float* c, - float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* rcond, float* rpvgrw, float* berr, - lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_zgbsvxx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, - lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* afb, lapack_int* ldafb, - lapack_int* ipiv, char* equed, double* r, double* c, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cgbsvxx( char* fact, char* trans, lapack_int* n, lapack_int* kl, - lapack_int* ku, lapack_int* nrhs, lapack_complex_float* ab, - lapack_int* ldab, lapack_complex_float* afb, - lapack_int* ldafb, lapack_int* ipiv, char* equed, float* r, - float* c, lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sgtsv( lapack_int* n, lapack_int* nrhs, float* dl, float* d, - float* du, float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_dgtsv( lapack_int* n, lapack_int* nrhs, double* dl, double* d, - double* du, double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_cgtsv( lapack_int* n, lapack_int* nrhs, lapack_complex_float* dl, - lapack_complex_float* d, lapack_complex_float* du, - lapack_complex_float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_zgtsv( lapack_int* n, lapack_int* nrhs, lapack_complex_double* dl, - lapack_complex_double* d, lapack_complex_double* du, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_sgtsvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - const float* dl, const float* d, const float* du, - float* dlf, float* df, float* duf, float* du2, - lapack_int* ipiv, const float* b, lapack_int* ldb, float* x, - lapack_int* ldx, float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dgtsvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - const double* dl, const double* d, const double* du, - double* dlf, double* df, double* duf, double* du2, - lapack_int* ipiv, const double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* rcond, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cgtsvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* dl, - const lapack_complex_float* d, - const lapack_complex_float* du, lapack_complex_float* dlf, - lapack_complex_float* df, lapack_complex_float* duf, - lapack_complex_float* du2, lapack_int* ipiv, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zgtsvx( char* fact, char* trans, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* dl, - const lapack_complex_double* d, - const lapack_complex_double* du, lapack_complex_double* dlf, - lapack_complex_double* df, lapack_complex_double* duf, - lapack_complex_double* du2, lapack_int* ipiv, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_sposv( char* uplo, lapack_int* n, lapack_int* nrhs, float* a, - lapack_int* lda, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dposv( char* uplo, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cposv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_zposv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dsposv( char* uplo, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* work, float* swork, - lapack_int* iter, lapack_int *info ); -void LAPACK_zcposv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, - lapack_complex_double* work, lapack_complex_float* swork, - double* rwork, lapack_int* iter, lapack_int *info ); -void LAPACK_sposvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* af, lapack_int* ldaf, - char* equed, float* s, float* b, lapack_int* ldb, float* x, - lapack_int* ldx, float* rcond, float* ferr, float* berr, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dposvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* af, lapack_int* ldaf, - char* equed, double* s, double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* rcond, double* ferr, - double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cposvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, char* equed, - float* s, lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zposvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, char* equed, - double* s, lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_dposvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* af, lapack_int* ldaf, - char* equed, double* s, double* b, lapack_int* ldb, - double* x, lapack_int* ldx, double* rcond, double* rpvgrw, - double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_sposvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* af, lapack_int* ldaf, - char* equed, float* s, float* b, lapack_int* ldb, float* x, - lapack_int* ldx, float* rcond, float* rpvgrw, float* berr, - lapack_int* n_err_bnds, float* err_bnds_norm, - float* err_bnds_comp, lapack_int* nparams, float* params, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_zposvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, char* equed, - double* s, lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_cposvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, char* equed, - float* s, lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sppsv( char* uplo, lapack_int* n, lapack_int* nrhs, float* ap, - float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_dppsv( char* uplo, lapack_int* n, lapack_int* nrhs, double* ap, - double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_cppsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* ap, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zppsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* ap, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_sppsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - float* ap, float* afp, char* equed, float* s, float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dppsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - double* ap, double* afp, char* equed, double* s, double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cppsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* ap, lapack_complex_float* afp, - char* equed, float* s, lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zppsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* ap, lapack_complex_double* afp, - char* equed, double* s, lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_spbsv( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - float* ab, lapack_int* ldab, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dpbsv( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - double* ab, lapack_int* ldab, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cpbsv( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_zpbsv( char* uplo, lapack_int* n, lapack_int* kd, lapack_int* nrhs, - lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_spbsvx( char* fact, char* uplo, lapack_int* n, lapack_int* kd, - lapack_int* nrhs, float* ab, lapack_int* ldab, float* afb, - lapack_int* ldafb, char* equed, float* s, float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dpbsvx( char* fact, char* uplo, lapack_int* n, lapack_int* kd, - lapack_int* nrhs, double* ab, lapack_int* ldab, double* afb, - lapack_int* ldafb, char* equed, double* s, double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_cpbsvx( char* fact, char* uplo, lapack_int* n, lapack_int* kd, - lapack_int* nrhs, lapack_complex_float* ab, - lapack_int* ldab, lapack_complex_float* afb, - lapack_int* ldafb, char* equed, float* s, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zpbsvx( char* fact, char* uplo, lapack_int* n, lapack_int* kd, - lapack_int* nrhs, lapack_complex_double* ab, - lapack_int* ldab, lapack_complex_double* afb, - lapack_int* ldafb, char* equed, double* s, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_sptsv( lapack_int* n, lapack_int* nrhs, float* d, float* e, - float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_dptsv( lapack_int* n, lapack_int* nrhs, double* d, double* e, - double* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_cptsv( lapack_int* n, lapack_int* nrhs, float* d, - lapack_complex_float* e, lapack_complex_float* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_zptsv( lapack_int* n, lapack_int* nrhs, double* d, - lapack_complex_double* e, lapack_complex_double* b, - lapack_int* ldb, lapack_int *info ); -void LAPACK_sptsvx( char* fact, lapack_int* n, lapack_int* nrhs, const float* d, - const float* e, float* df, float* ef, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int *info ); -void LAPACK_dptsvx( char* fact, lapack_int* n, lapack_int* nrhs, - const double* d, const double* e, double* df, double* ef, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* rcond, double* ferr, double* berr, - double* work, lapack_int *info ); -void LAPACK_cptsvx( char* fact, lapack_int* n, lapack_int* nrhs, const float* d, - const lapack_complex_float* e, float* df, - lapack_complex_float* ef, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zptsvx( char* fact, lapack_int* n, lapack_int* nrhs, - const double* d, const lapack_complex_double* e, double* df, - lapack_complex_double* ef, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_ssysv( char* uplo, lapack_int* n, lapack_int* nrhs, float* a, - lapack_int* lda, lapack_int* ipiv, float* b, lapack_int* ldb, - float* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_dsysv( char* uplo, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, lapack_int* ipiv, double* b, - lapack_int* ldb, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_csysv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zsysv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ssysvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const float* a, lapack_int* lda, float* af, - lapack_int* ldaf, lapack_int* ipiv, const float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* ferr, float* berr, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dsysvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const double* a, lapack_int* lda, double* af, - lapack_int* ldaf, lapack_int* ipiv, const double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* rcond, - double* ferr, double* berr, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_csysvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zsysvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_dsysvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* s, double* b, - lapack_int* ldb, double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ssysvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* s, float* b, - lapack_int* ldb, float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_zsysvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_csysvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_chesv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zhesv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_chesvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zhesvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_zhesvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, double* s, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* x, lapack_int* ldx, double* rcond, - double* rpvgrw, double* berr, lapack_int* n_err_bnds, - double* err_bnds_norm, double* err_bnds_comp, - lapack_int* nparams, double* params, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_chesvxx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* af, lapack_int* ldaf, - lapack_int* ipiv, char* equed, float* s, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* x, lapack_int* ldx, float* rcond, - float* rpvgrw, float* berr, lapack_int* n_err_bnds, - float* err_bnds_norm, float* err_bnds_comp, - lapack_int* nparams, float* params, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_sspsv( char* uplo, lapack_int* n, lapack_int* nrhs, float* ap, - lapack_int* ipiv, float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dspsv( char* uplo, lapack_int* n, lapack_int* nrhs, double* ap, - lapack_int* ipiv, double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_cspsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* ap, lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_zspsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* ap, lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_sspsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const float* ap, float* afp, lapack_int* ipiv, - const float* b, lapack_int* ldb, float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, float* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dspsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const double* ap, double* afp, lapack_int* ipiv, - const double* b, lapack_int* ldb, double* x, - lapack_int* ldx, double* rcond, double* ferr, double* berr, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cspsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, lapack_complex_float* afp, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zspsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, lapack_complex_double* afp, - lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_chpsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* ap, lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, lapack_int *info ); -void LAPACK_zhpsv( char* uplo, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* ap, lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_chpsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_float* ap, lapack_complex_float* afp, - lapack_int* ipiv, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* x, lapack_int* ldx, - float* rcond, float* ferr, float* berr, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zhpsvx( char* fact, char* uplo, lapack_int* n, lapack_int* nrhs, - const lapack_complex_double* ap, lapack_complex_double* afp, - lapack_int* ipiv, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* x, lapack_int* ldx, - double* rcond, double* ferr, double* berr, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sgeqrf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgeqrf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgeqrf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgeqrf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgeqpf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - lapack_int* jpvt, float* tau, float* work, - lapack_int *info ); -void LAPACK_dgeqpf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - lapack_int* jpvt, double* tau, double* work, - lapack_int *info ); -void LAPACK_cgeqpf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* jpvt, - lapack_complex_float* tau, lapack_complex_float* work, - float* rwork, lapack_int *info ); -void LAPACK_zgeqpf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* jpvt, - lapack_complex_double* tau, lapack_complex_double* work, - double* rwork, lapack_int *info ); -void LAPACK_sgeqp3( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - lapack_int* jpvt, float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dgeqp3( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - lapack_int* jpvt, double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cgeqp3( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* jpvt, - lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int *info ); -void LAPACK_zgeqp3( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* jpvt, - lapack_complex_double* tau, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int *info ); -void LAPACK_sorgqr( lapack_int* m, lapack_int* n, lapack_int* k, float* a, - lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorgqr( lapack_int* m, lapack_int* n, lapack_int* k, double* a, - lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sormqr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const float* a, lapack_int* lda, - const float* tau, float* c, lapack_int* ldc, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dormqr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const double* a, lapack_int* lda, - const double* tau, double* c, lapack_int* ldc, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cungqr( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zungqr( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmqr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmqr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgelqf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgelqf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgelqf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgelqf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sorglq( lapack_int* m, lapack_int* n, lapack_int* k, float* a, - lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorglq( lapack_int* m, lapack_int* n, lapack_int* k, double* a, - lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sormlq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const float* a, lapack_int* lda, - const float* tau, float* c, lapack_int* ldc, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dormlq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const double* a, lapack_int* lda, - const double* tau, double* c, lapack_int* ldc, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cunglq( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zunglq( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmlq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmlq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgeqlf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgeqlf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgeqlf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgeqlf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sorgql( lapack_int* m, lapack_int* n, lapack_int* k, float* a, - lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorgql( lapack_int* m, lapack_int* n, lapack_int* k, double* a, - lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cungql( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zungql( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sormql( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const float* a, lapack_int* lda, - const float* tau, float* c, lapack_int* ldc, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dormql( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const double* a, lapack_int* lda, - const double* tau, double* c, lapack_int* ldc, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cunmql( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmql( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgerqf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgerqf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgerqf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgerqf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sorgrq( lapack_int* m, lapack_int* n, lapack_int* k, float* a, - lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorgrq( lapack_int* m, lapack_int* n, lapack_int* k, double* a, - lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cungrq( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zungrq( lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sormrq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const float* a, lapack_int* lda, - const float* tau, float* c, lapack_int* ldc, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dormrq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const double* a, lapack_int* lda, - const double* tau, double* c, lapack_int* ldc, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cunmrq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmrq( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_stzrzf( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dtzrzf( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ctzrzf( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ztzrzf( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sormrz( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, const float* a, - lapack_int* lda, const float* tau, float* c, - lapack_int* ldc, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dormrz( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, const double* a, - lapack_int* lda, const double* tau, double* c, - lapack_int* ldc, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmrz( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmrz( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, lapack_complex_double* c, - lapack_int* ldc, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sggqrf( lapack_int* n, lapack_int* m, lapack_int* p, float* a, - lapack_int* lda, float* taua, float* b, lapack_int* ldb, - float* taub, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dggqrf( lapack_int* n, lapack_int* m, lapack_int* p, double* a, - lapack_int* lda, double* taua, double* b, lapack_int* ldb, - double* taub, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cggqrf( lapack_int* n, lapack_int* m, lapack_int* p, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* taua, lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* taub, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zggqrf( lapack_int* n, lapack_int* m, lapack_int* p, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* taua, lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* taub, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sggrqf( lapack_int* m, lapack_int* p, lapack_int* n, float* a, - lapack_int* lda, float* taua, float* b, lapack_int* ldb, - float* taub, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dggrqf( lapack_int* m, lapack_int* p, lapack_int* n, double* a, - lapack_int* lda, double* taua, double* b, lapack_int* ldb, - double* taub, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cggrqf( lapack_int* m, lapack_int* p, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* taua, lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* taub, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zggrqf( lapack_int* m, lapack_int* p, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* taua, lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* taub, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgebrd( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* d, float* e, float* tauq, float* taup, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dgebrd( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* d, double* e, double* tauq, double* taup, - double* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_cgebrd( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, float* d, float* e, - lapack_complex_float* tauq, lapack_complex_float* taup, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgebrd( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, double* d, double* e, - lapack_complex_double* tauq, lapack_complex_double* taup, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgbbrd( char* vect, lapack_int* m, lapack_int* n, lapack_int* ncc, - lapack_int* kl, lapack_int* ku, float* ab, lapack_int* ldab, - float* d, float* e, float* q, lapack_int* ldq, float* pt, - lapack_int* ldpt, float* c, lapack_int* ldc, float* work, - lapack_int *info ); -void LAPACK_dgbbrd( char* vect, lapack_int* m, lapack_int* n, lapack_int* ncc, - lapack_int* kl, lapack_int* ku, double* ab, - lapack_int* ldab, double* d, double* e, double* q, - lapack_int* ldq, double* pt, lapack_int* ldpt, double* c, - lapack_int* ldc, double* work, lapack_int *info ); -void LAPACK_cgbbrd( char* vect, lapack_int* m, lapack_int* n, lapack_int* ncc, - lapack_int* kl, lapack_int* ku, lapack_complex_float* ab, - lapack_int* ldab, float* d, float* e, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* pt, lapack_int* ldpt, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zgbbrd( char* vect, lapack_int* m, lapack_int* n, lapack_int* ncc, - lapack_int* kl, lapack_int* ku, lapack_complex_double* ab, - lapack_int* ldab, double* d, double* e, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* pt, lapack_int* ldpt, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sorgbr( char* vect, lapack_int* m, lapack_int* n, lapack_int* k, - float* a, lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorgbr( char* vect, lapack_int* m, lapack_int* n, lapack_int* k, - double* a, lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sormbr( char* vect, char* side, char* trans, lapack_int* m, - lapack_int* n, lapack_int* k, const float* a, - lapack_int* lda, const float* tau, float* c, - lapack_int* ldc, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dormbr( char* vect, char* side, char* trans, lapack_int* m, - lapack_int* n, lapack_int* k, const double* a, - lapack_int* lda, const double* tau, double* c, - lapack_int* ldc, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cungbr( char* vect, lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zungbr( char* vect, lapack_int* m, lapack_int* n, lapack_int* k, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmbr( char* vect, char* side, char* trans, lapack_int* m, - lapack_int* n, lapack_int* k, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmbr( char* vect, char* side, char* trans, lapack_int* m, - lapack_int* n, lapack_int* k, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, lapack_complex_double* c, - lapack_int* ldc, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sbdsqr( char* uplo, lapack_int* n, lapack_int* ncvt, - lapack_int* nru, lapack_int* ncc, float* d, float* e, - float* vt, lapack_int* ldvt, float* u, lapack_int* ldu, - float* c, lapack_int* ldc, float* work, lapack_int *info ); -void LAPACK_dbdsqr( char* uplo, lapack_int* n, lapack_int* ncvt, - lapack_int* nru, lapack_int* ncc, double* d, double* e, - double* vt, lapack_int* ldvt, double* u, lapack_int* ldu, - double* c, lapack_int* ldc, double* work, - lapack_int *info ); -void LAPACK_cbdsqr( char* uplo, lapack_int* n, lapack_int* ncvt, - lapack_int* nru, lapack_int* ncc, float* d, float* e, - lapack_complex_float* vt, lapack_int* ldvt, - lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* c, lapack_int* ldc, float* work, - lapack_int *info ); -void LAPACK_zbdsqr( char* uplo, lapack_int* n, lapack_int* ncvt, - lapack_int* nru, lapack_int* ncc, double* d, double* e, - lapack_complex_double* vt, lapack_int* ldvt, - lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* c, lapack_int* ldc, double* work, - lapack_int *info ); -void LAPACK_sbdsdc( char* uplo, char* compq, lapack_int* n, float* d, float* e, - float* u, lapack_int* ldu, float* vt, lapack_int* ldvt, - float* q, lapack_int* iq, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dbdsdc( char* uplo, char* compq, lapack_int* n, double* d, - double* e, double* u, lapack_int* ldu, double* vt, - lapack_int* ldvt, double* q, lapack_int* iq, double* work, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ssytrd( char* uplo, lapack_int* n, float* a, lapack_int* lda, - float* d, float* e, float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dsytrd( char* uplo, lapack_int* n, double* a, lapack_int* lda, - double* d, double* e, double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sorgtr( char* uplo, lapack_int* n, float* a, lapack_int* lda, - const float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dorgtr( char* uplo, lapack_int* n, double* a, lapack_int* lda, - const double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sormtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const float* a, lapack_int* lda, - const float* tau, float* c, lapack_int* ldc, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dormtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const double* a, lapack_int* lda, - const double* tau, double* c, lapack_int* ldc, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_chetrd( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, float* d, float* e, - lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zhetrd( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, double* d, double* e, - lapack_complex_double* tau, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cungtr( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zungtr( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zunmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ssptrd( char* uplo, lapack_int* n, float* ap, float* d, float* e, - float* tau, lapack_int *info ); -void LAPACK_dsptrd( char* uplo, lapack_int* n, double* ap, double* d, double* e, - double* tau, lapack_int *info ); -void LAPACK_sopgtr( char* uplo, lapack_int* n, const float* ap, - const float* tau, float* q, lapack_int* ldq, float* work, - lapack_int *info ); -void LAPACK_dopgtr( char* uplo, lapack_int* n, const double* ap, - const double* tau, double* q, lapack_int* ldq, double* work, - lapack_int *info ); -void LAPACK_sopmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const float* ap, const float* tau, float* c, - lapack_int* ldc, float* work, lapack_int *info ); -void LAPACK_dopmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const double* ap, const double* tau, - double* c, lapack_int* ldc, double* work, - lapack_int *info ); -void LAPACK_chptrd( char* uplo, lapack_int* n, lapack_complex_float* ap, - float* d, float* e, lapack_complex_float* tau, - lapack_int *info ); -void LAPACK_zhptrd( char* uplo, lapack_int* n, lapack_complex_double* ap, - double* d, double* e, lapack_complex_double* tau, - lapack_int *info ); -void LAPACK_cupgtr( char* uplo, lapack_int* n, const lapack_complex_float* ap, - const lapack_complex_float* tau, lapack_complex_float* q, - lapack_int* ldq, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zupgtr( char* uplo, lapack_int* n, const lapack_complex_double* ap, - const lapack_complex_double* tau, lapack_complex_double* q, - lapack_int* ldq, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_cupmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const lapack_complex_float* ap, - const lapack_complex_float* tau, lapack_complex_float* c, - lapack_int* ldc, lapack_complex_float* work, - lapack_int *info ); -void LAPACK_zupmtr( char* side, char* uplo, char* trans, lapack_int* m, - lapack_int* n, const lapack_complex_double* ap, - const lapack_complex_double* tau, lapack_complex_double* c, - lapack_int* ldc, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_ssbtrd( char* vect, char* uplo, lapack_int* n, lapack_int* kd, - float* ab, lapack_int* ldab, float* d, float* e, float* q, - lapack_int* ldq, float* work, lapack_int *info ); -void LAPACK_dsbtrd( char* vect, char* uplo, lapack_int* n, lapack_int* kd, - double* ab, lapack_int* ldab, double* d, double* e, - double* q, lapack_int* ldq, double* work, - lapack_int *info ); -void LAPACK_chbtrd( char* vect, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_float* ab, lapack_int* ldab, float* d, - float* e, lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zhbtrd( char* vect, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_double* ab, lapack_int* ldab, double* d, - double* e, lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_ssterf( lapack_int* n, float* d, float* e, lapack_int *info ); -void LAPACK_dsterf( lapack_int* n, double* d, double* e, lapack_int *info ); -void LAPACK_ssteqr( char* compz, lapack_int* n, float* d, float* e, float* z, - lapack_int* ldz, float* work, lapack_int *info ); -void LAPACK_dsteqr( char* compz, lapack_int* n, double* d, double* e, double* z, - lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_csteqr( char* compz, lapack_int* n, float* d, float* e, - lapack_complex_float* z, lapack_int* ldz, float* work, - lapack_int *info ); -void LAPACK_zsteqr( char* compz, lapack_int* n, double* d, double* e, - lapack_complex_double* z, lapack_int* ldz, double* work, - lapack_int *info ); -void LAPACK_sstemr( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - lapack_int* m, float* w, float* z, lapack_int* ldz, - lapack_int* nzc, lapack_int* isuppz, lapack_logical* tryrac, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_dstemr( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, lapack_int* m, double* w, double* z, - lapack_int* ldz, lapack_int* nzc, lapack_int* isuppz, - lapack_logical* tryrac, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_cstemr( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_int* nzc, lapack_int* isuppz, - lapack_logical* tryrac, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_zstemr( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int* ldz, lapack_int* nzc, - lapack_int* isuppz, lapack_logical* tryrac, double* work, - lapack_int* lwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_sstedc( char* compz, lapack_int* n, float* d, float* e, float* z, - lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dstedc( char* compz, lapack_int* n, double* d, double* e, double* z, - lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_cstedc( char* compz, lapack_int* n, float* d, float* e, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zstedc( char* compz, lapack_int* n, double* d, double* e, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_sstegr( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, float* z, - lapack_int* ldz, lapack_int* isuppz, float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_dstegr( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, lapack_int* isuppz, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_cstegr( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_int* isuppz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_zstegr( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_int* isuppz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_spteqr( char* compz, lapack_int* n, float* d, float* e, float* z, - lapack_int* ldz, float* work, lapack_int *info ); -void LAPACK_dpteqr( char* compz, lapack_int* n, double* d, double* e, double* z, - lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_cpteqr( char* compz, lapack_int* n, float* d, float* e, - lapack_complex_float* z, lapack_int* ldz, float* work, - lapack_int *info ); -void LAPACK_zpteqr( char* compz, lapack_int* n, double* d, double* e, - lapack_complex_double* z, lapack_int* ldz, double* work, - lapack_int *info ); -void LAPACK_sstebz( char* range, char* order, lapack_int* n, float* vl, - float* vu, lapack_int* il, lapack_int* iu, float* abstol, - const float* d, const float* e, lapack_int* m, - lapack_int* nsplit, float* w, lapack_int* iblock, - lapack_int* isplit, float* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dstebz( char* range, char* order, lapack_int* n, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - const double* d, const double* e, lapack_int* m, - lapack_int* nsplit, double* w, lapack_int* iblock, - lapack_int* isplit, double* work, lapack_int* iwork, - lapack_int *info ); -void LAPACK_sstein( lapack_int* n, const float* d, const float* e, - lapack_int* m, const float* w, const lapack_int* iblock, - const lapack_int* isplit, float* z, lapack_int* ldz, - float* work, lapack_int* iwork, lapack_int* ifailv, - lapack_int *info ); -void LAPACK_dstein( lapack_int* n, const double* d, const double* e, - lapack_int* m, const double* w, const lapack_int* iblock, - const lapack_int* isplit, double* z, lapack_int* ldz, - double* work, lapack_int* iwork, lapack_int* ifailv, - lapack_int *info ); -void LAPACK_cstein( lapack_int* n, const float* d, const float* e, - lapack_int* m, const float* w, const lapack_int* iblock, - const lapack_int* isplit, lapack_complex_float* z, - lapack_int* ldz, float* work, lapack_int* iwork, - lapack_int* ifailv, lapack_int *info ); -void LAPACK_zstein( lapack_int* n, const double* d, const double* e, - lapack_int* m, const double* w, const lapack_int* iblock, - const lapack_int* isplit, lapack_complex_double* z, - lapack_int* ldz, double* work, lapack_int* iwork, - lapack_int* ifailv, lapack_int *info ); -void LAPACK_sdisna( char* job, lapack_int* m, lapack_int* n, const float* d, - float* sep, lapack_int *info ); -void LAPACK_ddisna( char* job, lapack_int* m, lapack_int* n, const double* d, - double* sep, lapack_int *info ); -void LAPACK_ssygst( lapack_int* itype, char* uplo, lapack_int* n, float* a, - lapack_int* lda, const float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_dsygst( lapack_int* itype, char* uplo, lapack_int* n, double* a, - lapack_int* lda, const double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_chegst( lapack_int* itype, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_zhegst( lapack_int* itype, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* b, lapack_int* ldb, - lapack_int *info ); -void LAPACK_sspgst( lapack_int* itype, char* uplo, lapack_int* n, float* ap, - const float* bp, lapack_int *info ); -void LAPACK_dspgst( lapack_int* itype, char* uplo, lapack_int* n, double* ap, - const double* bp, lapack_int *info ); -void LAPACK_chpgst( lapack_int* itype, char* uplo, lapack_int* n, - lapack_complex_float* ap, const lapack_complex_float* bp, - lapack_int *info ); -void LAPACK_zhpgst( lapack_int* itype, char* uplo, lapack_int* n, - lapack_complex_double* ap, const lapack_complex_double* bp, - lapack_int *info ); -void LAPACK_ssbgst( char* vect, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, float* ab, lapack_int* ldab, - const float* bb, lapack_int* ldbb, float* x, - lapack_int* ldx, float* work, lapack_int *info ); -void LAPACK_dsbgst( char* vect, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, double* ab, lapack_int* ldab, - const double* bb, lapack_int* ldbb, double* x, - lapack_int* ldx, double* work, lapack_int *info ); -void LAPACK_chbgst( char* vect, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_float* ab, lapack_int* ldab, - const lapack_complex_float* bb, lapack_int* ldbb, - lapack_complex_float* x, lapack_int* ldx, - lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zhbgst( char* vect, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_double* ab, lapack_int* ldab, - const lapack_complex_double* bb, lapack_int* ldbb, - lapack_complex_double* x, lapack_int* ldx, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_spbstf( char* uplo, lapack_int* n, lapack_int* kb, float* bb, - lapack_int* ldbb, lapack_int *info ); -void LAPACK_dpbstf( char* uplo, lapack_int* n, lapack_int* kb, double* bb, - lapack_int* ldbb, lapack_int *info ); -void LAPACK_cpbstf( char* uplo, lapack_int* n, lapack_int* kb, - lapack_complex_float* bb, lapack_int* ldbb, - lapack_int *info ); -void LAPACK_zpbstf( char* uplo, lapack_int* n, lapack_int* kb, - lapack_complex_double* bb, lapack_int* ldbb, - lapack_int *info ); -void LAPACK_sgehrd( lapack_int* n, lapack_int* ilo, lapack_int* ihi, float* a, - lapack_int* lda, float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgehrd( lapack_int* n, lapack_int* ilo, lapack_int* ihi, double* a, - lapack_int* lda, double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cgehrd( lapack_int* n, lapack_int* ilo, lapack_int* ihi, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zgehrd( lapack_int* n, lapack_int* ilo, lapack_int* ihi, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* tau, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sorghr( lapack_int* n, lapack_int* ilo, lapack_int* ihi, float* a, - lapack_int* lda, const float* tau, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dorghr( lapack_int* n, lapack_int* ilo, lapack_int* ihi, double* a, - lapack_int* lda, const double* tau, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sormhr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, const float* a, - lapack_int* lda, const float* tau, float* c, - lapack_int* ldc, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dormhr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, const double* a, - lapack_int* lda, const double* tau, double* c, - lapack_int* ldc, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunghr( lapack_int* n, lapack_int* ilo, lapack_int* ihi, - lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zunghr( lapack_int* n, lapack_int* ilo, lapack_int* ihi, - lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cunmhr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* tau, lapack_complex_float* c, - lapack_int* ldc, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zunmhr( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* tau, lapack_complex_double* c, - lapack_int* ldc, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sgebal( char* job, lapack_int* n, float* a, lapack_int* lda, - lapack_int* ilo, lapack_int* ihi, float* scale, - lapack_int *info ); -void LAPACK_dgebal( char* job, lapack_int* n, double* a, lapack_int* lda, - lapack_int* ilo, lapack_int* ihi, double* scale, - lapack_int *info ); -void LAPACK_cgebal( char* job, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* ilo, lapack_int* ihi, - float* scale, lapack_int *info ); -void LAPACK_zgebal( char* job, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* ilo, lapack_int* ihi, - double* scale, lapack_int *info ); -void LAPACK_sgebak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const float* scale, lapack_int* m, - float* v, lapack_int* ldv, lapack_int *info ); -void LAPACK_dgebak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const double* scale, lapack_int* m, - double* v, lapack_int* ldv, lapack_int *info ); -void LAPACK_cgebak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const float* scale, lapack_int* m, - lapack_complex_float* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_zgebak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const double* scale, lapack_int* m, - lapack_complex_double* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_shseqr( char* job, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, float* h, lapack_int* ldh, float* wr, - float* wi, float* z, lapack_int* ldz, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dhseqr( char* job, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, double* h, lapack_int* ldh, double* wr, - double* wi, double* z, lapack_int* ldz, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_chseqr( char* job, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, lapack_complex_float* h, lapack_int* ldh, - lapack_complex_float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zhseqr( char* job, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, lapack_complex_double* h, lapack_int* ldh, - lapack_complex_double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_shsein( char* job, char* eigsrc, char* initv, - lapack_logical* select, lapack_int* n, const float* h, - lapack_int* ldh, float* wr, const float* wi, float* vl, - lapack_int* ldvl, float* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, float* work, - lapack_int* ifaill, lapack_int* ifailr, lapack_int *info ); -void LAPACK_dhsein( char* job, char* eigsrc, char* initv, - lapack_logical* select, lapack_int* n, const double* h, - lapack_int* ldh, double* wr, const double* wi, double* vl, - lapack_int* ldvl, double* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, double* work, - lapack_int* ifaill, lapack_int* ifailr, lapack_int *info ); -void LAPACK_chsein( char* job, char* eigsrc, char* initv, - const lapack_logical* select, lapack_int* n, - const lapack_complex_float* h, lapack_int* ldh, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int* ldvl, lapack_complex_float* vr, - lapack_int* ldvr, lapack_int* mm, lapack_int* m, - lapack_complex_float* work, float* rwork, - lapack_int* ifaill, lapack_int* ifailr, lapack_int *info ); -void LAPACK_zhsein( char* job, char* eigsrc, char* initv, - const lapack_logical* select, lapack_int* n, - const lapack_complex_double* h, lapack_int* ldh, - lapack_complex_double* w, lapack_complex_double* vl, - lapack_int* ldvl, lapack_complex_double* vr, - lapack_int* ldvr, lapack_int* mm, lapack_int* m, - lapack_complex_double* work, double* rwork, - lapack_int* ifaill, lapack_int* ifailr, lapack_int *info ); -void LAPACK_strevc( char* side, char* howmny, lapack_logical* select, - lapack_int* n, const float* t, lapack_int* ldt, float* vl, - lapack_int* ldvl, float* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, float* work, - lapack_int *info ); -void LAPACK_dtrevc( char* side, char* howmny, lapack_logical* select, - lapack_int* n, const double* t, lapack_int* ldt, double* vl, - lapack_int* ldvl, double* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, double* work, - lapack_int *info ); -void LAPACK_ctrevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* vl, lapack_int* ldvl, - lapack_complex_float* vr, lapack_int* ldvr, lapack_int* mm, - lapack_int* m, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztrevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* vl, lapack_int* ldvl, - lapack_complex_double* vr, lapack_int* ldvr, lapack_int* mm, - lapack_int* m, lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_strsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const float* t, lapack_int* ldt, - const float* vl, lapack_int* ldvl, const float* vr, - lapack_int* ldvr, float* s, float* sep, lapack_int* mm, - lapack_int* m, float* work, lapack_int* ldwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dtrsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const double* t, lapack_int* ldt, - const double* vl, lapack_int* ldvl, const double* vr, - lapack_int* ldvr, double* s, double* sep, lapack_int* mm, - lapack_int* m, double* work, lapack_int* ldwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ctrsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_float* t, - lapack_int* ldt, const lapack_complex_float* vl, - lapack_int* ldvl, const lapack_complex_float* vr, - lapack_int* ldvr, float* s, float* sep, lapack_int* mm, - lapack_int* m, lapack_complex_float* work, - lapack_int* ldwork, float* rwork, lapack_int *info ); -void LAPACK_ztrsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_double* t, - lapack_int* ldt, const lapack_complex_double* vl, - lapack_int* ldvl, const lapack_complex_double* vr, - lapack_int* ldvr, double* s, double* sep, lapack_int* mm, - lapack_int* m, lapack_complex_double* work, - lapack_int* ldwork, double* rwork, lapack_int *info ); -void LAPACK_strexc( char* compq, lapack_int* n, float* t, lapack_int* ldt, - float* q, lapack_int* ldq, lapack_int* ifst, - lapack_int* ilst, float* work, lapack_int *info ); -void LAPACK_dtrexc( char* compq, lapack_int* n, double* t, lapack_int* ldt, - double* q, lapack_int* ldq, lapack_int* ifst, - lapack_int* ilst, double* work, lapack_int *info ); -void LAPACK_ctrexc( char* compq, lapack_int* n, lapack_complex_float* t, - lapack_int* ldt, lapack_complex_float* q, lapack_int* ldq, - lapack_int* ifst, lapack_int* ilst, lapack_int *info ); -void LAPACK_ztrexc( char* compq, lapack_int* n, lapack_complex_double* t, - lapack_int* ldt, lapack_complex_double* q, lapack_int* ldq, - lapack_int* ifst, lapack_int* ilst, lapack_int *info ); -void LAPACK_strsen( char* job, char* compq, const lapack_logical* select, - lapack_int* n, float* t, lapack_int* ldt, float* q, - lapack_int* ldq, float* wr, float* wi, lapack_int* m, - float* s, float* sep, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dtrsen( char* job, char* compq, const lapack_logical* select, - lapack_int* n, double* t, lapack_int* ldt, double* q, - lapack_int* ldq, double* wr, double* wi, lapack_int* m, - double* s, double* sep, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_ctrsen( char* job, char* compq, const lapack_logical* select, - lapack_int* n, lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* w, lapack_int* m, float* s, - float* sep, lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_ztrsen( char* job, char* compq, const lapack_logical* select, - lapack_int* n, lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* w, lapack_int* m, double* s, - double* sep, lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_strsyl( char* trana, char* tranb, lapack_int* isgn, lapack_int* m, - lapack_int* n, const float* a, lapack_int* lda, - const float* b, lapack_int* ldb, float* c, lapack_int* ldc, - float* scale, lapack_int *info ); -void LAPACK_dtrsyl( char* trana, char* tranb, lapack_int* isgn, lapack_int* m, - lapack_int* n, const double* a, lapack_int* lda, - const double* b, lapack_int* ldb, double* c, - lapack_int* ldc, double* scale, lapack_int *info ); -void LAPACK_ctrsyl( char* trana, char* tranb, lapack_int* isgn, lapack_int* m, - lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* b, - lapack_int* ldb, lapack_complex_float* c, lapack_int* ldc, - float* scale, lapack_int *info ); -void LAPACK_ztrsyl( char* trana, char* tranb, lapack_int* isgn, lapack_int* m, - lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* b, - lapack_int* ldb, lapack_complex_double* c, lapack_int* ldc, - double* scale, lapack_int *info ); -void LAPACK_sgghrd( char* compq, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, float* a, lapack_int* lda, float* b, - lapack_int* ldb, float* q, lapack_int* ldq, float* z, - lapack_int* ldz, lapack_int *info ); -void LAPACK_dgghrd( char* compq, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, double* a, lapack_int* lda, double* b, - lapack_int* ldb, double* q, lapack_int* ldq, double* z, - lapack_int* ldz, lapack_int *info ); -void LAPACK_cgghrd( char* compq, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* z, lapack_int* ldz, - lapack_int *info ); -void LAPACK_zgghrd( char* compq, char* compz, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* z, lapack_int* ldz, - lapack_int *info ); -void LAPACK_sggbal( char* job, lapack_int* n, float* a, lapack_int* lda, - float* b, lapack_int* ldb, lapack_int* ilo, lapack_int* ihi, - float* lscale, float* rscale, float* work, - lapack_int *info ); -void LAPACK_dggbal( char* job, lapack_int* n, double* a, lapack_int* lda, - double* b, lapack_int* ldb, lapack_int* ilo, - lapack_int* ihi, double* lscale, double* rscale, - double* work, lapack_int *info ); -void LAPACK_cggbal( char* job, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* b, lapack_int* ldb, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale, float* work, lapack_int *info ); -void LAPACK_zggbal( char* job, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* b, lapack_int* ldb, - lapack_int* ilo, lapack_int* ihi, double* lscale, - double* rscale, double* work, lapack_int *info ); -void LAPACK_sggbak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const float* lscale, const float* rscale, - lapack_int* m, float* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_dggbak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const double* lscale, const double* rscale, - lapack_int* m, double* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_cggbak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const float* lscale, const float* rscale, - lapack_int* m, lapack_complex_float* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_zggbak( char* job, char* side, lapack_int* n, lapack_int* ilo, - lapack_int* ihi, const double* lscale, const double* rscale, - lapack_int* m, lapack_complex_double* v, lapack_int* ldv, - lapack_int *info ); -void LAPACK_shgeqz( char* job, char* compq, char* compz, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, float* h, lapack_int* ldh, - float* t, lapack_int* ldt, float* alphar, float* alphai, - float* beta, float* q, lapack_int* ldq, float* z, - lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dhgeqz( char* job, char* compq, char* compz, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, double* h, - lapack_int* ldh, double* t, lapack_int* ldt, double* alphar, - double* alphai, double* beta, double* q, lapack_int* ldq, - double* z, lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_chgeqz( char* job, char* compq, char* compz, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, lapack_complex_float* h, - lapack_int* ldh, lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zhgeqz( char* job, char* compq, char* compz, lapack_int* n, - lapack_int* ilo, lapack_int* ihi, lapack_complex_double* h, - lapack_int* ldh, lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_stgevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, const float* s, lapack_int* lds, - const float* p, lapack_int* ldp, float* vl, - lapack_int* ldvl, float* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, float* work, - lapack_int *info ); -void LAPACK_dtgevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, const double* s, lapack_int* lds, - const double* p, lapack_int* ldp, double* vl, - lapack_int* ldvl, double* vr, lapack_int* ldvr, - lapack_int* mm, lapack_int* m, double* work, - lapack_int *info ); -void LAPACK_ctgevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_float* s, - lapack_int* lds, const lapack_complex_float* p, - lapack_int* ldp, lapack_complex_float* vl, lapack_int* ldvl, - lapack_complex_float* vr, lapack_int* ldvr, lapack_int* mm, - lapack_int* m, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_ztgevc( char* side, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_double* s, - lapack_int* lds, const lapack_complex_double* p, - lapack_int* ldp, lapack_complex_double* vl, - lapack_int* ldvl, lapack_complex_double* vr, - lapack_int* ldvr, lapack_int* mm, lapack_int* m, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_stgexc( lapack_logical* wantq, lapack_logical* wantz, lapack_int* n, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* q, lapack_int* ldq, float* z, lapack_int* ldz, - lapack_int* ifst, lapack_int* ilst, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dtgexc( lapack_logical* wantq, lapack_logical* wantz, lapack_int* n, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* q, lapack_int* ldq, double* z, lapack_int* ldz, - lapack_int* ifst, lapack_int* ilst, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_ctgexc( lapack_logical* wantq, lapack_logical* wantz, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* z, lapack_int* ldz, lapack_int* ifst, - lapack_int* ilst, lapack_int *info ); -void LAPACK_ztgexc( lapack_logical* wantq, lapack_logical* wantz, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* z, lapack_int* ldz, lapack_int* ifst, - lapack_int* ilst, lapack_int *info ); -void LAPACK_stgsen( lapack_int* ijob, lapack_logical* wantq, - lapack_logical* wantz, const lapack_logical* select, - lapack_int* n, float* a, lapack_int* lda, float* b, - lapack_int* ldb, float* alphar, float* alphai, float* beta, - float* q, lapack_int* ldq, float* z, lapack_int* ldz, - lapack_int* m, float* pl, float* pr, float* dif, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_dtgsen( lapack_int* ijob, lapack_logical* wantq, - lapack_logical* wantz, const lapack_logical* select, - lapack_int* n, double* a, lapack_int* lda, double* b, - lapack_int* ldb, double* alphar, double* alphai, - double* beta, double* q, lapack_int* ldq, double* z, - lapack_int* ldz, lapack_int* m, double* pl, double* pr, - double* dif, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_ctgsen( lapack_int* ijob, lapack_logical* wantq, - lapack_logical* wantz, const lapack_logical* select, - lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* z, lapack_int* ldz, lapack_int* m, - float* pl, float* pr, float* dif, - lapack_complex_float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_ztgsen( lapack_int* ijob, lapack_logical* wantq, - lapack_logical* wantz, const lapack_logical* select, - lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* z, lapack_int* ldz, lapack_int* m, - double* pl, double* pr, double* dif, - lapack_complex_double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_stgsyl( char* trans, lapack_int* ijob, lapack_int* m, lapack_int* n, - const float* a, lapack_int* lda, const float* b, - lapack_int* ldb, float* c, lapack_int* ldc, const float* d, - lapack_int* ldd, const float* e, lapack_int* lde, float* f, - lapack_int* ldf, float* scale, float* dif, float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_dtgsyl( char* trans, lapack_int* ijob, lapack_int* m, lapack_int* n, - const double* a, lapack_int* lda, const double* b, - lapack_int* ldb, double* c, lapack_int* ldc, - const double* d, lapack_int* ldd, const double* e, - lapack_int* lde, double* f, lapack_int* ldf, double* scale, - double* dif, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ctgsyl( char* trans, lapack_int* ijob, lapack_int* m, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, - const lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* c, lapack_int* ldc, - const lapack_complex_float* d, lapack_int* ldd, - const lapack_complex_float* e, lapack_int* lde, - lapack_complex_float* f, lapack_int* ldf, float* scale, - float* dif, lapack_complex_float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ztgsyl( char* trans, lapack_int* ijob, lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, - const lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* c, lapack_int* ldc, - const lapack_complex_double* d, lapack_int* ldd, - const lapack_complex_double* e, lapack_int* lde, - lapack_complex_double* f, lapack_int* ldf, double* scale, - double* dif, lapack_complex_double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_stgsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const float* a, lapack_int* lda, - const float* b, lapack_int* ldb, const float* vl, - lapack_int* ldvl, const float* vr, lapack_int* ldvr, - float* s, float* dif, lapack_int* mm, lapack_int* m, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dtgsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const double* a, lapack_int* lda, - const double* b, lapack_int* ldb, const double* vl, - lapack_int* ldvl, const double* vr, lapack_int* ldvr, - double* s, double* dif, lapack_int* mm, lapack_int* m, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int *info ); -void LAPACK_ctgsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, const lapack_complex_float* b, - lapack_int* ldb, const lapack_complex_float* vl, - lapack_int* ldvl, const lapack_complex_float* vr, - lapack_int* ldvr, float* s, float* dif, lapack_int* mm, - lapack_int* m, lapack_complex_float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_ztgsna( char* job, char* howmny, const lapack_logical* select, - lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, const lapack_complex_double* b, - lapack_int* ldb, const lapack_complex_double* vl, - lapack_int* ldvl, const lapack_complex_double* vr, - lapack_int* ldvr, double* s, double* dif, lapack_int* mm, - lapack_int* m, lapack_complex_double* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_sggsvp( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, float* a, lapack_int* lda, - float* b, lapack_int* ldb, float* tola, float* tolb, - lapack_int* k, lapack_int* l, float* u, lapack_int* ldu, - float* v, lapack_int* ldv, float* q, lapack_int* ldq, - lapack_int* iwork, float* tau, float* work, - lapack_int *info ); -void LAPACK_dggsvp( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, double* a, lapack_int* lda, - double* b, lapack_int* ldb, double* tola, double* tolb, - lapack_int* k, lapack_int* l, double* u, lapack_int* ldu, - double* v, lapack_int* ldv, double* q, lapack_int* ldq, - lapack_int* iwork, double* tau, double* work, - lapack_int *info ); -void LAPACK_cggsvp( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* b, lapack_int* ldb, - float* tola, float* tolb, lapack_int* k, lapack_int* l, - lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* v, lapack_int* ldv, - lapack_complex_float* q, lapack_int* ldq, lapack_int* iwork, - float* rwork, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zggsvp( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* b, lapack_int* ldb, - double* tola, double* tolb, lapack_int* k, lapack_int* l, - lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* v, lapack_int* ldv, - lapack_complex_double* q, lapack_int* ldq, - lapack_int* iwork, double* rwork, - lapack_complex_double* tau, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_stgsja( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_int* k, lapack_int* l, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* tola, float* tolb, float* alpha, float* beta, - float* u, lapack_int* ldu, float* v, lapack_int* ldv, - float* q, lapack_int* ldq, float* work, lapack_int* ncycle, - lapack_int *info ); -void LAPACK_dtgsja( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_int* k, lapack_int* l, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* tola, double* tolb, double* alpha, double* beta, - double* u, lapack_int* ldu, double* v, lapack_int* ldv, - double* q, lapack_int* ldq, double* work, - lapack_int* ncycle, lapack_int *info ); -void LAPACK_ctgsja( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_int* k, lapack_int* l, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* tola, - float* tolb, float* alpha, float* beta, - lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* v, lapack_int* ldv, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* work, lapack_int* ncycle, - lapack_int *info ); -void LAPACK_ztgsja( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* p, lapack_int* n, lapack_int* k, lapack_int* l, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* tola, - double* tolb, double* alpha, double* beta, - lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* v, lapack_int* ldv, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* work, lapack_int* ncycle, - lapack_int *info ); -void LAPACK_sgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_cgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_sgelsy( lapack_int* m, lapack_int* n, lapack_int* nrhs, float* a, - lapack_int* lda, float* b, lapack_int* ldb, - lapack_int* jpvt, float* rcond, lapack_int* rank, - float* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_dgelsy( lapack_int* m, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, double* b, lapack_int* ldb, - lapack_int* jpvt, double* rcond, lapack_int* rank, - double* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_cgelsy( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, lapack_int* jpvt, - float* rcond, lapack_int* rank, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int *info ); -void LAPACK_zgelsy( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, lapack_int* jpvt, - double* rcond, lapack_int* rank, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_sgelss( lapack_int* m, lapack_int* n, lapack_int* nrhs, float* a, - lapack_int* lda, float* b, lapack_int* ldb, float* s, - float* rcond, lapack_int* rank, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dgelss( lapack_int* m, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* s, - double* rcond, lapack_int* rank, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cgelss( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* s, - float* rcond, lapack_int* rank, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int *info ); -void LAPACK_zgelss( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* s, - double* rcond, lapack_int* rank, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_sgelsd( lapack_int* m, lapack_int* n, lapack_int* nrhs, float* a, - lapack_int* lda, float* b, lapack_int* ldb, float* s, - float* rcond, lapack_int* rank, float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_dgelsd( lapack_int* m, lapack_int* n, lapack_int* nrhs, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* s, - double* rcond, lapack_int* rank, double* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_cgelsd( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* s, - float* rcond, lapack_int* rank, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int* iwork, - lapack_int *info ); -void LAPACK_zgelsd( lapack_int* m, lapack_int* n, lapack_int* nrhs, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* s, - double* rcond, lapack_int* rank, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_sgglse( lapack_int* m, lapack_int* n, lapack_int* p, float* a, - lapack_int* lda, float* b, lapack_int* ldb, float* c, - float* d, float* x, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgglse( lapack_int* m, lapack_int* n, lapack_int* p, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* c, - double* d, double* x, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgglse( lapack_int* m, lapack_int* n, lapack_int* p, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* c, lapack_complex_float* d, - lapack_complex_float* x, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zgglse( lapack_int* m, lapack_int* n, lapack_int* p, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* c, lapack_complex_double* d, - lapack_complex_double* x, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sggglm( lapack_int* n, lapack_int* m, lapack_int* p, float* a, - lapack_int* lda, float* b, lapack_int* ldb, float* d, - float* x, float* y, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dggglm( lapack_int* n, lapack_int* m, lapack_int* p, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* d, - double* x, double* y, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cggglm( lapack_int* n, lapack_int* m, lapack_int* p, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* d, lapack_complex_float* x, - lapack_complex_float* y, lapack_complex_float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_zggglm( lapack_int* n, lapack_int* m, lapack_int* p, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* d, lapack_complex_double* x, - lapack_complex_double* y, lapack_complex_double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_ssyev( char* jobz, char* uplo, lapack_int* n, float* a, - lapack_int* lda, float* w, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dsyev( char* jobz, char* uplo, lapack_int* n, double* a, - lapack_int* lda, double* w, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cheev( char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* w, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zheev( char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* w, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_ssyevd( char* jobz, char* uplo, lapack_int* n, float* a, - lapack_int* lda, float* w, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dsyevd( char* jobz, char* uplo, lapack_int* n, double* a, - lapack_int* lda, double* w, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_cheevd( char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* w, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zheevd( char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* w, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_ssyevx( char* jobz, char* range, char* uplo, lapack_int* n, - float* a, lapack_int* lda, float* vl, float* vu, - lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, float* z, lapack_int* ldz, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_dsyevx( char* jobz, char* range, char* uplo, lapack_int* n, - double* a, lapack_int* lda, double* vl, double* vu, - lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, double* z, lapack_int* ldz, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_cheevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* vl, - float* vu, lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_zheevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_ssyevr( char* jobz, char* range, char* uplo, lapack_int* n, - float* a, lapack_int* lda, float* vl, float* vu, - lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, float* z, lapack_int* ldz, - lapack_int* isuppz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dsyevr( char* jobz, char* range, char* uplo, lapack_int* n, - double* a, lapack_int* lda, double* vl, double* vu, - lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, double* z, lapack_int* ldz, - lapack_int* isuppz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_cheevr( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* vl, - float* vu, lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_int* isuppz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zheevr( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_int* isuppz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_sspev( char* jobz, char* uplo, lapack_int* n, float* ap, float* w, - float* z, lapack_int* ldz, float* work, lapack_int *info ); -void LAPACK_dspev( char* jobz, char* uplo, lapack_int* n, double* ap, double* w, - double* z, lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_chpev( char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* ap, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, float* rwork, - lapack_int *info ); -void LAPACK_zhpev( char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* ap, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sspevd( char* jobz, char* uplo, lapack_int* n, float* ap, float* w, - float* z, lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dspevd( char* jobz, char* uplo, lapack_int* n, double* ap, - double* w, double* z, lapack_int* ldz, double* work, - lapack_int* lwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_chpevd( char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* ap, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int* lrwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_zhpevd( char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* ap, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_sspevx( char* jobz, char* range, char* uplo, lapack_int* n, - float* ap, float* vl, float* vu, lapack_int* il, - lapack_int* iu, float* abstol, lapack_int* m, float* w, - float* z, lapack_int* ldz, float* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_dspevx( char* jobz, char* range, char* uplo, lapack_int* n, - double* ap, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, double* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_chpevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_float* ap, float* vl, float* vu, - lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_zhpevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_complex_double* ap, double* vl, double* vu, - lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_ssbev( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - float* ab, lapack_int* ldab, float* w, float* z, - lapack_int* ldz, float* work, lapack_int *info ); -void LAPACK_dsbev( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - double* ab, lapack_int* ldab, double* w, double* z, - lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_chbev( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_float* ab, lapack_int* ldab, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, float* rwork, lapack_int *info ); -void LAPACK_zhbev( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_double* ab, lapack_int* ldab, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_ssbevd( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - float* ab, lapack_int* ldab, float* w, float* z, - lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dsbevd( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - double* ab, lapack_int* ldab, double* w, double* z, - lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_chbevd( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_float* ab, lapack_int* ldab, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zhbevd( char* jobz, char* uplo, lapack_int* n, lapack_int* kd, - lapack_complex_double* ab, lapack_int* ldab, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_ssbevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* kd, float* ab, lapack_int* ldab, float* q, - lapack_int* ldq, float* vl, float* vu, lapack_int* il, - lapack_int* iu, float* abstol, lapack_int* m, float* w, - float* z, lapack_int* ldz, float* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_dsbevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* kd, double* ab, lapack_int* ldab, double* q, - lapack_int* ldq, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, double* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_chbevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* kd, lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* q, lapack_int* ldq, float* vl, - float* vu, lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_zhbevx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* kd, lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* q, lapack_int* ldq, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_sstev( char* jobz, lapack_int* n, float* d, float* e, float* z, - lapack_int* ldz, float* work, lapack_int *info ); -void LAPACK_dstev( char* jobz, lapack_int* n, double* d, double* e, double* z, - lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_sstevd( char* jobz, lapack_int* n, float* d, float* e, float* z, - lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_dstevd( char* jobz, lapack_int* n, double* d, double* e, double* z, - lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_sstevx( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, float* z, - lapack_int* ldz, float* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_dstevx( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, double* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_sstevr( char* jobz, char* range, lapack_int* n, float* d, float* e, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, float* z, - lapack_int* ldz, lapack_int* isuppz, float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_dstevr( char* jobz, char* range, lapack_int* n, double* d, - double* e, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, lapack_int* isuppz, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_sgees( char* jobvs, char* sort, LAPACK_S_SELECT2 select, - lapack_int* n, float* a, lapack_int* lda, lapack_int* sdim, - float* wr, float* wi, float* vs, lapack_int* ldvs, - float* work, lapack_int* lwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_dgees( char* jobvs, char* sort, LAPACK_D_SELECT2 select, - lapack_int* n, double* a, lapack_int* lda, lapack_int* sdim, - double* wr, double* wi, double* vs, lapack_int* ldvs, - double* work, lapack_int* lwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_cgees( char* jobvs, char* sort, LAPACK_C_SELECT1 select, - lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_int* sdim, lapack_complex_float* w, - lapack_complex_float* vs, lapack_int* ldvs, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_logical* bwork, lapack_int *info ); -void LAPACK_zgees( char* jobvs, char* sort, LAPACK_Z_SELECT1 select, - lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_int* sdim, lapack_complex_double* w, - lapack_complex_double* vs, lapack_int* ldvs, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_logical* bwork, lapack_int *info ); -void LAPACK_sgeesx( char* jobvs, char* sort, LAPACK_S_SELECT2 select, - char* sense, lapack_int* n, float* a, lapack_int* lda, - lapack_int* sdim, float* wr, float* wi, float* vs, - lapack_int* ldvs, float* rconde, float* rcondv, float* work, - lapack_int* lwork, lapack_int* iwork, lapack_int* liwork, - lapack_logical* bwork, lapack_int *info ); -void LAPACK_dgeesx( char* jobvs, char* sort, LAPACK_D_SELECT2 select, - char* sense, lapack_int* n, double* a, lapack_int* lda, - lapack_int* sdim, double* wr, double* wi, double* vs, - lapack_int* ldvs, double* rconde, double* rcondv, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_cgeesx( char* jobvs, char* sort, LAPACK_C_SELECT1 select, - char* sense, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* sdim, lapack_complex_float* w, - lapack_complex_float* vs, lapack_int* ldvs, float* rconde, - float* rcondv, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_zgeesx( char* jobvs, char* sort, LAPACK_Z_SELECT1 select, - char* sense, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* sdim, lapack_complex_double* w, - lapack_complex_double* vs, lapack_int* ldvs, double* rconde, - double* rcondv, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_sgeev( char* jobvl, char* jobvr, lapack_int* n, float* a, - lapack_int* lda, float* wr, float* wi, float* vl, - lapack_int* ldvl, float* vr, lapack_int* ldvr, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dgeev( char* jobvl, char* jobvr, lapack_int* n, double* a, - lapack_int* lda, double* wr, double* wi, double* vl, - lapack_int* ldvl, double* vr, lapack_int* ldvr, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cgeev( char* jobvl, char* jobvr, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int* ldvl, lapack_complex_float* vr, lapack_int* ldvr, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zgeev( char* jobvl, char* jobvr, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* w, lapack_complex_double* vl, - lapack_int* ldvl, lapack_complex_double* vr, - lapack_int* ldvr, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int *info ); -void LAPACK_sgeevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, float* a, lapack_int* lda, float* wr, - float* wi, float* vl, lapack_int* ldvl, float* vr, - lapack_int* ldvr, lapack_int* ilo, lapack_int* ihi, - float* scale, float* abnrm, float* rconde, float* rcondv, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int *info ); -void LAPACK_dgeevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, double* a, lapack_int* lda, double* wr, - double* wi, double* vl, lapack_int* ldvl, double* vr, - lapack_int* ldvr, lapack_int* ilo, lapack_int* ihi, - double* scale, double* abnrm, double* rconde, - double* rcondv, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_cgeevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* w, lapack_complex_float* vl, - lapack_int* ldvl, lapack_complex_float* vr, - lapack_int* ldvr, lapack_int* ilo, lapack_int* ihi, - float* scale, float* abnrm, float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zgeevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* w, lapack_complex_double* vl, - lapack_int* ldvl, lapack_complex_double* vr, - lapack_int* ldvr, lapack_int* ilo, lapack_int* ihi, - double* scale, double* abnrm, double* rconde, - double* rcondv, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int *info ); -void LAPACK_sgesvd( char* jobu, char* jobvt, lapack_int* m, lapack_int* n, - float* a, lapack_int* lda, float* s, float* u, - lapack_int* ldu, float* vt, lapack_int* ldvt, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_dgesvd( char* jobu, char* jobvt, lapack_int* m, lapack_int* n, - double* a, lapack_int* lda, double* s, double* u, - lapack_int* ldu, double* vt, lapack_int* ldvt, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cgesvd( char* jobu, char* jobvt, lapack_int* m, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* s, - lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* vt, lapack_int* ldvt, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zgesvd( char* jobu, char* jobvt, lapack_int* m, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* s, - lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* vt, lapack_int* ldvt, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_sgesdd( char* jobz, lapack_int* m, lapack_int* n, float* a, - lapack_int* lda, float* s, float* u, lapack_int* ldu, - float* vt, lapack_int* ldvt, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dgesdd( char* jobz, lapack_int* m, lapack_int* n, double* a, - lapack_int* lda, double* s, double* u, lapack_int* ldu, - double* vt, lapack_int* ldvt, double* work, - lapack_int* lwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_cgesdd( char* jobz, lapack_int* m, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, float* s, - lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* vt, lapack_int* ldvt, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_zgesdd( char* jobz, lapack_int* m, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, double* s, - lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* vt, lapack_int* ldvt, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* iwork, lapack_int *info ); -void LAPACK_dgejsv( char* joba, char* jobu, char* jobv, char* jobr, char* jobt, - char* jobp, lapack_int* m, lapack_int* n, double* a, - lapack_int* lda, double* sva, double* u, lapack_int* ldu, - double* v, lapack_int* ldv, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_sgejsv( char* joba, char* jobu, char* jobv, char* jobr, char* jobt, - char* jobp, lapack_int* m, lapack_int* n, float* a, - lapack_int* lda, float* sva, float* u, lapack_int* ldu, - float* v, lapack_int* ldv, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_dgesvj( char* joba, char* jobu, char* jobv, lapack_int* m, - lapack_int* n, double* a, lapack_int* lda, double* sva, - lapack_int* mv, double* v, lapack_int* ldv, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sgesvj( char* joba, char* jobu, char* jobv, lapack_int* m, - lapack_int* n, float* a, lapack_int* lda, float* sva, - lapack_int* mv, float* v, lapack_int* ldv, float* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_sggsvd( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* n, lapack_int* p, lapack_int* k, lapack_int* l, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* alpha, float* beta, float* u, lapack_int* ldu, - float* v, lapack_int* ldv, float* q, lapack_int* ldq, - float* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_dggsvd( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* n, lapack_int* p, lapack_int* k, lapack_int* l, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* alpha, double* beta, double* u, lapack_int* ldu, - double* v, lapack_int* ldv, double* q, lapack_int* ldq, - double* work, lapack_int* iwork, lapack_int *info ); -void LAPACK_cggsvd( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* n, lapack_int* p, lapack_int* k, lapack_int* l, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* alpha, - float* beta, lapack_complex_float* u, lapack_int* ldu, - lapack_complex_float* v, lapack_int* ldv, - lapack_complex_float* q, lapack_int* ldq, - lapack_complex_float* work, float* rwork, lapack_int* iwork, - lapack_int *info ); -void LAPACK_zggsvd( char* jobu, char* jobv, char* jobq, lapack_int* m, - lapack_int* n, lapack_int* p, lapack_int* k, lapack_int* l, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* alpha, - double* beta, lapack_complex_double* u, lapack_int* ldu, - lapack_complex_double* v, lapack_int* ldv, - lapack_complex_double* q, lapack_int* ldq, - lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int *info ); -void LAPACK_ssygv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* w, float* work, lapack_int* lwork, lapack_int *info ); -void LAPACK_dsygv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* w, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_chegv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* w, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zhegv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* w, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_ssygvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - float* w, float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_dsygvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* w, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_chegvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* w, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zhegvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* w, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_ssygvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, float* a, lapack_int* lda, float* b, - lapack_int* ldb, float* vl, float* vu, lapack_int* il, - lapack_int* iu, float* abstol, lapack_int* m, float* w, - float* z, lapack_int* ldz, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_dsygvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, double* a, lapack_int* lda, double* b, - lapack_int* ldb, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_chegvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, float* vl, - float* vu, lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_zhegvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_sspgv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - float* ap, float* bp, float* w, float* z, lapack_int* ldz, - float* work, lapack_int *info ); -void LAPACK_dspgv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - double* ap, double* bp, double* w, double* z, - lapack_int* ldz, double* work, lapack_int *info ); -void LAPACK_chpgv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* ap, lapack_complex_float* bp, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, float* rwork, lapack_int *info ); -void LAPACK_zhpgv( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* ap, lapack_complex_double* bp, - double* w, lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_sspgvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - float* ap, float* bp, float* w, float* z, lapack_int* ldz, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_dspgvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - double* ap, double* bp, double* w, double* z, - lapack_int* ldz, double* work, lapack_int* lwork, - lapack_int* iwork, lapack_int* liwork, lapack_int *info ); -void LAPACK_chpgvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_float* ap, lapack_complex_float* bp, - float* w, lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zhpgvd( lapack_int* itype, char* jobz, char* uplo, lapack_int* n, - lapack_complex_double* ap, lapack_complex_double* bp, - double* w, lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_sspgvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, float* ap, float* bp, float* vl, float* vu, - lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, float* z, lapack_int* ldz, - float* work, lapack_int* iwork, lapack_int* ifail, - lapack_int *info ); -void LAPACK_dspgvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, double* ap, double* bp, double* vl, - double* vu, lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, double* z, lapack_int* ldz, - double* work, lapack_int* iwork, lapack_int* ifail, - lapack_int *info ); -void LAPACK_chpgvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, lapack_complex_float* ap, - lapack_complex_float* bp, float* vl, float* vu, - lapack_int* il, lapack_int* iu, float* abstol, - lapack_int* m, float* w, lapack_complex_float* z, - lapack_int* ldz, lapack_complex_float* work, float* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_zhpgvx( lapack_int* itype, char* jobz, char* range, char* uplo, - lapack_int* n, lapack_complex_double* ap, - lapack_complex_double* bp, double* vl, double* vu, - lapack_int* il, lapack_int* iu, double* abstol, - lapack_int* m, double* w, lapack_complex_double* z, - lapack_int* ldz, lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_ssbgv( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, float* ab, lapack_int* ldab, float* bb, - lapack_int* ldbb, float* w, float* z, lapack_int* ldz, - float* work, lapack_int *info ); -void LAPACK_dsbgv( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, double* ab, lapack_int* ldab, double* bb, - lapack_int* ldbb, double* w, double* z, lapack_int* ldz, - double* work, lapack_int *info ); -void LAPACK_chbgv( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* bb, lapack_int* ldbb, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, float* rwork, lapack_int *info ); -void LAPACK_zhbgv( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* bb, lapack_int* ldbb, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, double* rwork, - lapack_int *info ); -void LAPACK_ssbgvd( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, float* ab, lapack_int* ldab, float* bb, - lapack_int* ldbb, float* w, float* z, lapack_int* ldz, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_dsbgvd( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, double* ab, lapack_int* ldab, double* bb, - lapack_int* ldbb, double* w, double* z, lapack_int* ldz, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_chbgvd( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_float* ab, lapack_int* ldab, - lapack_complex_float* bb, lapack_int* ldbb, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* lrwork, lapack_int* iwork, lapack_int* liwork, - lapack_int *info ); -void LAPACK_zhbgvd( char* jobz, char* uplo, lapack_int* n, lapack_int* ka, - lapack_int* kb, lapack_complex_double* ab, lapack_int* ldab, - lapack_complex_double* bb, lapack_int* ldbb, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, lapack_int* iwork, - lapack_int* liwork, lapack_int *info ); -void LAPACK_ssbgvx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* ka, lapack_int* kb, float* ab, lapack_int* ldab, - float* bb, lapack_int* ldbb, float* q, lapack_int* ldq, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, float* z, - lapack_int* ldz, float* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_dsbgvx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* ka, lapack_int* kb, double* ab, - lapack_int* ldab, double* bb, lapack_int* ldbb, double* q, - lapack_int* ldq, double* vl, double* vu, lapack_int* il, - lapack_int* iu, double* abstol, lapack_int* m, double* w, - double* z, lapack_int* ldz, double* work, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_chbgvx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* ka, lapack_int* kb, lapack_complex_float* ab, - lapack_int* ldab, lapack_complex_float* bb, - lapack_int* ldbb, lapack_complex_float* q, lapack_int* ldq, - float* vl, float* vu, lapack_int* il, lapack_int* iu, - float* abstol, lapack_int* m, float* w, - lapack_complex_float* z, lapack_int* ldz, - lapack_complex_float* work, float* rwork, lapack_int* iwork, - lapack_int* ifail, lapack_int *info ); -void LAPACK_zhbgvx( char* jobz, char* range, char* uplo, lapack_int* n, - lapack_int* ka, lapack_int* kb, lapack_complex_double* ab, - lapack_int* ldab, lapack_complex_double* bb, - lapack_int* ldbb, lapack_complex_double* q, lapack_int* ldq, - double* vl, double* vu, lapack_int* il, lapack_int* iu, - double* abstol, lapack_int* m, double* w, - lapack_complex_double* z, lapack_int* ldz, - lapack_complex_double* work, double* rwork, - lapack_int* iwork, lapack_int* ifail, lapack_int *info ); -void LAPACK_sgges( char* jobvsl, char* jobvsr, char* sort, - LAPACK_S_SELECT3 selctg, lapack_int* n, float* a, - lapack_int* lda, float* b, lapack_int* ldb, lapack_int* sdim, - float* alphar, float* alphai, float* beta, float* vsl, - lapack_int* ldvsl, float* vsr, lapack_int* ldvsr, - float* work, lapack_int* lwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_dgges( char* jobvsl, char* jobvsr, char* sort, - LAPACK_D_SELECT3 selctg, lapack_int* n, double* a, - lapack_int* lda, double* b, lapack_int* ldb, - lapack_int* sdim, double* alphar, double* alphai, - double* beta, double* vsl, lapack_int* ldvsl, double* vsr, - lapack_int* ldvsr, double* work, lapack_int* lwork, - lapack_logical* bwork, lapack_int *info ); -void LAPACK_cgges( char* jobvsl, char* jobvsr, char* sort, - LAPACK_C_SELECT2 selctg, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, lapack_int* sdim, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* vsl, lapack_int* ldvsl, - lapack_complex_float* vsr, lapack_int* ldvsr, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_logical* bwork, lapack_int *info ); -void LAPACK_zgges( char* jobvsl, char* jobvsr, char* sort, - LAPACK_Z_SELECT2 selctg, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, lapack_int* sdim, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int* ldvsl, - lapack_complex_double* vsr, lapack_int* ldvsr, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_logical* bwork, lapack_int *info ); -void LAPACK_sggesx( char* jobvsl, char* jobvsr, char* sort, - LAPACK_S_SELECT3 selctg, char* sense, lapack_int* n, - float* a, lapack_int* lda, float* b, lapack_int* ldb, - lapack_int* sdim, float* alphar, float* alphai, float* beta, - float* vsl, lapack_int* ldvsl, float* vsr, - lapack_int* ldvsr, float* rconde, float* rcondv, - float* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_dggesx( char* jobvsl, char* jobvsr, char* sort, - LAPACK_D_SELECT3 selctg, char* sense, lapack_int* n, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - lapack_int* sdim, double* alphar, double* alphai, - double* beta, double* vsl, lapack_int* ldvsl, double* vsr, - lapack_int* ldvsr, double* rconde, double* rcondv, - double* work, lapack_int* lwork, lapack_int* iwork, - lapack_int* liwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_cggesx( char* jobvsl, char* jobvsr, char* sort, - LAPACK_C_SELECT2 selctg, char* sense, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, lapack_int* sdim, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* vsl, lapack_int* ldvsl, - lapack_complex_float* vsr, lapack_int* ldvsr, float* rconde, - float* rcondv, lapack_complex_float* work, - lapack_int* lwork, float* rwork, lapack_int* iwork, - lapack_int* liwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_zggesx( char* jobvsl, char* jobvsr, char* sort, - LAPACK_Z_SELECT2 selctg, char* sense, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, lapack_int* sdim, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* vsl, lapack_int* ldvsl, - lapack_complex_double* vsr, lapack_int* ldvsr, - double* rconde, double* rcondv, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int* iwork, - lapack_int* liwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_sggev( char* jobvl, char* jobvr, lapack_int* n, float* a, - lapack_int* lda, float* b, lapack_int* ldb, float* alphar, - float* alphai, float* beta, float* vl, lapack_int* ldvl, - float* vr, lapack_int* ldvr, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dggev( char* jobvl, char* jobvr, lapack_int* n, double* a, - lapack_int* lda, double* b, lapack_int* ldb, double* alphar, - double* alphai, double* beta, double* vl, lapack_int* ldvl, - double* vr, lapack_int* ldvr, double* work, - lapack_int* lwork, lapack_int *info ); -void LAPACK_cggev( char* jobvl, char* jobvr, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* vl, lapack_int* ldvl, - lapack_complex_float* vr, lapack_int* ldvr, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int *info ); -void LAPACK_zggev( char* jobvl, char* jobvr, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int* ldvl, - lapack_complex_double* vr, lapack_int* ldvr, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int *info ); -void LAPACK_sggevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, float* a, lapack_int* lda, float* b, - lapack_int* ldb, float* alphar, float* alphai, float* beta, - float* vl, lapack_int* ldvl, float* vr, lapack_int* ldvr, - lapack_int* ilo, lapack_int* ihi, float* lscale, - float* rscale, float* abnrm, float* bbnrm, float* rconde, - float* rcondv, float* work, lapack_int* lwork, - lapack_int* iwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_dggevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, double* a, lapack_int* lda, double* b, - lapack_int* ldb, double* alphar, double* alphai, - double* beta, double* vl, lapack_int* ldvl, double* vr, - lapack_int* ldvr, lapack_int* ilo, lapack_int* ihi, - double* lscale, double* rscale, double* abnrm, - double* bbnrm, double* rconde, double* rcondv, double* work, - lapack_int* lwork, lapack_int* iwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_cggevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* vl, lapack_int* ldvl, - lapack_complex_float* vr, lapack_int* ldvr, lapack_int* ilo, - lapack_int* ihi, float* lscale, float* rscale, float* abnrm, - float* bbnrm, float* rconde, float* rcondv, - lapack_complex_float* work, lapack_int* lwork, float* rwork, - lapack_int* iwork, lapack_logical* bwork, - lapack_int *info ); -void LAPACK_zggevx( char* balanc, char* jobvl, char* jobvr, char* sense, - lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* vl, lapack_int* ldvl, - lapack_complex_double* vr, lapack_int* ldvr, - lapack_int* ilo, lapack_int* ihi, double* lscale, - double* rscale, double* abnrm, double* bbnrm, - double* rconde, double* rcondv, lapack_complex_double* work, - lapack_int* lwork, double* rwork, lapack_int* iwork, - lapack_logical* bwork, lapack_int *info ); -void LAPACK_dsfrk( char* transr, char* uplo, char* trans, lapack_int* n, - lapack_int* k, double* alpha, const double* a, - lapack_int* lda, double* beta, double* c ); -void LAPACK_ssfrk( char* transr, char* uplo, char* trans, lapack_int* n, - lapack_int* k, float* alpha, const float* a, lapack_int* lda, - float* beta, float* c ); -void LAPACK_zhfrk( char* transr, char* uplo, char* trans, lapack_int* n, - lapack_int* k, double* alpha, const lapack_complex_double* a, - lapack_int* lda, double* beta, lapack_complex_double* c ); -void LAPACK_chfrk( char* transr, char* uplo, char* trans, lapack_int* n, - lapack_int* k, float* alpha, const lapack_complex_float* a, - lapack_int* lda, float* beta, lapack_complex_float* c ); -void LAPACK_dtfsm( char* transr, char* side, char* uplo, char* trans, - char* diag, lapack_int* m, lapack_int* n, double* alpha, - const double* a, double* b, lapack_int* ldb ); -void LAPACK_stfsm( char* transr, char* side, char* uplo, char* trans, - char* diag, lapack_int* m, lapack_int* n, float* alpha, - const float* a, float* b, lapack_int* ldb ); -void LAPACK_ztfsm( char* transr, char* side, char* uplo, char* trans, - char* diag, lapack_int* m, lapack_int* n, - lapack_complex_double* alpha, const lapack_complex_double* a, - lapack_complex_double* b, lapack_int* ldb ); -void LAPACK_ctfsm( char* transr, char* side, char* uplo, char* trans, - char* diag, lapack_int* m, lapack_int* n, - lapack_complex_float* alpha, const lapack_complex_float* a, - lapack_complex_float* b, lapack_int* ldb ); -void LAPACK_dtfttp( char* transr, char* uplo, lapack_int* n, const double* arf, - double* ap, lapack_int *info ); -void LAPACK_stfttp( char* transr, char* uplo, lapack_int* n, const float* arf, - float* ap, lapack_int *info ); -void LAPACK_ztfttp( char* transr, char* uplo, lapack_int* n, - const lapack_complex_double* arf, lapack_complex_double* ap, - lapack_int *info ); -void LAPACK_ctfttp( char* transr, char* uplo, lapack_int* n, - const lapack_complex_float* arf, lapack_complex_float* ap, - lapack_int *info ); -void LAPACK_dtfttr( char* transr, char* uplo, lapack_int* n, const double* arf, - double* a, lapack_int* lda, lapack_int *info ); -void LAPACK_stfttr( char* transr, char* uplo, lapack_int* n, const float* arf, - float* a, lapack_int* lda, lapack_int *info ); -void LAPACK_ztfttr( char* transr, char* uplo, lapack_int* n, - const lapack_complex_double* arf, lapack_complex_double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_ctfttr( char* transr, char* uplo, lapack_int* n, - const lapack_complex_float* arf, lapack_complex_float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_dtpttf( char* transr, char* uplo, lapack_int* n, const double* ap, - double* arf, lapack_int *info ); -void LAPACK_stpttf( char* transr, char* uplo, lapack_int* n, const float* ap, - float* arf, lapack_int *info ); -void LAPACK_ztpttf( char* transr, char* uplo, lapack_int* n, - const lapack_complex_double* ap, lapack_complex_double* arf, - lapack_int *info ); -void LAPACK_ctpttf( char* transr, char* uplo, lapack_int* n, - const lapack_complex_float* ap, lapack_complex_float* arf, - lapack_int *info ); -void LAPACK_dtpttr( char* uplo, lapack_int* n, const double* ap, double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_stpttr( char* uplo, lapack_int* n, const float* ap, float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_ztpttr( char* uplo, lapack_int* n, const lapack_complex_double* ap, - lapack_complex_double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_ctpttr( char* uplo, lapack_int* n, const lapack_complex_float* ap, - lapack_complex_float* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dtrttf( char* transr, char* uplo, lapack_int* n, const double* a, - lapack_int* lda, double* arf, lapack_int *info ); -void LAPACK_strttf( char* transr, char* uplo, lapack_int* n, const float* a, - lapack_int* lda, float* arf, lapack_int *info ); -void LAPACK_ztrttf( char* transr, char* uplo, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* arf, lapack_int *info ); -void LAPACK_ctrttf( char* transr, char* uplo, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* arf, lapack_int *info ); -void LAPACK_dtrttp( char* uplo, lapack_int* n, const double* a, lapack_int* lda, - double* ap, lapack_int *info ); -void LAPACK_strttp( char* uplo, lapack_int* n, const float* a, lapack_int* lda, - float* ap, lapack_int *info ); -void LAPACK_ztrttp( char* uplo, lapack_int* n, const lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* ap, - lapack_int *info ); -void LAPACK_ctrttp( char* uplo, lapack_int* n, const lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* ap, - lapack_int *info ); -void LAPACK_sgeqrfp( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_dgeqrfp( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_cgeqrfp( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_zgeqrfp( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int* lwork, - lapack_int *info ); -void LAPACK_clacgv( lapack_int* n, lapack_complex_float* x, lapack_int* incx ); -void LAPACK_zlacgv( lapack_int* n, lapack_complex_double* x, lapack_int* incx ); -void LAPACK_slarnv( lapack_int* idist, lapack_int* iseed, lapack_int* n, - float* x ); -void LAPACK_dlarnv( lapack_int* idist, lapack_int* iseed, lapack_int* n, - double* x ); -void LAPACK_clarnv( lapack_int* idist, lapack_int* iseed, lapack_int* n, - lapack_complex_float* x ); -void LAPACK_zlarnv( lapack_int* idist, lapack_int* iseed, lapack_int* n, - lapack_complex_double* x ); -void LAPACK_sgeqr2( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int *info ); -void LAPACK_dgeqr2( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int *info ); -void LAPACK_cgeqr2( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zgeqr2( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_slacpy( char* uplo, lapack_int* m, lapack_int* n, const float* a, - lapack_int* lda, float* b, lapack_int* ldb ); -void LAPACK_dlacpy( char* uplo, lapack_int* m, lapack_int* n, const double* a, - lapack_int* lda, double* b, lapack_int* ldb ); -void LAPACK_clacpy( char* uplo, lapack_int* m, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb ); -void LAPACK_zlacpy( char* uplo, lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb ); -void LAPACK_sgetf2( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_dgetf2( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - lapack_int* ipiv, lapack_int *info ); -void LAPACK_cgetf2( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int* ipiv, lapack_int *info ); -void LAPACK_zgetf2( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int* ipiv, lapack_int *info ); -void LAPACK_slaswp( lapack_int* n, float* a, lapack_int* lda, lapack_int* k1, - lapack_int* k2, const lapack_int* ipiv, lapack_int* incx ); -void LAPACK_dlaswp( lapack_int* n, double* a, lapack_int* lda, lapack_int* k1, - lapack_int* k2, const lapack_int* ipiv, lapack_int* incx ); -void LAPACK_claswp( lapack_int* n, lapack_complex_float* a, lapack_int* lda, - lapack_int* k1, lapack_int* k2, const lapack_int* ipiv, - lapack_int* incx ); -void LAPACK_zlaswp( lapack_int* n, lapack_complex_double* a, lapack_int* lda, - lapack_int* k1, lapack_int* k2, const lapack_int* ipiv, - lapack_int* incx ); -float LAPACK_slange( char* norm, lapack_int* m, lapack_int* n, const float* a, - lapack_int* lda, float* work ); -double LAPACK_dlange( char* norm, lapack_int* m, lapack_int* n, const double* a, - lapack_int* lda, double* work ); -float LAPACK_clange( char* norm, lapack_int* m, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, float* work ); -double LAPACK_zlange( char* norm, lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, double* work ); -float LAPACK_clanhe( char* norm, char* uplo, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, float* work ); -double LAPACK_zlanhe( char* norm, char* uplo, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, double* work ); -float LAPACK_slansy( char* norm, char* uplo, lapack_int* n, const float* a, - lapack_int* lda, float* work ); -double LAPACK_dlansy( char* norm, char* uplo, lapack_int* n, const double* a, - lapack_int* lda, double* work ); -float LAPACK_clansy( char* norm, char* uplo, lapack_int* n, - const lapack_complex_float* a, lapack_int* lda, float* work ); -double LAPACK_zlansy( char* norm, char* uplo, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, double* work ); -float LAPACK_slantr( char* norm, char* uplo, char* diag, lapack_int* m, - lapack_int* n, const float* a, lapack_int* lda, float* work ); -double LAPACK_dlantr( char* norm, char* uplo, char* diag, lapack_int* m, - lapack_int* n, const double* a, lapack_int* lda, double* work ); -float LAPACK_clantr( char* norm, char* uplo, char* diag, lapack_int* m, - lapack_int* n, const lapack_complex_float* a, lapack_int* lda, - float* work ); -double LAPACK_zlantr( char* norm, char* uplo, char* diag, lapack_int* m, - lapack_int* n, const lapack_complex_double* a, lapack_int* lda, - double* work ); -float LAPACK_slamch( char* cmach ); -double LAPACK_dlamch( char* cmach ); -void LAPACK_sgelq2( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* tau, float* work, lapack_int *info ); -void LAPACK_dgelq2( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* tau, double* work, lapack_int *info ); -void LAPACK_cgelq2( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* tau, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zgelq2( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* tau, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_slarfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, const float* v, - lapack_int* ldv, const float* t, lapack_int* ldt, float* c, - lapack_int* ldc, float* work, lapack_int* ldwork ); -void LAPACK_dlarfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, - const double* v, lapack_int* ldv, const double* t, - lapack_int* ldt, double* c, lapack_int* ldc, double* work, - lapack_int* ldwork ); -void LAPACK_clarfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, - const lapack_complex_float* v, lapack_int* ldv, - const lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int* ldwork ); -void LAPACK_zlarfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, - const lapack_complex_double* v, lapack_int* ldv, - const lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int* ldwork ); -void LAPACK_slarfg( lapack_int* n, float* alpha, float* x, lapack_int* incx, - float* tau ); -void LAPACK_dlarfg( lapack_int* n, double* alpha, double* x, lapack_int* incx, - double* tau ); -void LAPACK_clarfg( lapack_int* n, lapack_complex_float* alpha, - lapack_complex_float* x, lapack_int* incx, - lapack_complex_float* tau ); -void LAPACK_zlarfg( lapack_int* n, lapack_complex_double* alpha, - lapack_complex_double* x, lapack_int* incx, - lapack_complex_double* tau ); -void LAPACK_slarft( char* direct, char* storev, lapack_int* n, lapack_int* k, - const float* v, lapack_int* ldv, const float* tau, float* t, - lapack_int* ldt ); -void LAPACK_dlarft( char* direct, char* storev, lapack_int* n, lapack_int* k, - const double* v, lapack_int* ldv, const double* tau, - double* t, lapack_int* ldt ); -void LAPACK_clarft( char* direct, char* storev, lapack_int* n, lapack_int* k, - const lapack_complex_float* v, lapack_int* ldv, - const lapack_complex_float* tau, lapack_complex_float* t, - lapack_int* ldt ); -void LAPACK_zlarft( char* direct, char* storev, lapack_int* n, lapack_int* k, - const lapack_complex_double* v, lapack_int* ldv, - const lapack_complex_double* tau, lapack_complex_double* t, - lapack_int* ldt ); -void LAPACK_slarfx( char* side, lapack_int* m, lapack_int* n, const float* v, - float* tau, float* c, lapack_int* ldc, float* work ); -void LAPACK_dlarfx( char* side, lapack_int* m, lapack_int* n, const double* v, - double* tau, double* c, lapack_int* ldc, double* work ); -void LAPACK_clarfx( char* side, lapack_int* m, lapack_int* n, - const lapack_complex_float* v, lapack_complex_float* tau, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work ); -void LAPACK_zlarfx( char* side, lapack_int* m, lapack_int* n, - const lapack_complex_double* v, lapack_complex_double* tau, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work ); -void LAPACK_slatms( lapack_int* m, lapack_int* n, char* dist, lapack_int* iseed, - char* sym, float* d, lapack_int* mode, float* cond, - float* dmax, lapack_int* kl, lapack_int* ku, char* pack, - float* a, lapack_int* lda, float* work, lapack_int *info ); -void LAPACK_dlatms( lapack_int* m, lapack_int* n, char* dist, lapack_int* iseed, - char* sym, double* d, lapack_int* mode, double* cond, - double* dmax, lapack_int* kl, lapack_int* ku, char* pack, - double* a, lapack_int* lda, double* work, - lapack_int *info ); -void LAPACK_clatms( lapack_int* m, lapack_int* n, char* dist, lapack_int* iseed, - char* sym, float* d, lapack_int* mode, float* cond, - float* dmax, lapack_int* kl, lapack_int* ku, char* pack, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zlatms( lapack_int* m, lapack_int* n, char* dist, lapack_int* iseed, - char* sym, double* d, lapack_int* mode, double* cond, - double* dmax, lapack_int* kl, lapack_int* ku, char* pack, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_slag2d( lapack_int* m, lapack_int* n, const float* sa, - lapack_int* ldsa, double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dlag2s( lapack_int* m, lapack_int* n, const double* a, - lapack_int* lda, float* sa, lapack_int* ldsa, - lapack_int *info ); -void LAPACK_clag2z( lapack_int* m, lapack_int* n, - const lapack_complex_float* sa, lapack_int* ldsa, - lapack_complex_double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_zlag2c( lapack_int* m, lapack_int* n, - const lapack_complex_double* a, lapack_int* lda, - lapack_complex_float* sa, lapack_int* ldsa, - lapack_int *info ); -void LAPACK_slauum( char* uplo, lapack_int* n, float* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_dlauum( char* uplo, lapack_int* n, double* a, lapack_int* lda, - lapack_int *info ); -void LAPACK_clauum( char* uplo, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_zlauum( char* uplo, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_int *info ); -void LAPACK_slagge( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const float* d, float* a, lapack_int* lda, - lapack_int* iseed, float* work, lapack_int *info ); -void LAPACK_dlagge( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const double* d, double* a, lapack_int* lda, - lapack_int* iseed, double* work, lapack_int *info ); -void LAPACK_clagge( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const float* d, lapack_complex_float* a, - lapack_int* lda, lapack_int* iseed, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zlagge( lapack_int* m, lapack_int* n, lapack_int* kl, - lapack_int* ku, const double* d, lapack_complex_double* a, - lapack_int* lda, lapack_int* iseed, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_slaset( char* uplo, lapack_int* m, lapack_int* n, float* alpha, - float* beta, float* a, lapack_int* lda ); -void LAPACK_dlaset( char* uplo, lapack_int* m, lapack_int* n, double* alpha, - double* beta, double* a, lapack_int* lda ); -void LAPACK_claset( char* uplo, lapack_int* m, lapack_int* n, - lapack_complex_float* alpha, lapack_complex_float* beta, - lapack_complex_float* a, lapack_int* lda ); -void LAPACK_zlaset( char* uplo, lapack_int* m, lapack_int* n, - lapack_complex_double* alpha, lapack_complex_double* beta, - lapack_complex_double* a, lapack_int* lda ); -void LAPACK_slasrt( char* id, lapack_int* n, float* d, lapack_int *info ); -void LAPACK_dlasrt( char* id, lapack_int* n, double* d, lapack_int *info ); -void LAPACK_claghe( lapack_int* n, lapack_int* k, const float* d, - lapack_complex_float* a, lapack_int* lda, lapack_int* iseed, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zlaghe( lapack_int* n, lapack_int* k, const double* d, - lapack_complex_double* a, lapack_int* lda, - lapack_int* iseed, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_slagsy( lapack_int* n, lapack_int* k, const float* d, float* a, - lapack_int* lda, lapack_int* iseed, float* work, - lapack_int *info ); -void LAPACK_dlagsy( lapack_int* n, lapack_int* k, const double* d, double* a, - lapack_int* lda, lapack_int* iseed, double* work, - lapack_int *info ); -void LAPACK_clagsy( lapack_int* n, lapack_int* k, const float* d, - lapack_complex_float* a, lapack_int* lda, lapack_int* iseed, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zlagsy( lapack_int* n, lapack_int* k, const double* d, - lapack_complex_double* a, lapack_int* lda, - lapack_int* iseed, lapack_complex_double* work, - lapack_int *info ); -void LAPACK_slapmr( lapack_logical* forwrd, lapack_int* m, lapack_int* n, - float* x, lapack_int* ldx, lapack_int* k ); -void LAPACK_dlapmr( lapack_logical* forwrd, lapack_int* m, lapack_int* n, - double* x, lapack_int* ldx, lapack_int* k ); -void LAPACK_clapmr( lapack_logical* forwrd, lapack_int* m, lapack_int* n, - lapack_complex_float* x, lapack_int* ldx, lapack_int* k ); -void LAPACK_zlapmr( lapack_logical* forwrd, lapack_int* m, lapack_int* n, - lapack_complex_double* x, lapack_int* ldx, lapack_int* k ); -float LAPACK_slapy2( float* x, float* y ); -double LAPACK_dlapy2( double* x, double* y ); -float LAPACK_slapy3( float* x, float* y, float* z ); -double LAPACK_dlapy3( double* x, double* y, double* z ); -void LAPACK_slartgp( float* f, float* g, float* cs, float* sn, float* r ); -void LAPACK_dlartgp( double* f, double* g, double* cs, double* sn, double* r ); -void LAPACK_slartgs( float* x, float* y, float* sigma, float* cs, float* sn ); -void LAPACK_dlartgs( double* x, double* y, double* sigma, double* cs, - double* sn ); -// LAPACK 3.3.0 -void LAPACK_cbbcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - lapack_int* m, lapack_int* p, lapack_int* q, - float* theta, float* phi, - lapack_complex_float* u1, lapack_int* ldu1, - lapack_complex_float* u2, lapack_int* ldu2, - lapack_complex_float* v1t, lapack_int* ldv1t, - lapack_complex_float* v2t, lapack_int* ldv2t, - float* b11d, float* b11e, float* b12d, - float* b12e, float* b21d, float* b21e, - float* b22d, float* b22e, float* rwork, - lapack_int* lrwork , lapack_int *info ); -void LAPACK_cheswapr( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* i1, - lapack_int* i2 ); -void LAPACK_chetri2( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_chetri2x( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int* nb , lapack_int *info ); -void LAPACK_chetrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work , lapack_int *info ); -void LAPACK_csyconv( char* uplo, char* way, - lapack_int* n, lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_float* work , lapack_int *info ); -void LAPACK_csyswapr( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* i1, - lapack_int* i2 ); -void LAPACK_csytri2( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_csytri2x( char* uplo, lapack_int* n, - lapack_complex_float* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int* nb , lapack_int *info ); -void LAPACK_csytrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, const lapack_complex_float* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work , lapack_int *info ); -void LAPACK_cunbdb( char* trans, char* signs, - lapack_int* m, lapack_int* p, lapack_int* q, - lapack_complex_float* x11, lapack_int* ldx11, - lapack_complex_float* x12, lapack_int* ldx12, - lapack_complex_float* x21, lapack_int* ldx21, - lapack_complex_float* x22, lapack_int* ldx22, - float* theta, float* phi, - lapack_complex_float* taup1, - lapack_complex_float* taup2, - lapack_complex_float* tauq1, - lapack_complex_float* tauq2, - lapack_complex_float* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_cuncsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - char* signs, lapack_int* m, lapack_int* p, - lapack_int* q, lapack_complex_float* x11, - lapack_int* ldx11, lapack_complex_float* x12, - lapack_int* ldx12, lapack_complex_float* x21, - lapack_int* ldx21, lapack_complex_float* x22, - lapack_int* ldx22, float* theta, - lapack_complex_float* u1, lapack_int* ldu1, - lapack_complex_float* u2, lapack_int* ldu2, - lapack_complex_float* v1t, lapack_int* ldv1t, - lapack_complex_float* v2t, lapack_int* ldv2t, - lapack_complex_float* work, lapack_int* lwork, - float* rwork, lapack_int* lrwork, - lapack_int* iwork , lapack_int *info ); -void LAPACK_dbbcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - lapack_int* m, lapack_int* p, lapack_int* q, - double* theta, double* phi, double* u1, - lapack_int* ldu1, double* u2, lapack_int* ldu2, - double* v1t, lapack_int* ldv1t, double* v2t, - lapack_int* ldv2t, double* b11d, double* b11e, - double* b12d, double* b12e, double* b21d, - double* b21e, double* b22d, double* b22e, - double* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_dorbdb( char* trans, char* signs, - lapack_int* m, lapack_int* p, lapack_int* q, - double* x11, lapack_int* ldx11, double* x12, - lapack_int* ldx12, double* x21, lapack_int* ldx21, - double* x22, lapack_int* ldx22, double* theta, - double* phi, double* taup1, double* taup2, - double* tauq1, double* tauq2, double* work, - lapack_int* lwork , lapack_int *info ); -void LAPACK_dorcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - char* signs, lapack_int* m, lapack_int* p, - lapack_int* q, double* x11, lapack_int* ldx11, - double* x12, lapack_int* ldx12, double* x21, - lapack_int* ldx21, double* x22, lapack_int* ldx22, - double* theta, double* u1, lapack_int* ldu1, - double* u2, lapack_int* ldu2, double* v1t, - lapack_int* ldv1t, double* v2t, lapack_int* ldv2t, - double* work, lapack_int* lwork, - lapack_int* iwork , lapack_int *info ); -void LAPACK_dsyconv( char* uplo, char* way, - lapack_int* n, double* a, lapack_int* lda, - const lapack_int* ipiv, double* work , lapack_int *info ); -void LAPACK_dsyswapr( char* uplo, lapack_int* n, - double* a, lapack_int* i1, lapack_int* i2 ); -void LAPACK_dsytri2( char* uplo, lapack_int* n, - double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_dsytri2x( char* uplo, lapack_int* n, - double* a, lapack_int* lda, - const lapack_int* ipiv, double* work, - lapack_int* nb , lapack_int *info ); -void LAPACK_dsytrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, const double* a, - lapack_int* lda, const lapack_int* ipiv, - double* b, lapack_int* ldb, double* work , lapack_int *info ); -void LAPACK_sbbcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - lapack_int* m, lapack_int* p, lapack_int* q, - float* theta, float* phi, float* u1, - lapack_int* ldu1, float* u2, lapack_int* ldu2, - float* v1t, lapack_int* ldv1t, float* v2t, - lapack_int* ldv2t, float* b11d, float* b11e, - float* b12d, float* b12e, float* b21d, - float* b21e, float* b22d, float* b22e, - float* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_sorbdb( char* trans, char* signs, - lapack_int* m, lapack_int* p, lapack_int* q, - float* x11, lapack_int* ldx11, float* x12, - lapack_int* ldx12, float* x21, lapack_int* ldx21, - float* x22, lapack_int* ldx22, float* theta, - float* phi, float* taup1, float* taup2, - float* tauq1, float* tauq2, float* work, - lapack_int* lwork , lapack_int *info ); -void LAPACK_sorcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - char* signs, lapack_int* m, lapack_int* p, - lapack_int* q, float* x11, lapack_int* ldx11, - float* x12, lapack_int* ldx12, float* x21, - lapack_int* ldx21, float* x22, lapack_int* ldx22, - float* theta, float* u1, lapack_int* ldu1, - float* u2, lapack_int* ldu2, float* v1t, - lapack_int* ldv1t, float* v2t, lapack_int* ldv2t, - float* work, lapack_int* lwork, - lapack_int* iwork , lapack_int *info ); -void LAPACK_ssyconv( char* uplo, char* way, - lapack_int* n, float* a, lapack_int* lda, - const lapack_int* ipiv, float* work , lapack_int *info ); -void LAPACK_ssyswapr( char* uplo, lapack_int* n, - float* a, lapack_int* i1, lapack_int* i2 ); -void LAPACK_ssytri2( char* uplo, lapack_int* n, - float* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_float* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_ssytri2x( char* uplo, lapack_int* n, - float* a, lapack_int* lda, - const lapack_int* ipiv, float* work, - lapack_int* nb , lapack_int *info ); -void LAPACK_ssytrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, const float* a, - lapack_int* lda, const lapack_int* ipiv, - float* b, lapack_int* ldb, float* work , lapack_int *info ); -void LAPACK_zbbcsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - lapack_int* m, lapack_int* p, lapack_int* q, - double* theta, double* phi, - lapack_complex_double* u1, lapack_int* ldu1, - lapack_complex_double* u2, lapack_int* ldu2, - lapack_complex_double* v1t, lapack_int* ldv1t, - lapack_complex_double* v2t, lapack_int* ldv2t, - double* b11d, double* b11e, double* b12d, - double* b12e, double* b21d, double* b21e, - double* b22d, double* b22e, double* rwork, - lapack_int* lrwork , lapack_int *info ); -void LAPACK_zheswapr( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* i1, - lapack_int* i2 ); -void LAPACK_zhetri2( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_zhetri2x( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int* nb , lapack_int *info ); -void LAPACK_zhetrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work , lapack_int *info ); -void LAPACK_zsyconv( char* uplo, char* way, - lapack_int* n, lapack_complex_double* a, - lapack_int* lda, const lapack_int* ipiv, - lapack_complex_double* work , lapack_int *info ); -void LAPACK_zsyswapr( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* i1, - lapack_int* i2 ); -void LAPACK_zsytri2( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_zsytri2x( char* uplo, lapack_int* n, - lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* work, lapack_int* nb , lapack_int *info ); -void LAPACK_zsytrs2( char* uplo, lapack_int* n, - lapack_int* nrhs, - const lapack_complex_double* a, lapack_int* lda, - const lapack_int* ipiv, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work , lapack_int *info ); -void LAPACK_zunbdb( char* trans, char* signs, - lapack_int* m, lapack_int* p, lapack_int* q, - lapack_complex_double* x11, lapack_int* ldx11, - lapack_complex_double* x12, lapack_int* ldx12, - lapack_complex_double* x21, lapack_int* ldx21, - lapack_complex_double* x22, lapack_int* ldx22, - double* theta, double* phi, - lapack_complex_double* taup1, - lapack_complex_double* taup2, - lapack_complex_double* tauq1, - lapack_complex_double* tauq2, - lapack_complex_double* work, lapack_int* lwork , lapack_int *info ); -void LAPACK_zuncsd( char* jobu1, char* jobu2, - char* jobv1t, char* jobv2t, char* trans, - char* signs, lapack_int* m, lapack_int* p, - lapack_int* q, lapack_complex_double* x11, - lapack_int* ldx11, lapack_complex_double* x12, - lapack_int* ldx12, lapack_complex_double* x21, - lapack_int* ldx21, lapack_complex_double* x22, - lapack_int* ldx22, double* theta, - lapack_complex_double* u1, lapack_int* ldu1, - lapack_complex_double* u2, lapack_int* ldu2, - lapack_complex_double* v1t, lapack_int* ldv1t, - lapack_complex_double* v2t, lapack_int* ldv2t, - lapack_complex_double* work, lapack_int* lwork, - double* rwork, lapack_int* lrwork, - lapack_int* iwork , lapack_int *info ); -// LAPACK 3.4.0 -void LAPACK_sgemqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* nb, const float* v, - lapack_int* ldv, const float* t, lapack_int* ldt, float* c, - lapack_int* ldc, float* work, lapack_int *info ); -void LAPACK_dgemqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* nb, const double* v, - lapack_int* ldv, const double* t, lapack_int* ldt, - double* c, lapack_int* ldc, double* work, - lapack_int *info ); -void LAPACK_cgemqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* nb, - const lapack_complex_float* v, lapack_int* ldv, - const lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* c, lapack_int* ldc, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zgemqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* nb, - const lapack_complex_double* v, lapack_int* ldv, - const lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* c, lapack_int* ldc, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_sgeqrt( lapack_int* m, lapack_int* n, lapack_int* nb, float* a, - lapack_int* lda, float* t, lapack_int* ldt, float* work, - lapack_int *info ); -void LAPACK_dgeqrt( lapack_int* m, lapack_int* n, lapack_int* nb, double* a, - lapack_int* lda, double* t, lapack_int* ldt, double* work, - lapack_int *info ); -void LAPACK_cgeqrt( lapack_int* m, lapack_int* n, lapack_int* nb, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_zgeqrt( lapack_int* m, lapack_int* n, lapack_int* nb, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_sgeqrt2( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* t, lapack_int* ldt, lapack_int *info ); -void LAPACK_dgeqrt2( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* t, lapack_int* ldt, lapack_int *info ); -void LAPACK_cgeqrt2( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_zgeqrt2( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_sgeqrt3( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* t, lapack_int* ldt, lapack_int *info ); -void LAPACK_dgeqrt3( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* t, lapack_int* ldt, lapack_int *info ); -void LAPACK_cgeqrt3( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_zgeqrt3( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_stpmqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, lapack_int* nb, - const float* v, lapack_int* ldv, const float* t, - lapack_int* ldt, float* a, lapack_int* lda, float* b, - lapack_int* ldb, float* work, lapack_int *info ); -void LAPACK_dtpmqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, lapack_int* nb, - const double* v, lapack_int* ldv, const double* t, - lapack_int* ldt, double* a, lapack_int* lda, double* b, - lapack_int* ldb, double* work, lapack_int *info ); -void LAPACK_ctpmqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, lapack_int* nb, - const lapack_complex_float* v, lapack_int* ldv, - const lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_ztpmqrt( char* side, char* trans, lapack_int* m, lapack_int* n, - lapack_int* k, lapack_int* l, lapack_int* nb, - const lapack_complex_double* v, lapack_int* ldv, - const lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_dtpqrt( lapack_int* m, lapack_int* n, lapack_int* l, lapack_int* nb, - double* a, lapack_int* lda, double* b, lapack_int* ldb, - double* t, lapack_int* ldt, double* work, - lapack_int *info ); -void LAPACK_ctpqrt( lapack_int* m, lapack_int* n, lapack_int* l, lapack_int* nb, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* t, lapack_complex_float* b, - lapack_int* ldb, lapack_int* ldt, - lapack_complex_float* work, lapack_int *info ); -void LAPACK_ztpqrt( lapack_int* m, lapack_int* n, lapack_int* l, lapack_int* nb, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* work, lapack_int *info ); -void LAPACK_stpqrt2( lapack_int* m, lapack_int* n, float* a, lapack_int* lda, - float* b, lapack_int* ldb, float* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_dtpqrt2( lapack_int* m, lapack_int* n, double* a, lapack_int* lda, - double* b, lapack_int* ldb, double* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_ctpqrt2( lapack_int* m, lapack_int* n, lapack_complex_float* a, - lapack_int* lda, lapack_complex_float* b, lapack_int* ldb, - lapack_complex_float* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_ztpqrt2( lapack_int* m, lapack_int* n, lapack_complex_double* a, - lapack_int* lda, lapack_complex_double* b, lapack_int* ldb, - lapack_complex_double* t, lapack_int* ldt, - lapack_int *info ); -void LAPACK_stprfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, lapack_int* l, - const float* v, lapack_int* ldv, const float* t, - lapack_int* ldt, float* a, lapack_int* lda, float* b, - lapack_int* ldb, const float* mywork, - lapack_int* myldwork ); -void LAPACK_dtprfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, lapack_int* l, - const double* v, lapack_int* ldv, const double* t, - lapack_int* ldt, double* a, lapack_int* lda, double* b, - lapack_int* ldb, const double* mywork, - lapack_int* myldwork ); -void LAPACK_ctprfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, lapack_int* l, - const lapack_complex_float* v, lapack_int* ldv, - const lapack_complex_float* t, lapack_int* ldt, - lapack_complex_float* a, lapack_int* lda, - lapack_complex_float* b, lapack_int* ldb, - const float* mywork, lapack_int* myldwork ); -void LAPACK_ztprfb( char* side, char* trans, char* direct, char* storev, - lapack_int* m, lapack_int* n, lapack_int* k, lapack_int* l, - const lapack_complex_double* v, lapack_int* ldv, - const lapack_complex_double* t, lapack_int* ldt, - lapack_complex_double* a, lapack_int* lda, - lapack_complex_double* b, lapack_int* ldb, - const double* mywork, lapack_int* myldwork ); -// LAPACK 3.X.X -void LAPACK_csyr( char* uplo, lapack_int* n, lapack_complex_float* alpha, - const lapack_complex_float* x, lapack_int* incx, - lapack_complex_float* a, lapack_int* lda ); -void LAPACK_zsyr( char* uplo, lapack_int* n, lapack_complex_double* alpha, - const lapack_complex_double* x, lapack_int* incx, - lapack_complex_double* a, lapack_int* lda ); - -#ifdef __cplusplus -} -#endif /* __cplusplus */ - -#endif /* _LAPACKE_H_ */ - -#endif /* _MKL_LAPACKE_H_ */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke_mangling.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke_mangling.h deleted file mode 100644 index 6211fd144d3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/misc/lapacke_mangling.h +++ /dev/null @@ -1,17 +0,0 @@ -#ifndef LAPACK_HEADER_INCLUDED -#define LAPACK_HEADER_INCLUDED - -#ifndef LAPACK_GLOBAL -#if defined(LAPACK_GLOBAL_PATTERN_LC) || defined(ADD_) -#define LAPACK_GLOBAL(lcname,UCNAME) lcname##_ -#elif defined(LAPACK_GLOBAL_PATTERN_UC) || defined(UPPER) -#define LAPACK_GLOBAL(lcname,UCNAME) UCNAME -#elif defined(LAPACK_GLOBAL_PATTERN_MC) || defined(NOCHANGE) -#define LAPACK_GLOBAL(lcname,UCNAME) lcname -#else -#define LAPACK_GLOBAL(lcname,UCNAME) lcname##_ -#endif -#endif - -#endif - diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h deleted file mode 100644 index 05a7449bc9e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h +++ /dev/null @@ -1,332 +0,0 @@ - -/** \returns an expression of the coefficient wise product of \c *this and \a other - * - * \sa MatrixBase::cwiseProduct - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product) -operator*(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product)(derived(), other.derived()); -} - -/** \returns an expression of the coefficient wise quotient of \c *this and \a other - * - * \sa MatrixBase::cwiseQuotient - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -operator/(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of \c *this and \a other - * - * Example: \include Cwise_min.cpp - * Output: \verbinclude Cwise_min.out - * - * \sa max() - */ -EIGEN_MAKE_CWISE_BINARY_OP(min,min) - -/** \returns an expression of the coefficient-wise min of \c *this and scalar \a other - * - * \sa max() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, - const CwiseNullaryOp, PlainObject> > -#ifdef EIGEN_PARSED_BY_DOXYGEN -min -#else -(min) -#endif -(const Scalar &other) const -{ - return (min)(Derived::PlainObject::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise max of \c *this and \a other - * - * Example: \include Cwise_max.cpp - * Output: \verbinclude Cwise_max.out - * - * \sa min() - */ -EIGEN_MAKE_CWISE_BINARY_OP(max,max) - -/** \returns an expression of the coefficient-wise max of \c *this and scalar \a other - * - * \sa min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, - const CwiseNullaryOp, PlainObject> > -#ifdef EIGEN_PARSED_BY_DOXYGEN -max -#else -(max) -#endif -(const Scalar &other) const -{ - return (max)(Derived::PlainObject::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise power of \c *this to the given array of \a exponents. - * - * This function computes the coefficient-wise power. - * - * Example: \include Cwise_array_power_array.cpp - * Output: \verbinclude Cwise_array_power_array.out - */ -EIGEN_MAKE_CWISE_BINARY_OP(pow,pow) - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(pow,pow) -#else -/** \returns an expression of the coefficients of \c *this rasied to the constant power \a exponent - * - * \tparam T is the scalar type of \a exponent. It must be compatible with the scalar type of the given expression. - * - * This function computes the coefficient-wise power. The function MatrixBase::pow() in the - * unsupported module MatrixFunctions computes the matrix power. - * - * Example: \include Cwise_pow.cpp - * Output: \verbinclude Cwise_pow.out - * - * \sa ArrayBase::pow(ArrayBase), square(), cube(), exp(), log() - */ -template -const CwiseBinaryOp,Derived,Constant > pow(const T& exponent) const; -#endif - - -// TODO code generating macros could be moved to Macros.h and could include generation of documentation -#define EIGEN_MAKE_CWISE_COMP_OP(OP, COMPARATOR) \ -template \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> \ -OP(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ -{ \ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); \ -}\ -typedef CwiseBinaryOp, const Derived, const CwiseNullaryOp, PlainObject> > Cmp ## COMPARATOR ## ReturnType; \ -typedef CwiseBinaryOp, const CwiseNullaryOp, PlainObject>, const Derived > RCmp ## COMPARATOR ## ReturnType; \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Cmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s) const { \ - return this->OP(Derived::PlainObject::Constant(rows(), cols(), s)); \ -} \ -EIGEN_DEVICE_FUNC friend EIGEN_STRONG_INLINE const RCmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s, const EIGEN_CURRENT_STORAGE_BASE_CLASS& d) { \ - return Derived::PlainObject::Constant(d.rows(), d.cols(), s).OP(d); \ -} - -#define EIGEN_MAKE_CWISE_COMP_R_OP(OP, R_OP, RCOMPARATOR) \ -template \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseBinaryOp, const OtherDerived, const Derived> \ -OP(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ -{ \ - return CwiseBinaryOp, const OtherDerived, const Derived>(other.derived(), derived()); \ -} \ -EIGEN_DEVICE_FUNC \ -inline const RCmp ## RCOMPARATOR ## ReturnType \ -OP(const Scalar& s) const { \ - return Derived::PlainObject::Constant(rows(), cols(), s).R_OP(*this); \ -} \ -friend inline const Cmp ## RCOMPARATOR ## ReturnType \ -OP(const Scalar& s, const Derived& d) { \ - return d.R_OP(Derived::PlainObject::Constant(d.rows(), d.cols(), s)); \ -} - - - -/** \returns an expression of the coefficient-wise \< operator of *this and \a other - * - * Example: \include Cwise_less.cpp - * Output: \verbinclude Cwise_less.out - * - * \sa all(), any(), operator>(), operator<=() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator<, LT) - -/** \returns an expression of the coefficient-wise \<= operator of *this and \a other - * - * Example: \include Cwise_less_equal.cpp - * Output: \verbinclude Cwise_less_equal.out - * - * \sa all(), any(), operator>=(), operator<() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator<=, LE) - -/** \returns an expression of the coefficient-wise \> operator of *this and \a other - * - * Example: \include Cwise_greater.cpp - * Output: \verbinclude Cwise_greater.out - * - * \sa all(), any(), operator>=(), operator<() - */ -EIGEN_MAKE_CWISE_COMP_R_OP(operator>, operator<, LT) - -/** \returns an expression of the coefficient-wise \>= operator of *this and \a other - * - * Example: \include Cwise_greater_equal.cpp - * Output: \verbinclude Cwise_greater_equal.out - * - * \sa all(), any(), operator>(), operator<=() - */ -EIGEN_MAKE_CWISE_COMP_R_OP(operator>=, operator<=, LE) - -/** \returns an expression of the coefficient-wise == operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include Cwise_equal_equal.cpp - * Output: \verbinclude Cwise_equal_equal.out - * - * \sa all(), any(), isApprox(), isMuchSmallerThan() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator==, EQ) - -/** \returns an expression of the coefficient-wise != operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include Cwise_not_equal.cpp - * Output: \verbinclude Cwise_not_equal.out - * - * \sa all(), any(), isApprox(), isMuchSmallerThan() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator!=, NEQ) - - -#undef EIGEN_MAKE_CWISE_COMP_OP -#undef EIGEN_MAKE_CWISE_COMP_R_OP - -// scalar addition -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator+,sum) -#else -/** \returns an expression of \c *this with each coeff incremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - * - * Example: \include Cwise_plus.cpp - * Output: \verbinclude Cwise_plus.out - * - * \sa operator+=(), operator-() - */ -template -const CwiseBinaryOp,Derived,Constant > operator+(const T& scalar) const; -/** \returns an expression of \a expr with each coeff incremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator+(const T& scalar, const StorageBaseType& expr); -#endif - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator-,difference) -#else -/** \returns an expression of \c *this with each coeff decremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - * - * Example: \include Cwise_minus.cpp - * Output: \verbinclude Cwise_minus.out - * - * \sa operator+=(), operator-() - */ -template -const CwiseBinaryOp,Derived,Constant > operator-(const T& scalar) const; -/** \returns an expression of the constant matrix of value \a scalar decremented by the coefficients of \a expr - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator-(const T& scalar, const StorageBaseType& expr); -#endif - - -#ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(operator/,quotient) -#else - /** - * \brief Component-wise division of the scalar \a s by array elements of \a a. - * - * \tparam Scalar is the scalar type of \a x. It must be compatible with the scalar type of the given array expression (\c Derived::Scalar). - */ - template friend - inline const CwiseBinaryOp,Constant,Derived> - operator/(const T& s,const StorageBaseType& a); -#endif - -/** \returns an expression of the coefficient-wise ^ operator of *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_xor.cpp - * Output: \verbinclude Cwise_boolean_xor.out - * - * \sa operator&&(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator^(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} - -// NOTE disabled until we agree on argument order -#if 0 -/** \cpp11 \returns an expression of the coefficient-wise polygamma function. - * - * \specialfunctions_module - * - * It returns the \a n -th derivative of the digamma(psi) evaluated at \c *this. - * - * \warning Be careful with the order of the parameters: x.polygamma(n) is equivalent to polygamma(n,x) - * - * \sa Eigen::polygamma() - */ -template -inline const CwiseBinaryOp, const DerivedN, const Derived> -polygamma(const EIGEN_CURRENT_STORAGE_BASE_CLASS &n) const -{ - return CwiseBinaryOp, const DerivedN, const Derived>(n.derived(), this->derived()); -} -#endif - -/** \returns an expression of the coefficient-wise zeta function. - * - * \specialfunctions_module - * - * It returns the Riemann zeta function of two arguments \c *this and \a q: - * - * \param *this is the exposent, it must be > 1 - * \param q is the shift, it must be > 0 - * - * \note This function supports only float and double scalar types. To support other scalar types, the user has - * to provide implementations of zeta(T,T) for any scalar type T to be supported. - * - * This method is an alias for zeta(*this,q); - * - * \sa Eigen::zeta() - */ -template -inline const CwiseBinaryOp, const Derived, const DerivedQ> -zeta(const EIGEN_CURRENT_STORAGE_BASE_CLASS &q) const -{ - return CwiseBinaryOp, const Derived, const DerivedQ>(this->derived(), q.derived()); -} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h deleted file mode 100644 index ebaa3f192b1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h +++ /dev/null @@ -1,552 +0,0 @@ - - -typedef CwiseUnaryOp, const Derived> AbsReturnType; -typedef CwiseUnaryOp, const Derived> ArgReturnType; -typedef CwiseUnaryOp, const Derived> Abs2ReturnType; -typedef CwiseUnaryOp, const Derived> SqrtReturnType; -typedef CwiseUnaryOp, const Derived> RsqrtReturnType; -typedef CwiseUnaryOp, const Derived> SignReturnType; -typedef CwiseUnaryOp, const Derived> InverseReturnType; -typedef CwiseUnaryOp, const Derived> BooleanNotReturnType; - -typedef CwiseUnaryOp, const Derived> ExpReturnType; -typedef CwiseUnaryOp, const Derived> LogReturnType; -typedef CwiseUnaryOp, const Derived> Log1pReturnType; -typedef CwiseUnaryOp, const Derived> Log10ReturnType; -typedef CwiseUnaryOp, const Derived> CosReturnType; -typedef CwiseUnaryOp, const Derived> SinReturnType; -typedef CwiseUnaryOp, const Derived> TanReturnType; -typedef CwiseUnaryOp, const Derived> AcosReturnType; -typedef CwiseUnaryOp, const Derived> AsinReturnType; -typedef CwiseUnaryOp, const Derived> AtanReturnType; -typedef CwiseUnaryOp, const Derived> TanhReturnType; -typedef CwiseUnaryOp, const Derived> SinhReturnType; -typedef CwiseUnaryOp, const Derived> CoshReturnType; -typedef CwiseUnaryOp, const Derived> SquareReturnType; -typedef CwiseUnaryOp, const Derived> CubeReturnType; -typedef CwiseUnaryOp, const Derived> RoundReturnType; -typedef CwiseUnaryOp, const Derived> FloorReturnType; -typedef CwiseUnaryOp, const Derived> CeilReturnType; -typedef CwiseUnaryOp, const Derived> IsNaNReturnType; -typedef CwiseUnaryOp, const Derived> IsInfReturnType; -typedef CwiseUnaryOp, const Derived> IsFiniteReturnType; - -/** \returns an expression of the coefficient-wise absolute value of \c *this - * - * Example: \include Cwise_abs.cpp - * Output: \verbinclude Cwise_abs.out - * - * \sa Math functions, abs2() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const AbsReturnType -abs() const -{ - return AbsReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise phase angle of \c *this - * - * Example: \include Cwise_arg.cpp - * Output: \verbinclude Cwise_arg.out - * - * \sa abs() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const ArgReturnType -arg() const -{ - return ArgReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise squared absolute value of \c *this - * - * Example: \include Cwise_abs2.cpp - * Output: \verbinclude Cwise_abs2.out - * - * \sa Math functions, abs(), square() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const Abs2ReturnType -abs2() const -{ - return Abs2ReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise exponential of *this. - * - * This function computes the coefficient-wise exponential. The function MatrixBase::exp() in the - * unsupported module MatrixFunctions computes the matrix exponential. - * - * Example: \include Cwise_exp.cpp - * Output: \verbinclude Cwise_exp.out - * - * \sa Math functions, pow(), log(), sin(), cos() - */ -EIGEN_DEVICE_FUNC -inline const ExpReturnType -exp() const -{ - return ExpReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise logarithm of *this. - * - * This function computes the coefficient-wise logarithm. The function MatrixBase::log() in the - * unsupported module MatrixFunctions computes the matrix logarithm. - * - * Example: \include Cwise_log.cpp - * Output: \verbinclude Cwise_log.out - * - * \sa Math functions, exp() - */ -EIGEN_DEVICE_FUNC -inline const LogReturnType -log() const -{ - return LogReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise logarithm of 1 plus \c *this. - * - * In exact arithmetic, \c x.log() is equivalent to \c (x+1).log(), - * however, with finite precision, this function is much more accurate when \c x is close to zero. - * - * \sa Math functions, log() - */ -EIGEN_DEVICE_FUNC -inline const Log1pReturnType -log1p() const -{ - return Log1pReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise base-10 logarithm of *this. - * - * This function computes the coefficient-wise base-10 logarithm. - * - * Example: \include Cwise_log10.cpp - * Output: \verbinclude Cwise_log10.out - * - * \sa Math functions, log() - */ -EIGEN_DEVICE_FUNC -inline const Log10ReturnType -log10() const -{ - return Log10ReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise square root of *this. - * - * This function computes the coefficient-wise square root. The function MatrixBase::sqrt() in the - * unsupported module MatrixFunctions computes the matrix square root. - * - * Example: \include Cwise_sqrt.cpp - * Output: \verbinclude Cwise_sqrt.out - * - * \sa Math functions, pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const SqrtReturnType -sqrt() const -{ - return SqrtReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise inverse square root of *this. - * - * This function computes the coefficient-wise inverse square root. - * - * Example: \include Cwise_sqrt.cpp - * Output: \verbinclude Cwise_sqrt.out - * - * \sa pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const RsqrtReturnType -rsqrt() const -{ - return RsqrtReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise signum of *this. - * - * This function computes the coefficient-wise signum. - * - * Example: \include Cwise_sign.cpp - * Output: \verbinclude Cwise_sign.out - * - * \sa pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const SignReturnType -sign() const -{ - return SignReturnType(derived()); -} - - -/** \returns an expression of the coefficient-wise cosine of *this. - * - * This function computes the coefficient-wise cosine. The function MatrixBase::cos() in the - * unsupported module MatrixFunctions computes the matrix cosine. - * - * Example: \include Cwise_cos.cpp - * Output: \verbinclude Cwise_cos.out - * - * \sa Math functions, sin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const CosReturnType -cos() const -{ - return CosReturnType(derived()); -} - - -/** \returns an expression of the coefficient-wise sine of *this. - * - * This function computes the coefficient-wise sine. The function MatrixBase::sin() in the - * unsupported module MatrixFunctions computes the matrix sine. - * - * Example: \include Cwise_sin.cpp - * Output: \verbinclude Cwise_sin.out - * - * \sa Math functions, cos(), asin() - */ -EIGEN_DEVICE_FUNC -inline const SinReturnType -sin() const -{ - return SinReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise tan of *this. - * - * Example: \include Cwise_tan.cpp - * Output: \verbinclude Cwise_tan.out - * - * \sa Math functions, cos(), sin() - */ -EIGEN_DEVICE_FUNC -inline const TanReturnType -tan() const -{ - return TanReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc tan of *this. - * - * Example: \include Cwise_atan.cpp - * Output: \verbinclude Cwise_atan.out - * - * \sa Math functions, tan(), asin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const AtanReturnType -atan() const -{ - return AtanReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc cosine of *this. - * - * Example: \include Cwise_acos.cpp - * Output: \verbinclude Cwise_acos.out - * - * \sa Math functions, cos(), asin() - */ -EIGEN_DEVICE_FUNC -inline const AcosReturnType -acos() const -{ - return AcosReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc sine of *this. - * - * Example: \include Cwise_asin.cpp - * Output: \verbinclude Cwise_asin.out - * - * \sa Math functions, sin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const AsinReturnType -asin() const -{ - return AsinReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic tan of *this. - * - * Example: \include Cwise_tanh.cpp - * Output: \verbinclude Cwise_tanh.out - * - * \sa Math functions, tan(), sinh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const TanhReturnType -tanh() const -{ - return TanhReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic sin of *this. - * - * Example: \include Cwise_sinh.cpp - * Output: \verbinclude Cwise_sinh.out - * - * \sa Math functions, sin(), tanh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const SinhReturnType -sinh() const -{ - return SinhReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic cos of *this. - * - * Example: \include Cwise_cosh.cpp - * Output: \verbinclude Cwise_cosh.out - * - * \sa Math functions, tan(), sinh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const CoshReturnType -cosh() const -{ - return CoshReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise inverse of *this. - * - * Example: \include Cwise_inverse.cpp - * Output: \verbinclude Cwise_inverse.out - * - * \sa operator/(), operator*() - */ -EIGEN_DEVICE_FUNC -inline const InverseReturnType -inverse() const -{ - return InverseReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise square of *this. - * - * Example: \include Cwise_square.cpp - * Output: \verbinclude Cwise_square.out - * - * \sa Math functions, abs2(), cube(), pow() - */ -EIGEN_DEVICE_FUNC -inline const SquareReturnType -square() const -{ - return SquareReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise cube of *this. - * - * Example: \include Cwise_cube.cpp - * Output: \verbinclude Cwise_cube.out - * - * \sa Math functions, square(), pow() - */ -EIGEN_DEVICE_FUNC -inline const CubeReturnType -cube() const -{ - return CubeReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise round of *this. - * - * Example: \include Cwise_round.cpp - * Output: \verbinclude Cwise_round.out - * - * \sa Math functions, ceil(), floor() - */ -EIGEN_DEVICE_FUNC -inline const RoundReturnType -round() const -{ - return RoundReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise floor of *this. - * - * Example: \include Cwise_floor.cpp - * Output: \verbinclude Cwise_floor.out - * - * \sa Math functions, ceil(), round() - */ -EIGEN_DEVICE_FUNC -inline const FloorReturnType -floor() const -{ - return FloorReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise ceil of *this. - * - * Example: \include Cwise_ceil.cpp - * Output: \verbinclude Cwise_ceil.out - * - * \sa Math functions, floor(), round() - */ -EIGEN_DEVICE_FUNC -inline const CeilReturnType -ceil() const -{ - return CeilReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isnan of *this. - * - * Example: \include Cwise_isNaN.cpp - * Output: \verbinclude Cwise_isNaN.out - * - * \sa isfinite(), isinf() - */ -EIGEN_DEVICE_FUNC -inline const IsNaNReturnType -isNaN() const -{ - return IsNaNReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isinf of *this. - * - * Example: \include Cwise_isInf.cpp - * Output: \verbinclude Cwise_isInf.out - * - * \sa isnan(), isfinite() - */ -EIGEN_DEVICE_FUNC -inline const IsInfReturnType -isInf() const -{ - return IsInfReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isfinite of *this. - * - * Example: \include Cwise_isFinite.cpp - * Output: \verbinclude Cwise_isFinite.out - * - * \sa isnan(), isinf() - */ -EIGEN_DEVICE_FUNC -inline const IsFiniteReturnType -isFinite() const -{ - return IsFiniteReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise ! operator of *this - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_not.cpp - * Output: \verbinclude Cwise_boolean_not.out - * - * \sa operator!=() - */ -EIGEN_DEVICE_FUNC -inline const BooleanNotReturnType -operator!() const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return BooleanNotReturnType(derived()); -} - - -// --- SpecialFunctions module --- - -typedef CwiseUnaryOp, const Derived> LgammaReturnType; -typedef CwiseUnaryOp, const Derived> DigammaReturnType; -typedef CwiseUnaryOp, const Derived> ErfReturnType; -typedef CwiseUnaryOp, const Derived> ErfcReturnType; - -/** \cpp11 \returns an expression of the coefficient-wise ln(|gamma(*this)|). - * - * \specialfunctions_module - * - * Example: \include Cwise_lgamma.cpp - * Output: \verbinclude Cwise_lgamma.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of lgamma(T) for any scalar - * type T to be supported. - * - * \sa Math functions, digamma() - */ -EIGEN_DEVICE_FUNC -inline const LgammaReturnType -lgamma() const -{ - return LgammaReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise digamma (psi, derivative of lgamma). - * - * \specialfunctions_module - * - * \note This function supports only float and double scalar types. To support other scalar types, - * the user has to provide implementations of digamma(T) for any scalar - * type T to be supported. - * - * \sa Math functions, Eigen::digamma(), Eigen::polygamma(), lgamma() - */ -EIGEN_DEVICE_FUNC -inline const DigammaReturnType -digamma() const -{ - return DigammaReturnType(derived()); -} - -/** \cpp11 \returns an expression of the coefficient-wise Gauss error - * function of *this. - * - * \specialfunctions_module - * - * Example: \include Cwise_erf.cpp - * Output: \verbinclude Cwise_erf.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of erf(T) for any scalar - * type T to be supported. - * - * \sa Math functions, erfc() - */ -EIGEN_DEVICE_FUNC -inline const ErfReturnType -erf() const -{ - return ErfReturnType(derived()); -} - -/** \cpp11 \returns an expression of the coefficient-wise Complementary error - * function of *this. - * - * \specialfunctions_module - * - * Example: \include Cwise_erfc.cpp - * Output: \verbinclude Cwise_erfc.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of erfc(T) for any scalar - * type T to be supported. - * - * \sa Math functions, erf() - */ -EIGEN_DEVICE_FUNC -inline const ErfcReturnType -erfc() const -{ - return ErfcReturnType(derived()); -} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h deleted file mode 100644 index ac35a0086cf..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h +++ /dev/null @@ -1,1058 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARSED_BY_DOXYGEN - -/// \internal expression type of a column */ -typedef Block::RowsAtCompileTime, 1, !IsRowMajor> ColXpr; -typedef const Block::RowsAtCompileTime, 1, !IsRowMajor> ConstColXpr; -/// \internal expression type of a row */ -typedef Block::ColsAtCompileTime, IsRowMajor> RowXpr; -typedef const Block::ColsAtCompileTime, IsRowMajor> ConstRowXpr; -/// \internal expression type of a block of whole columns */ -typedef Block::RowsAtCompileTime, Dynamic, !IsRowMajor> ColsBlockXpr; -typedef const Block::RowsAtCompileTime, Dynamic, !IsRowMajor> ConstColsBlockXpr; -/// \internal expression type of a block of whole rows */ -typedef Block::ColsAtCompileTime, IsRowMajor> RowsBlockXpr; -typedef const Block::ColsAtCompileTime, IsRowMajor> ConstRowsBlockXpr; -/// \internal expression type of a block of whole columns */ -template struct NColsBlockXpr { typedef Block::RowsAtCompileTime, N, !IsRowMajor> Type; }; -template struct ConstNColsBlockXpr { typedef const Block::RowsAtCompileTime, N, !IsRowMajor> Type; }; -/// \internal expression type of a block of whole rows */ -template struct NRowsBlockXpr { typedef Block::ColsAtCompileTime, IsRowMajor> Type; }; -template struct ConstNRowsBlockXpr { typedef const Block::ColsAtCompileTime, IsRowMajor> Type; }; -/// \internal expression of a block */ -typedef Block BlockXpr; -typedef const Block ConstBlockXpr; -/// \internal expression of a block of fixed sizes */ -template struct FixedBlockXpr { typedef Block Type; }; -template struct ConstFixedBlockXpr { typedef Block Type; }; - -typedef VectorBlock SegmentReturnType; -typedef const VectorBlock ConstSegmentReturnType; -template struct FixedSegmentReturnType { typedef VectorBlock Type; }; -template struct ConstFixedSegmentReturnType { typedef const VectorBlock Type; }; - -#endif // not EIGEN_PARSED_BY_DOXYGEN - -/// \returns a dynamic-size expression of a block in *this. -/// -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// \param blockRows the number of rows in the block -/// \param blockCols the number of columns in the block -/// -/// Example: \include MatrixBase_block_int_int_int_int.cpp -/// Output: \verbinclude MatrixBase_block_int_int_int_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size matrix, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr block(Index startRow, Index startCol, Index blockRows, Index blockCols) -{ - return BlockXpr(derived(), startRow, startCol, blockRows, blockCols); -} - -/// This is the const version of block(Index,Index,Index,Index). */ -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr block(Index startRow, Index startCol, Index blockRows, Index blockCols) const -{ - return ConstBlockXpr(derived(), startRow, startCol, blockRows, blockCols); -} - - - - -/// \returns a dynamic-size expression of a top-right corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_topRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_topRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr topRightCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// This is the const version of topRightCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr topRightCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// \returns an expression of a fixed-size top-right corner of *this. -/// -/// \tparam CRows the number of rows in the corner -/// \tparam CCols the number of columns in the corner -/// -/// Example: \include MatrixBase_template_int_int_topRightCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topRightCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type topRightCorner() -{ - return typename FixedBlockXpr::Type(derived(), 0, cols() - CCols); -} - -/// This is the const version of topRightCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type topRightCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, cols() - CCols); -} - -/// \returns an expression of a top-right corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_topRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type topRightCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// This is the const version of topRightCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type topRightCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, cols() - cCols, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a top-left corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_topLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_topLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr topLeftCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), 0, 0, cRows, cCols); -} - -/// This is the const version of topLeftCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr topLeftCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), 0, 0, cRows, cCols); -} - -/// \returns an expression of a fixed-size top-left corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_topLeftCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topLeftCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type topLeftCorner() -{ - return typename FixedBlockXpr::Type(derived(), 0, 0); -} - -/// This is the const version of topLeftCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type topLeftCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, 0); -} - -/// \returns an expression of a top-left corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_topLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type topLeftCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), 0, 0, cRows, cCols); -} - -/// This is the const version of topLeftCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type topLeftCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, 0, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a bottom-right corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_bottomRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_bottomRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr bottomRightCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// This is the const version of bottomRightCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr bottomRightCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// \returns an expression of a fixed-size bottom-right corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_bottomRightCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomRightCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type bottomRightCorner() -{ - return typename FixedBlockXpr::Type(derived(), rows() - CRows, cols() - CCols); -} - -/// This is the const version of bottomRightCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type bottomRightCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - CRows, cols() - CCols); -} - -/// \returns an expression of a bottom-right corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_bottomRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type bottomRightCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// This is the const version of bottomRightCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type bottomRightCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a bottom-left corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_bottomLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_bottomLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr bottomLeftCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// This is the const version of bottomLeftCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr bottomLeftCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// \returns an expression of a fixed-size bottom-left corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_bottomLeftCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomLeftCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type bottomLeftCorner() -{ - return typename FixedBlockXpr::Type(derived(), rows() - CRows, 0); -} - -/// This is the const version of bottomLeftCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type bottomLeftCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - CRows, 0); -} - -/// \returns an expression of a bottom-left corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_bottomLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type bottomLeftCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// This is the const version of bottomLeftCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type bottomLeftCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - cRows, 0, cRows, cCols); -} - - - -/// \returns a block consisting of the top rows of *this. -/// -/// \param n the number of rows in the block -/// -/// Example: \include MatrixBase_topRows_int.cpp -/// Output: \verbinclude MatrixBase_topRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr topRows(Index n) -{ - return RowsBlockXpr(derived(), 0, 0, n, cols()); -} - -/// This is the const version of topRows(Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr topRows(Index n) const -{ - return ConstRowsBlockXpr(derived(), 0, 0, n, cols()); -} - -/// \returns a block consisting of the top rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_topRows.cpp -/// Output: \verbinclude MatrixBase_template_int_topRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type topRows(Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), 0, 0, n, cols()); -} - -/// This is the const version of topRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type topRows(Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), 0, 0, n, cols()); -} - - - -/// \returns a block consisting of the bottom rows of *this. -/// -/// \param n the number of rows in the block -/// -/// Example: \include MatrixBase_bottomRows_int.cpp -/// Output: \verbinclude MatrixBase_bottomRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr bottomRows(Index n) -{ - return RowsBlockXpr(derived(), rows() - n, 0, n, cols()); -} - -/// This is the const version of bottomRows(Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr bottomRows(Index n) const -{ - return ConstRowsBlockXpr(derived(), rows() - n, 0, n, cols()); -} - -/// \returns a block consisting of the bottom rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_bottomRows.cpp -/// Output: \verbinclude MatrixBase_template_int_bottomRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type bottomRows(Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), rows() - n, 0, n, cols()); -} - -/// This is the const version of bottomRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type bottomRows(Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), rows() - n, 0, n, cols()); -} - - - -/// \returns a block consisting of a range of rows of *this. -/// -/// \param startRow the index of the first row in the block -/// \param n the number of rows in the block -/// -/// Example: \include DenseBase_middleRows_int.cpp -/// Output: \verbinclude DenseBase_middleRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr middleRows(Index startRow, Index n) -{ - return RowsBlockXpr(derived(), startRow, 0, n, cols()); -} - -/// This is the const version of middleRows(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr middleRows(Index startRow, Index n) const -{ - return ConstRowsBlockXpr(derived(), startRow, 0, n, cols()); -} - -/// \returns a block consisting of a range of rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param startRow the index of the first row in the block -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include DenseBase_template_int_middleRows.cpp -/// Output: \verbinclude DenseBase_template_int_middleRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type middleRows(Index startRow, Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), startRow, 0, n, cols()); -} - -/// This is the const version of middleRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type middleRows(Index startRow, Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), startRow, 0, n, cols()); -} - - - -/// \returns a block consisting of the left columns of *this. -/// -/// \param n the number of columns in the block -/// -/// Example: \include MatrixBase_leftCols_int.cpp -/// Output: \verbinclude MatrixBase_leftCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr leftCols(Index n) -{ - return ColsBlockXpr(derived(), 0, 0, rows(), n); -} - -/// This is the const version of leftCols(Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr leftCols(Index n) const -{ - return ConstColsBlockXpr(derived(), 0, 0, rows(), n); -} - -/// \returns a block consisting of the left columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_leftCols.cpp -/// Output: \verbinclude MatrixBase_template_int_leftCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type leftCols(Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, 0, rows(), n); -} - -/// This is the const version of leftCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type leftCols(Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, 0, rows(), n); -} - - - -/// \returns a block consisting of the right columns of *this. -/// -/// \param n the number of columns in the block -/// -/// Example: \include MatrixBase_rightCols_int.cpp -/// Output: \verbinclude MatrixBase_rightCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr rightCols(Index n) -{ - return ColsBlockXpr(derived(), 0, cols() - n, rows(), n); -} - -/// This is the const version of rightCols(Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr rightCols(Index n) const -{ - return ConstColsBlockXpr(derived(), 0, cols() - n, rows(), n); -} - -/// \returns a block consisting of the right columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_rightCols.cpp -/// Output: \verbinclude MatrixBase_template_int_rightCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type rightCols(Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, cols() - n, rows(), n); -} - -/// This is the const version of rightCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type rightCols(Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, cols() - n, rows(), n); -} - - - -/// \returns a block consisting of a range of columns of *this. -/// -/// \param startCol the index of the first column in the block -/// \param numCols the number of columns in the block -/// -/// Example: \include DenseBase_middleCols_int.cpp -/// Output: \verbinclude DenseBase_middleCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr middleCols(Index startCol, Index numCols) -{ - return ColsBlockXpr(derived(), 0, startCol, rows(), numCols); -} - -/// This is the const version of middleCols(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr middleCols(Index startCol, Index numCols) const -{ - return ConstColsBlockXpr(derived(), 0, startCol, rows(), numCols); -} - -/// \returns a block consisting of a range of columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param startCol the index of the first column in the block -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include DenseBase_template_int_middleCols.cpp -/// Output: \verbinclude DenseBase_template_int_middleCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type middleCols(Index startCol, Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, startCol, rows(), n); -} - -/// This is the const version of middleCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type middleCols(Index startCol, Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, startCol, rows(), n); -} - - - -/// \returns a fixed-size expression of a block in *this. -/// -/// The template parameters \a NRows and \a NCols are the number of -/// rows and columns in the block. -/// -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// -/// Example: \include MatrixBase_block_int_int.cpp -/// Output: \verbinclude MatrixBase_block_int_int.out -/// -/// \note since block is a templated member, the keyword template has to be used -/// if the matrix type is also a template parameter: \code m.template block<3,3>(1,1); \endcode -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type block(Index startRow, Index startCol) -{ - return typename FixedBlockXpr::Type(derived(), startRow, startCol); -} - -/// This is the const version of block<>(Index, Index). */ -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type block(Index startRow, Index startCol) const -{ - return typename ConstFixedBlockXpr::Type(derived(), startRow, startCol); -} - -/// \returns an expression of a block in *this. -/// -/// \tparam NRows number of rows in block as specified at compile-time -/// \tparam NCols number of columns in block as specified at compile-time -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// \param blockRows number of rows in block as specified at run-time -/// \param blockCols number of columns in block as specified at run-time -/// -/// This function is mainly useful for blocks where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a blockRows should equal \a NRows unless -/// \a NRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_block_int_int_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_block_int_int_int_int.cpp -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -inline typename FixedBlockXpr::Type block(Index startRow, Index startCol, - Index blockRows, Index blockCols) -{ - return typename FixedBlockXpr::Type(derived(), startRow, startCol, blockRows, blockCols); -} - -/// This is the const version of block<>(Index, Index, Index, Index). -template -inline const typename ConstFixedBlockXpr::Type block(Index startRow, Index startCol, - Index blockRows, Index blockCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), startRow, startCol, blockRows, blockCols); -} - -/// \returns an expression of the \a i-th column of *this. Note that the numbering starts at 0. -/// -/// Example: \include MatrixBase_col.cpp -/// Output: \verbinclude MatrixBase_col.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/** - * \sa row(), class Block */ -EIGEN_DEVICE_FUNC -inline ColXpr col(Index i) -{ - return ColXpr(derived(), i); -} - -/// This is the const version of col(). -EIGEN_DEVICE_FUNC -inline ConstColXpr col(Index i) const -{ - return ConstColXpr(derived(), i); -} - -/// \returns an expression of the \a i-th row of *this. Note that the numbering starts at 0. -/// -/// Example: \include MatrixBase_row.cpp -/// Output: \verbinclude MatrixBase_row.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/** - * \sa col(), class Block */ -EIGEN_DEVICE_FUNC -inline RowXpr row(Index i) -{ - return RowXpr(derived(), i); -} - -/// This is the const version of row(). */ -EIGEN_DEVICE_FUNC -inline ConstRowXpr row(Index i) const -{ - return ConstRowXpr(derived(), i); -} - -/// \returns a dynamic-size expression of a segment (i.e. a vector block) in *this. -/// -/// \only_for_vectors -/// -/// \param start the first coefficient in the segment -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_segment_int_int.cpp -/// Output: \verbinclude MatrixBase_segment_int_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, segment(Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType segment(Index start, Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), start, n); -} - - -/// This is the const version of segment(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType segment(Index start, Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), start, n); -} - -/// \returns a dynamic-size expression of the first coefficients of *this. -/// -/// \only_for_vectors -/// -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_start_int.cpp -/// Output: \verbinclude MatrixBase_start_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType head(Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), 0, n); -} - -/// This is the const version of head(Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType head(Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), 0, n); -} - -/// \returns a dynamic-size expression of the last coefficients of *this. -/// -/// \only_for_vectors -/// -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_end_int.cpp -/// Output: \verbinclude MatrixBase_end_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType tail(Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), this->size() - n, n); -} - -/// This is the const version of tail(Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType tail(Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), this->size() - n, n); -} - -/// \returns a fixed-size expression of a segment (i.e. a vector block) in \c *this -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param start the index of the first element in the segment -/// \param n the number of coefficients in the segment as specified at compile-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_segment.cpp -/// Output: \verbinclude MatrixBase_template_int_segment.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type segment(Index start, Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), start, n); -} - -/// This is the const version of segment(Index). -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type segment(Index start, Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), start, n); -} - -/// \returns a fixed-size expression of the first coefficients of *this. -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param n the number of coefficients in the segment as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_start.cpp -/// Output: \verbinclude MatrixBase_template_int_start.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type head(Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), 0, n); -} - -/// This is the const version of head(). -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type head(Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), 0, n); -} - -/// \returns a fixed-size expression of the last coefficients of *this. -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param n the number of coefficients in the segment as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_end.cpp -/// Output: \verbinclude MatrixBase_template_int_end.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type tail(Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), size() - n); -} - -/// This is the const version of tail. -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type tail(Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), size() - n); -} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h deleted file mode 100644 index 8b6730ede02..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h +++ /dev/null @@ -1,115 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing common coefficient wise functions. - -/** \returns an expression of the difference of \c *this and \a other - * - * \note If you want to substract a given scalar from all coefficients, see Cwise::operator-(). - * - * \sa class CwiseBinaryOp, operator-=() - */ -EIGEN_MAKE_CWISE_BINARY_OP(operator-,difference) - -/** \returns an expression of the sum of \c *this and \a other - * - * \note If you want to add a given scalar to all coefficients, see Cwise::operator+(). - * - * \sa class CwiseBinaryOp, operator+=() - */ -EIGEN_MAKE_CWISE_BINARY_OP(operator+,sum) - -/** \returns an expression of a custom coefficient-wise operator \a func of *this and \a other - * - * The template parameter \a CustomBinaryOp is the type of the functor - * of the custom operator (see class CwiseBinaryOp for an example) - * - * Here is an example illustrating the use of custom functors: - * \include class_CwiseBinaryOp.cpp - * Output: \verbinclude class_CwiseBinaryOp.out - * - * \sa class CwiseBinaryOp, operator+(), operator-(), cwiseProduct() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp -binaryExpr(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other, const CustomBinaryOp& func = CustomBinaryOp()) const -{ - return CwiseBinaryOp(derived(), other.derived(), func); -} - - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator*,product) -#else -/** \returns an expression of \c *this scaled by the scalar factor \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template -const CwiseBinaryOp,Derived,Constant > operator*(const T& scalar) const; -/** \returns an expression of \a expr scaled by the scalar factor \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator*(const T& scalar, const StorageBaseType& expr); -#endif - - - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(operator/,quotient) -#else -/** \returns an expression of \c *this divided by the scalar value \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template -const CwiseBinaryOp,Derived,Constant > operator/(const T& scalar) const; -#endif - -/** \returns an expression of the coefficient-wise boolean \b and operator of \c *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_and.cpp - * Output: \verbinclude Cwise_boolean_and.out - * - * \sa operator||(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator&&(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} - -/** \returns an expression of the coefficient-wise boolean \b or operator of \c *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_or.cpp - * Output: \verbinclude Cwise_boolean_or.out - * - * \sa operator&&(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator||(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h deleted file mode 100644 index 89f4faaac6b..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h +++ /dev/null @@ -1,163 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing common coefficient wise functions. - -#ifndef EIGEN_PARSED_BY_DOXYGEN - -/** \internal the return type of conjugate() */ -typedef typename internal::conditional::IsComplex, - const CwiseUnaryOp, const Derived>, - const Derived& - >::type ConjugateReturnType; -/** \internal the return type of real() const */ -typedef typename internal::conditional::IsComplex, - const CwiseUnaryOp, const Derived>, - const Derived& - >::type RealReturnType; -/** \internal the return type of real() */ -typedef typename internal::conditional::IsComplex, - CwiseUnaryView, Derived>, - Derived& - >::type NonConstRealReturnType; -/** \internal the return type of imag() const */ -typedef CwiseUnaryOp, const Derived> ImagReturnType; -/** \internal the return type of imag() */ -typedef CwiseUnaryView, Derived> NonConstImagReturnType; - -typedef CwiseUnaryOp, const Derived> NegativeReturnType; - -#endif // not EIGEN_PARSED_BY_DOXYGEN - -/// \returns an expression of the opposite of \c *this -/// -EIGEN_DOC_UNARY_ADDONS(operator-,opposite) -/// -EIGEN_DEVICE_FUNC -inline const NegativeReturnType -operator-() const { return NegativeReturnType(derived()); } - - -template struct CastXpr { typedef typename internal::cast_return_type, const Derived> >::type Type; }; - -/// \returns an expression of \c *this with the \a Scalar type casted to -/// \a NewScalar. -/// -/// The template parameter \a NewScalar is the type we are casting the scalars to. -/// -EIGEN_DOC_UNARY_ADDONS(cast,conversion function) -/// -/// \sa class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -typename CastXpr::Type -cast() const -{ - return typename CastXpr::Type(derived()); -} - -/// \returns an expression of the complex conjugate of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(conjugate,complex conjugate) -/// -/// \sa Math functions, MatrixBase::adjoint() -EIGEN_DEVICE_FUNC -inline ConjugateReturnType -conjugate() const -{ - return ConjugateReturnType(derived()); -} - -/// \returns a read-only expression of the real part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(real,real part function) -/// -/// \sa imag() -EIGEN_DEVICE_FUNC -inline RealReturnType -real() const { return RealReturnType(derived()); } - -/// \returns an read-only expression of the imaginary part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(imag,imaginary part function) -/// -/// \sa real() -EIGEN_DEVICE_FUNC -inline const ImagReturnType -imag() const { return ImagReturnType(derived()); } - -/// \brief Apply a unary operator coefficient-wise -/// \param[in] func Functor implementing the unary operator -/// \tparam CustomUnaryOp Type of \a func -/// \returns An expression of a custom coefficient-wise unary operator \a func of *this -/// -/// The function \c ptr_fun() from the C++ standard library can be used to make functors out of normal functions. -/// -/// Example: -/// \include class_CwiseUnaryOp_ptrfun.cpp -/// Output: \verbinclude class_CwiseUnaryOp_ptrfun.out -/// -/// Genuine functors allow for more possibilities, for instance it may contain a state. -/// -/// Example: -/// \include class_CwiseUnaryOp.cpp -/// Output: \verbinclude class_CwiseUnaryOp.out -/// -EIGEN_DOC_UNARY_ADDONS(unaryExpr,unary function) -/// -/// \sa unaryViewExpr, binaryExpr, class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -inline const CwiseUnaryOp -unaryExpr(const CustomUnaryOp& func = CustomUnaryOp()) const -{ - return CwiseUnaryOp(derived(), func); -} - -/// \returns an expression of a custom coefficient-wise unary operator \a func of *this -/// -/// The template parameter \a CustomUnaryOp is the type of the functor -/// of the custom unary operator. -/// -/// Example: -/// \include class_CwiseUnaryOp.cpp -/// Output: \verbinclude class_CwiseUnaryOp.out -/// -EIGEN_DOC_UNARY_ADDONS(unaryViewExpr,unary function) -/// -/// \sa unaryExpr, binaryExpr class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -inline const CwiseUnaryView -unaryViewExpr(const CustomViewOp& func = CustomViewOp()) const -{ - return CwiseUnaryView(derived(), func); -} - -/// \returns a non const expression of the real part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(real,real part function) -/// -/// \sa imag() -EIGEN_DEVICE_FUNC -inline NonConstRealReturnType -real() { return NonConstRealReturnType(derived()); } - -/// \returns a non const expression of the imaginary part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(imag,imaginary part function) -/// -/// \sa real() -EIGEN_DEVICE_FUNC -inline NonConstImagReturnType -imag() { return NonConstImagReturnType(derived()); } diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h deleted file mode 100644 index f1084abefbc..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h +++ /dev/null @@ -1,152 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing matrix specifics coefficient wise functions. - -/** \returns an expression of the Schur product (coefficient wise product) of *this and \a other - * - * Example: \include MatrixBase_cwiseProduct.cpp - * Output: \verbinclude MatrixBase_cwiseProduct.out - * - * \sa class CwiseBinaryOp, cwiseAbs2 - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product) -cwiseProduct(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product)(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise == operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include MatrixBase_cwiseEqual.cpp - * Output: \verbinclude MatrixBase_cwiseEqual.out - * - * \sa cwiseNotEqual(), isApprox(), isMuchSmallerThan() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseEqual(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise != operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include MatrixBase_cwiseNotEqual.cpp - * Output: \verbinclude MatrixBase_cwiseNotEqual.out - * - * \sa cwiseEqual(), isApprox(), isMuchSmallerThan() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseNotEqual(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of *this and \a other - * - * Example: \include MatrixBase_cwiseMin.cpp - * Output: \verbinclude MatrixBase_cwiseMin.out - * - * \sa class CwiseBinaryOp, max() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseMin(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of *this and scalar \a other - * - * \sa class CwiseBinaryOp, min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const ConstantReturnType> -cwiseMin(const Scalar &other) const -{ - return cwiseMin(Derived::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise max of *this and \a other - * - * Example: \include MatrixBase_cwiseMax.cpp - * Output: \verbinclude MatrixBase_cwiseMax.out - * - * \sa class CwiseBinaryOp, min() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseMax(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise max of *this and scalar \a other - * - * \sa class CwiseBinaryOp, min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const ConstantReturnType> -cwiseMax(const Scalar &other) const -{ - return cwiseMax(Derived::Constant(rows(), cols(), other)); -} - - -/** \returns an expression of the coefficient-wise quotient of *this and \a other - * - * Example: \include MatrixBase_cwiseQuotient.cpp - * Output: \verbinclude MatrixBase_cwiseQuotient.out - * - * \sa class CwiseBinaryOp, cwiseProduct(), cwiseInverse() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseQuotient(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -typedef CwiseBinaryOp, const Derived, const ConstantReturnType> CwiseScalarEqualReturnType; - -/** \returns an expression of the coefficient-wise == operator of \c *this and a scalar \a s - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * \sa cwiseEqual(const MatrixBase &) const - */ -EIGEN_DEVICE_FUNC -inline const CwiseScalarEqualReturnType -cwiseEqual(const Scalar& s) const -{ - return CwiseScalarEqualReturnType(derived(), Derived::Constant(rows(), cols(), s), internal::scalar_cmp_op()); -} diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h deleted file mode 100644 index b1be3d566c6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h +++ /dev/null @@ -1,85 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is included into the body of the base classes supporting matrix specific coefficient-wise functions. -// This include MatrixBase and SparseMatrixBase. - - -typedef CwiseUnaryOp, const Derived> CwiseAbsReturnType; -typedef CwiseUnaryOp, const Derived> CwiseAbs2ReturnType; -typedef CwiseUnaryOp, const Derived> CwiseSqrtReturnType; -typedef CwiseUnaryOp, const Derived> CwiseSignReturnType; -typedef CwiseUnaryOp, const Derived> CwiseInverseReturnType; - -/// \returns an expression of the coefficient-wise absolute value of \c *this -/// -/// Example: \include MatrixBase_cwiseAbs.cpp -/// Output: \verbinclude MatrixBase_cwiseAbs.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseAbs,absolute value) -/// -/// \sa cwiseAbs2() -/// -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseAbsReturnType -cwiseAbs() const { return CwiseAbsReturnType(derived()); } - -/// \returns an expression of the coefficient-wise squared absolute value of \c *this -/// -/// Example: \include MatrixBase_cwiseAbs2.cpp -/// Output: \verbinclude MatrixBase_cwiseAbs2.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseAbs2,squared absolute value) -/// -/// \sa cwiseAbs() -/// -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseAbs2ReturnType -cwiseAbs2() const { return CwiseAbs2ReturnType(derived()); } - -/// \returns an expression of the coefficient-wise square root of *this. -/// -/// Example: \include MatrixBase_cwiseSqrt.cpp -/// Output: \verbinclude MatrixBase_cwiseSqrt.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseSqrt,square-root) -/// -/// \sa cwisePow(), cwiseSquare() -/// -EIGEN_DEVICE_FUNC -inline const CwiseSqrtReturnType -cwiseSqrt() const { return CwiseSqrtReturnType(derived()); } - -/// \returns an expression of the coefficient-wise signum of *this. -/// -/// Example: \include MatrixBase_cwiseSign.cpp -/// Output: \verbinclude MatrixBase_cwiseSign.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseSign,sign function) -/// -EIGEN_DEVICE_FUNC -inline const CwiseSignReturnType -cwiseSign() const { return CwiseSignReturnType(derived()); } - - -/// \returns an expression of the coefficient-wise inverse of *this. -/// -/// Example: \include MatrixBase_cwiseInverse.cpp -/// Output: \verbinclude MatrixBase_cwiseInverse.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseInverse,inverse) -/// -/// \sa cwiseProduct() -/// -EIGEN_DEVICE_FUNC -inline const CwiseInverseReturnType -cwiseInverse() const { return CwiseInverseReturnType(derived()); } - - diff --git a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/README.md b/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/README.md deleted file mode 100644 index 4654a81c372..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/eigen-3.3.9/README.md +++ /dev/null @@ -1,3 +0,0 @@ -**Eigen is a C++ template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms.** - -For more information go to http://eigen.tuxfamily.org/. diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/adl_serializer.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/adl_serializer.hpp deleted file mode 100644 index 4af1c4bb1d0..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/adl_serializer.hpp +++ /dev/null @@ -1,49 +0,0 @@ -#pragma once - -#include - -#include -#include - -namespace nlohmann -{ - -template -struct adl_serializer -{ - /*! - @brief convert a JSON value to any value type - - This function is usually called by the `get()` function of the - @ref basic_json class (either explicit or via conversion operators). - - @param[in] j JSON value to read from - @param[in,out] val value to write to - */ - template - static auto from_json(BasicJsonType&& j, ValueType& val) noexcept( - noexcept(::nlohmann::from_json(std::forward(j), val))) - -> decltype(::nlohmann::from_json(std::forward(j), val), void()) - { - ::nlohmann::from_json(std::forward(j), val); - } - - /*! - @brief convert any value type to a JSON value - - This function is usually called by the constructors of the @ref basic_json - class. - - @param[in,out] j JSON value to write to - @param[in] val value to read from - */ - template - static auto to_json(BasicJsonType& j, ValueType&& val) noexcept( - noexcept(::nlohmann::to_json(j, std::forward(val)))) - -> decltype(::nlohmann::to_json(j, std::forward(val)), void()) - { - ::nlohmann::to_json(j, std::forward(val)); - } -}; - -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/byte_container_with_subtype.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/byte_container_with_subtype.hpp deleted file mode 100644 index 69f9feb2127..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/byte_container_with_subtype.hpp +++ /dev/null @@ -1,166 +0,0 @@ -#pragma once - -#include // uint8_t -#include // tie -#include // move - -namespace nlohmann -{ - -/*! -@brief an internal type for a backed binary type - -This type extends the template parameter @a BinaryType provided to `basic_json` -with a subtype used by BSON and MessagePack. This type exists so that the user -does not have to specify a type themselves with a specific naming scheme in -order to override the binary type. - -@tparam BinaryType container to store bytes (`std::vector` by - default) - -@since version 3.8.0 -*/ -template -class byte_container_with_subtype : public BinaryType -{ - public: - /// the type of the underlying container - using container_type = BinaryType; - - byte_container_with_subtype() noexcept(noexcept(container_type())) - : container_type() - {} - - byte_container_with_subtype(const container_type& b) noexcept(noexcept(container_type(b))) - : container_type(b) - {} - - byte_container_with_subtype(container_type&& b) noexcept(noexcept(container_type(std::move(b)))) - : container_type(std::move(b)) - {} - - byte_container_with_subtype(const container_type& b, std::uint8_t subtype) noexcept(noexcept(container_type(b))) - : container_type(b) - , m_subtype(subtype) - , m_has_subtype(true) - {} - - byte_container_with_subtype(container_type&& b, std::uint8_t subtype) noexcept(noexcept(container_type(std::move(b)))) - : container_type(std::move(b)) - , m_subtype(subtype) - , m_has_subtype(true) - {} - - bool operator==(const byte_container_with_subtype& rhs) const - { - return std::tie(static_cast(*this), m_subtype, m_has_subtype) == - std::tie(static_cast(rhs), rhs.m_subtype, rhs.m_has_subtype); - } - - bool operator!=(const byte_container_with_subtype& rhs) const - { - return !(rhs == *this); - } - - /*! - @brief sets the binary subtype - - Sets the binary subtype of the value, also flags a binary JSON value as - having a subtype, which has implications for serialization. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @sa @ref subtype() -- return the binary subtype - @sa @ref clear_subtype() -- clears the binary subtype - @sa @ref has_subtype() -- returns whether or not the binary value has a - subtype - - @since version 3.8.0 - */ - void set_subtype(std::uint8_t subtype) noexcept - { - m_subtype = subtype; - m_has_subtype = true; - } - - /*! - @brief return the binary subtype - - Returns the numerical subtype of the value if it has a subtype. If it does - not have a subtype, this function will return size_t(-1) as a sentinel - value. - - @return the numerical subtype of the binary value - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @sa @ref set_subtype() -- sets the binary subtype - @sa @ref clear_subtype() -- clears the binary subtype - @sa @ref has_subtype() -- returns whether or not the binary value has a - subtype - - @since version 3.8.0 - */ - constexpr std::uint8_t subtype() const noexcept - { - return m_subtype; - } - - /*! - @brief return whether the value has a subtype - - @return whether the value has a subtype - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @sa @ref subtype() -- return the binary subtype - @sa @ref set_subtype() -- sets the binary subtype - @sa @ref clear_subtype() -- clears the binary subtype - - @since version 3.8.0 - */ - constexpr bool has_subtype() const noexcept - { - return m_has_subtype; - } - - /*! - @brief clears the binary subtype - - Clears the binary subtype and flags the value as not having a subtype, which - has implications for serialization; for instance MessagePack will prefer the - bin family over the ext family. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @sa @ref subtype() -- return the binary subtype - @sa @ref set_subtype() -- sets the binary subtype - @sa @ref has_subtype() -- returns whether or not the binary value has a - subtype - - @since version 3.8.0 - */ - void clear_subtype() noexcept - { - m_subtype = 0; - m_has_subtype = false; - } - - private: - std::uint8_t m_subtype = 0; - bool m_has_subtype = false; -}; - -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/from_json.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/from_json.hpp deleted file mode 100644 index 438b84a2e1f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/from_json.hpp +++ /dev/null @@ -1,403 +0,0 @@ -#pragma once - -#include // transform -#include // array -#include // forward_list -#include // inserter, front_inserter, end -#include // map -#include // string -#include // tuple, make_tuple -#include // is_arithmetic, is_same, is_enum, underlying_type, is_convertible -#include // unordered_map -#include // pair, declval -#include // valarray - -#include -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -template -void from_json(const BasicJsonType& j, typename std::nullptr_t& n) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_null())) - { - JSON_THROW(type_error::create(302, "type must be null, but is " + std::string(j.type_name()))); - } - n = nullptr; -} - -// overloads for basic_json template parameters -template < typename BasicJsonType, typename ArithmeticType, - enable_if_t < std::is_arithmetic::value&& - !std::is_same::value, - int > = 0 > -void get_arithmetic_value(const BasicJsonType& j, ArithmeticType& val) -{ - switch (static_cast(j)) - { - case value_t::number_unsigned: - { - val = static_cast(*j.template get_ptr()); - break; - } - case value_t::number_integer: - { - val = static_cast(*j.template get_ptr()); - break; - } - case value_t::number_float: - { - val = static_cast(*j.template get_ptr()); - break; - } - - default: - JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name()))); - } -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::boolean_t& b) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_boolean())) - { - JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(j.type_name()))); - } - b = *j.template get_ptr(); -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::string_t& s) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_string())) - { - JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name()))); - } - s = *j.template get_ptr(); -} - -template < - typename BasicJsonType, typename ConstructibleStringType, - enable_if_t < - is_constructible_string_type::value&& - !std::is_same::value, - int > = 0 > -void from_json(const BasicJsonType& j, ConstructibleStringType& s) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_string())) - { - JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name()))); - } - - s = *j.template get_ptr(); -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::number_float_t& val) -{ - get_arithmetic_value(j, val); -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::number_unsigned_t& val) -{ - get_arithmetic_value(j, val); -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::number_integer_t& val) -{ - get_arithmetic_value(j, val); -} - -template::value, int> = 0> -void from_json(const BasicJsonType& j, EnumType& e) -{ - typename std::underlying_type::type val; - get_arithmetic_value(j, val); - e = static_cast(val); -} - -// forward_list doesn't have an insert method -template::value, int> = 0> -void from_json(const BasicJsonType& j, std::forward_list& l) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()))); - } - l.clear(); - std::transform(j.rbegin(), j.rend(), - std::front_inserter(l), [](const BasicJsonType & i) - { - return i.template get(); - }); -} - -// valarray doesn't have an insert method -template::value, int> = 0> -void from_json(const BasicJsonType& j, std::valarray& l) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()))); - } - l.resize(j.size()); - std::transform(j.begin(), j.end(), std::begin(l), - [](const BasicJsonType & elem) - { - return elem.template get(); - }); -} - -template -auto from_json(const BasicJsonType& j, T (&arr)[N]) --> decltype(j.template get(), void()) -{ - for (std::size_t i = 0; i < N; ++i) - { - arr[i] = j.at(i).template get(); - } -} - -template -void from_json_array_impl(const BasicJsonType& j, typename BasicJsonType::array_t& arr, priority_tag<3> /*unused*/) -{ - arr = *j.template get_ptr(); -} - -template -auto from_json_array_impl(const BasicJsonType& j, std::array& arr, - priority_tag<2> /*unused*/) --> decltype(j.template get(), void()) -{ - for (std::size_t i = 0; i < N; ++i) - { - arr[i] = j.at(i).template get(); - } -} - -template -auto from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr, priority_tag<1> /*unused*/) --> decltype( - arr.reserve(std::declval()), - j.template get(), - void()) -{ - using std::end; - - ConstructibleArrayType ret; - ret.reserve(j.size()); - std::transform(j.begin(), j.end(), - std::inserter(ret, end(ret)), [](const BasicJsonType & i) - { - // get() returns *this, this won't call a from_json - // method when value_type is BasicJsonType - return i.template get(); - }); - arr = std::move(ret); -} - -template -void from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr, - priority_tag<0> /*unused*/) -{ - using std::end; - - ConstructibleArrayType ret; - std::transform( - j.begin(), j.end(), std::inserter(ret, end(ret)), - [](const BasicJsonType & i) - { - // get() returns *this, this won't call a from_json - // method when value_type is BasicJsonType - return i.template get(); - }); - arr = std::move(ret); -} - -template < typename BasicJsonType, typename ConstructibleArrayType, - enable_if_t < - is_constructible_array_type::value&& - !is_constructible_object_type::value&& - !is_constructible_string_type::value&& - !std::is_same::value&& - !is_basic_json::value, - int > = 0 > -auto from_json(const BasicJsonType& j, ConstructibleArrayType& arr) --> decltype(from_json_array_impl(j, arr, priority_tag<3> {}), -j.template get(), -void()) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + - std::string(j.type_name()))); - } - - from_json_array_impl(j, arr, priority_tag<3> {}); -} - -template -void from_json(const BasicJsonType& j, typename BasicJsonType::binary_t& bin) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_binary())) - { - JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(j.type_name()))); - } - - bin = *j.template get_ptr(); -} - -template::value, int> = 0> -void from_json(const BasicJsonType& j, ConstructibleObjectType& obj) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_object())) - { - JSON_THROW(type_error::create(302, "type must be object, but is " + std::string(j.type_name()))); - } - - ConstructibleObjectType ret; - auto inner_object = j.template get_ptr(); - using value_type = typename ConstructibleObjectType::value_type; - std::transform( - inner_object->begin(), inner_object->end(), - std::inserter(ret, ret.begin()), - [](typename BasicJsonType::object_t::value_type const & p) - { - return value_type(p.first, p.second.template get()); - }); - obj = std::move(ret); -} - -// overload for arithmetic types, not chosen for basic_json template arguments -// (BooleanType, etc..); note: Is it really necessary to provide explicit -// overloads for boolean_t etc. in case of a custom BooleanType which is not -// an arithmetic type? -template < typename BasicJsonType, typename ArithmeticType, - enable_if_t < - std::is_arithmetic::value&& - !std::is_same::value&& - !std::is_same::value&& - !std::is_same::value&& - !std::is_same::value, - int > = 0 > -void from_json(const BasicJsonType& j, ArithmeticType& val) -{ - switch (static_cast(j)) - { - case value_t::number_unsigned: - { - val = static_cast(*j.template get_ptr()); - break; - } - case value_t::number_integer: - { - val = static_cast(*j.template get_ptr()); - break; - } - case value_t::number_float: - { - val = static_cast(*j.template get_ptr()); - break; - } - case value_t::boolean: - { - val = static_cast(*j.template get_ptr()); - break; - } - - default: - JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name()))); - } -} - -template -void from_json(const BasicJsonType& j, std::pair& p) -{ - p = {j.at(0).template get(), j.at(1).template get()}; -} - -template -void from_json_tuple_impl(const BasicJsonType& j, Tuple& t, index_sequence /*unused*/) -{ - t = std::make_tuple(j.at(Idx).template get::type>()...); -} - -template -void from_json(const BasicJsonType& j, std::tuple& t) -{ - from_json_tuple_impl(j, t, index_sequence_for {}); -} - -template < typename BasicJsonType, typename Key, typename Value, typename Compare, typename Allocator, - typename = enable_if_t < !std::is_constructible < - typename BasicJsonType::string_t, Key >::value >> -void from_json(const BasicJsonType& j, std::map& m) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()))); - } - m.clear(); - for (const auto& p : j) - { - if (JSON_HEDLEY_UNLIKELY(!p.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name()))); - } - m.emplace(p.at(0).template get(), p.at(1).template get()); - } -} - -template < typename BasicJsonType, typename Key, typename Value, typename Hash, typename KeyEqual, typename Allocator, - typename = enable_if_t < !std::is_constructible < - typename BasicJsonType::string_t, Key >::value >> -void from_json(const BasicJsonType& j, std::unordered_map& m) -{ - if (JSON_HEDLEY_UNLIKELY(!j.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()))); - } - m.clear(); - for (const auto& p : j) - { - if (JSON_HEDLEY_UNLIKELY(!p.is_array())) - { - JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name()))); - } - m.emplace(p.at(0).template get(), p.at(1).template get()); - } -} - -struct from_json_fn -{ - template - auto operator()(const BasicJsonType& j, T& val) const - noexcept(noexcept(from_json(j, val))) - -> decltype(from_json(j, val), void()) - { - return from_json(j, val); - } -}; -} // namespace detail - -/// namespace to hold default `from_json` function -/// to see why this is required: -/// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4381.html -namespace -{ -constexpr const auto& from_json = detail::static_const::value; -} // namespace -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_chars.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_chars.hpp deleted file mode 100644 index c632ff2bea6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_chars.hpp +++ /dev/null @@ -1,1105 +0,0 @@ -#pragma once - -#include // array -#include // signbit, isfinite -#include // intN_t, uintN_t -#include // memcpy, memmove -#include // numeric_limits -#include // conditional - -#include - -namespace nlohmann -{ -namespace detail -{ - -/*! -@brief implements the Grisu2 algorithm for binary to decimal floating-point -conversion. - -This implementation is a slightly modified version of the reference -implementation which may be obtained from -http://florian.loitsch.com/publications (bench.tar.gz). - -The code is distributed under the MIT license, Copyright (c) 2009 Florian Loitsch. - -For a detailed description of the algorithm see: - -[1] Loitsch, "Printing Floating-Point Numbers Quickly and Accurately with - Integers", Proceedings of the ACM SIGPLAN 2010 Conference on Programming - Language Design and Implementation, PLDI 2010 -[2] Burger, Dybvig, "Printing Floating-Point Numbers Quickly and Accurately", - Proceedings of the ACM SIGPLAN 1996 Conference on Programming Language - Design and Implementation, PLDI 1996 -*/ -namespace dtoa_impl -{ - -template -Target reinterpret_bits(const Source source) -{ - static_assert(sizeof(Target) == sizeof(Source), "size mismatch"); - - Target target; - std::memcpy(&target, &source, sizeof(Source)); - return target; -} - -struct diyfp // f * 2^e -{ - static constexpr int kPrecision = 64; // = q - - std::uint64_t f = 0; - int e = 0; - - constexpr diyfp(std::uint64_t f_, int e_) noexcept : f(f_), e(e_) {} - - /*! - @brief returns x - y - @pre x.e == y.e and x.f >= y.f - */ - static diyfp sub(const diyfp& x, const diyfp& y) noexcept - { - JSON_ASSERT(x.e == y.e); - JSON_ASSERT(x.f >= y.f); - - return {x.f - y.f, x.e}; - } - - /*! - @brief returns x * y - @note The result is rounded. (Only the upper q bits are returned.) - */ - static diyfp mul(const diyfp& x, const diyfp& y) noexcept - { - static_assert(kPrecision == 64, "internal error"); - - // Computes: - // f = round((x.f * y.f) / 2^q) - // e = x.e + y.e + q - - // Emulate the 64-bit * 64-bit multiplication: - // - // p = u * v - // = (u_lo + 2^32 u_hi) (v_lo + 2^32 v_hi) - // = (u_lo v_lo ) + 2^32 ((u_lo v_hi ) + (u_hi v_lo )) + 2^64 (u_hi v_hi ) - // = (p0 ) + 2^32 ((p1 ) + (p2 )) + 2^64 (p3 ) - // = (p0_lo + 2^32 p0_hi) + 2^32 ((p1_lo + 2^32 p1_hi) + (p2_lo + 2^32 p2_hi)) + 2^64 (p3 ) - // = (p0_lo ) + 2^32 (p0_hi + p1_lo + p2_lo ) + 2^64 (p1_hi + p2_hi + p3) - // = (p0_lo ) + 2^32 (Q ) + 2^64 (H ) - // = (p0_lo ) + 2^32 (Q_lo + 2^32 Q_hi ) + 2^64 (H ) - // - // (Since Q might be larger than 2^32 - 1) - // - // = (p0_lo + 2^32 Q_lo) + 2^64 (Q_hi + H) - // - // (Q_hi + H does not overflow a 64-bit int) - // - // = p_lo + 2^64 p_hi - - const std::uint64_t u_lo = x.f & 0xFFFFFFFFu; - const std::uint64_t u_hi = x.f >> 32u; - const std::uint64_t v_lo = y.f & 0xFFFFFFFFu; - const std::uint64_t v_hi = y.f >> 32u; - - const std::uint64_t p0 = u_lo * v_lo; - const std::uint64_t p1 = u_lo * v_hi; - const std::uint64_t p2 = u_hi * v_lo; - const std::uint64_t p3 = u_hi * v_hi; - - const std::uint64_t p0_hi = p0 >> 32u; - const std::uint64_t p1_lo = p1 & 0xFFFFFFFFu; - const std::uint64_t p1_hi = p1 >> 32u; - const std::uint64_t p2_lo = p2 & 0xFFFFFFFFu; - const std::uint64_t p2_hi = p2 >> 32u; - - std::uint64_t Q = p0_hi + p1_lo + p2_lo; - - // The full product might now be computed as - // - // p_hi = p3 + p2_hi + p1_hi + (Q >> 32) - // p_lo = p0_lo + (Q << 32) - // - // But in this particular case here, the full p_lo is not required. - // Effectively we only need to add the highest bit in p_lo to p_hi (and - // Q_hi + 1 does not overflow). - - Q += std::uint64_t{1} << (64u - 32u - 1u); // round, ties up - - const std::uint64_t h = p3 + p2_hi + p1_hi + (Q >> 32u); - - return {h, x.e + y.e + 64}; - } - - /*! - @brief normalize x such that the significand is >= 2^(q-1) - @pre x.f != 0 - */ - static diyfp normalize(diyfp x) noexcept - { - JSON_ASSERT(x.f != 0); - - while ((x.f >> 63u) == 0) - { - x.f <<= 1u; - x.e--; - } - - return x; - } - - /*! - @brief normalize x such that the result has the exponent E - @pre e >= x.e and the upper e - x.e bits of x.f must be zero. - */ - static diyfp normalize_to(const diyfp& x, const int target_exponent) noexcept - { - const int delta = x.e - target_exponent; - - JSON_ASSERT(delta >= 0); - JSON_ASSERT(((x.f << delta) >> delta) == x.f); - - return {x.f << delta, target_exponent}; - } -}; - -struct boundaries -{ - diyfp w; - diyfp minus; - diyfp plus; -}; - -/*! -Compute the (normalized) diyfp representing the input number 'value' and its -boundaries. - -@pre value must be finite and positive -*/ -template -boundaries compute_boundaries(FloatType value) -{ - JSON_ASSERT(std::isfinite(value)); - JSON_ASSERT(value > 0); - - // Convert the IEEE representation into a diyfp. - // - // If v is denormal: - // value = 0.F * 2^(1 - bias) = ( F) * 2^(1 - bias - (p-1)) - // If v is normalized: - // value = 1.F * 2^(E - bias) = (2^(p-1) + F) * 2^(E - bias - (p-1)) - - static_assert(std::numeric_limits::is_iec559, - "internal error: dtoa_short requires an IEEE-754 floating-point implementation"); - - constexpr int kPrecision = std::numeric_limits::digits; // = p (includes the hidden bit) - constexpr int kBias = std::numeric_limits::max_exponent - 1 + (kPrecision - 1); - constexpr int kMinExp = 1 - kBias; - constexpr std::uint64_t kHiddenBit = std::uint64_t{1} << (kPrecision - 1); // = 2^(p-1) - - using bits_type = typename std::conditional::type; - - const std::uint64_t bits = reinterpret_bits(value); - const std::uint64_t E = bits >> (kPrecision - 1); - const std::uint64_t F = bits & (kHiddenBit - 1); - - const bool is_denormal = E == 0; - const diyfp v = is_denormal - ? diyfp(F, kMinExp) - : diyfp(F + kHiddenBit, static_cast(E) - kBias); - - // Compute the boundaries m- and m+ of the floating-point value - // v = f * 2^e. - // - // Determine v- and v+, the floating-point predecessor and successor if v, - // respectively. - // - // v- = v - 2^e if f != 2^(p-1) or e == e_min (A) - // = v - 2^(e-1) if f == 2^(p-1) and e > e_min (B) - // - // v+ = v + 2^e - // - // Let m- = (v- + v) / 2 and m+ = (v + v+) / 2. All real numbers _strictly_ - // between m- and m+ round to v, regardless of how the input rounding - // algorithm breaks ties. - // - // ---+-------------+-------------+-------------+-------------+--- (A) - // v- m- v m+ v+ - // - // -----------------+------+------+-------------+-------------+--- (B) - // v- m- v m+ v+ - - const bool lower_boundary_is_closer = F == 0 && E > 1; - const diyfp m_plus = diyfp(2 * v.f + 1, v.e - 1); - const diyfp m_minus = lower_boundary_is_closer - ? diyfp(4 * v.f - 1, v.e - 2) // (B) - : diyfp(2 * v.f - 1, v.e - 1); // (A) - - // Determine the normalized w+ = m+. - const diyfp w_plus = diyfp::normalize(m_plus); - - // Determine w- = m- such that e_(w-) = e_(w+). - const diyfp w_minus = diyfp::normalize_to(m_minus, w_plus.e); - - return {diyfp::normalize(v), w_minus, w_plus}; -} - -// Given normalized diyfp w, Grisu needs to find a (normalized) cached -// power-of-ten c, such that the exponent of the product c * w = f * 2^e lies -// within a certain range [alpha, gamma] (Definition 3.2 from [1]) -// -// alpha <= e = e_c + e_w + q <= gamma -// -// or -// -// f_c * f_w * 2^alpha <= f_c 2^(e_c) * f_w 2^(e_w) * 2^q -// <= f_c * f_w * 2^gamma -// -// Since c and w are normalized, i.e. 2^(q-1) <= f < 2^q, this implies -// -// 2^(q-1) * 2^(q-1) * 2^alpha <= c * w * 2^q < 2^q * 2^q * 2^gamma -// -// or -// -// 2^(q - 2 + alpha) <= c * w < 2^(q + gamma) -// -// The choice of (alpha,gamma) determines the size of the table and the form of -// the digit generation procedure. Using (alpha,gamma)=(-60,-32) works out well -// in practice: -// -// The idea is to cut the number c * w = f * 2^e into two parts, which can be -// processed independently: An integral part p1, and a fractional part p2: -// -// f * 2^e = ( (f div 2^-e) * 2^-e + (f mod 2^-e) ) * 2^e -// = (f div 2^-e) + (f mod 2^-e) * 2^e -// = p1 + p2 * 2^e -// -// The conversion of p1 into decimal form requires a series of divisions and -// modulos by (a power of) 10. These operations are faster for 32-bit than for -// 64-bit integers, so p1 should ideally fit into a 32-bit integer. This can be -// achieved by choosing -// -// -e >= 32 or e <= -32 := gamma -// -// In order to convert the fractional part -// -// p2 * 2^e = p2 / 2^-e = d[-1] / 10^1 + d[-2] / 10^2 + ... -// -// into decimal form, the fraction is repeatedly multiplied by 10 and the digits -// d[-i] are extracted in order: -// -// (10 * p2) div 2^-e = d[-1] -// (10 * p2) mod 2^-e = d[-2] / 10^1 + ... -// -// The multiplication by 10 must not overflow. It is sufficient to choose -// -// 10 * p2 < 16 * p2 = 2^4 * p2 <= 2^64. -// -// Since p2 = f mod 2^-e < 2^-e, -// -// -e <= 60 or e >= -60 := alpha - -constexpr int kAlpha = -60; -constexpr int kGamma = -32; - -struct cached_power // c = f * 2^e ~= 10^k -{ - std::uint64_t f; - int e; - int k; -}; - -/*! -For a normalized diyfp w = f * 2^e, this function returns a (normalized) cached -power-of-ten c = f_c * 2^e_c, such that the exponent of the product w * c -satisfies (Definition 3.2 from [1]) - - alpha <= e_c + e + q <= gamma. -*/ -inline cached_power get_cached_power_for_binary_exponent(int e) -{ - // Now - // - // alpha <= e_c + e + q <= gamma (1) - // ==> f_c * 2^alpha <= c * 2^e * 2^q - // - // and since the c's are normalized, 2^(q-1) <= f_c, - // - // ==> 2^(q - 1 + alpha) <= c * 2^(e + q) - // ==> 2^(alpha - e - 1) <= c - // - // If c were an exact power of ten, i.e. c = 10^k, one may determine k as - // - // k = ceil( log_10( 2^(alpha - e - 1) ) ) - // = ceil( (alpha - e - 1) * log_10(2) ) - // - // From the paper: - // "In theory the result of the procedure could be wrong since c is rounded, - // and the computation itself is approximated [...]. In practice, however, - // this simple function is sufficient." - // - // For IEEE double precision floating-point numbers converted into - // normalized diyfp's w = f * 2^e, with q = 64, - // - // e >= -1022 (min IEEE exponent) - // -52 (p - 1) - // -52 (p - 1, possibly normalize denormal IEEE numbers) - // -11 (normalize the diyfp) - // = -1137 - // - // and - // - // e <= +1023 (max IEEE exponent) - // -52 (p - 1) - // -11 (normalize the diyfp) - // = 960 - // - // This binary exponent range [-1137,960] results in a decimal exponent - // range [-307,324]. One does not need to store a cached power for each - // k in this range. For each such k it suffices to find a cached power - // such that the exponent of the product lies in [alpha,gamma]. - // This implies that the difference of the decimal exponents of adjacent - // table entries must be less than or equal to - // - // floor( (gamma - alpha) * log_10(2) ) = 8. - // - // (A smaller distance gamma-alpha would require a larger table.) - - // NB: - // Actually this function returns c, such that -60 <= e_c + e + 64 <= -34. - - constexpr int kCachedPowersMinDecExp = -300; - constexpr int kCachedPowersDecStep = 8; - - static constexpr std::array kCachedPowers = - { - { - { 0xAB70FE17C79AC6CA, -1060, -300 }, - { 0xFF77B1FCBEBCDC4F, -1034, -292 }, - { 0xBE5691EF416BD60C, -1007, -284 }, - { 0x8DD01FAD907FFC3C, -980, -276 }, - { 0xD3515C2831559A83, -954, -268 }, - { 0x9D71AC8FADA6C9B5, -927, -260 }, - { 0xEA9C227723EE8BCB, -901, -252 }, - { 0xAECC49914078536D, -874, -244 }, - { 0x823C12795DB6CE57, -847, -236 }, - { 0xC21094364DFB5637, -821, -228 }, - { 0x9096EA6F3848984F, -794, -220 }, - { 0xD77485CB25823AC7, -768, -212 }, - { 0xA086CFCD97BF97F4, -741, -204 }, - { 0xEF340A98172AACE5, -715, -196 }, - { 0xB23867FB2A35B28E, -688, -188 }, - { 0x84C8D4DFD2C63F3B, -661, -180 }, - { 0xC5DD44271AD3CDBA, -635, -172 }, - { 0x936B9FCEBB25C996, -608, -164 }, - { 0xDBAC6C247D62A584, -582, -156 }, - { 0xA3AB66580D5FDAF6, -555, -148 }, - { 0xF3E2F893DEC3F126, -529, -140 }, - { 0xB5B5ADA8AAFF80B8, -502, -132 }, - { 0x87625F056C7C4A8B, -475, -124 }, - { 0xC9BCFF6034C13053, -449, -116 }, - { 0x964E858C91BA2655, -422, -108 }, - { 0xDFF9772470297EBD, -396, -100 }, - { 0xA6DFBD9FB8E5B88F, -369, -92 }, - { 0xF8A95FCF88747D94, -343, -84 }, - { 0xB94470938FA89BCF, -316, -76 }, - { 0x8A08F0F8BF0F156B, -289, -68 }, - { 0xCDB02555653131B6, -263, -60 }, - { 0x993FE2C6D07B7FAC, -236, -52 }, - { 0xE45C10C42A2B3B06, -210, -44 }, - { 0xAA242499697392D3, -183, -36 }, - { 0xFD87B5F28300CA0E, -157, -28 }, - { 0xBCE5086492111AEB, -130, -20 }, - { 0x8CBCCC096F5088CC, -103, -12 }, - { 0xD1B71758E219652C, -77, -4 }, - { 0x9C40000000000000, -50, 4 }, - { 0xE8D4A51000000000, -24, 12 }, - { 0xAD78EBC5AC620000, 3, 20 }, - { 0x813F3978F8940984, 30, 28 }, - { 0xC097CE7BC90715B3, 56, 36 }, - { 0x8F7E32CE7BEA5C70, 83, 44 }, - { 0xD5D238A4ABE98068, 109, 52 }, - { 0x9F4F2726179A2245, 136, 60 }, - { 0xED63A231D4C4FB27, 162, 68 }, - { 0xB0DE65388CC8ADA8, 189, 76 }, - { 0x83C7088E1AAB65DB, 216, 84 }, - { 0xC45D1DF942711D9A, 242, 92 }, - { 0x924D692CA61BE758, 269, 100 }, - { 0xDA01EE641A708DEA, 295, 108 }, - { 0xA26DA3999AEF774A, 322, 116 }, - { 0xF209787BB47D6B85, 348, 124 }, - { 0xB454E4A179DD1877, 375, 132 }, - { 0x865B86925B9BC5C2, 402, 140 }, - { 0xC83553C5C8965D3D, 428, 148 }, - { 0x952AB45CFA97A0B3, 455, 156 }, - { 0xDE469FBD99A05FE3, 481, 164 }, - { 0xA59BC234DB398C25, 508, 172 }, - { 0xF6C69A72A3989F5C, 534, 180 }, - { 0xB7DCBF5354E9BECE, 561, 188 }, - { 0x88FCF317F22241E2, 588, 196 }, - { 0xCC20CE9BD35C78A5, 614, 204 }, - { 0x98165AF37B2153DF, 641, 212 }, - { 0xE2A0B5DC971F303A, 667, 220 }, - { 0xA8D9D1535CE3B396, 694, 228 }, - { 0xFB9B7CD9A4A7443C, 720, 236 }, - { 0xBB764C4CA7A44410, 747, 244 }, - { 0x8BAB8EEFB6409C1A, 774, 252 }, - { 0xD01FEF10A657842C, 800, 260 }, - { 0x9B10A4E5E9913129, 827, 268 }, - { 0xE7109BFBA19C0C9D, 853, 276 }, - { 0xAC2820D9623BF429, 880, 284 }, - { 0x80444B5E7AA7CF85, 907, 292 }, - { 0xBF21E44003ACDD2D, 933, 300 }, - { 0x8E679C2F5E44FF8F, 960, 308 }, - { 0xD433179D9C8CB841, 986, 316 }, - { 0x9E19DB92B4E31BA9, 1013, 324 }, - } - }; - - // This computation gives exactly the same results for k as - // k = ceil((kAlpha - e - 1) * 0.30102999566398114) - // for |e| <= 1500, but doesn't require floating-point operations. - // NB: log_10(2) ~= 78913 / 2^18 - JSON_ASSERT(e >= -1500); - JSON_ASSERT(e <= 1500); - const int f = kAlpha - e - 1; - const int k = (f * 78913) / (1 << 18) + static_cast(f > 0); - - const int index = (-kCachedPowersMinDecExp + k + (kCachedPowersDecStep - 1)) / kCachedPowersDecStep; - JSON_ASSERT(index >= 0); - JSON_ASSERT(static_cast(index) < kCachedPowers.size()); - - const cached_power cached = kCachedPowers[static_cast(index)]; - JSON_ASSERT(kAlpha <= cached.e + e + 64); - JSON_ASSERT(kGamma >= cached.e + e + 64); - - return cached; -} - -/*! -For n != 0, returns k, such that pow10 := 10^(k-1) <= n < 10^k. -For n == 0, returns 1 and sets pow10 := 1. -*/ -inline int find_largest_pow10(const std::uint32_t n, std::uint32_t& pow10) -{ - // LCOV_EXCL_START - if (n >= 1000000000) - { - pow10 = 1000000000; - return 10; - } - // LCOV_EXCL_STOP - else if (n >= 100000000) - { - pow10 = 100000000; - return 9; - } - else if (n >= 10000000) - { - pow10 = 10000000; - return 8; - } - else if (n >= 1000000) - { - pow10 = 1000000; - return 7; - } - else if (n >= 100000) - { - pow10 = 100000; - return 6; - } - else if (n >= 10000) - { - pow10 = 10000; - return 5; - } - else if (n >= 1000) - { - pow10 = 1000; - return 4; - } - else if (n >= 100) - { - pow10 = 100; - return 3; - } - else if (n >= 10) - { - pow10 = 10; - return 2; - } - else - { - pow10 = 1; - return 1; - } -} - -inline void grisu2_round(char* buf, int len, std::uint64_t dist, std::uint64_t delta, - std::uint64_t rest, std::uint64_t ten_k) -{ - JSON_ASSERT(len >= 1); - JSON_ASSERT(dist <= delta); - JSON_ASSERT(rest <= delta); - JSON_ASSERT(ten_k > 0); - - // <--------------------------- delta ----> - // <---- dist ---------> - // --------------[------------------+-------------------]-------------- - // M- w M+ - // - // ten_k - // <------> - // <---- rest ----> - // --------------[------------------+----+--------------]-------------- - // w V - // = buf * 10^k - // - // ten_k represents a unit-in-the-last-place in the decimal representation - // stored in buf. - // Decrement buf by ten_k while this takes buf closer to w. - - // The tests are written in this order to avoid overflow in unsigned - // integer arithmetic. - - while (rest < dist - && delta - rest >= ten_k - && (rest + ten_k < dist || dist - rest > rest + ten_k - dist)) - { - JSON_ASSERT(buf[len - 1] != '0'); - buf[len - 1]--; - rest += ten_k; - } -} - -/*! -Generates V = buffer * 10^decimal_exponent, such that M- <= V <= M+. -M- and M+ must be normalized and share the same exponent -60 <= e <= -32. -*/ -inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent, - diyfp M_minus, diyfp w, diyfp M_plus) -{ - static_assert(kAlpha >= -60, "internal error"); - static_assert(kGamma <= -32, "internal error"); - - // Generates the digits (and the exponent) of a decimal floating-point - // number V = buffer * 10^decimal_exponent in the range [M-, M+]. The diyfp's - // w, M- and M+ share the same exponent e, which satisfies alpha <= e <= gamma. - // - // <--------------------------- delta ----> - // <---- dist ---------> - // --------------[------------------+-------------------]-------------- - // M- w M+ - // - // Grisu2 generates the digits of M+ from left to right and stops as soon as - // V is in [M-,M+]. - - JSON_ASSERT(M_plus.e >= kAlpha); - JSON_ASSERT(M_plus.e <= kGamma); - - std::uint64_t delta = diyfp::sub(M_plus, M_minus).f; // (significand of (M+ - M-), implicit exponent is e) - std::uint64_t dist = diyfp::sub(M_plus, w ).f; // (significand of (M+ - w ), implicit exponent is e) - - // Split M+ = f * 2^e into two parts p1 and p2 (note: e < 0): - // - // M+ = f * 2^e - // = ((f div 2^-e) * 2^-e + (f mod 2^-e)) * 2^e - // = ((p1 ) * 2^-e + (p2 )) * 2^e - // = p1 + p2 * 2^e - - const diyfp one(std::uint64_t{1} << -M_plus.e, M_plus.e); - - auto p1 = static_cast(M_plus.f >> -one.e); // p1 = f div 2^-e (Since -e >= 32, p1 fits into a 32-bit int.) - std::uint64_t p2 = M_plus.f & (one.f - 1); // p2 = f mod 2^-e - - // 1) - // - // Generate the digits of the integral part p1 = d[n-1]...d[1]d[0] - - JSON_ASSERT(p1 > 0); - - std::uint32_t pow10; - const int k = find_largest_pow10(p1, pow10); - - // 10^(k-1) <= p1 < 10^k, pow10 = 10^(k-1) - // - // p1 = (p1 div 10^(k-1)) * 10^(k-1) + (p1 mod 10^(k-1)) - // = (d[k-1] ) * 10^(k-1) + (p1 mod 10^(k-1)) - // - // M+ = p1 + p2 * 2^e - // = d[k-1] * 10^(k-1) + (p1 mod 10^(k-1)) + p2 * 2^e - // = d[k-1] * 10^(k-1) + ((p1 mod 10^(k-1)) * 2^-e + p2) * 2^e - // = d[k-1] * 10^(k-1) + ( rest) * 2^e - // - // Now generate the digits d[n] of p1 from left to right (n = k-1,...,0) - // - // p1 = d[k-1]...d[n] * 10^n + d[n-1]...d[0] - // - // but stop as soon as - // - // rest * 2^e = (d[n-1]...d[0] * 2^-e + p2) * 2^e <= delta * 2^e - - int n = k; - while (n > 0) - { - // Invariants: - // M+ = buffer * 10^n + (p1 + p2 * 2^e) (buffer = 0 for n = k) - // pow10 = 10^(n-1) <= p1 < 10^n - // - const std::uint32_t d = p1 / pow10; // d = p1 div 10^(n-1) - const std::uint32_t r = p1 % pow10; // r = p1 mod 10^(n-1) - // - // M+ = buffer * 10^n + (d * 10^(n-1) + r) + p2 * 2^e - // = (buffer * 10 + d) * 10^(n-1) + (r + p2 * 2^e) - // - JSON_ASSERT(d <= 9); - buffer[length++] = static_cast('0' + d); // buffer := buffer * 10 + d - // - // M+ = buffer * 10^(n-1) + (r + p2 * 2^e) - // - p1 = r; - n--; - // - // M+ = buffer * 10^n + (p1 + p2 * 2^e) - // pow10 = 10^n - // - - // Now check if enough digits have been generated. - // Compute - // - // p1 + p2 * 2^e = (p1 * 2^-e + p2) * 2^e = rest * 2^e - // - // Note: - // Since rest and delta share the same exponent e, it suffices to - // compare the significands. - const std::uint64_t rest = (std::uint64_t{p1} << -one.e) + p2; - if (rest <= delta) - { - // V = buffer * 10^n, with M- <= V <= M+. - - decimal_exponent += n; - - // We may now just stop. But instead look if the buffer could be - // decremented to bring V closer to w. - // - // pow10 = 10^n is now 1 ulp in the decimal representation V. - // The rounding procedure works with diyfp's with an implicit - // exponent of e. - // - // 10^n = (10^n * 2^-e) * 2^e = ulp * 2^e - // - const std::uint64_t ten_n = std::uint64_t{pow10} << -one.e; - grisu2_round(buffer, length, dist, delta, rest, ten_n); - - return; - } - - pow10 /= 10; - // - // pow10 = 10^(n-1) <= p1 < 10^n - // Invariants restored. - } - - // 2) - // - // The digits of the integral part have been generated: - // - // M+ = d[k-1]...d[1]d[0] + p2 * 2^e - // = buffer + p2 * 2^e - // - // Now generate the digits of the fractional part p2 * 2^e. - // - // Note: - // No decimal point is generated: the exponent is adjusted instead. - // - // p2 actually represents the fraction - // - // p2 * 2^e - // = p2 / 2^-e - // = d[-1] / 10^1 + d[-2] / 10^2 + ... - // - // Now generate the digits d[-m] of p1 from left to right (m = 1,2,...) - // - // p2 * 2^e = d[-1]d[-2]...d[-m] * 10^-m - // + 10^-m * (d[-m-1] / 10^1 + d[-m-2] / 10^2 + ...) - // - // using - // - // 10^m * p2 = ((10^m * p2) div 2^-e) * 2^-e + ((10^m * p2) mod 2^-e) - // = ( d) * 2^-e + ( r) - // - // or - // 10^m * p2 * 2^e = d + r * 2^e - // - // i.e. - // - // M+ = buffer + p2 * 2^e - // = buffer + 10^-m * (d + r * 2^e) - // = (buffer * 10^m + d) * 10^-m + 10^-m * r * 2^e - // - // and stop as soon as 10^-m * r * 2^e <= delta * 2^e - - JSON_ASSERT(p2 > delta); - - int m = 0; - for (;;) - { - // Invariant: - // M+ = buffer * 10^-m + 10^-m * (d[-m-1] / 10 + d[-m-2] / 10^2 + ...) * 2^e - // = buffer * 10^-m + 10^-m * (p2 ) * 2^e - // = buffer * 10^-m + 10^-m * (1/10 * (10 * p2) ) * 2^e - // = buffer * 10^-m + 10^-m * (1/10 * ((10*p2 div 2^-e) * 2^-e + (10*p2 mod 2^-e)) * 2^e - // - JSON_ASSERT(p2 <= (std::numeric_limits::max)() / 10); - p2 *= 10; - const std::uint64_t d = p2 >> -one.e; // d = (10 * p2) div 2^-e - const std::uint64_t r = p2 & (one.f - 1); // r = (10 * p2) mod 2^-e - // - // M+ = buffer * 10^-m + 10^-m * (1/10 * (d * 2^-e + r) * 2^e - // = buffer * 10^-m + 10^-m * (1/10 * (d + r * 2^e)) - // = (buffer * 10 + d) * 10^(-m-1) + 10^(-m-1) * r * 2^e - // - JSON_ASSERT(d <= 9); - buffer[length++] = static_cast('0' + d); // buffer := buffer * 10 + d - // - // M+ = buffer * 10^(-m-1) + 10^(-m-1) * r * 2^e - // - p2 = r; - m++; - // - // M+ = buffer * 10^-m + 10^-m * p2 * 2^e - // Invariant restored. - - // Check if enough digits have been generated. - // - // 10^-m * p2 * 2^e <= delta * 2^e - // p2 * 2^e <= 10^m * delta * 2^e - // p2 <= 10^m * delta - delta *= 10; - dist *= 10; - if (p2 <= delta) - { - break; - } - } - - // V = buffer * 10^-m, with M- <= V <= M+. - - decimal_exponent -= m; - - // 1 ulp in the decimal representation is now 10^-m. - // Since delta and dist are now scaled by 10^m, we need to do the - // same with ulp in order to keep the units in sync. - // - // 10^m * 10^-m = 1 = 2^-e * 2^e = ten_m * 2^e - // - const std::uint64_t ten_m = one.f; - grisu2_round(buffer, length, dist, delta, p2, ten_m); - - // By construction this algorithm generates the shortest possible decimal - // number (Loitsch, Theorem 6.2) which rounds back to w. - // For an input number of precision p, at least - // - // N = 1 + ceil(p * log_10(2)) - // - // decimal digits are sufficient to identify all binary floating-point - // numbers (Matula, "In-and-Out conversions"). - // This implies that the algorithm does not produce more than N decimal - // digits. - // - // N = 17 for p = 53 (IEEE double precision) - // N = 9 for p = 24 (IEEE single precision) -} - -/*! -v = buf * 10^decimal_exponent -len is the length of the buffer (number of decimal digits) -The buffer must be large enough, i.e. >= max_digits10. -*/ -JSON_HEDLEY_NON_NULL(1) -inline void grisu2(char* buf, int& len, int& decimal_exponent, - diyfp m_minus, diyfp v, diyfp m_plus) -{ - JSON_ASSERT(m_plus.e == m_minus.e); - JSON_ASSERT(m_plus.e == v.e); - - // --------(-----------------------+-----------------------)-------- (A) - // m- v m+ - // - // --------------------(-----------+-----------------------)-------- (B) - // m- v m+ - // - // First scale v (and m- and m+) such that the exponent is in the range - // [alpha, gamma]. - - const cached_power cached = get_cached_power_for_binary_exponent(m_plus.e); - - const diyfp c_minus_k(cached.f, cached.e); // = c ~= 10^-k - - // The exponent of the products is = v.e + c_minus_k.e + q and is in the range [alpha,gamma] - const diyfp w = diyfp::mul(v, c_minus_k); - const diyfp w_minus = diyfp::mul(m_minus, c_minus_k); - const diyfp w_plus = diyfp::mul(m_plus, c_minus_k); - - // ----(---+---)---------------(---+---)---------------(---+---)---- - // w- w w+ - // = c*m- = c*v = c*m+ - // - // diyfp::mul rounds its result and c_minus_k is approximated too. w, w- and - // w+ are now off by a small amount. - // In fact: - // - // w - v * 10^k < 1 ulp - // - // To account for this inaccuracy, add resp. subtract 1 ulp. - // - // --------+---[---------------(---+---)---------------]---+-------- - // w- M- w M+ w+ - // - // Now any number in [M-, M+] (bounds included) will round to w when input, - // regardless of how the input rounding algorithm breaks ties. - // - // And digit_gen generates the shortest possible such number in [M-, M+]. - // Note that this does not mean that Grisu2 always generates the shortest - // possible number in the interval (m-, m+). - const diyfp M_minus(w_minus.f + 1, w_minus.e); - const diyfp M_plus (w_plus.f - 1, w_plus.e ); - - decimal_exponent = -cached.k; // = -(-k) = k - - grisu2_digit_gen(buf, len, decimal_exponent, M_minus, w, M_plus); -} - -/*! -v = buf * 10^decimal_exponent -len is the length of the buffer (number of decimal digits) -The buffer must be large enough, i.e. >= max_digits10. -*/ -template -JSON_HEDLEY_NON_NULL(1) -void grisu2(char* buf, int& len, int& decimal_exponent, FloatType value) -{ - static_assert(diyfp::kPrecision >= std::numeric_limits::digits + 3, - "internal error: not enough precision"); - - JSON_ASSERT(std::isfinite(value)); - JSON_ASSERT(value > 0); - - // If the neighbors (and boundaries) of 'value' are always computed for double-precision - // numbers, all float's can be recovered using strtod (and strtof). However, the resulting - // decimal representations are not exactly "short". - // - // The documentation for 'std::to_chars' (https://en.cppreference.com/w/cpp/utility/to_chars) - // says "value is converted to a string as if by std::sprintf in the default ("C") locale" - // and since sprintf promotes float's to double's, I think this is exactly what 'std::to_chars' - // does. - // On the other hand, the documentation for 'std::to_chars' requires that "parsing the - // representation using the corresponding std::from_chars function recovers value exactly". That - // indicates that single precision floating-point numbers should be recovered using - // 'std::strtof'. - // - // NB: If the neighbors are computed for single-precision numbers, there is a single float - // (7.0385307e-26f) which can't be recovered using strtod. The resulting double precision - // value is off by 1 ulp. -#if 0 - const boundaries w = compute_boundaries(static_cast(value)); -#else - const boundaries w = compute_boundaries(value); -#endif - - grisu2(buf, len, decimal_exponent, w.minus, w.w, w.plus); -} - -/*! -@brief appends a decimal representation of e to buf -@return a pointer to the element following the exponent. -@pre -1000 < e < 1000 -*/ -JSON_HEDLEY_NON_NULL(1) -JSON_HEDLEY_RETURNS_NON_NULL -inline char* append_exponent(char* buf, int e) -{ - JSON_ASSERT(e > -1000); - JSON_ASSERT(e < 1000); - - if (e < 0) - { - e = -e; - *buf++ = '-'; - } - else - { - *buf++ = '+'; - } - - auto k = static_cast(e); - if (k < 10) - { - // Always print at least two digits in the exponent. - // This is for compatibility with printf("%g"). - *buf++ = '0'; - *buf++ = static_cast('0' + k); - } - else if (k < 100) - { - *buf++ = static_cast('0' + k / 10); - k %= 10; - *buf++ = static_cast('0' + k); - } - else - { - *buf++ = static_cast('0' + k / 100); - k %= 100; - *buf++ = static_cast('0' + k / 10); - k %= 10; - *buf++ = static_cast('0' + k); - } - - return buf; -} - -/*! -@brief prettify v = buf * 10^decimal_exponent - -If v is in the range [10^min_exp, 10^max_exp) it will be printed in fixed-point -notation. Otherwise it will be printed in exponential notation. - -@pre min_exp < 0 -@pre max_exp > 0 -*/ -JSON_HEDLEY_NON_NULL(1) -JSON_HEDLEY_RETURNS_NON_NULL -inline char* format_buffer(char* buf, int len, int decimal_exponent, - int min_exp, int max_exp) -{ - JSON_ASSERT(min_exp < 0); - JSON_ASSERT(max_exp > 0); - - const int k = len; - const int n = len + decimal_exponent; - - // v = buf * 10^(n-k) - // k is the length of the buffer (number of decimal digits) - // n is the position of the decimal point relative to the start of the buffer. - - if (k <= n && n <= max_exp) - { - // digits[000] - // len <= max_exp + 2 - - std::memset(buf + k, '0', static_cast(n) - static_cast(k)); - // Make it look like a floating-point number (#362, #378) - buf[n + 0] = '.'; - buf[n + 1] = '0'; - return buf + (static_cast(n) + 2); - } - - if (0 < n && n <= max_exp) - { - // dig.its - // len <= max_digits10 + 1 - - JSON_ASSERT(k > n); - - std::memmove(buf + (static_cast(n) + 1), buf + n, static_cast(k) - static_cast(n)); - buf[n] = '.'; - return buf + (static_cast(k) + 1U); - } - - if (min_exp < n && n <= 0) - { - // 0.[000]digits - // len <= 2 + (-min_exp - 1) + max_digits10 - - std::memmove(buf + (2 + static_cast(-n)), buf, static_cast(k)); - buf[0] = '0'; - buf[1] = '.'; - std::memset(buf + 2, '0', static_cast(-n)); - return buf + (2U + static_cast(-n) + static_cast(k)); - } - - if (k == 1) - { - // dE+123 - // len <= 1 + 5 - - buf += 1; - } - else - { - // d.igitsE+123 - // len <= max_digits10 + 1 + 5 - - std::memmove(buf + 2, buf + 1, static_cast(k) - 1); - buf[1] = '.'; - buf += 1 + static_cast(k); - } - - *buf++ = 'e'; - return append_exponent(buf, n - 1); -} - -} // namespace dtoa_impl - -/*! -@brief generates a decimal representation of the floating-point number value in [first, last). - -The format of the resulting decimal representation is similar to printf's %g -format. Returns an iterator pointing past-the-end of the decimal representation. - -@note The input number must be finite, i.e. NaN's and Inf's are not supported. -@note The buffer must be large enough. -@note The result is NOT null-terminated. -*/ -template -JSON_HEDLEY_NON_NULL(1, 2) -JSON_HEDLEY_RETURNS_NON_NULL -char* to_chars(char* first, const char* last, FloatType value) -{ - static_cast(last); // maybe unused - fix warning - JSON_ASSERT(std::isfinite(value)); - - // Use signbit(value) instead of (value < 0) since signbit works for -0. - if (std::signbit(value)) - { - value = -value; - *first++ = '-'; - } - - if (value == 0) // +-0 - { - *first++ = '0'; - // Make it look like a floating-point number (#362, #378) - *first++ = '.'; - *first++ = '0'; - return first; - } - - JSON_ASSERT(last - first >= std::numeric_limits::max_digits10); - - // Compute v = buffer * 10^decimal_exponent. - // The decimal digits are stored in the buffer, which needs to be interpreted - // as an unsigned decimal integer. - // len is the length of the buffer, i.e. the number of decimal digits. - int len = 0; - int decimal_exponent = 0; - dtoa_impl::grisu2(first, len, decimal_exponent, value); - - JSON_ASSERT(len <= std::numeric_limits::max_digits10); - - // Format the buffer like printf("%.*g", prec, value) - constexpr int kMinExp = -4; - // Use digits10 here to increase compatibility with version 2. - constexpr int kMaxExp = std::numeric_limits::digits10; - - JSON_ASSERT(last - first >= kMaxExp + 2); - JSON_ASSERT(last - first >= 2 + (-kMinExp - 1) + std::numeric_limits::max_digits10); - JSON_ASSERT(last - first >= std::numeric_limits::max_digits10 + 6); - - return dtoa_impl::format_buffer(first, len, decimal_exponent, kMinExp, kMaxExp); -} - -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_json.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_json.hpp deleted file mode 100644 index b45004fd428..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/conversions/to_json.hpp +++ /dev/null @@ -1,374 +0,0 @@ -#pragma once - -#include // copy -#include // begin, end -#include // string -#include // tuple, get -#include // is_same, is_constructible, is_floating_point, is_enum, underlying_type -#include // move, forward, declval, pair -#include // valarray -#include // vector - -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -////////////////// -// constructors // -////////////////// - -template struct external_constructor; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, typename BasicJsonType::boolean_t b) noexcept - { - j.m_type = value_t::boolean; - j.m_value = b; - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, const typename BasicJsonType::string_t& s) - { - j.m_type = value_t::string; - j.m_value = s; - j.assert_invariant(); - } - - template - static void construct(BasicJsonType& j, typename BasicJsonType::string_t&& s) - { - j.m_type = value_t::string; - j.m_value = std::move(s); - j.assert_invariant(); - } - - template < typename BasicJsonType, typename CompatibleStringType, - enable_if_t < !std::is_same::value, - int > = 0 > - static void construct(BasicJsonType& j, const CompatibleStringType& str) - { - j.m_type = value_t::string; - j.m_value.string = j.template create(str); - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, const typename BasicJsonType::binary_t& b) - { - j.m_type = value_t::binary; - typename BasicJsonType::binary_t value{b}; - j.m_value = value; - j.assert_invariant(); - } - - template - static void construct(BasicJsonType& j, typename BasicJsonType::binary_t&& b) - { - j.m_type = value_t::binary; - typename BasicJsonType::binary_t value{std::move(b)}; - j.m_value = value; - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, typename BasicJsonType::number_float_t val) noexcept - { - j.m_type = value_t::number_float; - j.m_value = val; - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, typename BasicJsonType::number_unsigned_t val) noexcept - { - j.m_type = value_t::number_unsigned; - j.m_value = val; - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, typename BasicJsonType::number_integer_t val) noexcept - { - j.m_type = value_t::number_integer; - j.m_value = val; - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, const typename BasicJsonType::array_t& arr) - { - j.m_type = value_t::array; - j.m_value = arr; - j.assert_invariant(); - } - - template - static void construct(BasicJsonType& j, typename BasicJsonType::array_t&& arr) - { - j.m_type = value_t::array; - j.m_value = std::move(arr); - j.assert_invariant(); - } - - template < typename BasicJsonType, typename CompatibleArrayType, - enable_if_t < !std::is_same::value, - int > = 0 > - static void construct(BasicJsonType& j, const CompatibleArrayType& arr) - { - using std::begin; - using std::end; - j.m_type = value_t::array; - j.m_value.array = j.template create(begin(arr), end(arr)); - j.assert_invariant(); - } - - template - static void construct(BasicJsonType& j, const std::vector& arr) - { - j.m_type = value_t::array; - j.m_value = value_t::array; - j.m_value.array->reserve(arr.size()); - for (const bool x : arr) - { - j.m_value.array->push_back(x); - } - j.assert_invariant(); - } - - template::value, int> = 0> - static void construct(BasicJsonType& j, const std::valarray& arr) - { - j.m_type = value_t::array; - j.m_value = value_t::array; - j.m_value.array->resize(arr.size()); - if (arr.size() > 0) - { - std::copy(std::begin(arr), std::end(arr), j.m_value.array->begin()); - } - j.assert_invariant(); - } -}; - -template<> -struct external_constructor -{ - template - static void construct(BasicJsonType& j, const typename BasicJsonType::object_t& obj) - { - j.m_type = value_t::object; - j.m_value = obj; - j.assert_invariant(); - } - - template - static void construct(BasicJsonType& j, typename BasicJsonType::object_t&& obj) - { - j.m_type = value_t::object; - j.m_value = std::move(obj); - j.assert_invariant(); - } - - template < typename BasicJsonType, typename CompatibleObjectType, - enable_if_t < !std::is_same::value, int > = 0 > - static void construct(BasicJsonType& j, const CompatibleObjectType& obj) - { - using std::begin; - using std::end; - - j.m_type = value_t::object; - j.m_value.object = j.template create(begin(obj), end(obj)); - j.assert_invariant(); - } -}; - -///////////// -// to_json // -///////////// - -template::value, int> = 0> -void to_json(BasicJsonType& j, T b) noexcept -{ - external_constructor::construct(j, b); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, const CompatibleString& s) -{ - external_constructor::construct(j, s); -} - -template -void to_json(BasicJsonType& j, typename BasicJsonType::string_t&& s) -{ - external_constructor::construct(j, std::move(s)); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, FloatType val) noexcept -{ - external_constructor::construct(j, static_cast(val)); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, CompatibleNumberUnsignedType val) noexcept -{ - external_constructor::construct(j, static_cast(val)); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, CompatibleNumberIntegerType val) noexcept -{ - external_constructor::construct(j, static_cast(val)); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, EnumType e) noexcept -{ - using underlying_type = typename std::underlying_type::type; - external_constructor::construct(j, static_cast(e)); -} - -template -void to_json(BasicJsonType& j, const std::vector& e) -{ - external_constructor::construct(j, e); -} - -template < typename BasicJsonType, typename CompatibleArrayType, - enable_if_t < is_compatible_array_type::value&& - !is_compatible_object_type::value&& - !is_compatible_string_type::value&& - !std::is_same::value&& - !is_basic_json::value, - int > = 0 > -void to_json(BasicJsonType& j, const CompatibleArrayType& arr) -{ - external_constructor::construct(j, arr); -} - -template -void to_json(BasicJsonType& j, const typename BasicJsonType::binary_t& bin) -{ - external_constructor::construct(j, bin); -} - -template::value, int> = 0> -void to_json(BasicJsonType& j, const std::valarray& arr) -{ - external_constructor::construct(j, std::move(arr)); -} - -template -void to_json(BasicJsonType& j, typename BasicJsonType::array_t&& arr) -{ - external_constructor::construct(j, std::move(arr)); -} - -template < typename BasicJsonType, typename CompatibleObjectType, - enable_if_t < is_compatible_object_type::value&& !is_basic_json::value, int > = 0 > -void to_json(BasicJsonType& j, const CompatibleObjectType& obj) -{ - external_constructor::construct(j, obj); -} - -template -void to_json(BasicJsonType& j, typename BasicJsonType::object_t&& obj) -{ - external_constructor::construct(j, std::move(obj)); -} - -template < - typename BasicJsonType, typename T, std::size_t N, - enable_if_t < !std::is_constructible::value, - int > = 0 > -void to_json(BasicJsonType& j, const T(&arr)[N]) -{ - external_constructor::construct(j, arr); -} - -template < typename BasicJsonType, typename T1, typename T2, enable_if_t < std::is_constructible::value&& std::is_constructible::value, int > = 0 > -void to_json(BasicJsonType& j, const std::pair& p) -{ - j = { p.first, p.second }; -} - -// for https://github.com/nlohmann/json/pull/1134 -template>::value, int> = 0> -void to_json(BasicJsonType& j, const T& b) -{ - j = { {b.key(), b.value()} }; -} - -template -void to_json_tuple_impl(BasicJsonType& j, const Tuple& t, index_sequence /*unused*/) -{ - j = { std::get(t)... }; -} - -template::value, int > = 0> -void to_json(BasicJsonType& j, const T& t) -{ - to_json_tuple_impl(j, t, make_index_sequence::value> {}); -} - -struct to_json_fn -{ - template - auto operator()(BasicJsonType& j, T&& val) const noexcept(noexcept(to_json(j, std::forward(val)))) - -> decltype(to_json(j, std::forward(val)), void()) - { - return to_json(j, std::forward(val)); - } -}; -} // namespace detail - -/// namespace to hold default `to_json` function -namespace -{ -constexpr const auto& to_json = detail::static_const::value; -} // namespace -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/exceptions.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/exceptions.hpp deleted file mode 100644 index dd92897d5ae..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/exceptions.hpp +++ /dev/null @@ -1,357 +0,0 @@ -#pragma once - -#include // exception -#include // runtime_error -#include // to_string - -#include -#include - -namespace nlohmann -{ -namespace detail -{ -//////////////// -// exceptions // -//////////////// - -/*! -@brief general exception of the @ref basic_json class - -This class is an extension of `std::exception` objects with a member @a id for -exception ids. It is used as the base class for all exceptions thrown by the -@ref basic_json class. This class can hence be used as "wildcard" to catch -exceptions. - -Subclasses: -- @ref parse_error for exceptions indicating a parse error -- @ref invalid_iterator for exceptions indicating errors with iterators -- @ref type_error for exceptions indicating executing a member function with - a wrong type -- @ref out_of_range for exceptions indicating access out of the defined range -- @ref other_error for exceptions indicating other library errors - -@internal -@note To have nothrow-copy-constructible exceptions, we internally use - `std::runtime_error` which can cope with arbitrary-length error messages. - Intermediate strings are built with static functions and then passed to - the actual constructor. -@endinternal - -@liveexample{The following code shows how arbitrary library exceptions can be -caught.,exception} - -@since version 3.0.0 -*/ -class exception : public std::exception -{ - public: - /// returns the explanatory string - JSON_HEDLEY_RETURNS_NON_NULL - const char* what() const noexcept override - { - return m.what(); - } - - /// the id of the exception - const int id; - - protected: - JSON_HEDLEY_NON_NULL(3) - exception(int id_, const char* what_arg) : id(id_), m(what_arg) {} - - static std::string name(const std::string& ename, int id_) - { - return "[json.exception." + ename + "." + std::to_string(id_) + "] "; - } - - private: - /// an exception object as storage for error messages - std::runtime_error m; -}; - -/*! -@brief exception indicating a parse error - -This exception is thrown by the library when a parse error occurs. Parse errors -can occur during the deserialization of JSON text, CBOR, MessagePack, as well -as when using JSON Patch. - -Member @a byte holds the byte index of the last read character in the input -file. - -Exceptions have ids 1xx. - -name / id | example message | description ------------------------------- | --------------- | ------------------------- -json.exception.parse_error.101 | parse error at 2: unexpected end of input; expected string literal | This error indicates a syntax error while deserializing a JSON text. The error message describes that an unexpected token (character) was encountered, and the member @a byte indicates the error position. -json.exception.parse_error.102 | parse error at 14: missing or wrong low surrogate | JSON uses the `\uxxxx` format to describe Unicode characters. Code points above above 0xFFFF are split into two `\uxxxx` entries ("surrogate pairs"). This error indicates that the surrogate pair is incomplete or contains an invalid code point. -json.exception.parse_error.103 | parse error: code points above 0x10FFFF are invalid | Unicode supports code points up to 0x10FFFF. Code points above 0x10FFFF are invalid. -json.exception.parse_error.104 | parse error: JSON patch must be an array of objects | [RFC 6902](https://tools.ietf.org/html/rfc6902) requires a JSON Patch document to be a JSON document that represents an array of objects. -json.exception.parse_error.105 | parse error: operation must have string member 'op' | An operation of a JSON Patch document must contain exactly one "op" member, whose value indicates the operation to perform. Its value must be one of "add", "remove", "replace", "move", "copy", or "test"; other values are errors. -json.exception.parse_error.106 | parse error: array index '01' must not begin with '0' | An array index in a JSON Pointer ([RFC 6901](https://tools.ietf.org/html/rfc6901)) may be `0` or any number without a leading `0`. -json.exception.parse_error.107 | parse error: JSON pointer must be empty or begin with '/' - was: 'foo' | A JSON Pointer must be a Unicode string containing a sequence of zero or more reference tokens, each prefixed by a `/` character. -json.exception.parse_error.108 | parse error: escape character '~' must be followed with '0' or '1' | In a JSON Pointer, only `~0` and `~1` are valid escape sequences. -json.exception.parse_error.109 | parse error: array index 'one' is not a number | A JSON Pointer array index must be a number. -json.exception.parse_error.110 | parse error at 1: cannot read 2 bytes from vector | When parsing CBOR or MessagePack, the byte vector ends before the complete value has been read. -json.exception.parse_error.112 | parse error at 1: error reading CBOR; last byte: 0xF8 | Not all types of CBOR or MessagePack are supported. This exception occurs if an unsupported byte was read. -json.exception.parse_error.113 | parse error at 2: expected a CBOR string; last byte: 0x98 | While parsing a map key, a value that is not a string has been read. -json.exception.parse_error.114 | parse error: Unsupported BSON record type 0x0F | The parsing of the corresponding BSON record type is not implemented (yet). -json.exception.parse_error.115 | parse error at byte 5: syntax error while parsing UBJSON high-precision number: invalid number text: 1A | A UBJSON high-precision number could not be parsed. - -@note For an input with n bytes, 1 is the index of the first character and n+1 - is the index of the terminating null byte or the end of file. This also - holds true when reading a byte vector (CBOR or MessagePack). - -@liveexample{The following code shows how a `parse_error` exception can be -caught.,parse_error} - -@sa - @ref exception for the base class of the library exceptions -@sa - @ref invalid_iterator for exceptions indicating errors with iterators -@sa - @ref type_error for exceptions indicating executing a member function with - a wrong type -@sa - @ref out_of_range for exceptions indicating access out of the defined range -@sa - @ref other_error for exceptions indicating other library errors - -@since version 3.0.0 -*/ -class parse_error : public exception -{ - public: - /*! - @brief create a parse error exception - @param[in] id_ the id of the exception - @param[in] pos the position where the error occurred (or with - chars_read_total=0 if the position cannot be - determined) - @param[in] what_arg the explanatory string - @return parse_error object - */ - static parse_error create(int id_, const position_t& pos, const std::string& what_arg) - { - std::string w = exception::name("parse_error", id_) + "parse error" + - position_string(pos) + ": " + what_arg; - return parse_error(id_, pos.chars_read_total, w.c_str()); - } - - static parse_error create(int id_, std::size_t byte_, const std::string& what_arg) - { - std::string w = exception::name("parse_error", id_) + "parse error" + - (byte_ != 0 ? (" at byte " + std::to_string(byte_)) : "") + - ": " + what_arg; - return parse_error(id_, byte_, w.c_str()); - } - - /*! - @brief byte index of the parse error - - The byte index of the last read character in the input file. - - @note For an input with n bytes, 1 is the index of the first character and - n+1 is the index of the terminating null byte or the end of file. - This also holds true when reading a byte vector (CBOR or MessagePack). - */ - const std::size_t byte; - - private: - parse_error(int id_, std::size_t byte_, const char* what_arg) - : exception(id_, what_arg), byte(byte_) {} - - static std::string position_string(const position_t& pos) - { - return " at line " + std::to_string(pos.lines_read + 1) + - ", column " + std::to_string(pos.chars_read_current_line); - } -}; - -/*! -@brief exception indicating errors with iterators - -This exception is thrown if iterators passed to a library function do not match -the expected semantics. - -Exceptions have ids 2xx. - -name / id | example message | description ------------------------------------ | --------------- | ------------------------- -json.exception.invalid_iterator.201 | iterators are not compatible | The iterators passed to constructor @ref basic_json(InputIT first, InputIT last) are not compatible, meaning they do not belong to the same container. Therefore, the range (@a first, @a last) is invalid. -json.exception.invalid_iterator.202 | iterator does not fit current value | In an erase or insert function, the passed iterator @a pos does not belong to the JSON value for which the function was called. It hence does not define a valid position for the deletion/insertion. -json.exception.invalid_iterator.203 | iterators do not fit current value | Either iterator passed to function @ref erase(IteratorType first, IteratorType last) does not belong to the JSON value from which values shall be erased. It hence does not define a valid range to delete values from. -json.exception.invalid_iterator.204 | iterators out of range | When an iterator range for a primitive type (number, boolean, or string) is passed to a constructor or an erase function, this range has to be exactly (@ref begin(), @ref end()), because this is the only way the single stored value is expressed. All other ranges are invalid. -json.exception.invalid_iterator.205 | iterator out of range | When an iterator for a primitive type (number, boolean, or string) is passed to an erase function, the iterator has to be the @ref begin() iterator, because it is the only way to address the stored value. All other iterators are invalid. -json.exception.invalid_iterator.206 | cannot construct with iterators from null | The iterators passed to constructor @ref basic_json(InputIT first, InputIT last) belong to a JSON null value and hence to not define a valid range. -json.exception.invalid_iterator.207 | cannot use key() for non-object iterators | The key() member function can only be used on iterators belonging to a JSON object, because other types do not have a concept of a key. -json.exception.invalid_iterator.208 | cannot use operator[] for object iterators | The operator[] to specify a concrete offset cannot be used on iterators belonging to a JSON object, because JSON objects are unordered. -json.exception.invalid_iterator.209 | cannot use offsets with object iterators | The offset operators (+, -, +=, -=) cannot be used on iterators belonging to a JSON object, because JSON objects are unordered. -json.exception.invalid_iterator.210 | iterators do not fit | The iterator range passed to the insert function are not compatible, meaning they do not belong to the same container. Therefore, the range (@a first, @a last) is invalid. -json.exception.invalid_iterator.211 | passed iterators may not belong to container | The iterator range passed to the insert function must not be a subrange of the container to insert to. -json.exception.invalid_iterator.212 | cannot compare iterators of different containers | When two iterators are compared, they must belong to the same container. -json.exception.invalid_iterator.213 | cannot compare order of object iterators | The order of object iterators cannot be compared, because JSON objects are unordered. -json.exception.invalid_iterator.214 | cannot get value | Cannot get value for iterator: Either the iterator belongs to a null value or it is an iterator to a primitive type (number, boolean, or string), but the iterator is different to @ref begin(). - -@liveexample{The following code shows how an `invalid_iterator` exception can be -caught.,invalid_iterator} - -@sa - @ref exception for the base class of the library exceptions -@sa - @ref parse_error for exceptions indicating a parse error -@sa - @ref type_error for exceptions indicating executing a member function with - a wrong type -@sa - @ref out_of_range for exceptions indicating access out of the defined range -@sa - @ref other_error for exceptions indicating other library errors - -@since version 3.0.0 -*/ -class invalid_iterator : public exception -{ - public: - static invalid_iterator create(int id_, const std::string& what_arg) - { - std::string w = exception::name("invalid_iterator", id_) + what_arg; - return invalid_iterator(id_, w.c_str()); - } - - private: - JSON_HEDLEY_NON_NULL(3) - invalid_iterator(int id_, const char* what_arg) - : exception(id_, what_arg) {} -}; - -/*! -@brief exception indicating executing a member function with a wrong type - -This exception is thrown in case of a type error; that is, a library function is -executed on a JSON value whose type does not match the expected semantics. - -Exceptions have ids 3xx. - -name / id | example message | description ------------------------------ | --------------- | ------------------------- -json.exception.type_error.301 | cannot create object from initializer list | To create an object from an initializer list, the initializer list must consist only of a list of pairs whose first element is a string. When this constraint is violated, an array is created instead. -json.exception.type_error.302 | type must be object, but is array | During implicit or explicit value conversion, the JSON type must be compatible to the target type. For instance, a JSON string can only be converted into string types, but not into numbers or boolean types. -json.exception.type_error.303 | incompatible ReferenceType for get_ref, actual type is object | To retrieve a reference to a value stored in a @ref basic_json object with @ref get_ref, the type of the reference must match the value type. For instance, for a JSON array, the @a ReferenceType must be @ref array_t &. -json.exception.type_error.304 | cannot use at() with string | The @ref at() member functions can only be executed for certain JSON types. -json.exception.type_error.305 | cannot use operator[] with string | The @ref operator[] member functions can only be executed for certain JSON types. -json.exception.type_error.306 | cannot use value() with string | The @ref value() member functions can only be executed for certain JSON types. -json.exception.type_error.307 | cannot use erase() with string | The @ref erase() member functions can only be executed for certain JSON types. -json.exception.type_error.308 | cannot use push_back() with string | The @ref push_back() and @ref operator+= member functions can only be executed for certain JSON types. -json.exception.type_error.309 | cannot use insert() with | The @ref insert() member functions can only be executed for certain JSON types. -json.exception.type_error.310 | cannot use swap() with number | The @ref swap() member functions can only be executed for certain JSON types. -json.exception.type_error.311 | cannot use emplace_back() with string | The @ref emplace_back() member function can only be executed for certain JSON types. -json.exception.type_error.312 | cannot use update() with string | The @ref update() member functions can only be executed for certain JSON types. -json.exception.type_error.313 | invalid value to unflatten | The @ref unflatten function converts an object whose keys are JSON Pointers back into an arbitrary nested JSON value. The JSON Pointers must not overlap, because then the resulting value would not be well defined. -json.exception.type_error.314 | only objects can be unflattened | The @ref unflatten function only works for an object whose keys are JSON Pointers. -json.exception.type_error.315 | values in object must be primitive | The @ref unflatten function only works for an object whose keys are JSON Pointers and whose values are primitive. -json.exception.type_error.316 | invalid UTF-8 byte at index 10: 0x7E | The @ref dump function only works with UTF-8 encoded strings; that is, if you assign a `std::string` to a JSON value, make sure it is UTF-8 encoded. | -json.exception.type_error.317 | JSON value cannot be serialized to requested format | The dynamic type of the object cannot be represented in the requested serialization format (e.g. a raw `true` or `null` JSON object cannot be serialized to BSON) | - -@liveexample{The following code shows how a `type_error` exception can be -caught.,type_error} - -@sa - @ref exception for the base class of the library exceptions -@sa - @ref parse_error for exceptions indicating a parse error -@sa - @ref invalid_iterator for exceptions indicating errors with iterators -@sa - @ref out_of_range for exceptions indicating access out of the defined range -@sa - @ref other_error for exceptions indicating other library errors - -@since version 3.0.0 -*/ -class type_error : public exception -{ - public: - static type_error create(int id_, const std::string& what_arg) - { - std::string w = exception::name("type_error", id_) + what_arg; - return type_error(id_, w.c_str()); - } - - private: - JSON_HEDLEY_NON_NULL(3) - type_error(int id_, const char* what_arg) : exception(id_, what_arg) {} -}; - -/*! -@brief exception indicating access out of the defined range - -This exception is thrown in case a library function is called on an input -parameter that exceeds the expected range, for instance in case of array -indices or nonexisting object keys. - -Exceptions have ids 4xx. - -name / id | example message | description -------------------------------- | --------------- | ------------------------- -json.exception.out_of_range.401 | array index 3 is out of range | The provided array index @a i is larger than @a size-1. -json.exception.out_of_range.402 | array index '-' (3) is out of range | The special array index `-` in a JSON Pointer never describes a valid element of the array, but the index past the end. That is, it can only be used to add elements at this position, but not to read it. -json.exception.out_of_range.403 | key 'foo' not found | The provided key was not found in the JSON object. -json.exception.out_of_range.404 | unresolved reference token 'foo' | A reference token in a JSON Pointer could not be resolved. -json.exception.out_of_range.405 | JSON pointer has no parent | The JSON Patch operations 'remove' and 'add' can not be applied to the root element of the JSON value. -json.exception.out_of_range.406 | number overflow parsing '10E1000' | A parsed number could not be stored as without changing it to NaN or INF. -json.exception.out_of_range.407 | number overflow serializing '9223372036854775808' | UBJSON and BSON only support integer numbers up to 9223372036854775807. (until version 3.8.0) | -json.exception.out_of_range.408 | excessive array size: 8658170730974374167 | The size (following `#`) of an UBJSON array or object exceeds the maximal capacity. | -json.exception.out_of_range.409 | BSON key cannot contain code point U+0000 (at byte 2) | Key identifiers to be serialized to BSON cannot contain code point U+0000, since the key is stored as zero-terminated c-string | - -@liveexample{The following code shows how an `out_of_range` exception can be -caught.,out_of_range} - -@sa - @ref exception for the base class of the library exceptions -@sa - @ref parse_error for exceptions indicating a parse error -@sa - @ref invalid_iterator for exceptions indicating errors with iterators -@sa - @ref type_error for exceptions indicating executing a member function with - a wrong type -@sa - @ref other_error for exceptions indicating other library errors - -@since version 3.0.0 -*/ -class out_of_range : public exception -{ - public: - static out_of_range create(int id_, const std::string& what_arg) - { - std::string w = exception::name("out_of_range", id_) + what_arg; - return out_of_range(id_, w.c_str()); - } - - private: - JSON_HEDLEY_NON_NULL(3) - out_of_range(int id_, const char* what_arg) : exception(id_, what_arg) {} -}; - -/*! -@brief exception indicating other library errors - -This exception is thrown in case of errors that cannot be classified with the -other exception types. - -Exceptions have ids 5xx. - -name / id | example message | description ------------------------------- | --------------- | ------------------------- -json.exception.other_error.501 | unsuccessful: {"op":"test","path":"/baz", "value":"bar"} | A JSON Patch operation 'test' failed. The unsuccessful operation is also printed. - -@sa - @ref exception for the base class of the library exceptions -@sa - @ref parse_error for exceptions indicating a parse error -@sa - @ref invalid_iterator for exceptions indicating errors with iterators -@sa - @ref type_error for exceptions indicating executing a member function with - a wrong type -@sa - @ref out_of_range for exceptions indicating access out of the defined range - -@liveexample{The following code shows how an `other_error` exception can be -caught.,other_error} - -@since version 3.0.0 -*/ -class other_error : public exception -{ - public: - static other_error create(int id_, const std::string& what_arg) - { - std::string w = exception::name("other_error", id_) + what_arg; - return other_error(id_, w.c_str()); - } - - private: - JSON_HEDLEY_NON_NULL(3) - other_error(int id_, const char* what_arg) : exception(id_, what_arg) {} -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/hash.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/hash.hpp deleted file mode 100644 index 4094cc94fc3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/hash.hpp +++ /dev/null @@ -1,117 +0,0 @@ -#pragma once - -#include // size_t, uint8_t -#include // hash - -namespace nlohmann -{ -namespace detail -{ - -// boost::hash_combine -inline std::size_t combine(std::size_t seed, std::size_t h) noexcept -{ - seed ^= h + 0x9e3779b9 + (seed << 6U) + (seed >> 2U); - return seed; -} - -/*! -@brief hash a JSON value - -The hash function tries to rely on std::hash where possible. Furthermore, the -type of the JSON value is taken into account to have different hash values for -null, 0, 0U, and false, etc. - -@tparam BasicJsonType basic_json specialization -@param j JSON value to hash -@return hash value of j -*/ -template -std::size_t hash(const BasicJsonType& j) -{ - using string_t = typename BasicJsonType::string_t; - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - - const auto type = static_cast(j.type()); - switch (j.type()) - { - case BasicJsonType::value_t::null: - case BasicJsonType::value_t::discarded: - { - return combine(type, 0); - } - - case BasicJsonType::value_t::object: - { - auto seed = combine(type, j.size()); - for (const auto& element : j.items()) - { - const auto h = std::hash {}(element.key()); - seed = combine(seed, h); - seed = combine(seed, hash(element.value())); - } - return seed; - } - - case BasicJsonType::value_t::array: - { - auto seed = combine(type, j.size()); - for (const auto& element : j) - { - seed = combine(seed, hash(element)); - } - return seed; - } - - case BasicJsonType::value_t::string: - { - const auto h = std::hash {}(j.template get_ref()); - return combine(type, h); - } - - case BasicJsonType::value_t::boolean: - { - const auto h = std::hash {}(j.template get()); - return combine(type, h); - } - - case BasicJsonType::value_t::number_integer: - { - const auto h = std::hash {}(j.template get()); - return combine(type, h); - } - - case nlohmann::detail::value_t::number_unsigned: - { - const auto h = std::hash {}(j.template get()); - return combine(type, h); - } - - case nlohmann::detail::value_t::number_float: - { - const auto h = std::hash {}(j.template get()); - return combine(type, h); - } - - case nlohmann::detail::value_t::binary: - { - auto seed = combine(type, j.get_binary().size()); - const auto h = std::hash {}(j.get_binary().has_subtype()); - seed = combine(seed, h); - seed = combine(seed, j.get_binary().subtype()); - for (const auto byte : j.get_binary()) - { - seed = combine(seed, std::hash {}(byte)); - } - return seed; - } - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } -} - -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/binary_reader.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/binary_reader.hpp deleted file mode 100644 index 4c94d1ccfd3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/binary_reader.hpp +++ /dev/null @@ -1,2459 +0,0 @@ -#pragma once - -#include // generate_n -#include // array -#include // ldexp -#include // size_t -#include // uint8_t, uint16_t, uint32_t, uint64_t -#include // snprintf -#include // memcpy -#include // back_inserter -#include // numeric_limits -#include // char_traits, string -#include // make_pair, move - -#include -#include -#include -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ - -/// how to treat CBOR tags -enum class cbor_tag_handler_t -{ - error, ///< throw a parse_error exception in case of a tag - ignore ///< ignore tags -}; - -/*! -@brief determine system byte order - -@return true if and only if system's byte order is little endian - -@note from https://stackoverflow.com/a/1001328/266378 -*/ -static inline bool little_endianess(int num = 1) noexcept -{ - return *reinterpret_cast(&num) == 1; -} - - -/////////////////// -// binary reader // -/////////////////// - -/*! -@brief deserialization of CBOR, MessagePack, and UBJSON values -*/ -template> -class binary_reader -{ - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - using json_sax_t = SAX; - using char_type = typename InputAdapterType::char_type; - using char_int_type = typename std::char_traits::int_type; - - public: - /*! - @brief create a binary reader - - @param[in] adapter input adapter to read from - */ - explicit binary_reader(InputAdapterType&& adapter) : ia(std::move(adapter)) - { - (void)detail::is_sax_static_asserts {}; - } - - // make class move-only - binary_reader(const binary_reader&) = delete; - binary_reader(binary_reader&&) = default; - binary_reader& operator=(const binary_reader&) = delete; - binary_reader& operator=(binary_reader&&) = default; - ~binary_reader() = default; - - /*! - @param[in] format the binary format to parse - @param[in] sax_ a SAX event processor - @param[in] strict whether to expect the input to be consumed completed - @param[in] tag_handler how to treat CBOR tags - - @return - */ - JSON_HEDLEY_NON_NULL(3) - bool sax_parse(const input_format_t format, - json_sax_t* sax_, - const bool strict = true, - const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error) - { - sax = sax_; - bool result = false; - - switch (format) - { - case input_format_t::bson: - result = parse_bson_internal(); - break; - - case input_format_t::cbor: - result = parse_cbor_internal(true, tag_handler); - break; - - case input_format_t::msgpack: - result = parse_msgpack_internal(); - break; - - case input_format_t::ubjson: - result = parse_ubjson_internal(); - break; - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - - // strict mode: next byte must be EOF - if (result && strict) - { - if (format == input_format_t::ubjson) - { - get_ignore_noop(); - } - else - { - get(); - } - - if (JSON_HEDLEY_UNLIKELY(current != std::char_traits::eof())) - { - return sax->parse_error(chars_read, get_token_string(), - parse_error::create(110, chars_read, exception_message(format, "expected end of input; last byte: 0x" + get_token_string(), "value"))); - } - } - - return result; - } - - private: - ////////// - // BSON // - ////////// - - /*! - @brief Reads in a BSON-object and passes it to the SAX-parser. - @return whether a valid BSON-value was passed to the SAX parser - */ - bool parse_bson_internal() - { - std::int32_t document_size{}; - get_number(input_format_t::bson, document_size); - - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(std::size_t(-1)))) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_list(/*is_array*/false))) - { - return false; - } - - return sax->end_object(); - } - - /*! - @brief Parses a C-style string from the BSON input. - @param[in, out] result A reference to the string variable where the read - string is to be stored. - @return `true` if the \x00-byte indicating the end of the string was - encountered before the EOF; false` indicates an unexpected EOF. - */ - bool get_bson_cstr(string_t& result) - { - auto out = std::back_inserter(result); - while (true) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::bson, "cstring"))) - { - return false; - } - if (current == 0x00) - { - return true; - } - *out++ = static_cast(current); - } - } - - /*! - @brief Parses a zero-terminated string of length @a len from the BSON - input. - @param[in] len The length (including the zero-byte at the end) of the - string to be read. - @param[in, out] result A reference to the string variable where the read - string is to be stored. - @tparam NumberType The type of the length @a len - @pre len >= 1 - @return `true` if the string was successfully parsed - */ - template - bool get_bson_string(const NumberType len, string_t& result) - { - if (JSON_HEDLEY_UNLIKELY(len < 1)) - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::bson, "string length must be at least 1, is " + std::to_string(len), "string"))); - } - - return get_string(input_format_t::bson, len - static_cast(1), result) && get() != std::char_traits::eof(); - } - - /*! - @brief Parses a byte array input of length @a len from the BSON input. - @param[in] len The length of the byte array to be read. - @param[in, out] result A reference to the binary variable where the read - array is to be stored. - @tparam NumberType The type of the length @a len - @pre len >= 0 - @return `true` if the byte array was successfully parsed - */ - template - bool get_bson_binary(const NumberType len, binary_t& result) - { - if (JSON_HEDLEY_UNLIKELY(len < 0)) - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::bson, "byte array length cannot be negative, is " + std::to_string(len), "binary"))); - } - - // All BSON binary values have a subtype - std::uint8_t subtype{}; - get_number(input_format_t::bson, subtype); - result.set_subtype(subtype); - - return get_binary(input_format_t::bson, len, result); - } - - /*! - @brief Read a BSON document element of the given @a element_type. - @param[in] element_type The BSON element type, c.f. http://bsonspec.org/spec.html - @param[in] element_type_parse_position The position in the input stream, - where the `element_type` was read. - @warning Not all BSON element types are supported yet. An unsupported - @a element_type will give rise to a parse_error.114: - Unsupported BSON record type 0x... - @return whether a valid BSON-object/array was passed to the SAX parser - */ - bool parse_bson_element_internal(const char_int_type element_type, - const std::size_t element_type_parse_position) - { - switch (element_type) - { - case 0x01: // double - { - double number{}; - return get_number(input_format_t::bson, number) && sax->number_float(static_cast(number), ""); - } - - case 0x02: // string - { - std::int32_t len{}; - string_t value; - return get_number(input_format_t::bson, len) && get_bson_string(len, value) && sax->string(value); - } - - case 0x03: // object - { - return parse_bson_internal(); - } - - case 0x04: // array - { - return parse_bson_array(); - } - - case 0x05: // binary - { - std::int32_t len{}; - binary_t value; - return get_number(input_format_t::bson, len) && get_bson_binary(len, value) && sax->binary(value); - } - - case 0x08: // boolean - { - return sax->boolean(get() != 0); - } - - case 0x0A: // null - { - return sax->null(); - } - - case 0x10: // int32 - { - std::int32_t value{}; - return get_number(input_format_t::bson, value) && sax->number_integer(value); - } - - case 0x12: // int64 - { - std::int64_t value{}; - return get_number(input_format_t::bson, value) && sax->number_integer(value); - } - - default: // anything else not supported (yet) - { - std::array cr{{}}; - (std::snprintf)(cr.data(), cr.size(), "%.2hhX", static_cast(element_type)); - return sax->parse_error(element_type_parse_position, std::string(cr.data()), parse_error::create(114, element_type_parse_position, "Unsupported BSON record type 0x" + std::string(cr.data()))); - } - } - } - - /*! - @brief Read a BSON element list (as specified in the BSON-spec) - - The same binary layout is used for objects and arrays, hence it must be - indicated with the argument @a is_array which one is expected - (true --> array, false --> object). - - @param[in] is_array Determines if the element list being read is to be - treated as an object (@a is_array == false), or as an - array (@a is_array == true). - @return whether a valid BSON-object/array was passed to the SAX parser - */ - bool parse_bson_element_list(const bool is_array) - { - string_t key; - - while (auto element_type = get()) - { - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::bson, "element list"))) - { - return false; - } - - const std::size_t element_type_parse_position = chars_read; - if (JSON_HEDLEY_UNLIKELY(!get_bson_cstr(key))) - { - return false; - } - - if (!is_array && !sax->key(key)) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_internal(element_type, element_type_parse_position))) - { - return false; - } - - // get_bson_cstr only appends - key.clear(); - } - - return true; - } - - /*! - @brief Reads an array from the BSON input and passes it to the SAX-parser. - @return whether a valid BSON-array was passed to the SAX parser - */ - bool parse_bson_array() - { - std::int32_t document_size{}; - get_number(input_format_t::bson, document_size); - - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(std::size_t(-1)))) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_list(/*is_array*/true))) - { - return false; - } - - return sax->end_array(); - } - - ////////// - // CBOR // - ////////// - - /*! - @param[in] get_char whether a new character should be retrieved from the - input (true) or whether the last read character should - be considered instead (false) - @param[in] tag_handler how CBOR tags should be treated - - @return whether a valid CBOR value was passed to the SAX parser - */ - bool parse_cbor_internal(const bool get_char, - const cbor_tag_handler_t tag_handler) - { - switch (get_char ? get() : current) - { - // EOF - case std::char_traits::eof(): - return unexpect_eof(input_format_t::cbor, "value"); - - // Integer 0x00..0x17 (0..23) - case 0x00: - case 0x01: - case 0x02: - case 0x03: - case 0x04: - case 0x05: - case 0x06: - case 0x07: - case 0x08: - case 0x09: - case 0x0A: - case 0x0B: - case 0x0C: - case 0x0D: - case 0x0E: - case 0x0F: - case 0x10: - case 0x11: - case 0x12: - case 0x13: - case 0x14: - case 0x15: - case 0x16: - case 0x17: - return sax->number_unsigned(static_cast(current)); - - case 0x18: // Unsigned integer (one-byte uint8_t follows) - { - std::uint8_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_unsigned(number); - } - - case 0x19: // Unsigned integer (two-byte uint16_t follows) - { - std::uint16_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_unsigned(number); - } - - case 0x1A: // Unsigned integer (four-byte uint32_t follows) - { - std::uint32_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_unsigned(number); - } - - case 0x1B: // Unsigned integer (eight-byte uint64_t follows) - { - std::uint64_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_unsigned(number); - } - - // Negative integer -1-0x00..-1-0x17 (-1..-24) - case 0x20: - case 0x21: - case 0x22: - case 0x23: - case 0x24: - case 0x25: - case 0x26: - case 0x27: - case 0x28: - case 0x29: - case 0x2A: - case 0x2B: - case 0x2C: - case 0x2D: - case 0x2E: - case 0x2F: - case 0x30: - case 0x31: - case 0x32: - case 0x33: - case 0x34: - case 0x35: - case 0x36: - case 0x37: - return sax->number_integer(static_cast(0x20 - 1 - current)); - - case 0x38: // Negative integer (one-byte uint8_t follows) - { - std::uint8_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast(-1) - number); - } - - case 0x39: // Negative integer -1-n (two-byte uint16_t follows) - { - std::uint16_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast(-1) - number); - } - - case 0x3A: // Negative integer -1-n (four-byte uint32_t follows) - { - std::uint32_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast(-1) - number); - } - - case 0x3B: // Negative integer -1-n (eight-byte uint64_t follows) - { - std::uint64_t number{}; - return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast(-1) - - static_cast(number)); - } - - // Binary data (0x00..0x17 bytes follow) - case 0x40: - case 0x41: - case 0x42: - case 0x43: - case 0x44: - case 0x45: - case 0x46: - case 0x47: - case 0x48: - case 0x49: - case 0x4A: - case 0x4B: - case 0x4C: - case 0x4D: - case 0x4E: - case 0x4F: - case 0x50: - case 0x51: - case 0x52: - case 0x53: - case 0x54: - case 0x55: - case 0x56: - case 0x57: - case 0x58: // Binary data (one-byte uint8_t for n follows) - case 0x59: // Binary data (two-byte uint16_t for n follow) - case 0x5A: // Binary data (four-byte uint32_t for n follow) - case 0x5B: // Binary data (eight-byte uint64_t for n follow) - case 0x5F: // Binary data (indefinite length) - { - binary_t b; - return get_cbor_binary(b) && sax->binary(b); - } - - // UTF-8 string (0x00..0x17 bytes follow) - case 0x60: - case 0x61: - case 0x62: - case 0x63: - case 0x64: - case 0x65: - case 0x66: - case 0x67: - case 0x68: - case 0x69: - case 0x6A: - case 0x6B: - case 0x6C: - case 0x6D: - case 0x6E: - case 0x6F: - case 0x70: - case 0x71: - case 0x72: - case 0x73: - case 0x74: - case 0x75: - case 0x76: - case 0x77: - case 0x78: // UTF-8 string (one-byte uint8_t for n follows) - case 0x79: // UTF-8 string (two-byte uint16_t for n follow) - case 0x7A: // UTF-8 string (four-byte uint32_t for n follow) - case 0x7B: // UTF-8 string (eight-byte uint64_t for n follow) - case 0x7F: // UTF-8 string (indefinite length) - { - string_t s; - return get_cbor_string(s) && sax->string(s); - } - - // array (0x00..0x17 data items follow) - case 0x80: - case 0x81: - case 0x82: - case 0x83: - case 0x84: - case 0x85: - case 0x86: - case 0x87: - case 0x88: - case 0x89: - case 0x8A: - case 0x8B: - case 0x8C: - case 0x8D: - case 0x8E: - case 0x8F: - case 0x90: - case 0x91: - case 0x92: - case 0x93: - case 0x94: - case 0x95: - case 0x96: - case 0x97: - return get_cbor_array(static_cast(static_cast(current) & 0x1Fu), tag_handler); - - case 0x98: // array (one-byte uint8_t for n follows) - { - std::uint8_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast(len), tag_handler); - } - - case 0x99: // array (two-byte uint16_t for n follow) - { - std::uint16_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast(len), tag_handler); - } - - case 0x9A: // array (four-byte uint32_t for n follow) - { - std::uint32_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast(len), tag_handler); - } - - case 0x9B: // array (eight-byte uint64_t for n follow) - { - std::uint64_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast(len), tag_handler); - } - - case 0x9F: // array (indefinite length) - return get_cbor_array(std::size_t(-1), tag_handler); - - // map (0x00..0x17 pairs of data items follow) - case 0xA0: - case 0xA1: - case 0xA2: - case 0xA3: - case 0xA4: - case 0xA5: - case 0xA6: - case 0xA7: - case 0xA8: - case 0xA9: - case 0xAA: - case 0xAB: - case 0xAC: - case 0xAD: - case 0xAE: - case 0xAF: - case 0xB0: - case 0xB1: - case 0xB2: - case 0xB3: - case 0xB4: - case 0xB5: - case 0xB6: - case 0xB7: - return get_cbor_object(static_cast(static_cast(current) & 0x1Fu), tag_handler); - - case 0xB8: // map (one-byte uint8_t for n follows) - { - std::uint8_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast(len), tag_handler); - } - - case 0xB9: // map (two-byte uint16_t for n follow) - { - std::uint16_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast(len), tag_handler); - } - - case 0xBA: // map (four-byte uint32_t for n follow) - { - std::uint32_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast(len), tag_handler); - } - - case 0xBB: // map (eight-byte uint64_t for n follow) - { - std::uint64_t len{}; - return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast(len), tag_handler); - } - - case 0xBF: // map (indefinite length) - return get_cbor_object(std::size_t(-1), tag_handler); - - case 0xC6: // tagged item - case 0xC7: - case 0xC8: - case 0xC9: - case 0xCA: - case 0xCB: - case 0xCC: - case 0xCD: - case 0xCE: - case 0xCF: - case 0xD0: - case 0xD1: - case 0xD2: - case 0xD3: - case 0xD4: - case 0xD8: // tagged item (1 bytes follow) - case 0xD9: // tagged item (2 bytes follow) - case 0xDA: // tagged item (4 bytes follow) - case 0xDB: // tagged item (8 bytes follow) - { - switch (tag_handler) - { - case cbor_tag_handler_t::error: - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::cbor, "invalid byte: 0x" + last_token, "value"))); - } - - case cbor_tag_handler_t::ignore: - { - switch (current) - { - case 0xD8: - { - std::uint8_t len{}; - get_number(input_format_t::cbor, len); - break; - } - case 0xD9: - { - std::uint16_t len{}; - get_number(input_format_t::cbor, len); - break; - } - case 0xDA: - { - std::uint32_t len{}; - get_number(input_format_t::cbor, len); - break; - } - case 0xDB: - { - std::uint64_t len{}; - get_number(input_format_t::cbor, len); - break; - } - default: - break; - } - return parse_cbor_internal(true, tag_handler); - } - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - } - - case 0xF4: // false - return sax->boolean(false); - - case 0xF5: // true - return sax->boolean(true); - - case 0xF6: // null - return sax->null(); - - case 0xF9: // Half-Precision Float (two-byte IEEE 754) - { - const auto byte1_raw = get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "number"))) - { - return false; - } - const auto byte2_raw = get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "number"))) - { - return false; - } - - const auto byte1 = static_cast(byte1_raw); - const auto byte2 = static_cast(byte2_raw); - - // code from RFC 7049, Appendix D, Figure 3: - // As half-precision floating-point numbers were only added - // to IEEE 754 in 2008, today's programming platforms often - // still only have limited support for them. It is very - // easy to include at least decoding support for them even - // without such support. An example of a small decoder for - // half-precision floating-point numbers in the C language - // is shown in Fig. 3. - const auto half = static_cast((byte1 << 8u) + byte2); - const double val = [&half] - { - const int exp = (half >> 10u) & 0x1Fu; - const unsigned int mant = half & 0x3FFu; - JSON_ASSERT(0 <= exp&& exp <= 32); - JSON_ASSERT(mant <= 1024); - switch (exp) - { - case 0: - return std::ldexp(mant, -24); - case 31: - return (mant == 0) - ? std::numeric_limits::infinity() - : std::numeric_limits::quiet_NaN(); - default: - return std::ldexp(mant + 1024, exp - 25); - } - }(); - return sax->number_float((half & 0x8000u) != 0 - ? static_cast(-val) - : static_cast(val), ""); - } - - case 0xFA: // Single-Precision Float (four-byte IEEE 754) - { - float number{}; - return get_number(input_format_t::cbor, number) && sax->number_float(static_cast(number), ""); - } - - case 0xFB: // Double-Precision Float (eight-byte IEEE 754) - { - double number{}; - return get_number(input_format_t::cbor, number) && sax->number_float(static_cast(number), ""); - } - - default: // anything else (0xFF is handled inside the other types) - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::cbor, "invalid byte: 0x" + last_token, "value"))); - } - } - } - - /*! - @brief reads a CBOR string - - This function first reads starting bytes to determine the expected - string length and then copies this number of bytes into a string. - Additionally, CBOR's strings with indefinite lengths are supported. - - @param[out] result created string - - @return whether string creation completed - */ - bool get_cbor_string(string_t& result) - { - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string"))) - { - return false; - } - - switch (current) - { - // UTF-8 string (0x00..0x17 bytes follow) - case 0x60: - case 0x61: - case 0x62: - case 0x63: - case 0x64: - case 0x65: - case 0x66: - case 0x67: - case 0x68: - case 0x69: - case 0x6A: - case 0x6B: - case 0x6C: - case 0x6D: - case 0x6E: - case 0x6F: - case 0x70: - case 0x71: - case 0x72: - case 0x73: - case 0x74: - case 0x75: - case 0x76: - case 0x77: - { - return get_string(input_format_t::cbor, static_cast(current) & 0x1Fu, result); - } - - case 0x78: // UTF-8 string (one-byte uint8_t for n follows) - { - std::uint8_t len{}; - return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result); - } - - case 0x79: // UTF-8 string (two-byte uint16_t for n follow) - { - std::uint16_t len{}; - return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result); - } - - case 0x7A: // UTF-8 string (four-byte uint32_t for n follow) - { - std::uint32_t len{}; - return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result); - } - - case 0x7B: // UTF-8 string (eight-byte uint64_t for n follow) - { - std::uint64_t len{}; - return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result); - } - - case 0x7F: // UTF-8 string (indefinite length) - { - while (get() != 0xFF) - { - string_t chunk; - if (!get_cbor_string(chunk)) - { - return false; - } - result.append(chunk); - } - return true; - } - - default: - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::cbor, "expected length specification (0x60-0x7B) or indefinite string type (0x7F); last byte: 0x" + last_token, "string"))); - } - } - } - - /*! - @brief reads a CBOR byte array - - This function first reads starting bytes to determine the expected - byte array length and then copies this number of bytes into the byte array. - Additionally, CBOR's byte arrays with indefinite lengths are supported. - - @param[out] result created byte array - - @return whether byte array creation completed - */ - bool get_cbor_binary(binary_t& result) - { - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary"))) - { - return false; - } - - switch (current) - { - // Binary data (0x00..0x17 bytes follow) - case 0x40: - case 0x41: - case 0x42: - case 0x43: - case 0x44: - case 0x45: - case 0x46: - case 0x47: - case 0x48: - case 0x49: - case 0x4A: - case 0x4B: - case 0x4C: - case 0x4D: - case 0x4E: - case 0x4F: - case 0x50: - case 0x51: - case 0x52: - case 0x53: - case 0x54: - case 0x55: - case 0x56: - case 0x57: - { - return get_binary(input_format_t::cbor, static_cast(current) & 0x1Fu, result); - } - - case 0x58: // Binary data (one-byte uint8_t for n follows) - { - std::uint8_t len{}; - return get_number(input_format_t::cbor, len) && - get_binary(input_format_t::cbor, len, result); - } - - case 0x59: // Binary data (two-byte uint16_t for n follow) - { - std::uint16_t len{}; - return get_number(input_format_t::cbor, len) && - get_binary(input_format_t::cbor, len, result); - } - - case 0x5A: // Binary data (four-byte uint32_t for n follow) - { - std::uint32_t len{}; - return get_number(input_format_t::cbor, len) && - get_binary(input_format_t::cbor, len, result); - } - - case 0x5B: // Binary data (eight-byte uint64_t for n follow) - { - std::uint64_t len{}; - return get_number(input_format_t::cbor, len) && - get_binary(input_format_t::cbor, len, result); - } - - case 0x5F: // Binary data (indefinite length) - { - while (get() != 0xFF) - { - binary_t chunk; - if (!get_cbor_binary(chunk)) - { - return false; - } - result.insert(result.end(), chunk.begin(), chunk.end()); - } - return true; - } - - default: - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::cbor, "expected length specification (0x40-0x5B) or indefinite binary array type (0x5F); last byte: 0x" + last_token, "binary"))); - } - } - } - - /*! - @param[in] len the length of the array or std::size_t(-1) for an - array of indefinite size - @param[in] tag_handler how CBOR tags should be treated - @return whether array creation completed - */ - bool get_cbor_array(const std::size_t len, - const cbor_tag_handler_t tag_handler) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(len))) - { - return false; - } - - if (len != std::size_t(-1)) - { - for (std::size_t i = 0; i < len; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler))) - { - return false; - } - } - } - else - { - while (get() != 0xFF) - { - if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(false, tag_handler))) - { - return false; - } - } - } - - return sax->end_array(); - } - - /*! - @param[in] len the length of the object or std::size_t(-1) for an - object of indefinite size - @param[in] tag_handler how CBOR tags should be treated - @return whether object creation completed - */ - bool get_cbor_object(const std::size_t len, - const cbor_tag_handler_t tag_handler) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(len))) - { - return false; - } - - string_t key; - if (len != std::size_t(-1)) - { - for (std::size_t i = 0; i < len; ++i) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!get_cbor_string(key) || !sax->key(key))) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler))) - { - return false; - } - key.clear(); - } - } - else - { - while (get() != 0xFF) - { - if (JSON_HEDLEY_UNLIKELY(!get_cbor_string(key) || !sax->key(key))) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler))) - { - return false; - } - key.clear(); - } - } - - return sax->end_object(); - } - - ///////////// - // MsgPack // - ///////////// - - /*! - @return whether a valid MessagePack value was passed to the SAX parser - */ - bool parse_msgpack_internal() - { - switch (get()) - { - // EOF - case std::char_traits::eof(): - return unexpect_eof(input_format_t::msgpack, "value"); - - // positive fixint - case 0x00: - case 0x01: - case 0x02: - case 0x03: - case 0x04: - case 0x05: - case 0x06: - case 0x07: - case 0x08: - case 0x09: - case 0x0A: - case 0x0B: - case 0x0C: - case 0x0D: - case 0x0E: - case 0x0F: - case 0x10: - case 0x11: - case 0x12: - case 0x13: - case 0x14: - case 0x15: - case 0x16: - case 0x17: - case 0x18: - case 0x19: - case 0x1A: - case 0x1B: - case 0x1C: - case 0x1D: - case 0x1E: - case 0x1F: - case 0x20: - case 0x21: - case 0x22: - case 0x23: - case 0x24: - case 0x25: - case 0x26: - case 0x27: - case 0x28: - case 0x29: - case 0x2A: - case 0x2B: - case 0x2C: - case 0x2D: - case 0x2E: - case 0x2F: - case 0x30: - case 0x31: - case 0x32: - case 0x33: - case 0x34: - case 0x35: - case 0x36: - case 0x37: - case 0x38: - case 0x39: - case 0x3A: - case 0x3B: - case 0x3C: - case 0x3D: - case 0x3E: - case 0x3F: - case 0x40: - case 0x41: - case 0x42: - case 0x43: - case 0x44: - case 0x45: - case 0x46: - case 0x47: - case 0x48: - case 0x49: - case 0x4A: - case 0x4B: - case 0x4C: - case 0x4D: - case 0x4E: - case 0x4F: - case 0x50: - case 0x51: - case 0x52: - case 0x53: - case 0x54: - case 0x55: - case 0x56: - case 0x57: - case 0x58: - case 0x59: - case 0x5A: - case 0x5B: - case 0x5C: - case 0x5D: - case 0x5E: - case 0x5F: - case 0x60: - case 0x61: - case 0x62: - case 0x63: - case 0x64: - case 0x65: - case 0x66: - case 0x67: - case 0x68: - case 0x69: - case 0x6A: - case 0x6B: - case 0x6C: - case 0x6D: - case 0x6E: - case 0x6F: - case 0x70: - case 0x71: - case 0x72: - case 0x73: - case 0x74: - case 0x75: - case 0x76: - case 0x77: - case 0x78: - case 0x79: - case 0x7A: - case 0x7B: - case 0x7C: - case 0x7D: - case 0x7E: - case 0x7F: - return sax->number_unsigned(static_cast(current)); - - // fixmap - case 0x80: - case 0x81: - case 0x82: - case 0x83: - case 0x84: - case 0x85: - case 0x86: - case 0x87: - case 0x88: - case 0x89: - case 0x8A: - case 0x8B: - case 0x8C: - case 0x8D: - case 0x8E: - case 0x8F: - return get_msgpack_object(static_cast(static_cast(current) & 0x0Fu)); - - // fixarray - case 0x90: - case 0x91: - case 0x92: - case 0x93: - case 0x94: - case 0x95: - case 0x96: - case 0x97: - case 0x98: - case 0x99: - case 0x9A: - case 0x9B: - case 0x9C: - case 0x9D: - case 0x9E: - case 0x9F: - return get_msgpack_array(static_cast(static_cast(current) & 0x0Fu)); - - // fixstr - case 0xA0: - case 0xA1: - case 0xA2: - case 0xA3: - case 0xA4: - case 0xA5: - case 0xA6: - case 0xA7: - case 0xA8: - case 0xA9: - case 0xAA: - case 0xAB: - case 0xAC: - case 0xAD: - case 0xAE: - case 0xAF: - case 0xB0: - case 0xB1: - case 0xB2: - case 0xB3: - case 0xB4: - case 0xB5: - case 0xB6: - case 0xB7: - case 0xB8: - case 0xB9: - case 0xBA: - case 0xBB: - case 0xBC: - case 0xBD: - case 0xBE: - case 0xBF: - case 0xD9: // str 8 - case 0xDA: // str 16 - case 0xDB: // str 32 - { - string_t s; - return get_msgpack_string(s) && sax->string(s); - } - - case 0xC0: // nil - return sax->null(); - - case 0xC2: // false - return sax->boolean(false); - - case 0xC3: // true - return sax->boolean(true); - - case 0xC4: // bin 8 - case 0xC5: // bin 16 - case 0xC6: // bin 32 - case 0xC7: // ext 8 - case 0xC8: // ext 16 - case 0xC9: // ext 32 - case 0xD4: // fixext 1 - case 0xD5: // fixext 2 - case 0xD6: // fixext 4 - case 0xD7: // fixext 8 - case 0xD8: // fixext 16 - { - binary_t b; - return get_msgpack_binary(b) && sax->binary(b); - } - - case 0xCA: // float 32 - { - float number{}; - return get_number(input_format_t::msgpack, number) && sax->number_float(static_cast(number), ""); - } - - case 0xCB: // float 64 - { - double number{}; - return get_number(input_format_t::msgpack, number) && sax->number_float(static_cast(number), ""); - } - - case 0xCC: // uint 8 - { - std::uint8_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number); - } - - case 0xCD: // uint 16 - { - std::uint16_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number); - } - - case 0xCE: // uint 32 - { - std::uint32_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number); - } - - case 0xCF: // uint 64 - { - std::uint64_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number); - } - - case 0xD0: // int 8 - { - std::int8_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_integer(number); - } - - case 0xD1: // int 16 - { - std::int16_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_integer(number); - } - - case 0xD2: // int 32 - { - std::int32_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_integer(number); - } - - case 0xD3: // int 64 - { - std::int64_t number{}; - return get_number(input_format_t::msgpack, number) && sax->number_integer(number); - } - - case 0xDC: // array 16 - { - std::uint16_t len{}; - return get_number(input_format_t::msgpack, len) && get_msgpack_array(static_cast(len)); - } - - case 0xDD: // array 32 - { - std::uint32_t len{}; - return get_number(input_format_t::msgpack, len) && get_msgpack_array(static_cast(len)); - } - - case 0xDE: // map 16 - { - std::uint16_t len{}; - return get_number(input_format_t::msgpack, len) && get_msgpack_object(static_cast(len)); - } - - case 0xDF: // map 32 - { - std::uint32_t len{}; - return get_number(input_format_t::msgpack, len) && get_msgpack_object(static_cast(len)); - } - - // negative fixint - case 0xE0: - case 0xE1: - case 0xE2: - case 0xE3: - case 0xE4: - case 0xE5: - case 0xE6: - case 0xE7: - case 0xE8: - case 0xE9: - case 0xEA: - case 0xEB: - case 0xEC: - case 0xED: - case 0xEE: - case 0xEF: - case 0xF0: - case 0xF1: - case 0xF2: - case 0xF3: - case 0xF4: - case 0xF5: - case 0xF6: - case 0xF7: - case 0xF8: - case 0xF9: - case 0xFA: - case 0xFB: - case 0xFC: - case 0xFD: - case 0xFE: - case 0xFF: - return sax->number_integer(static_cast(current)); - - default: // anything else - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::msgpack, "invalid byte: 0x" + last_token, "value"))); - } - } - } - - /*! - @brief reads a MessagePack string - - This function first reads starting bytes to determine the expected - string length and then copies this number of bytes into a string. - - @param[out] result created string - - @return whether string creation completed - */ - bool get_msgpack_string(string_t& result) - { - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::msgpack, "string"))) - { - return false; - } - - switch (current) - { - // fixstr - case 0xA0: - case 0xA1: - case 0xA2: - case 0xA3: - case 0xA4: - case 0xA5: - case 0xA6: - case 0xA7: - case 0xA8: - case 0xA9: - case 0xAA: - case 0xAB: - case 0xAC: - case 0xAD: - case 0xAE: - case 0xAF: - case 0xB0: - case 0xB1: - case 0xB2: - case 0xB3: - case 0xB4: - case 0xB5: - case 0xB6: - case 0xB7: - case 0xB8: - case 0xB9: - case 0xBA: - case 0xBB: - case 0xBC: - case 0xBD: - case 0xBE: - case 0xBF: - { - return get_string(input_format_t::msgpack, static_cast(current) & 0x1Fu, result); - } - - case 0xD9: // str 8 - { - std::uint8_t len{}; - return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result); - } - - case 0xDA: // str 16 - { - std::uint16_t len{}; - return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result); - } - - case 0xDB: // str 32 - { - std::uint32_t len{}; - return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result); - } - - default: - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::msgpack, "expected length specification (0xA0-0xBF, 0xD9-0xDB); last byte: 0x" + last_token, "string"))); - } - } - } - - /*! - @brief reads a MessagePack byte array - - This function first reads starting bytes to determine the expected - byte array length and then copies this number of bytes into a byte array. - - @param[out] result created byte array - - @return whether byte array creation completed - */ - bool get_msgpack_binary(binary_t& result) - { - // helper function to set the subtype - auto assign_and_return_true = [&result](std::int8_t subtype) - { - result.set_subtype(static_cast(subtype)); - return true; - }; - - switch (current) - { - case 0xC4: // bin 8 - { - std::uint8_t len{}; - return get_number(input_format_t::msgpack, len) && - get_binary(input_format_t::msgpack, len, result); - } - - case 0xC5: // bin 16 - { - std::uint16_t len{}; - return get_number(input_format_t::msgpack, len) && - get_binary(input_format_t::msgpack, len, result); - } - - case 0xC6: // bin 32 - { - std::uint32_t len{}; - return get_number(input_format_t::msgpack, len) && - get_binary(input_format_t::msgpack, len, result); - } - - case 0xC7: // ext 8 - { - std::uint8_t len{}; - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, len) && - get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, len, result) && - assign_and_return_true(subtype); - } - - case 0xC8: // ext 16 - { - std::uint16_t len{}; - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, len) && - get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, len, result) && - assign_and_return_true(subtype); - } - - case 0xC9: // ext 32 - { - std::uint32_t len{}; - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, len) && - get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, len, result) && - assign_and_return_true(subtype); - } - - case 0xD4: // fixext 1 - { - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, 1, result) && - assign_and_return_true(subtype); - } - - case 0xD5: // fixext 2 - { - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, 2, result) && - assign_and_return_true(subtype); - } - - case 0xD6: // fixext 4 - { - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, 4, result) && - assign_and_return_true(subtype); - } - - case 0xD7: // fixext 8 - { - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, 8, result) && - assign_and_return_true(subtype); - } - - case 0xD8: // fixext 16 - { - std::int8_t subtype{}; - return get_number(input_format_t::msgpack, subtype) && - get_binary(input_format_t::msgpack, 16, result) && - assign_and_return_true(subtype); - } - - default: // LCOV_EXCL_LINE - return false; // LCOV_EXCL_LINE - } - } - - /*! - @param[in] len the length of the array - @return whether array creation completed - */ - bool get_msgpack_array(const std::size_t len) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(len))) - { - return false; - } - - for (std::size_t i = 0; i < len; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!parse_msgpack_internal())) - { - return false; - } - } - - return sax->end_array(); - } - - /*! - @param[in] len the length of the object - @return whether object creation completed - */ - bool get_msgpack_object(const std::size_t len) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(len))) - { - return false; - } - - string_t key; - for (std::size_t i = 0; i < len; ++i) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!get_msgpack_string(key) || !sax->key(key))) - { - return false; - } - - if (JSON_HEDLEY_UNLIKELY(!parse_msgpack_internal())) - { - return false; - } - key.clear(); - } - - return sax->end_object(); - } - - //////////// - // UBJSON // - //////////// - - /*! - @param[in] get_char whether a new character should be retrieved from the - input (true, default) or whether the last read - character should be considered instead - - @return whether a valid UBJSON value was passed to the SAX parser - */ - bool parse_ubjson_internal(const bool get_char = true) - { - return get_ubjson_value(get_char ? get_ignore_noop() : current); - } - - /*! - @brief reads a UBJSON string - - This function is either called after reading the 'S' byte explicitly - indicating a string, or in case of an object key where the 'S' byte can be - left out. - - @param[out] result created string - @param[in] get_char whether a new character should be retrieved from the - input (true, default) or whether the last read - character should be considered instead - - @return whether string creation completed - */ - bool get_ubjson_string(string_t& result, const bool get_char = true) - { - if (get_char) - { - get(); // TODO(niels): may we ignore N here? - } - - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "value"))) - { - return false; - } - - switch (current) - { - case 'U': - { - std::uint8_t len{}; - return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result); - } - - case 'i': - { - std::int8_t len{}; - return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result); - } - - case 'I': - { - std::int16_t len{}; - return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result); - } - - case 'l': - { - std::int32_t len{}; - return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result); - } - - case 'L': - { - std::int64_t len{}; - return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result); - } - - default: - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "expected length type specification (U, i, I, l, L); last byte: 0x" + last_token, "string"))); - } - } - - /*! - @param[out] result determined size - @return whether size determination completed - */ - bool get_ubjson_size_value(std::size_t& result) - { - switch (get_ignore_noop()) - { - case 'U': - { - std::uint8_t number{}; - if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number))) - { - return false; - } - result = static_cast(number); - return true; - } - - case 'i': - { - std::int8_t number{}; - if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number))) - { - return false; - } - result = static_cast(number); - return true; - } - - case 'I': - { - std::int16_t number{}; - if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number))) - { - return false; - } - result = static_cast(number); - return true; - } - - case 'l': - { - std::int32_t number{}; - if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number))) - { - return false; - } - result = static_cast(number); - return true; - } - - case 'L': - { - std::int64_t number{}; - if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number))) - { - return false; - } - result = static_cast(number); - return true; - } - - default: - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "expected length type specification (U, i, I, l, L) after '#'; last byte: 0x" + last_token, "size"))); - } - } - } - - /*! - @brief determine the type and size for a container - - In the optimized UBJSON format, a type and a size can be provided to allow - for a more compact representation. - - @param[out] result pair of the size and the type - - @return whether pair creation completed - */ - bool get_ubjson_size_type(std::pair& result) - { - result.first = string_t::npos; // size - result.second = 0; // type - - get_ignore_noop(); - - if (current == '$') - { - result.second = get(); // must not ignore 'N', because 'N' maybe the type - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "type"))) - { - return false; - } - - get_ignore_noop(); - if (JSON_HEDLEY_UNLIKELY(current != '#')) - { - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "value"))) - { - return false; - } - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::ubjson, "expected '#' after type information; last byte: 0x" + last_token, "size"))); - } - - return get_ubjson_size_value(result.first); - } - - if (current == '#') - { - return get_ubjson_size_value(result.first); - } - - return true; - } - - /*! - @param prefix the previously read or set type prefix - @return whether value creation completed - */ - bool get_ubjson_value(const char_int_type prefix) - { - switch (prefix) - { - case std::char_traits::eof(): // EOF - return unexpect_eof(input_format_t::ubjson, "value"); - - case 'T': // true - return sax->boolean(true); - case 'F': // false - return sax->boolean(false); - - case 'Z': // null - return sax->null(); - - case 'U': - { - std::uint8_t number{}; - return get_number(input_format_t::ubjson, number) && sax->number_unsigned(number); - } - - case 'i': - { - std::int8_t number{}; - return get_number(input_format_t::ubjson, number) && sax->number_integer(number); - } - - case 'I': - { - std::int16_t number{}; - return get_number(input_format_t::ubjson, number) && sax->number_integer(number); - } - - case 'l': - { - std::int32_t number{}; - return get_number(input_format_t::ubjson, number) && sax->number_integer(number); - } - - case 'L': - { - std::int64_t number{}; - return get_number(input_format_t::ubjson, number) && sax->number_integer(number); - } - - case 'd': - { - float number{}; - return get_number(input_format_t::ubjson, number) && sax->number_float(static_cast(number), ""); - } - - case 'D': - { - double number{}; - return get_number(input_format_t::ubjson, number) && sax->number_float(static_cast(number), ""); - } - - case 'H': - { - return get_ubjson_high_precision_number(); - } - - case 'C': // char - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "char"))) - { - return false; - } - if (JSON_HEDLEY_UNLIKELY(current > 127)) - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "byte after 'C' must be in range 0x00..0x7F; last byte: 0x" + last_token, "char"))); - } - string_t s(1, static_cast(current)); - return sax->string(s); - } - - case 'S': // string - { - string_t s; - return get_ubjson_string(s) && sax->string(s); - } - - case '[': // array - return get_ubjson_array(); - - case '{': // object - return get_ubjson_object(); - - default: // anything else - { - auto last_token = get_token_string(); - return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::ubjson, "invalid byte: 0x" + last_token, "value"))); - } - } - } - - /*! - @return whether array creation completed - */ - bool get_ubjson_array() - { - std::pair size_and_type; - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_type(size_and_type))) - { - return false; - } - - if (size_and_type.first != string_t::npos) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(size_and_type.first))) - { - return false; - } - - if (size_and_type.second != 0) - { - if (size_and_type.second != 'N') - { - for (std::size_t i = 0; i < size_and_type.first; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_value(size_and_type.second))) - { - return false; - } - } - } - } - else - { - for (std::size_t i = 0; i < size_and_type.first; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal())) - { - return false; - } - } - } - } - else - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(std::size_t(-1)))) - { - return false; - } - - while (current != ']') - { - if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal(false))) - { - return false; - } - get_ignore_noop(); - } - } - - return sax->end_array(); - } - - /*! - @return whether object creation completed - */ - bool get_ubjson_object() - { - std::pair size_and_type; - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_type(size_and_type))) - { - return false; - } - - string_t key; - if (size_and_type.first != string_t::npos) - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(size_and_type.first))) - { - return false; - } - - if (size_and_type.second != 0) - { - for (std::size_t i = 0; i < size_and_type.first; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key) || !sax->key(key))) - { - return false; - } - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_value(size_and_type.second))) - { - return false; - } - key.clear(); - } - } - else - { - for (std::size_t i = 0; i < size_and_type.first; ++i) - { - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key) || !sax->key(key))) - { - return false; - } - if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal())) - { - return false; - } - key.clear(); - } - } - } - else - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(std::size_t(-1)))) - { - return false; - } - - while (current != '}') - { - if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key, false) || !sax->key(key))) - { - return false; - } - if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal())) - { - return false; - } - get_ignore_noop(); - key.clear(); - } - } - - return sax->end_object(); - } - - // Note, no reader for UBJSON binary types is implemented because they do - // not exist - - bool get_ubjson_high_precision_number() - { - // get size of following number string - std::size_t size{}; - auto res = get_ubjson_size_value(size); - if (JSON_HEDLEY_UNLIKELY(!res)) - { - return res; - } - - // get number string - std::vector number_vector; - for (std::size_t i = 0; i < size; ++i) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "number"))) - { - return false; - } - number_vector.push_back(static_cast(current)); - } - - // parse number string - auto number_ia = detail::input_adapter(std::forward(number_vector)); - auto number_lexer = detail::lexer(std::move(number_ia), false); - const auto result_number = number_lexer.scan(); - const auto number_string = number_lexer.get_token_string(); - const auto result_remainder = number_lexer.scan(); - - using token_type = typename detail::lexer_base::token_type; - - if (JSON_HEDLEY_UNLIKELY(result_remainder != token_type::end_of_input)) - { - return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number"))); - } - - switch (result_number) - { - case token_type::value_integer: - return sax->number_integer(number_lexer.get_number_integer()); - case token_type::value_unsigned: - return sax->number_unsigned(number_lexer.get_number_unsigned()); - case token_type::value_float: - return sax->number_float(number_lexer.get_number_float(), std::move(number_string)); - default: - return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number"))); - } - } - - /////////////////////// - // Utility functions // - /////////////////////// - - /*! - @brief get next character from the input - - This function provides the interface to the used input adapter. It does - not throw in case the input reached EOF, but returns a -'ve valued - `std::char_traits::eof()` in that case. - - @return character read from the input - */ - char_int_type get() - { - ++chars_read; - return current = ia.get_character(); - } - - /*! - @return character read from the input after ignoring all 'N' entries - */ - char_int_type get_ignore_noop() - { - do - { - get(); - } - while (current == 'N'); - - return current; - } - - /* - @brief read a number from the input - - @tparam NumberType the type of the number - @param[in] format the current format (for diagnostics) - @param[out] result number of type @a NumberType - - @return whether conversion completed - - @note This function needs to respect the system's endianess, because - bytes in CBOR, MessagePack, and UBJSON are stored in network order - (big endian) and therefore need reordering on little endian systems. - */ - template - bool get_number(const input_format_t format, NumberType& result) - { - // step 1: read input into array with system's byte order - std::array vec; - for (std::size_t i = 0; i < sizeof(NumberType); ++i) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "number"))) - { - return false; - } - - // reverse byte order prior to conversion if necessary - if (is_little_endian != InputIsLittleEndian) - { - vec[sizeof(NumberType) - i - 1] = static_cast(current); - } - else - { - vec[i] = static_cast(current); // LCOV_EXCL_LINE - } - } - - // step 2: convert array into number of type T and return - std::memcpy(&result, vec.data(), sizeof(NumberType)); - return true; - } - - /*! - @brief create a string by reading characters from the input - - @tparam NumberType the type of the number - @param[in] format the current format (for diagnostics) - @param[in] len number of characters to read - @param[out] result string created by reading @a len bytes - - @return whether string creation completed - - @note We can not reserve @a len bytes for the result, because @a len - may be too large. Usually, @ref unexpect_eof() detects the end of - the input before we run out of string memory. - */ - template - bool get_string(const input_format_t format, - const NumberType len, - string_t& result) - { - bool success = true; - for (NumberType i = 0; i < len; i++) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "string"))) - { - success = false; - break; - } - result.push_back(static_cast(current)); - }; - return success; - } - - /*! - @brief create a byte array by reading bytes from the input - - @tparam NumberType the type of the number - @param[in] format the current format (for diagnostics) - @param[in] len number of bytes to read - @param[out] result byte array created by reading @a len bytes - - @return whether byte array creation completed - - @note We can not reserve @a len bytes for the result, because @a len - may be too large. Usually, @ref unexpect_eof() detects the end of - the input before we run out of memory. - */ - template - bool get_binary(const input_format_t format, - const NumberType len, - binary_t& result) - { - bool success = true; - for (NumberType i = 0; i < len; i++) - { - get(); - if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "binary"))) - { - success = false; - break; - } - result.push_back(static_cast(current)); - } - return success; - } - - /*! - @param[in] format the current format (for diagnostics) - @param[in] context further context information (for diagnostics) - @return whether the last read character is not EOF - */ - JSON_HEDLEY_NON_NULL(3) - bool unexpect_eof(const input_format_t format, const char* context) const - { - if (JSON_HEDLEY_UNLIKELY(current == std::char_traits::eof())) - { - return sax->parse_error(chars_read, "", - parse_error::create(110, chars_read, exception_message(format, "unexpected end of input", context))); - } - return true; - } - - /*! - @return a string representation of the last read byte - */ - std::string get_token_string() const - { - std::array cr{{}}; - (std::snprintf)(cr.data(), cr.size(), "%.2hhX", static_cast(current)); - return std::string{cr.data()}; - } - - /*! - @param[in] format the current format - @param[in] detail a detailed error message - @param[in] context further context information - @return a message string to use in the parse_error exceptions - */ - std::string exception_message(const input_format_t format, - const std::string& detail, - const std::string& context) const - { - std::string error_msg = "syntax error while parsing "; - - switch (format) - { - case input_format_t::cbor: - error_msg += "CBOR"; - break; - - case input_format_t::msgpack: - error_msg += "MessagePack"; - break; - - case input_format_t::ubjson: - error_msg += "UBJSON"; - break; - - case input_format_t::bson: - error_msg += "BSON"; - break; - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - - return error_msg + " " + context + ": " + detail; - } - - private: - /// input adapter - InputAdapterType ia; - - /// the current character - char_int_type current = std::char_traits::eof(); - - /// the number of characters read - std::size_t chars_read = 0; - - /// whether we can assume little endianess - const bool is_little_endian = little_endianess(); - - /// the SAX parser - json_sax_t* sax = nullptr; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/input_adapters.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/input_adapters.hpp deleted file mode 100644 index 63921ca55ce..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/input_adapters.hpp +++ /dev/null @@ -1,454 +0,0 @@ -#pragma once - -#include // array -#include // size_t -#include //FILE * -#include // strlen -#include // istream -#include // begin, end, iterator_traits, random_access_iterator_tag, distance, next -#include // shared_ptr, make_shared, addressof -#include // accumulate -#include // string, char_traits -#include // enable_if, is_base_of, is_pointer, is_integral, remove_pointer -#include // pair, declval - -#include -#include - -namespace nlohmann -{ -namespace detail -{ -/// the supported input formats -enum class input_format_t { json, cbor, msgpack, ubjson, bson }; - -//////////////////// -// input adapters // -//////////////////// - -/*! -Input adapter for stdio file access. This adapter read only 1 byte and do not use any - buffer. This adapter is a very low level adapter. -*/ -class file_input_adapter -{ - public: - using char_type = char; - - JSON_HEDLEY_NON_NULL(2) - explicit file_input_adapter(std::FILE* f) noexcept - : m_file(f) - {} - - // make class move-only - file_input_adapter(const file_input_adapter&) = delete; - file_input_adapter(file_input_adapter&&) = default; - file_input_adapter& operator=(const file_input_adapter&) = delete; - file_input_adapter& operator=(file_input_adapter&&) = delete; - - std::char_traits::int_type get_character() noexcept - { - return std::fgetc(m_file); - } - - private: - /// the file pointer to read from - std::FILE* m_file; -}; - - -/*! -Input adapter for a (caching) istream. Ignores a UFT Byte Order Mark at -beginning of input. Does not support changing the underlying std::streambuf -in mid-input. Maintains underlying std::istream and std::streambuf to support -subsequent use of standard std::istream operations to process any input -characters following those used in parsing the JSON input. Clears the -std::istream flags; any input errors (e.g., EOF) will be detected by the first -subsequent call for input from the std::istream. -*/ -class input_stream_adapter -{ - public: - using char_type = char; - - ~input_stream_adapter() - { - // clear stream flags; we use underlying streambuf I/O, do not - // maintain ifstream flags, except eof - if (is != nullptr) - { - is->clear(is->rdstate() & std::ios::eofbit); - } - } - - explicit input_stream_adapter(std::istream& i) - : is(&i), sb(i.rdbuf()) - {} - - // delete because of pointer members - input_stream_adapter(const input_stream_adapter&) = delete; - input_stream_adapter& operator=(input_stream_adapter&) = delete; - input_stream_adapter& operator=(input_stream_adapter&& rhs) = delete; - - input_stream_adapter(input_stream_adapter&& rhs) noexcept : is(rhs.is), sb(rhs.sb) - { - rhs.is = nullptr; - rhs.sb = nullptr; - } - - // std::istream/std::streambuf use std::char_traits::to_int_type, to - // ensure that std::char_traits::eof() and the character 0xFF do not - // end up as the same value, eg. 0xFFFFFFFF. - std::char_traits::int_type get_character() - { - auto res = sb->sbumpc(); - // set eof manually, as we don't use the istream interface. - if (JSON_HEDLEY_UNLIKELY(res == EOF)) - { - is->clear(is->rdstate() | std::ios::eofbit); - } - return res; - } - - private: - /// the associated input stream - std::istream* is = nullptr; - std::streambuf* sb = nullptr; -}; - -// General-purpose iterator-based adapter. It might not be as fast as -// theoretically possible for some containers, but it is extremely versatile. -template -class iterator_input_adapter -{ - public: - using char_type = typename std::iterator_traits::value_type; - - iterator_input_adapter(IteratorType first, IteratorType last) - : current(std::move(first)), end(std::move(last)) {} - - typename std::char_traits::int_type get_character() - { - if (JSON_HEDLEY_LIKELY(current != end)) - { - auto result = std::char_traits::to_int_type(*current); - std::advance(current, 1); - return result; - } - else - { - return std::char_traits::eof(); - } - } - - private: - IteratorType current; - IteratorType end; - - template - friend struct wide_string_input_helper; - - bool empty() const - { - return current == end; - } - -}; - - -template -struct wide_string_input_helper; - -template -struct wide_string_input_helper -{ - // UTF-32 - static void fill_buffer(BaseInputAdapter& input, - std::array::int_type, 4>& utf8_bytes, - size_t& utf8_bytes_index, - size_t& utf8_bytes_filled) - { - utf8_bytes_index = 0; - - if (JSON_HEDLEY_UNLIKELY(input.empty())) - { - utf8_bytes[0] = std::char_traits::eof(); - utf8_bytes_filled = 1; - } - else - { - // get the current character - const auto wc = input.get_character(); - - // UTF-32 to UTF-8 encoding - if (wc < 0x80) - { - utf8_bytes[0] = static_cast::int_type>(wc); - utf8_bytes_filled = 1; - } - else if (wc <= 0x7FF) - { - utf8_bytes[0] = static_cast::int_type>(0xC0u | ((static_cast(wc) >> 6u) & 0x1Fu)); - utf8_bytes[1] = static_cast::int_type>(0x80u | (static_cast(wc) & 0x3Fu)); - utf8_bytes_filled = 2; - } - else if (wc <= 0xFFFF) - { - utf8_bytes[0] = static_cast::int_type>(0xE0u | ((static_cast(wc) >> 12u) & 0x0Fu)); - utf8_bytes[1] = static_cast::int_type>(0x80u | ((static_cast(wc) >> 6u) & 0x3Fu)); - utf8_bytes[2] = static_cast::int_type>(0x80u | (static_cast(wc) & 0x3Fu)); - utf8_bytes_filled = 3; - } - else if (wc <= 0x10FFFF) - { - utf8_bytes[0] = static_cast::int_type>(0xF0u | ((static_cast(wc) >> 18u) & 0x07u)); - utf8_bytes[1] = static_cast::int_type>(0x80u | ((static_cast(wc) >> 12u) & 0x3Fu)); - utf8_bytes[2] = static_cast::int_type>(0x80u | ((static_cast(wc) >> 6u) & 0x3Fu)); - utf8_bytes[3] = static_cast::int_type>(0x80u | (static_cast(wc) & 0x3Fu)); - utf8_bytes_filled = 4; - } - else - { - // unknown character - utf8_bytes[0] = static_cast::int_type>(wc); - utf8_bytes_filled = 1; - } - } - } -}; - -template -struct wide_string_input_helper -{ - // UTF-16 - static void fill_buffer(BaseInputAdapter& input, - std::array::int_type, 4>& utf8_bytes, - size_t& utf8_bytes_index, - size_t& utf8_bytes_filled) - { - utf8_bytes_index = 0; - - if (JSON_HEDLEY_UNLIKELY(input.empty())) - { - utf8_bytes[0] = std::char_traits::eof(); - utf8_bytes_filled = 1; - } - else - { - // get the current character - const auto wc = input.get_character(); - - // UTF-16 to UTF-8 encoding - if (wc < 0x80) - { - utf8_bytes[0] = static_cast::int_type>(wc); - utf8_bytes_filled = 1; - } - else if (wc <= 0x7FF) - { - utf8_bytes[0] = static_cast::int_type>(0xC0u | ((static_cast(wc) >> 6u))); - utf8_bytes[1] = static_cast::int_type>(0x80u | (static_cast(wc) & 0x3Fu)); - utf8_bytes_filled = 2; - } - else if (0xD800 > wc || wc >= 0xE000) - { - utf8_bytes[0] = static_cast::int_type>(0xE0u | ((static_cast(wc) >> 12u))); - utf8_bytes[1] = static_cast::int_type>(0x80u | ((static_cast(wc) >> 6u) & 0x3Fu)); - utf8_bytes[2] = static_cast::int_type>(0x80u | (static_cast(wc) & 0x3Fu)); - utf8_bytes_filled = 3; - } - else - { - if (JSON_HEDLEY_UNLIKELY(!input.empty())) - { - const auto wc2 = static_cast(input.get_character()); - const auto charcode = 0x10000u + (((static_cast(wc) & 0x3FFu) << 10u) | (wc2 & 0x3FFu)); - utf8_bytes[0] = static_cast::int_type>(0xF0u | (charcode >> 18u)); - utf8_bytes[1] = static_cast::int_type>(0x80u | ((charcode >> 12u) & 0x3Fu)); - utf8_bytes[2] = static_cast::int_type>(0x80u | ((charcode >> 6u) & 0x3Fu)); - utf8_bytes[3] = static_cast::int_type>(0x80u | (charcode & 0x3Fu)); - utf8_bytes_filled = 4; - } - else - { - utf8_bytes[0] = static_cast::int_type>(wc); - utf8_bytes_filled = 1; - } - } - } - } -}; - -// Wraps another input apdater to convert wide character types into individual bytes. -template -class wide_string_input_adapter -{ - public: - using char_type = char; - - wide_string_input_adapter(BaseInputAdapter base) - : base_adapter(base) {} - - typename std::char_traits::int_type get_character() noexcept - { - // check if buffer needs to be filled - if (utf8_bytes_index == utf8_bytes_filled) - { - fill_buffer(); - - JSON_ASSERT(utf8_bytes_filled > 0); - JSON_ASSERT(utf8_bytes_index == 0); - } - - // use buffer - JSON_ASSERT(utf8_bytes_filled > 0); - JSON_ASSERT(utf8_bytes_index < utf8_bytes_filled); - return utf8_bytes[utf8_bytes_index++]; - } - - private: - BaseInputAdapter base_adapter; - - template - void fill_buffer() - { - wide_string_input_helper::fill_buffer(base_adapter, utf8_bytes, utf8_bytes_index, utf8_bytes_filled); - } - - /// a buffer for UTF-8 bytes - std::array::int_type, 4> utf8_bytes = {{0, 0, 0, 0}}; - - /// index to the utf8_codes array for the next valid byte - std::size_t utf8_bytes_index = 0; - /// number of valid bytes in the utf8_codes array - std::size_t utf8_bytes_filled = 0; -}; - - -template -struct iterator_input_adapter_factory -{ - using iterator_type = IteratorType; - using char_type = typename std::iterator_traits::value_type; - using adapter_type = iterator_input_adapter; - - static adapter_type create(IteratorType first, IteratorType last) - { - return adapter_type(std::move(first), std::move(last)); - } -}; - -template -struct is_iterator_of_multibyte -{ - using value_type = typename std::iterator_traits::value_type; - enum - { - value = sizeof(value_type) > 1 - }; -}; - -template -struct iterator_input_adapter_factory::value>> -{ - using iterator_type = IteratorType; - using char_type = typename std::iterator_traits::value_type; - using base_adapter_type = iterator_input_adapter; - using adapter_type = wide_string_input_adapter; - - static adapter_type create(IteratorType first, IteratorType last) - { - return adapter_type(base_adapter_type(std::move(first), std::move(last))); - } -}; - -// General purpose iterator-based input -template -typename iterator_input_adapter_factory::adapter_type input_adapter(IteratorType first, IteratorType last) -{ - using factory_type = iterator_input_adapter_factory; - return factory_type::create(first, last); -} - -// Convenience shorthand from container to iterator -template -auto input_adapter(const ContainerType& container) -> decltype(input_adapter(begin(container), end(container))) -{ - // Enable ADL - using std::begin; - using std::end; - - return input_adapter(begin(container), end(container)); -} - -// Special cases with fast paths -inline file_input_adapter input_adapter(std::FILE* file) -{ - return file_input_adapter(file); -} - -inline input_stream_adapter input_adapter(std::istream& stream) -{ - return input_stream_adapter(stream); -} - -inline input_stream_adapter input_adapter(std::istream&& stream) -{ - return input_stream_adapter(stream); -} - -using contiguous_bytes_input_adapter = decltype(input_adapter(std::declval(), std::declval())); - -// Null-delimited strings, and the like. -template < typename CharT, - typename std::enable_if < - std::is_pointer::value&& - !std::is_array::value&& - std::is_integral::type>::value&& - sizeof(typename std::remove_pointer::type) == 1, - int >::type = 0 > -contiguous_bytes_input_adapter input_adapter(CharT b) -{ - auto length = std::strlen(reinterpret_cast(b)); - const auto* ptr = reinterpret_cast(b); - return input_adapter(ptr, ptr + length); -} - -template -auto input_adapter(T (&array)[N]) -> decltype(input_adapter(array, array + N)) -{ - return input_adapter(array, array + N); -} - -// This class only handles inputs of input_buffer_adapter type. -// It's required so that expressions like {ptr, len} can be implicitely casted -// to the correct adapter. -class span_input_adapter -{ - public: - template < typename CharT, - typename std::enable_if < - std::is_pointer::value&& - std::is_integral::type>::value&& - sizeof(typename std::remove_pointer::type) == 1, - int >::type = 0 > - span_input_adapter(CharT b, std::size_t l) - : ia(reinterpret_cast(b), reinterpret_cast(b) + l) {} - - template::iterator_category, std::random_access_iterator_tag>::value, - int>::type = 0> - span_input_adapter(IteratorType first, IteratorType last) - : ia(input_adapter(first, last)) {} - - contiguous_bytes_input_adapter&& get() - { - return std::move(ia); - } - - private: - contiguous_bytes_input_adapter ia; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/json_sax.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/json_sax.hpp deleted file mode 100644 index 223acd60ebd..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/json_sax.hpp +++ /dev/null @@ -1,695 +0,0 @@ -#pragma once - -#include -#include // string -#include // move -#include // vector - -#include -#include - -namespace nlohmann -{ - -/*! -@brief SAX interface - -This class describes the SAX interface used by @ref nlohmann::json::sax_parse. -Each function is called in different situations while the input is parsed. The -boolean return value informs the parser whether to continue processing the -input. -*/ -template -struct json_sax -{ - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - - /*! - @brief a null value was read - @return whether parsing should proceed - */ - virtual bool null() = 0; - - /*! - @brief a boolean value was read - @param[in] val boolean value - @return whether parsing should proceed - */ - virtual bool boolean(bool val) = 0; - - /*! - @brief an integer number was read - @param[in] val integer value - @return whether parsing should proceed - */ - virtual bool number_integer(number_integer_t val) = 0; - - /*! - @brief an unsigned integer number was read - @param[in] val unsigned integer value - @return whether parsing should proceed - */ - virtual bool number_unsigned(number_unsigned_t val) = 0; - - /*! - @brief an floating-point number was read - @param[in] val floating-point value - @param[in] s raw token value - @return whether parsing should proceed - */ - virtual bool number_float(number_float_t val, const string_t& s) = 0; - - /*! - @brief a string was read - @param[in] val string value - @return whether parsing should proceed - @note It is safe to move the passed string. - */ - virtual bool string(string_t& val) = 0; - - /*! - @brief a binary string was read - @param[in] val binary value - @return whether parsing should proceed - @note It is safe to move the passed binary. - */ - virtual bool binary(binary_t& val) = 0; - - /*! - @brief the beginning of an object was read - @param[in] elements number of object elements or -1 if unknown - @return whether parsing should proceed - @note binary formats may report the number of elements - */ - virtual bool start_object(std::size_t elements) = 0; - - /*! - @brief an object key was read - @param[in] val object key - @return whether parsing should proceed - @note It is safe to move the passed string. - */ - virtual bool key(string_t& val) = 0; - - /*! - @brief the end of an object was read - @return whether parsing should proceed - */ - virtual bool end_object() = 0; - - /*! - @brief the beginning of an array was read - @param[in] elements number of array elements or -1 if unknown - @return whether parsing should proceed - @note binary formats may report the number of elements - */ - virtual bool start_array(std::size_t elements) = 0; - - /*! - @brief the end of an array was read - @return whether parsing should proceed - */ - virtual bool end_array() = 0; - - /*! - @brief a parse error occurred - @param[in] position the position in the input where the error occurs - @param[in] last_token the last read token - @param[in] ex an exception object describing the error - @return whether parsing should proceed (must return false) - */ - virtual bool parse_error(std::size_t position, - const std::string& last_token, - const detail::exception& ex) = 0; - - virtual ~json_sax() = default; -}; - - -namespace detail -{ -/*! -@brief SAX implementation to create a JSON value from SAX events - -This class implements the @ref json_sax interface and processes the SAX events -to create a JSON value which makes it basically a DOM parser. The structure or -hierarchy of the JSON value is managed by the stack `ref_stack` which contains -a pointer to the respective array or object for each recursion depth. - -After successful parsing, the value that is passed by reference to the -constructor contains the parsed value. - -@tparam BasicJsonType the JSON type -*/ -template -class json_sax_dom_parser -{ - public: - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - - /*! - @param[in, out] r reference to a JSON value that is manipulated while - parsing - @param[in] allow_exceptions_ whether parse errors yield exceptions - */ - explicit json_sax_dom_parser(BasicJsonType& r, const bool allow_exceptions_ = true) - : root(r), allow_exceptions(allow_exceptions_) - {} - - // make class move-only - json_sax_dom_parser(const json_sax_dom_parser&) = delete; - json_sax_dom_parser(json_sax_dom_parser&&) = default; - json_sax_dom_parser& operator=(const json_sax_dom_parser&) = delete; - json_sax_dom_parser& operator=(json_sax_dom_parser&&) = default; - ~json_sax_dom_parser() = default; - - bool null() - { - handle_value(nullptr); - return true; - } - - bool boolean(bool val) - { - handle_value(val); - return true; - } - - bool number_integer(number_integer_t val) - { - handle_value(val); - return true; - } - - bool number_unsigned(number_unsigned_t val) - { - handle_value(val); - return true; - } - - bool number_float(number_float_t val, const string_t& /*unused*/) - { - handle_value(val); - return true; - } - - bool string(string_t& val) - { - handle_value(val); - return true; - } - - bool binary(binary_t& val) - { - handle_value(std::move(val)); - return true; - } - - bool start_object(std::size_t len) - { - ref_stack.push_back(handle_value(BasicJsonType::value_t::object)); - - if (JSON_HEDLEY_UNLIKELY(len != std::size_t(-1) && len > ref_stack.back()->max_size())) - { - JSON_THROW(out_of_range::create(408, - "excessive object size: " + std::to_string(len))); - } - - return true; - } - - bool key(string_t& val) - { - // add null at given key and store the reference for later - object_element = &(ref_stack.back()->m_value.object->operator[](val)); - return true; - } - - bool end_object() - { - ref_stack.pop_back(); - return true; - } - - bool start_array(std::size_t len) - { - ref_stack.push_back(handle_value(BasicJsonType::value_t::array)); - - if (JSON_HEDLEY_UNLIKELY(len != std::size_t(-1) && len > ref_stack.back()->max_size())) - { - JSON_THROW(out_of_range::create(408, - "excessive array size: " + std::to_string(len))); - } - - return true; - } - - bool end_array() - { - ref_stack.pop_back(); - return true; - } - - template - bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/, - const Exception& ex) - { - errored = true; - static_cast(ex); - if (allow_exceptions) - { - JSON_THROW(ex); - } - return false; - } - - constexpr bool is_errored() const - { - return errored; - } - - private: - /*! - @invariant If the ref stack is empty, then the passed value will be the new - root. - @invariant If the ref stack contains a value, then it is an array or an - object to which we can add elements - */ - template - JSON_HEDLEY_RETURNS_NON_NULL - BasicJsonType* handle_value(Value&& v) - { - if (ref_stack.empty()) - { - root = BasicJsonType(std::forward(v)); - return &root; - } - - JSON_ASSERT(ref_stack.back()->is_array() || ref_stack.back()->is_object()); - - if (ref_stack.back()->is_array()) - { - ref_stack.back()->m_value.array->emplace_back(std::forward(v)); - return &(ref_stack.back()->m_value.array->back()); - } - - JSON_ASSERT(ref_stack.back()->is_object()); - JSON_ASSERT(object_element); - *object_element = BasicJsonType(std::forward(v)); - return object_element; - } - - /// the parsed JSON value - BasicJsonType& root; - /// stack to model hierarchy of values - std::vector ref_stack {}; - /// helper to hold the reference for the next object element - BasicJsonType* object_element = nullptr; - /// whether a syntax error occurred - bool errored = false; - /// whether to throw exceptions in case of errors - const bool allow_exceptions = true; -}; - -template -class json_sax_dom_callback_parser -{ - public: - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - using parser_callback_t = typename BasicJsonType::parser_callback_t; - using parse_event_t = typename BasicJsonType::parse_event_t; - - json_sax_dom_callback_parser(BasicJsonType& r, - const parser_callback_t cb, - const bool allow_exceptions_ = true) - : root(r), callback(cb), allow_exceptions(allow_exceptions_) - { - keep_stack.push_back(true); - } - - // make class move-only - json_sax_dom_callback_parser(const json_sax_dom_callback_parser&) = delete; - json_sax_dom_callback_parser(json_sax_dom_callback_parser&&) = default; - json_sax_dom_callback_parser& operator=(const json_sax_dom_callback_parser&) = delete; - json_sax_dom_callback_parser& operator=(json_sax_dom_callback_parser&&) = default; - ~json_sax_dom_callback_parser() = default; - - bool null() - { - handle_value(nullptr); - return true; - } - - bool boolean(bool val) - { - handle_value(val); - return true; - } - - bool number_integer(number_integer_t val) - { - handle_value(val); - return true; - } - - bool number_unsigned(number_unsigned_t val) - { - handle_value(val); - return true; - } - - bool number_float(number_float_t val, const string_t& /*unused*/) - { - handle_value(val); - return true; - } - - bool string(string_t& val) - { - handle_value(val); - return true; - } - - bool binary(binary_t& val) - { - handle_value(std::move(val)); - return true; - } - - bool start_object(std::size_t len) - { - // check callback for object start - const bool keep = callback(static_cast(ref_stack.size()), parse_event_t::object_start, discarded); - keep_stack.push_back(keep); - - auto val = handle_value(BasicJsonType::value_t::object, true); - ref_stack.push_back(val.second); - - // check object limit - if (ref_stack.back() && JSON_HEDLEY_UNLIKELY(len != std::size_t(-1) && len > ref_stack.back()->max_size())) - { - JSON_THROW(out_of_range::create(408, "excessive object size: " + std::to_string(len))); - } - - return true; - } - - bool key(string_t& val) - { - BasicJsonType k = BasicJsonType(val); - - // check callback for key - const bool keep = callback(static_cast(ref_stack.size()), parse_event_t::key, k); - key_keep_stack.push_back(keep); - - // add discarded value at given key and store the reference for later - if (keep && ref_stack.back()) - { - object_element = &(ref_stack.back()->m_value.object->operator[](val) = discarded); - } - - return true; - } - - bool end_object() - { - if (ref_stack.back() && !callback(static_cast(ref_stack.size()) - 1, parse_event_t::object_end, *ref_stack.back())) - { - // discard object - *ref_stack.back() = discarded; - } - - JSON_ASSERT(!ref_stack.empty()); - JSON_ASSERT(!keep_stack.empty()); - ref_stack.pop_back(); - keep_stack.pop_back(); - - if (!ref_stack.empty() && ref_stack.back() && ref_stack.back()->is_structured()) - { - // remove discarded value - for (auto it = ref_stack.back()->begin(); it != ref_stack.back()->end(); ++it) - { - if (it->is_discarded()) - { - ref_stack.back()->erase(it); - break; - } - } - } - - return true; - } - - bool start_array(std::size_t len) - { - const bool keep = callback(static_cast(ref_stack.size()), parse_event_t::array_start, discarded); - keep_stack.push_back(keep); - - auto val = handle_value(BasicJsonType::value_t::array, true); - ref_stack.push_back(val.second); - - // check array limit - if (ref_stack.back() && JSON_HEDLEY_UNLIKELY(len != std::size_t(-1) && len > ref_stack.back()->max_size())) - { - JSON_THROW(out_of_range::create(408, "excessive array size: " + std::to_string(len))); - } - - return true; - } - - bool end_array() - { - bool keep = true; - - if (ref_stack.back()) - { - keep = callback(static_cast(ref_stack.size()) - 1, parse_event_t::array_end, *ref_stack.back()); - if (!keep) - { - // discard array - *ref_stack.back() = discarded; - } - } - - JSON_ASSERT(!ref_stack.empty()); - JSON_ASSERT(!keep_stack.empty()); - ref_stack.pop_back(); - keep_stack.pop_back(); - - // remove discarded value - if (!keep && !ref_stack.empty() && ref_stack.back()->is_array()) - { - ref_stack.back()->m_value.array->pop_back(); - } - - return true; - } - - template - bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/, - const Exception& ex) - { - errored = true; - static_cast(ex); - if (allow_exceptions) - { - JSON_THROW(ex); - } - return false; - } - - constexpr bool is_errored() const - { - return errored; - } - - private: - /*! - @param[in] v value to add to the JSON value we build during parsing - @param[in] skip_callback whether we should skip calling the callback - function; this is required after start_array() and - start_object() SAX events, because otherwise we would call the - callback function with an empty array or object, respectively. - - @invariant If the ref stack is empty, then the passed value will be the new - root. - @invariant If the ref stack contains a value, then it is an array or an - object to which we can add elements - - @return pair of boolean (whether value should be kept) and pointer (to the - passed value in the ref_stack hierarchy; nullptr if not kept) - */ - template - std::pair handle_value(Value&& v, const bool skip_callback = false) - { - JSON_ASSERT(!keep_stack.empty()); - - // do not handle this value if we know it would be added to a discarded - // container - if (!keep_stack.back()) - { - return {false, nullptr}; - } - - // create value - auto value = BasicJsonType(std::forward(v)); - - // check callback - const bool keep = skip_callback || callback(static_cast(ref_stack.size()), parse_event_t::value, value); - - // do not handle this value if we just learnt it shall be discarded - if (!keep) - { - return {false, nullptr}; - } - - if (ref_stack.empty()) - { - root = std::move(value); - return {true, &root}; - } - - // skip this value if we already decided to skip the parent - // (https://github.com/nlohmann/json/issues/971#issuecomment-413678360) - if (!ref_stack.back()) - { - return {false, nullptr}; - } - - // we now only expect arrays and objects - JSON_ASSERT(ref_stack.back()->is_array() || ref_stack.back()->is_object()); - - // array - if (ref_stack.back()->is_array()) - { - ref_stack.back()->m_value.array->push_back(std::move(value)); - return {true, &(ref_stack.back()->m_value.array->back())}; - } - - // object - JSON_ASSERT(ref_stack.back()->is_object()); - // check if we should store an element for the current key - JSON_ASSERT(!key_keep_stack.empty()); - const bool store_element = key_keep_stack.back(); - key_keep_stack.pop_back(); - - if (!store_element) - { - return {false, nullptr}; - } - - JSON_ASSERT(object_element); - *object_element = std::move(value); - return {true, object_element}; - } - - /// the parsed JSON value - BasicJsonType& root; - /// stack to model hierarchy of values - std::vector ref_stack {}; - /// stack to manage which values to keep - std::vector keep_stack {}; - /// stack to manage which object keys to keep - std::vector key_keep_stack {}; - /// helper to hold the reference for the next object element - BasicJsonType* object_element = nullptr; - /// whether a syntax error occurred - bool errored = false; - /// callback function - const parser_callback_t callback = nullptr; - /// whether to throw exceptions in case of errors - const bool allow_exceptions = true; - /// a discarded value for the callback - BasicJsonType discarded = BasicJsonType::value_t::discarded; -}; - -template -class json_sax_acceptor -{ - public: - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - - bool null() - { - return true; - } - - bool boolean(bool /*unused*/) - { - return true; - } - - bool number_integer(number_integer_t /*unused*/) - { - return true; - } - - bool number_unsigned(number_unsigned_t /*unused*/) - { - return true; - } - - bool number_float(number_float_t /*unused*/, const string_t& /*unused*/) - { - return true; - } - - bool string(string_t& /*unused*/) - { - return true; - } - - bool binary(binary_t& /*unused*/) - { - return true; - } - - bool start_object(std::size_t /*unused*/ = std::size_t(-1)) - { - return true; - } - - bool key(string_t& /*unused*/) - { - return true; - } - - bool end_object() - { - return true; - } - - bool start_array(std::size_t /*unused*/ = std::size_t(-1)) - { - return true; - } - - bool end_array() - { - return true; - } - - bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/, const detail::exception& /*unused*/) - { - return false; - } -}; -} // namespace detail - -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/lexer.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/lexer.hpp deleted file mode 100644 index 0a9601352e4..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/lexer.hpp +++ /dev/null @@ -1,1623 +0,0 @@ -#pragma once - -#include // array -#include // localeconv -#include // size_t -#include // snprintf -#include // strtof, strtod, strtold, strtoll, strtoull -#include // initializer_list -#include // char_traits, string -#include // move -#include // vector - -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -/////////// -// lexer // -/////////// - -template -class lexer_base -{ - public: - /// token types for the parser - enum class token_type - { - uninitialized, ///< indicating the scanner is uninitialized - literal_true, ///< the `true` literal - literal_false, ///< the `false` literal - literal_null, ///< the `null` literal - value_string, ///< a string -- use get_string() for actual value - value_unsigned, ///< an unsigned integer -- use get_number_unsigned() for actual value - value_integer, ///< a signed integer -- use get_number_integer() for actual value - value_float, ///< an floating point number -- use get_number_float() for actual value - begin_array, ///< the character for array begin `[` - begin_object, ///< the character for object begin `{` - end_array, ///< the character for array end `]` - end_object, ///< the character for object end `}` - name_separator, ///< the name separator `:` - value_separator, ///< the value separator `,` - parse_error, ///< indicating a parse error - end_of_input, ///< indicating the end of the input buffer - literal_or_value ///< a literal or the begin of a value (only for diagnostics) - }; - - /// return name of values of type token_type (only used for errors) - JSON_HEDLEY_RETURNS_NON_NULL - JSON_HEDLEY_CONST - static const char* token_type_name(const token_type t) noexcept - { - switch (t) - { - case token_type::uninitialized: - return ""; - case token_type::literal_true: - return "true literal"; - case token_type::literal_false: - return "false literal"; - case token_type::literal_null: - return "null literal"; - case token_type::value_string: - return "string literal"; - case token_type::value_unsigned: - case token_type::value_integer: - case token_type::value_float: - return "number literal"; - case token_type::begin_array: - return "'['"; - case token_type::begin_object: - return "'{'"; - case token_type::end_array: - return "']'"; - case token_type::end_object: - return "'}'"; - case token_type::name_separator: - return "':'"; - case token_type::value_separator: - return "','"; - case token_type::parse_error: - return ""; - case token_type::end_of_input: - return "end of input"; - case token_type::literal_or_value: - return "'[', '{', or a literal"; - // LCOV_EXCL_START - default: // catch non-enum values - return "unknown token"; - // LCOV_EXCL_STOP - } - } -}; -/*! -@brief lexical analysis - -This class organizes the lexical analysis during JSON deserialization. -*/ -template -class lexer : public lexer_base -{ - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using char_type = typename InputAdapterType::char_type; - using char_int_type = typename std::char_traits::int_type; - - public: - using token_type = typename lexer_base::token_type; - - explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false) - : ia(std::move(adapter)) - , ignore_comments(ignore_comments_) - , decimal_point_char(static_cast(get_decimal_point())) - {} - - // delete because of pointer members - lexer(const lexer&) = delete; - lexer(lexer&&) = default; - lexer& operator=(lexer&) = delete; - lexer& operator=(lexer&&) = default; - ~lexer() = default; - - private: - ///////////////////// - // locales - ///////////////////// - - /// return the locale-dependent decimal point - JSON_HEDLEY_PURE - static char get_decimal_point() noexcept - { - const auto* loc = localeconv(); - JSON_ASSERT(loc != nullptr); - return (loc->decimal_point == nullptr) ? '.' : *(loc->decimal_point); - } - - ///////////////////// - // scan functions - ///////////////////// - - /*! - @brief get codepoint from 4 hex characters following `\u` - - For input "\u c1 c2 c3 c4" the codepoint is: - (c1 * 0x1000) + (c2 * 0x0100) + (c3 * 0x0010) + c4 - = (c1 << 12) + (c2 << 8) + (c3 << 4) + (c4 << 0) - - Furthermore, the possible characters '0'..'9', 'A'..'F', and 'a'..'f' - must be converted to the integers 0x0..0x9, 0xA..0xF, 0xA..0xF, resp. The - conversion is done by subtracting the offset (0x30, 0x37, and 0x57) - between the ASCII value of the character and the desired integer value. - - @return codepoint (0x0000..0xFFFF) or -1 in case of an error (e.g. EOF or - non-hex character) - */ - int get_codepoint() - { - // this function only makes sense after reading `\u` - JSON_ASSERT(current == 'u'); - int codepoint = 0; - - const auto factors = { 12u, 8u, 4u, 0u }; - for (const auto factor : factors) - { - get(); - - if (current >= '0' && current <= '9') - { - codepoint += static_cast((static_cast(current) - 0x30u) << factor); - } - else if (current >= 'A' && current <= 'F') - { - codepoint += static_cast((static_cast(current) - 0x37u) << factor); - } - else if (current >= 'a' && current <= 'f') - { - codepoint += static_cast((static_cast(current) - 0x57u) << factor); - } - else - { - return -1; - } - } - - JSON_ASSERT(0x0000 <= codepoint && codepoint <= 0xFFFF); - return codepoint; - } - - /*! - @brief check if the next byte(s) are inside a given range - - Adds the current byte and, for each passed range, reads a new byte and - checks if it is inside the range. If a violation was detected, set up an - error message and return false. Otherwise, return true. - - @param[in] ranges list of integers; interpreted as list of pairs of - inclusive lower and upper bound, respectively - - @pre The passed list @a ranges must have 2, 4, or 6 elements; that is, - 1, 2, or 3 pairs. This precondition is enforced by an assertion. - - @return true if and only if no range violation was detected - */ - bool next_byte_in_range(std::initializer_list ranges) - { - JSON_ASSERT(ranges.size() == 2 || ranges.size() == 4 || ranges.size() == 6); - add(current); - - for (auto range = ranges.begin(); range != ranges.end(); ++range) - { - get(); - if (JSON_HEDLEY_LIKELY(*range <= current && current <= *(++range))) - { - add(current); - } - else - { - error_message = "invalid string: ill-formed UTF-8 byte"; - return false; - } - } - - return true; - } - - /*! - @brief scan a string literal - - This function scans a string according to Sect. 7 of RFC 7159. While - scanning, bytes are escaped and copied into buffer token_buffer. Then the - function returns successfully, token_buffer is *not* null-terminated (as it - may contain \0 bytes), and token_buffer.size() is the number of bytes in the - string. - - @return token_type::value_string if string could be successfully scanned, - token_type::parse_error otherwise - - @note In case of errors, variable error_message contains a textual - description. - */ - token_type scan_string() - { - // reset token_buffer (ignore opening quote) - reset(); - - // we entered the function by reading an open quote - JSON_ASSERT(current == '\"'); - - while (true) - { - // get next character - switch (get()) - { - // end of file while parsing string - case std::char_traits::eof(): - { - error_message = "invalid string: missing closing quote"; - return token_type::parse_error; - } - - // closing quote - case '\"': - { - return token_type::value_string; - } - - // escapes - case '\\': - { - switch (get()) - { - // quotation mark - case '\"': - add('\"'); - break; - // reverse solidus - case '\\': - add('\\'); - break; - // solidus - case '/': - add('/'); - break; - // backspace - case 'b': - add('\b'); - break; - // form feed - case 'f': - add('\f'); - break; - // line feed - case 'n': - add('\n'); - break; - // carriage return - case 'r': - add('\r'); - break; - // tab - case 't': - add('\t'); - break; - - // unicode escapes - case 'u': - { - const int codepoint1 = get_codepoint(); - int codepoint = codepoint1; // start with codepoint1 - - if (JSON_HEDLEY_UNLIKELY(codepoint1 == -1)) - { - error_message = "invalid string: '\\u' must be followed by 4 hex digits"; - return token_type::parse_error; - } - - // check if code point is a high surrogate - if (0xD800 <= codepoint1 && codepoint1 <= 0xDBFF) - { - // expect next \uxxxx entry - if (JSON_HEDLEY_LIKELY(get() == '\\' && get() == 'u')) - { - const int codepoint2 = get_codepoint(); - - if (JSON_HEDLEY_UNLIKELY(codepoint2 == -1)) - { - error_message = "invalid string: '\\u' must be followed by 4 hex digits"; - return token_type::parse_error; - } - - // check if codepoint2 is a low surrogate - if (JSON_HEDLEY_LIKELY(0xDC00 <= codepoint2 && codepoint2 <= 0xDFFF)) - { - // overwrite codepoint - codepoint = static_cast( - // high surrogate occupies the most significant 22 bits - (static_cast(codepoint1) << 10u) - // low surrogate occupies the least significant 15 bits - + static_cast(codepoint2) - // there is still the 0xD800, 0xDC00 and 0x10000 noise - // in the result so we have to subtract with: - // (0xD800 << 10) + DC00 - 0x10000 = 0x35FDC00 - - 0x35FDC00u); - } - else - { - error_message = "invalid string: surrogate U+D800..U+DBFF must be followed by U+DC00..U+DFFF"; - return token_type::parse_error; - } - } - else - { - error_message = "invalid string: surrogate U+D800..U+DBFF must be followed by U+DC00..U+DFFF"; - return token_type::parse_error; - } - } - else - { - if (JSON_HEDLEY_UNLIKELY(0xDC00 <= codepoint1 && codepoint1 <= 0xDFFF)) - { - error_message = "invalid string: surrogate U+DC00..U+DFFF must follow U+D800..U+DBFF"; - return token_type::parse_error; - } - } - - // result of the above calculation yields a proper codepoint - JSON_ASSERT(0x00 <= codepoint && codepoint <= 0x10FFFF); - - // translate codepoint into bytes - if (codepoint < 0x80) - { - // 1-byte characters: 0xxxxxxx (ASCII) - add(static_cast(codepoint)); - } - else if (codepoint <= 0x7FF) - { - // 2-byte characters: 110xxxxx 10xxxxxx - add(static_cast(0xC0u | (static_cast(codepoint) >> 6u))); - add(static_cast(0x80u | (static_cast(codepoint) & 0x3Fu))); - } - else if (codepoint <= 0xFFFF) - { - // 3-byte characters: 1110xxxx 10xxxxxx 10xxxxxx - add(static_cast(0xE0u | (static_cast(codepoint) >> 12u))); - add(static_cast(0x80u | ((static_cast(codepoint) >> 6u) & 0x3Fu))); - add(static_cast(0x80u | (static_cast(codepoint) & 0x3Fu))); - } - else - { - // 4-byte characters: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx - add(static_cast(0xF0u | (static_cast(codepoint) >> 18u))); - add(static_cast(0x80u | ((static_cast(codepoint) >> 12u) & 0x3Fu))); - add(static_cast(0x80u | ((static_cast(codepoint) >> 6u) & 0x3Fu))); - add(static_cast(0x80u | (static_cast(codepoint) & 0x3Fu))); - } - - break; - } - - // other characters after escape - default: - error_message = "invalid string: forbidden character after backslash"; - return token_type::parse_error; - } - - break; - } - - // invalid control characters - case 0x00: - { - error_message = "invalid string: control character U+0000 (NUL) must be escaped to \\u0000"; - return token_type::parse_error; - } - - case 0x01: - { - error_message = "invalid string: control character U+0001 (SOH) must be escaped to \\u0001"; - return token_type::parse_error; - } - - case 0x02: - { - error_message = "invalid string: control character U+0002 (STX) must be escaped to \\u0002"; - return token_type::parse_error; - } - - case 0x03: - { - error_message = "invalid string: control character U+0003 (ETX) must be escaped to \\u0003"; - return token_type::parse_error; - } - - case 0x04: - { - error_message = "invalid string: control character U+0004 (EOT) must be escaped to \\u0004"; - return token_type::parse_error; - } - - case 0x05: - { - error_message = "invalid string: control character U+0005 (ENQ) must be escaped to \\u0005"; - return token_type::parse_error; - } - - case 0x06: - { - error_message = "invalid string: control character U+0006 (ACK) must be escaped to \\u0006"; - return token_type::parse_error; - } - - case 0x07: - { - error_message = "invalid string: control character U+0007 (BEL) must be escaped to \\u0007"; - return token_type::parse_error; - } - - case 0x08: - { - error_message = "invalid string: control character U+0008 (BS) must be escaped to \\u0008 or \\b"; - return token_type::parse_error; - } - - case 0x09: - { - error_message = "invalid string: control character U+0009 (HT) must be escaped to \\u0009 or \\t"; - return token_type::parse_error; - } - - case 0x0A: - { - error_message = "invalid string: control character U+000A (LF) must be escaped to \\u000A or \\n"; - return token_type::parse_error; - } - - case 0x0B: - { - error_message = "invalid string: control character U+000B (VT) must be escaped to \\u000B"; - return token_type::parse_error; - } - - case 0x0C: - { - error_message = "invalid string: control character U+000C (FF) must be escaped to \\u000C or \\f"; - return token_type::parse_error; - } - - case 0x0D: - { - error_message = "invalid string: control character U+000D (CR) must be escaped to \\u000D or \\r"; - return token_type::parse_error; - } - - case 0x0E: - { - error_message = "invalid string: control character U+000E (SO) must be escaped to \\u000E"; - return token_type::parse_error; - } - - case 0x0F: - { - error_message = "invalid string: control character U+000F (SI) must be escaped to \\u000F"; - return token_type::parse_error; - } - - case 0x10: - { - error_message = "invalid string: control character U+0010 (DLE) must be escaped to \\u0010"; - return token_type::parse_error; - } - - case 0x11: - { - error_message = "invalid string: control character U+0011 (DC1) must be escaped to \\u0011"; - return token_type::parse_error; - } - - case 0x12: - { - error_message = "invalid string: control character U+0012 (DC2) must be escaped to \\u0012"; - return token_type::parse_error; - } - - case 0x13: - { - error_message = "invalid string: control character U+0013 (DC3) must be escaped to \\u0013"; - return token_type::parse_error; - } - - case 0x14: - { - error_message = "invalid string: control character U+0014 (DC4) must be escaped to \\u0014"; - return token_type::parse_error; - } - - case 0x15: - { - error_message = "invalid string: control character U+0015 (NAK) must be escaped to \\u0015"; - return token_type::parse_error; - } - - case 0x16: - { - error_message = "invalid string: control character U+0016 (SYN) must be escaped to \\u0016"; - return token_type::parse_error; - } - - case 0x17: - { - error_message = "invalid string: control character U+0017 (ETB) must be escaped to \\u0017"; - return token_type::parse_error; - } - - case 0x18: - { - error_message = "invalid string: control character U+0018 (CAN) must be escaped to \\u0018"; - return token_type::parse_error; - } - - case 0x19: - { - error_message = "invalid string: control character U+0019 (EM) must be escaped to \\u0019"; - return token_type::parse_error; - } - - case 0x1A: - { - error_message = "invalid string: control character U+001A (SUB) must be escaped to \\u001A"; - return token_type::parse_error; - } - - case 0x1B: - { - error_message = "invalid string: control character U+001B (ESC) must be escaped to \\u001B"; - return token_type::parse_error; - } - - case 0x1C: - { - error_message = "invalid string: control character U+001C (FS) must be escaped to \\u001C"; - return token_type::parse_error; - } - - case 0x1D: - { - error_message = "invalid string: control character U+001D (GS) must be escaped to \\u001D"; - return token_type::parse_error; - } - - case 0x1E: - { - error_message = "invalid string: control character U+001E (RS) must be escaped to \\u001E"; - return token_type::parse_error; - } - - case 0x1F: - { - error_message = "invalid string: control character U+001F (US) must be escaped to \\u001F"; - return token_type::parse_error; - } - - // U+0020..U+007F (except U+0022 (quote) and U+005C (backspace)) - case 0x20: - case 0x21: - case 0x23: - case 0x24: - case 0x25: - case 0x26: - case 0x27: - case 0x28: - case 0x29: - case 0x2A: - case 0x2B: - case 0x2C: - case 0x2D: - case 0x2E: - case 0x2F: - case 0x30: - case 0x31: - case 0x32: - case 0x33: - case 0x34: - case 0x35: - case 0x36: - case 0x37: - case 0x38: - case 0x39: - case 0x3A: - case 0x3B: - case 0x3C: - case 0x3D: - case 0x3E: - case 0x3F: - case 0x40: - case 0x41: - case 0x42: - case 0x43: - case 0x44: - case 0x45: - case 0x46: - case 0x47: - case 0x48: - case 0x49: - case 0x4A: - case 0x4B: - case 0x4C: - case 0x4D: - case 0x4E: - case 0x4F: - case 0x50: - case 0x51: - case 0x52: - case 0x53: - case 0x54: - case 0x55: - case 0x56: - case 0x57: - case 0x58: - case 0x59: - case 0x5A: - case 0x5B: - case 0x5D: - case 0x5E: - case 0x5F: - case 0x60: - case 0x61: - case 0x62: - case 0x63: - case 0x64: - case 0x65: - case 0x66: - case 0x67: - case 0x68: - case 0x69: - case 0x6A: - case 0x6B: - case 0x6C: - case 0x6D: - case 0x6E: - case 0x6F: - case 0x70: - case 0x71: - case 0x72: - case 0x73: - case 0x74: - case 0x75: - case 0x76: - case 0x77: - case 0x78: - case 0x79: - case 0x7A: - case 0x7B: - case 0x7C: - case 0x7D: - case 0x7E: - case 0x7F: - { - add(current); - break; - } - - // U+0080..U+07FF: bytes C2..DF 80..BF - case 0xC2: - case 0xC3: - case 0xC4: - case 0xC5: - case 0xC6: - case 0xC7: - case 0xC8: - case 0xC9: - case 0xCA: - case 0xCB: - case 0xCC: - case 0xCD: - case 0xCE: - case 0xCF: - case 0xD0: - case 0xD1: - case 0xD2: - case 0xD3: - case 0xD4: - case 0xD5: - case 0xD6: - case 0xD7: - case 0xD8: - case 0xD9: - case 0xDA: - case 0xDB: - case 0xDC: - case 0xDD: - case 0xDE: - case 0xDF: - { - if (JSON_HEDLEY_UNLIKELY(!next_byte_in_range({0x80, 0xBF}))) - { - return token_type::parse_error; - } - break; - } - - // U+0800..U+0FFF: bytes E0 A0..BF 80..BF - case 0xE0: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0xA0, 0xBF, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // U+1000..U+CFFF: bytes E1..EC 80..BF 80..BF - // U+E000..U+FFFF: bytes EE..EF 80..BF 80..BF - case 0xE1: - case 0xE2: - case 0xE3: - case 0xE4: - case 0xE5: - case 0xE6: - case 0xE7: - case 0xE8: - case 0xE9: - case 0xEA: - case 0xEB: - case 0xEC: - case 0xEE: - case 0xEF: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0xBF, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // U+D000..U+D7FF: bytes ED 80..9F 80..BF - case 0xED: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0x9F, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // U+10000..U+3FFFF F0 90..BF 80..BF 80..BF - case 0xF0: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x90, 0xBF, 0x80, 0xBF, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF - case 0xF1: - case 0xF2: - case 0xF3: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0xBF, 0x80, 0xBF, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // U+100000..U+10FFFF F4 80..8F 80..BF 80..BF - case 0xF4: - { - if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0x8F, 0x80, 0xBF, 0x80, 0xBF})))) - { - return token_type::parse_error; - } - break; - } - - // remaining bytes (80..C1 and F5..FF) are ill-formed - default: - { - error_message = "invalid string: ill-formed UTF-8 byte"; - return token_type::parse_error; - } - } - } - } - - /*! - * @brief scan a comment - * @return whether comment could be scanned successfully - */ - bool scan_comment() - { - switch (get()) - { - // single-line comments skip input until a newline or EOF is read - case '/': - { - while (true) - { - switch (get()) - { - case '\n': - case '\r': - case std::char_traits::eof(): - case '\0': - return true; - - default: - break; - } - } - } - - // multi-line comments skip input until */ is read - case '*': - { - while (true) - { - switch (get()) - { - case std::char_traits::eof(): - case '\0': - { - error_message = "invalid comment; missing closing '*/'"; - return false; - } - - case '*': - { - switch (get()) - { - case '/': - return true; - - default: - { - unget(); - continue; - } - } - } - - default: - continue; - } - } - } - - // unexpected character after reading '/' - default: - { - error_message = "invalid comment; expecting '/' or '*' after '/'"; - return false; - } - } - } - - JSON_HEDLEY_NON_NULL(2) - static void strtof(float& f, const char* str, char** endptr) noexcept - { - f = std::strtof(str, endptr); - } - - JSON_HEDLEY_NON_NULL(2) - static void strtof(double& f, const char* str, char** endptr) noexcept - { - f = std::strtod(str, endptr); - } - - JSON_HEDLEY_NON_NULL(2) - static void strtof(long double& f, const char* str, char** endptr) noexcept - { - f = std::strtold(str, endptr); - } - - /*! - @brief scan a number literal - - This function scans a string according to Sect. 6 of RFC 7159. - - The function is realized with a deterministic finite state machine derived - from the grammar described in RFC 7159. Starting in state "init", the - input is read and used to determined the next state. Only state "done" - accepts the number. State "error" is a trap state to model errors. In the - table below, "anything" means any character but the ones listed before. - - state | 0 | 1-9 | e E | + | - | . | anything - ---------|----------|----------|----------|---------|---------|----------|----------- - init | zero | any1 | [error] | [error] | minus | [error] | [error] - minus | zero | any1 | [error] | [error] | [error] | [error] | [error] - zero | done | done | exponent | done | done | decimal1 | done - any1 | any1 | any1 | exponent | done | done | decimal1 | done - decimal1 | decimal2 | decimal2 | [error] | [error] | [error] | [error] | [error] - decimal2 | decimal2 | decimal2 | exponent | done | done | done | done - exponent | any2 | any2 | [error] | sign | sign | [error] | [error] - sign | any2 | any2 | [error] | [error] | [error] | [error] | [error] - any2 | any2 | any2 | done | done | done | done | done - - The state machine is realized with one label per state (prefixed with - "scan_number_") and `goto` statements between them. The state machine - contains cycles, but any cycle can be left when EOF is read. Therefore, - the function is guaranteed to terminate. - - During scanning, the read bytes are stored in token_buffer. This string is - then converted to a signed integer, an unsigned integer, or a - floating-point number. - - @return token_type::value_unsigned, token_type::value_integer, or - token_type::value_float if number could be successfully scanned, - token_type::parse_error otherwise - - @note The scanner is independent of the current locale. Internally, the - locale's decimal point is used instead of `.` to work with the - locale-dependent converters. - */ - token_type scan_number() // lgtm [cpp/use-of-goto] - { - // reset token_buffer to store the number's bytes - reset(); - - // the type of the parsed number; initially set to unsigned; will be - // changed if minus sign, decimal point or exponent is read - token_type number_type = token_type::value_unsigned; - - // state (init): we just found out we need to scan a number - switch (current) - { - case '-': - { - add(current); - goto scan_number_minus; - } - - case '0': - { - add(current); - goto scan_number_zero; - } - - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any1; - } - - // all other characters are rejected outside scan_number() - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - -scan_number_minus: - // state: we just parsed a leading minus sign - number_type = token_type::value_integer; - switch (get()) - { - case '0': - { - add(current); - goto scan_number_zero; - } - - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any1; - } - - default: - { - error_message = "invalid number; expected digit after '-'"; - return token_type::parse_error; - } - } - -scan_number_zero: - // state: we just parse a zero (maybe with a leading minus sign) - switch (get()) - { - case '.': - { - add(decimal_point_char); - goto scan_number_decimal1; - } - - case 'e': - case 'E': - { - add(current); - goto scan_number_exponent; - } - - default: - goto scan_number_done; - } - -scan_number_any1: - // state: we just parsed a number 0-9 (maybe with a leading minus sign) - switch (get()) - { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any1; - } - - case '.': - { - add(decimal_point_char); - goto scan_number_decimal1; - } - - case 'e': - case 'E': - { - add(current); - goto scan_number_exponent; - } - - default: - goto scan_number_done; - } - -scan_number_decimal1: - // state: we just parsed a decimal point - number_type = token_type::value_float; - switch (get()) - { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_decimal2; - } - - default: - { - error_message = "invalid number; expected digit after '.'"; - return token_type::parse_error; - } - } - -scan_number_decimal2: - // we just parsed at least one number after a decimal point - switch (get()) - { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_decimal2; - } - - case 'e': - case 'E': - { - add(current); - goto scan_number_exponent; - } - - default: - goto scan_number_done; - } - -scan_number_exponent: - // we just parsed an exponent - number_type = token_type::value_float; - switch (get()) - { - case '+': - case '-': - { - add(current); - goto scan_number_sign; - } - - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any2; - } - - default: - { - error_message = - "invalid number; expected '+', '-', or digit after exponent"; - return token_type::parse_error; - } - } - -scan_number_sign: - // we just parsed an exponent sign - switch (get()) - { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any2; - } - - default: - { - error_message = "invalid number; expected digit after exponent sign"; - return token_type::parse_error; - } - } - -scan_number_any2: - // we just parsed a number after the exponent or exponent sign - switch (get()) - { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - { - add(current); - goto scan_number_any2; - } - - default: - goto scan_number_done; - } - -scan_number_done: - // unget the character after the number (we only read it to know that - // we are done scanning a number) - unget(); - - char* endptr = nullptr; - errno = 0; - - // try to parse integers first and fall back to floats - if (number_type == token_type::value_unsigned) - { - const auto x = std::strtoull(token_buffer.data(), &endptr, 10); - - // we checked the number format before - JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size()); - - if (errno == 0) - { - value_unsigned = static_cast(x); - if (value_unsigned == x) - { - return token_type::value_unsigned; - } - } - } - else if (number_type == token_type::value_integer) - { - const auto x = std::strtoll(token_buffer.data(), &endptr, 10); - - // we checked the number format before - JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size()); - - if (errno == 0) - { - value_integer = static_cast(x); - if (value_integer == x) - { - return token_type::value_integer; - } - } - } - - // this code is reached if we parse a floating-point number or if an - // integer conversion above failed - strtof(value_float, token_buffer.data(), &endptr); - - // we checked the number format before - JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size()); - - return token_type::value_float; - } - - /*! - @param[in] literal_text the literal text to expect - @param[in] length the length of the passed literal text - @param[in] return_type the token type to return on success - */ - JSON_HEDLEY_NON_NULL(2) - token_type scan_literal(const char_type* literal_text, const std::size_t length, - token_type return_type) - { - JSON_ASSERT(std::char_traits::to_char_type(current) == literal_text[0]); - for (std::size_t i = 1; i < length; ++i) - { - if (JSON_HEDLEY_UNLIKELY(std::char_traits::to_char_type(get()) != literal_text[i])) - { - error_message = "invalid literal"; - return token_type::parse_error; - } - } - return return_type; - } - - ///////////////////// - // input management - ///////////////////// - - /// reset token_buffer; current character is beginning of token - void reset() noexcept - { - token_buffer.clear(); - token_string.clear(); - token_string.push_back(std::char_traits::to_char_type(current)); - } - - /* - @brief get next character from the input - - This function provides the interface to the used input adapter. It does - not throw in case the input reached EOF, but returns a - `std::char_traits::eof()` in that case. Stores the scanned characters - for use in error messages. - - @return character read from the input - */ - char_int_type get() - { - ++position.chars_read_total; - ++position.chars_read_current_line; - - if (next_unget) - { - // just reset the next_unget variable and work with current - next_unget = false; - } - else - { - current = ia.get_character(); - } - - if (JSON_HEDLEY_LIKELY(current != std::char_traits::eof())) - { - token_string.push_back(std::char_traits::to_char_type(current)); - } - - if (current == '\n') - { - ++position.lines_read; - position.chars_read_current_line = 0; - } - - return current; - } - - /*! - @brief unget current character (read it again on next get) - - We implement unget by setting variable next_unget to true. The input is not - changed - we just simulate ungetting by modifying chars_read_total, - chars_read_current_line, and token_string. The next call to get() will - behave as if the unget character is read again. - */ - void unget() - { - next_unget = true; - - --position.chars_read_total; - - // in case we "unget" a newline, we have to also decrement the lines_read - if (position.chars_read_current_line == 0) - { - if (position.lines_read > 0) - { - --position.lines_read; - } - } - else - { - --position.chars_read_current_line; - } - - if (JSON_HEDLEY_LIKELY(current != std::char_traits::eof())) - { - JSON_ASSERT(!token_string.empty()); - token_string.pop_back(); - } - } - - /// add a character to token_buffer - void add(char_int_type c) - { - token_buffer.push_back(static_cast(c)); - } - - public: - ///////////////////// - // value getters - ///////////////////// - - /// return integer value - constexpr number_integer_t get_number_integer() const noexcept - { - return value_integer; - } - - /// return unsigned integer value - constexpr number_unsigned_t get_number_unsigned() const noexcept - { - return value_unsigned; - } - - /// return floating-point value - constexpr number_float_t get_number_float() const noexcept - { - return value_float; - } - - /// return current string value (implicitly resets the token; useful only once) - string_t& get_string() - { - return token_buffer; - } - - ///////////////////// - // diagnostics - ///////////////////// - - /// return position of last read token - constexpr position_t get_position() const noexcept - { - return position; - } - - /// return the last read token (for errors only). Will never contain EOF - /// (an arbitrary value that is not a valid char value, often -1), because - /// 255 may legitimately occur. May contain NUL, which should be escaped. - std::string get_token_string() const - { - // escape control characters - std::string result; - for (const auto c : token_string) - { - if (static_cast(c) <= '\x1F') - { - // escape control characters - std::array cs{{}}; - (std::snprintf)(cs.data(), cs.size(), "", static_cast(c)); - result += cs.data(); - } - else - { - // add character as is - result.push_back(static_cast(c)); - } - } - - return result; - } - - /// return syntax error message - JSON_HEDLEY_RETURNS_NON_NULL - constexpr const char* get_error_message() const noexcept - { - return error_message; - } - - ///////////////////// - // actual scanner - ///////////////////// - - /*! - @brief skip the UTF-8 byte order mark - @return true iff there is no BOM or the correct BOM has been skipped - */ - bool skip_bom() - { - if (get() == 0xEF) - { - // check if we completely parse the BOM - return get() == 0xBB && get() == 0xBF; - } - - // the first character is not the beginning of the BOM; unget it to - // process is later - unget(); - return true; - } - - void skip_whitespace() - { - do - { - get(); - } - while (current == ' ' || current == '\t' || current == '\n' || current == '\r'); - } - - token_type scan() - { - // initially, skip the BOM - if (position.chars_read_total == 0 && !skip_bom()) - { - error_message = "invalid BOM; must be 0xEF 0xBB 0xBF if given"; - return token_type::parse_error; - } - - // read next character and ignore whitespace - skip_whitespace(); - - // ignore comments - while (ignore_comments && current == '/') - { - if (!scan_comment()) - { - return token_type::parse_error; - } - - // skip following whitespace - skip_whitespace(); - } - - switch (current) - { - // structural characters - case '[': - return token_type::begin_array; - case ']': - return token_type::end_array; - case '{': - return token_type::begin_object; - case '}': - return token_type::end_object; - case ':': - return token_type::name_separator; - case ',': - return token_type::value_separator; - - // literals - case 't': - { - std::array true_literal = {{'t', 'r', 'u', 'e'}}; - return scan_literal(true_literal.data(), true_literal.size(), token_type::literal_true); - } - case 'f': - { - std::array false_literal = {{'f', 'a', 'l', 's', 'e'}}; - return scan_literal(false_literal.data(), false_literal.size(), token_type::literal_false); - } - case 'n': - { - std::array null_literal = {{'n', 'u', 'l', 'l'}}; - return scan_literal(null_literal.data(), null_literal.size(), token_type::literal_null); - } - - // string - case '\"': - return scan_string(); - - // number - case '-': - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - return scan_number(); - - // end of input (the null byte is needed when parsing from - // string literals) - case '\0': - case std::char_traits::eof(): - return token_type::end_of_input; - - // error - default: - error_message = "invalid literal"; - return token_type::parse_error; - } - } - - private: - /// input adapter - InputAdapterType ia; - - /// whether comments should be ignored (true) or signaled as errors (false) - const bool ignore_comments = false; - - /// the current character - char_int_type current = std::char_traits::eof(); - - /// whether the next get() call should just return current - bool next_unget = false; - - /// the start position of the current token - position_t position {}; - - /// raw input token string (for error messages) - std::vector token_string {}; - - /// buffer for variable-length tokens (numbers, strings) - string_t token_buffer {}; - - /// a description of occurred lexer errors - const char* error_message = ""; - - // number values - number_integer_t value_integer = 0; - number_unsigned_t value_unsigned = 0; - number_float_t value_float = 0; - - /// the decimal point - const char_int_type decimal_point_char = '.'; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/parser.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/parser.hpp deleted file mode 100644 index ffe483aa1e7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/parser.hpp +++ /dev/null @@ -1,501 +0,0 @@ -#pragma once - -#include // isfinite -#include // uint8_t -#include // function -#include // string -#include // move -#include // vector - -#include -#include -#include -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -//////////// -// parser // -//////////// - -enum class parse_event_t : uint8_t -{ - /// the parser read `{` and started to process a JSON object - object_start, - /// the parser read `}` and finished processing a JSON object - object_end, - /// the parser read `[` and started to process a JSON array - array_start, - /// the parser read `]` and finished processing a JSON array - array_end, - /// the parser read a key of a value in an object - key, - /// the parser finished reading a JSON value - value -}; - -template -using parser_callback_t = - std::function; - -/*! -@brief syntax analysis - -This class implements a recursive descent parser. -*/ -template -class parser -{ - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using lexer_t = lexer; - using token_type = typename lexer_t::token_type; - - public: - /// a parser reading from an input adapter - explicit parser(InputAdapterType&& adapter, - const parser_callback_t cb = nullptr, - const bool allow_exceptions_ = true, - const bool skip_comments = false) - : callback(cb) - , m_lexer(std::move(adapter), skip_comments) - , allow_exceptions(allow_exceptions_) - { - // read first token - get_token(); - } - - /*! - @brief public parser interface - - @param[in] strict whether to expect the last token to be EOF - @param[in,out] result parsed JSON value - - @throw parse_error.101 in case of an unexpected token - @throw parse_error.102 if to_unicode fails or surrogate error - @throw parse_error.103 if to_unicode fails - */ - void parse(const bool strict, BasicJsonType& result) - { - if (callback) - { - json_sax_dom_callback_parser sdp(result, callback, allow_exceptions); - sax_parse_internal(&sdp); - result.assert_invariant(); - - // in strict mode, input must be completely read - if (strict && (get_token() != token_type::end_of_input)) - { - sdp.parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::end_of_input, "value"))); - } - - // in case of an error, return discarded value - if (sdp.is_errored()) - { - result = value_t::discarded; - return; - } - - // set top-level value to null if it was discarded by the callback - // function - if (result.is_discarded()) - { - result = nullptr; - } - } - else - { - json_sax_dom_parser sdp(result, allow_exceptions); - sax_parse_internal(&sdp); - result.assert_invariant(); - - // in strict mode, input must be completely read - if (strict && (get_token() != token_type::end_of_input)) - { - sdp.parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::end_of_input, "value"))); - } - - // in case of an error, return discarded value - if (sdp.is_errored()) - { - result = value_t::discarded; - return; - } - } - } - - /*! - @brief public accept interface - - @param[in] strict whether to expect the last token to be EOF - @return whether the input is a proper JSON text - */ - bool accept(const bool strict = true) - { - json_sax_acceptor sax_acceptor; - return sax_parse(&sax_acceptor, strict); - } - - template - JSON_HEDLEY_NON_NULL(2) - bool sax_parse(SAX* sax, const bool strict = true) - { - (void)detail::is_sax_static_asserts {}; - const bool result = sax_parse_internal(sax); - - // strict mode: next byte must be EOF - if (result && strict && (get_token() != token_type::end_of_input)) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::end_of_input, "value"))); - } - - return result; - } - - private: - template - JSON_HEDLEY_NON_NULL(2) - bool sax_parse_internal(SAX* sax) - { - // stack to remember the hierarchy of structured values we are parsing - // true = array; false = object - std::vector states; - // value to avoid a goto (see comment where set to true) - bool skip_to_state_evaluation = false; - - while (true) - { - if (!skip_to_state_evaluation) - { - // invariant: get_token() was called before each iteration - switch (last_token) - { - case token_type::begin_object: - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_object(std::size_t(-1)))) - { - return false; - } - - // closing } -> we are done - if (get_token() == token_type::end_object) - { - if (JSON_HEDLEY_UNLIKELY(!sax->end_object())) - { - return false; - } - break; - } - - // parse key - if (JSON_HEDLEY_UNLIKELY(last_token != token_type::value_string)) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::value_string, "object key"))); - } - if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string()))) - { - return false; - } - - // parse separator (:) - if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::name_separator)) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::name_separator, "object separator"))); - } - - // remember we are now inside an object - states.push_back(false); - - // parse values - get_token(); - continue; - } - - case token_type::begin_array: - { - if (JSON_HEDLEY_UNLIKELY(!sax->start_array(std::size_t(-1)))) - { - return false; - } - - // closing ] -> we are done - if (get_token() == token_type::end_array) - { - if (JSON_HEDLEY_UNLIKELY(!sax->end_array())) - { - return false; - } - break; - } - - // remember we are now inside an array - states.push_back(true); - - // parse values (no need to call get_token) - continue; - } - - case token_type::value_float: - { - const auto res = m_lexer.get_number_float(); - - if (JSON_HEDLEY_UNLIKELY(!std::isfinite(res))) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - out_of_range::create(406, "number overflow parsing '" + m_lexer.get_token_string() + "'")); - } - - if (JSON_HEDLEY_UNLIKELY(!sax->number_float(res, m_lexer.get_string()))) - { - return false; - } - - break; - } - - case token_type::literal_false: - { - if (JSON_HEDLEY_UNLIKELY(!sax->boolean(false))) - { - return false; - } - break; - } - - case token_type::literal_null: - { - if (JSON_HEDLEY_UNLIKELY(!sax->null())) - { - return false; - } - break; - } - - case token_type::literal_true: - { - if (JSON_HEDLEY_UNLIKELY(!sax->boolean(true))) - { - return false; - } - break; - } - - case token_type::value_integer: - { - if (JSON_HEDLEY_UNLIKELY(!sax->number_integer(m_lexer.get_number_integer()))) - { - return false; - } - break; - } - - case token_type::value_string: - { - if (JSON_HEDLEY_UNLIKELY(!sax->string(m_lexer.get_string()))) - { - return false; - } - break; - } - - case token_type::value_unsigned: - { - if (JSON_HEDLEY_UNLIKELY(!sax->number_unsigned(m_lexer.get_number_unsigned()))) - { - return false; - } - break; - } - - case token_type::parse_error: - { - // using "uninitialized" to avoid "expected" message - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::uninitialized, "value"))); - } - - default: // the last token was unexpected - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::literal_or_value, "value"))); - } - } - } - else - { - skip_to_state_evaluation = false; - } - - // we reached this line after we successfully parsed a value - if (states.empty()) - { - // empty stack: we reached the end of the hierarchy: done - return true; - } - - if (states.back()) // array - { - // comma -> next value - if (get_token() == token_type::value_separator) - { - // parse a new value - get_token(); - continue; - } - - // closing ] - if (JSON_HEDLEY_LIKELY(last_token == token_type::end_array)) - { - if (JSON_HEDLEY_UNLIKELY(!sax->end_array())) - { - return false; - } - - // We are done with this array. Before we can parse a - // new value, we need to evaluate the new state first. - // By setting skip_to_state_evaluation to false, we - // are effectively jumping to the beginning of this if. - JSON_ASSERT(!states.empty()); - states.pop_back(); - skip_to_state_evaluation = true; - continue; - } - - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::end_array, "array"))); - } - else // object - { - // comma -> next value - if (get_token() == token_type::value_separator) - { - // parse key - if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::value_string)) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::value_string, "object key"))); - } - - if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string()))) - { - return false; - } - - // parse separator (:) - if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::name_separator)) - { - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::name_separator, "object separator"))); - } - - // parse values - get_token(); - continue; - } - - // closing } - if (JSON_HEDLEY_LIKELY(last_token == token_type::end_object)) - { - if (JSON_HEDLEY_UNLIKELY(!sax->end_object())) - { - return false; - } - - // We are done with this object. Before we can parse a - // new value, we need to evaluate the new state first. - // By setting skip_to_state_evaluation to false, we - // are effectively jumping to the beginning of this if. - JSON_ASSERT(!states.empty()); - states.pop_back(); - skip_to_state_evaluation = true; - continue; - } - - return sax->parse_error(m_lexer.get_position(), - m_lexer.get_token_string(), - parse_error::create(101, m_lexer.get_position(), - exception_message(token_type::end_object, "object"))); - } - } - } - - /// get next token from lexer - token_type get_token() - { - return last_token = m_lexer.scan(); - } - - std::string exception_message(const token_type expected, const std::string& context) - { - std::string error_msg = "syntax error "; - - if (!context.empty()) - { - error_msg += "while parsing " + context + " "; - } - - error_msg += "- "; - - if (last_token == token_type::parse_error) - { - error_msg += std::string(m_lexer.get_error_message()) + "; last read: '" + - m_lexer.get_token_string() + "'"; - } - else - { - error_msg += "unexpected " + std::string(lexer_t::token_type_name(last_token)); - } - - if (expected != token_type::uninitialized) - { - error_msg += "; expected " + std::string(lexer_t::token_type_name(expected)); - } - - return error_msg; - } - - private: - /// callback function - const parser_callback_t callback = nullptr; - /// the type of the last read token - token_type last_token = token_type::uninitialized; - /// the lexer - lexer_t m_lexer; - /// whether to throw exceptions in case of errors - const bool allow_exceptions = true; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/position_t.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/position_t.hpp deleted file mode 100644 index 14e9649fb29..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/input/position_t.hpp +++ /dev/null @@ -1,27 +0,0 @@ -#pragma once - -#include // size_t - -namespace nlohmann -{ -namespace detail -{ -/// struct to capture the start position of the current token -struct position_t -{ - /// the total number of characters read - std::size_t chars_read_total = 0; - /// the number of characters read in the current line - std::size_t chars_read_current_line = 0; - /// the number of lines read - std::size_t lines_read = 0; - - /// conversion to size_t to preserve SAX interface - constexpr operator size_t() const - { - return chars_read_total; - } -}; - -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/internal_iterator.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/internal_iterator.hpp deleted file mode 100644 index 2c81f723fd2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/internal_iterator.hpp +++ /dev/null @@ -1,25 +0,0 @@ -#pragma once - -#include - -namespace nlohmann -{ -namespace detail -{ -/*! -@brief an iterator value - -@note This structure could easily be a union, but MSVC currently does not allow -unions members with complex constructors, see https://github.com/nlohmann/json/pull/105. -*/ -template struct internal_iterator -{ - /// iterator for JSON objects - typename BasicJsonType::object_t::iterator object_iterator {}; - /// iterator for JSON arrays - typename BasicJsonType::array_t::iterator array_iterator {}; - /// generic iterator for all other types - primitive_iterator_t primitive_iterator {}; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iter_impl.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iter_impl.hpp deleted file mode 100644 index e9a394d4cc1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iter_impl.hpp +++ /dev/null @@ -1,637 +0,0 @@ -#pragma once - -#include // iterator, random_access_iterator_tag, bidirectional_iterator_tag, advance, next -#include // conditional, is_const, remove_const - -#include -#include -#include -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -// forward declare, to be able to friend it later on -template class iteration_proxy; -template class iteration_proxy_value; - -/*! -@brief a template for a bidirectional iterator for the @ref basic_json class -This class implements a both iterators (iterator and const_iterator) for the -@ref basic_json class. -@note An iterator is called *initialized* when a pointer to a JSON value has - been set (e.g., by a constructor or a copy assignment). If the iterator is - default-constructed, it is *uninitialized* and most methods are undefined. - **The library uses assertions to detect calls on uninitialized iterators.** -@requirement The class satisfies the following concept requirements: -- -[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator): - The iterator that can be moved can be moved in both directions (i.e. - incremented and decremented). -@since version 1.0.0, simplified in version 2.0.9, change to bidirectional - iterators in version 3.0.0 (see https://github.com/nlohmann/json/issues/593) -*/ -template -class iter_impl -{ - /// allow basic_json to access private members - friend iter_impl::value, typename std::remove_const::type, const BasicJsonType>::type>; - friend BasicJsonType; - friend iteration_proxy; - friend iteration_proxy_value; - - using object_t = typename BasicJsonType::object_t; - using array_t = typename BasicJsonType::array_t; - // make sure BasicJsonType is basic_json or const basic_json - static_assert(is_basic_json::type>::value, - "iter_impl only accepts (const) basic_json"); - - public: - - /// The std::iterator class template (used as a base class to provide typedefs) is deprecated in C++17. - /// The C++ Standard has never required user-defined iterators to derive from std::iterator. - /// A user-defined iterator should provide publicly accessible typedefs named - /// iterator_category, value_type, difference_type, pointer, and reference. - /// Note that value_type is required to be non-const, even for constant iterators. - using iterator_category = std::bidirectional_iterator_tag; - - /// the type of the values when the iterator is dereferenced - using value_type = typename BasicJsonType::value_type; - /// a type to represent differences between iterators - using difference_type = typename BasicJsonType::difference_type; - /// defines a pointer to the type iterated over (value_type) - using pointer = typename std::conditional::value, - typename BasicJsonType::const_pointer, - typename BasicJsonType::pointer>::type; - /// defines a reference to the type iterated over (value_type) - using reference = - typename std::conditional::value, - typename BasicJsonType::const_reference, - typename BasicJsonType::reference>::type; - - /// default constructor - iter_impl() = default; - - /*! - @brief constructor for a given JSON instance - @param[in] object pointer to a JSON object for this iterator - @pre object != nullptr - @post The iterator is initialized; i.e. `m_object != nullptr`. - */ - explicit iter_impl(pointer object) noexcept : m_object(object) - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - m_it.object_iterator = typename object_t::iterator(); - break; - } - - case value_t::array: - { - m_it.array_iterator = typename array_t::iterator(); - break; - } - - default: - { - m_it.primitive_iterator = primitive_iterator_t(); - break; - } - } - } - - /*! - @note The conventional copy constructor and copy assignment are implicitly - defined. Combined with the following converting constructor and - assignment, they support: (1) copy from iterator to iterator, (2) - copy from const iterator to const iterator, and (3) conversion from - iterator to const iterator. However conversion from const iterator - to iterator is not defined. - */ - - /*! - @brief const copy constructor - @param[in] other const iterator to copy from - @note This copy constructor had to be defined explicitly to circumvent a bug - occurring on msvc v19.0 compiler (VS 2015) debug build. For more - information refer to: https://github.com/nlohmann/json/issues/1608 - */ - iter_impl(const iter_impl& other) noexcept - : m_object(other.m_object), m_it(other.m_it) - {} - - /*! - @brief converting assignment - @param[in] other const iterator to copy from - @return const/non-const iterator - @note It is not checked whether @a other is initialized. - */ - iter_impl& operator=(const iter_impl& other) noexcept - { - m_object = other.m_object; - m_it = other.m_it; - return *this; - } - - /*! - @brief converting constructor - @param[in] other non-const iterator to copy from - @note It is not checked whether @a other is initialized. - */ - iter_impl(const iter_impl::type>& other) noexcept - : m_object(other.m_object), m_it(other.m_it) - {} - - /*! - @brief converting assignment - @param[in] other non-const iterator to copy from - @return const/non-const iterator - @note It is not checked whether @a other is initialized. - */ - iter_impl& operator=(const iter_impl::type>& other) noexcept - { - m_object = other.m_object; - m_it = other.m_it; - return *this; - } - - private: - /*! - @brief set the iterator to the first value - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - void set_begin() noexcept - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - m_it.object_iterator = m_object->m_value.object->begin(); - break; - } - - case value_t::array: - { - m_it.array_iterator = m_object->m_value.array->begin(); - break; - } - - case value_t::null: - { - // set to end so begin()==end() is true: null is empty - m_it.primitive_iterator.set_end(); - break; - } - - default: - { - m_it.primitive_iterator.set_begin(); - break; - } - } - } - - /*! - @brief set the iterator past the last value - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - void set_end() noexcept - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - m_it.object_iterator = m_object->m_value.object->end(); - break; - } - - case value_t::array: - { - m_it.array_iterator = m_object->m_value.array->end(); - break; - } - - default: - { - m_it.primitive_iterator.set_end(); - break; - } - } - } - - public: - /*! - @brief return a reference to the value pointed to by the iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - reference operator*() const - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - JSON_ASSERT(m_it.object_iterator != m_object->m_value.object->end()); - return m_it.object_iterator->second; - } - - case value_t::array: - { - JSON_ASSERT(m_it.array_iterator != m_object->m_value.array->end()); - return *m_it.array_iterator; - } - - case value_t::null: - JSON_THROW(invalid_iterator::create(214, "cannot get value")); - - default: - { - if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.is_begin())) - { - return *m_object; - } - - JSON_THROW(invalid_iterator::create(214, "cannot get value")); - } - } - } - - /*! - @brief dereference the iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - pointer operator->() const - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - JSON_ASSERT(m_it.object_iterator != m_object->m_value.object->end()); - return &(m_it.object_iterator->second); - } - - case value_t::array: - { - JSON_ASSERT(m_it.array_iterator != m_object->m_value.array->end()); - return &*m_it.array_iterator; - } - - default: - { - if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.is_begin())) - { - return m_object; - } - - JSON_THROW(invalid_iterator::create(214, "cannot get value")); - } - } - } - - /*! - @brief post-increment (it++) - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl const operator++(int) - { - auto result = *this; - ++(*this); - return result; - } - - /*! - @brief pre-increment (++it) - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl& operator++() - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - std::advance(m_it.object_iterator, 1); - break; - } - - case value_t::array: - { - std::advance(m_it.array_iterator, 1); - break; - } - - default: - { - ++m_it.primitive_iterator; - break; - } - } - - return *this; - } - - /*! - @brief post-decrement (it--) - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl const operator--(int) - { - auto result = *this; - --(*this); - return result; - } - - /*! - @brief pre-decrement (--it) - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl& operator--() - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - { - std::advance(m_it.object_iterator, -1); - break; - } - - case value_t::array: - { - std::advance(m_it.array_iterator, -1); - break; - } - - default: - { - --m_it.primitive_iterator; - break; - } - } - - return *this; - } - - /*! - @brief comparison: equal - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator==(const iter_impl& other) const - { - // if objects are not the same, the comparison is undefined - if (JSON_HEDLEY_UNLIKELY(m_object != other.m_object)) - { - JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers")); - } - - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - return (m_it.object_iterator == other.m_it.object_iterator); - - case value_t::array: - return (m_it.array_iterator == other.m_it.array_iterator); - - default: - return (m_it.primitive_iterator == other.m_it.primitive_iterator); - } - } - - /*! - @brief comparison: not equal - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator!=(const iter_impl& other) const - { - return !operator==(other); - } - - /*! - @brief comparison: smaller - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator<(const iter_impl& other) const - { - // if objects are not the same, the comparison is undefined - if (JSON_HEDLEY_UNLIKELY(m_object != other.m_object)) - { - JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers")); - } - - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - JSON_THROW(invalid_iterator::create(213, "cannot compare order of object iterators")); - - case value_t::array: - return (m_it.array_iterator < other.m_it.array_iterator); - - default: - return (m_it.primitive_iterator < other.m_it.primitive_iterator); - } - } - - /*! - @brief comparison: less than or equal - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator<=(const iter_impl& other) const - { - return !other.operator < (*this); - } - - /*! - @brief comparison: greater than - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator>(const iter_impl& other) const - { - return !operator<=(other); - } - - /*! - @brief comparison: greater than or equal - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - bool operator>=(const iter_impl& other) const - { - return !operator<(other); - } - - /*! - @brief add to iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl& operator+=(difference_type i) - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators")); - - case value_t::array: - { - std::advance(m_it.array_iterator, i); - break; - } - - default: - { - m_it.primitive_iterator += i; - break; - } - } - - return *this; - } - - /*! - @brief subtract from iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl& operator-=(difference_type i) - { - return operator+=(-i); - } - - /*! - @brief add to iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl operator+(difference_type i) const - { - auto result = *this; - result += i; - return result; - } - - /*! - @brief addition of distance and iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - friend iter_impl operator+(difference_type i, const iter_impl& it) - { - auto result = it; - result += i; - return result; - } - - /*! - @brief subtract from iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - iter_impl operator-(difference_type i) const - { - auto result = *this; - result -= i; - return result; - } - - /*! - @brief return difference - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - difference_type operator-(const iter_impl& other) const - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators")); - - case value_t::array: - return m_it.array_iterator - other.m_it.array_iterator; - - default: - return m_it.primitive_iterator - other.m_it.primitive_iterator; - } - } - - /*! - @brief access to successor - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - reference operator[](difference_type n) const - { - JSON_ASSERT(m_object != nullptr); - - switch (m_object->m_type) - { - case value_t::object: - JSON_THROW(invalid_iterator::create(208, "cannot use operator[] for object iterators")); - - case value_t::array: - return *std::next(m_it.array_iterator, n); - - case value_t::null: - JSON_THROW(invalid_iterator::create(214, "cannot get value")); - - default: - { - if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.get_value() == -n)) - { - return *m_object; - } - - JSON_THROW(invalid_iterator::create(214, "cannot get value")); - } - } - } - - /*! - @brief return the key of an object iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - const typename object_t::key_type& key() const - { - JSON_ASSERT(m_object != nullptr); - - if (JSON_HEDLEY_LIKELY(m_object->is_object())) - { - return m_it.object_iterator->first; - } - - JSON_THROW(invalid_iterator::create(207, "cannot use key() for non-object iterators")); - } - - /*! - @brief return the value of an iterator - @pre The iterator is initialized; i.e. `m_object != nullptr`. - */ - reference value() const - { - return operator*(); - } - - private: - /// associated JSON instance - pointer m_object = nullptr; - /// the actual iterator of the associated instance - internal_iterator::type> m_it {}; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iteration_proxy.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iteration_proxy.hpp deleted file mode 100644 index 74b4eb347f7..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iteration_proxy.hpp +++ /dev/null @@ -1,178 +0,0 @@ -#pragma once - -#include // size_t -#include // input_iterator_tag -#include // string, to_string -#include // tuple_size, get, tuple_element - -#include -#include - -namespace nlohmann -{ -namespace detail -{ -template -void int_to_string( string_type& target, std::size_t value ) -{ - // For ADL - using std::to_string; - target = to_string(value); -} -template class iteration_proxy_value -{ - public: - using difference_type = std::ptrdiff_t; - using value_type = iteration_proxy_value; - using pointer = value_type * ; - using reference = value_type & ; - using iterator_category = std::input_iterator_tag; - using string_type = typename std::remove_cv< typename std::remove_reference().key() ) >::type >::type; - - private: - /// the iterator - IteratorType anchor; - /// an index for arrays (used to create key names) - std::size_t array_index = 0; - /// last stringified array index - mutable std::size_t array_index_last = 0; - /// a string representation of the array index - mutable string_type array_index_str = "0"; - /// an empty string (to return a reference for primitive values) - const string_type empty_str = ""; - - public: - explicit iteration_proxy_value(IteratorType it) noexcept : anchor(it) {} - - /// dereference operator (needed for range-based for) - iteration_proxy_value& operator*() - { - return *this; - } - - /// increment operator (needed for range-based for) - iteration_proxy_value& operator++() - { - ++anchor; - ++array_index; - - return *this; - } - - /// equality operator (needed for InputIterator) - bool operator==(const iteration_proxy_value& o) const - { - return anchor == o.anchor; - } - - /// inequality operator (needed for range-based for) - bool operator!=(const iteration_proxy_value& o) const - { - return anchor != o.anchor; - } - - /// return key of the iterator - const string_type& key() const - { - JSON_ASSERT(anchor.m_object != nullptr); - - switch (anchor.m_object->type()) - { - // use integer array index as key - case value_t::array: - { - if (array_index != array_index_last) - { - int_to_string( array_index_str, array_index ); - array_index_last = array_index; - } - return array_index_str; - } - - // use key from the object - case value_t::object: - return anchor.key(); - - // use an empty key for all primitive types - default: - return empty_str; - } - } - - /// return value of the iterator - typename IteratorType::reference value() const - { - return anchor.value(); - } -}; - -/// proxy class for the items() function -template class iteration_proxy -{ - private: - /// the container to iterate - typename IteratorType::reference container; - - public: - /// construct iteration proxy from a container - explicit iteration_proxy(typename IteratorType::reference cont) noexcept - : container(cont) {} - - /// return iterator begin (needed for range-based for) - iteration_proxy_value begin() noexcept - { - return iteration_proxy_value(container.begin()); - } - - /// return iterator end (needed for range-based for) - iteration_proxy_value end() noexcept - { - return iteration_proxy_value(container.end()); - } -}; -// Structured Bindings Support -// For further reference see https://blog.tartanllama.xyz/structured-bindings/ -// And see https://github.com/nlohmann/json/pull/1391 -template = 0> -auto get(const nlohmann::detail::iteration_proxy_value& i) -> decltype(i.key()) -{ - return i.key(); -} -// Structured Bindings Support -// For further reference see https://blog.tartanllama.xyz/structured-bindings/ -// And see https://github.com/nlohmann/json/pull/1391 -template = 0> -auto get(const nlohmann::detail::iteration_proxy_value& i) -> decltype(i.value()) -{ - return i.value(); -} -} // namespace detail -} // namespace nlohmann - -// The Addition to the STD Namespace is required to add -// Structured Bindings Support to the iteration_proxy_value class -// For further reference see https://blog.tartanllama.xyz/structured-bindings/ -// And see https://github.com/nlohmann/json/pull/1391 -namespace std -{ -#if defined(__clang__) - // Fix: https://github.com/nlohmann/json/issues/1401 - #pragma clang diagnostic push - #pragma clang diagnostic ignored "-Wmismatched-tags" -#endif -template -class tuple_size<::nlohmann::detail::iteration_proxy_value> - : public std::integral_constant {}; - -template -class tuple_element> -{ - public: - using type = decltype( - get(std::declval < - ::nlohmann::detail::iteration_proxy_value> ())); -}; -#if defined(__clang__) - #pragma clang diagnostic pop -#endif -} // namespace std diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iterator_traits.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iterator_traits.hpp deleted file mode 100644 index da5636188f5..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/iterator_traits.hpp +++ /dev/null @@ -1,51 +0,0 @@ -#pragma once - -#include // random_access_iterator_tag - -#include -#include - -namespace nlohmann -{ -namespace detail -{ -template -struct iterator_types {}; - -template -struct iterator_types < - It, - void_t> -{ - using difference_type = typename It::difference_type; - using value_type = typename It::value_type; - using pointer = typename It::pointer; - using reference = typename It::reference; - using iterator_category = typename It::iterator_category; -}; - -// This is required as some compilers implement std::iterator_traits in a way that -// doesn't work with SFINAE. See https://github.com/nlohmann/json/issues/1341. -template -struct iterator_traits -{ -}; - -template -struct iterator_traits < T, enable_if_t < !std::is_pointer::value >> - : iterator_types -{ -}; - -template -struct iterator_traits::value>> -{ - using iterator_category = std::random_access_iterator_tag; - using value_type = T; - using difference_type = ptrdiff_t; - using pointer = T*; - using reference = T&; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/json_reverse_iterator.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/json_reverse_iterator.hpp deleted file mode 100644 index f3b5b5db6b8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/json_reverse_iterator.hpp +++ /dev/null @@ -1,119 +0,0 @@ -#pragma once - -#include // ptrdiff_t -#include // reverse_iterator -#include // declval - -namespace nlohmann -{ -namespace detail -{ -////////////////////// -// reverse_iterator // -////////////////////// - -/*! -@brief a template for a reverse iterator class - -@tparam Base the base iterator type to reverse. Valid types are @ref -iterator (to create @ref reverse_iterator) and @ref const_iterator (to -create @ref const_reverse_iterator). - -@requirement The class satisfies the following concept requirements: -- -[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator): - The iterator that can be moved can be moved in both directions (i.e. - incremented and decremented). -- [OutputIterator](https://en.cppreference.com/w/cpp/named_req/OutputIterator): - It is possible to write to the pointed-to element (only if @a Base is - @ref iterator). - -@since version 1.0.0 -*/ -template -class json_reverse_iterator : public std::reverse_iterator -{ - public: - using difference_type = std::ptrdiff_t; - /// shortcut to the reverse iterator adapter - using base_iterator = std::reverse_iterator; - /// the reference type for the pointed-to element - using reference = typename Base::reference; - - /// create reverse iterator from iterator - explicit json_reverse_iterator(const typename base_iterator::iterator_type& it) noexcept - : base_iterator(it) {} - - /// create reverse iterator from base class - explicit json_reverse_iterator(const base_iterator& it) noexcept : base_iterator(it) {} - - /// post-increment (it++) - json_reverse_iterator const operator++(int) - { - return static_cast(base_iterator::operator++(1)); - } - - /// pre-increment (++it) - json_reverse_iterator& operator++() - { - return static_cast(base_iterator::operator++()); - } - - /// post-decrement (it--) - json_reverse_iterator const operator--(int) - { - return static_cast(base_iterator::operator--(1)); - } - - /// pre-decrement (--it) - json_reverse_iterator& operator--() - { - return static_cast(base_iterator::operator--()); - } - - /// add to iterator - json_reverse_iterator& operator+=(difference_type i) - { - return static_cast(base_iterator::operator+=(i)); - } - - /// add to iterator - json_reverse_iterator operator+(difference_type i) const - { - return static_cast(base_iterator::operator+(i)); - } - - /// subtract from iterator - json_reverse_iterator operator-(difference_type i) const - { - return static_cast(base_iterator::operator-(i)); - } - - /// return difference - difference_type operator-(const json_reverse_iterator& other) const - { - return base_iterator(*this) - base_iterator(other); - } - - /// access to successor - reference operator[](difference_type n) const - { - return *(this->operator+(n)); - } - - /// return the key of an object iterator - auto key() const -> decltype(std::declval().key()) - { - auto it = --this->base(); - return it.key(); - } - - /// return the value of an iterator - reference value() const - { - auto it = --this->base(); - return it.operator * (); - } -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/primitive_iterator.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/primitive_iterator.hpp deleted file mode 100644 index 28d6f1a65df..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/iterators/primitive_iterator.hpp +++ /dev/null @@ -1,120 +0,0 @@ -#pragma once - -#include // ptrdiff_t -#include // numeric_limits - -namespace nlohmann -{ -namespace detail -{ -/* -@brief an iterator for primitive JSON types - -This class models an iterator for primitive JSON types (boolean, number, -string). It's only purpose is to allow the iterator/const_iterator classes -to "iterate" over primitive values. Internally, the iterator is modeled by -a `difference_type` variable. Value begin_value (`0`) models the begin, -end_value (`1`) models past the end. -*/ -class primitive_iterator_t -{ - private: - using difference_type = std::ptrdiff_t; - static constexpr difference_type begin_value = 0; - static constexpr difference_type end_value = begin_value + 1; - - /// iterator as signed integer type - difference_type m_it = (std::numeric_limits::min)(); - - public: - constexpr difference_type get_value() const noexcept - { - return m_it; - } - - /// set iterator to a defined beginning - void set_begin() noexcept - { - m_it = begin_value; - } - - /// set iterator to a defined past the end - void set_end() noexcept - { - m_it = end_value; - } - - /// return whether the iterator can be dereferenced - constexpr bool is_begin() const noexcept - { - return m_it == begin_value; - } - - /// return whether the iterator is at end - constexpr bool is_end() const noexcept - { - return m_it == end_value; - } - - friend constexpr bool operator==(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept - { - return lhs.m_it == rhs.m_it; - } - - friend constexpr bool operator<(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept - { - return lhs.m_it < rhs.m_it; - } - - primitive_iterator_t operator+(difference_type n) noexcept - { - auto result = *this; - result += n; - return result; - } - - friend constexpr difference_type operator-(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept - { - return lhs.m_it - rhs.m_it; - } - - primitive_iterator_t& operator++() noexcept - { - ++m_it; - return *this; - } - - primitive_iterator_t const operator++(int) noexcept - { - auto result = *this; - ++m_it; - return result; - } - - primitive_iterator_t& operator--() noexcept - { - --m_it; - return *this; - } - - primitive_iterator_t const operator--(int) noexcept - { - auto result = *this; - --m_it; - return result; - } - - primitive_iterator_t& operator+=(difference_type n) noexcept - { - m_it += n; - return *this; - } - - primitive_iterator_t& operator-=(difference_type n) noexcept - { - m_it -= n; - return *this; - } -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_pointer.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_pointer.hpp deleted file mode 100644 index 78bc3a3a38d..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_pointer.hpp +++ /dev/null @@ -1,975 +0,0 @@ -#pragma once - -#include // all_of -#include // isdigit -#include // max -#include // accumulate -#include // string -#include // move -#include // vector - -#include -#include -#include - -namespace nlohmann -{ -template -class json_pointer -{ - // allow basic_json to access private members - NLOHMANN_BASIC_JSON_TPL_DECLARATION - friend class basic_json; - - public: - /*! - @brief create JSON pointer - - Create a JSON pointer according to the syntax described in - [Section 3 of RFC6901](https://tools.ietf.org/html/rfc6901#section-3). - - @param[in] s string representing the JSON pointer; if omitted, the empty - string is assumed which references the whole JSON value - - @throw parse_error.107 if the given JSON pointer @a s is nonempty and does - not begin with a slash (`/`); see example below - - @throw parse_error.108 if a tilde (`~`) in the given JSON pointer @a s is - not followed by `0` (representing `~`) or `1` (representing `/`); see - example below - - @liveexample{The example shows the construction several valid JSON pointers - as well as the exceptional behavior.,json_pointer} - - @since version 2.0.0 - */ - explicit json_pointer(const std::string& s = "") - : reference_tokens(split(s)) - {} - - /*! - @brief return a string representation of the JSON pointer - - @invariant For each JSON pointer `ptr`, it holds: - @code {.cpp} - ptr == json_pointer(ptr.to_string()); - @endcode - - @return a string representation of the JSON pointer - - @liveexample{The example shows the result of `to_string`.,json_pointer__to_string} - - @since version 2.0.0 - */ - std::string to_string() const - { - return std::accumulate(reference_tokens.begin(), reference_tokens.end(), - std::string{}, - [](const std::string & a, const std::string & b) - { - return a + "/" + escape(b); - }); - } - - /// @copydoc to_string() - operator std::string() const - { - return to_string(); - } - - /*! - @brief append another JSON pointer at the end of this JSON pointer - - @param[in] ptr JSON pointer to append - @return JSON pointer with @a ptr appended - - @liveexample{The example shows the usage of `operator/=`.,json_pointer__operator_add} - - @complexity Linear in the length of @a ptr. - - @sa @ref operator/=(std::string) to append a reference token - @sa @ref operator/=(std::size_t) to append an array index - @sa @ref operator/(const json_pointer&, const json_pointer&) for a binary operator - - @since version 3.6.0 - */ - json_pointer& operator/=(const json_pointer& ptr) - { - reference_tokens.insert(reference_tokens.end(), - ptr.reference_tokens.begin(), - ptr.reference_tokens.end()); - return *this; - } - - /*! - @brief append an unescaped reference token at the end of this JSON pointer - - @param[in] token reference token to append - @return JSON pointer with @a token appended without escaping @a token - - @liveexample{The example shows the usage of `operator/=`.,json_pointer__operator_add} - - @complexity Amortized constant. - - @sa @ref operator/=(const json_pointer&) to append a JSON pointer - @sa @ref operator/=(std::size_t) to append an array index - @sa @ref operator/(const json_pointer&, std::size_t) for a binary operator - - @since version 3.6.0 - */ - json_pointer& operator/=(std::string token) - { - push_back(std::move(token)); - return *this; - } - - /*! - @brief append an array index at the end of this JSON pointer - - @param[in] array_idx array index to append - @return JSON pointer with @a array_idx appended - - @liveexample{The example shows the usage of `operator/=`.,json_pointer__operator_add} - - @complexity Amortized constant. - - @sa @ref operator/=(const json_pointer&) to append a JSON pointer - @sa @ref operator/=(std::string) to append a reference token - @sa @ref operator/(const json_pointer&, std::string) for a binary operator - - @since version 3.6.0 - */ - json_pointer& operator/=(std::size_t array_idx) - { - return *this /= std::to_string(array_idx); - } - - /*! - @brief create a new JSON pointer by appending the right JSON pointer at the end of the left JSON pointer - - @param[in] lhs JSON pointer - @param[in] rhs JSON pointer - @return a new JSON pointer with @a rhs appended to @a lhs - - @liveexample{The example shows the usage of `operator/`.,json_pointer__operator_add_binary} - - @complexity Linear in the length of @a lhs and @a rhs. - - @sa @ref operator/=(const json_pointer&) to append a JSON pointer - - @since version 3.6.0 - */ - friend json_pointer operator/(const json_pointer& lhs, - const json_pointer& rhs) - { - return json_pointer(lhs) /= rhs; - } - - /*! - @brief create a new JSON pointer by appending the unescaped token at the end of the JSON pointer - - @param[in] ptr JSON pointer - @param[in] token reference token - @return a new JSON pointer with unescaped @a token appended to @a ptr - - @liveexample{The example shows the usage of `operator/`.,json_pointer__operator_add_binary} - - @complexity Linear in the length of @a ptr. - - @sa @ref operator/=(std::string) to append a reference token - - @since version 3.6.0 - */ - friend json_pointer operator/(const json_pointer& ptr, std::string token) - { - return json_pointer(ptr) /= std::move(token); - } - - /*! - @brief create a new JSON pointer by appending the array-index-token at the end of the JSON pointer - - @param[in] ptr JSON pointer - @param[in] array_idx array index - @return a new JSON pointer with @a array_idx appended to @a ptr - - @liveexample{The example shows the usage of `operator/`.,json_pointer__operator_add_binary} - - @complexity Linear in the length of @a ptr. - - @sa @ref operator/=(std::size_t) to append an array index - - @since version 3.6.0 - */ - friend json_pointer operator/(const json_pointer& ptr, std::size_t array_idx) - { - return json_pointer(ptr) /= array_idx; - } - - /*! - @brief returns the parent of this JSON pointer - - @return parent of this JSON pointer; in case this JSON pointer is the root, - the root itself is returned - - @complexity Linear in the length of the JSON pointer. - - @liveexample{The example shows the result of `parent_pointer` for different - JSON Pointers.,json_pointer__parent_pointer} - - @since version 3.6.0 - */ - json_pointer parent_pointer() const - { - if (empty()) - { - return *this; - } - - json_pointer res = *this; - res.pop_back(); - return res; - } - - /*! - @brief remove last reference token - - @pre not `empty()` - - @liveexample{The example shows the usage of `pop_back`.,json_pointer__pop_back} - - @complexity Constant. - - @throw out_of_range.405 if JSON pointer has no parent - - @since version 3.6.0 - */ - void pop_back() - { - if (JSON_HEDLEY_UNLIKELY(empty())) - { - JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent")); - } - - reference_tokens.pop_back(); - } - - /*! - @brief return last reference token - - @pre not `empty()` - @return last reference token - - @liveexample{The example shows the usage of `back`.,json_pointer__back} - - @complexity Constant. - - @throw out_of_range.405 if JSON pointer has no parent - - @since version 3.6.0 - */ - const std::string& back() const - { - if (JSON_HEDLEY_UNLIKELY(empty())) - { - JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent")); - } - - return reference_tokens.back(); - } - - /*! - @brief append an unescaped token at the end of the reference pointer - - @param[in] token token to add - - @complexity Amortized constant. - - @liveexample{The example shows the result of `push_back` for different - JSON Pointers.,json_pointer__push_back} - - @since version 3.6.0 - */ - void push_back(const std::string& token) - { - reference_tokens.push_back(token); - } - - /// @copydoc push_back(const std::string&) - void push_back(std::string&& token) - { - reference_tokens.push_back(std::move(token)); - } - - /*! - @brief return whether pointer points to the root document - - @return true iff the JSON pointer points to the root document - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example shows the result of `empty` for different JSON - Pointers.,json_pointer__empty} - - @since version 3.6.0 - */ - bool empty() const noexcept - { - return reference_tokens.empty(); - } - - private: - /*! - @param[in] s reference token to be converted into an array index - - @return integer representation of @a s - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index begins not with a digit - @throw out_of_range.404 if string @a s could not be converted to an integer - @throw out_of_range.410 if an array index exceeds size_type - */ - static typename BasicJsonType::size_type array_index(const std::string& s) - { - using size_type = typename BasicJsonType::size_type; - - // error condition (cf. RFC 6901, Sect. 4) - if (JSON_HEDLEY_UNLIKELY(s.size() > 1 && s[0] == '0')) - { - JSON_THROW(detail::parse_error::create(106, 0, - "array index '" + s + - "' must not begin with '0'")); - } - - // error condition (cf. RFC 6901, Sect. 4) - if (JSON_HEDLEY_UNLIKELY(s.size() > 1 && !(s[0] >= '1' && s[0] <= '9'))) - { - JSON_THROW(detail::parse_error::create(109, 0, "array index '" + s + "' is not a number")); - } - - std::size_t processed_chars = 0; - unsigned long long res = 0; - JSON_TRY - { - res = std::stoull(s, &processed_chars); - } - JSON_CATCH(std::out_of_range&) - { - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + s + "'")); - } - - // check if the string was completely read - if (JSON_HEDLEY_UNLIKELY(processed_chars != s.size())) - { - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + s + "'")); - } - - // only triggered on special platforms (like 32bit), see also - // https://github.com/nlohmann/json/pull/2203 - if (res >= static_cast((std::numeric_limits::max)())) - { - JSON_THROW(detail::out_of_range::create(410, "array index " + s + " exceeds size_type")); // LCOV_EXCL_LINE - } - - return static_cast(res); - } - - json_pointer top() const - { - if (JSON_HEDLEY_UNLIKELY(empty())) - { - JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent")); - } - - json_pointer result = *this; - result.reference_tokens = {reference_tokens[0]}; - return result; - } - - /*! - @brief create and return a reference to the pointed to value - - @complexity Linear in the number of reference tokens. - - @throw parse_error.109 if array index is not a number - @throw type_error.313 if value cannot be unflattened - */ - BasicJsonType& get_and_create(BasicJsonType& j) const - { - auto result = &j; - - // in case no reference tokens exist, return a reference to the JSON value - // j which will be overwritten by a primitive value - for (const auto& reference_token : reference_tokens) - { - switch (result->type()) - { - case detail::value_t::null: - { - if (reference_token == "0") - { - // start a new array if reference token is 0 - result = &result->operator[](0); - } - else - { - // start a new object otherwise - result = &result->operator[](reference_token); - } - break; - } - - case detail::value_t::object: - { - // create an entry in the object - result = &result->operator[](reference_token); - break; - } - - case detail::value_t::array: - { - // create an entry in the array - result = &result->operator[](array_index(reference_token)); - break; - } - - /* - The following code is only reached if there exists a reference - token _and_ the current value is primitive. In this case, we have - an error situation, because primitive values may only occur as - single value; that is, with an empty list of reference tokens. - */ - default: - JSON_THROW(detail::type_error::create(313, "invalid value to unflatten")); - } - } - - return *result; - } - - /*! - @brief return a reference to the pointed to value - - @note This version does not throw if a value is not present, but tries to - create nested values instead. For instance, calling this function - with pointer `"/this/that"` on a null value is equivalent to calling - `operator[]("this").operator[]("that")` on that value, effectively - changing the null value to an object. - - @param[in] ptr a JSON value - - @return reference to the JSON value pointed to by the JSON pointer - - @complexity Linear in the length of the JSON pointer. - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.404 if the JSON pointer can not be resolved - */ - BasicJsonType& get_unchecked(BasicJsonType* ptr) const - { - for (const auto& reference_token : reference_tokens) - { - // convert null values to arrays or objects before continuing - if (ptr->is_null()) - { - // check if reference token is a number - const bool nums = - std::all_of(reference_token.begin(), reference_token.end(), - [](const unsigned char x) - { - return std::isdigit(x); - }); - - // change value to array for numbers or "-" or to object otherwise - *ptr = (nums || reference_token == "-") - ? detail::value_t::array - : detail::value_t::object; - } - - switch (ptr->type()) - { - case detail::value_t::object: - { - // use unchecked object access - ptr = &ptr->operator[](reference_token); - break; - } - - case detail::value_t::array: - { - if (reference_token == "-") - { - // explicitly treat "-" as index beyond the end - ptr = &ptr->operator[](ptr->m_value.array->size()); - } - else - { - // convert array index to number; unchecked access - ptr = &ptr->operator[](array_index(reference_token)); - } - break; - } - - default: - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'")); - } - } - - return *ptr; - } - - /*! - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.402 if the array index '-' is used - @throw out_of_range.404 if the JSON pointer can not be resolved - */ - BasicJsonType& get_checked(BasicJsonType* ptr) const - { - for (const auto& reference_token : reference_tokens) - { - switch (ptr->type()) - { - case detail::value_t::object: - { - // note: at performs range check - ptr = &ptr->at(reference_token); - break; - } - - case detail::value_t::array: - { - if (JSON_HEDLEY_UNLIKELY(reference_token == "-")) - { - // "-" always fails the range check - JSON_THROW(detail::out_of_range::create(402, - "array index '-' (" + std::to_string(ptr->m_value.array->size()) + - ") is out of range")); - } - - // note: at performs range check - ptr = &ptr->at(array_index(reference_token)); - break; - } - - default: - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'")); - } - } - - return *ptr; - } - - /*! - @brief return a const reference to the pointed to value - - @param[in] ptr a JSON value - - @return const reference to the JSON value pointed to by the JSON - pointer - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.402 if the array index '-' is used - @throw out_of_range.404 if the JSON pointer can not be resolved - */ - const BasicJsonType& get_unchecked(const BasicJsonType* ptr) const - { - for (const auto& reference_token : reference_tokens) - { - switch (ptr->type()) - { - case detail::value_t::object: - { - // use unchecked object access - ptr = &ptr->operator[](reference_token); - break; - } - - case detail::value_t::array: - { - if (JSON_HEDLEY_UNLIKELY(reference_token == "-")) - { - // "-" cannot be used for const access - JSON_THROW(detail::out_of_range::create(402, - "array index '-' (" + std::to_string(ptr->m_value.array->size()) + - ") is out of range")); - } - - // use unchecked array access - ptr = &ptr->operator[](array_index(reference_token)); - break; - } - - default: - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'")); - } - } - - return *ptr; - } - - /*! - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.402 if the array index '-' is used - @throw out_of_range.404 if the JSON pointer can not be resolved - */ - const BasicJsonType& get_checked(const BasicJsonType* ptr) const - { - for (const auto& reference_token : reference_tokens) - { - switch (ptr->type()) - { - case detail::value_t::object: - { - // note: at performs range check - ptr = &ptr->at(reference_token); - break; - } - - case detail::value_t::array: - { - if (JSON_HEDLEY_UNLIKELY(reference_token == "-")) - { - // "-" always fails the range check - JSON_THROW(detail::out_of_range::create(402, - "array index '-' (" + std::to_string(ptr->m_value.array->size()) + - ") is out of range")); - } - - // note: at performs range check - ptr = &ptr->at(array_index(reference_token)); - break; - } - - default: - JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'")); - } - } - - return *ptr; - } - - /*! - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - */ - bool contains(const BasicJsonType* ptr) const - { - for (const auto& reference_token : reference_tokens) - { - switch (ptr->type()) - { - case detail::value_t::object: - { - if (!ptr->contains(reference_token)) - { - // we did not find the key in the object - return false; - } - - ptr = &ptr->operator[](reference_token); - break; - } - - case detail::value_t::array: - { - if (JSON_HEDLEY_UNLIKELY(reference_token == "-")) - { - // "-" always fails the range check - return false; - } - if (JSON_HEDLEY_UNLIKELY(reference_token.size() == 1 && !("0" <= reference_token && reference_token <= "9"))) - { - // invalid char - return false; - } - if (JSON_HEDLEY_UNLIKELY(reference_token.size() > 1)) - { - if (JSON_HEDLEY_UNLIKELY(!('1' <= reference_token[0] && reference_token[0] <= '9'))) - { - // first char should be between '1' and '9' - return false; - } - for (std::size_t i = 1; i < reference_token.size(); i++) - { - if (JSON_HEDLEY_UNLIKELY(!('0' <= reference_token[i] && reference_token[i] <= '9'))) - { - // other char should be between '0' and '9' - return false; - } - } - } - - const auto idx = array_index(reference_token); - if (idx >= ptr->size()) - { - // index out of range - return false; - } - - ptr = &ptr->operator[](idx); - break; - } - - default: - { - // we do not expect primitive values if there is still a - // reference token to process - return false; - } - } - } - - // no reference token left means we found a primitive value - return true; - } - - /*! - @brief split the string input to reference tokens - - @note This function is only called by the json_pointer constructor. - All exceptions below are documented there. - - @throw parse_error.107 if the pointer is not empty or begins with '/' - @throw parse_error.108 if character '~' is not followed by '0' or '1' - */ - static std::vector split(const std::string& reference_string) - { - std::vector result; - - // special case: empty reference string -> no reference tokens - if (reference_string.empty()) - { - return result; - } - - // check if nonempty reference string begins with slash - if (JSON_HEDLEY_UNLIKELY(reference_string[0] != '/')) - { - JSON_THROW(detail::parse_error::create(107, 1, - "JSON pointer must be empty or begin with '/' - was: '" + - reference_string + "'")); - } - - // extract the reference tokens: - // - slash: position of the last read slash (or end of string) - // - start: position after the previous slash - for ( - // search for the first slash after the first character - std::size_t slash = reference_string.find_first_of('/', 1), - // set the beginning of the first reference token - start = 1; - // we can stop if start == 0 (if slash == std::string::npos) - start != 0; - // set the beginning of the next reference token - // (will eventually be 0 if slash == std::string::npos) - start = (slash == std::string::npos) ? 0 : slash + 1, - // find next slash - slash = reference_string.find_first_of('/', start)) - { - // use the text between the beginning of the reference token - // (start) and the last slash (slash). - auto reference_token = reference_string.substr(start, slash - start); - - // check reference tokens are properly escaped - for (std::size_t pos = reference_token.find_first_of('~'); - pos != std::string::npos; - pos = reference_token.find_first_of('~', pos + 1)) - { - JSON_ASSERT(reference_token[pos] == '~'); - - // ~ must be followed by 0 or 1 - if (JSON_HEDLEY_UNLIKELY(pos == reference_token.size() - 1 || - (reference_token[pos + 1] != '0' && - reference_token[pos + 1] != '1'))) - { - JSON_THROW(detail::parse_error::create(108, 0, "escape character '~' must be followed with '0' or '1'")); - } - } - - // finally, store the reference token - unescape(reference_token); - result.push_back(reference_token); - } - - return result; - } - - /*! - @brief replace all occurrences of a substring by another string - - @param[in,out] s the string to manipulate; changed so that all - occurrences of @a f are replaced with @a t - @param[in] f the substring to replace with @a t - @param[in] t the string to replace @a f - - @pre The search string @a f must not be empty. **This precondition is - enforced with an assertion.** - - @since version 2.0.0 - */ - static void replace_substring(std::string& s, const std::string& f, - const std::string& t) - { - JSON_ASSERT(!f.empty()); - for (auto pos = s.find(f); // find first occurrence of f - pos != std::string::npos; // make sure f was found - s.replace(pos, f.size(), t), // replace with t, and - pos = s.find(f, pos + t.size())) // find next occurrence of f - {} - } - - /// escape "~" to "~0" and "/" to "~1" - static std::string escape(std::string s) - { - replace_substring(s, "~", "~0"); - replace_substring(s, "/", "~1"); - return s; - } - - /// unescape "~1" to tilde and "~0" to slash (order is important!) - static void unescape(std::string& s) - { - replace_substring(s, "~1", "/"); - replace_substring(s, "~0", "~"); - } - - /*! - @param[in] reference_string the reference string to the current value - @param[in] value the value to consider - @param[in,out] result the result object to insert values to - - @note Empty objects or arrays are flattened to `null`. - */ - static void flatten(const std::string& reference_string, - const BasicJsonType& value, - BasicJsonType& result) - { - switch (value.type()) - { - case detail::value_t::array: - { - if (value.m_value.array->empty()) - { - // flatten empty array as null - result[reference_string] = nullptr; - } - else - { - // iterate array and use index as reference string - for (std::size_t i = 0; i < value.m_value.array->size(); ++i) - { - flatten(reference_string + "/" + std::to_string(i), - value.m_value.array->operator[](i), result); - } - } - break; - } - - case detail::value_t::object: - { - if (value.m_value.object->empty()) - { - // flatten empty object as null - result[reference_string] = nullptr; - } - else - { - // iterate object and use keys as reference string - for (const auto& element : *value.m_value.object) - { - flatten(reference_string + "/" + escape(element.first), element.second, result); - } - } - break; - } - - default: - { - // add primitive value with its reference string - result[reference_string] = value; - break; - } - } - } - - /*! - @param[in] value flattened JSON - - @return unflattened JSON - - @throw parse_error.109 if array index is not a number - @throw type_error.314 if value is not an object - @throw type_error.315 if object values are not primitive - @throw type_error.313 if value cannot be unflattened - */ - static BasicJsonType - unflatten(const BasicJsonType& value) - { - if (JSON_HEDLEY_UNLIKELY(!value.is_object())) - { - JSON_THROW(detail::type_error::create(314, "only objects can be unflattened")); - } - - BasicJsonType result; - - // iterate the JSON object values - for (const auto& element : *value.m_value.object) - { - if (JSON_HEDLEY_UNLIKELY(!element.second.is_primitive())) - { - JSON_THROW(detail::type_error::create(315, "values in object must be primitive")); - } - - // assign value to reference pointed to by JSON pointer; Note that if - // the JSON pointer is "" (i.e., points to the whole value), function - // get_and_create returns a reference to result itself. An assignment - // will then create a primitive value. - json_pointer(element.first).get_and_create(result) = element.second; - } - - return result; - } - - /*! - @brief compares two JSON pointers for equality - - @param[in] lhs JSON pointer to compare - @param[in] rhs JSON pointer to compare - @return whether @a lhs is equal to @a rhs - - @complexity Linear in the length of the JSON pointer - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - */ - friend bool operator==(json_pointer const& lhs, - json_pointer const& rhs) noexcept - { - return lhs.reference_tokens == rhs.reference_tokens; - } - - /*! - @brief compares two JSON pointers for inequality - - @param[in] lhs JSON pointer to compare - @param[in] rhs JSON pointer to compare - @return whether @a lhs is not equal @a rhs - - @complexity Linear in the length of the JSON pointer - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - */ - friend bool operator!=(json_pointer const& lhs, - json_pointer const& rhs) noexcept - { - return !(lhs == rhs); - } - - /// the reference tokens - std::vector reference_tokens; -}; -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_ref.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_ref.hpp deleted file mode 100644 index c9bf6cb228c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/json_ref.hpp +++ /dev/null @@ -1,76 +0,0 @@ -#pragma once - -#include -#include - -#include - -namespace nlohmann -{ -namespace detail -{ -template -class json_ref -{ - public: - using value_type = BasicJsonType; - - json_ref(value_type&& value) - : owned_value(std::move(value)) - , value_ref(&owned_value) - , is_rvalue(true) - {} - - json_ref(const value_type& value) - : value_ref(const_cast(&value)) - , is_rvalue(false) - {} - - json_ref(std::initializer_list init) - : owned_value(init) - , value_ref(&owned_value) - , is_rvalue(true) - {} - - template < - class... Args, - enable_if_t::value, int> = 0 > - json_ref(Args && ... args) - : owned_value(std::forward(args)...) - , value_ref(&owned_value) - , is_rvalue(true) - {} - - // class should be movable only - json_ref(json_ref&&) = default; - json_ref(const json_ref&) = delete; - json_ref& operator=(const json_ref&) = delete; - json_ref& operator=(json_ref&&) = delete; - ~json_ref() = default; - - value_type moved_or_copied() const - { - if (is_rvalue) - { - return std::move(*value_ref); - } - return *value_ref; - } - - value_type const& operator*() const - { - return *static_cast(value_ref); - } - - value_type const* operator->() const - { - return static_cast(value_ref); - } - - private: - mutable value_type owned_value = nullptr; - value_type* value_ref = nullptr; - const bool is_rvalue = true; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_scope.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_scope.hpp deleted file mode 100644 index 8c9f63296bc..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_scope.hpp +++ /dev/null @@ -1,296 +0,0 @@ -#pragma once - -#include // pair -#include - -// This file contains all internal macro definitions -// You MUST include macro_unscope.hpp at the end of json.hpp to undef all of them - -// exclude unsupported compilers -#if !defined(JSON_SKIP_UNSUPPORTED_COMPILER_CHECK) - #if defined(__clang__) - #if (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__) < 30400 - #error "unsupported Clang version - see https://github.com/nlohmann/json#supported-compilers" - #endif - #elif defined(__GNUC__) && !(defined(__ICC) || defined(__INTEL_COMPILER)) - #if (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) < 40800 - #error "unsupported GCC version - see https://github.com/nlohmann/json#supported-compilers" - #endif - #endif -#endif - -// C++ language standard detection -#if (defined(__cplusplus) && __cplusplus >= 202002L) || (defined(_MSVC_LANG) && _MSVC_LANG >= 202002L) - #define JSON_HAS_CPP_20 - #define JSON_HAS_CPP_17 - #define JSON_HAS_CPP_14 -#elif (defined(__cplusplus) && __cplusplus >= 201703L) || (defined(_HAS_CXX17) && _HAS_CXX17 == 1) // fix for issue #464 - #define JSON_HAS_CPP_17 - #define JSON_HAS_CPP_14 -#elif (defined(__cplusplus) && __cplusplus >= 201402L) || (defined(_HAS_CXX14) && _HAS_CXX14 == 1) - #define JSON_HAS_CPP_14 -#endif - -// disable float-equal warnings on GCC/clang -#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__) - #pragma GCC diagnostic push - #pragma GCC diagnostic ignored "-Wfloat-equal" -#endif - -// disable documentation warnings on clang -#if defined(__clang__) - #pragma GCC diagnostic push - #pragma GCC diagnostic ignored "-Wdocumentation" -#endif - -// allow to disable exceptions -#if (defined(__cpp_exceptions) || defined(__EXCEPTIONS) || defined(_CPPUNWIND)) && !defined(JSON_NOEXCEPTION) - #define JSON_THROW(exception) throw exception - #define JSON_TRY try - #define JSON_CATCH(exception) catch(exception) - #define JSON_INTERNAL_CATCH(exception) catch(exception) -#else - #include - #define JSON_THROW(exception) std::abort() - #define JSON_TRY if(true) - #define JSON_CATCH(exception) if(false) - #define JSON_INTERNAL_CATCH(exception) if(false) -#endif - -// override exception macros -#if defined(JSON_THROW_USER) - #undef JSON_THROW - #define JSON_THROW JSON_THROW_USER -#endif -#if defined(JSON_TRY_USER) - #undef JSON_TRY - #define JSON_TRY JSON_TRY_USER -#endif -#if defined(JSON_CATCH_USER) - #undef JSON_CATCH - #define JSON_CATCH JSON_CATCH_USER - #undef JSON_INTERNAL_CATCH - #define JSON_INTERNAL_CATCH JSON_CATCH_USER -#endif -#if defined(JSON_INTERNAL_CATCH_USER) - #undef JSON_INTERNAL_CATCH - #define JSON_INTERNAL_CATCH JSON_INTERNAL_CATCH_USER -#endif - -// allow to override assert -#if !defined(JSON_ASSERT) - #include // assert - #define JSON_ASSERT(x) assert(x) -#endif - -/*! -@brief macro to briefly define a mapping between an enum and JSON -@def NLOHMANN_JSON_SERIALIZE_ENUM -@since version 3.4.0 -*/ -#define NLOHMANN_JSON_SERIALIZE_ENUM(ENUM_TYPE, ...) \ - template \ - inline void to_json(BasicJsonType& j, const ENUM_TYPE& e) \ - { \ - static_assert(std::is_enum::value, #ENUM_TYPE " must be an enum!"); \ - static const std::pair m[] = __VA_ARGS__; \ - auto it = std::find_if(std::begin(m), std::end(m), \ - [e](const std::pair& ej_pair) -> bool \ - { \ - return ej_pair.first == e; \ - }); \ - j = ((it != std::end(m)) ? it : std::begin(m))->second; \ - } \ - template \ - inline void from_json(const BasicJsonType& j, ENUM_TYPE& e) \ - { \ - static_assert(std::is_enum::value, #ENUM_TYPE " must be an enum!"); \ - static const std::pair m[] = __VA_ARGS__; \ - auto it = std::find_if(std::begin(m), std::end(m), \ - [&j](const std::pair& ej_pair) -> bool \ - { \ - return ej_pair.second == j; \ - }); \ - e = ((it != std::end(m)) ? it : std::begin(m))->first; \ - } - -// Ugly macros to avoid uglier copy-paste when specializing basic_json. They -// may be removed in the future once the class is split. - -#define NLOHMANN_BASIC_JSON_TPL_DECLARATION \ - template class ObjectType, \ - template class ArrayType, \ - class StringType, class BooleanType, class NumberIntegerType, \ - class NumberUnsignedType, class NumberFloatType, \ - template class AllocatorType, \ - template class JSONSerializer, \ - class BinaryType> - -#define NLOHMANN_BASIC_JSON_TPL \ - basic_json - -// Macros to simplify conversion from/to types - -#define NLOHMANN_JSON_EXPAND( x ) x -#define NLOHMANN_JSON_GET_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15, _16, _17, _18, _19, _20, _21, _22, _23, _24, _25, _26, _27, _28, _29, _30, _31, _32, _33, _34, _35, _36, _37, _38, _39, _40, _41, _42, _43, _44, _45, _46, _47, _48, _49, _50, _51, _52, _53, _54, _55, _56, _57, _58, _59, _60, _61, _62, _63, _64, NAME,...) NAME -#define NLOHMANN_JSON_PASTE(...) NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_GET_MACRO(__VA_ARGS__, \ - NLOHMANN_JSON_PASTE64, \ - NLOHMANN_JSON_PASTE63, \ - NLOHMANN_JSON_PASTE62, \ - NLOHMANN_JSON_PASTE61, \ - NLOHMANN_JSON_PASTE60, \ - NLOHMANN_JSON_PASTE59, \ - NLOHMANN_JSON_PASTE58, \ - NLOHMANN_JSON_PASTE57, \ - NLOHMANN_JSON_PASTE56, \ - NLOHMANN_JSON_PASTE55, \ - NLOHMANN_JSON_PASTE54, \ - NLOHMANN_JSON_PASTE53, \ - NLOHMANN_JSON_PASTE52, \ - NLOHMANN_JSON_PASTE51, \ - NLOHMANN_JSON_PASTE50, \ - NLOHMANN_JSON_PASTE49, \ - NLOHMANN_JSON_PASTE48, \ - NLOHMANN_JSON_PASTE47, \ - NLOHMANN_JSON_PASTE46, \ - NLOHMANN_JSON_PASTE45, \ - NLOHMANN_JSON_PASTE44, \ - NLOHMANN_JSON_PASTE43, \ - NLOHMANN_JSON_PASTE42, \ - NLOHMANN_JSON_PASTE41, \ - NLOHMANN_JSON_PASTE40, \ - NLOHMANN_JSON_PASTE39, \ - NLOHMANN_JSON_PASTE38, \ - NLOHMANN_JSON_PASTE37, \ - NLOHMANN_JSON_PASTE36, \ - NLOHMANN_JSON_PASTE35, \ - NLOHMANN_JSON_PASTE34, \ - NLOHMANN_JSON_PASTE33, \ - NLOHMANN_JSON_PASTE32, \ - NLOHMANN_JSON_PASTE31, \ - NLOHMANN_JSON_PASTE30, \ - NLOHMANN_JSON_PASTE29, \ - NLOHMANN_JSON_PASTE28, \ - NLOHMANN_JSON_PASTE27, \ - NLOHMANN_JSON_PASTE26, \ - NLOHMANN_JSON_PASTE25, \ - NLOHMANN_JSON_PASTE24, \ - NLOHMANN_JSON_PASTE23, \ - NLOHMANN_JSON_PASTE22, \ - NLOHMANN_JSON_PASTE21, \ - NLOHMANN_JSON_PASTE20, \ - NLOHMANN_JSON_PASTE19, \ - NLOHMANN_JSON_PASTE18, \ - NLOHMANN_JSON_PASTE17, \ - NLOHMANN_JSON_PASTE16, \ - NLOHMANN_JSON_PASTE15, \ - NLOHMANN_JSON_PASTE14, \ - NLOHMANN_JSON_PASTE13, \ - NLOHMANN_JSON_PASTE12, \ - NLOHMANN_JSON_PASTE11, \ - NLOHMANN_JSON_PASTE10, \ - NLOHMANN_JSON_PASTE9, \ - NLOHMANN_JSON_PASTE8, \ - NLOHMANN_JSON_PASTE7, \ - NLOHMANN_JSON_PASTE6, \ - NLOHMANN_JSON_PASTE5, \ - NLOHMANN_JSON_PASTE4, \ - NLOHMANN_JSON_PASTE3, \ - NLOHMANN_JSON_PASTE2, \ - NLOHMANN_JSON_PASTE1)(__VA_ARGS__)) -#define NLOHMANN_JSON_PASTE2(func, v1) func(v1) -#define NLOHMANN_JSON_PASTE3(func, v1, v2) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE2(func, v2) -#define NLOHMANN_JSON_PASTE4(func, v1, v2, v3) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE3(func, v2, v3) -#define NLOHMANN_JSON_PASTE5(func, v1, v2, v3, v4) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE4(func, v2, v3, v4) -#define NLOHMANN_JSON_PASTE6(func, v1, v2, v3, v4, v5) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE5(func, v2, v3, v4, v5) -#define NLOHMANN_JSON_PASTE7(func, v1, v2, v3, v4, v5, v6) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE6(func, v2, v3, v4, v5, v6) -#define NLOHMANN_JSON_PASTE8(func, v1, v2, v3, v4, v5, v6, v7) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE7(func, v2, v3, v4, v5, v6, v7) -#define NLOHMANN_JSON_PASTE9(func, v1, v2, v3, v4, v5, v6, v7, v8) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE8(func, v2, v3, v4, v5, v6, v7, v8) -#define NLOHMANN_JSON_PASTE10(func, v1, v2, v3, v4, v5, v6, v7, v8, v9) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE9(func, v2, v3, v4, v5, v6, v7, v8, v9) -#define NLOHMANN_JSON_PASTE11(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE10(func, v2, v3, v4, v5, v6, v7, v8, v9, v10) -#define NLOHMANN_JSON_PASTE12(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE11(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11) -#define NLOHMANN_JSON_PASTE13(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE12(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12) -#define NLOHMANN_JSON_PASTE14(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE13(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13) -#define NLOHMANN_JSON_PASTE15(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE14(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14) -#define NLOHMANN_JSON_PASTE16(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE15(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15) -#define NLOHMANN_JSON_PASTE17(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE16(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16) -#define NLOHMANN_JSON_PASTE18(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE17(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17) -#define NLOHMANN_JSON_PASTE19(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE18(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18) -#define NLOHMANN_JSON_PASTE20(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE19(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19) -#define NLOHMANN_JSON_PASTE21(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE20(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20) -#define NLOHMANN_JSON_PASTE22(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE21(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21) -#define NLOHMANN_JSON_PASTE23(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE22(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22) -#define NLOHMANN_JSON_PASTE24(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE23(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23) -#define NLOHMANN_JSON_PASTE25(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE24(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24) -#define NLOHMANN_JSON_PASTE26(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE25(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25) -#define NLOHMANN_JSON_PASTE27(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE26(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26) -#define NLOHMANN_JSON_PASTE28(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE27(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27) -#define NLOHMANN_JSON_PASTE29(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE28(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28) -#define NLOHMANN_JSON_PASTE30(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE29(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29) -#define NLOHMANN_JSON_PASTE31(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE30(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30) -#define NLOHMANN_JSON_PASTE32(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE31(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31) -#define NLOHMANN_JSON_PASTE33(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE32(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32) -#define NLOHMANN_JSON_PASTE34(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE33(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33) -#define NLOHMANN_JSON_PASTE35(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE34(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34) -#define NLOHMANN_JSON_PASTE36(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE35(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35) -#define NLOHMANN_JSON_PASTE37(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE36(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36) -#define NLOHMANN_JSON_PASTE38(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE37(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37) -#define NLOHMANN_JSON_PASTE39(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE38(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38) -#define NLOHMANN_JSON_PASTE40(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE39(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39) -#define NLOHMANN_JSON_PASTE41(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE40(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40) -#define NLOHMANN_JSON_PASTE42(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE41(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41) -#define NLOHMANN_JSON_PASTE43(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE42(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42) -#define NLOHMANN_JSON_PASTE44(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE43(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43) -#define NLOHMANN_JSON_PASTE45(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE44(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44) -#define NLOHMANN_JSON_PASTE46(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE45(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45) -#define NLOHMANN_JSON_PASTE47(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE46(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46) -#define NLOHMANN_JSON_PASTE48(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE47(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47) -#define NLOHMANN_JSON_PASTE49(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE48(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48) -#define NLOHMANN_JSON_PASTE50(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE49(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49) -#define NLOHMANN_JSON_PASTE51(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE50(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50) -#define NLOHMANN_JSON_PASTE52(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE51(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51) -#define NLOHMANN_JSON_PASTE53(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE52(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52) -#define NLOHMANN_JSON_PASTE54(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE53(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53) -#define NLOHMANN_JSON_PASTE55(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE54(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54) -#define NLOHMANN_JSON_PASTE56(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE55(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55) -#define NLOHMANN_JSON_PASTE57(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE56(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56) -#define NLOHMANN_JSON_PASTE58(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE57(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57) -#define NLOHMANN_JSON_PASTE59(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE58(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58) -#define NLOHMANN_JSON_PASTE60(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE59(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59) -#define NLOHMANN_JSON_PASTE61(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE60(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60) -#define NLOHMANN_JSON_PASTE62(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE61(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61) -#define NLOHMANN_JSON_PASTE63(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE62(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62) -#define NLOHMANN_JSON_PASTE64(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE63(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63) - -#define NLOHMANN_JSON_TO(v1) nlohmann_json_j[#v1] = nlohmann_json_t.v1; -#define NLOHMANN_JSON_FROM(v1) nlohmann_json_j.at(#v1).get_to(nlohmann_json_t.v1); - -/*! -@brief macro -@def NLOHMANN_DEFINE_TYPE_INTRUSIVE -@since version 3.9.0 -*/ -#define NLOHMANN_DEFINE_TYPE_INTRUSIVE(Type, ...) \ - friend void to_json(nlohmann::json& nlohmann_json_j, const Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_TO, __VA_ARGS__)) } \ - friend void from_json(const nlohmann::json& nlohmann_json_j, Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_FROM, __VA_ARGS__)) } - -/*! -@brief macro -@def NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE -@since version 3.9.0 -*/ -#define NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(Type, ...) \ - inline void to_json(nlohmann::json& nlohmann_json_j, const Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_TO, __VA_ARGS__)) } \ - inline void from_json(const nlohmann::json& nlohmann_json_j, Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_FROM, __VA_ARGS__)) } - -#ifndef JSON_USE_IMPLICIT_CONVERSIONS - #define JSON_USE_IMPLICIT_CONVERSIONS 1 -#endif - -#if JSON_USE_IMPLICIT_CONVERSIONS - #define JSON_EXPLICIT -#else - #define JSON_EXPLICIT explicit -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_unscope.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_unscope.hpp deleted file mode 100644 index eb7065113db..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/macro_unscope.hpp +++ /dev/null @@ -1,23 +0,0 @@ -#pragma once - -// restore GCC/clang diagnostic settings -#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__) - #pragma GCC diagnostic pop -#endif -#if defined(__clang__) - #pragma GCC diagnostic pop -#endif - -// clean up -#undef JSON_ASSERT -#undef JSON_INTERNAL_CATCH -#undef JSON_CATCH -#undef JSON_THROW -#undef JSON_TRY -#undef JSON_HAS_CPP_14 -#undef JSON_HAS_CPP_17 -#undef NLOHMANN_BASIC_JSON_TPL_DECLARATION -#undef NLOHMANN_BASIC_JSON_TPL -#undef JSON_EXPLICIT - -#include diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/cpp_future.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/cpp_future.hpp deleted file mode 100644 index dd929ee1420..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/cpp_future.hpp +++ /dev/null @@ -1,62 +0,0 @@ -#pragma once - -#include // size_t -#include // conditional, enable_if, false_type, integral_constant, is_constructible, is_integral, is_same, remove_cv, remove_reference, true_type - -namespace nlohmann -{ -namespace detail -{ -// alias templates to reduce boilerplate -template -using enable_if_t = typename std::enable_if::type; - -template -using uncvref_t = typename std::remove_cv::type>::type; - -// implementation of C++14 index_sequence and affiliates -// source: https://stackoverflow.com/a/32223343 -template -struct index_sequence -{ - using type = index_sequence; - using value_type = std::size_t; - static constexpr std::size_t size() noexcept - { - return sizeof...(Ints); - } -}; - -template -struct merge_and_renumber; - -template -struct merge_and_renumber, index_sequence> - : index_sequence < I1..., (sizeof...(I1) + I2)... > {}; - -template -struct make_index_sequence - : merge_and_renumber < typename make_index_sequence < N / 2 >::type, - typename make_index_sequence < N - N / 2 >::type > {}; - -template<> struct make_index_sequence<0> : index_sequence<> {}; -template<> struct make_index_sequence<1> : index_sequence<0> {}; - -template -using index_sequence_for = make_index_sequence; - -// dispatch utility (taken from ranges-v3) -template struct priority_tag : priority_tag < N - 1 > {}; -template<> struct priority_tag<0> {}; - -// taken from ranges-v3 -template -struct static_const -{ - static constexpr T value{}; -}; - -template -constexpr T static_const::value; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/detected.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/detected.hpp deleted file mode 100644 index 7b5a0035370..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/detected.hpp +++ /dev/null @@ -1,58 +0,0 @@ -#pragma once - -#include - -#include - -// https://en.cppreference.com/w/cpp/experimental/is_detected -namespace nlohmann -{ -namespace detail -{ -struct nonesuch -{ - nonesuch() = delete; - ~nonesuch() = delete; - nonesuch(nonesuch const&) = delete; - nonesuch(nonesuch const&&) = delete; - void operator=(nonesuch const&) = delete; - void operator=(nonesuch&&) = delete; -}; - -template class Op, - class... Args> -struct detector -{ - using value_t = std::false_type; - using type = Default; -}; - -template class Op, class... Args> -struct detector>, Op, Args...> -{ - using value_t = std::true_type; - using type = Op; -}; - -template class Op, class... Args> -using is_detected = typename detector::value_t; - -template class Op, class... Args> -using detected_t = typename detector::type; - -template class Op, class... Args> -using detected_or = detector; - -template class Op, class... Args> -using detected_or_t = typename detected_or::type; - -template class Op, class... Args> -using is_detected_exact = std::is_same>; - -template class Op, class... Args> -using is_detected_convertible = - std::is_convertible, To>; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/is_sax.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/is_sax.hpp deleted file mode 100644 index e1e48a0bae1..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/is_sax.hpp +++ /dev/null @@ -1,149 +0,0 @@ -#pragma once - -#include // size_t -#include // declval -#include // string - -#include -#include - -namespace nlohmann -{ -namespace detail -{ -template -using null_function_t = decltype(std::declval().null()); - -template -using boolean_function_t = - decltype(std::declval().boolean(std::declval())); - -template -using number_integer_function_t = - decltype(std::declval().number_integer(std::declval())); - -template -using number_unsigned_function_t = - decltype(std::declval().number_unsigned(std::declval())); - -template -using number_float_function_t = decltype(std::declval().number_float( - std::declval(), std::declval())); - -template -using string_function_t = - decltype(std::declval().string(std::declval())); - -template -using binary_function_t = - decltype(std::declval().binary(std::declval())); - -template -using start_object_function_t = - decltype(std::declval().start_object(std::declval())); - -template -using key_function_t = - decltype(std::declval().key(std::declval())); - -template -using end_object_function_t = decltype(std::declval().end_object()); - -template -using start_array_function_t = - decltype(std::declval().start_array(std::declval())); - -template -using end_array_function_t = decltype(std::declval().end_array()); - -template -using parse_error_function_t = decltype(std::declval().parse_error( - std::declval(), std::declval(), - std::declval())); - -template -struct is_sax -{ - private: - static_assert(is_basic_json::value, - "BasicJsonType must be of type basic_json<...>"); - - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - using exception_t = typename BasicJsonType::exception; - - public: - static constexpr bool value = - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value && - is_detected_exact::value; -}; - -template -struct is_sax_static_asserts -{ - private: - static_assert(is_basic_json::value, - "BasicJsonType must be of type basic_json<...>"); - - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using number_float_t = typename BasicJsonType::number_float_t; - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - using exception_t = typename BasicJsonType::exception; - - public: - static_assert(is_detected_exact::value, - "Missing/invalid function: bool null()"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool boolean(bool)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool boolean(bool)"); - static_assert( - is_detected_exact::value, - "Missing/invalid function: bool number_integer(number_integer_t)"); - static_assert( - is_detected_exact::value, - "Missing/invalid function: bool number_unsigned(number_unsigned_t)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool number_float(number_float_t, const string_t&)"); - static_assert( - is_detected_exact::value, - "Missing/invalid function: bool string(string_t&)"); - static_assert( - is_detected_exact::value, - "Missing/invalid function: bool binary(binary_t&)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool start_object(std::size_t)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool key(string_t&)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool end_object()"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool start_array(std::size_t)"); - static_assert(is_detected_exact::value, - "Missing/invalid function: bool end_array()"); - static_assert( - is_detected_exact::value, - "Missing/invalid function: bool parse_error(std::size_t, const " - "std::string&, const exception&)"); -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/type_traits.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/type_traits.hpp deleted file mode 100644 index ac143becf58..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/type_traits.hpp +++ /dev/null @@ -1,396 +0,0 @@ -#pragma once - -#include // numeric_limits -#include // false_type, is_constructible, is_integral, is_same, true_type -#include // declval - -#include -#include -#include -#include -#include - -namespace nlohmann -{ -/*! -@brief detail namespace with internal helper functions - -This namespace collects functions that should not be exposed, -implementations of some @ref basic_json methods, and meta-programming helpers. - -@since version 2.1.0 -*/ -namespace detail -{ -///////////// -// helpers // -///////////// - -// Note to maintainers: -// -// Every trait in this file expects a non CV-qualified type. -// The only exceptions are in the 'aliases for detected' section -// (i.e. those of the form: decltype(T::member_function(std::declval()))) -// -// In this case, T has to be properly CV-qualified to constraint the function arguments -// (e.g. to_json(BasicJsonType&, const T&)) - -template struct is_basic_json : std::false_type {}; - -NLOHMANN_BASIC_JSON_TPL_DECLARATION -struct is_basic_json : std::true_type {}; - -////////////////////// -// json_ref helpers // -////////////////////// - -template -class json_ref; - -template -struct is_json_ref : std::false_type {}; - -template -struct is_json_ref> : std::true_type {}; - -////////////////////////// -// aliases for detected // -////////////////////////// - -template -using mapped_type_t = typename T::mapped_type; - -template -using key_type_t = typename T::key_type; - -template -using value_type_t = typename T::value_type; - -template -using difference_type_t = typename T::difference_type; - -template -using pointer_t = typename T::pointer; - -template -using reference_t = typename T::reference; - -template -using iterator_category_t = typename T::iterator_category; - -template -using iterator_t = typename T::iterator; - -template -using to_json_function = decltype(T::to_json(std::declval()...)); - -template -using from_json_function = decltype(T::from_json(std::declval()...)); - -template -using get_template_function = decltype(std::declval().template get()); - -// trait checking if JSONSerializer::from_json(json const&, udt&) exists -template -struct has_from_json : std::false_type {}; - -// trait checking if j.get is valid -// use this trait instead of std::is_constructible or std::is_convertible, -// both rely on, or make use of implicit conversions, and thus fail when T -// has several constructors/operator= (see https://github.com/nlohmann/json/issues/958) -template -struct is_getable -{ - static constexpr bool value = is_detected::value; -}; - -template -struct has_from_json < BasicJsonType, T, - enable_if_t < !is_basic_json::value >> -{ - using serializer = typename BasicJsonType::template json_serializer; - - static constexpr bool value = - is_detected_exact::value; -}; - -// This trait checks if JSONSerializer::from_json(json const&) exists -// this overload is used for non-default-constructible user-defined-types -template -struct has_non_default_from_json : std::false_type {}; - -template -struct has_non_default_from_json < BasicJsonType, T, enable_if_t < !is_basic_json::value >> -{ - using serializer = typename BasicJsonType::template json_serializer; - - static constexpr bool value = - is_detected_exact::value; -}; - -// This trait checks if BasicJsonType::json_serializer::to_json exists -// Do not evaluate the trait when T is a basic_json type, to avoid template instantiation infinite recursion. -template -struct has_to_json : std::false_type {}; - -template -struct has_to_json < BasicJsonType, T, enable_if_t < !is_basic_json::value >> -{ - using serializer = typename BasicJsonType::template json_serializer; - - static constexpr bool value = - is_detected_exact::value; -}; - - -/////////////////// -// is_ functions // -/////////////////// - -template -struct is_iterator_traits : std::false_type {}; - -template -struct is_iterator_traits> -{ - private: - using traits = iterator_traits; - - public: - static constexpr auto value = - is_detected::value && - is_detected::value && - is_detected::value && - is_detected::value && - is_detected::value; -}; - -// source: https://stackoverflow.com/a/37193089/4116453 - -template -struct is_complete_type : std::false_type {}; - -template -struct is_complete_type : std::true_type {}; - -template -struct is_compatible_object_type_impl : std::false_type {}; - -template -struct is_compatible_object_type_impl < - BasicJsonType, CompatibleObjectType, - enable_if_t < is_detected::value&& - is_detected::value >> -{ - - using object_t = typename BasicJsonType::object_t; - - // macOS's is_constructible does not play well with nonesuch... - static constexpr bool value = - std::is_constructible::value && - std::is_constructible::value; -}; - -template -struct is_compatible_object_type - : is_compatible_object_type_impl {}; - -template -struct is_constructible_object_type_impl : std::false_type {}; - -template -struct is_constructible_object_type_impl < - BasicJsonType, ConstructibleObjectType, - enable_if_t < is_detected::value&& - is_detected::value >> -{ - using object_t = typename BasicJsonType::object_t; - - static constexpr bool value = - (std::is_default_constructible::value && - (std::is_move_assignable::value || - std::is_copy_assignable::value) && - (std::is_constructible::value && - std::is_same < - typename object_t::mapped_type, - typename ConstructibleObjectType::mapped_type >::value)) || - (has_from_json::value || - has_non_default_from_json < - BasicJsonType, - typename ConstructibleObjectType::mapped_type >::value); -}; - -template -struct is_constructible_object_type - : is_constructible_object_type_impl {}; - -template -struct is_compatible_string_type_impl : std::false_type {}; - -template -struct is_compatible_string_type_impl < - BasicJsonType, CompatibleStringType, - enable_if_t::value >> -{ - static constexpr auto value = - std::is_constructible::value; -}; - -template -struct is_compatible_string_type - : is_compatible_string_type_impl {}; - -template -struct is_constructible_string_type_impl : std::false_type {}; - -template -struct is_constructible_string_type_impl < - BasicJsonType, ConstructibleStringType, - enable_if_t::value >> -{ - static constexpr auto value = - std::is_constructible::value; -}; - -template -struct is_constructible_string_type - : is_constructible_string_type_impl {}; - -template -struct is_compatible_array_type_impl : std::false_type {}; - -template -struct is_compatible_array_type_impl < - BasicJsonType, CompatibleArrayType, - enable_if_t < is_detected::value&& - is_detected::value&& -// This is needed because json_reverse_iterator has a ::iterator type... -// Therefore it is detected as a CompatibleArrayType. -// The real fix would be to have an Iterable concept. - !is_iterator_traits < - iterator_traits>::value >> -{ - static constexpr bool value = - std::is_constructible::value; -}; - -template -struct is_compatible_array_type - : is_compatible_array_type_impl {}; - -template -struct is_constructible_array_type_impl : std::false_type {}; - -template -struct is_constructible_array_type_impl < - BasicJsonType, ConstructibleArrayType, - enable_if_t::value >> - : std::true_type {}; - -template -struct is_constructible_array_type_impl < - BasicJsonType, ConstructibleArrayType, - enable_if_t < !std::is_same::value&& - std::is_default_constructible::value&& -(std::is_move_assignable::value || - std::is_copy_assignable::value)&& -is_detected::value&& -is_detected::value&& -is_complete_type < -detected_t>::value >> -{ - static constexpr bool value = - // This is needed because json_reverse_iterator has a ::iterator type, - // furthermore, std::back_insert_iterator (and other iterators) have a - // base class `iterator`... Therefore it is detected as a - // ConstructibleArrayType. The real fix would be to have an Iterable - // concept. - !is_iterator_traits>::value && - - (std::is_same::value || - has_from_json::value || - has_non_default_from_json < - BasicJsonType, typename ConstructibleArrayType::value_type >::value); -}; - -template -struct is_constructible_array_type - : is_constructible_array_type_impl {}; - -template -struct is_compatible_integer_type_impl : std::false_type {}; - -template -struct is_compatible_integer_type_impl < - RealIntegerType, CompatibleNumberIntegerType, - enable_if_t < std::is_integral::value&& - std::is_integral::value&& - !std::is_same::value >> -{ - // is there an assert somewhere on overflows? - using RealLimits = std::numeric_limits; - using CompatibleLimits = std::numeric_limits; - - static constexpr auto value = - std::is_constructible::value && - CompatibleLimits::is_integer && - RealLimits::is_signed == CompatibleLimits::is_signed; -}; - -template -struct is_compatible_integer_type - : is_compatible_integer_type_impl {}; - -template -struct is_compatible_type_impl: std::false_type {}; - -template -struct is_compatible_type_impl < - BasicJsonType, CompatibleType, - enable_if_t::value >> -{ - static constexpr bool value = - has_to_json::value; -}; - -template -struct is_compatible_type - : is_compatible_type_impl {}; - -// https://en.cppreference.com/w/cpp/types/conjunction -template struct conjunction : std::true_type { }; -template struct conjunction : B1 { }; -template -struct conjunction -: std::conditional, B1>::type {}; - -template -struct is_constructible_tuple : std::false_type {}; - -template -struct is_constructible_tuple> : conjunction...> {}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/void_t.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/void_t.hpp deleted file mode 100644 index 4ee2c86f0d0..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/meta/void_t.hpp +++ /dev/null @@ -1,13 +0,0 @@ -#pragma once - -namespace nlohmann -{ -namespace detail -{ -template struct make_void -{ - using type = void; -}; -template using void_t = typename make_void::type; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/binary_writer.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/binary_writer.hpp deleted file mode 100644 index 0c6185e0483..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/binary_writer.hpp +++ /dev/null @@ -1,1595 +0,0 @@ -#pragma once - -#include // reverse -#include // array -#include // uint8_t, uint16_t, uint32_t, uint64_t -#include // memcpy -#include // numeric_limits -#include // string -#include // isnan, isinf - -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -/////////////////// -// binary writer // -/////////////////// - -/*! -@brief serialization to CBOR and MessagePack values -*/ -template -class binary_writer -{ - using string_t = typename BasicJsonType::string_t; - using binary_t = typename BasicJsonType::binary_t; - using number_float_t = typename BasicJsonType::number_float_t; - - public: - /*! - @brief create a binary writer - - @param[in] adapter output adapter to write to - */ - explicit binary_writer(output_adapter_t adapter) : oa(adapter) - { - JSON_ASSERT(oa); - } - - /*! - @param[in] j JSON value to serialize - @pre j.type() == value_t::object - */ - void write_bson(const BasicJsonType& j) - { - switch (j.type()) - { - case value_t::object: - { - write_bson_object(*j.m_value.object); - break; - } - - default: - { - JSON_THROW(type_error::create(317, "to serialize to BSON, top-level type must be object, but is " + std::string(j.type_name()))); - } - } - } - - /*! - @param[in] j JSON value to serialize - */ - void write_cbor(const BasicJsonType& j) - { - switch (j.type()) - { - case value_t::null: - { - oa->write_character(to_char_type(0xF6)); - break; - } - - case value_t::boolean: - { - oa->write_character(j.m_value.boolean - ? to_char_type(0xF5) - : to_char_type(0xF4)); - break; - } - - case value_t::number_integer: - { - if (j.m_value.number_integer >= 0) - { - // CBOR does not differentiate between positive signed - // integers and unsigned integers. Therefore, we used the - // code from the value_t::number_unsigned case here. - if (j.m_value.number_integer <= 0x17) - { - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x18)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x19)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x1A)); - write_number(static_cast(j.m_value.number_integer)); - } - else - { - oa->write_character(to_char_type(0x1B)); - write_number(static_cast(j.m_value.number_integer)); - } - } - else - { - // The conversions below encode the sign in the first - // byte, and the value is converted to a positive number. - const auto positive_number = -1 - j.m_value.number_integer; - if (j.m_value.number_integer >= -24) - { - write_number(static_cast(0x20 + positive_number)); - } - else if (positive_number <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x38)); - write_number(static_cast(positive_number)); - } - else if (positive_number <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x39)); - write_number(static_cast(positive_number)); - } - else if (positive_number <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x3A)); - write_number(static_cast(positive_number)); - } - else - { - oa->write_character(to_char_type(0x3B)); - write_number(static_cast(positive_number)); - } - } - break; - } - - case value_t::number_unsigned: - { - if (j.m_value.number_unsigned <= 0x17) - { - write_number(static_cast(j.m_value.number_unsigned)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x18)); - write_number(static_cast(j.m_value.number_unsigned)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x19)); - write_number(static_cast(j.m_value.number_unsigned)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x1A)); - write_number(static_cast(j.m_value.number_unsigned)); - } - else - { - oa->write_character(to_char_type(0x1B)); - write_number(static_cast(j.m_value.number_unsigned)); - } - break; - } - - case value_t::number_float: - { - if (std::isnan(j.m_value.number_float)) - { - // NaN is 0xf97e00 in CBOR - oa->write_character(to_char_type(0xF9)); - oa->write_character(to_char_type(0x7E)); - oa->write_character(to_char_type(0x00)); - } - else if (std::isinf(j.m_value.number_float)) - { - // Infinity is 0xf97c00, -Infinity is 0xf9fc00 - oa->write_character(to_char_type(0xf9)); - oa->write_character(j.m_value.number_float > 0 ? to_char_type(0x7C) : to_char_type(0xFC)); - oa->write_character(to_char_type(0x00)); - } - else - { - write_compact_float(j.m_value.number_float, detail::input_format_t::cbor); - } - break; - } - - case value_t::string: - { - // step 1: write control byte and the string length - const auto N = j.m_value.string->size(); - if (N <= 0x17) - { - write_number(static_cast(0x60 + N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x78)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x79)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x7A)); - write_number(static_cast(N)); - } - // LCOV_EXCL_START - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x7B)); - write_number(static_cast(N)); - } - // LCOV_EXCL_STOP - - // step 2: write the string - oa->write_characters( - reinterpret_cast(j.m_value.string->c_str()), - j.m_value.string->size()); - break; - } - - case value_t::array: - { - // step 1: write control byte and the array size - const auto N = j.m_value.array->size(); - if (N <= 0x17) - { - write_number(static_cast(0x80 + N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x98)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x99)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x9A)); - write_number(static_cast(N)); - } - // LCOV_EXCL_START - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x9B)); - write_number(static_cast(N)); - } - // LCOV_EXCL_STOP - - // step 2: write each element - for (const auto& el : *j.m_value.array) - { - write_cbor(el); - } - break; - } - - case value_t::binary: - { - if (j.m_value.binary->has_subtype()) - { - write_number(static_cast(0xd8)); - write_number(j.m_value.binary->subtype()); - } - - // step 1: write control byte and the binary array size - const auto N = j.m_value.binary->size(); - if (N <= 0x17) - { - write_number(static_cast(0x40 + N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x58)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x59)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x5A)); - write_number(static_cast(N)); - } - // LCOV_EXCL_START - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0x5B)); - write_number(static_cast(N)); - } - // LCOV_EXCL_STOP - - // step 2: write each element - oa->write_characters( - reinterpret_cast(j.m_value.binary->data()), - N); - - break; - } - - case value_t::object: - { - // step 1: write control byte and the object size - const auto N = j.m_value.object->size(); - if (N <= 0x17) - { - write_number(static_cast(0xA0 + N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0xB8)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0xB9)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0xBA)); - write_number(static_cast(N)); - } - // LCOV_EXCL_START - else if (N <= (std::numeric_limits::max)()) - { - oa->write_character(to_char_type(0xBB)); - write_number(static_cast(N)); - } - // LCOV_EXCL_STOP - - // step 2: write each element - for (const auto& el : *j.m_value.object) - { - write_cbor(el.first); - write_cbor(el.second); - } - break; - } - - default: - break; - } - } - - /*! - @param[in] j JSON value to serialize - */ - void write_msgpack(const BasicJsonType& j) - { - switch (j.type()) - { - case value_t::null: // nil - { - oa->write_character(to_char_type(0xC0)); - break; - } - - case value_t::boolean: // true and false - { - oa->write_character(j.m_value.boolean - ? to_char_type(0xC3) - : to_char_type(0xC2)); - break; - } - - case value_t::number_integer: - { - if (j.m_value.number_integer >= 0) - { - // MessagePack does not differentiate between positive - // signed integers and unsigned integers. Therefore, we used - // the code from the value_t::number_unsigned case here. - if (j.m_value.number_unsigned < 128) - { - // positive fixnum - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 8 - oa->write_character(to_char_type(0xCC)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 16 - oa->write_character(to_char_type(0xCD)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 32 - oa->write_character(to_char_type(0xCE)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 64 - oa->write_character(to_char_type(0xCF)); - write_number(static_cast(j.m_value.number_integer)); - } - } - else - { - if (j.m_value.number_integer >= -32) - { - // negative fixnum - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer >= (std::numeric_limits::min)() && - j.m_value.number_integer <= (std::numeric_limits::max)()) - { - // int 8 - oa->write_character(to_char_type(0xD0)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer >= (std::numeric_limits::min)() && - j.m_value.number_integer <= (std::numeric_limits::max)()) - { - // int 16 - oa->write_character(to_char_type(0xD1)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer >= (std::numeric_limits::min)() && - j.m_value.number_integer <= (std::numeric_limits::max)()) - { - // int 32 - oa->write_character(to_char_type(0xD2)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_integer >= (std::numeric_limits::min)() && - j.m_value.number_integer <= (std::numeric_limits::max)()) - { - // int 64 - oa->write_character(to_char_type(0xD3)); - write_number(static_cast(j.m_value.number_integer)); - } - } - break; - } - - case value_t::number_unsigned: - { - if (j.m_value.number_unsigned < 128) - { - // positive fixnum - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 8 - oa->write_character(to_char_type(0xCC)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 16 - oa->write_character(to_char_type(0xCD)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 32 - oa->write_character(to_char_type(0xCE)); - write_number(static_cast(j.m_value.number_integer)); - } - else if (j.m_value.number_unsigned <= (std::numeric_limits::max)()) - { - // uint 64 - oa->write_character(to_char_type(0xCF)); - write_number(static_cast(j.m_value.number_integer)); - } - break; - } - - case value_t::number_float: - { - write_compact_float(j.m_value.number_float, detail::input_format_t::msgpack); - break; - } - - case value_t::string: - { - // step 1: write control byte and the string length - const auto N = j.m_value.string->size(); - if (N <= 31) - { - // fixstr - write_number(static_cast(0xA0 | N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // str 8 - oa->write_character(to_char_type(0xD9)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // str 16 - oa->write_character(to_char_type(0xDA)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // str 32 - oa->write_character(to_char_type(0xDB)); - write_number(static_cast(N)); - } - - // step 2: write the string - oa->write_characters( - reinterpret_cast(j.m_value.string->c_str()), - j.m_value.string->size()); - break; - } - - case value_t::array: - { - // step 1: write control byte and the array size - const auto N = j.m_value.array->size(); - if (N <= 15) - { - // fixarray - write_number(static_cast(0x90 | N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // array 16 - oa->write_character(to_char_type(0xDC)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // array 32 - oa->write_character(to_char_type(0xDD)); - write_number(static_cast(N)); - } - - // step 2: write each element - for (const auto& el : *j.m_value.array) - { - write_msgpack(el); - } - break; - } - - case value_t::binary: - { - // step 0: determine if the binary type has a set subtype to - // determine whether or not to use the ext or fixext types - const bool use_ext = j.m_value.binary->has_subtype(); - - // step 1: write control byte and the byte string length - const auto N = j.m_value.binary->size(); - if (N <= (std::numeric_limits::max)()) - { - std::uint8_t output_type{}; - bool fixed = true; - if (use_ext) - { - switch (N) - { - case 1: - output_type = 0xD4; // fixext 1 - break; - case 2: - output_type = 0xD5; // fixext 2 - break; - case 4: - output_type = 0xD6; // fixext 4 - break; - case 8: - output_type = 0xD7; // fixext 8 - break; - case 16: - output_type = 0xD8; // fixext 16 - break; - default: - output_type = 0xC7; // ext 8 - fixed = false; - break; - } - - } - else - { - output_type = 0xC4; // bin 8 - fixed = false; - } - - oa->write_character(to_char_type(output_type)); - if (!fixed) - { - write_number(static_cast(N)); - } - } - else if (N <= (std::numeric_limits::max)()) - { - std::uint8_t output_type = use_ext - ? 0xC8 // ext 16 - : 0xC5; // bin 16 - - oa->write_character(to_char_type(output_type)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - std::uint8_t output_type = use_ext - ? 0xC9 // ext 32 - : 0xC6; // bin 32 - - oa->write_character(to_char_type(output_type)); - write_number(static_cast(N)); - } - - // step 1.5: if this is an ext type, write the subtype - if (use_ext) - { - write_number(static_cast(j.m_value.binary->subtype())); - } - - // step 2: write the byte string - oa->write_characters( - reinterpret_cast(j.m_value.binary->data()), - N); - - break; - } - - case value_t::object: - { - // step 1: write control byte and the object size - const auto N = j.m_value.object->size(); - if (N <= 15) - { - // fixmap - write_number(static_cast(0x80 | (N & 0xF))); - } - else if (N <= (std::numeric_limits::max)()) - { - // map 16 - oa->write_character(to_char_type(0xDE)); - write_number(static_cast(N)); - } - else if (N <= (std::numeric_limits::max)()) - { - // map 32 - oa->write_character(to_char_type(0xDF)); - write_number(static_cast(N)); - } - - // step 2: write each element - for (const auto& el : *j.m_value.object) - { - write_msgpack(el.first); - write_msgpack(el.second); - } - break; - } - - default: - break; - } - } - - /*! - @param[in] j JSON value to serialize - @param[in] use_count whether to use '#' prefixes (optimized format) - @param[in] use_type whether to use '$' prefixes (optimized format) - @param[in] add_prefix whether prefixes need to be used for this value - */ - void write_ubjson(const BasicJsonType& j, const bool use_count, - const bool use_type, const bool add_prefix = true) - { - switch (j.type()) - { - case value_t::null: - { - if (add_prefix) - { - oa->write_character(to_char_type('Z')); - } - break; - } - - case value_t::boolean: - { - if (add_prefix) - { - oa->write_character(j.m_value.boolean - ? to_char_type('T') - : to_char_type('F')); - } - break; - } - - case value_t::number_integer: - { - write_number_with_ubjson_prefix(j.m_value.number_integer, add_prefix); - break; - } - - case value_t::number_unsigned: - { - write_number_with_ubjson_prefix(j.m_value.number_unsigned, add_prefix); - break; - } - - case value_t::number_float: - { - write_number_with_ubjson_prefix(j.m_value.number_float, add_prefix); - break; - } - - case value_t::string: - { - if (add_prefix) - { - oa->write_character(to_char_type('S')); - } - write_number_with_ubjson_prefix(j.m_value.string->size(), true); - oa->write_characters( - reinterpret_cast(j.m_value.string->c_str()), - j.m_value.string->size()); - break; - } - - case value_t::array: - { - if (add_prefix) - { - oa->write_character(to_char_type('[')); - } - - bool prefix_required = true; - if (use_type && !j.m_value.array->empty()) - { - JSON_ASSERT(use_count); - const CharType first_prefix = ubjson_prefix(j.front()); - const bool same_prefix = std::all_of(j.begin() + 1, j.end(), - [this, first_prefix](const BasicJsonType & v) - { - return ubjson_prefix(v) == first_prefix; - }); - - if (same_prefix) - { - prefix_required = false; - oa->write_character(to_char_type('$')); - oa->write_character(first_prefix); - } - } - - if (use_count) - { - oa->write_character(to_char_type('#')); - write_number_with_ubjson_prefix(j.m_value.array->size(), true); - } - - for (const auto& el : *j.m_value.array) - { - write_ubjson(el, use_count, use_type, prefix_required); - } - - if (!use_count) - { - oa->write_character(to_char_type(']')); - } - - break; - } - - case value_t::binary: - { - if (add_prefix) - { - oa->write_character(to_char_type('[')); - } - - if (use_type && !j.m_value.binary->empty()) - { - JSON_ASSERT(use_count); - oa->write_character(to_char_type('$')); - oa->write_character('U'); - } - - if (use_count) - { - oa->write_character(to_char_type('#')); - write_number_with_ubjson_prefix(j.m_value.binary->size(), true); - } - - if (use_type) - { - oa->write_characters( - reinterpret_cast(j.m_value.binary->data()), - j.m_value.binary->size()); - } - else - { - for (size_t i = 0; i < j.m_value.binary->size(); ++i) - { - oa->write_character(to_char_type('U')); - oa->write_character(j.m_value.binary->data()[i]); - } - } - - if (!use_count) - { - oa->write_character(to_char_type(']')); - } - - break; - } - - case value_t::object: - { - if (add_prefix) - { - oa->write_character(to_char_type('{')); - } - - bool prefix_required = true; - if (use_type && !j.m_value.object->empty()) - { - JSON_ASSERT(use_count); - const CharType first_prefix = ubjson_prefix(j.front()); - const bool same_prefix = std::all_of(j.begin(), j.end(), - [this, first_prefix](const BasicJsonType & v) - { - return ubjson_prefix(v) == first_prefix; - }); - - if (same_prefix) - { - prefix_required = false; - oa->write_character(to_char_type('$')); - oa->write_character(first_prefix); - } - } - - if (use_count) - { - oa->write_character(to_char_type('#')); - write_number_with_ubjson_prefix(j.m_value.object->size(), true); - } - - for (const auto& el : *j.m_value.object) - { - write_number_with_ubjson_prefix(el.first.size(), true); - oa->write_characters( - reinterpret_cast(el.first.c_str()), - el.first.size()); - write_ubjson(el.second, use_count, use_type, prefix_required); - } - - if (!use_count) - { - oa->write_character(to_char_type('}')); - } - - break; - } - - default: - break; - } - } - - private: - ////////// - // BSON // - ////////// - - /*! - @return The size of a BSON document entry header, including the id marker - and the entry name size (and its null-terminator). - */ - static std::size_t calc_bson_entry_header_size(const string_t& name) - { - const auto it = name.find(static_cast(0)); - if (JSON_HEDLEY_UNLIKELY(it != BasicJsonType::string_t::npos)) - { - JSON_THROW(out_of_range::create(409, - "BSON key cannot contain code point U+0000 (at byte " + std::to_string(it) + ")")); - } - - return /*id*/ 1ul + name.size() + /*zero-terminator*/1u; - } - - /*! - @brief Writes the given @a element_type and @a name to the output adapter - */ - void write_bson_entry_header(const string_t& name, - const std::uint8_t element_type) - { - oa->write_character(to_char_type(element_type)); // boolean - oa->write_characters( - reinterpret_cast(name.c_str()), - name.size() + 1u); - } - - /*! - @brief Writes a BSON element with key @a name and boolean value @a value - */ - void write_bson_boolean(const string_t& name, - const bool value) - { - write_bson_entry_header(name, 0x08); - oa->write_character(value ? to_char_type(0x01) : to_char_type(0x00)); - } - - /*! - @brief Writes a BSON element with key @a name and double value @a value - */ - void write_bson_double(const string_t& name, - const double value) - { - write_bson_entry_header(name, 0x01); - write_number(value); - } - - /*! - @return The size of the BSON-encoded string in @a value - */ - static std::size_t calc_bson_string_size(const string_t& value) - { - return sizeof(std::int32_t) + value.size() + 1ul; - } - - /*! - @brief Writes a BSON element with key @a name and string value @a value - */ - void write_bson_string(const string_t& name, - const string_t& value) - { - write_bson_entry_header(name, 0x02); - - write_number(static_cast(value.size() + 1ul)); - oa->write_characters( - reinterpret_cast(value.c_str()), - value.size() + 1); - } - - /*! - @brief Writes a BSON element with key @a name and null value - */ - void write_bson_null(const string_t& name) - { - write_bson_entry_header(name, 0x0A); - } - - /*! - @return The size of the BSON-encoded integer @a value - */ - static std::size_t calc_bson_integer_size(const std::int64_t value) - { - return (std::numeric_limits::min)() <= value && value <= (std::numeric_limits::max)() - ? sizeof(std::int32_t) - : sizeof(std::int64_t); - } - - /*! - @brief Writes a BSON element with key @a name and integer @a value - */ - void write_bson_integer(const string_t& name, - const std::int64_t value) - { - if ((std::numeric_limits::min)() <= value && value <= (std::numeric_limits::max)()) - { - write_bson_entry_header(name, 0x10); // int32 - write_number(static_cast(value)); - } - else - { - write_bson_entry_header(name, 0x12); // int64 - write_number(static_cast(value)); - } - } - - /*! - @return The size of the BSON-encoded unsigned integer in @a j - */ - static constexpr std::size_t calc_bson_unsigned_size(const std::uint64_t value) noexcept - { - return (value <= static_cast((std::numeric_limits::max)())) - ? sizeof(std::int32_t) - : sizeof(std::int64_t); - } - - /*! - @brief Writes a BSON element with key @a name and unsigned @a value - */ - void write_bson_unsigned(const string_t& name, - const std::uint64_t value) - { - if (value <= static_cast((std::numeric_limits::max)())) - { - write_bson_entry_header(name, 0x10 /* int32 */); - write_number(static_cast(value)); - } - else if (value <= static_cast((std::numeric_limits::max)())) - { - write_bson_entry_header(name, 0x12 /* int64 */); - write_number(static_cast(value)); - } - else - { - JSON_THROW(out_of_range::create(407, "integer number " + std::to_string(value) + " cannot be represented by BSON as it does not fit int64")); - } - } - - /*! - @brief Writes a BSON element with key @a name and object @a value - */ - void write_bson_object_entry(const string_t& name, - const typename BasicJsonType::object_t& value) - { - write_bson_entry_header(name, 0x03); // object - write_bson_object(value); - } - - /*! - @return The size of the BSON-encoded array @a value - */ - static std::size_t calc_bson_array_size(const typename BasicJsonType::array_t& value) - { - std::size_t array_index = 0ul; - - const std::size_t embedded_document_size = std::accumulate(std::begin(value), std::end(value), std::size_t(0), [&array_index](std::size_t result, const typename BasicJsonType::array_t::value_type & el) - { - return result + calc_bson_element_size(std::to_string(array_index++), el); - }); - - return sizeof(std::int32_t) + embedded_document_size + 1ul; - } - - /*! - @return The size of the BSON-encoded binary array @a value - */ - static std::size_t calc_bson_binary_size(const typename BasicJsonType::binary_t& value) - { - return sizeof(std::int32_t) + value.size() + 1ul; - } - - /*! - @brief Writes a BSON element with key @a name and array @a value - */ - void write_bson_array(const string_t& name, - const typename BasicJsonType::array_t& value) - { - write_bson_entry_header(name, 0x04); // array - write_number(static_cast(calc_bson_array_size(value))); - - std::size_t array_index = 0ul; - - for (const auto& el : value) - { - write_bson_element(std::to_string(array_index++), el); - } - - oa->write_character(to_char_type(0x00)); - } - - /*! - @brief Writes a BSON element with key @a name and binary value @a value - */ - void write_bson_binary(const string_t& name, - const binary_t& value) - { - write_bson_entry_header(name, 0x05); - - write_number(static_cast(value.size())); - write_number(value.has_subtype() ? value.subtype() : std::uint8_t(0x00)); - - oa->write_characters(reinterpret_cast(value.data()), value.size()); - } - - /*! - @brief Calculates the size necessary to serialize the JSON value @a j with its @a name - @return The calculated size for the BSON document entry for @a j with the given @a name. - */ - static std::size_t calc_bson_element_size(const string_t& name, - const BasicJsonType& j) - { - const auto header_size = calc_bson_entry_header_size(name); - switch (j.type()) - { - case value_t::object: - return header_size + calc_bson_object_size(*j.m_value.object); - - case value_t::array: - return header_size + calc_bson_array_size(*j.m_value.array); - - case value_t::binary: - return header_size + calc_bson_binary_size(*j.m_value.binary); - - case value_t::boolean: - return header_size + 1ul; - - case value_t::number_float: - return header_size + 8ul; - - case value_t::number_integer: - return header_size + calc_bson_integer_size(j.m_value.number_integer); - - case value_t::number_unsigned: - return header_size + calc_bson_unsigned_size(j.m_value.number_unsigned); - - case value_t::string: - return header_size + calc_bson_string_size(*j.m_value.string); - - case value_t::null: - return header_size + 0ul; - - // LCOV_EXCL_START - default: - JSON_ASSERT(false); - return 0ul; - // LCOV_EXCL_STOP - } - } - - /*! - @brief Serializes the JSON value @a j to BSON and associates it with the - key @a name. - @param name The name to associate with the JSON entity @a j within the - current BSON document - @return The size of the BSON entry - */ - void write_bson_element(const string_t& name, - const BasicJsonType& j) - { - switch (j.type()) - { - case value_t::object: - return write_bson_object_entry(name, *j.m_value.object); - - case value_t::array: - return write_bson_array(name, *j.m_value.array); - - case value_t::binary: - return write_bson_binary(name, *j.m_value.binary); - - case value_t::boolean: - return write_bson_boolean(name, j.m_value.boolean); - - case value_t::number_float: - return write_bson_double(name, j.m_value.number_float); - - case value_t::number_integer: - return write_bson_integer(name, j.m_value.number_integer); - - case value_t::number_unsigned: - return write_bson_unsigned(name, j.m_value.number_unsigned); - - case value_t::string: - return write_bson_string(name, *j.m_value.string); - - case value_t::null: - return write_bson_null(name); - - // LCOV_EXCL_START - default: - JSON_ASSERT(false); - return; - // LCOV_EXCL_STOP - } - } - - /*! - @brief Calculates the size of the BSON serialization of the given - JSON-object @a j. - @param[in] j JSON value to serialize - @pre j.type() == value_t::object - */ - static std::size_t calc_bson_object_size(const typename BasicJsonType::object_t& value) - { - std::size_t document_size = std::accumulate(value.begin(), value.end(), std::size_t(0), - [](size_t result, const typename BasicJsonType::object_t::value_type & el) - { - return result += calc_bson_element_size(el.first, el.second); - }); - - return sizeof(std::int32_t) + document_size + 1ul; - } - - /*! - @param[in] j JSON value to serialize - @pre j.type() == value_t::object - */ - void write_bson_object(const typename BasicJsonType::object_t& value) - { - write_number(static_cast(calc_bson_object_size(value))); - - for (const auto& el : value) - { - write_bson_element(el.first, el.second); - } - - oa->write_character(to_char_type(0x00)); - } - - ////////// - // CBOR // - ////////// - - static constexpr CharType get_cbor_float_prefix(float /*unused*/) - { - return to_char_type(0xFA); // Single-Precision Float - } - - static constexpr CharType get_cbor_float_prefix(double /*unused*/) - { - return to_char_type(0xFB); // Double-Precision Float - } - - ///////////// - // MsgPack // - ///////////// - - static constexpr CharType get_msgpack_float_prefix(float /*unused*/) - { - return to_char_type(0xCA); // float 32 - } - - static constexpr CharType get_msgpack_float_prefix(double /*unused*/) - { - return to_char_type(0xCB); // float 64 - } - - //////////// - // UBJSON // - //////////// - - // UBJSON: write number (floating point) - template::value, int>::type = 0> - void write_number_with_ubjson_prefix(const NumberType n, - const bool add_prefix) - { - if (add_prefix) - { - oa->write_character(get_ubjson_float_prefix(n)); - } - write_number(n); - } - - // UBJSON: write number (unsigned integer) - template::value, int>::type = 0> - void write_number_with_ubjson_prefix(const NumberType n, - const bool add_prefix) - { - if (n <= static_cast((std::numeric_limits::max)())) - { - if (add_prefix) - { - oa->write_character(to_char_type('i')); // int8 - } - write_number(static_cast(n)); - } - else if (n <= (std::numeric_limits::max)()) - { - if (add_prefix) - { - oa->write_character(to_char_type('U')); // uint8 - } - write_number(static_cast(n)); - } - else if (n <= static_cast((std::numeric_limits::max)())) - { - if (add_prefix) - { - oa->write_character(to_char_type('I')); // int16 - } - write_number(static_cast(n)); - } - else if (n <= static_cast((std::numeric_limits::max)())) - { - if (add_prefix) - { - oa->write_character(to_char_type('l')); // int32 - } - write_number(static_cast(n)); - } - else if (n <= static_cast((std::numeric_limits::max)())) - { - if (add_prefix) - { - oa->write_character(to_char_type('L')); // int64 - } - write_number(static_cast(n)); - } - else - { - if (add_prefix) - { - oa->write_character(to_char_type('H')); // high-precision number - } - - const auto number = BasicJsonType(n).dump(); - write_number_with_ubjson_prefix(number.size(), true); - for (std::size_t i = 0; i < number.size(); ++i) - { - oa->write_character(to_char_type(static_cast(number[i]))); - } - } - } - - // UBJSON: write number (signed integer) - template < typename NumberType, typename std::enable_if < - std::is_signed::value&& - !std::is_floating_point::value, int >::type = 0 > - void write_number_with_ubjson_prefix(const NumberType n, - const bool add_prefix) - { - if ((std::numeric_limits::min)() <= n && n <= (std::numeric_limits::max)()) - { - if (add_prefix) - { - oa->write_character(to_char_type('i')); // int8 - } - write_number(static_cast(n)); - } - else if (static_cast((std::numeric_limits::min)()) <= n && n <= static_cast((std::numeric_limits::max)())) - { - if (add_prefix) - { - oa->write_character(to_char_type('U')); // uint8 - } - write_number(static_cast(n)); - } - else if ((std::numeric_limits::min)() <= n && n <= (std::numeric_limits::max)()) - { - if (add_prefix) - { - oa->write_character(to_char_type('I')); // int16 - } - write_number(static_cast(n)); - } - else if ((std::numeric_limits::min)() <= n && n <= (std::numeric_limits::max)()) - { - if (add_prefix) - { - oa->write_character(to_char_type('l')); // int32 - } - write_number(static_cast(n)); - } - else if ((std::numeric_limits::min)() <= n && n <= (std::numeric_limits::max)()) - { - if (add_prefix) - { - oa->write_character(to_char_type('L')); // int64 - } - write_number(static_cast(n)); - } - // LCOV_EXCL_START - else - { - if (add_prefix) - { - oa->write_character(to_char_type('H')); // high-precision number - } - - const auto number = BasicJsonType(n).dump(); - write_number_with_ubjson_prefix(number.size(), true); - for (std::size_t i = 0; i < number.size(); ++i) - { - oa->write_character(to_char_type(static_cast(number[i]))); - } - } - // LCOV_EXCL_STOP - } - - /*! - @brief determine the type prefix of container values - */ - CharType ubjson_prefix(const BasicJsonType& j) const noexcept - { - switch (j.type()) - { - case value_t::null: - return 'Z'; - - case value_t::boolean: - return j.m_value.boolean ? 'T' : 'F'; - - case value_t::number_integer: - { - if ((std::numeric_limits::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits::max)()) - { - return 'i'; - } - if ((std::numeric_limits::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits::max)()) - { - return 'U'; - } - if ((std::numeric_limits::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits::max)()) - { - return 'I'; - } - if ((std::numeric_limits::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits::max)()) - { - return 'l'; - } - if ((std::numeric_limits::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits::max)()) - { - return 'L'; - } - // anything else is treated as high-precision number - return 'H'; // LCOV_EXCL_LINE - } - - case value_t::number_unsigned: - { - if (j.m_value.number_unsigned <= static_cast((std::numeric_limits::max)())) - { - return 'i'; - } - if (j.m_value.number_unsigned <= static_cast((std::numeric_limits::max)())) - { - return 'U'; - } - if (j.m_value.number_unsigned <= static_cast((std::numeric_limits::max)())) - { - return 'I'; - } - if (j.m_value.number_unsigned <= static_cast((std::numeric_limits::max)())) - { - return 'l'; - } - if (j.m_value.number_unsigned <= static_cast((std::numeric_limits::max)())) - { - return 'L'; - } - // anything else is treated as high-precision number - return 'H'; // LCOV_EXCL_LINE - } - - case value_t::number_float: - return get_ubjson_float_prefix(j.m_value.number_float); - - case value_t::string: - return 'S'; - - case value_t::array: // fallthrough - case value_t::binary: - return '['; - - case value_t::object: - return '{'; - - default: // discarded values - return 'N'; - } - } - - static constexpr CharType get_ubjson_float_prefix(float /*unused*/) - { - return 'd'; // float 32 - } - - static constexpr CharType get_ubjson_float_prefix(double /*unused*/) - { - return 'D'; // float 64 - } - - /////////////////////// - // Utility functions // - /////////////////////// - - /* - @brief write a number to output input - @param[in] n number of type @a NumberType - @tparam NumberType the type of the number - @tparam OutputIsLittleEndian Set to true if output data is - required to be little endian - - @note This function needs to respect the system's endianess, because bytes - in CBOR, MessagePack, and UBJSON are stored in network order (big - endian) and therefore need reordering on little endian systems. - */ - template - void write_number(const NumberType n) - { - // step 1: write number to array of length NumberType - std::array vec; - std::memcpy(vec.data(), &n, sizeof(NumberType)); - - // step 2: write array to output (with possible reordering) - if (is_little_endian != OutputIsLittleEndian) - { - // reverse byte order prior to conversion if necessary - std::reverse(vec.begin(), vec.end()); - } - - oa->write_characters(vec.data(), sizeof(NumberType)); - } - - void write_compact_float(const number_float_t n, detail::input_format_t format) - { - if (static_cast(n) >= static_cast(std::numeric_limits::lowest()) && - static_cast(n) <= static_cast((std::numeric_limits::max)()) && - static_cast(static_cast(n)) == static_cast(n)) - { - oa->write_character(format == detail::input_format_t::cbor - ? get_cbor_float_prefix(static_cast(n)) - : get_msgpack_float_prefix(static_cast(n))); - write_number(static_cast(n)); - } - else - { - oa->write_character(format == detail::input_format_t::cbor - ? get_cbor_float_prefix(n) - : get_msgpack_float_prefix(n)); - write_number(n); - } - } - - public: - // The following to_char_type functions are implement the conversion - // between uint8_t and CharType. In case CharType is not unsigned, - // such a conversion is required to allow values greater than 128. - // See for a discussion. - template < typename C = CharType, - enable_if_t < std::is_signed::value && std::is_signed::value > * = nullptr > - static constexpr CharType to_char_type(std::uint8_t x) noexcept - { - return *reinterpret_cast(&x); - } - - template < typename C = CharType, - enable_if_t < std::is_signed::value && std::is_unsigned::value > * = nullptr > - static CharType to_char_type(std::uint8_t x) noexcept - { - static_assert(sizeof(std::uint8_t) == sizeof(CharType), "size of CharType must be equal to std::uint8_t"); - static_assert(std::is_trivial::value, "CharType must be trivial"); - CharType result; - std::memcpy(&result, &x, sizeof(x)); - return result; - } - - template::value>* = nullptr> - static constexpr CharType to_char_type(std::uint8_t x) noexcept - { - return x; - } - - template < typename InputCharType, typename C = CharType, - enable_if_t < - std::is_signed::value && - std::is_signed::value && - std::is_same::type>::value - > * = nullptr > - static constexpr CharType to_char_type(InputCharType x) noexcept - { - return x; - } - - private: - /// whether we can assume little endianess - const bool is_little_endian = little_endianess(); - - /// the output - output_adapter_t oa = nullptr; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/output_adapters.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/output_adapters.hpp deleted file mode 100644 index 71ca65b92d6..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/output_adapters.hpp +++ /dev/null @@ -1,123 +0,0 @@ -#pragma once - -#include // copy -#include // size_t -#include // streamsize -#include // back_inserter -#include // shared_ptr, make_shared -#include // basic_ostream -#include // basic_string -#include // vector -#include - -namespace nlohmann -{ -namespace detail -{ -/// abstract output adapter interface -template struct output_adapter_protocol -{ - virtual void write_character(CharType c) = 0; - virtual void write_characters(const CharType* s, std::size_t length) = 0; - virtual ~output_adapter_protocol() = default; -}; - -/// a type to simplify interfaces -template -using output_adapter_t = std::shared_ptr>; - -/// output adapter for byte vectors -template -class output_vector_adapter : public output_adapter_protocol -{ - public: - explicit output_vector_adapter(std::vector& vec) noexcept - : v(vec) - {} - - void write_character(CharType c) override - { - v.push_back(c); - } - - JSON_HEDLEY_NON_NULL(2) - void write_characters(const CharType* s, std::size_t length) override - { - std::copy(s, s + length, std::back_inserter(v)); - } - - private: - std::vector& v; -}; - -/// output adapter for output streams -template -class output_stream_adapter : public output_adapter_protocol -{ - public: - explicit output_stream_adapter(std::basic_ostream& s) noexcept - : stream(s) - {} - - void write_character(CharType c) override - { - stream.put(c); - } - - JSON_HEDLEY_NON_NULL(2) - void write_characters(const CharType* s, std::size_t length) override - { - stream.write(s, static_cast(length)); - } - - private: - std::basic_ostream& stream; -}; - -/// output adapter for basic_string -template> -class output_string_adapter : public output_adapter_protocol -{ - public: - explicit output_string_adapter(StringType& s) noexcept - : str(s) - {} - - void write_character(CharType c) override - { - str.push_back(c); - } - - JSON_HEDLEY_NON_NULL(2) - void write_characters(const CharType* s, std::size_t length) override - { - str.append(s, length); - } - - private: - StringType& str; -}; - -template> -class output_adapter -{ - public: - output_adapter(std::vector& vec) - : oa(std::make_shared>(vec)) {} - - output_adapter(std::basic_ostream& s) - : oa(std::make_shared>(s)) {} - - output_adapter(StringType& s) - : oa(std::make_shared>(s)) {} - - operator output_adapter_t() - { - return oa; - } - - private: - output_adapter_t oa = nullptr; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/serializer.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/serializer.hpp deleted file mode 100644 index 865b8904e2a..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/output/serializer.hpp +++ /dev/null @@ -1,947 +0,0 @@ -#pragma once - -#include // reverse, remove, fill, find, none_of -#include // array -#include // localeconv, lconv -#include // labs, isfinite, isnan, signbit -#include // size_t, ptrdiff_t -#include // uint8_t -#include // snprintf -#include // numeric_limits -#include // string, char_traits -#include // is_same -#include // move - -#include -#include -#include -#include -#include -#include -#include - -namespace nlohmann -{ -namespace detail -{ -/////////////////// -// serialization // -/////////////////// - -/// how to treat decoding errors -enum class error_handler_t -{ - strict, ///< throw a type_error exception in case of invalid UTF-8 - replace, ///< replace invalid UTF-8 sequences with U+FFFD - ignore ///< ignore invalid UTF-8 sequences -}; - -template -class serializer -{ - using string_t = typename BasicJsonType::string_t; - using number_float_t = typename BasicJsonType::number_float_t; - using number_integer_t = typename BasicJsonType::number_integer_t; - using number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using binary_char_t = typename BasicJsonType::binary_t::value_type; - static constexpr std::uint8_t UTF8_ACCEPT = 0; - static constexpr std::uint8_t UTF8_REJECT = 1; - - public: - /*! - @param[in] s output stream to serialize to - @param[in] ichar indentation character to use - @param[in] error_handler_ how to react on decoding errors - */ - serializer(output_adapter_t s, const char ichar, - error_handler_t error_handler_ = error_handler_t::strict) - : o(std::move(s)) - , loc(std::localeconv()) - , thousands_sep(loc->thousands_sep == nullptr ? '\0' : std::char_traits::to_char_type(* (loc->thousands_sep))) - , decimal_point(loc->decimal_point == nullptr ? '\0' : std::char_traits::to_char_type(* (loc->decimal_point))) - , indent_char(ichar) - , indent_string(512, indent_char) - , error_handler(error_handler_) - {} - - // delete because of pointer members - serializer(const serializer&) = delete; - serializer& operator=(const serializer&) = delete; - serializer(serializer&&) = delete; - serializer& operator=(serializer&&) = delete; - ~serializer() = default; - - /*! - @brief internal implementation of the serialization function - - This function is called by the public member function dump and organizes - the serialization internally. The indentation level is propagated as - additional parameter. In case of arrays and objects, the function is - called recursively. - - - strings and object keys are escaped using `escape_string()` - - integer numbers are converted implicitly via `operator<<` - - floating-point numbers are converted to a string using `"%g"` format - - binary values are serialized as objects containing the subtype and the - byte array - - @param[in] val value to serialize - @param[in] pretty_print whether the output shall be pretty-printed - @param[in] ensure_ascii If @a ensure_ascii is true, all non-ASCII characters - in the output are escaped with `\uXXXX` sequences, and the result consists - of ASCII characters only. - @param[in] indent_step the indent level - @param[in] current_indent the current indent level (only used internally) - */ - void dump(const BasicJsonType& val, - const bool pretty_print, - const bool ensure_ascii, - const unsigned int indent_step, - const unsigned int current_indent = 0) - { - switch (val.m_type) - { - case value_t::object: - { - if (val.m_value.object->empty()) - { - o->write_characters("{}", 2); - return; - } - - if (pretty_print) - { - o->write_characters("{\n", 2); - - // variable to hold indentation for recursive calls - const auto new_indent = current_indent + indent_step; - if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent)) - { - indent_string.resize(indent_string.size() * 2, ' '); - } - - // first n-1 elements - auto i = val.m_value.object->cbegin(); - for (std::size_t cnt = 0; cnt < val.m_value.object->size() - 1; ++cnt, ++i) - { - o->write_characters(indent_string.c_str(), new_indent); - o->write_character('\"'); - dump_escaped(i->first, ensure_ascii); - o->write_characters("\": ", 3); - dump(i->second, true, ensure_ascii, indent_step, new_indent); - o->write_characters(",\n", 2); - } - - // last element - JSON_ASSERT(i != val.m_value.object->cend()); - JSON_ASSERT(std::next(i) == val.m_value.object->cend()); - o->write_characters(indent_string.c_str(), new_indent); - o->write_character('\"'); - dump_escaped(i->first, ensure_ascii); - o->write_characters("\": ", 3); - dump(i->second, true, ensure_ascii, indent_step, new_indent); - - o->write_character('\n'); - o->write_characters(indent_string.c_str(), current_indent); - o->write_character('}'); - } - else - { - o->write_character('{'); - - // first n-1 elements - auto i = val.m_value.object->cbegin(); - for (std::size_t cnt = 0; cnt < val.m_value.object->size() - 1; ++cnt, ++i) - { - o->write_character('\"'); - dump_escaped(i->first, ensure_ascii); - o->write_characters("\":", 2); - dump(i->second, false, ensure_ascii, indent_step, current_indent); - o->write_character(','); - } - - // last element - JSON_ASSERT(i != val.m_value.object->cend()); - JSON_ASSERT(std::next(i) == val.m_value.object->cend()); - o->write_character('\"'); - dump_escaped(i->first, ensure_ascii); - o->write_characters("\":", 2); - dump(i->second, false, ensure_ascii, indent_step, current_indent); - - o->write_character('}'); - } - - return; - } - - case value_t::array: - { - if (val.m_value.array->empty()) - { - o->write_characters("[]", 2); - return; - } - - if (pretty_print) - { - o->write_characters("[\n", 2); - - // variable to hold indentation for recursive calls - const auto new_indent = current_indent + indent_step; - if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent)) - { - indent_string.resize(indent_string.size() * 2, ' '); - } - - // first n-1 elements - for (auto i = val.m_value.array->cbegin(); - i != val.m_value.array->cend() - 1; ++i) - { - o->write_characters(indent_string.c_str(), new_indent); - dump(*i, true, ensure_ascii, indent_step, new_indent); - o->write_characters(",\n", 2); - } - - // last element - JSON_ASSERT(!val.m_value.array->empty()); - o->write_characters(indent_string.c_str(), new_indent); - dump(val.m_value.array->back(), true, ensure_ascii, indent_step, new_indent); - - o->write_character('\n'); - o->write_characters(indent_string.c_str(), current_indent); - o->write_character(']'); - } - else - { - o->write_character('['); - - // first n-1 elements - for (auto i = val.m_value.array->cbegin(); - i != val.m_value.array->cend() - 1; ++i) - { - dump(*i, false, ensure_ascii, indent_step, current_indent); - o->write_character(','); - } - - // last element - JSON_ASSERT(!val.m_value.array->empty()); - dump(val.m_value.array->back(), false, ensure_ascii, indent_step, current_indent); - - o->write_character(']'); - } - - return; - } - - case value_t::string: - { - o->write_character('\"'); - dump_escaped(*val.m_value.string, ensure_ascii); - o->write_character('\"'); - return; - } - - case value_t::binary: - { - if (pretty_print) - { - o->write_characters("{\n", 2); - - // variable to hold indentation for recursive calls - const auto new_indent = current_indent + indent_step; - if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent)) - { - indent_string.resize(indent_string.size() * 2, ' '); - } - - o->write_characters(indent_string.c_str(), new_indent); - - o->write_characters("\"bytes\": [", 10); - - if (!val.m_value.binary->empty()) - { - for (auto i = val.m_value.binary->cbegin(); - i != val.m_value.binary->cend() - 1; ++i) - { - dump_integer(*i); - o->write_characters(", ", 2); - } - dump_integer(val.m_value.binary->back()); - } - - o->write_characters("],\n", 3); - o->write_characters(indent_string.c_str(), new_indent); - - o->write_characters("\"subtype\": ", 11); - if (val.m_value.binary->has_subtype()) - { - dump_integer(val.m_value.binary->subtype()); - } - else - { - o->write_characters("null", 4); - } - o->write_character('\n'); - o->write_characters(indent_string.c_str(), current_indent); - o->write_character('}'); - } - else - { - o->write_characters("{\"bytes\":[", 10); - - if (!val.m_value.binary->empty()) - { - for (auto i = val.m_value.binary->cbegin(); - i != val.m_value.binary->cend() - 1; ++i) - { - dump_integer(*i); - o->write_character(','); - } - dump_integer(val.m_value.binary->back()); - } - - o->write_characters("],\"subtype\":", 12); - if (val.m_value.binary->has_subtype()) - { - dump_integer(val.m_value.binary->subtype()); - o->write_character('}'); - } - else - { - o->write_characters("null}", 5); - } - } - return; - } - - case value_t::boolean: - { - if (val.m_value.boolean) - { - o->write_characters("true", 4); - } - else - { - o->write_characters("false", 5); - } - return; - } - - case value_t::number_integer: - { - dump_integer(val.m_value.number_integer); - return; - } - - case value_t::number_unsigned: - { - dump_integer(val.m_value.number_unsigned); - return; - } - - case value_t::number_float: - { - dump_float(val.m_value.number_float); - return; - } - - case value_t::discarded: - { - o->write_characters("", 11); - return; - } - - case value_t::null: - { - o->write_characters("null", 4); - return; - } - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - } - - private: - /*! - @brief dump escaped string - - Escape a string by replacing certain special characters by a sequence of an - escape character (backslash) and another character and other control - characters by a sequence of "\u" followed by a four-digit hex - representation. The escaped string is written to output stream @a o. - - @param[in] s the string to escape - @param[in] ensure_ascii whether to escape non-ASCII characters with - \uXXXX sequences - - @complexity Linear in the length of string @a s. - */ - void dump_escaped(const string_t& s, const bool ensure_ascii) - { - std::uint32_t codepoint; - std::uint8_t state = UTF8_ACCEPT; - std::size_t bytes = 0; // number of bytes written to string_buffer - - // number of bytes written at the point of the last valid byte - std::size_t bytes_after_last_accept = 0; - std::size_t undumped_chars = 0; - - for (std::size_t i = 0; i < s.size(); ++i) - { - const auto byte = static_cast(s[i]); - - switch (decode(state, codepoint, byte)) - { - case UTF8_ACCEPT: // decode found a new code point - { - switch (codepoint) - { - case 0x08: // backspace - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 'b'; - break; - } - - case 0x09: // horizontal tab - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 't'; - break; - } - - case 0x0A: // newline - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 'n'; - break; - } - - case 0x0C: // formfeed - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 'f'; - break; - } - - case 0x0D: // carriage return - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 'r'; - break; - } - - case 0x22: // quotation mark - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = '\"'; - break; - } - - case 0x5C: // reverse solidus - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = '\\'; - break; - } - - default: - { - // escape control characters (0x00..0x1F) or, if - // ensure_ascii parameter is used, non-ASCII characters - if ((codepoint <= 0x1F) || (ensure_ascii && (codepoint >= 0x7F))) - { - if (codepoint <= 0xFFFF) - { - (std::snprintf)(string_buffer.data() + bytes, 7, "\\u%04x", - static_cast(codepoint)); - bytes += 6; - } - else - { - (std::snprintf)(string_buffer.data() + bytes, 13, "\\u%04x\\u%04x", - static_cast(0xD7C0u + (codepoint >> 10u)), - static_cast(0xDC00u + (codepoint & 0x3FFu))); - bytes += 12; - } - } - else - { - // copy byte to buffer (all previous bytes - // been copied have in default case above) - string_buffer[bytes++] = s[i]; - } - break; - } - } - - // write buffer and reset index; there must be 13 bytes - // left, as this is the maximal number of bytes to be - // written ("\uxxxx\uxxxx\0") for one code point - if (string_buffer.size() - bytes < 13) - { - o->write_characters(string_buffer.data(), bytes); - bytes = 0; - } - - // remember the byte position of this accept - bytes_after_last_accept = bytes; - undumped_chars = 0; - break; - } - - case UTF8_REJECT: // decode found invalid UTF-8 byte - { - switch (error_handler) - { - case error_handler_t::strict: - { - std::string sn(3, '\0'); - (std::snprintf)(&sn[0], sn.size(), "%.2X", byte); - JSON_THROW(type_error::create(316, "invalid UTF-8 byte at index " + std::to_string(i) + ": 0x" + sn)); - } - - case error_handler_t::ignore: - case error_handler_t::replace: - { - // in case we saw this character the first time, we - // would like to read it again, because the byte - // may be OK for itself, but just not OK for the - // previous sequence - if (undumped_chars > 0) - { - --i; - } - - // reset length buffer to the last accepted index; - // thus removing/ignoring the invalid characters - bytes = bytes_after_last_accept; - - if (error_handler == error_handler_t::replace) - { - // add a replacement character - if (ensure_ascii) - { - string_buffer[bytes++] = '\\'; - string_buffer[bytes++] = 'u'; - string_buffer[bytes++] = 'f'; - string_buffer[bytes++] = 'f'; - string_buffer[bytes++] = 'f'; - string_buffer[bytes++] = 'd'; - } - else - { - string_buffer[bytes++] = detail::binary_writer::to_char_type('\xEF'); - string_buffer[bytes++] = detail::binary_writer::to_char_type('\xBF'); - string_buffer[bytes++] = detail::binary_writer::to_char_type('\xBD'); - } - - // write buffer and reset index; there must be 13 bytes - // left, as this is the maximal number of bytes to be - // written ("\uxxxx\uxxxx\0") for one code point - if (string_buffer.size() - bytes < 13) - { - o->write_characters(string_buffer.data(), bytes); - bytes = 0; - } - - bytes_after_last_accept = bytes; - } - - undumped_chars = 0; - - // continue processing the string - state = UTF8_ACCEPT; - break; - } - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - break; - } - - default: // decode found yet incomplete multi-byte code point - { - if (!ensure_ascii) - { - // code point will not be escaped - copy byte to buffer - string_buffer[bytes++] = s[i]; - } - ++undumped_chars; - break; - } - } - } - - // we finished processing the string - if (JSON_HEDLEY_LIKELY(state == UTF8_ACCEPT)) - { - // write buffer - if (bytes > 0) - { - o->write_characters(string_buffer.data(), bytes); - } - } - else - { - // we finish reading, but do not accept: string was incomplete - switch (error_handler) - { - case error_handler_t::strict: - { - std::string sn(3, '\0'); - (std::snprintf)(&sn[0], sn.size(), "%.2X", static_cast(s.back())); - JSON_THROW(type_error::create(316, "incomplete UTF-8 string; last byte: 0x" + sn)); - } - - case error_handler_t::ignore: - { - // write all accepted bytes - o->write_characters(string_buffer.data(), bytes_after_last_accept); - break; - } - - case error_handler_t::replace: - { - // write all accepted bytes - o->write_characters(string_buffer.data(), bytes_after_last_accept); - // add a replacement character - if (ensure_ascii) - { - o->write_characters("\\ufffd", 6); - } - else - { - o->write_characters("\xEF\xBF\xBD", 3); - } - break; - } - - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - } - } - - /*! - @brief count digits - - Count the number of decimal (base 10) digits for an input unsigned integer. - - @param[in] x unsigned integer number to count its digits - @return number of decimal digits - */ - inline unsigned int count_digits(number_unsigned_t x) noexcept - { - unsigned int n_digits = 1; - for (;;) - { - if (x < 10) - { - return n_digits; - } - if (x < 100) - { - return n_digits + 1; - } - if (x < 1000) - { - return n_digits + 2; - } - if (x < 10000) - { - return n_digits + 3; - } - x = x / 10000u; - n_digits += 4; - } - } - - /*! - @brief dump an integer - - Dump a given integer to output stream @a o. Works internally with - @a number_buffer. - - @param[in] x integer number (signed or unsigned) to dump - @tparam NumberType either @a number_integer_t or @a number_unsigned_t - */ - template < typename NumberType, detail::enable_if_t < - std::is_same::value || - std::is_same::value || - std::is_same::value, - int > = 0 > - void dump_integer(NumberType x) - { - static constexpr std::array, 100> digits_to_99 - { - { - {{'0', '0'}}, {{'0', '1'}}, {{'0', '2'}}, {{'0', '3'}}, {{'0', '4'}}, {{'0', '5'}}, {{'0', '6'}}, {{'0', '7'}}, {{'0', '8'}}, {{'0', '9'}}, - {{'1', '0'}}, {{'1', '1'}}, {{'1', '2'}}, {{'1', '3'}}, {{'1', '4'}}, {{'1', '5'}}, {{'1', '6'}}, {{'1', '7'}}, {{'1', '8'}}, {{'1', '9'}}, - {{'2', '0'}}, {{'2', '1'}}, {{'2', '2'}}, {{'2', '3'}}, {{'2', '4'}}, {{'2', '5'}}, {{'2', '6'}}, {{'2', '7'}}, {{'2', '8'}}, {{'2', '9'}}, - {{'3', '0'}}, {{'3', '1'}}, {{'3', '2'}}, {{'3', '3'}}, {{'3', '4'}}, {{'3', '5'}}, {{'3', '6'}}, {{'3', '7'}}, {{'3', '8'}}, {{'3', '9'}}, - {{'4', '0'}}, {{'4', '1'}}, {{'4', '2'}}, {{'4', '3'}}, {{'4', '4'}}, {{'4', '5'}}, {{'4', '6'}}, {{'4', '7'}}, {{'4', '8'}}, {{'4', '9'}}, - {{'5', '0'}}, {{'5', '1'}}, {{'5', '2'}}, {{'5', '3'}}, {{'5', '4'}}, {{'5', '5'}}, {{'5', '6'}}, {{'5', '7'}}, {{'5', '8'}}, {{'5', '9'}}, - {{'6', '0'}}, {{'6', '1'}}, {{'6', '2'}}, {{'6', '3'}}, {{'6', '4'}}, {{'6', '5'}}, {{'6', '6'}}, {{'6', '7'}}, {{'6', '8'}}, {{'6', '9'}}, - {{'7', '0'}}, {{'7', '1'}}, {{'7', '2'}}, {{'7', '3'}}, {{'7', '4'}}, {{'7', '5'}}, {{'7', '6'}}, {{'7', '7'}}, {{'7', '8'}}, {{'7', '9'}}, - {{'8', '0'}}, {{'8', '1'}}, {{'8', '2'}}, {{'8', '3'}}, {{'8', '4'}}, {{'8', '5'}}, {{'8', '6'}}, {{'8', '7'}}, {{'8', '8'}}, {{'8', '9'}}, - {{'9', '0'}}, {{'9', '1'}}, {{'9', '2'}}, {{'9', '3'}}, {{'9', '4'}}, {{'9', '5'}}, {{'9', '6'}}, {{'9', '7'}}, {{'9', '8'}}, {{'9', '9'}}, - } - }; - - // special case for "0" - if (x == 0) - { - o->write_character('0'); - return; - } - - // use a pointer to fill the buffer - auto buffer_ptr = number_buffer.begin(); - - const bool is_negative = std::is_same::value && !(x >= 0); // see issue #755 - number_unsigned_t abs_value; - - unsigned int n_chars; - - if (is_negative) - { - *buffer_ptr = '-'; - abs_value = remove_sign(static_cast(x)); - - // account one more byte for the minus sign - n_chars = 1 + count_digits(abs_value); - } - else - { - abs_value = static_cast(x); - n_chars = count_digits(abs_value); - } - - // spare 1 byte for '\0' - JSON_ASSERT(n_chars < number_buffer.size() - 1); - - // jump to the end to generate the string from backward - // so we later avoid reversing the result - buffer_ptr += n_chars; - - // Fast int2ascii implementation inspired by "Fastware" talk by Andrei Alexandrescu - // See: https://www.youtube.com/watch?v=o4-CwDo2zpg - while (abs_value >= 100) - { - const auto digits_index = static_cast((abs_value % 100)); - abs_value /= 100; - *(--buffer_ptr) = digits_to_99[digits_index][1]; - *(--buffer_ptr) = digits_to_99[digits_index][0]; - } - - if (abs_value >= 10) - { - const auto digits_index = static_cast(abs_value); - *(--buffer_ptr) = digits_to_99[digits_index][1]; - *(--buffer_ptr) = digits_to_99[digits_index][0]; - } - else - { - *(--buffer_ptr) = static_cast('0' + abs_value); - } - - o->write_characters(number_buffer.data(), n_chars); - } - - /*! - @brief dump a floating-point number - - Dump a given floating-point number to output stream @a o. Works internally - with @a number_buffer. - - @param[in] x floating-point number to dump - */ - void dump_float(number_float_t x) - { - // NaN / inf - if (!std::isfinite(x)) - { - o->write_characters("null", 4); - return; - } - - // If number_float_t is an IEEE-754 single or double precision number, - // use the Grisu2 algorithm to produce short numbers which are - // guaranteed to round-trip, using strtof and strtod, resp. - // - // NB: The test below works if == . - static constexpr bool is_ieee_single_or_double - = (std::numeric_limits::is_iec559 && std::numeric_limits::digits == 24 && std::numeric_limits::max_exponent == 128) || - (std::numeric_limits::is_iec559 && std::numeric_limits::digits == 53 && std::numeric_limits::max_exponent == 1024); - - dump_float(x, std::integral_constant()); - } - - void dump_float(number_float_t x, std::true_type /*is_ieee_single_or_double*/) - { - char* begin = number_buffer.data(); - char* end = ::nlohmann::detail::to_chars(begin, begin + number_buffer.size(), x); - - o->write_characters(begin, static_cast(end - begin)); - } - - void dump_float(number_float_t x, std::false_type /*is_ieee_single_or_double*/) - { - // get number of digits for a float -> text -> float round-trip - static constexpr auto d = std::numeric_limits::max_digits10; - - // the actual conversion - std::ptrdiff_t len = (std::snprintf)(number_buffer.data(), number_buffer.size(), "%.*g", d, x); - - // negative value indicates an error - JSON_ASSERT(len > 0); - // check if buffer was large enough - JSON_ASSERT(static_cast(len) < number_buffer.size()); - - // erase thousands separator - if (thousands_sep != '\0') - { - const auto end = std::remove(number_buffer.begin(), - number_buffer.begin() + len, thousands_sep); - std::fill(end, number_buffer.end(), '\0'); - JSON_ASSERT((end - number_buffer.begin()) <= len); - len = (end - number_buffer.begin()); - } - - // convert decimal point to '.' - if (decimal_point != '\0' && decimal_point != '.') - { - const auto dec_pos = std::find(number_buffer.begin(), number_buffer.end(), decimal_point); - if (dec_pos != number_buffer.end()) - { - *dec_pos = '.'; - } - } - - o->write_characters(number_buffer.data(), static_cast(len)); - - // determine if need to append ".0" - const bool value_is_int_like = - std::none_of(number_buffer.begin(), number_buffer.begin() + len + 1, - [](char c) - { - return c == '.' || c == 'e'; - }); - - if (value_is_int_like) - { - o->write_characters(".0", 2); - } - } - - /*! - @brief check whether a string is UTF-8 encoded - - The function checks each byte of a string whether it is UTF-8 encoded. The - result of the check is stored in the @a state parameter. The function must - be called initially with state 0 (accept). State 1 means the string must - be rejected, because the current byte is not allowed. If the string is - completely processed, but the state is non-zero, the string ended - prematurely; that is, the last byte indicated more bytes should have - followed. - - @param[in,out] state the state of the decoding - @param[in,out] codep codepoint (valid only if resulting state is UTF8_ACCEPT) - @param[in] byte next byte to decode - @return new state - - @note The function has been edited: a std::array is used. - - @copyright Copyright (c) 2008-2009 Bjoern Hoehrmann - @sa http://bjoern.hoehrmann.de/utf-8/decoder/dfa/ - */ - static std::uint8_t decode(std::uint8_t& state, std::uint32_t& codep, const std::uint8_t byte) noexcept - { - static const std::array utf8d = - { - { - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 00..1F - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 20..3F - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 40..5F - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 60..7F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, // 80..9F - 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, // A0..BF - 8, 8, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // C0..DF - 0xA, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x4, 0x3, 0x3, // E0..EF - 0xB, 0x6, 0x6, 0x6, 0x5, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, // F0..FF - 0x0, 0x1, 0x2, 0x3, 0x5, 0x8, 0x7, 0x1, 0x1, 0x1, 0x4, 0x6, 0x1, 0x1, 0x1, 0x1, // s0..s0 - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, // s1..s2 - 1, 2, 1, 1, 1, 1, 1, 2, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, // s3..s4 - 1, 2, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 1, 3, 1, 1, 1, 1, 1, 1, // s5..s6 - 1, 3, 1, 1, 1, 1, 1, 3, 1, 3, 1, 1, 1, 1, 1, 1, 1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 // s7..s8 - } - }; - - const std::uint8_t type = utf8d[byte]; - - codep = (state != UTF8_ACCEPT) - ? (byte & 0x3fu) | (codep << 6u) - : (0xFFu >> type) & (byte); - - std::size_t index = 256u + static_cast(state) * 16u + static_cast(type); - JSON_ASSERT(index < 400); - state = utf8d[index]; - return state; - } - - /* - * Overload to make the compiler happy while it is instantiating - * dump_integer for number_unsigned_t. - * Must never be called. - */ - number_unsigned_t remove_sign(number_unsigned_t x) - { - JSON_ASSERT(false); // LCOV_EXCL_LINE - return x; // LCOV_EXCL_LINE - } - - /* - * Helper function for dump_integer - * - * This function takes a negative signed integer and returns its absolute - * value as unsigned integer. The plus/minus shuffling is necessary as we can - * not directly remove the sign of an arbitrary signed integer as the - * absolute values of INT_MIN and INT_MAX are usually not the same. See - * #1708 for details. - */ - inline number_unsigned_t remove_sign(number_integer_t x) noexcept - { - JSON_ASSERT(x < 0 && x < (std::numeric_limits::max)()); - return static_cast(-(x + 1)) + 1; - } - - private: - /// the output of the serializer - output_adapter_t o = nullptr; - - /// a (hopefully) large enough character buffer - std::array number_buffer{{}}; - - /// the locale - const std::lconv* loc = nullptr; - /// the locale's thousand separator character - const char thousands_sep = '\0'; - /// the locale's decimal point character - const char decimal_point = '\0'; - - /// string buffer - std::array string_buffer{{}}; - - /// the indentation character - const char indent_char; - /// the indentation string - string_t indent_string; - - /// error_handler how to react on decoding errors - const error_handler_t error_handler; -}; -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/value_t.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/value_t.hpp deleted file mode 100644 index 0383df06f8f..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/detail/value_t.hpp +++ /dev/null @@ -1,81 +0,0 @@ -#pragma once - -#include // array -#include // size_t -#include // uint8_t -#include // string - -namespace nlohmann -{ -namespace detail -{ -/////////////////////////// -// JSON type enumeration // -/////////////////////////// - -/*! -@brief the JSON type enumeration - -This enumeration collects the different JSON types. It is internally used to -distinguish the stored values, and the functions @ref basic_json::is_null(), -@ref basic_json::is_object(), @ref basic_json::is_array(), -@ref basic_json::is_string(), @ref basic_json::is_boolean(), -@ref basic_json::is_number() (with @ref basic_json::is_number_integer(), -@ref basic_json::is_number_unsigned(), and @ref basic_json::is_number_float()), -@ref basic_json::is_discarded(), @ref basic_json::is_primitive(), and -@ref basic_json::is_structured() rely on it. - -@note There are three enumeration entries (number_integer, number_unsigned, and -number_float), because the library distinguishes these three types for numbers: -@ref basic_json::number_unsigned_t is used for unsigned integers, -@ref basic_json::number_integer_t is used for signed integers, and -@ref basic_json::number_float_t is used for floating-point numbers or to -approximate integers which do not fit in the limits of their respective type. - -@sa @ref basic_json::basic_json(const value_t value_type) -- create a JSON -value with the default value for a given type - -@since version 1.0.0 -*/ -enum class value_t : std::uint8_t -{ - null, ///< null value - object, ///< object (unordered set of name/value pairs) - array, ///< array (ordered collection of values) - string, ///< string value - boolean, ///< boolean value - number_integer, ///< number value (signed integer) - number_unsigned, ///< number value (unsigned integer) - number_float, ///< number value (floating-point) - binary, ///< binary array (ordered collection of bytes) - discarded ///< discarded by the parser callback function -}; - -/*! -@brief comparison operator for JSON types - -Returns an ordering that is similar to Python: -- order: null < boolean < number < object < array < string < binary -- furthermore, each type is not smaller than itself -- discarded values are not comparable -- binary is represented as a b"" string in python and directly comparable to a - string; however, making a binary array directly comparable with a string would - be surprising behavior in a JSON file. - -@since version 1.0.0 -*/ -inline bool operator<(const value_t lhs, const value_t rhs) noexcept -{ - static constexpr std::array order = {{ - 0 /* null */, 3 /* object */, 4 /* array */, 5 /* string */, - 1 /* boolean */, 2 /* integer */, 2 /* unsigned */, 2 /* float */, - 6 /* binary */ - } - }; - - const auto l_index = static_cast(lhs); - const auto r_index = static_cast(rhs); - return l_index < order.size() && r_index < order.size() && order[l_index] < order[r_index]; -} -} // namespace detail -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json.hpp deleted file mode 100644 index a9e6c76d114..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json.hpp +++ /dev/null @@ -1,8791 +0,0 @@ -/* - __ _____ _____ _____ - __| | __| | | | JSON for Modern C++ -| | |__ | | | | | | version 3.9.1 -|_____|_____|_____|_|___| https://github.com/nlohmann/json - -Licensed under the MIT License . -SPDX-License-Identifier: MIT -Copyright (c) 2013-2019 Niels Lohmann . - -Permission is hereby granted, free of charge, to any person obtaining a copy -of this software and associated documentation files (the "Software"), to deal -in the Software without restriction, including without limitation the rights -to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -copies of the Software, and to permit persons to whom the Software is -furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice shall be included in all -copies or substantial portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE -SOFTWARE. -*/ - -#ifndef INCLUDE_NLOHMANN_JSON_HPP_ -#define INCLUDE_NLOHMANN_JSON_HPP_ - -#define NLOHMANN_JSON_VERSION_MAJOR 3 -#define NLOHMANN_JSON_VERSION_MINOR 9 -#define NLOHMANN_JSON_VERSION_PATCH 1 - -#include // all_of, find, for_each -#include // nullptr_t, ptrdiff_t, size_t -#include // hash, less -#include // initializer_list -#include // istream, ostream -#include // random_access_iterator_tag -#include // unique_ptr -#include // accumulate -#include // string, stoi, to_string -#include // declval, forward, move, pair, swap -#include // vector - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -/*! -@brief namespace for Niels Lohmann -@see https://github.com/nlohmann -@since version 1.0.0 -*/ -namespace nlohmann -{ - -/*! -@brief a class to store JSON values - -@tparam ObjectType type for JSON objects (`std::map` by default; will be used -in @ref object_t) -@tparam ArrayType type for JSON arrays (`std::vector` by default; will be used -in @ref array_t) -@tparam StringType type for JSON strings and object keys (`std::string` by -default; will be used in @ref string_t) -@tparam BooleanType type for JSON booleans (`bool` by default; will be used -in @ref boolean_t) -@tparam NumberIntegerType type for JSON integer numbers (`int64_t` by -default; will be used in @ref number_integer_t) -@tparam NumberUnsignedType type for JSON unsigned integer numbers (@c -`uint64_t` by default; will be used in @ref number_unsigned_t) -@tparam NumberFloatType type for JSON floating-point numbers (`double` by -default; will be used in @ref number_float_t) -@tparam BinaryType type for packed binary data for compatibility with binary -serialization formats (`std::vector` by default; will be used in -@ref binary_t) -@tparam AllocatorType type of the allocator to use (`std::allocator` by -default) -@tparam JSONSerializer the serializer to resolve internal calls to `to_json()` -and `from_json()` (@ref adl_serializer by default) - -@requirement The class satisfies the following concept requirements: -- Basic - - [DefaultConstructible](https://en.cppreference.com/w/cpp/named_req/DefaultConstructible): - JSON values can be default constructed. The result will be a JSON null - value. - - [MoveConstructible](https://en.cppreference.com/w/cpp/named_req/MoveConstructible): - A JSON value can be constructed from an rvalue argument. - - [CopyConstructible](https://en.cppreference.com/w/cpp/named_req/CopyConstructible): - A JSON value can be copy-constructed from an lvalue expression. - - [MoveAssignable](https://en.cppreference.com/w/cpp/named_req/MoveAssignable): - A JSON value van be assigned from an rvalue argument. - - [CopyAssignable](https://en.cppreference.com/w/cpp/named_req/CopyAssignable): - A JSON value can be copy-assigned from an lvalue expression. - - [Destructible](https://en.cppreference.com/w/cpp/named_req/Destructible): - JSON values can be destructed. -- Layout - - [StandardLayoutType](https://en.cppreference.com/w/cpp/named_req/StandardLayoutType): - JSON values have - [standard layout](https://en.cppreference.com/w/cpp/language/data_members#Standard_layout): - All non-static data members are private and standard layout types, the - class has no virtual functions or (virtual) base classes. -- Library-wide - - [EqualityComparable](https://en.cppreference.com/w/cpp/named_req/EqualityComparable): - JSON values can be compared with `==`, see @ref - operator==(const_reference,const_reference). - - [LessThanComparable](https://en.cppreference.com/w/cpp/named_req/LessThanComparable): - JSON values can be compared with `<`, see @ref - operator<(const_reference,const_reference). - - [Swappable](https://en.cppreference.com/w/cpp/named_req/Swappable): - Any JSON lvalue or rvalue of can be swapped with any lvalue or rvalue of - other compatible types, using unqualified function call @ref swap(). - - [NullablePointer](https://en.cppreference.com/w/cpp/named_req/NullablePointer): - JSON values can be compared against `std::nullptr_t` objects which are used - to model the `null` value. -- Container - - [Container](https://en.cppreference.com/w/cpp/named_req/Container): - JSON values can be used like STL containers and provide iterator access. - - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer); - JSON values can be used like STL containers and provide reverse iterator - access. - -@invariant The member variables @a m_value and @a m_type have the following -relationship: -- If `m_type == value_t::object`, then `m_value.object != nullptr`. -- If `m_type == value_t::array`, then `m_value.array != nullptr`. -- If `m_type == value_t::string`, then `m_value.string != nullptr`. -The invariants are checked by member function assert_invariant(). - -@internal -@note ObjectType trick from https://stackoverflow.com/a/9860911 -@endinternal - -@see [RFC 7159: The JavaScript Object Notation (JSON) Data Interchange -Format](http://rfc7159.net/rfc7159) - -@since version 1.0.0 - -@nosubgrouping -*/ -NLOHMANN_BASIC_JSON_TPL_DECLARATION -class basic_json -{ - private: - template friend struct detail::external_constructor; - friend ::nlohmann::json_pointer; - - template - friend class ::nlohmann::detail::parser; - friend ::nlohmann::detail::serializer; - template - friend class ::nlohmann::detail::iter_impl; - template - friend class ::nlohmann::detail::binary_writer; - template - friend class ::nlohmann::detail::binary_reader; - template - friend class ::nlohmann::detail::json_sax_dom_parser; - template - friend class ::nlohmann::detail::json_sax_dom_callback_parser; - - /// workaround type for MSVC - using basic_json_t = NLOHMANN_BASIC_JSON_TPL; - - // convenience aliases for types residing in namespace detail; - using lexer = ::nlohmann::detail::lexer_base; - - template - static ::nlohmann::detail::parser parser( - InputAdapterType adapter, - detail::parser_callback_tcb = nullptr, - const bool allow_exceptions = true, - const bool ignore_comments = false - ) - { - return ::nlohmann::detail::parser(std::move(adapter), - std::move(cb), allow_exceptions, ignore_comments); - } - - using primitive_iterator_t = ::nlohmann::detail::primitive_iterator_t; - template - using internal_iterator = ::nlohmann::detail::internal_iterator; - template - using iter_impl = ::nlohmann::detail::iter_impl; - template - using iteration_proxy = ::nlohmann::detail::iteration_proxy; - template using json_reverse_iterator = ::nlohmann::detail::json_reverse_iterator; - - template - using output_adapter_t = ::nlohmann::detail::output_adapter_t; - - template - using binary_reader = ::nlohmann::detail::binary_reader; - template using binary_writer = ::nlohmann::detail::binary_writer; - - using serializer = ::nlohmann::detail::serializer; - - public: - using value_t = detail::value_t; - /// JSON Pointer, see @ref nlohmann::json_pointer - using json_pointer = ::nlohmann::json_pointer; - template - using json_serializer = JSONSerializer; - /// how to treat decoding errors - using error_handler_t = detail::error_handler_t; - /// how to treat CBOR tags - using cbor_tag_handler_t = detail::cbor_tag_handler_t; - /// helper type for initializer lists of basic_json values - using initializer_list_t = std::initializer_list>; - - using input_format_t = detail::input_format_t; - /// SAX interface type, see @ref nlohmann::json_sax - using json_sax_t = json_sax; - - //////////////// - // exceptions // - //////////////// - - /// @name exceptions - /// Classes to implement user-defined exceptions. - /// @{ - - /// @copydoc detail::exception - using exception = detail::exception; - /// @copydoc detail::parse_error - using parse_error = detail::parse_error; - /// @copydoc detail::invalid_iterator - using invalid_iterator = detail::invalid_iterator; - /// @copydoc detail::type_error - using type_error = detail::type_error; - /// @copydoc detail::out_of_range - using out_of_range = detail::out_of_range; - /// @copydoc detail::other_error - using other_error = detail::other_error; - - /// @} - - - ///////////////////// - // container types // - ///////////////////// - - /// @name container types - /// The canonic container types to use @ref basic_json like any other STL - /// container. - /// @{ - - /// the type of elements in a basic_json container - using value_type = basic_json; - - /// the type of an element reference - using reference = value_type&; - /// the type of an element const reference - using const_reference = const value_type&; - - /// a type to represent differences between iterators - using difference_type = std::ptrdiff_t; - /// a type to represent container sizes - using size_type = std::size_t; - - /// the allocator type - using allocator_type = AllocatorType; - - /// the type of an element pointer - using pointer = typename std::allocator_traits::pointer; - /// the type of an element const pointer - using const_pointer = typename std::allocator_traits::const_pointer; - - /// an iterator for a basic_json container - using iterator = iter_impl; - /// a const iterator for a basic_json container - using const_iterator = iter_impl; - /// a reverse iterator for a basic_json container - using reverse_iterator = json_reverse_iterator; - /// a const reverse iterator for a basic_json container - using const_reverse_iterator = json_reverse_iterator; - - /// @} - - - /*! - @brief returns the allocator associated with the container - */ - static allocator_type get_allocator() - { - return allocator_type(); - } - - /*! - @brief returns version information on the library - - This function returns a JSON object with information about the library, - including the version number and information on the platform and compiler. - - @return JSON object holding version information - key | description - ----------- | --------------- - `compiler` | Information on the used compiler. It is an object with the following keys: `c++` (the used C++ standard), `family` (the compiler family; possible values are `clang`, `icc`, `gcc`, `ilecpp`, `msvc`, `pgcpp`, `sunpro`, and `unknown`), and `version` (the compiler version). - `copyright` | The copyright line for the library as string. - `name` | The name of the library as string. - `platform` | The used platform as string. Possible values are `win32`, `linux`, `apple`, `unix`, and `unknown`. - `url` | The URL of the project as string. - `version` | The version of the library. It is an object with the following keys: `major`, `minor`, and `patch` as defined by [Semantic Versioning](http://semver.org), and `string` (the version string). - - @liveexample{The following code shows an example output of the `meta()` - function.,meta} - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @complexity Constant. - - @since 2.1.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json meta() - { - basic_json result; - - result["copyright"] = "(C) 2013-2020 Niels Lohmann"; - result["name"] = "JSON for Modern C++"; - result["url"] = "https://github.com/nlohmann/json"; - result["version"]["string"] = - std::to_string(NLOHMANN_JSON_VERSION_MAJOR) + "." + - std::to_string(NLOHMANN_JSON_VERSION_MINOR) + "." + - std::to_string(NLOHMANN_JSON_VERSION_PATCH); - result["version"]["major"] = NLOHMANN_JSON_VERSION_MAJOR; - result["version"]["minor"] = NLOHMANN_JSON_VERSION_MINOR; - result["version"]["patch"] = NLOHMANN_JSON_VERSION_PATCH; - -#ifdef _WIN32 - result["platform"] = "win32"; -#elif defined __linux__ - result["platform"] = "linux"; -#elif defined __APPLE__ - result["platform"] = "apple"; -#elif defined __unix__ - result["platform"] = "unix"; -#else - result["platform"] = "unknown"; -#endif - -#if defined(__ICC) || defined(__INTEL_COMPILER) - result["compiler"] = {{"family", "icc"}, {"version", __INTEL_COMPILER}}; -#elif defined(__clang__) - result["compiler"] = {{"family", "clang"}, {"version", __clang_version__}}; -#elif defined(__GNUC__) || defined(__GNUG__) - result["compiler"] = {{"family", "gcc"}, {"version", std::to_string(__GNUC__) + "." + std::to_string(__GNUC_MINOR__) + "." + std::to_string(__GNUC_PATCHLEVEL__)}}; -#elif defined(__HP_cc) || defined(__HP_aCC) - result["compiler"] = "hp" -#elif defined(__IBMCPP__) - result["compiler"] = {{"family", "ilecpp"}, {"version", __IBMCPP__}}; -#elif defined(_MSC_VER) - result["compiler"] = {{"family", "msvc"}, {"version", _MSC_VER}}; -#elif defined(__PGI) - result["compiler"] = {{"family", "pgcpp"}, {"version", __PGI}}; -#elif defined(__SUNPRO_CC) - result["compiler"] = {{"family", "sunpro"}, {"version", __SUNPRO_CC}}; -#else - result["compiler"] = {{"family", "unknown"}, {"version", "unknown"}}; -#endif - -#ifdef __cplusplus - result["compiler"]["c++"] = std::to_string(__cplusplus); -#else - result["compiler"]["c++"] = "unknown"; -#endif - return result; - } - - - /////////////////////////// - // JSON value data types // - /////////////////////////// - - /// @name JSON value data types - /// The data types to store a JSON value. These types are derived from - /// the template arguments passed to class @ref basic_json. - /// @{ - -#if defined(JSON_HAS_CPP_14) - // Use transparent comparator if possible, combined with perfect forwarding - // on find() and count() calls prevents unnecessary string construction. - using object_comparator_t = std::less<>; -#else - using object_comparator_t = std::less; -#endif - - /*! - @brief a type for an object - - [RFC 7159](http://rfc7159.net/rfc7159) describes JSON objects as follows: - > An object is an unordered collection of zero or more name/value pairs, - > where a name is a string and a value is a string, number, boolean, null, - > object, or array. - - To store objects in C++, a type is defined by the template parameters - described below. - - @tparam ObjectType the container to store objects (e.g., `std::map` or - `std::unordered_map`) - @tparam StringType the type of the keys or names (e.g., `std::string`). - The comparison function `std::less` is used to order elements - inside the container. - @tparam AllocatorType the allocator to use for objects (e.g., - `std::allocator`) - - #### Default type - - With the default values for @a ObjectType (`std::map`), @a StringType - (`std::string`), and @a AllocatorType (`std::allocator`), the default - value for @a object_t is: - - @code {.cpp} - std::map< - std::string, // key_type - basic_json, // value_type - std::less, // key_compare - std::allocator> // allocator_type - > - @endcode - - #### Behavior - - The choice of @a object_t influences the behavior of the JSON class. With - the default type, objects have the following behavior: - - - When all names are unique, objects will be interoperable in the sense - that all software implementations receiving that object will agree on - the name-value mappings. - - When the names within an object are not unique, it is unspecified which - one of the values for a given key will be chosen. For instance, - `{"key": 2, "key": 1}` could be equal to either `{"key": 1}` or - `{"key": 2}`. - - Internally, name/value pairs are stored in lexicographical order of the - names. Objects will also be serialized (see @ref dump) in this order. - For instance, `{"b": 1, "a": 2}` and `{"a": 2, "b": 1}` will be stored - and serialized as `{"a": 2, "b": 1}`. - - When comparing objects, the order of the name/value pairs is irrelevant. - This makes objects interoperable in the sense that they will not be - affected by these differences. For instance, `{"b": 1, "a": 2}` and - `{"a": 2, "b": 1}` will be treated as equal. - - #### Limits - - [RFC 7159](http://rfc7159.net/rfc7159) specifies: - > An implementation may set limits on the maximum depth of nesting. - - In this class, the object's limit of nesting is not explicitly constrained. - However, a maximum depth of nesting may be introduced by the compiler or - runtime environment. A theoretical limit can be queried by calling the - @ref max_size function of a JSON object. - - #### Storage - - Objects are stored as pointers in a @ref basic_json type. That is, for any - access to object values, a pointer of type `object_t*` must be - dereferenced. - - @sa @ref array_t -- type for an array value - - @since version 1.0.0 - - @note The order name/value pairs are added to the object is *not* - preserved by the library. Therefore, iterating an object may return - name/value pairs in a different order than they were originally stored. In - fact, keys will be traversed in alphabetical order as `std::map` with - `std::less` is used by default. Please note this behavior conforms to [RFC - 7159](http://rfc7159.net/rfc7159), because any order implements the - specified "unordered" nature of JSON objects. - */ - using object_t = ObjectType>>; - - /*! - @brief a type for an array - - [RFC 7159](http://rfc7159.net/rfc7159) describes JSON arrays as follows: - > An array is an ordered sequence of zero or more values. - - To store objects in C++, a type is defined by the template parameters - explained below. - - @tparam ArrayType container type to store arrays (e.g., `std::vector` or - `std::list`) - @tparam AllocatorType allocator to use for arrays (e.g., `std::allocator`) - - #### Default type - - With the default values for @a ArrayType (`std::vector`) and @a - AllocatorType (`std::allocator`), the default value for @a array_t is: - - @code {.cpp} - std::vector< - basic_json, // value_type - std::allocator // allocator_type - > - @endcode - - #### Limits - - [RFC 7159](http://rfc7159.net/rfc7159) specifies: - > An implementation may set limits on the maximum depth of nesting. - - In this class, the array's limit of nesting is not explicitly constrained. - However, a maximum depth of nesting may be introduced by the compiler or - runtime environment. A theoretical limit can be queried by calling the - @ref max_size function of a JSON array. - - #### Storage - - Arrays are stored as pointers in a @ref basic_json type. That is, for any - access to array values, a pointer of type `array_t*` must be dereferenced. - - @sa @ref object_t -- type for an object value - - @since version 1.0.0 - */ - using array_t = ArrayType>; - - /*! - @brief a type for a string - - [RFC 7159](http://rfc7159.net/rfc7159) describes JSON strings as follows: - > A string is a sequence of zero or more Unicode characters. - - To store objects in C++, a type is defined by the template parameter - described below. Unicode values are split by the JSON class into - byte-sized characters during deserialization. - - @tparam StringType the container to store strings (e.g., `std::string`). - Note this container is used for keys/names in objects, see @ref object_t. - - #### Default type - - With the default values for @a StringType (`std::string`), the default - value for @a string_t is: - - @code {.cpp} - std::string - @endcode - - #### Encoding - - Strings are stored in UTF-8 encoding. Therefore, functions like - `std::string::size()` or `std::string::length()` return the number of - bytes in the string rather than the number of characters or glyphs. - - #### String comparison - - [RFC 7159](http://rfc7159.net/rfc7159) states: - > Software implementations are typically required to test names of object - > members for equality. Implementations that transform the textual - > representation into sequences of Unicode code units and then perform the - > comparison numerically, code unit by code unit, are interoperable in the - > sense that implementations will agree in all cases on equality or - > inequality of two strings. For example, implementations that compare - > strings with escaped characters unconverted may incorrectly find that - > `"a\\b"` and `"a\u005Cb"` are not equal. - - This implementation is interoperable as it does compare strings code unit - by code unit. - - #### Storage - - String values are stored as pointers in a @ref basic_json type. That is, - for any access to string values, a pointer of type `string_t*` must be - dereferenced. - - @since version 1.0.0 - */ - using string_t = StringType; - - /*! - @brief a type for a boolean - - [RFC 7159](http://rfc7159.net/rfc7159) implicitly describes a boolean as a - type which differentiates the two literals `true` and `false`. - - To store objects in C++, a type is defined by the template parameter @a - BooleanType which chooses the type to use. - - #### Default type - - With the default values for @a BooleanType (`bool`), the default value for - @a boolean_t is: - - @code {.cpp} - bool - @endcode - - #### Storage - - Boolean values are stored directly inside a @ref basic_json type. - - @since version 1.0.0 - */ - using boolean_t = BooleanType; - - /*! - @brief a type for a number (integer) - - [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows: - > The representation of numbers is similar to that used in most - > programming languages. A number is represented in base 10 using decimal - > digits. It contains an integer component that may be prefixed with an - > optional minus sign, which may be followed by a fraction part and/or an - > exponent part. Leading zeros are not allowed. (...) Numeric values that - > cannot be represented in the grammar below (such as Infinity and NaN) - > are not permitted. - - This description includes both integer and floating-point numbers. - However, C++ allows more precise storage if it is known whether the number - is a signed integer, an unsigned integer or a floating-point number. - Therefore, three different types, @ref number_integer_t, @ref - number_unsigned_t and @ref number_float_t are used. - - To store integer numbers in C++, a type is defined by the template - parameter @a NumberIntegerType which chooses the type to use. - - #### Default type - - With the default values for @a NumberIntegerType (`int64_t`), the default - value for @a number_integer_t is: - - @code {.cpp} - int64_t - @endcode - - #### Default behavior - - - The restrictions about leading zeros is not enforced in C++. Instead, - leading zeros in integer literals lead to an interpretation as octal - number. Internally, the value will be stored as decimal number. For - instance, the C++ integer literal `010` will be serialized to `8`. - During deserialization, leading zeros yield an error. - - Not-a-number (NaN) values will be serialized to `null`. - - #### Limits - - [RFC 7159](http://rfc7159.net/rfc7159) specifies: - > An implementation may set limits on the range and precision of numbers. - - When the default type is used, the maximal integer number that can be - stored is `9223372036854775807` (INT64_MAX) and the minimal integer number - that can be stored is `-9223372036854775808` (INT64_MIN). Integer numbers - that are out of range will yield over/underflow when used in a - constructor. During deserialization, too large or small integer numbers - will be automatically be stored as @ref number_unsigned_t or @ref - number_float_t. - - [RFC 7159](http://rfc7159.net/rfc7159) further states: - > Note that when such software is used, numbers that are integers and are - > in the range \f$[-2^{53}+1, 2^{53}-1]\f$ are interoperable in the sense - > that implementations will agree exactly on their numeric values. - - As this range is a subrange of the exactly supported range [INT64_MIN, - INT64_MAX], this class's integer type is interoperable. - - #### Storage - - Integer number values are stored directly inside a @ref basic_json type. - - @sa @ref number_float_t -- type for number values (floating-point) - - @sa @ref number_unsigned_t -- type for number values (unsigned integer) - - @since version 1.0.0 - */ - using number_integer_t = NumberIntegerType; - - /*! - @brief a type for a number (unsigned) - - [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows: - > The representation of numbers is similar to that used in most - > programming languages. A number is represented in base 10 using decimal - > digits. It contains an integer component that may be prefixed with an - > optional minus sign, which may be followed by a fraction part and/or an - > exponent part. Leading zeros are not allowed. (...) Numeric values that - > cannot be represented in the grammar below (such as Infinity and NaN) - > are not permitted. - - This description includes both integer and floating-point numbers. - However, C++ allows more precise storage if it is known whether the number - is a signed integer, an unsigned integer or a floating-point number. - Therefore, three different types, @ref number_integer_t, @ref - number_unsigned_t and @ref number_float_t are used. - - To store unsigned integer numbers in C++, a type is defined by the - template parameter @a NumberUnsignedType which chooses the type to use. - - #### Default type - - With the default values for @a NumberUnsignedType (`uint64_t`), the - default value for @a number_unsigned_t is: - - @code {.cpp} - uint64_t - @endcode - - #### Default behavior - - - The restrictions about leading zeros is not enforced in C++. Instead, - leading zeros in integer literals lead to an interpretation as octal - number. Internally, the value will be stored as decimal number. For - instance, the C++ integer literal `010` will be serialized to `8`. - During deserialization, leading zeros yield an error. - - Not-a-number (NaN) values will be serialized to `null`. - - #### Limits - - [RFC 7159](http://rfc7159.net/rfc7159) specifies: - > An implementation may set limits on the range and precision of numbers. - - When the default type is used, the maximal integer number that can be - stored is `18446744073709551615` (UINT64_MAX) and the minimal integer - number that can be stored is `0`. Integer numbers that are out of range - will yield over/underflow when used in a constructor. During - deserialization, too large or small integer numbers will be automatically - be stored as @ref number_integer_t or @ref number_float_t. - - [RFC 7159](http://rfc7159.net/rfc7159) further states: - > Note that when such software is used, numbers that are integers and are - > in the range \f$[-2^{53}+1, 2^{53}-1]\f$ are interoperable in the sense - > that implementations will agree exactly on their numeric values. - - As this range is a subrange (when considered in conjunction with the - number_integer_t type) of the exactly supported range [0, UINT64_MAX], - this class's integer type is interoperable. - - #### Storage - - Integer number values are stored directly inside a @ref basic_json type. - - @sa @ref number_float_t -- type for number values (floating-point) - @sa @ref number_integer_t -- type for number values (integer) - - @since version 2.0.0 - */ - using number_unsigned_t = NumberUnsignedType; - - /*! - @brief a type for a number (floating-point) - - [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows: - > The representation of numbers is similar to that used in most - > programming languages. A number is represented in base 10 using decimal - > digits. It contains an integer component that may be prefixed with an - > optional minus sign, which may be followed by a fraction part and/or an - > exponent part. Leading zeros are not allowed. (...) Numeric values that - > cannot be represented in the grammar below (such as Infinity and NaN) - > are not permitted. - - This description includes both integer and floating-point numbers. - However, C++ allows more precise storage if it is known whether the number - is a signed integer, an unsigned integer or a floating-point number. - Therefore, three different types, @ref number_integer_t, @ref - number_unsigned_t and @ref number_float_t are used. - - To store floating-point numbers in C++, a type is defined by the template - parameter @a NumberFloatType which chooses the type to use. - - #### Default type - - With the default values for @a NumberFloatType (`double`), the default - value for @a number_float_t is: - - @code {.cpp} - double - @endcode - - #### Default behavior - - - The restrictions about leading zeros is not enforced in C++. Instead, - leading zeros in floating-point literals will be ignored. Internally, - the value will be stored as decimal number. For instance, the C++ - floating-point literal `01.2` will be serialized to `1.2`. During - deserialization, leading zeros yield an error. - - Not-a-number (NaN) values will be serialized to `null`. - - #### Limits - - [RFC 7159](http://rfc7159.net/rfc7159) states: - > This specification allows implementations to set limits on the range and - > precision of numbers accepted. Since software that implements IEEE - > 754-2008 binary64 (double precision) numbers is generally available and - > widely used, good interoperability can be achieved by implementations - > that expect no more precision or range than these provide, in the sense - > that implementations will approximate JSON numbers within the expected - > precision. - - This implementation does exactly follow this approach, as it uses double - precision floating-point numbers. Note values smaller than - `-1.79769313486232e+308` and values greater than `1.79769313486232e+308` - will be stored as NaN internally and be serialized to `null`. - - #### Storage - - Floating-point number values are stored directly inside a @ref basic_json - type. - - @sa @ref number_integer_t -- type for number values (integer) - - @sa @ref number_unsigned_t -- type for number values (unsigned integer) - - @since version 1.0.0 - */ - using number_float_t = NumberFloatType; - - /*! - @brief a type for a packed binary type - - This type is a type designed to carry binary data that appears in various - serialized formats, such as CBOR's Major Type 2, MessagePack's bin, and - BSON's generic binary subtype. This type is NOT a part of standard JSON and - exists solely for compatibility with these binary types. As such, it is - simply defined as an ordered sequence of zero or more byte values. - - Additionally, as an implementation detail, the subtype of the binary data is - carried around as a `std::uint8_t`, which is compatible with both of the - binary data formats that use binary subtyping, (though the specific - numbering is incompatible with each other, and it is up to the user to - translate between them). - - [CBOR's RFC 7049](https://tools.ietf.org/html/rfc7049) describes this type - as: - > Major type 2: a byte string. The string's length in bytes is represented - > following the rules for positive integers (major type 0). - - [MessagePack's documentation on the bin type - family](https://github.com/msgpack/msgpack/blob/master/spec.md#bin-format-family) - describes this type as: - > Bin format family stores an byte array in 2, 3, or 5 bytes of extra bytes - > in addition to the size of the byte array. - - [BSON's specifications](http://bsonspec.org/spec.html) describe several - binary types; however, this type is intended to represent the generic binary - type which has the description: - > Generic binary subtype - This is the most commonly used binary subtype and - > should be the 'default' for drivers and tools. - - None of these impose any limitations on the internal representation other - than the basic unit of storage be some type of array whose parts are - decomposable into bytes. - - The default representation of this binary format is a - `std::vector`, which is a very common way to represent a byte - array in modern C++. - - #### Default type - - The default values for @a BinaryType is `std::vector` - - #### Storage - - Binary Arrays are stored as pointers in a @ref basic_json type. That is, - for any access to array values, a pointer of the type `binary_t*` must be - dereferenced. - - #### Notes on subtypes - - - CBOR - - Binary values are represented as byte strings. No subtypes are - supported and will be ignored when CBOR is written. - - MessagePack - - If a subtype is given and the binary array contains exactly 1, 2, 4, 8, - or 16 elements, the fixext family (fixext1, fixext2, fixext4, fixext8) - is used. For other sizes, the ext family (ext8, ext16, ext32) is used. - The subtype is then added as singed 8-bit integer. - - If no subtype is given, the bin family (bin8, bin16, bin32) is used. - - BSON - - If a subtype is given, it is used and added as unsigned 8-bit integer. - - If no subtype is given, the generic binary subtype 0x00 is used. - - @sa @ref binary -- create a binary array - - @since version 3.8.0 - */ - using binary_t = nlohmann::byte_container_with_subtype; - /// @} - - private: - - /// helper for exception-safe object creation - template - JSON_HEDLEY_RETURNS_NON_NULL - static T* create(Args&& ... args) - { - AllocatorType alloc; - using AllocatorTraits = std::allocator_traits>; - - auto deleter = [&](T * object) - { - AllocatorTraits::deallocate(alloc, object, 1); - }; - std::unique_ptr object(AllocatorTraits::allocate(alloc, 1), deleter); - AllocatorTraits::construct(alloc, object.get(), std::forward(args)...); - JSON_ASSERT(object != nullptr); - return object.release(); - } - - //////////////////////// - // JSON value storage // - //////////////////////// - - /*! - @brief a JSON value - - The actual storage for a JSON value of the @ref basic_json class. This - union combines the different storage types for the JSON value types - defined in @ref value_t. - - JSON type | value_t type | used type - --------- | --------------- | ------------------------ - object | object | pointer to @ref object_t - array | array | pointer to @ref array_t - string | string | pointer to @ref string_t - boolean | boolean | @ref boolean_t - number | number_integer | @ref number_integer_t - number | number_unsigned | @ref number_unsigned_t - number | number_float | @ref number_float_t - binary | binary | pointer to @ref binary_t - null | null | *no value is stored* - - @note Variable-length types (objects, arrays, and strings) are stored as - pointers. The size of the union should not exceed 64 bits if the default - value types are used. - - @since version 1.0.0 - */ - union json_value - { - /// object (stored with pointer to save storage) - object_t* object; - /// array (stored with pointer to save storage) - array_t* array; - /// string (stored with pointer to save storage) - string_t* string; - /// binary (stored with pointer to save storage) - binary_t* binary; - /// boolean - boolean_t boolean; - /// number (integer) - number_integer_t number_integer; - /// number (unsigned integer) - number_unsigned_t number_unsigned; - /// number (floating-point) - number_float_t number_float; - - /// default constructor (for null values) - json_value() = default; - /// constructor for booleans - json_value(boolean_t v) noexcept : boolean(v) {} - /// constructor for numbers (integer) - json_value(number_integer_t v) noexcept : number_integer(v) {} - /// constructor for numbers (unsigned) - json_value(number_unsigned_t v) noexcept : number_unsigned(v) {} - /// constructor for numbers (floating-point) - json_value(number_float_t v) noexcept : number_float(v) {} - /// constructor for empty values of a given type - json_value(value_t t) - { - switch (t) - { - case value_t::object: - { - object = create(); - break; - } - - case value_t::array: - { - array = create(); - break; - } - - case value_t::string: - { - string = create(""); - break; - } - - case value_t::binary: - { - binary = create(); - break; - } - - case value_t::boolean: - { - boolean = boolean_t(false); - break; - } - - case value_t::number_integer: - { - number_integer = number_integer_t(0); - break; - } - - case value_t::number_unsigned: - { - number_unsigned = number_unsigned_t(0); - break; - } - - case value_t::number_float: - { - number_float = number_float_t(0.0); - break; - } - - case value_t::null: - { - object = nullptr; // silence warning, see #821 - break; - } - - default: - { - object = nullptr; // silence warning, see #821 - if (JSON_HEDLEY_UNLIKELY(t == value_t::null)) - { - JSON_THROW(other_error::create(500, "961c151d2e87f2686a955a9be24d316f1362bf21 3.9.1")); // LCOV_EXCL_LINE - } - break; - } - } - } - - /// constructor for strings - json_value(const string_t& value) - { - string = create(value); - } - - /// constructor for rvalue strings - json_value(string_t&& value) - { - string = create(std::move(value)); - } - - /// constructor for objects - json_value(const object_t& value) - { - object = create(value); - } - - /// constructor for rvalue objects - json_value(object_t&& value) - { - object = create(std::move(value)); - } - - /// constructor for arrays - json_value(const array_t& value) - { - array = create(value); - } - - /// constructor for rvalue arrays - json_value(array_t&& value) - { - array = create(std::move(value)); - } - - /// constructor for binary arrays - json_value(const typename binary_t::container_type& value) - { - binary = create(value); - } - - /// constructor for rvalue binary arrays - json_value(typename binary_t::container_type&& value) - { - binary = create(std::move(value)); - } - - /// constructor for binary arrays (internal type) - json_value(const binary_t& value) - { - binary = create(value); - } - - /// constructor for rvalue binary arrays (internal type) - json_value(binary_t&& value) - { - binary = create(std::move(value)); - } - - void destroy(value_t t) noexcept - { - // flatten the current json_value to a heap-allocated stack - std::vector stack; - - // move the top-level items to stack - if (t == value_t::array) - { - stack.reserve(array->size()); - std::move(array->begin(), array->end(), std::back_inserter(stack)); - } - else if (t == value_t::object) - { - stack.reserve(object->size()); - for (auto&& it : *object) - { - stack.push_back(std::move(it.second)); - } - } - - while (!stack.empty()) - { - // move the last item to local variable to be processed - basic_json current_item(std::move(stack.back())); - stack.pop_back(); - - // if current_item is array/object, move - // its children to the stack to be processed later - if (current_item.is_array()) - { - std::move(current_item.m_value.array->begin(), current_item.m_value.array->end(), - std::back_inserter(stack)); - - current_item.m_value.array->clear(); - } - else if (current_item.is_object()) - { - for (auto&& it : *current_item.m_value.object) - { - stack.push_back(std::move(it.second)); - } - - current_item.m_value.object->clear(); - } - - // it's now safe that current_item get destructed - // since it doesn't have any children - } - - switch (t) - { - case value_t::object: - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, object); - std::allocator_traits::deallocate(alloc, object, 1); - break; - } - - case value_t::array: - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, array); - std::allocator_traits::deallocate(alloc, array, 1); - break; - } - - case value_t::string: - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, string); - std::allocator_traits::deallocate(alloc, string, 1); - break; - } - - case value_t::binary: - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, binary); - std::allocator_traits::deallocate(alloc, binary, 1); - break; - } - - default: - { - break; - } - } - } - }; - - /*! - @brief checks the class invariants - - This function asserts the class invariants. It needs to be called at the - end of every constructor to make sure that created objects respect the - invariant. Furthermore, it has to be called each time the type of a JSON - value is changed, because the invariant expresses a relationship between - @a m_type and @a m_value. - */ - void assert_invariant() const noexcept - { - JSON_ASSERT(m_type != value_t::object || m_value.object != nullptr); - JSON_ASSERT(m_type != value_t::array || m_value.array != nullptr); - JSON_ASSERT(m_type != value_t::string || m_value.string != nullptr); - JSON_ASSERT(m_type != value_t::binary || m_value.binary != nullptr); - } - - public: - ////////////////////////// - // JSON parser callback // - ////////////////////////// - - /*! - @brief parser event types - - The parser callback distinguishes the following events: - - `object_start`: the parser read `{` and started to process a JSON object - - `key`: the parser read a key of a value in an object - - `object_end`: the parser read `}` and finished processing a JSON object - - `array_start`: the parser read `[` and started to process a JSON array - - `array_end`: the parser read `]` and finished processing a JSON array - - `value`: the parser finished reading a JSON value - - @image html callback_events.png "Example when certain parse events are triggered" - - @sa @ref parser_callback_t for more information and examples - */ - using parse_event_t = detail::parse_event_t; - - /*! - @brief per-element parser callback type - - With a parser callback function, the result of parsing a JSON text can be - influenced. When passed to @ref parse, it is called on certain events - (passed as @ref parse_event_t via parameter @a event) with a set recursion - depth @a depth and context JSON value @a parsed. The return value of the - callback function is a boolean indicating whether the element that emitted - the callback shall be kept or not. - - We distinguish six scenarios (determined by the event type) in which the - callback function can be called. The following table describes the values - of the parameters @a depth, @a event, and @a parsed. - - parameter @a event | description | parameter @a depth | parameter @a parsed - ------------------ | ----------- | ------------------ | ------------------- - parse_event_t::object_start | the parser read `{` and started to process a JSON object | depth of the parent of the JSON object | a JSON value with type discarded - parse_event_t::key | the parser read a key of a value in an object | depth of the currently parsed JSON object | a JSON string containing the key - parse_event_t::object_end | the parser read `}` and finished processing a JSON object | depth of the parent of the JSON object | the parsed JSON object - parse_event_t::array_start | the parser read `[` and started to process a JSON array | depth of the parent of the JSON array | a JSON value with type discarded - parse_event_t::array_end | the parser read `]` and finished processing a JSON array | depth of the parent of the JSON array | the parsed JSON array - parse_event_t::value | the parser finished reading a JSON value | depth of the value | the parsed JSON value - - @image html callback_events.png "Example when certain parse events are triggered" - - Discarding a value (i.e., returning `false`) has different effects - depending on the context in which function was called: - - - Discarded values in structured types are skipped. That is, the parser - will behave as if the discarded value was never read. - - In case a value outside a structured type is skipped, it is replaced - with `null`. This case happens if the top-level element is skipped. - - @param[in] depth the depth of the recursion during parsing - - @param[in] event an event of type parse_event_t indicating the context in - the callback function has been called - - @param[in,out] parsed the current intermediate parse result; note that - writing to this value has no effect for parse_event_t::key events - - @return Whether the JSON value which called the function during parsing - should be kept (`true`) or not (`false`). In the latter case, it is either - skipped completely or replaced by an empty discarded object. - - @sa @ref parse for examples - - @since version 1.0.0 - */ - using parser_callback_t = detail::parser_callback_t; - - ////////////////// - // constructors // - ////////////////// - - /// @name constructors and destructors - /// Constructors of class @ref basic_json, copy/move constructor, copy - /// assignment, static functions creating objects, and the destructor. - /// @{ - - /*! - @brief create an empty value with a given type - - Create an empty JSON value with a given type. The value will be default - initialized with an empty value which depends on the type: - - Value type | initial value - ----------- | ------------- - null | `null` - boolean | `false` - string | `""` - number | `0` - object | `{}` - array | `[]` - binary | empty array - - @param[in] v the type of the value to create - - @complexity Constant. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The following code shows the constructor for different @ref - value_t values,basic_json__value_t} - - @sa @ref clear() -- restores the postcondition of this constructor - - @since version 1.0.0 - */ - basic_json(const value_t v) - : m_type(v), m_value(v) - { - assert_invariant(); - } - - /*! - @brief create a null object - - Create a `null` JSON value. It either takes a null pointer as parameter - (explicitly creating `null`) or no parameter (implicitly creating `null`). - The passed null pointer itself is not read -- it is only used to choose - the right constructor. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this constructor never throws - exceptions. - - @liveexample{The following code shows the constructor with and without a - null pointer parameter.,basic_json__nullptr_t} - - @since version 1.0.0 - */ - basic_json(std::nullptr_t = nullptr) noexcept - : basic_json(value_t::null) - { - assert_invariant(); - } - - /*! - @brief create a JSON value - - This is a "catch all" constructor for all compatible JSON types; that is, - types for which a `to_json()` method exists. The constructor forwards the - parameter @a val to that method (to `json_serializer::to_json` method - with `U = uncvref_t`, to be exact). - - Template type @a CompatibleType includes, but is not limited to, the - following types: - - **arrays**: @ref array_t and all kinds of compatible containers such as - `std::vector`, `std::deque`, `std::list`, `std::forward_list`, - `std::array`, `std::valarray`, `std::set`, `std::unordered_set`, - `std::multiset`, and `std::unordered_multiset` with a `value_type` from - which a @ref basic_json value can be constructed. - - **objects**: @ref object_t and all kinds of compatible associative - containers such as `std::map`, `std::unordered_map`, `std::multimap`, - and `std::unordered_multimap` with a `key_type` compatible to - @ref string_t and a `value_type` from which a @ref basic_json value can - be constructed. - - **strings**: @ref string_t, string literals, and all compatible string - containers can be used. - - **numbers**: @ref number_integer_t, @ref number_unsigned_t, - @ref number_float_t, and all convertible number types such as `int`, - `size_t`, `int64_t`, `float` or `double` can be used. - - **boolean**: @ref boolean_t / `bool` can be used. - - **binary**: @ref binary_t / `std::vector` may be used, - unfortunately because string literals cannot be distinguished from binary - character arrays by the C++ type system, all types compatible with `const - char*` will be directed to the string constructor instead. This is both - for backwards compatibility, and due to the fact that a binary type is not - a standard JSON type. - - See the examples below. - - @tparam CompatibleType a type such that: - - @a CompatibleType is not derived from `std::istream`, - - @a CompatibleType is not @ref basic_json (to avoid hijacking copy/move - constructors), - - @a CompatibleType is not a different @ref basic_json type (i.e. with different template arguments) - - @a CompatibleType is not a @ref basic_json nested type (e.g., - @ref json_pointer, @ref iterator, etc ...) - - @ref @ref json_serializer has a - `to_json(basic_json_t&, CompatibleType&&)` method - - @tparam U = `uncvref_t` - - @param[in] val the value to be forwarded to the respective constructor - - @complexity Usually linear in the size of the passed @a val, also - depending on the implementation of the called `to_json()` - method. - - @exceptionsafety Depends on the called constructor. For types directly - supported by the library (i.e., all types for which no `to_json()` function - was provided), strong guarantee holds: if an exception is thrown, there are - no changes to any JSON value. - - @liveexample{The following code shows the constructor with several - compatible types.,basic_json__CompatibleType} - - @since version 2.1.0 - */ - template < typename CompatibleType, - typename U = detail::uncvref_t, - detail::enable_if_t < - !detail::is_basic_json::value && detail::is_compatible_type::value, int > = 0 > - basic_json(CompatibleType && val) noexcept(noexcept( - JSONSerializer::to_json(std::declval(), - std::forward(val)))) - { - JSONSerializer::to_json(*this, std::forward(val)); - assert_invariant(); - } - - /*! - @brief create a JSON value from an existing one - - This is a constructor for existing @ref basic_json types. - It does not hijack copy/move constructors, since the parameter has different - template arguments than the current ones. - - The constructor tries to convert the internal @ref m_value of the parameter. - - @tparam BasicJsonType a type such that: - - @a BasicJsonType is a @ref basic_json type. - - @a BasicJsonType has different template arguments than @ref basic_json_t. - - @param[in] val the @ref basic_json value to be converted. - - @complexity Usually linear in the size of the passed @a val, also - depending on the implementation of the called `to_json()` - method. - - @exceptionsafety Depends on the called constructor. For types directly - supported by the library (i.e., all types for which no `to_json()` function - was provided), strong guarantee holds: if an exception is thrown, there are - no changes to any JSON value. - - @since version 3.2.0 - */ - template < typename BasicJsonType, - detail::enable_if_t < - detail::is_basic_json::value&& !std::is_same::value, int > = 0 > - basic_json(const BasicJsonType& val) - { - using other_boolean_t = typename BasicJsonType::boolean_t; - using other_number_float_t = typename BasicJsonType::number_float_t; - using other_number_integer_t = typename BasicJsonType::number_integer_t; - using other_number_unsigned_t = typename BasicJsonType::number_unsigned_t; - using other_string_t = typename BasicJsonType::string_t; - using other_object_t = typename BasicJsonType::object_t; - using other_array_t = typename BasicJsonType::array_t; - using other_binary_t = typename BasicJsonType::binary_t; - - switch (val.type()) - { - case value_t::boolean: - JSONSerializer::to_json(*this, val.template get()); - break; - case value_t::number_float: - JSONSerializer::to_json(*this, val.template get()); - break; - case value_t::number_integer: - JSONSerializer::to_json(*this, val.template get()); - break; - case value_t::number_unsigned: - JSONSerializer::to_json(*this, val.template get()); - break; - case value_t::string: - JSONSerializer::to_json(*this, val.template get_ref()); - break; - case value_t::object: - JSONSerializer::to_json(*this, val.template get_ref()); - break; - case value_t::array: - JSONSerializer::to_json(*this, val.template get_ref()); - break; - case value_t::binary: - JSONSerializer::to_json(*this, val.template get_ref()); - break; - case value_t::null: - *this = nullptr; - break; - case value_t::discarded: - m_type = value_t::discarded; - break; - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - assert_invariant(); - } - - /*! - @brief create a container (array or object) from an initializer list - - Creates a JSON value of type array or object from the passed initializer - list @a init. In case @a type_deduction is `true` (default), the type of - the JSON value to be created is deducted from the initializer list @a init - according to the following rules: - - 1. If the list is empty, an empty JSON object value `{}` is created. - 2. If the list consists of pairs whose first element is a string, a JSON - object value is created where the first elements of the pairs are - treated as keys and the second elements are as values. - 3. In all other cases, an array is created. - - The rules aim to create the best fit between a C++ initializer list and - JSON values. The rationale is as follows: - - 1. The empty initializer list is written as `{}` which is exactly an empty - JSON object. - 2. C++ has no way of describing mapped types other than to list a list of - pairs. As JSON requires that keys must be of type string, rule 2 is the - weakest constraint one can pose on initializer lists to interpret them - as an object. - 3. In all other cases, the initializer list could not be interpreted as - JSON object type, so interpreting it as JSON array type is safe. - - With the rules described above, the following JSON values cannot be - expressed by an initializer list: - - - the empty array (`[]`): use @ref array(initializer_list_t) - with an empty initializer list in this case - - arrays whose elements satisfy rule 2: use @ref - array(initializer_list_t) with the same initializer list - in this case - - @note When used without parentheses around an empty initializer list, @ref - basic_json() is called instead of this function, yielding the JSON null - value. - - @param[in] init initializer list with JSON values - - @param[in] type_deduction internal parameter; when set to `true`, the type - of the JSON value is deducted from the initializer list @a init; when set - to `false`, the type provided via @a manual_type is forced. This mode is - used by the functions @ref array(initializer_list_t) and - @ref object(initializer_list_t). - - @param[in] manual_type internal parameter; when @a type_deduction is set - to `false`, the created JSON value will use the provided type (only @ref - value_t::array and @ref value_t::object are valid); when @a type_deduction - is set to `true`, this parameter has no effect - - @throw type_error.301 if @a type_deduction is `false`, @a manual_type is - `value_t::object`, but @a init contains an element which is not a pair - whose first element is a string. In this case, the constructor could not - create an object. If @a type_deduction would have be `true`, an array - would have been created. See @ref object(initializer_list_t) - for an example. - - @complexity Linear in the size of the initializer list @a init. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The example below shows how JSON values are created from - initializer lists.,basic_json__list_init_t} - - @sa @ref array(initializer_list_t) -- create a JSON array - value from an initializer list - @sa @ref object(initializer_list_t) -- create a JSON object - value from an initializer list - - @since version 1.0.0 - */ - basic_json(initializer_list_t init, - bool type_deduction = true, - value_t manual_type = value_t::array) - { - // check if each element is an array with two elements whose first - // element is a string - bool is_an_object = std::all_of(init.begin(), init.end(), - [](const detail::json_ref& element_ref) - { - return element_ref->is_array() && element_ref->size() == 2 && (*element_ref)[0].is_string(); - }); - - // adjust type if type deduction is not wanted - if (!type_deduction) - { - // if array is wanted, do not create an object though possible - if (manual_type == value_t::array) - { - is_an_object = false; - } - - // if object is wanted but impossible, throw an exception - if (JSON_HEDLEY_UNLIKELY(manual_type == value_t::object && !is_an_object)) - { - JSON_THROW(type_error::create(301, "cannot create object from initializer list")); - } - } - - if (is_an_object) - { - // the initializer list is a list of pairs -> create object - m_type = value_t::object; - m_value = value_t::object; - - std::for_each(init.begin(), init.end(), [this](const detail::json_ref& element_ref) - { - auto element = element_ref.moved_or_copied(); - m_value.object->emplace( - std::move(*((*element.m_value.array)[0].m_value.string)), - std::move((*element.m_value.array)[1])); - }); - } - else - { - // the initializer list describes an array -> create array - m_type = value_t::array; - m_value.array = create(init.begin(), init.end()); - } - - assert_invariant(); - } - - /*! - @brief explicitly create a binary array (without subtype) - - Creates a JSON binary array value from a given binary container. Binary - values are part of various binary formats, such as CBOR, MessagePack, and - BSON. This constructor is used to create a value for serialization to those - formats. - - @note Note, this function exists because of the difficulty in correctly - specifying the correct template overload in the standard value ctor, as both - JSON arrays and JSON binary arrays are backed with some form of a - `std::vector`. Because JSON binary arrays are a non-standard extension it - was decided that it would be best to prevent automatic initialization of a - binary array type, for backwards compatibility and so it does not happen on - accident. - - @param[in] init container containing bytes to use as binary type - - @return JSON binary array value - - @complexity Linear in the size of @a init. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @since version 3.8.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json binary(const typename binary_t::container_type& init) - { - auto res = basic_json(); - res.m_type = value_t::binary; - res.m_value = init; - return res; - } - - /*! - @brief explicitly create a binary array (with subtype) - - Creates a JSON binary array value from a given binary container. Binary - values are part of various binary formats, such as CBOR, MessagePack, and - BSON. This constructor is used to create a value for serialization to those - formats. - - @note Note, this function exists because of the difficulty in correctly - specifying the correct template overload in the standard value ctor, as both - JSON arrays and JSON binary arrays are backed with some form of a - `std::vector`. Because JSON binary arrays are a non-standard extension it - was decided that it would be best to prevent automatic initialization of a - binary array type, for backwards compatibility and so it does not happen on - accident. - - @param[in] init container containing bytes to use as binary type - @param[in] subtype subtype to use in MessagePack and BSON - - @return JSON binary array value - - @complexity Linear in the size of @a init. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @since version 3.8.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json binary(const typename binary_t::container_type& init, std::uint8_t subtype) - { - auto res = basic_json(); - res.m_type = value_t::binary; - res.m_value = binary_t(init, subtype); - return res; - } - - /// @copydoc binary(const typename binary_t::container_type&) - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json binary(typename binary_t::container_type&& init) - { - auto res = basic_json(); - res.m_type = value_t::binary; - res.m_value = std::move(init); - return res; - } - - /// @copydoc binary(const typename binary_t::container_type&, std::uint8_t) - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json binary(typename binary_t::container_type&& init, std::uint8_t subtype) - { - auto res = basic_json(); - res.m_type = value_t::binary; - res.m_value = binary_t(std::move(init), subtype); - return res; - } - - /*! - @brief explicitly create an array from an initializer list - - Creates a JSON array value from a given initializer list. That is, given a - list of values `a, b, c`, creates the JSON value `[a, b, c]`. If the - initializer list is empty, the empty array `[]` is created. - - @note This function is only needed to express two edge cases that cannot - be realized with the initializer list constructor (@ref - basic_json(initializer_list_t, bool, value_t)). These cases - are: - 1. creating an array whose elements are all pairs whose first element is a - string -- in this case, the initializer list constructor would create an - object, taking the first elements as keys - 2. creating an empty array -- passing the empty initializer list to the - initializer list constructor yields an empty object - - @param[in] init initializer list with JSON values to create an array from - (optional) - - @return JSON array value - - @complexity Linear in the size of @a init. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The following code shows an example for the `array` - function.,array} - - @sa @ref basic_json(initializer_list_t, bool, value_t) -- - create a JSON value from an initializer list - @sa @ref object(initializer_list_t) -- create a JSON object - value from an initializer list - - @since version 1.0.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json array(initializer_list_t init = {}) - { - return basic_json(init, false, value_t::array); - } - - /*! - @brief explicitly create an object from an initializer list - - Creates a JSON object value from a given initializer list. The initializer - lists elements must be pairs, and their first elements must be strings. If - the initializer list is empty, the empty object `{}` is created. - - @note This function is only added for symmetry reasons. In contrast to the - related function @ref array(initializer_list_t), there are - no cases which can only be expressed by this function. That is, any - initializer list @a init can also be passed to the initializer list - constructor @ref basic_json(initializer_list_t, bool, value_t). - - @param[in] init initializer list to create an object from (optional) - - @return JSON object value - - @throw type_error.301 if @a init is not a list of pairs whose first - elements are strings. In this case, no object can be created. When such a - value is passed to @ref basic_json(initializer_list_t, bool, value_t), - an array would have been created from the passed initializer list @a init. - See example below. - - @complexity Linear in the size of @a init. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The following code shows an example for the `object` - function.,object} - - @sa @ref basic_json(initializer_list_t, bool, value_t) -- - create a JSON value from an initializer list - @sa @ref array(initializer_list_t) -- create a JSON array - value from an initializer list - - @since version 1.0.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json object(initializer_list_t init = {}) - { - return basic_json(init, false, value_t::object); - } - - /*! - @brief construct an array with count copies of given value - - Constructs a JSON array value by creating @a cnt copies of a passed value. - In case @a cnt is `0`, an empty array is created. - - @param[in] cnt the number of JSON copies of @a val to create - @param[in] val the JSON value to copy - - @post `std::distance(begin(),end()) == cnt` holds. - - @complexity Linear in @a cnt. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The following code shows examples for the @ref - basic_json(size_type\, const basic_json&) - constructor.,basic_json__size_type_basic_json} - - @since version 1.0.0 - */ - basic_json(size_type cnt, const basic_json& val) - : m_type(value_t::array) - { - m_value.array = create(cnt, val); - assert_invariant(); - } - - /*! - @brief construct a JSON container given an iterator range - - Constructs the JSON value with the contents of the range `[first, last)`. - The semantics depends on the different types a JSON value can have: - - In case of a null type, invalid_iterator.206 is thrown. - - In case of other primitive types (number, boolean, or string), @a first - must be `begin()` and @a last must be `end()`. In this case, the value is - copied. Otherwise, invalid_iterator.204 is thrown. - - In case of structured types (array, object), the constructor behaves as - similar versions for `std::vector` or `std::map`; that is, a JSON array - or object is constructed from the values in the range. - - @tparam InputIT an input iterator type (@ref iterator or @ref - const_iterator) - - @param[in] first begin of the range to copy from (included) - @param[in] last end of the range to copy from (excluded) - - @pre Iterators @a first and @a last must be initialized. **This - precondition is enforced with an assertion (see warning).** If - assertions are switched off, a violation of this precondition yields - undefined behavior. - - @pre Range `[first, last)` is valid. Usually, this precondition cannot be - checked efficiently. Only certain edge cases are detected; see the - description of the exceptions below. A violation of this precondition - yields undefined behavior. - - @warning A precondition is enforced with a runtime assertion that will - result in calling `std::abort` if this precondition is not met. - Assertions can be disabled by defining `NDEBUG` at compile time. - See https://en.cppreference.com/w/cpp/error/assert for more - information. - - @throw invalid_iterator.201 if iterators @a first and @a last are not - compatible (i.e., do not belong to the same JSON value). In this case, - the range `[first, last)` is undefined. - @throw invalid_iterator.204 if iterators @a first and @a last belong to a - primitive type (number, boolean, or string), but @a first does not point - to the first element any more. In this case, the range `[first, last)` is - undefined. See example code below. - @throw invalid_iterator.206 if iterators @a first and @a last belong to a - null value. In this case, the range `[first, last)` is undefined. - - @complexity Linear in distance between @a first and @a last. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @liveexample{The example below shows several ways to create JSON values by - specifying a subrange with iterators.,basic_json__InputIt_InputIt} - - @since version 1.0.0 - */ - template < class InputIT, typename std::enable_if < - std::is_same::value || - std::is_same::value, int >::type = 0 > - basic_json(InputIT first, InputIT last) - { - JSON_ASSERT(first.m_object != nullptr); - JSON_ASSERT(last.m_object != nullptr); - - // make sure iterator fits the current value - if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object)) - { - JSON_THROW(invalid_iterator::create(201, "iterators are not compatible")); - } - - // copy type from first iterator - m_type = first.m_object->m_type; - - // check if iterator range is complete for primitive values - switch (m_type) - { - case value_t::boolean: - case value_t::number_float: - case value_t::number_integer: - case value_t::number_unsigned: - case value_t::string: - { - if (JSON_HEDLEY_UNLIKELY(!first.m_it.primitive_iterator.is_begin() - || !last.m_it.primitive_iterator.is_end())) - { - JSON_THROW(invalid_iterator::create(204, "iterators out of range")); - } - break; - } - - default: - break; - } - - switch (m_type) - { - case value_t::number_integer: - { - m_value.number_integer = first.m_object->m_value.number_integer; - break; - } - - case value_t::number_unsigned: - { - m_value.number_unsigned = first.m_object->m_value.number_unsigned; - break; - } - - case value_t::number_float: - { - m_value.number_float = first.m_object->m_value.number_float; - break; - } - - case value_t::boolean: - { - m_value.boolean = first.m_object->m_value.boolean; - break; - } - - case value_t::string: - { - m_value = *first.m_object->m_value.string; - break; - } - - case value_t::object: - { - m_value.object = create(first.m_it.object_iterator, - last.m_it.object_iterator); - break; - } - - case value_t::array: - { - m_value.array = create(first.m_it.array_iterator, - last.m_it.array_iterator); - break; - } - - case value_t::binary: - { - m_value = *first.m_object->m_value.binary; - break; - } - - default: - JSON_THROW(invalid_iterator::create(206, "cannot construct with iterators from " + - std::string(first.m_object->type_name()))); - } - - assert_invariant(); - } - - - /////////////////////////////////////// - // other constructors and destructor // - /////////////////////////////////////// - - template, - std::is_same>::value, int> = 0 > - basic_json(const JsonRef& ref) : basic_json(ref.moved_or_copied()) {} - - /*! - @brief copy constructor - - Creates a copy of a given JSON value. - - @param[in] other the JSON value to copy - - @post `*this == other` - - @complexity Linear in the size of @a other. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes to any JSON value. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is linear. - - As postcondition, it holds: `other == basic_json(other)`. - - @liveexample{The following code shows an example for the copy - constructor.,basic_json__basic_json} - - @since version 1.0.0 - */ - basic_json(const basic_json& other) - : m_type(other.m_type) - { - // check of passed value is valid - other.assert_invariant(); - - switch (m_type) - { - case value_t::object: - { - m_value = *other.m_value.object; - break; - } - - case value_t::array: - { - m_value = *other.m_value.array; - break; - } - - case value_t::string: - { - m_value = *other.m_value.string; - break; - } - - case value_t::boolean: - { - m_value = other.m_value.boolean; - break; - } - - case value_t::number_integer: - { - m_value = other.m_value.number_integer; - break; - } - - case value_t::number_unsigned: - { - m_value = other.m_value.number_unsigned; - break; - } - - case value_t::number_float: - { - m_value = other.m_value.number_float; - break; - } - - case value_t::binary: - { - m_value = *other.m_value.binary; - break; - } - - default: - break; - } - - assert_invariant(); - } - - /*! - @brief move constructor - - Move constructor. Constructs a JSON value with the contents of the given - value @a other using move semantics. It "steals" the resources from @a - other and leaves it as JSON null value. - - @param[in,out] other value to move to this object - - @post `*this` has the same value as @a other before the call. - @post @a other is a JSON null value. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this constructor never throws - exceptions. - - @requirement This function helps `basic_json` satisfying the - [MoveConstructible](https://en.cppreference.com/w/cpp/named_req/MoveConstructible) - requirements. - - @liveexample{The code below shows the move constructor explicitly called - via std::move.,basic_json__moveconstructor} - - @since version 1.0.0 - */ - basic_json(basic_json&& other) noexcept - : m_type(std::move(other.m_type)), - m_value(std::move(other.m_value)) - { - // check that passed value is valid - other.assert_invariant(); - - // invalidate payload - other.m_type = value_t::null; - other.m_value = {}; - - assert_invariant(); - } - - /*! - @brief copy assignment - - Copy assignment operator. Copies a JSON value via the "copy and swap" - strategy: It is expressed in terms of the copy constructor, destructor, - and the `swap()` member function. - - @param[in] other value to copy from - - @complexity Linear. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is linear. - - @liveexample{The code below shows and example for the copy assignment. It - creates a copy of value `a` which is then swapped with `b`. Finally\, the - copy of `a` (which is the null value after the swap) is - destroyed.,basic_json__copyassignment} - - @since version 1.0.0 - */ - basic_json& operator=(basic_json other) noexcept ( - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value&& - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value - ) - { - // check that passed value is valid - other.assert_invariant(); - - using std::swap; - swap(m_type, other.m_type); - swap(m_value, other.m_value); - - assert_invariant(); - return *this; - } - - /*! - @brief destructor - - Destroys the JSON value and frees all allocated memory. - - @complexity Linear. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is linear. - - All stored elements are destroyed and all memory is freed. - - @since version 1.0.0 - */ - ~basic_json() noexcept - { - assert_invariant(); - m_value.destroy(m_type); - } - - /// @} - - public: - /////////////////////// - // object inspection // - /////////////////////// - - /// @name object inspection - /// Functions to inspect the type of a JSON value. - /// @{ - - /*! - @brief serialization - - Serialization function for JSON values. The function tries to mimic - Python's `json.dumps()` function, and currently supports its @a indent - and @a ensure_ascii parameters. - - @param[in] indent If indent is nonnegative, then array elements and object - members will be pretty-printed with that indent level. An indent level of - `0` will only insert newlines. `-1` (the default) selects the most compact - representation. - @param[in] indent_char The character to use for indentation if @a indent is - greater than `0`. The default is ` ` (space). - @param[in] ensure_ascii If @a ensure_ascii is true, all non-ASCII characters - in the output are escaped with `\uXXXX` sequences, and the result consists - of ASCII characters only. - @param[in] error_handler how to react on decoding errors; there are three - possible values: `strict` (throws and exception in case a decoding error - occurs; default), `replace` (replace invalid UTF-8 sequences with U+FFFD), - and `ignore` (ignore invalid UTF-8 sequences during serialization; all - bytes are copied to the output unchanged). - - @return string containing the serialization of the JSON value - - @throw type_error.316 if a string stored inside the JSON value is not - UTF-8 encoded and @a error_handler is set to strict - - @note Binary values are serialized as object containing two keys: - - "bytes": an array of bytes as integers - - "subtype": the subtype as integer or "null" if the binary has no subtype - - @complexity Linear. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @liveexample{The following example shows the effect of different @a indent\, - @a indent_char\, and @a ensure_ascii parameters to the result of the - serialization.,dump} - - @see https://docs.python.org/2/library/json.html#json.dump - - @since version 1.0.0; indentation character @a indent_char, option - @a ensure_ascii and exceptions added in version 3.0.0; error - handlers added in version 3.4.0; serialization of binary values added - in version 3.8.0. - */ - string_t dump(const int indent = -1, - const char indent_char = ' ', - const bool ensure_ascii = false, - const error_handler_t error_handler = error_handler_t::strict) const - { - string_t result; - serializer s(detail::output_adapter(result), indent_char, error_handler); - - if (indent >= 0) - { - s.dump(*this, true, ensure_ascii, static_cast(indent)); - } - else - { - s.dump(*this, false, ensure_ascii, 0); - } - - return result; - } - - /*! - @brief return the type of the JSON value (explicit) - - Return the type of the JSON value as a value from the @ref value_t - enumeration. - - @return the type of the JSON value - Value type | return value - ------------------------- | ------------------------- - null | value_t::null - boolean | value_t::boolean - string | value_t::string - number (integer) | value_t::number_integer - number (unsigned integer) | value_t::number_unsigned - number (floating-point) | value_t::number_float - object | value_t::object - array | value_t::array - binary | value_t::binary - discarded | value_t::discarded - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `type()` for all JSON - types.,type} - - @sa @ref operator value_t() -- return the type of the JSON value (implicit) - @sa @ref type_name() -- return the type as string - - @since version 1.0.0 - */ - constexpr value_t type() const noexcept - { - return m_type; - } - - /*! - @brief return whether type is primitive - - This function returns true if and only if the JSON type is primitive - (string, number, boolean, or null). - - @return `true` if type is primitive (string, number, boolean, or null), - `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_primitive()` for all JSON - types.,is_primitive} - - @sa @ref is_structured() -- returns whether JSON value is structured - @sa @ref is_null() -- returns whether JSON value is `null` - @sa @ref is_string() -- returns whether JSON value is a string - @sa @ref is_boolean() -- returns whether JSON value is a boolean - @sa @ref is_number() -- returns whether JSON value is a number - @sa @ref is_binary() -- returns whether JSON value is a binary array - - @since version 1.0.0 - */ - constexpr bool is_primitive() const noexcept - { - return is_null() || is_string() || is_boolean() || is_number() || is_binary(); - } - - /*! - @brief return whether type is structured - - This function returns true if and only if the JSON type is structured - (array or object). - - @return `true` if type is structured (array or object), `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_structured()` for all JSON - types.,is_structured} - - @sa @ref is_primitive() -- returns whether value is primitive - @sa @ref is_array() -- returns whether value is an array - @sa @ref is_object() -- returns whether value is an object - - @since version 1.0.0 - */ - constexpr bool is_structured() const noexcept - { - return is_array() || is_object(); - } - - /*! - @brief return whether value is null - - This function returns true if and only if the JSON value is null. - - @return `true` if type is null, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_null()` for all JSON - types.,is_null} - - @since version 1.0.0 - */ - constexpr bool is_null() const noexcept - { - return m_type == value_t::null; - } - - /*! - @brief return whether value is a boolean - - This function returns true if and only if the JSON value is a boolean. - - @return `true` if type is boolean, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_boolean()` for all JSON - types.,is_boolean} - - @since version 1.0.0 - */ - constexpr bool is_boolean() const noexcept - { - return m_type == value_t::boolean; - } - - /*! - @brief return whether value is a number - - This function returns true if and only if the JSON value is a number. This - includes both integer (signed and unsigned) and floating-point values. - - @return `true` if type is number (regardless whether integer, unsigned - integer or floating-type), `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_number()` for all JSON - types.,is_number} - - @sa @ref is_number_integer() -- check if value is an integer or unsigned - integer number - @sa @ref is_number_unsigned() -- check if value is an unsigned integer - number - @sa @ref is_number_float() -- check if value is a floating-point number - - @since version 1.0.0 - */ - constexpr bool is_number() const noexcept - { - return is_number_integer() || is_number_float(); - } - - /*! - @brief return whether value is an integer number - - This function returns true if and only if the JSON value is a signed or - unsigned integer number. This excludes floating-point values. - - @return `true` if type is an integer or unsigned integer number, `false` - otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_number_integer()` for all - JSON types.,is_number_integer} - - @sa @ref is_number() -- check if value is a number - @sa @ref is_number_unsigned() -- check if value is an unsigned integer - number - @sa @ref is_number_float() -- check if value is a floating-point number - - @since version 1.0.0 - */ - constexpr bool is_number_integer() const noexcept - { - return m_type == value_t::number_integer || m_type == value_t::number_unsigned; - } - - /*! - @brief return whether value is an unsigned integer number - - This function returns true if and only if the JSON value is an unsigned - integer number. This excludes floating-point and signed integer values. - - @return `true` if type is an unsigned integer number, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_number_unsigned()` for all - JSON types.,is_number_unsigned} - - @sa @ref is_number() -- check if value is a number - @sa @ref is_number_integer() -- check if value is an integer or unsigned - integer number - @sa @ref is_number_float() -- check if value is a floating-point number - - @since version 2.0.0 - */ - constexpr bool is_number_unsigned() const noexcept - { - return m_type == value_t::number_unsigned; - } - - /*! - @brief return whether value is a floating-point number - - This function returns true if and only if the JSON value is a - floating-point number. This excludes signed and unsigned integer values. - - @return `true` if type is a floating-point number, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_number_float()` for all - JSON types.,is_number_float} - - @sa @ref is_number() -- check if value is number - @sa @ref is_number_integer() -- check if value is an integer number - @sa @ref is_number_unsigned() -- check if value is an unsigned integer - number - - @since version 1.0.0 - */ - constexpr bool is_number_float() const noexcept - { - return m_type == value_t::number_float; - } - - /*! - @brief return whether value is an object - - This function returns true if and only if the JSON value is an object. - - @return `true` if type is object, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_object()` for all JSON - types.,is_object} - - @since version 1.0.0 - */ - constexpr bool is_object() const noexcept - { - return m_type == value_t::object; - } - - /*! - @brief return whether value is an array - - This function returns true if and only if the JSON value is an array. - - @return `true` if type is array, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_array()` for all JSON - types.,is_array} - - @since version 1.0.0 - */ - constexpr bool is_array() const noexcept - { - return m_type == value_t::array; - } - - /*! - @brief return whether value is a string - - This function returns true if and only if the JSON value is a string. - - @return `true` if type is string, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_string()` for all JSON - types.,is_string} - - @since version 1.0.0 - */ - constexpr bool is_string() const noexcept - { - return m_type == value_t::string; - } - - /*! - @brief return whether value is a binary array - - This function returns true if and only if the JSON value is a binary array. - - @return `true` if type is binary array, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_binary()` for all JSON - types.,is_binary} - - @since version 3.8.0 - */ - constexpr bool is_binary() const noexcept - { - return m_type == value_t::binary; - } - - /*! - @brief return whether value is discarded - - This function returns true if and only if the JSON value was discarded - during parsing with a callback function (see @ref parser_callback_t). - - @note This function will always be `false` for JSON values after parsing. - That is, discarded values can only occur during parsing, but will be - removed when inside a structured value or replaced by null in other cases. - - @return `true` if type is discarded, `false` otherwise. - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies `is_discarded()` for all JSON - types.,is_discarded} - - @since version 1.0.0 - */ - constexpr bool is_discarded() const noexcept - { - return m_type == value_t::discarded; - } - - /*! - @brief return the type of the JSON value (implicit) - - Implicitly return the type of the JSON value as a value from the @ref - value_t enumeration. - - @return the type of the JSON value - - @complexity Constant. - - @exceptionsafety No-throw guarantee: this member function never throws - exceptions. - - @liveexample{The following code exemplifies the @ref value_t operator for - all JSON types.,operator__value_t} - - @sa @ref type() -- return the type of the JSON value (explicit) - @sa @ref type_name() -- return the type as string - - @since version 1.0.0 - */ - constexpr operator value_t() const noexcept - { - return m_type; - } - - /// @} - - private: - ////////////////// - // value access // - ////////////////// - - /// get a boolean (explicit) - boolean_t get_impl(boolean_t* /*unused*/) const - { - if (JSON_HEDLEY_LIKELY(is_boolean())) - { - return m_value.boolean; - } - - JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(type_name()))); - } - - /// get a pointer to the value (object) - object_t* get_impl_ptr(object_t* /*unused*/) noexcept - { - return is_object() ? m_value.object : nullptr; - } - - /// get a pointer to the value (object) - constexpr const object_t* get_impl_ptr(const object_t* /*unused*/) const noexcept - { - return is_object() ? m_value.object : nullptr; - } - - /// get a pointer to the value (array) - array_t* get_impl_ptr(array_t* /*unused*/) noexcept - { - return is_array() ? m_value.array : nullptr; - } - - /// get a pointer to the value (array) - constexpr const array_t* get_impl_ptr(const array_t* /*unused*/) const noexcept - { - return is_array() ? m_value.array : nullptr; - } - - /// get a pointer to the value (string) - string_t* get_impl_ptr(string_t* /*unused*/) noexcept - { - return is_string() ? m_value.string : nullptr; - } - - /// get a pointer to the value (string) - constexpr const string_t* get_impl_ptr(const string_t* /*unused*/) const noexcept - { - return is_string() ? m_value.string : nullptr; - } - - /// get a pointer to the value (boolean) - boolean_t* get_impl_ptr(boolean_t* /*unused*/) noexcept - { - return is_boolean() ? &m_value.boolean : nullptr; - } - - /// get a pointer to the value (boolean) - constexpr const boolean_t* get_impl_ptr(const boolean_t* /*unused*/) const noexcept - { - return is_boolean() ? &m_value.boolean : nullptr; - } - - /// get a pointer to the value (integer number) - number_integer_t* get_impl_ptr(number_integer_t* /*unused*/) noexcept - { - return is_number_integer() ? &m_value.number_integer : nullptr; - } - - /// get a pointer to the value (integer number) - constexpr const number_integer_t* get_impl_ptr(const number_integer_t* /*unused*/) const noexcept - { - return is_number_integer() ? &m_value.number_integer : nullptr; - } - - /// get a pointer to the value (unsigned number) - number_unsigned_t* get_impl_ptr(number_unsigned_t* /*unused*/) noexcept - { - return is_number_unsigned() ? &m_value.number_unsigned : nullptr; - } - - /// get a pointer to the value (unsigned number) - constexpr const number_unsigned_t* get_impl_ptr(const number_unsigned_t* /*unused*/) const noexcept - { - return is_number_unsigned() ? &m_value.number_unsigned : nullptr; - } - - /// get a pointer to the value (floating-point number) - number_float_t* get_impl_ptr(number_float_t* /*unused*/) noexcept - { - return is_number_float() ? &m_value.number_float : nullptr; - } - - /// get a pointer to the value (floating-point number) - constexpr const number_float_t* get_impl_ptr(const number_float_t* /*unused*/) const noexcept - { - return is_number_float() ? &m_value.number_float : nullptr; - } - - /// get a pointer to the value (binary) - binary_t* get_impl_ptr(binary_t* /*unused*/) noexcept - { - return is_binary() ? m_value.binary : nullptr; - } - - /// get a pointer to the value (binary) - constexpr const binary_t* get_impl_ptr(const binary_t* /*unused*/) const noexcept - { - return is_binary() ? m_value.binary : nullptr; - } - - /*! - @brief helper function to implement get_ref() - - This function helps to implement get_ref() without code duplication for - const and non-const overloads - - @tparam ThisType will be deduced as `basic_json` or `const basic_json` - - @throw type_error.303 if ReferenceType does not match underlying value - type of the current JSON - */ - template - static ReferenceType get_ref_impl(ThisType& obj) - { - // delegate the call to get_ptr<>() - auto ptr = obj.template get_ptr::type>(); - - if (JSON_HEDLEY_LIKELY(ptr != nullptr)) - { - return *ptr; - } - - JSON_THROW(type_error::create(303, "incompatible ReferenceType for get_ref, actual type is " + std::string(obj.type_name()))); - } - - public: - /// @name value access - /// Direct access to the stored value of a JSON value. - /// @{ - - /*! - @brief get special-case overload - - This overloads avoids a lot of template boilerplate, it can be seen as the - identity method - - @tparam BasicJsonType == @ref basic_json - - @return a copy of *this - - @complexity Constant. - - @since version 2.1.0 - */ - template::type, basic_json_t>::value, - int> = 0> - basic_json get() const - { - return *this; - } - - /*! - @brief get special-case overload - - This overloads converts the current @ref basic_json in a different - @ref basic_json type - - @tparam BasicJsonType == @ref basic_json - - @return a copy of *this, converted into @tparam BasicJsonType - - @complexity Depending on the implementation of the called `from_json()` - method. - - @since version 3.2.0 - */ - template < typename BasicJsonType, detail::enable_if_t < - !std::is_same::value&& - detail::is_basic_json::value, int > = 0 > - BasicJsonType get() const - { - return *this; - } - - /*! - @brief get a value (explicit) - - Explicit type conversion between the JSON value and a compatible value - which is [CopyConstructible](https://en.cppreference.com/w/cpp/named_req/CopyConstructible) - and [DefaultConstructible](https://en.cppreference.com/w/cpp/named_req/DefaultConstructible). - The value is converted by calling the @ref json_serializer - `from_json()` method. - - The function is equivalent to executing - @code {.cpp} - ValueType ret; - JSONSerializer::from_json(*this, ret); - return ret; - @endcode - - This overloads is chosen if: - - @a ValueType is not @ref basic_json, - - @ref json_serializer has a `from_json()` method of the form - `void from_json(const basic_json&, ValueType&)`, and - - @ref json_serializer does not have a `from_json()` method of - the form `ValueType from_json(const basic_json&)` - - @tparam ValueTypeCV the provided value type - @tparam ValueType the returned value type - - @return copy of the JSON value, converted to @a ValueType - - @throw what @ref json_serializer `from_json()` method throws - - @liveexample{The example below shows several conversions from JSON values - to other types. There a few things to note: (1) Floating-point numbers can - be converted to integers\, (2) A JSON array can be converted to a standard - `std::vector`\, (3) A JSON object can be converted to C++ - associative containers such as `std::unordered_map`.,get__ValueType_const} - - @since version 2.1.0 - */ - template < typename ValueTypeCV, typename ValueType = detail::uncvref_t, - detail::enable_if_t < - !detail::is_basic_json::value && - detail::has_from_json::value && - !detail::has_non_default_from_json::value, - int > = 0 > - ValueType get() const noexcept(noexcept( - JSONSerializer::from_json(std::declval(), std::declval()))) - { - // we cannot static_assert on ValueTypeCV being non-const, because - // there is support for get(), which is why we - // still need the uncvref - static_assert(!std::is_reference::value, - "get() cannot be used with reference types, you might want to use get_ref()"); - static_assert(std::is_default_constructible::value, - "types must be DefaultConstructible when used with get()"); - - ValueType ret; - JSONSerializer::from_json(*this, ret); - return ret; - } - - /*! - @brief get a value (explicit); special case - - Explicit type conversion between the JSON value and a compatible value - which is **not** [CopyConstructible](https://en.cppreference.com/w/cpp/named_req/CopyConstructible) - and **not** [DefaultConstructible](https://en.cppreference.com/w/cpp/named_req/DefaultConstructible). - The value is converted by calling the @ref json_serializer - `from_json()` method. - - The function is equivalent to executing - @code {.cpp} - return JSONSerializer::from_json(*this); - @endcode - - This overloads is chosen if: - - @a ValueType is not @ref basic_json and - - @ref json_serializer has a `from_json()` method of the form - `ValueType from_json(const basic_json&)` - - @note If @ref json_serializer has both overloads of - `from_json()`, this one is chosen. - - @tparam ValueTypeCV the provided value type - @tparam ValueType the returned value type - - @return copy of the JSON value, converted to @a ValueType - - @throw what @ref json_serializer `from_json()` method throws - - @since version 2.1.0 - */ - template < typename ValueTypeCV, typename ValueType = detail::uncvref_t, - detail::enable_if_t < !std::is_same::value && - detail::has_non_default_from_json::value, - int > = 0 > - ValueType get() const noexcept(noexcept( - JSONSerializer::from_json(std::declval()))) - { - static_assert(!std::is_reference::value, - "get() cannot be used with reference types, you might want to use get_ref()"); - return JSONSerializer::from_json(*this); - } - - /*! - @brief get a value (explicit) - - Explicit type conversion between the JSON value and a compatible value. - The value is filled into the input parameter by calling the @ref json_serializer - `from_json()` method. - - The function is equivalent to executing - @code {.cpp} - ValueType v; - JSONSerializer::from_json(*this, v); - @endcode - - This overloads is chosen if: - - @a ValueType is not @ref basic_json, - - @ref json_serializer has a `from_json()` method of the form - `void from_json(const basic_json&, ValueType&)`, and - - @tparam ValueType the input parameter type. - - @return the input parameter, allowing chaining calls. - - @throw what @ref json_serializer `from_json()` method throws - - @liveexample{The example below shows several conversions from JSON values - to other types. There a few things to note: (1) Floating-point numbers can - be converted to integers\, (2) A JSON array can be converted to a standard - `std::vector`\, (3) A JSON object can be converted to C++ - associative containers such as `std::unordered_map`.,get_to} - - @since version 3.3.0 - */ - template < typename ValueType, - detail::enable_if_t < - !detail::is_basic_json::value&& - detail::has_from_json::value, - int > = 0 > - ValueType & get_to(ValueType& v) const noexcept(noexcept( - JSONSerializer::from_json(std::declval(), v))) - { - JSONSerializer::from_json(*this, v); - return v; - } - - // specialization to allow to call get_to with a basic_json value - // see https://github.com/nlohmann/json/issues/2175 - template::value, - int> = 0> - ValueType & get_to(ValueType& v) const - { - v = *this; - return v; - } - - template < - typename T, std::size_t N, - typename Array = T (&)[N], - detail::enable_if_t < - detail::has_from_json::value, int > = 0 > - Array get_to(T (&v)[N]) const - noexcept(noexcept(JSONSerializer::from_json( - std::declval(), v))) - { - JSONSerializer::from_json(*this, v); - return v; - } - - - /*! - @brief get a pointer value (implicit) - - Implicit pointer access to the internally stored JSON value. No copies are - made. - - @warning Writing data to the pointee of the result yields an undefined - state. - - @tparam PointerType pointer type; must be a pointer to @ref array_t, @ref - object_t, @ref string_t, @ref boolean_t, @ref number_integer_t, - @ref number_unsigned_t, or @ref number_float_t. Enforced by a static - assertion. - - @return pointer to the internally stored JSON value if the requested - pointer type @a PointerType fits to the JSON value; `nullptr` otherwise - - @complexity Constant. - - @liveexample{The example below shows how pointers to internal values of a - JSON value can be requested. Note that no type conversions are made and a - `nullptr` is returned if the value and the requested pointer type does not - match.,get_ptr} - - @since version 1.0.0 - */ - template::value, int>::type = 0> - auto get_ptr() noexcept -> decltype(std::declval().get_impl_ptr(std::declval())) - { - // delegate the call to get_impl_ptr<>() - return get_impl_ptr(static_cast(nullptr)); - } - - /*! - @brief get a pointer value (implicit) - @copydoc get_ptr() - */ - template < typename PointerType, typename std::enable_if < - std::is_pointer::value&& - std::is_const::type>::value, int >::type = 0 > - constexpr auto get_ptr() const noexcept -> decltype(std::declval().get_impl_ptr(std::declval())) - { - // delegate the call to get_impl_ptr<>() const - return get_impl_ptr(static_cast(nullptr)); - } - - /*! - @brief get a pointer value (explicit) - - Explicit pointer access to the internally stored JSON value. No copies are - made. - - @warning The pointer becomes invalid if the underlying JSON object - changes. - - @tparam PointerType pointer type; must be a pointer to @ref array_t, @ref - object_t, @ref string_t, @ref boolean_t, @ref number_integer_t, - @ref number_unsigned_t, or @ref number_float_t. - - @return pointer to the internally stored JSON value if the requested - pointer type @a PointerType fits to the JSON value; `nullptr` otherwise - - @complexity Constant. - - @liveexample{The example below shows how pointers to internal values of a - JSON value can be requested. Note that no type conversions are made and a - `nullptr` is returned if the value and the requested pointer type does not - match.,get__PointerType} - - @sa @ref get_ptr() for explicit pointer-member access - - @since version 1.0.0 - */ - template::value, int>::type = 0> - auto get() noexcept -> decltype(std::declval().template get_ptr()) - { - // delegate the call to get_ptr - return get_ptr(); - } - - /*! - @brief get a pointer value (explicit) - @copydoc get() - */ - template::value, int>::type = 0> - constexpr auto get() const noexcept -> decltype(std::declval().template get_ptr()) - { - // delegate the call to get_ptr - return get_ptr(); - } - - /*! - @brief get a reference value (implicit) - - Implicit reference access to the internally stored JSON value. No copies - are made. - - @warning Writing data to the referee of the result yields an undefined - state. - - @tparam ReferenceType reference type; must be a reference to @ref array_t, - @ref object_t, @ref string_t, @ref boolean_t, @ref number_integer_t, or - @ref number_float_t. Enforced by static assertion. - - @return reference to the internally stored JSON value if the requested - reference type @a ReferenceType fits to the JSON value; throws - type_error.303 otherwise - - @throw type_error.303 in case passed type @a ReferenceType is incompatible - with the stored JSON value; see example below - - @complexity Constant. - - @liveexample{The example shows several calls to `get_ref()`.,get_ref} - - @since version 1.1.0 - */ - template::value, int>::type = 0> - ReferenceType get_ref() - { - // delegate call to get_ref_impl - return get_ref_impl(*this); - } - - /*! - @brief get a reference value (implicit) - @copydoc get_ref() - */ - template < typename ReferenceType, typename std::enable_if < - std::is_reference::value&& - std::is_const::type>::value, int >::type = 0 > - ReferenceType get_ref() const - { - // delegate call to get_ref_impl - return get_ref_impl(*this); - } - - /*! - @brief get a value (implicit) - - Implicit type conversion between the JSON value and a compatible value. - The call is realized by calling @ref get() const. - - @tparam ValueType non-pointer type compatible to the JSON value, for - instance `int` for JSON integer numbers, `bool` for JSON booleans, or - `std::vector` types for JSON arrays. The character type of @ref string_t - as well as an initializer list of this type is excluded to avoid - ambiguities as these types implicitly convert to `std::string`. - - @return copy of the JSON value, converted to type @a ValueType - - @throw type_error.302 in case passed type @a ValueType is incompatible - to the JSON value type (e.g., the JSON value is of type boolean, but a - string is requested); see example below - - @complexity Linear in the size of the JSON value. - - @liveexample{The example below shows several conversions from JSON values - to other types. There a few things to note: (1) Floating-point numbers can - be converted to integers\, (2) A JSON array can be converted to a standard - `std::vector`\, (3) A JSON object can be converted to C++ - associative containers such as `std::unordered_map`.,operator__ValueType} - - @since version 1.0.0 - */ - template < typename ValueType, typename std::enable_if < - !std::is_pointer::value&& - !std::is_same>::value&& - !std::is_same::value&& - !detail::is_basic_json::value - && !std::is_same>::value -#if defined(JSON_HAS_CPP_17) && (defined(__GNUC__) || (defined(_MSC_VER) && _MSC_VER >= 1910 && _MSC_VER <= 1914)) - && !std::is_same::value -#endif - && detail::is_detected::value - , int >::type = 0 > - JSON_EXPLICIT operator ValueType() const - { - // delegate the call to get<>() const - return get(); - } - - /*! - @return reference to the binary value - - @throw type_error.302 if the value is not binary - - @sa @ref is_binary() to check if the value is binary - - @since version 3.8.0 - */ - binary_t& get_binary() - { - if (!is_binary()) - { - JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(type_name()))); - } - - return *get_ptr(); - } - - /// @copydoc get_binary() - const binary_t& get_binary() const - { - if (!is_binary()) - { - JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(type_name()))); - } - - return *get_ptr(); - } - - /// @} - - - //////////////////// - // element access // - //////////////////// - - /// @name element access - /// Access to the JSON value. - /// @{ - - /*! - @brief access specified array element with bounds checking - - Returns a reference to the element at specified location @a idx, with - bounds checking. - - @param[in] idx index of the element to access - - @return reference to the element at index @a idx - - @throw type_error.304 if the JSON value is not an array; in this case, - calling `at` with an index makes no sense. See example below. - @throw out_of_range.401 if the index @a idx is out of range of the array; - that is, `idx >= size()`. See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @since version 1.0.0 - - @liveexample{The example below shows how array elements can be read and - written using `at()`. It also demonstrates the different exceptions that - can be thrown.,at__size_type} - */ - reference at(size_type idx) - { - // at only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - JSON_TRY - { - return m_value.array->at(idx); - } - JSON_CATCH (std::out_of_range&) - { - // create better exception explanation - JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range")); - } - } - else - { - JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()))); - } - } - - /*! - @brief access specified array element with bounds checking - - Returns a const reference to the element at specified location @a idx, - with bounds checking. - - @param[in] idx index of the element to access - - @return const reference to the element at index @a idx - - @throw type_error.304 if the JSON value is not an array; in this case, - calling `at` with an index makes no sense. See example below. - @throw out_of_range.401 if the index @a idx is out of range of the array; - that is, `idx >= size()`. See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @since version 1.0.0 - - @liveexample{The example below shows how array elements can be read using - `at()`. It also demonstrates the different exceptions that can be thrown., - at__size_type_const} - */ - const_reference at(size_type idx) const - { - // at only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - JSON_TRY - { - return m_value.array->at(idx); - } - JSON_CATCH (std::out_of_range&) - { - // create better exception explanation - JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range")); - } - } - else - { - JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()))); - } - } - - /*! - @brief access specified object element with bounds checking - - Returns a reference to the element at with specified key @a key, with - bounds checking. - - @param[in] key key of the element to access - - @return reference to the element at key @a key - - @throw type_error.304 if the JSON value is not an object; in this case, - calling `at` with a key makes no sense. See example below. - @throw out_of_range.403 if the key @a key is is not stored in the object; - that is, `find(key) == end()`. See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Logarithmic in the size of the container. - - @sa @ref operator[](const typename object_t::key_type&) for unchecked - access by reference - @sa @ref value() for access by value with a default value - - @since version 1.0.0 - - @liveexample{The example below shows how object elements can be read and - written using `at()`. It also demonstrates the different exceptions that - can be thrown.,at__object_t_key_type} - */ - reference at(const typename object_t::key_type& key) - { - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - JSON_TRY - { - return m_value.object->at(key); - } - JSON_CATCH (std::out_of_range&) - { - // create better exception explanation - JSON_THROW(out_of_range::create(403, "key '" + key + "' not found")); - } - } - else - { - JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()))); - } - } - - /*! - @brief access specified object element with bounds checking - - Returns a const reference to the element at with specified key @a key, - with bounds checking. - - @param[in] key key of the element to access - - @return const reference to the element at key @a key - - @throw type_error.304 if the JSON value is not an object; in this case, - calling `at` with a key makes no sense. See example below. - @throw out_of_range.403 if the key @a key is is not stored in the object; - that is, `find(key) == end()`. See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Logarithmic in the size of the container. - - @sa @ref operator[](const typename object_t::key_type&) for unchecked - access by reference - @sa @ref value() for access by value with a default value - - @since version 1.0.0 - - @liveexample{The example below shows how object elements can be read using - `at()`. It also demonstrates the different exceptions that can be thrown., - at__object_t_key_type_const} - */ - const_reference at(const typename object_t::key_type& key) const - { - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - JSON_TRY - { - return m_value.object->at(key); - } - JSON_CATCH (std::out_of_range&) - { - // create better exception explanation - JSON_THROW(out_of_range::create(403, "key '" + key + "' not found")); - } - } - else - { - JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()))); - } - } - - /*! - @brief access specified array element - - Returns a reference to the element at specified location @a idx. - - @note If @a idx is beyond the range of the array (i.e., `idx >= size()`), - then the array is silently filled up with `null` values to make `idx` a - valid reference to the last stored element. - - @param[in] idx index of the element to access - - @return reference to the element at index @a idx - - @throw type_error.305 if the JSON value is not an array or null; in that - cases, using the [] operator with an index makes no sense. - - @complexity Constant if @a idx is in the range of the array. Otherwise - linear in `idx - size()`. - - @liveexample{The example below shows how array elements can be read and - written using `[]` operator. Note the addition of `null` - values.,operatorarray__size_type} - - @since version 1.0.0 - */ - reference operator[](size_type idx) - { - // implicitly convert null value to an empty array - if (is_null()) - { - m_type = value_t::array; - m_value.array = create(); - assert_invariant(); - } - - // operator[] only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - // fill up array with null values if given idx is outside range - if (idx >= m_value.array->size()) - { - m_value.array->insert(m_value.array->end(), - idx - m_value.array->size() + 1, - basic_json()); - } - - return m_value.array->operator[](idx); - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name()))); - } - - /*! - @brief access specified array element - - Returns a const reference to the element at specified location @a idx. - - @param[in] idx index of the element to access - - @return const reference to the element at index @a idx - - @throw type_error.305 if the JSON value is not an array; in that case, - using the [] operator with an index makes no sense. - - @complexity Constant. - - @liveexample{The example below shows how array elements can be read using - the `[]` operator.,operatorarray__size_type_const} - - @since version 1.0.0 - */ - const_reference operator[](size_type idx) const - { - // const operator[] only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - return m_value.array->operator[](idx); - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name()))); - } - - /*! - @brief access specified object element - - Returns a reference to the element at with specified key @a key. - - @note If @a key is not found in the object, then it is silently added to - the object and filled with a `null` value to make `key` a valid reference. - In case the value was `null` before, it is converted to an object. - - @param[in] key key of the element to access - - @return reference to the element at key @a key - - @throw type_error.305 if the JSON value is not an object or null; in that - cases, using the [] operator with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be read and - written using the `[]` operator.,operatorarray__key_type} - - @sa @ref at(const typename object_t::key_type&) for access by reference - with range checking - @sa @ref value() for access by value with a default value - - @since version 1.0.0 - */ - reference operator[](const typename object_t::key_type& key) - { - // implicitly convert null value to an empty object - if (is_null()) - { - m_type = value_t::object; - m_value.object = create(); - assert_invariant(); - } - - // operator[] only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - return m_value.object->operator[](key); - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()))); - } - - /*! - @brief read-only access specified object element - - Returns a const reference to the element at with specified key @a key. No - bounds checking is performed. - - @warning If the element with key @a key does not exist, the behavior is - undefined. - - @param[in] key key of the element to access - - @return const reference to the element at key @a key - - @pre The element with key @a key must exist. **This precondition is - enforced with an assertion.** - - @throw type_error.305 if the JSON value is not an object; in that case, - using the [] operator with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be read using - the `[]` operator.,operatorarray__key_type_const} - - @sa @ref at(const typename object_t::key_type&) for access by reference - with range checking - @sa @ref value() for access by value with a default value - - @since version 1.0.0 - */ - const_reference operator[](const typename object_t::key_type& key) const - { - // const operator[] only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - JSON_ASSERT(m_value.object->find(key) != m_value.object->end()); - return m_value.object->find(key)->second; - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()))); - } - - /*! - @brief access specified object element - - Returns a reference to the element at with specified key @a key. - - @note If @a key is not found in the object, then it is silently added to - the object and filled with a `null` value to make `key` a valid reference. - In case the value was `null` before, it is converted to an object. - - @param[in] key key of the element to access - - @return reference to the element at key @a key - - @throw type_error.305 if the JSON value is not an object or null; in that - cases, using the [] operator with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be read and - written using the `[]` operator.,operatorarray__key_type} - - @sa @ref at(const typename object_t::key_type&) for access by reference - with range checking - @sa @ref value() for access by value with a default value - - @since version 1.1.0 - */ - template - JSON_HEDLEY_NON_NULL(2) - reference operator[](T* key) - { - // implicitly convert null to object - if (is_null()) - { - m_type = value_t::object; - m_value = value_t::object; - assert_invariant(); - } - - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - return m_value.object->operator[](key); - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()))); - } - - /*! - @brief read-only access specified object element - - Returns a const reference to the element at with specified key @a key. No - bounds checking is performed. - - @warning If the element with key @a key does not exist, the behavior is - undefined. - - @param[in] key key of the element to access - - @return const reference to the element at key @a key - - @pre The element with key @a key must exist. **This precondition is - enforced with an assertion.** - - @throw type_error.305 if the JSON value is not an object; in that case, - using the [] operator with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be read using - the `[]` operator.,operatorarray__key_type_const} - - @sa @ref at(const typename object_t::key_type&) for access by reference - with range checking - @sa @ref value() for access by value with a default value - - @since version 1.1.0 - */ - template - JSON_HEDLEY_NON_NULL(2) - const_reference operator[](T* key) const - { - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - JSON_ASSERT(m_value.object->find(key) != m_value.object->end()); - return m_value.object->find(key)->second; - } - - JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()))); - } - - /*! - @brief access specified object element with default value - - Returns either a copy of an object's element at the specified key @a key - or a given default value if no element with key @a key exists. - - The function is basically equivalent to executing - @code {.cpp} - try { - return at(key); - } catch(out_of_range) { - return default_value; - } - @endcode - - @note Unlike @ref at(const typename object_t::key_type&), this function - does not throw if the given key @a key was not found. - - @note Unlike @ref operator[](const typename object_t::key_type& key), this - function does not implicitly add an element to the position defined by @a - key. This function is furthermore also applicable to const objects. - - @param[in] key key of the element to access - @param[in] default_value the value to return if @a key is not found - - @tparam ValueType type compatible to JSON values, for instance `int` for - JSON integer numbers, `bool` for JSON booleans, or `std::vector` types for - JSON arrays. Note the type of the expected value at @a key and the default - value @a default_value must be compatible. - - @return copy of the element at key @a key or @a default_value if @a key - is not found - - @throw type_error.302 if @a default_value does not match the type of the - value at @a key - @throw type_error.306 if the JSON value is not an object; in that case, - using `value()` with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be queried - with a default value.,basic_json__value} - - @sa @ref at(const typename object_t::key_type&) for access by reference - with range checking - @sa @ref operator[](const typename object_t::key_type&) for unchecked - access by reference - - @since version 1.0.0 - */ - // using std::is_convertible in a std::enable_if will fail when using explicit conversions - template < class ValueType, typename std::enable_if < - detail::is_getable::value - && !std::is_same::value, int >::type = 0 > - ValueType value(const typename object_t::key_type& key, const ValueType& default_value) const - { - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - // if key is found, return value and given default value otherwise - const auto it = find(key); - if (it != end()) - { - return it->template get(); - } - - return default_value; - } - - JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name()))); - } - - /*! - @brief overload for a default value of type const char* - @copydoc basic_json::value(const typename object_t::key_type&, const ValueType&) const - */ - string_t value(const typename object_t::key_type& key, const char* default_value) const - { - return value(key, string_t(default_value)); - } - - /*! - @brief access specified object element via JSON Pointer with default value - - Returns either a copy of an object's element at the specified key @a key - or a given default value if no element with key @a key exists. - - The function is basically equivalent to executing - @code {.cpp} - try { - return at(ptr); - } catch(out_of_range) { - return default_value; - } - @endcode - - @note Unlike @ref at(const json_pointer&), this function does not throw - if the given key @a key was not found. - - @param[in] ptr a JSON pointer to the element to access - @param[in] default_value the value to return if @a ptr found no value - - @tparam ValueType type compatible to JSON values, for instance `int` for - JSON integer numbers, `bool` for JSON booleans, or `std::vector` types for - JSON arrays. Note the type of the expected value at @a key and the default - value @a default_value must be compatible. - - @return copy of the element at key @a key or @a default_value if @a key - is not found - - @throw type_error.302 if @a default_value does not match the type of the - value at @a ptr - @throw type_error.306 if the JSON value is not an object; in that case, - using `value()` with a key makes no sense. - - @complexity Logarithmic in the size of the container. - - @liveexample{The example below shows how object elements can be queried - with a default value.,basic_json__value_ptr} - - @sa @ref operator[](const json_pointer&) for unchecked access by reference - - @since version 2.0.2 - */ - template::value, int>::type = 0> - ValueType value(const json_pointer& ptr, const ValueType& default_value) const - { - // at only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - // if pointer resolves a value, return it or use default value - JSON_TRY - { - return ptr.get_checked(this).template get(); - } - JSON_INTERNAL_CATCH (out_of_range&) - { - return default_value; - } - } - - JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name()))); - } - - /*! - @brief overload for a default value of type const char* - @copydoc basic_json::value(const json_pointer&, ValueType) const - */ - JSON_HEDLEY_NON_NULL(3) - string_t value(const json_pointer& ptr, const char* default_value) const - { - return value(ptr, string_t(default_value)); - } - - /*! - @brief access the first element - - Returns a reference to the first element in the container. For a JSON - container `c`, the expression `c.front()` is equivalent to `*c.begin()`. - - @return In case of a structured type (array or object), a reference to the - first element is returned. In case of number, string, boolean, or binary - values, a reference to the value is returned. - - @complexity Constant. - - @pre The JSON value must not be `null` (would throw `std::out_of_range`) - or an empty array or object (undefined behavior, **guarded by - assertions**). - @post The JSON value remains unchanged. - - @throw invalid_iterator.214 when called on `null` value - - @liveexample{The following code shows an example for `front()`.,front} - - @sa @ref back() -- access the last element - - @since version 1.0.0 - */ - reference front() - { - return *begin(); - } - - /*! - @copydoc basic_json::front() - */ - const_reference front() const - { - return *cbegin(); - } - - /*! - @brief access the last element - - Returns a reference to the last element in the container. For a JSON - container `c`, the expression `c.back()` is equivalent to - @code {.cpp} - auto tmp = c.end(); - --tmp; - return *tmp; - @endcode - - @return In case of a structured type (array or object), a reference to the - last element is returned. In case of number, string, boolean, or binary - values, a reference to the value is returned. - - @complexity Constant. - - @pre The JSON value must not be `null` (would throw `std::out_of_range`) - or an empty array or object (undefined behavior, **guarded by - assertions**). - @post The JSON value remains unchanged. - - @throw invalid_iterator.214 when called on a `null` value. See example - below. - - @liveexample{The following code shows an example for `back()`.,back} - - @sa @ref front() -- access the first element - - @since version 1.0.0 - */ - reference back() - { - auto tmp = end(); - --tmp; - return *tmp; - } - - /*! - @copydoc basic_json::back() - */ - const_reference back() const - { - auto tmp = cend(); - --tmp; - return *tmp; - } - - /*! - @brief remove element given an iterator - - Removes the element specified by iterator @a pos. The iterator @a pos must - be valid and dereferenceable. Thus the `end()` iterator (which is valid, - but is not dereferenceable) cannot be used as a value for @a pos. - - If called on a primitive type other than `null`, the resulting JSON value - will be `null`. - - @param[in] pos iterator to the element to remove - @return Iterator following the last removed element. If the iterator @a - pos refers to the last element, the `end()` iterator is returned. - - @tparam IteratorType an @ref iterator or @ref const_iterator - - @post Invalidates iterators and references at or after the point of the - erase, including the `end()` iterator. - - @throw type_error.307 if called on a `null` value; example: `"cannot use - erase() with null"` - @throw invalid_iterator.202 if called on an iterator which does not belong - to the current JSON value; example: `"iterator does not fit current - value"` - @throw invalid_iterator.205 if called on a primitive type with invalid - iterator (i.e., any iterator which is not `begin()`); example: `"iterator - out of range"` - - @complexity The complexity depends on the type: - - objects: amortized constant - - arrays: linear in distance between @a pos and the end of the container - - strings and binary: linear in the length of the member - - other types: constant - - @liveexample{The example shows the result of `erase()` for different JSON - types.,erase__IteratorType} - - @sa @ref erase(IteratorType, IteratorType) -- removes the elements in - the given range - @sa @ref erase(const typename object_t::key_type&) -- removes the element - from an object at the given key - @sa @ref erase(const size_type) -- removes the element from an array at - the given index - - @since version 1.0.0 - */ - template < class IteratorType, typename std::enable_if < - std::is_same::value || - std::is_same::value, int >::type - = 0 > - IteratorType erase(IteratorType pos) - { - // make sure iterator fits the current value - if (JSON_HEDLEY_UNLIKELY(this != pos.m_object)) - { - JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value")); - } - - IteratorType result = end(); - - switch (m_type) - { - case value_t::boolean: - case value_t::number_float: - case value_t::number_integer: - case value_t::number_unsigned: - case value_t::string: - case value_t::binary: - { - if (JSON_HEDLEY_UNLIKELY(!pos.m_it.primitive_iterator.is_begin())) - { - JSON_THROW(invalid_iterator::create(205, "iterator out of range")); - } - - if (is_string()) - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, m_value.string); - std::allocator_traits::deallocate(alloc, m_value.string, 1); - m_value.string = nullptr; - } - else if (is_binary()) - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, m_value.binary); - std::allocator_traits::deallocate(alloc, m_value.binary, 1); - m_value.binary = nullptr; - } - - m_type = value_t::null; - assert_invariant(); - break; - } - - case value_t::object: - { - result.m_it.object_iterator = m_value.object->erase(pos.m_it.object_iterator); - break; - } - - case value_t::array: - { - result.m_it.array_iterator = m_value.array->erase(pos.m_it.array_iterator); - break; - } - - default: - JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()))); - } - - return result; - } - - /*! - @brief remove elements given an iterator range - - Removes the element specified by the range `[first; last)`. The iterator - @a first does not need to be dereferenceable if `first == last`: erasing - an empty range is a no-op. - - If called on a primitive type other than `null`, the resulting JSON value - will be `null`. - - @param[in] first iterator to the beginning of the range to remove - @param[in] last iterator past the end of the range to remove - @return Iterator following the last removed element. If the iterator @a - second refers to the last element, the `end()` iterator is returned. - - @tparam IteratorType an @ref iterator or @ref const_iterator - - @post Invalidates iterators and references at or after the point of the - erase, including the `end()` iterator. - - @throw type_error.307 if called on a `null` value; example: `"cannot use - erase() with null"` - @throw invalid_iterator.203 if called on iterators which does not belong - to the current JSON value; example: `"iterators do not fit current value"` - @throw invalid_iterator.204 if called on a primitive type with invalid - iterators (i.e., if `first != begin()` and `last != end()`); example: - `"iterators out of range"` - - @complexity The complexity depends on the type: - - objects: `log(size()) + std::distance(first, last)` - - arrays: linear in the distance between @a first and @a last, plus linear - in the distance between @a last and end of the container - - strings and binary: linear in the length of the member - - other types: constant - - @liveexample{The example shows the result of `erase()` for different JSON - types.,erase__IteratorType_IteratorType} - - @sa @ref erase(IteratorType) -- removes the element at a given position - @sa @ref erase(const typename object_t::key_type&) -- removes the element - from an object at the given key - @sa @ref erase(const size_type) -- removes the element from an array at - the given index - - @since version 1.0.0 - */ - template < class IteratorType, typename std::enable_if < - std::is_same::value || - std::is_same::value, int >::type - = 0 > - IteratorType erase(IteratorType first, IteratorType last) - { - // make sure iterator fits the current value - if (JSON_HEDLEY_UNLIKELY(this != first.m_object || this != last.m_object)) - { - JSON_THROW(invalid_iterator::create(203, "iterators do not fit current value")); - } - - IteratorType result = end(); - - switch (m_type) - { - case value_t::boolean: - case value_t::number_float: - case value_t::number_integer: - case value_t::number_unsigned: - case value_t::string: - case value_t::binary: - { - if (JSON_HEDLEY_LIKELY(!first.m_it.primitive_iterator.is_begin() - || !last.m_it.primitive_iterator.is_end())) - { - JSON_THROW(invalid_iterator::create(204, "iterators out of range")); - } - - if (is_string()) - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, m_value.string); - std::allocator_traits::deallocate(alloc, m_value.string, 1); - m_value.string = nullptr; - } - else if (is_binary()) - { - AllocatorType alloc; - std::allocator_traits::destroy(alloc, m_value.binary); - std::allocator_traits::deallocate(alloc, m_value.binary, 1); - m_value.binary = nullptr; - } - - m_type = value_t::null; - assert_invariant(); - break; - } - - case value_t::object: - { - result.m_it.object_iterator = m_value.object->erase(first.m_it.object_iterator, - last.m_it.object_iterator); - break; - } - - case value_t::array: - { - result.m_it.array_iterator = m_value.array->erase(first.m_it.array_iterator, - last.m_it.array_iterator); - break; - } - - default: - JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()))); - } - - return result; - } - - /*! - @brief remove element from a JSON object given a key - - Removes elements from a JSON object with the key value @a key. - - @param[in] key value of the elements to remove - - @return Number of elements removed. If @a ObjectType is the default - `std::map` type, the return value will always be `0` (@a key was not - found) or `1` (@a key was found). - - @post References and iterators to the erased elements are invalidated. - Other references and iterators are not affected. - - @throw type_error.307 when called on a type other than JSON object; - example: `"cannot use erase() with null"` - - @complexity `log(size()) + count(key)` - - @liveexample{The example shows the effect of `erase()`.,erase__key_type} - - @sa @ref erase(IteratorType) -- removes the element at a given position - @sa @ref erase(IteratorType, IteratorType) -- removes the elements in - the given range - @sa @ref erase(const size_type) -- removes the element from an array at - the given index - - @since version 1.0.0 - */ - size_type erase(const typename object_t::key_type& key) - { - // this erase only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - return m_value.object->erase(key); - } - - JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()))); - } - - /*! - @brief remove element from a JSON array given an index - - Removes element from a JSON array at the index @a idx. - - @param[in] idx index of the element to remove - - @throw type_error.307 when called on a type other than JSON object; - example: `"cannot use erase() with null"` - @throw out_of_range.401 when `idx >= size()`; example: `"array index 17 - is out of range"` - - @complexity Linear in distance between @a idx and the end of the container. - - @liveexample{The example shows the effect of `erase()`.,erase__size_type} - - @sa @ref erase(IteratorType) -- removes the element at a given position - @sa @ref erase(IteratorType, IteratorType) -- removes the elements in - the given range - @sa @ref erase(const typename object_t::key_type&) -- removes the element - from an object at the given key - - @since version 1.0.0 - */ - void erase(const size_type idx) - { - // this erase only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - if (JSON_HEDLEY_UNLIKELY(idx >= size())) - { - JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range")); - } - - m_value.array->erase(m_value.array->begin() + static_cast(idx)); - } - else - { - JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()))); - } - } - - /// @} - - - //////////// - // lookup // - //////////// - - /// @name lookup - /// @{ - - /*! - @brief find an element in a JSON object - - Finds an element in a JSON object with key equivalent to @a key. If the - element is not found or the JSON value is not an object, end() is - returned. - - @note This method always returns @ref end() when executed on a JSON type - that is not an object. - - @param[in] key key value of the element to search for. - - @return Iterator to an element with key equivalent to @a key. If no such - element is found or the JSON value is not an object, past-the-end (see - @ref end()) iterator is returned. - - @complexity Logarithmic in the size of the JSON object. - - @liveexample{The example shows how `find()` is used.,find__key_type} - - @sa @ref contains(KeyT&&) const -- checks whether a key exists - - @since version 1.0.0 - */ - template - iterator find(KeyT&& key) - { - auto result = end(); - - if (is_object()) - { - result.m_it.object_iterator = m_value.object->find(std::forward(key)); - } - - return result; - } - - /*! - @brief find an element in a JSON object - @copydoc find(KeyT&&) - */ - template - const_iterator find(KeyT&& key) const - { - auto result = cend(); - - if (is_object()) - { - result.m_it.object_iterator = m_value.object->find(std::forward(key)); - } - - return result; - } - - /*! - @brief returns the number of occurrences of a key in a JSON object - - Returns the number of elements with key @a key. If ObjectType is the - default `std::map` type, the return value will always be `0` (@a key was - not found) or `1` (@a key was found). - - @note This method always returns `0` when executed on a JSON type that is - not an object. - - @param[in] key key value of the element to count - - @return Number of elements with key @a key. If the JSON value is not an - object, the return value will be `0`. - - @complexity Logarithmic in the size of the JSON object. - - @liveexample{The example shows how `count()` is used.,count} - - @since version 1.0.0 - */ - template - size_type count(KeyT&& key) const - { - // return 0 for all nonobject types - return is_object() ? m_value.object->count(std::forward(key)) : 0; - } - - /*! - @brief check the existence of an element in a JSON object - - Check whether an element exists in a JSON object with key equivalent to - @a key. If the element is not found or the JSON value is not an object, - false is returned. - - @note This method always returns false when executed on a JSON type - that is not an object. - - @param[in] key key value to check its existence. - - @return true if an element with specified @a key exists. If no such - element with such key is found or the JSON value is not an object, - false is returned. - - @complexity Logarithmic in the size of the JSON object. - - @liveexample{The following code shows an example for `contains()`.,contains} - - @sa @ref find(KeyT&&) -- returns an iterator to an object element - @sa @ref contains(const json_pointer&) const -- checks the existence for a JSON pointer - - @since version 3.6.0 - */ - template < typename KeyT, typename std::enable_if < - !std::is_same::type, json_pointer>::value, int >::type = 0 > - bool contains(KeyT && key) const - { - return is_object() && m_value.object->find(std::forward(key)) != m_value.object->end(); - } - - /*! - @brief check the existence of an element in a JSON object given a JSON pointer - - Check whether the given JSON pointer @a ptr can be resolved in the current - JSON value. - - @note This method can be executed on any JSON value type. - - @param[in] ptr JSON pointer to check its existence. - - @return true if the JSON pointer can be resolved to a stored value, false - otherwise. - - @post If `j.contains(ptr)` returns true, it is safe to call `j[ptr]`. - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - - @complexity Logarithmic in the size of the JSON object. - - @liveexample{The following code shows an example for `contains()`.,contains_json_pointer} - - @sa @ref contains(KeyT &&) const -- checks the existence of a key - - @since version 3.7.0 - */ - bool contains(const json_pointer& ptr) const - { - return ptr.contains(this); - } - - /// @} - - - /////////////// - // iterators // - /////////////// - - /// @name iterators - /// @{ - - /*! - @brief returns an iterator to the first element - - Returns an iterator to the first element. - - @image html range-begin-end.svg "Illustration from cppreference.com" - - @return iterator to the first element - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - @liveexample{The following code shows an example for `begin()`.,begin} - - @sa @ref cbegin() -- returns a const iterator to the beginning - @sa @ref end() -- returns an iterator to the end - @sa @ref cend() -- returns a const iterator to the end - - @since version 1.0.0 - */ - iterator begin() noexcept - { - iterator result(this); - result.set_begin(); - return result; - } - - /*! - @copydoc basic_json::cbegin() - */ - const_iterator begin() const noexcept - { - return cbegin(); - } - - /*! - @brief returns a const iterator to the first element - - Returns a const iterator to the first element. - - @image html range-begin-end.svg "Illustration from cppreference.com" - - @return const iterator to the first element - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - Has the semantics of `const_cast(*this).begin()`. - - @liveexample{The following code shows an example for `cbegin()`.,cbegin} - - @sa @ref begin() -- returns an iterator to the beginning - @sa @ref end() -- returns an iterator to the end - @sa @ref cend() -- returns a const iterator to the end - - @since version 1.0.0 - */ - const_iterator cbegin() const noexcept - { - const_iterator result(this); - result.set_begin(); - return result; - } - - /*! - @brief returns an iterator to one past the last element - - Returns an iterator to one past the last element. - - @image html range-begin-end.svg "Illustration from cppreference.com" - - @return iterator one past the last element - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - @liveexample{The following code shows an example for `end()`.,end} - - @sa @ref cend() -- returns a const iterator to the end - @sa @ref begin() -- returns an iterator to the beginning - @sa @ref cbegin() -- returns a const iterator to the beginning - - @since version 1.0.0 - */ - iterator end() noexcept - { - iterator result(this); - result.set_end(); - return result; - } - - /*! - @copydoc basic_json::cend() - */ - const_iterator end() const noexcept - { - return cend(); - } - - /*! - @brief returns a const iterator to one past the last element - - Returns a const iterator to one past the last element. - - @image html range-begin-end.svg "Illustration from cppreference.com" - - @return const iterator one past the last element - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - Has the semantics of `const_cast(*this).end()`. - - @liveexample{The following code shows an example for `cend()`.,cend} - - @sa @ref end() -- returns an iterator to the end - @sa @ref begin() -- returns an iterator to the beginning - @sa @ref cbegin() -- returns a const iterator to the beginning - - @since version 1.0.0 - */ - const_iterator cend() const noexcept - { - const_iterator result(this); - result.set_end(); - return result; - } - - /*! - @brief returns an iterator to the reverse-beginning - - Returns an iterator to the reverse-beginning; that is, the last element. - - @image html range-rbegin-rend.svg "Illustration from cppreference.com" - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer) - requirements: - - The complexity is constant. - - Has the semantics of `reverse_iterator(end())`. - - @liveexample{The following code shows an example for `rbegin()`.,rbegin} - - @sa @ref crbegin() -- returns a const reverse iterator to the beginning - @sa @ref rend() -- returns a reverse iterator to the end - @sa @ref crend() -- returns a const reverse iterator to the end - - @since version 1.0.0 - */ - reverse_iterator rbegin() noexcept - { - return reverse_iterator(end()); - } - - /*! - @copydoc basic_json::crbegin() - */ - const_reverse_iterator rbegin() const noexcept - { - return crbegin(); - } - - /*! - @brief returns an iterator to the reverse-end - - Returns an iterator to the reverse-end; that is, one before the first - element. - - @image html range-rbegin-rend.svg "Illustration from cppreference.com" - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer) - requirements: - - The complexity is constant. - - Has the semantics of `reverse_iterator(begin())`. - - @liveexample{The following code shows an example for `rend()`.,rend} - - @sa @ref crend() -- returns a const reverse iterator to the end - @sa @ref rbegin() -- returns a reverse iterator to the beginning - @sa @ref crbegin() -- returns a const reverse iterator to the beginning - - @since version 1.0.0 - */ - reverse_iterator rend() noexcept - { - return reverse_iterator(begin()); - } - - /*! - @copydoc basic_json::crend() - */ - const_reverse_iterator rend() const noexcept - { - return crend(); - } - - /*! - @brief returns a const reverse iterator to the last element - - Returns a const iterator to the reverse-beginning; that is, the last - element. - - @image html range-rbegin-rend.svg "Illustration from cppreference.com" - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer) - requirements: - - The complexity is constant. - - Has the semantics of `const_cast(*this).rbegin()`. - - @liveexample{The following code shows an example for `crbegin()`.,crbegin} - - @sa @ref rbegin() -- returns a reverse iterator to the beginning - @sa @ref rend() -- returns a reverse iterator to the end - @sa @ref crend() -- returns a const reverse iterator to the end - - @since version 1.0.0 - */ - const_reverse_iterator crbegin() const noexcept - { - return const_reverse_iterator(cend()); - } - - /*! - @brief returns a const reverse iterator to one before the first - - Returns a const reverse iterator to the reverse-end; that is, one before - the first element. - - @image html range-rbegin-rend.svg "Illustration from cppreference.com" - - @complexity Constant. - - @requirement This function helps `basic_json` satisfying the - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer) - requirements: - - The complexity is constant. - - Has the semantics of `const_cast(*this).rend()`. - - @liveexample{The following code shows an example for `crend()`.,crend} - - @sa @ref rend() -- returns a reverse iterator to the end - @sa @ref rbegin() -- returns a reverse iterator to the beginning - @sa @ref crbegin() -- returns a const reverse iterator to the beginning - - @since version 1.0.0 - */ - const_reverse_iterator crend() const noexcept - { - return const_reverse_iterator(cbegin()); - } - - public: - /*! - @brief wrapper to access iterator member functions in range-based for - - This function allows to access @ref iterator::key() and @ref - iterator::value() during range-based for loops. In these loops, a - reference to the JSON values is returned, so there is no access to the - underlying iterator. - - For loop without iterator_wrapper: - - @code{cpp} - for (auto it = j_object.begin(); it != j_object.end(); ++it) - { - std::cout << "key: " << it.key() << ", value:" << it.value() << '\n'; - } - @endcode - - Range-based for loop without iterator proxy: - - @code{cpp} - for (auto it : j_object) - { - // "it" is of type json::reference and has no key() member - std::cout << "value: " << it << '\n'; - } - @endcode - - Range-based for loop with iterator proxy: - - @code{cpp} - for (auto it : json::iterator_wrapper(j_object)) - { - std::cout << "key: " << it.key() << ", value:" << it.value() << '\n'; - } - @endcode - - @note When iterating over an array, `key()` will return the index of the - element as string (see example). - - @param[in] ref reference to a JSON value - @return iteration proxy object wrapping @a ref with an interface to use in - range-based for loops - - @liveexample{The following code shows how the wrapper is used,iterator_wrapper} - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @note The name of this function is not yet final and may change in the - future. - - @deprecated This stream operator is deprecated and will be removed in - future 4.0.0 of the library. Please use @ref items() instead; - that is, replace `json::iterator_wrapper(j)` with `j.items()`. - */ - JSON_HEDLEY_DEPRECATED_FOR(3.1.0, items()) - static iteration_proxy iterator_wrapper(reference ref) noexcept - { - return ref.items(); - } - - /*! - @copydoc iterator_wrapper(reference) - */ - JSON_HEDLEY_DEPRECATED_FOR(3.1.0, items()) - static iteration_proxy iterator_wrapper(const_reference ref) noexcept - { - return ref.items(); - } - - /*! - @brief helper to access iterator member functions in range-based for - - This function allows to access @ref iterator::key() and @ref - iterator::value() during range-based for loops. In these loops, a - reference to the JSON values is returned, so there is no access to the - underlying iterator. - - For loop without `items()` function: - - @code{cpp} - for (auto it = j_object.begin(); it != j_object.end(); ++it) - { - std::cout << "key: " << it.key() << ", value:" << it.value() << '\n'; - } - @endcode - - Range-based for loop without `items()` function: - - @code{cpp} - for (auto it : j_object) - { - // "it" is of type json::reference and has no key() member - std::cout << "value: " << it << '\n'; - } - @endcode - - Range-based for loop with `items()` function: - - @code{cpp} - for (auto& el : j_object.items()) - { - std::cout << "key: " << el.key() << ", value:" << el.value() << '\n'; - } - @endcode - - The `items()` function also allows to use - [structured bindings](https://en.cppreference.com/w/cpp/language/structured_binding) - (C++17): - - @code{cpp} - for (auto& [key, val] : j_object.items()) - { - std::cout << "key: " << key << ", value:" << val << '\n'; - } - @endcode - - @note When iterating over an array, `key()` will return the index of the - element as string (see example). For primitive types (e.g., numbers), - `key()` returns an empty string. - - @warning Using `items()` on temporary objects is dangerous. Make sure the - object's lifetime exeeds the iteration. See - for more - information. - - @return iteration proxy object wrapping @a ref with an interface to use in - range-based for loops - - @liveexample{The following code shows how the function is used.,items} - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @since version 3.1.0, structured bindings support since 3.5.0. - */ - iteration_proxy items() noexcept - { - return iteration_proxy(*this); - } - - /*! - @copydoc items() - */ - iteration_proxy items() const noexcept - { - return iteration_proxy(*this); - } - - /// @} - - - ////////////// - // capacity // - ////////////// - - /// @name capacity - /// @{ - - /*! - @brief checks whether the container is empty. - - Checks if a JSON value has no elements (i.e. whether its @ref size is `0`). - - @return The return value depends on the different types and is - defined as follows: - Value type | return value - ----------- | ------------- - null | `true` - boolean | `false` - string | `false` - number | `false` - binary | `false` - object | result of function `object_t::empty()` - array | result of function `array_t::empty()` - - @liveexample{The following code uses `empty()` to check if a JSON - object contains any elements.,empty} - - @complexity Constant, as long as @ref array_t and @ref object_t satisfy - the Container concept; that is, their `empty()` functions have constant - complexity. - - @iterators No changes. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @note This function does not return whether a string stored as JSON value - is empty - it returns whether the JSON container itself is empty which is - false in the case of a string. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - Has the semantics of `begin() == end()`. - - @sa @ref size() -- returns the number of elements - - @since version 1.0.0 - */ - bool empty() const noexcept - { - switch (m_type) - { - case value_t::null: - { - // null values are empty - return true; - } - - case value_t::array: - { - // delegate call to array_t::empty() - return m_value.array->empty(); - } - - case value_t::object: - { - // delegate call to object_t::empty() - return m_value.object->empty(); - } - - default: - { - // all other types are nonempty - return false; - } - } - } - - /*! - @brief returns the number of elements - - Returns the number of elements in a JSON value. - - @return The return value depends on the different types and is - defined as follows: - Value type | return value - ----------- | ------------- - null | `0` - boolean | `1` - string | `1` - number | `1` - binary | `1` - object | result of function object_t::size() - array | result of function array_t::size() - - @liveexample{The following code calls `size()` on the different value - types.,size} - - @complexity Constant, as long as @ref array_t and @ref object_t satisfy - the Container concept; that is, their size() functions have constant - complexity. - - @iterators No changes. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @note This function does not return the length of a string stored as JSON - value - it returns the number of elements in the JSON value which is 1 in - the case of a string. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - Has the semantics of `std::distance(begin(), end())`. - - @sa @ref empty() -- checks whether the container is empty - @sa @ref max_size() -- returns the maximal number of elements - - @since version 1.0.0 - */ - size_type size() const noexcept - { - switch (m_type) - { - case value_t::null: - { - // null values are empty - return 0; - } - - case value_t::array: - { - // delegate call to array_t::size() - return m_value.array->size(); - } - - case value_t::object: - { - // delegate call to object_t::size() - return m_value.object->size(); - } - - default: - { - // all other types have size 1 - return 1; - } - } - } - - /*! - @brief returns the maximum possible number of elements - - Returns the maximum number of elements a JSON value is able to hold due to - system or library implementation limitations, i.e. `std::distance(begin(), - end())` for the JSON value. - - @return The return value depends on the different types and is - defined as follows: - Value type | return value - ----------- | ------------- - null | `0` (same as `size()`) - boolean | `1` (same as `size()`) - string | `1` (same as `size()`) - number | `1` (same as `size()`) - binary | `1` (same as `size()`) - object | result of function `object_t::max_size()` - array | result of function `array_t::max_size()` - - @liveexample{The following code calls `max_size()` on the different value - types. Note the output is implementation specific.,max_size} - - @complexity Constant, as long as @ref array_t and @ref object_t satisfy - the Container concept; that is, their `max_size()` functions have constant - complexity. - - @iterators No changes. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @requirement This function helps `basic_json` satisfying the - [Container](https://en.cppreference.com/w/cpp/named_req/Container) - requirements: - - The complexity is constant. - - Has the semantics of returning `b.size()` where `b` is the largest - possible JSON value. - - @sa @ref size() -- returns the number of elements - - @since version 1.0.0 - */ - size_type max_size() const noexcept - { - switch (m_type) - { - case value_t::array: - { - // delegate call to array_t::max_size() - return m_value.array->max_size(); - } - - case value_t::object: - { - // delegate call to object_t::max_size() - return m_value.object->max_size(); - } - - default: - { - // all other types have max_size() == size() - return size(); - } - } - } - - /// @} - - - /////////////// - // modifiers // - /////////////// - - /// @name modifiers - /// @{ - - /*! - @brief clears the contents - - Clears the content of a JSON value and resets it to the default value as - if @ref basic_json(value_t) would have been called with the current value - type from @ref type(): - - Value type | initial value - ----------- | ------------- - null | `null` - boolean | `false` - string | `""` - number | `0` - binary | An empty byte vector - object | `{}` - array | `[]` - - @post Has the same effect as calling - @code {.cpp} - *this = basic_json(type()); - @endcode - - @liveexample{The example below shows the effect of `clear()` to different - JSON types.,clear} - - @complexity Linear in the size of the JSON value. - - @iterators All iterators, pointers and references related to this container - are invalidated. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @sa @ref basic_json(value_t) -- constructor that creates an object with the - same value than calling `clear()` - - @since version 1.0.0 - */ - void clear() noexcept - { - switch (m_type) - { - case value_t::number_integer: - { - m_value.number_integer = 0; - break; - } - - case value_t::number_unsigned: - { - m_value.number_unsigned = 0; - break; - } - - case value_t::number_float: - { - m_value.number_float = 0.0; - break; - } - - case value_t::boolean: - { - m_value.boolean = false; - break; - } - - case value_t::string: - { - m_value.string->clear(); - break; - } - - case value_t::binary: - { - m_value.binary->clear(); - break; - } - - case value_t::array: - { - m_value.array->clear(); - break; - } - - case value_t::object: - { - m_value.object->clear(); - break; - } - - default: - break; - } - } - - /*! - @brief add an object to an array - - Appends the given element @a val to the end of the JSON value. If the - function is called on a JSON null value, an empty array is created before - appending @a val. - - @param[in] val the value to add to the JSON array - - @throw type_error.308 when called on a type other than JSON array or - null; example: `"cannot use push_back() with number"` - - @complexity Amortized constant. - - @liveexample{The example shows how `push_back()` and `+=` can be used to - add elements to a JSON array. Note how the `null` value was silently - converted to a JSON array.,push_back} - - @since version 1.0.0 - */ - void push_back(basic_json&& val) - { - // push_back only works for null objects or arrays - if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array()))) - { - JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()))); - } - - // transform null object into an array - if (is_null()) - { - m_type = value_t::array; - m_value = value_t::array; - assert_invariant(); - } - - // add element to array (move semantics) - m_value.array->push_back(std::move(val)); - // if val is moved from, basic_json move constructor marks it null so we do not call the destructor - } - - /*! - @brief add an object to an array - @copydoc push_back(basic_json&&) - */ - reference operator+=(basic_json&& val) - { - push_back(std::move(val)); - return *this; - } - - /*! - @brief add an object to an array - @copydoc push_back(basic_json&&) - */ - void push_back(const basic_json& val) - { - // push_back only works for null objects or arrays - if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array()))) - { - JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()))); - } - - // transform null object into an array - if (is_null()) - { - m_type = value_t::array; - m_value = value_t::array; - assert_invariant(); - } - - // add element to array - m_value.array->push_back(val); - } - - /*! - @brief add an object to an array - @copydoc push_back(basic_json&&) - */ - reference operator+=(const basic_json& val) - { - push_back(val); - return *this; - } - - /*! - @brief add an object to an object - - Inserts the given element @a val to the JSON object. If the function is - called on a JSON null value, an empty object is created before inserting - @a val. - - @param[in] val the value to add to the JSON object - - @throw type_error.308 when called on a type other than JSON object or - null; example: `"cannot use push_back() with number"` - - @complexity Logarithmic in the size of the container, O(log(`size()`)). - - @liveexample{The example shows how `push_back()` and `+=` can be used to - add elements to a JSON object. Note how the `null` value was silently - converted to a JSON object.,push_back__object_t__value} - - @since version 1.0.0 - */ - void push_back(const typename object_t::value_type& val) - { - // push_back only works for null objects or objects - if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_object()))) - { - JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()))); - } - - // transform null object into an object - if (is_null()) - { - m_type = value_t::object; - m_value = value_t::object; - assert_invariant(); - } - - // add element to array - m_value.object->insert(val); - } - - /*! - @brief add an object to an object - @copydoc push_back(const typename object_t::value_type&) - */ - reference operator+=(const typename object_t::value_type& val) - { - push_back(val); - return *this; - } - - /*! - @brief add an object to an object - - This function allows to use `push_back` with an initializer list. In case - - 1. the current value is an object, - 2. the initializer list @a init contains only two elements, and - 3. the first element of @a init is a string, - - @a init is converted into an object element and added using - @ref push_back(const typename object_t::value_type&). Otherwise, @a init - is converted to a JSON value and added using @ref push_back(basic_json&&). - - @param[in] init an initializer list - - @complexity Linear in the size of the initializer list @a init. - - @note This function is required to resolve an ambiguous overload error, - because pairs like `{"key", "value"}` can be both interpreted as - `object_t::value_type` or `std::initializer_list`, see - https://github.com/nlohmann/json/issues/235 for more information. - - @liveexample{The example shows how initializer lists are treated as - objects when possible.,push_back__initializer_list} - */ - void push_back(initializer_list_t init) - { - if (is_object() && init.size() == 2 && (*init.begin())->is_string()) - { - basic_json&& key = init.begin()->moved_or_copied(); - push_back(typename object_t::value_type( - std::move(key.get_ref()), (init.begin() + 1)->moved_or_copied())); - } - else - { - push_back(basic_json(init)); - } - } - - /*! - @brief add an object to an object - @copydoc push_back(initializer_list_t) - */ - reference operator+=(initializer_list_t init) - { - push_back(init); - return *this; - } - - /*! - @brief add an object to an array - - Creates a JSON value from the passed parameters @a args to the end of the - JSON value. If the function is called on a JSON null value, an empty array - is created before appending the value created from @a args. - - @param[in] args arguments to forward to a constructor of @ref basic_json - @tparam Args compatible types to create a @ref basic_json object - - @return reference to the inserted element - - @throw type_error.311 when called on a type other than JSON array or - null; example: `"cannot use emplace_back() with number"` - - @complexity Amortized constant. - - @liveexample{The example shows how `push_back()` can be used to add - elements to a JSON array. Note how the `null` value was silently converted - to a JSON array.,emplace_back} - - @since version 2.0.8, returns reference since 3.7.0 - */ - template - reference emplace_back(Args&& ... args) - { - // emplace_back only works for null objects or arrays - if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array()))) - { - JSON_THROW(type_error::create(311, "cannot use emplace_back() with " + std::string(type_name()))); - } - - // transform null object into an array - if (is_null()) - { - m_type = value_t::array; - m_value = value_t::array; - assert_invariant(); - } - - // add element to array (perfect forwarding) -#ifdef JSON_HAS_CPP_17 - return m_value.array->emplace_back(std::forward(args)...); -#else - m_value.array->emplace_back(std::forward(args)...); - return m_value.array->back(); -#endif - } - - /*! - @brief add an object to an object if key does not exist - - Inserts a new element into a JSON object constructed in-place with the - given @a args if there is no element with the key in the container. If the - function is called on a JSON null value, an empty object is created before - appending the value created from @a args. - - @param[in] args arguments to forward to a constructor of @ref basic_json - @tparam Args compatible types to create a @ref basic_json object - - @return a pair consisting of an iterator to the inserted element, or the - already-existing element if no insertion happened, and a bool - denoting whether the insertion took place. - - @throw type_error.311 when called on a type other than JSON object or - null; example: `"cannot use emplace() with number"` - - @complexity Logarithmic in the size of the container, O(log(`size()`)). - - @liveexample{The example shows how `emplace()` can be used to add elements - to a JSON object. Note how the `null` value was silently converted to a - JSON object. Further note how no value is added if there was already one - value stored with the same key.,emplace} - - @since version 2.0.8 - */ - template - std::pair emplace(Args&& ... args) - { - // emplace only works for null objects or arrays - if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_object()))) - { - JSON_THROW(type_error::create(311, "cannot use emplace() with " + std::string(type_name()))); - } - - // transform null object into an object - if (is_null()) - { - m_type = value_t::object; - m_value = value_t::object; - assert_invariant(); - } - - // add element to array (perfect forwarding) - auto res = m_value.object->emplace(std::forward(args)...); - // create result iterator and set iterator to the result of emplace - auto it = begin(); - it.m_it.object_iterator = res.first; - - // return pair of iterator and boolean - return {it, res.second}; - } - - /// Helper for insertion of an iterator - /// @note: This uses std::distance to support GCC 4.8, - /// see https://github.com/nlohmann/json/pull/1257 - template - iterator insert_iterator(const_iterator pos, Args&& ... args) - { - iterator result(this); - JSON_ASSERT(m_value.array != nullptr); - - auto insert_pos = std::distance(m_value.array->begin(), pos.m_it.array_iterator); - m_value.array->insert(pos.m_it.array_iterator, std::forward(args)...); - result.m_it.array_iterator = m_value.array->begin() + insert_pos; - - // This could have been written as: - // result.m_it.array_iterator = m_value.array->insert(pos.m_it.array_iterator, cnt, val); - // but the return value of insert is missing in GCC 4.8, so it is written this way instead. - - return result; - } - - /*! - @brief inserts element - - Inserts element @a val before iterator @a pos. - - @param[in] pos iterator before which the content will be inserted; may be - the end() iterator - @param[in] val element to insert - @return iterator pointing to the inserted @a val. - - @throw type_error.309 if called on JSON values other than arrays; - example: `"cannot use insert() with string"` - @throw invalid_iterator.202 if @a pos is not an iterator of *this; - example: `"iterator does not fit current value"` - - @complexity Constant plus linear in the distance between @a pos and end of - the container. - - @liveexample{The example shows how `insert()` is used.,insert} - - @since version 1.0.0 - */ - iterator insert(const_iterator pos, const basic_json& val) - { - // insert only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - // check if iterator pos fits to this JSON value - if (JSON_HEDLEY_UNLIKELY(pos.m_object != this)) - { - JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value")); - } - - // insert to array and return iterator - return insert_iterator(pos, val); - } - - JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()))); - } - - /*! - @brief inserts element - @copydoc insert(const_iterator, const basic_json&) - */ - iterator insert(const_iterator pos, basic_json&& val) - { - return insert(pos, val); - } - - /*! - @brief inserts elements - - Inserts @a cnt copies of @a val before iterator @a pos. - - @param[in] pos iterator before which the content will be inserted; may be - the end() iterator - @param[in] cnt number of copies of @a val to insert - @param[in] val element to insert - @return iterator pointing to the first element inserted, or @a pos if - `cnt==0` - - @throw type_error.309 if called on JSON values other than arrays; example: - `"cannot use insert() with string"` - @throw invalid_iterator.202 if @a pos is not an iterator of *this; - example: `"iterator does not fit current value"` - - @complexity Linear in @a cnt plus linear in the distance between @a pos - and end of the container. - - @liveexample{The example shows how `insert()` is used.,insert__count} - - @since version 1.0.0 - */ - iterator insert(const_iterator pos, size_type cnt, const basic_json& val) - { - // insert only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - // check if iterator pos fits to this JSON value - if (JSON_HEDLEY_UNLIKELY(pos.m_object != this)) - { - JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value")); - } - - // insert to array and return iterator - return insert_iterator(pos, cnt, val); - } - - JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()))); - } - - /*! - @brief inserts elements - - Inserts elements from range `[first, last)` before iterator @a pos. - - @param[in] pos iterator before which the content will be inserted; may be - the end() iterator - @param[in] first begin of the range of elements to insert - @param[in] last end of the range of elements to insert - - @throw type_error.309 if called on JSON values other than arrays; example: - `"cannot use insert() with string"` - @throw invalid_iterator.202 if @a pos is not an iterator of *this; - example: `"iterator does not fit current value"` - @throw invalid_iterator.210 if @a first and @a last do not belong to the - same JSON value; example: `"iterators do not fit"` - @throw invalid_iterator.211 if @a first or @a last are iterators into - container for which insert is called; example: `"passed iterators may not - belong to container"` - - @return iterator pointing to the first element inserted, or @a pos if - `first==last` - - @complexity Linear in `std::distance(first, last)` plus linear in the - distance between @a pos and end of the container. - - @liveexample{The example shows how `insert()` is used.,insert__range} - - @since version 1.0.0 - */ - iterator insert(const_iterator pos, const_iterator first, const_iterator last) - { - // insert only works for arrays - if (JSON_HEDLEY_UNLIKELY(!is_array())) - { - JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()))); - } - - // check if iterator pos fits to this JSON value - if (JSON_HEDLEY_UNLIKELY(pos.m_object != this)) - { - JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value")); - } - - // check if range iterators belong to the same JSON object - if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object)) - { - JSON_THROW(invalid_iterator::create(210, "iterators do not fit")); - } - - if (JSON_HEDLEY_UNLIKELY(first.m_object == this)) - { - JSON_THROW(invalid_iterator::create(211, "passed iterators may not belong to container")); - } - - // insert to array and return iterator - return insert_iterator(pos, first.m_it.array_iterator, last.m_it.array_iterator); - } - - /*! - @brief inserts elements - - Inserts elements from initializer list @a ilist before iterator @a pos. - - @param[in] pos iterator before which the content will be inserted; may be - the end() iterator - @param[in] ilist initializer list to insert the values from - - @throw type_error.309 if called on JSON values other than arrays; example: - `"cannot use insert() with string"` - @throw invalid_iterator.202 if @a pos is not an iterator of *this; - example: `"iterator does not fit current value"` - - @return iterator pointing to the first element inserted, or @a pos if - `ilist` is empty - - @complexity Linear in `ilist.size()` plus linear in the distance between - @a pos and end of the container. - - @liveexample{The example shows how `insert()` is used.,insert__ilist} - - @since version 1.0.0 - */ - iterator insert(const_iterator pos, initializer_list_t ilist) - { - // insert only works for arrays - if (JSON_HEDLEY_UNLIKELY(!is_array())) - { - JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()))); - } - - // check if iterator pos fits to this JSON value - if (JSON_HEDLEY_UNLIKELY(pos.m_object != this)) - { - JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value")); - } - - // insert to array and return iterator - return insert_iterator(pos, ilist.begin(), ilist.end()); - } - - /*! - @brief inserts elements - - Inserts elements from range `[first, last)`. - - @param[in] first begin of the range of elements to insert - @param[in] last end of the range of elements to insert - - @throw type_error.309 if called on JSON values other than objects; example: - `"cannot use insert() with string"` - @throw invalid_iterator.202 if iterator @a first or @a last does does not - point to an object; example: `"iterators first and last must point to - objects"` - @throw invalid_iterator.210 if @a first and @a last do not belong to the - same JSON value; example: `"iterators do not fit"` - - @complexity Logarithmic: `O(N*log(size() + N))`, where `N` is the number - of elements to insert. - - @liveexample{The example shows how `insert()` is used.,insert__range_object} - - @since version 3.0.0 - */ - void insert(const_iterator first, const_iterator last) - { - // insert only works for objects - if (JSON_HEDLEY_UNLIKELY(!is_object())) - { - JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()))); - } - - // check if range iterators belong to the same JSON object - if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object)) - { - JSON_THROW(invalid_iterator::create(210, "iterators do not fit")); - } - - // passed iterators must belong to objects - if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_object())) - { - JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to objects")); - } - - m_value.object->insert(first.m_it.object_iterator, last.m_it.object_iterator); - } - - /*! - @brief updates a JSON object from another object, overwriting existing keys - - Inserts all values from JSON object @a j and overwrites existing keys. - - @param[in] j JSON object to read values from - - @throw type_error.312 if called on JSON values other than objects; example: - `"cannot use update() with string"` - - @complexity O(N*log(size() + N)), where N is the number of elements to - insert. - - @liveexample{The example shows how `update()` is used.,update} - - @sa https://docs.python.org/3.6/library/stdtypes.html#dict.update - - @since version 3.0.0 - */ - void update(const_reference j) - { - // implicitly convert null value to an empty object - if (is_null()) - { - m_type = value_t::object; - m_value.object = create(); - assert_invariant(); - } - - if (JSON_HEDLEY_UNLIKELY(!is_object())) - { - JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(type_name()))); - } - if (JSON_HEDLEY_UNLIKELY(!j.is_object())) - { - JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(j.type_name()))); - } - - for (auto it = j.cbegin(); it != j.cend(); ++it) - { - m_value.object->operator[](it.key()) = it.value(); - } - } - - /*! - @brief updates a JSON object from another object, overwriting existing keys - - Inserts all values from from range `[first, last)` and overwrites existing - keys. - - @param[in] first begin of the range of elements to insert - @param[in] last end of the range of elements to insert - - @throw type_error.312 if called on JSON values other than objects; example: - `"cannot use update() with string"` - @throw invalid_iterator.202 if iterator @a first or @a last does does not - point to an object; example: `"iterators first and last must point to - objects"` - @throw invalid_iterator.210 if @a first and @a last do not belong to the - same JSON value; example: `"iterators do not fit"` - - @complexity O(N*log(size() + N)), where N is the number of elements to - insert. - - @liveexample{The example shows how `update()` is used__range.,update} - - @sa https://docs.python.org/3.6/library/stdtypes.html#dict.update - - @since version 3.0.0 - */ - void update(const_iterator first, const_iterator last) - { - // implicitly convert null value to an empty object - if (is_null()) - { - m_type = value_t::object; - m_value.object = create(); - assert_invariant(); - } - - if (JSON_HEDLEY_UNLIKELY(!is_object())) - { - JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(type_name()))); - } - - // check if range iterators belong to the same JSON object - if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object)) - { - JSON_THROW(invalid_iterator::create(210, "iterators do not fit")); - } - - // passed iterators must belong to objects - if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_object() - || !last.m_object->is_object())) - { - JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to objects")); - } - - for (auto it = first; it != last; ++it) - { - m_value.object->operator[](it.key()) = it.value(); - } - } - - /*! - @brief exchanges the values - - Exchanges the contents of the JSON value with those of @a other. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. - - @param[in,out] other JSON value to exchange the contents with - - @complexity Constant. - - @liveexample{The example below shows how JSON values can be swapped with - `swap()`.,swap__reference} - - @since version 1.0.0 - */ - void swap(reference other) noexcept ( - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value&& - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value - ) - { - std::swap(m_type, other.m_type); - std::swap(m_value, other.m_value); - assert_invariant(); - } - - /*! - @brief exchanges the values - - Exchanges the contents of the JSON value from @a left with those of @a right. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. implemented as a friend function callable via ADL. - - @param[in,out] left JSON value to exchange the contents with - @param[in,out] right JSON value to exchange the contents with - - @complexity Constant. - - @liveexample{The example below shows how JSON values can be swapped with - `swap()`.,swap__reference} - - @since version 1.0.0 - */ - friend void swap(reference left, reference right) noexcept ( - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value&& - std::is_nothrow_move_constructible::value&& - std::is_nothrow_move_assignable::value - ) - { - left.swap(right); - } - - /*! - @brief exchanges the values - - Exchanges the contents of a JSON array with those of @a other. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. - - @param[in,out] other array to exchange the contents with - - @throw type_error.310 when JSON value is not an array; example: `"cannot - use swap() with string"` - - @complexity Constant. - - @liveexample{The example below shows how arrays can be swapped with - `swap()`.,swap__array_t} - - @since version 1.0.0 - */ - void swap(array_t& other) - { - // swap only works for arrays - if (JSON_HEDLEY_LIKELY(is_array())) - { - std::swap(*(m_value.array), other); - } - else - { - JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()))); - } - } - - /*! - @brief exchanges the values - - Exchanges the contents of a JSON object with those of @a other. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. - - @param[in,out] other object to exchange the contents with - - @throw type_error.310 when JSON value is not an object; example: - `"cannot use swap() with string"` - - @complexity Constant. - - @liveexample{The example below shows how objects can be swapped with - `swap()`.,swap__object_t} - - @since version 1.0.0 - */ - void swap(object_t& other) - { - // swap only works for objects - if (JSON_HEDLEY_LIKELY(is_object())) - { - std::swap(*(m_value.object), other); - } - else - { - JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()))); - } - } - - /*! - @brief exchanges the values - - Exchanges the contents of a JSON string with those of @a other. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. - - @param[in,out] other string to exchange the contents with - - @throw type_error.310 when JSON value is not a string; example: `"cannot - use swap() with boolean"` - - @complexity Constant. - - @liveexample{The example below shows how strings can be swapped with - `swap()`.,swap__string_t} - - @since version 1.0.0 - */ - void swap(string_t& other) - { - // swap only works for strings - if (JSON_HEDLEY_LIKELY(is_string())) - { - std::swap(*(m_value.string), other); - } - else - { - JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()))); - } - } - - /*! - @brief exchanges the values - - Exchanges the contents of a JSON string with those of @a other. Does not - invoke any move, copy, or swap operations on individual elements. All - iterators and references remain valid. The past-the-end iterator is - invalidated. - - @param[in,out] other binary to exchange the contents with - - @throw type_error.310 when JSON value is not a string; example: `"cannot - use swap() with boolean"` - - @complexity Constant. - - @liveexample{The example below shows how strings can be swapped with - `swap()`.,swap__binary_t} - - @since version 3.8.0 - */ - void swap(binary_t& other) - { - // swap only works for strings - if (JSON_HEDLEY_LIKELY(is_binary())) - { - std::swap(*(m_value.binary), other); - } - else - { - JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()))); - } - } - - /// @copydoc swap(binary_t) - void swap(typename binary_t::container_type& other) - { - // swap only works for strings - if (JSON_HEDLEY_LIKELY(is_binary())) - { - std::swap(*(m_value.binary), other); - } - else - { - JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()))); - } - } - - /// @} - - public: - ////////////////////////////////////////// - // lexicographical comparison operators // - ////////////////////////////////////////// - - /// @name lexicographical comparison operators - /// @{ - - /*! - @brief comparison: equal - - Compares two JSON values for equality according to the following rules: - - Two JSON values are equal if (1) they are from the same type and (2) - their stored values are the same according to their respective - `operator==`. - - Integer and floating-point numbers are automatically converted before - comparison. Note that two NaN values are always treated as unequal. - - Two JSON null values are equal. - - @note Floating-point inside JSON values numbers are compared with - `json::number_float_t::operator==` which is `double::operator==` by - default. To compare floating-point while respecting an epsilon, an alternative - [comparison function](https://github.com/mariokonrad/marnav/blob/master/include/marnav/math/floatingpoint.hpp#L34-#L39) - could be used, for instance - @code {.cpp} - template::value, T>::type> - inline bool is_same(T a, T b, T epsilon = std::numeric_limits::epsilon()) noexcept - { - return std::abs(a - b) <= epsilon; - } - @endcode - Or you can self-defined operator equal function like this: - @code {.cpp} - bool my_equal(const_reference lhs, const_reference rhs) { - const auto lhs_type lhs.type(); - const auto rhs_type rhs.type(); - if (lhs_type == rhs_type) { - switch(lhs_type) - // self_defined case - case value_t::number_float: - return std::abs(lhs - rhs) <= std::numeric_limits::epsilon(); - // other cases remain the same with the original - ... - } - ... - } - @endcode - - @note NaN values never compare equal to themselves or to other NaN values. - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether the values @a lhs and @a rhs are equal - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @complexity Linear. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__equal} - - @since version 1.0.0 - */ - friend bool operator==(const_reference lhs, const_reference rhs) noexcept - { - const auto lhs_type = lhs.type(); - const auto rhs_type = rhs.type(); - - if (lhs_type == rhs_type) - { - switch (lhs_type) - { - case value_t::array: - return *lhs.m_value.array == *rhs.m_value.array; - - case value_t::object: - return *lhs.m_value.object == *rhs.m_value.object; - - case value_t::null: - return true; - - case value_t::string: - return *lhs.m_value.string == *rhs.m_value.string; - - case value_t::boolean: - return lhs.m_value.boolean == rhs.m_value.boolean; - - case value_t::number_integer: - return lhs.m_value.number_integer == rhs.m_value.number_integer; - - case value_t::number_unsigned: - return lhs.m_value.number_unsigned == rhs.m_value.number_unsigned; - - case value_t::number_float: - return lhs.m_value.number_float == rhs.m_value.number_float; - - case value_t::binary: - return *lhs.m_value.binary == *rhs.m_value.binary; - - default: - return false; - } - } - else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float) - { - return static_cast(lhs.m_value.number_integer) == rhs.m_value.number_float; - } - else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer) - { - return lhs.m_value.number_float == static_cast(rhs.m_value.number_integer); - } - else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float) - { - return static_cast(lhs.m_value.number_unsigned) == rhs.m_value.number_float; - } - else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned) - { - return lhs.m_value.number_float == static_cast(rhs.m_value.number_unsigned); - } - else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer) - { - return static_cast(lhs.m_value.number_unsigned) == rhs.m_value.number_integer; - } - else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_unsigned) - { - return lhs.m_value.number_integer == static_cast(rhs.m_value.number_unsigned); - } - - return false; - } - - /*! - @brief comparison: equal - @copydoc operator==(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator==(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs == basic_json(rhs); - } - - /*! - @brief comparison: equal - @copydoc operator==(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator==(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) == rhs; - } - - /*! - @brief comparison: not equal - - Compares two JSON values for inequality by calculating `not (lhs == rhs)`. - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether the values @a lhs and @a rhs are not equal - - @complexity Linear. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__notequal} - - @since version 1.0.0 - */ - friend bool operator!=(const_reference lhs, const_reference rhs) noexcept - { - return !(lhs == rhs); - } - - /*! - @brief comparison: not equal - @copydoc operator!=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator!=(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs != basic_json(rhs); - } - - /*! - @brief comparison: not equal - @copydoc operator!=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator!=(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) != rhs; - } - - /*! - @brief comparison: less than - - Compares whether one JSON value @a lhs is less than another JSON value @a - rhs according to the following rules: - - If @a lhs and @a rhs have the same type, the values are compared using - the default `<` operator. - - Integer and floating-point numbers are automatically converted before - comparison - - In case @a lhs and @a rhs have different types, the values are ignored - and the order of the types is considered, see - @ref operator<(const value_t, const value_t). - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether @a lhs is less than @a rhs - - @complexity Linear. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__less} - - @since version 1.0.0 - */ - friend bool operator<(const_reference lhs, const_reference rhs) noexcept - { - const auto lhs_type = lhs.type(); - const auto rhs_type = rhs.type(); - - if (lhs_type == rhs_type) - { - switch (lhs_type) - { - case value_t::array: - // note parentheses are necessary, see - // https://github.com/nlohmann/json/issues/1530 - return (*lhs.m_value.array) < (*rhs.m_value.array); - - case value_t::object: - return (*lhs.m_value.object) < (*rhs.m_value.object); - - case value_t::null: - return false; - - case value_t::string: - return (*lhs.m_value.string) < (*rhs.m_value.string); - - case value_t::boolean: - return (lhs.m_value.boolean) < (rhs.m_value.boolean); - - case value_t::number_integer: - return (lhs.m_value.number_integer) < (rhs.m_value.number_integer); - - case value_t::number_unsigned: - return (lhs.m_value.number_unsigned) < (rhs.m_value.number_unsigned); - - case value_t::number_float: - return (lhs.m_value.number_float) < (rhs.m_value.number_float); - - case value_t::binary: - return (*lhs.m_value.binary) < (*rhs.m_value.binary); - - default: - return false; - } - } - else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float) - { - return static_cast(lhs.m_value.number_integer) < rhs.m_value.number_float; - } - else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer) - { - return lhs.m_value.number_float < static_cast(rhs.m_value.number_integer); - } - else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float) - { - return static_cast(lhs.m_value.number_unsigned) < rhs.m_value.number_float; - } - else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned) - { - return lhs.m_value.number_float < static_cast(rhs.m_value.number_unsigned); - } - else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_unsigned) - { - return lhs.m_value.number_integer < static_cast(rhs.m_value.number_unsigned); - } - else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer) - { - return static_cast(lhs.m_value.number_unsigned) < rhs.m_value.number_integer; - } - - // We only reach this line if we cannot compare values. In that case, - // we compare types. Note we have to call the operator explicitly, - // because MSVC has problems otherwise. - return operator<(lhs_type, rhs_type); - } - - /*! - @brief comparison: less than - @copydoc operator<(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator<(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs < basic_json(rhs); - } - - /*! - @brief comparison: less than - @copydoc operator<(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator<(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) < rhs; - } - - /*! - @brief comparison: less than or equal - - Compares whether one JSON value @a lhs is less than or equal to another - JSON value by calculating `not (rhs < lhs)`. - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether @a lhs is less than or equal to @a rhs - - @complexity Linear. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__greater} - - @since version 1.0.0 - */ - friend bool operator<=(const_reference lhs, const_reference rhs) noexcept - { - return !(rhs < lhs); - } - - /*! - @brief comparison: less than or equal - @copydoc operator<=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator<=(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs <= basic_json(rhs); - } - - /*! - @brief comparison: less than or equal - @copydoc operator<=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator<=(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) <= rhs; - } - - /*! - @brief comparison: greater than - - Compares whether one JSON value @a lhs is greater than another - JSON value by calculating `not (lhs <= rhs)`. - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether @a lhs is greater than to @a rhs - - @complexity Linear. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__lessequal} - - @since version 1.0.0 - */ - friend bool operator>(const_reference lhs, const_reference rhs) noexcept - { - return !(lhs <= rhs); - } - - /*! - @brief comparison: greater than - @copydoc operator>(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator>(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs > basic_json(rhs); - } - - /*! - @brief comparison: greater than - @copydoc operator>(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator>(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) > rhs; - } - - /*! - @brief comparison: greater than or equal - - Compares whether one JSON value @a lhs is greater than or equal to another - JSON value by calculating `not (lhs < rhs)`. - - @param[in] lhs first JSON value to consider - @param[in] rhs second JSON value to consider - @return whether @a lhs is greater than or equal to @a rhs - - @complexity Linear. - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @liveexample{The example demonstrates comparing several JSON - types.,operator__greaterequal} - - @since version 1.0.0 - */ - friend bool operator>=(const_reference lhs, const_reference rhs) noexcept - { - return !(lhs < rhs); - } - - /*! - @brief comparison: greater than or equal - @copydoc operator>=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator>=(const_reference lhs, const ScalarType rhs) noexcept - { - return lhs >= basic_json(rhs); - } - - /*! - @brief comparison: greater than or equal - @copydoc operator>=(const_reference, const_reference) - */ - template::value, int>::type = 0> - friend bool operator>=(const ScalarType lhs, const_reference rhs) noexcept - { - return basic_json(lhs) >= rhs; - } - - /// @} - - /////////////////// - // serialization // - /////////////////// - - /// @name serialization - /// @{ - - /*! - @brief serialize to stream - - Serialize the given JSON value @a j to the output stream @a o. The JSON - value will be serialized using the @ref dump member function. - - - The indentation of the output can be controlled with the member variable - `width` of the output stream @a o. For instance, using the manipulator - `std::setw(4)` on @a o sets the indentation level to `4` and the - serialization result is the same as calling `dump(4)`. - - - The indentation character can be controlled with the member variable - `fill` of the output stream @a o. For instance, the manipulator - `std::setfill('\\t')` sets indentation to use a tab character rather than - the default space character. - - @param[in,out] o stream to serialize to - @param[in] j JSON value to serialize - - @return the stream @a o - - @throw type_error.316 if a string stored inside the JSON value is not - UTF-8 encoded - - @complexity Linear. - - @liveexample{The example below shows the serialization with different - parameters to `width` to adjust the indentation level.,operator_serialize} - - @since version 1.0.0; indentation character added in version 3.0.0 - */ - friend std::ostream& operator<<(std::ostream& o, const basic_json& j) - { - // read width member and use it as indentation parameter if nonzero - const bool pretty_print = o.width() > 0; - const auto indentation = pretty_print ? o.width() : 0; - - // reset width to 0 for subsequent calls to this stream - o.width(0); - - // do the actual serialization - serializer s(detail::output_adapter(o), o.fill()); - s.dump(j, pretty_print, false, static_cast(indentation)); - return o; - } - - /*! - @brief serialize to stream - @deprecated This stream operator is deprecated and will be removed in - future 4.0.0 of the library. Please use - @ref operator<<(std::ostream&, const basic_json&) - instead; that is, replace calls like `j >> o;` with `o << j;`. - @since version 1.0.0; deprecated since version 3.0.0 - */ - JSON_HEDLEY_DEPRECATED_FOR(3.0.0, operator<<(std::ostream&, const basic_json&)) - friend std::ostream& operator>>(const basic_json& j, std::ostream& o) - { - return o << j; - } - - /// @} - - - ///////////////////// - // deserialization // - ///////////////////// - - /// @name deserialization - /// @{ - - /*! - @brief deserialize from a compatible input - - @tparam InputType A compatible input, for instance - - an std::istream object - - a FILE pointer - - a C-style array of characters - - a pointer to a null-terminated string of single byte characters - - an object obj for which begin(obj) and end(obj) produces a valid pair of - iterators. - - @param[in] i input to read from - @param[in] cb a parser callback function of type @ref parser_callback_t - which is used to control the deserialization by filtering unwanted values - (optional) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - @param[in] ignore_comments whether comments should be ignored and treated - like whitespace (true) or yield a parse error (true); (optional, false by - default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.101 if a parse error occurs; example: `""unexpected end - of input; expected string literal""` - @throw parse_error.102 if to_unicode fails or surrogate error - @throw parse_error.103 if to_unicode fails - - @complexity Linear in the length of the input. The parser is a predictive - LL(1) parser. The complexity can be higher if the parser callback function - @a cb or reading from the input @a i has a super-linear complexity. - - @note A UTF-8 byte order mark is silently ignored. - - @liveexample{The example below demonstrates the `parse()` function reading - from an array.,parse__array__parser_callback_t} - - @liveexample{The example below demonstrates the `parse()` function with - and without callback function.,parse__string__parser_callback_t} - - @liveexample{The example below demonstrates the `parse()` function with - and without callback function.,parse__istream__parser_callback_t} - - @liveexample{The example below demonstrates the `parse()` function reading - from a contiguous container.,parse__contiguouscontainer__parser_callback_t} - - @since version 2.0.3 (contiguous containers); version 3.9.0 allowed to - ignore comments. - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json parse(InputType&& i, - const parser_callback_t cb = nullptr, - const bool allow_exceptions = true, - const bool ignore_comments = false) - { - basic_json result; - parser(detail::input_adapter(std::forward(i)), cb, allow_exceptions, ignore_comments).parse(true, result); - return result; - } - - /*! - @brief deserialize from a pair of character iterators - - The value_type of the iterator must be a integral type with size of 1, 2 or - 4 bytes, which will be interpreted respectively as UTF-8, UTF-16 and UTF-32. - - @param[in] first iterator to start of character range - @param[in] last iterator to end of character range - @param[in] cb a parser callback function of type @ref parser_callback_t - which is used to control the deserialization by filtering unwanted values - (optional) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - @param[in] ignore_comments whether comments should be ignored and treated - like whitespace (true) or yield a parse error (true); (optional, false by - default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.101 if a parse error occurs; example: `""unexpected end - of input; expected string literal""` - @throw parse_error.102 if to_unicode fails or surrogate error - @throw parse_error.103 if to_unicode fails - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json parse(IteratorType first, - IteratorType last, - const parser_callback_t cb = nullptr, - const bool allow_exceptions = true, - const bool ignore_comments = false) - { - basic_json result; - parser(detail::input_adapter(std::move(first), std::move(last)), cb, allow_exceptions, ignore_comments).parse(true, result); - return result; - } - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, parse(ptr, ptr + len)) - static basic_json parse(detail::span_input_adapter&& i, - const parser_callback_t cb = nullptr, - const bool allow_exceptions = true, - const bool ignore_comments = false) - { - basic_json result; - parser(i.get(), cb, allow_exceptions, ignore_comments).parse(true, result); - return result; - } - - /*! - @brief check if the input is valid JSON - - Unlike the @ref parse(InputType&&, const parser_callback_t,const bool) - function, this function neither throws an exception in case of invalid JSON - input (i.e., a parse error) nor creates diagnostic information. - - @tparam InputType A compatible input, for instance - - an std::istream object - - a FILE pointer - - a C-style array of characters - - a pointer to a null-terminated string of single byte characters - - an object obj for which begin(obj) and end(obj) produces a valid pair of - iterators. - - @param[in] i input to read from - @param[in] ignore_comments whether comments should be ignored and treated - like whitespace (true) or yield a parse error (true); (optional, false by - default) - - @return Whether the input read from @a i is valid JSON. - - @complexity Linear in the length of the input. The parser is a predictive - LL(1) parser. - - @note A UTF-8 byte order mark is silently ignored. - - @liveexample{The example below demonstrates the `accept()` function reading - from a string.,accept__string} - */ - template - static bool accept(InputType&& i, - const bool ignore_comments = false) - { - return parser(detail::input_adapter(std::forward(i)), nullptr, false, ignore_comments).accept(true); - } - - template - static bool accept(IteratorType first, IteratorType last, - const bool ignore_comments = false) - { - return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments).accept(true); - } - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, accept(ptr, ptr + len)) - static bool accept(detail::span_input_adapter&& i, - const bool ignore_comments = false) - { - return parser(i.get(), nullptr, false, ignore_comments).accept(true); - } - - /*! - @brief generate SAX events - - The SAX event lister must follow the interface of @ref json_sax. - - This function reads from a compatible input. Examples are: - - an std::istream object - - a FILE pointer - - a C-style array of characters - - a pointer to a null-terminated string of single byte characters - - an object obj for which begin(obj) and end(obj) produces a valid pair of - iterators. - - @param[in] i input to read from - @param[in,out] sax SAX event listener - @param[in] format the format to parse (JSON, CBOR, MessagePack, or UBJSON) - @param[in] strict whether the input has to be consumed completely - @param[in] ignore_comments whether comments should be ignored and treated - like whitespace (true) or yield a parse error (true); (optional, false by - default); only applies to the JSON file format. - - @return return value of the last processed SAX event - - @throw parse_error.101 if a parse error occurs; example: `""unexpected end - of input; expected string literal""` - @throw parse_error.102 if to_unicode fails or surrogate error - @throw parse_error.103 if to_unicode fails - - @complexity Linear in the length of the input. The parser is a predictive - LL(1) parser. The complexity can be higher if the SAX consumer @a sax has - a super-linear complexity. - - @note A UTF-8 byte order mark is silently ignored. - - @liveexample{The example below demonstrates the `sax_parse()` function - reading from string and processing the events with a user-defined SAX - event consumer.,sax_parse} - - @since version 3.2.0 - */ - template - JSON_HEDLEY_NON_NULL(2) - static bool sax_parse(InputType&& i, SAX* sax, - input_format_t format = input_format_t::json, - const bool strict = true, - const bool ignore_comments = false) - { - auto ia = detail::input_adapter(std::forward(i)); - return format == input_format_t::json - ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict) - : detail::binary_reader(std::move(ia)).sax_parse(format, sax, strict); - } - - template - JSON_HEDLEY_NON_NULL(3) - static bool sax_parse(IteratorType first, IteratorType last, SAX* sax, - input_format_t format = input_format_t::json, - const bool strict = true, - const bool ignore_comments = false) - { - auto ia = detail::input_adapter(std::move(first), std::move(last)); - return format == input_format_t::json - ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict) - : detail::binary_reader(std::move(ia)).sax_parse(format, sax, strict); - } - - template - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, sax_parse(ptr, ptr + len, ...)) - JSON_HEDLEY_NON_NULL(2) - static bool sax_parse(detail::span_input_adapter&& i, SAX* sax, - input_format_t format = input_format_t::json, - const bool strict = true, - const bool ignore_comments = false) - { - auto ia = i.get(); - return format == input_format_t::json - ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict) - : detail::binary_reader(std::move(ia)).sax_parse(format, sax, strict); - } - - /*! - @brief deserialize from stream - @deprecated This stream operator is deprecated and will be removed in - version 4.0.0 of the library. Please use - @ref operator>>(std::istream&, basic_json&) - instead; that is, replace calls like `j << i;` with `i >> j;`. - @since version 1.0.0; deprecated since version 3.0.0 - */ - JSON_HEDLEY_DEPRECATED_FOR(3.0.0, operator>>(std::istream&, basic_json&)) - friend std::istream& operator<<(basic_json& j, std::istream& i) - { - return operator>>(i, j); - } - - /*! - @brief deserialize from stream - - Deserializes an input stream to a JSON value. - - @param[in,out] i input stream to read a serialized JSON value from - @param[in,out] j JSON value to write the deserialized input to - - @throw parse_error.101 in case of an unexpected token - @throw parse_error.102 if to_unicode fails or surrogate error - @throw parse_error.103 if to_unicode fails - - @complexity Linear in the length of the input. The parser is a predictive - LL(1) parser. - - @note A UTF-8 byte order mark is silently ignored. - - @liveexample{The example below shows how a JSON value is constructed by - reading a serialization from a stream.,operator_deserialize} - - @sa parse(std::istream&, const parser_callback_t) for a variant with a - parser callback function to filter values while parsing - - @since version 1.0.0 - */ - friend std::istream& operator>>(std::istream& i, basic_json& j) - { - parser(detail::input_adapter(i)).parse(false, j); - return i; - } - - /// @} - - /////////////////////////// - // convenience functions // - /////////////////////////// - - /*! - @brief return the type as string - - Returns the type name as string to be used in error messages - usually to - indicate that a function was called on a wrong JSON type. - - @return a string representation of a the @a m_type member: - Value type | return value - ----------- | ------------- - null | `"null"` - boolean | `"boolean"` - string | `"string"` - number | `"number"` (for all number types) - object | `"object"` - array | `"array"` - binary | `"binary"` - discarded | `"discarded"` - - @exceptionsafety No-throw guarantee: this function never throws exceptions. - - @complexity Constant. - - @liveexample{The following code exemplifies `type_name()` for all JSON - types.,type_name} - - @sa @ref type() -- return the type of the JSON value - @sa @ref operator value_t() -- return the type of the JSON value (implicit) - - @since version 1.0.0, public since 2.1.0, `const char*` and `noexcept` - since 3.0.0 - */ - JSON_HEDLEY_RETURNS_NON_NULL - const char* type_name() const noexcept - { - { - switch (m_type) - { - case value_t::null: - return "null"; - case value_t::object: - return "object"; - case value_t::array: - return "array"; - case value_t::string: - return "string"; - case value_t::boolean: - return "boolean"; - case value_t::binary: - return "binary"; - case value_t::discarded: - return "discarded"; - default: - return "number"; - } - } - } - - - private: - ////////////////////// - // member variables // - ////////////////////// - - /// the type of the current element - value_t m_type = value_t::null; - - /// the value of the current element - json_value m_value = {}; - - ////////////////////////////////////////// - // binary serialization/deserialization // - ////////////////////////////////////////// - - /// @name binary serialization/deserialization support - /// @{ - - public: - /*! - @brief create a CBOR serialization of a given JSON value - - Serializes a given JSON value @a j to a byte vector using the CBOR (Concise - Binary Object Representation) serialization format. CBOR is a binary - serialization format which aims to be more compact than JSON itself, yet - more efficient to parse. - - The library uses the following mapping from JSON values types to - CBOR types according to the CBOR specification (RFC 7049): - - JSON value type | value/range | CBOR type | first byte - --------------- | ------------------------------------------ | ---------------------------------- | --------------- - null | `null` | Null | 0xF6 - boolean | `true` | True | 0xF5 - boolean | `false` | False | 0xF4 - number_integer | -9223372036854775808..-2147483649 | Negative integer (8 bytes follow) | 0x3B - number_integer | -2147483648..-32769 | Negative integer (4 bytes follow) | 0x3A - number_integer | -32768..-129 | Negative integer (2 bytes follow) | 0x39 - number_integer | -128..-25 | Negative integer (1 byte follow) | 0x38 - number_integer | -24..-1 | Negative integer | 0x20..0x37 - number_integer | 0..23 | Integer | 0x00..0x17 - number_integer | 24..255 | Unsigned integer (1 byte follow) | 0x18 - number_integer | 256..65535 | Unsigned integer (2 bytes follow) | 0x19 - number_integer | 65536..4294967295 | Unsigned integer (4 bytes follow) | 0x1A - number_integer | 4294967296..18446744073709551615 | Unsigned integer (8 bytes follow) | 0x1B - number_unsigned | 0..23 | Integer | 0x00..0x17 - number_unsigned | 24..255 | Unsigned integer (1 byte follow) | 0x18 - number_unsigned | 256..65535 | Unsigned integer (2 bytes follow) | 0x19 - number_unsigned | 65536..4294967295 | Unsigned integer (4 bytes follow) | 0x1A - number_unsigned | 4294967296..18446744073709551615 | Unsigned integer (8 bytes follow) | 0x1B - number_float | *any value representable by a float* | Single-Precision Float | 0xFA - number_float | *any value NOT representable by a float* | Double-Precision Float | 0xFB - string | *length*: 0..23 | UTF-8 string | 0x60..0x77 - string | *length*: 23..255 | UTF-8 string (1 byte follow) | 0x78 - string | *length*: 256..65535 | UTF-8 string (2 bytes follow) | 0x79 - string | *length*: 65536..4294967295 | UTF-8 string (4 bytes follow) | 0x7A - string | *length*: 4294967296..18446744073709551615 | UTF-8 string (8 bytes follow) | 0x7B - array | *size*: 0..23 | array | 0x80..0x97 - array | *size*: 23..255 | array (1 byte follow) | 0x98 - array | *size*: 256..65535 | array (2 bytes follow) | 0x99 - array | *size*: 65536..4294967295 | array (4 bytes follow) | 0x9A - array | *size*: 4294967296..18446744073709551615 | array (8 bytes follow) | 0x9B - object | *size*: 0..23 | map | 0xA0..0xB7 - object | *size*: 23..255 | map (1 byte follow) | 0xB8 - object | *size*: 256..65535 | map (2 bytes follow) | 0xB9 - object | *size*: 65536..4294967295 | map (4 bytes follow) | 0xBA - object | *size*: 4294967296..18446744073709551615 | map (8 bytes follow) | 0xBB - binary | *size*: 0..23 | byte string | 0x40..0x57 - binary | *size*: 23..255 | byte string (1 byte follow) | 0x58 - binary | *size*: 256..65535 | byte string (2 bytes follow) | 0x59 - binary | *size*: 65536..4294967295 | byte string (4 bytes follow) | 0x5A - binary | *size*: 4294967296..18446744073709551615 | byte string (8 bytes follow) | 0x5B - - @note The mapping is **complete** in the sense that any JSON value type - can be converted to a CBOR value. - - @note If NaN or Infinity are stored inside a JSON number, they are - serialized properly. This behavior differs from the @ref dump() - function which serializes NaN or Infinity to `null`. - - @note The following CBOR types are not used in the conversion: - - UTF-8 strings terminated by "break" (0x7F) - - arrays terminated by "break" (0x9F) - - maps terminated by "break" (0xBF) - - byte strings terminated by "break" (0x5F) - - date/time (0xC0..0xC1) - - bignum (0xC2..0xC3) - - decimal fraction (0xC4) - - bigfloat (0xC5) - - expected conversions (0xD5..0xD7) - - simple values (0xE0..0xF3, 0xF8) - - undefined (0xF7) - - half-precision floats (0xF9) - - break (0xFF) - - @param[in] j JSON value to serialize - @return CBOR serialization as byte vector - - @complexity Linear in the size of the JSON value @a j. - - @liveexample{The example shows the serialization of a JSON value to a byte - vector in CBOR format.,to_cbor} - - @sa http://cbor.io - @sa @ref from_cbor(detail::input_adapter&&, const bool, const bool, const cbor_tag_handler_t) for the - analogous deserialization - @sa @ref to_msgpack(const basic_json&) for the related MessagePack format - @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the - related UBJSON format - - @since version 2.0.9; compact representation of floating-point numbers - since version 3.8.0 - */ - static std::vector to_cbor(const basic_json& j) - { - std::vector result; - to_cbor(j, result); - return result; - } - - static void to_cbor(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_cbor(j); - } - - static void to_cbor(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_cbor(j); - } - - /*! - @brief create a MessagePack serialization of a given JSON value - - Serializes a given JSON value @a j to a byte vector using the MessagePack - serialization format. MessagePack is a binary serialization format which - aims to be more compact than JSON itself, yet more efficient to parse. - - The library uses the following mapping from JSON values types to - MessagePack types according to the MessagePack specification: - - JSON value type | value/range | MessagePack type | first byte - --------------- | --------------------------------- | ---------------- | ---------- - null | `null` | nil | 0xC0 - boolean | `true` | true | 0xC3 - boolean | `false` | false | 0xC2 - number_integer | -9223372036854775808..-2147483649 | int64 | 0xD3 - number_integer | -2147483648..-32769 | int32 | 0xD2 - number_integer | -32768..-129 | int16 | 0xD1 - number_integer | -128..-33 | int8 | 0xD0 - number_integer | -32..-1 | negative fixint | 0xE0..0xFF - number_integer | 0..127 | positive fixint | 0x00..0x7F - number_integer | 128..255 | uint 8 | 0xCC - number_integer | 256..65535 | uint 16 | 0xCD - number_integer | 65536..4294967295 | uint 32 | 0xCE - number_integer | 4294967296..18446744073709551615 | uint 64 | 0xCF - number_unsigned | 0..127 | positive fixint | 0x00..0x7F - number_unsigned | 128..255 | uint 8 | 0xCC - number_unsigned | 256..65535 | uint 16 | 0xCD - number_unsigned | 65536..4294967295 | uint 32 | 0xCE - number_unsigned | 4294967296..18446744073709551615 | uint 64 | 0xCF - number_float | *any value representable by a float* | float 32 | 0xCA - number_float | *any value NOT representable by a float* | float 64 | 0xCB - string | *length*: 0..31 | fixstr | 0xA0..0xBF - string | *length*: 32..255 | str 8 | 0xD9 - string | *length*: 256..65535 | str 16 | 0xDA - string | *length*: 65536..4294967295 | str 32 | 0xDB - array | *size*: 0..15 | fixarray | 0x90..0x9F - array | *size*: 16..65535 | array 16 | 0xDC - array | *size*: 65536..4294967295 | array 32 | 0xDD - object | *size*: 0..15 | fix map | 0x80..0x8F - object | *size*: 16..65535 | map 16 | 0xDE - object | *size*: 65536..4294967295 | map 32 | 0xDF - binary | *size*: 0..255 | bin 8 | 0xC4 - binary | *size*: 256..65535 | bin 16 | 0xC5 - binary | *size*: 65536..4294967295 | bin 32 | 0xC6 - - @note The mapping is **complete** in the sense that any JSON value type - can be converted to a MessagePack value. - - @note The following values can **not** be converted to a MessagePack value: - - strings with more than 4294967295 bytes - - byte strings with more than 4294967295 bytes - - arrays with more than 4294967295 elements - - objects with more than 4294967295 elements - - @note Any MessagePack output created @ref to_msgpack can be successfully - parsed by @ref from_msgpack. - - @note If NaN or Infinity are stored inside a JSON number, they are - serialized properly. This behavior differs from the @ref dump() - function which serializes NaN or Infinity to `null`. - - @param[in] j JSON value to serialize - @return MessagePack serialization as byte vector - - @complexity Linear in the size of the JSON value @a j. - - @liveexample{The example shows the serialization of a JSON value to a byte - vector in MessagePack format.,to_msgpack} - - @sa http://msgpack.org - @sa @ref from_msgpack for the analogous deserialization - @sa @ref to_cbor(const basic_json& for the related CBOR format - @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the - related UBJSON format - - @since version 2.0.9 - */ - static std::vector to_msgpack(const basic_json& j) - { - std::vector result; - to_msgpack(j, result); - return result; - } - - static void to_msgpack(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_msgpack(j); - } - - static void to_msgpack(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_msgpack(j); - } - - /*! - @brief create a UBJSON serialization of a given JSON value - - Serializes a given JSON value @a j to a byte vector using the UBJSON - (Universal Binary JSON) serialization format. UBJSON aims to be more compact - than JSON itself, yet more efficient to parse. - - The library uses the following mapping from JSON values types to - UBJSON types according to the UBJSON specification: - - JSON value type | value/range | UBJSON type | marker - --------------- | --------------------------------- | ----------- | ------ - null | `null` | null | `Z` - boolean | `true` | true | `T` - boolean | `false` | false | `F` - number_integer | -9223372036854775808..-2147483649 | int64 | `L` - number_integer | -2147483648..-32769 | int32 | `l` - number_integer | -32768..-129 | int16 | `I` - number_integer | -128..127 | int8 | `i` - number_integer | 128..255 | uint8 | `U` - number_integer | 256..32767 | int16 | `I` - number_integer | 32768..2147483647 | int32 | `l` - number_integer | 2147483648..9223372036854775807 | int64 | `L` - number_unsigned | 0..127 | int8 | `i` - number_unsigned | 128..255 | uint8 | `U` - number_unsigned | 256..32767 | int16 | `I` - number_unsigned | 32768..2147483647 | int32 | `l` - number_unsigned | 2147483648..9223372036854775807 | int64 | `L` - number_unsigned | 2147483649..18446744073709551615 | high-precision | `H` - number_float | *any value* | float64 | `D` - string | *with shortest length indicator* | string | `S` - array | *see notes on optimized format* | array | `[` - object | *see notes on optimized format* | map | `{` - - @note The mapping is **complete** in the sense that any JSON value type - can be converted to a UBJSON value. - - @note The following values can **not** be converted to a UBJSON value: - - strings with more than 9223372036854775807 bytes (theoretical) - - @note The following markers are not used in the conversion: - - `Z`: no-op values are not created. - - `C`: single-byte strings are serialized with `S` markers. - - @note Any UBJSON output created @ref to_ubjson can be successfully parsed - by @ref from_ubjson. - - @note If NaN or Infinity are stored inside a JSON number, they are - serialized properly. This behavior differs from the @ref dump() - function which serializes NaN or Infinity to `null`. - - @note The optimized formats for containers are supported: Parameter - @a use_size adds size information to the beginning of a container and - removes the closing marker. Parameter @a use_type further checks - whether all elements of a container have the same type and adds the - type marker to the beginning of the container. The @a use_type - parameter must only be used together with @a use_size = true. Note - that @a use_size = true alone may result in larger representations - - the benefit of this parameter is that the receiving side is - immediately informed on the number of elements of the container. - - @note If the JSON data contains the binary type, the value stored is a list - of integers, as suggested by the UBJSON documentation. In particular, - this means that serialization and the deserialization of a JSON - containing binary values into UBJSON and back will result in a - different JSON object. - - @param[in] j JSON value to serialize - @param[in] use_size whether to add size annotations to container types - @param[in] use_type whether to add type annotations to container types - (must be combined with @a use_size = true) - @return UBJSON serialization as byte vector - - @complexity Linear in the size of the JSON value @a j. - - @liveexample{The example shows the serialization of a JSON value to a byte - vector in UBJSON format.,to_ubjson} - - @sa http://ubjson.org - @sa @ref from_ubjson(detail::input_adapter&&, const bool, const bool) for the - analogous deserialization - @sa @ref to_cbor(const basic_json& for the related CBOR format - @sa @ref to_msgpack(const basic_json&) for the related MessagePack format - - @since version 3.1.0 - */ - static std::vector to_ubjson(const basic_json& j, - const bool use_size = false, - const bool use_type = false) - { - std::vector result; - to_ubjson(j, result, use_size, use_type); - return result; - } - - static void to_ubjson(const basic_json& j, detail::output_adapter o, - const bool use_size = false, const bool use_type = false) - { - binary_writer(o).write_ubjson(j, use_size, use_type); - } - - static void to_ubjson(const basic_json& j, detail::output_adapter o, - const bool use_size = false, const bool use_type = false) - { - binary_writer(o).write_ubjson(j, use_size, use_type); - } - - - /*! - @brief Serializes the given JSON object `j` to BSON and returns a vector - containing the corresponding BSON-representation. - - BSON (Binary JSON) is a binary format in which zero or more ordered key/value pairs are - stored as a single entity (a so-called document). - - The library uses the following mapping from JSON values types to BSON types: - - JSON value type | value/range | BSON type | marker - --------------- | --------------------------------- | ----------- | ------ - null | `null` | null | 0x0A - boolean | `true`, `false` | boolean | 0x08 - number_integer | -9223372036854775808..-2147483649 | int64 | 0x12 - number_integer | -2147483648..2147483647 | int32 | 0x10 - number_integer | 2147483648..9223372036854775807 | int64 | 0x12 - number_unsigned | 0..2147483647 | int32 | 0x10 - number_unsigned | 2147483648..9223372036854775807 | int64 | 0x12 - number_unsigned | 9223372036854775808..18446744073709551615| -- | -- - number_float | *any value* | double | 0x01 - string | *any value* | string | 0x02 - array | *any value* | document | 0x04 - object | *any value* | document | 0x03 - binary | *any value* | binary | 0x05 - - @warning The mapping is **incomplete**, since only JSON-objects (and things - contained therein) can be serialized to BSON. - Also, integers larger than 9223372036854775807 cannot be serialized to BSON, - and the keys may not contain U+0000, since they are serialized a - zero-terminated c-strings. - - @throw out_of_range.407 if `j.is_number_unsigned() && j.get() > 9223372036854775807` - @throw out_of_range.409 if a key in `j` contains a NULL (U+0000) - @throw type_error.317 if `!j.is_object()` - - @pre The input `j` is required to be an object: `j.is_object() == true`. - - @note Any BSON output created via @ref to_bson can be successfully parsed - by @ref from_bson. - - @param[in] j JSON value to serialize - @return BSON serialization as byte vector - - @complexity Linear in the size of the JSON value @a j. - - @liveexample{The example shows the serialization of a JSON value to a byte - vector in BSON format.,to_bson} - - @sa http://bsonspec.org/spec.html - @sa @ref from_bson(detail::input_adapter&&, const bool strict) for the - analogous deserialization - @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the - related UBJSON format - @sa @ref to_cbor(const basic_json&) for the related CBOR format - @sa @ref to_msgpack(const basic_json&) for the related MessagePack format - */ - static std::vector to_bson(const basic_json& j) - { - std::vector result; - to_bson(j, result); - return result; - } - - /*! - @brief Serializes the given JSON object `j` to BSON and forwards the - corresponding BSON-representation to the given output_adapter `o`. - @param j The JSON object to convert to BSON. - @param o The output adapter that receives the binary BSON representation. - @pre The input `j` shall be an object: `j.is_object() == true` - @sa @ref to_bson(const basic_json&) - */ - static void to_bson(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_bson(j); - } - - /*! - @copydoc to_bson(const basic_json&, detail::output_adapter) - */ - static void to_bson(const basic_json& j, detail::output_adapter o) - { - binary_writer(o).write_bson(j); - } - - - /*! - @brief create a JSON value from an input in CBOR format - - Deserializes a given input @a i to a JSON value using the CBOR (Concise - Binary Object Representation) serialization format. - - The library maps CBOR types to JSON value types as follows: - - CBOR type | JSON value type | first byte - ---------------------- | --------------- | ---------- - Integer | number_unsigned | 0x00..0x17 - Unsigned integer | number_unsigned | 0x18 - Unsigned integer | number_unsigned | 0x19 - Unsigned integer | number_unsigned | 0x1A - Unsigned integer | number_unsigned | 0x1B - Negative integer | number_integer | 0x20..0x37 - Negative integer | number_integer | 0x38 - Negative integer | number_integer | 0x39 - Negative integer | number_integer | 0x3A - Negative integer | number_integer | 0x3B - Byte string | binary | 0x40..0x57 - Byte string | binary | 0x58 - Byte string | binary | 0x59 - Byte string | binary | 0x5A - Byte string | binary | 0x5B - UTF-8 string | string | 0x60..0x77 - UTF-8 string | string | 0x78 - UTF-8 string | string | 0x79 - UTF-8 string | string | 0x7A - UTF-8 string | string | 0x7B - UTF-8 string | string | 0x7F - array | array | 0x80..0x97 - array | array | 0x98 - array | array | 0x99 - array | array | 0x9A - array | array | 0x9B - array | array | 0x9F - map | object | 0xA0..0xB7 - map | object | 0xB8 - map | object | 0xB9 - map | object | 0xBA - map | object | 0xBB - map | object | 0xBF - False | `false` | 0xF4 - True | `true` | 0xF5 - Null | `null` | 0xF6 - Half-Precision Float | number_float | 0xF9 - Single-Precision Float | number_float | 0xFA - Double-Precision Float | number_float | 0xFB - - @warning The mapping is **incomplete** in the sense that not all CBOR - types can be converted to a JSON value. The following CBOR types - are not supported and will yield parse errors (parse_error.112): - - date/time (0xC0..0xC1) - - bignum (0xC2..0xC3) - - decimal fraction (0xC4) - - bigfloat (0xC5) - - expected conversions (0xD5..0xD7) - - simple values (0xE0..0xF3, 0xF8) - - undefined (0xF7) - - @warning CBOR allows map keys of any type, whereas JSON only allows - strings as keys in object values. Therefore, CBOR maps with keys - other than UTF-8 strings are rejected (parse_error.113). - - @note Any CBOR output created @ref to_cbor can be successfully parsed by - @ref from_cbor. - - @param[in] i an input in CBOR format convertible to an input adapter - @param[in] strict whether to expect the input to be consumed until EOF - (true by default) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - @param[in] tag_handler how to treat CBOR tags (optional, error by default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.110 if the given input ends prematurely or the end of - file was not reached when @a strict was set to true - @throw parse_error.112 if unsupported features from CBOR were - used in the given input @a v or if the input is not valid CBOR - @throw parse_error.113 if a string was expected as map key, but not found - - @complexity Linear in the size of the input @a i. - - @liveexample{The example shows the deserialization of a byte vector in CBOR - format to a JSON value.,from_cbor} - - @sa http://cbor.io - @sa @ref to_cbor(const basic_json&) for the analogous serialization - @sa @ref from_msgpack(detail::input_adapter&&, const bool, const bool) for the - related MessagePack format - @sa @ref from_ubjson(detail::input_adapter&&, const bool, const bool) for the - related UBJSON format - - @since version 2.0.9; parameter @a start_index since 2.1.1; changed to - consume input adapters, removed start_index parameter, and added - @a strict parameter since 3.0.0; added @a allow_exceptions parameter - since 3.2.0; added @a tag_handler parameter since 3.9.0. - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_cbor(InputType&& i, - const bool strict = true, - const bool allow_exceptions = true, - const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::forward(i)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); - return res ? result : basic_json(value_t::discarded); - } - - /*! - @copydoc from_cbor(detail::input_adapter&&, const bool, const bool, const cbor_tag_handler_t) - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_cbor(IteratorType first, IteratorType last, - const bool strict = true, - const bool allow_exceptions = true, - const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::move(first), std::move(last)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); - return res ? result : basic_json(value_t::discarded); - } - - template - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_cbor(ptr, ptr + len)) - static basic_json from_cbor(const T* ptr, std::size_t len, - const bool strict = true, - const bool allow_exceptions = true, - const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error) - { - return from_cbor(ptr, ptr + len, strict, allow_exceptions, tag_handler); - } - - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_cbor(ptr, ptr + len)) - static basic_json from_cbor(detail::span_input_adapter&& i, - const bool strict = true, - const bool allow_exceptions = true, - const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = i.get(); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); - return res ? result : basic_json(value_t::discarded); - } - - /*! - @brief create a JSON value from an input in MessagePack format - - Deserializes a given input @a i to a JSON value using the MessagePack - serialization format. - - The library maps MessagePack types to JSON value types as follows: - - MessagePack type | JSON value type | first byte - ---------------- | --------------- | ---------- - positive fixint | number_unsigned | 0x00..0x7F - fixmap | object | 0x80..0x8F - fixarray | array | 0x90..0x9F - fixstr | string | 0xA0..0xBF - nil | `null` | 0xC0 - false | `false` | 0xC2 - true | `true` | 0xC3 - float 32 | number_float | 0xCA - float 64 | number_float | 0xCB - uint 8 | number_unsigned | 0xCC - uint 16 | number_unsigned | 0xCD - uint 32 | number_unsigned | 0xCE - uint 64 | number_unsigned | 0xCF - int 8 | number_integer | 0xD0 - int 16 | number_integer | 0xD1 - int 32 | number_integer | 0xD2 - int 64 | number_integer | 0xD3 - str 8 | string | 0xD9 - str 16 | string | 0xDA - str 32 | string | 0xDB - array 16 | array | 0xDC - array 32 | array | 0xDD - map 16 | object | 0xDE - map 32 | object | 0xDF - bin 8 | binary | 0xC4 - bin 16 | binary | 0xC5 - bin 32 | binary | 0xC6 - ext 8 | binary | 0xC7 - ext 16 | binary | 0xC8 - ext 32 | binary | 0xC9 - fixext 1 | binary | 0xD4 - fixext 2 | binary | 0xD5 - fixext 4 | binary | 0xD6 - fixext 8 | binary | 0xD7 - fixext 16 | binary | 0xD8 - negative fixint | number_integer | 0xE0-0xFF - - @note Any MessagePack output created @ref to_msgpack can be successfully - parsed by @ref from_msgpack. - - @param[in] i an input in MessagePack format convertible to an input - adapter - @param[in] strict whether to expect the input to be consumed until EOF - (true by default) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.110 if the given input ends prematurely or the end of - file was not reached when @a strict was set to true - @throw parse_error.112 if unsupported features from MessagePack were - used in the given input @a i or if the input is not valid MessagePack - @throw parse_error.113 if a string was expected as map key, but not found - - @complexity Linear in the size of the input @a i. - - @liveexample{The example shows the deserialization of a byte vector in - MessagePack format to a JSON value.,from_msgpack} - - @sa http://msgpack.org - @sa @ref to_msgpack(const basic_json&) for the analogous serialization - @sa @ref from_cbor(detail::input_adapter&&, const bool, const bool, const cbor_tag_handler_t) for the - related CBOR format - @sa @ref from_ubjson(detail::input_adapter&&, const bool, const bool) for - the related UBJSON format - @sa @ref from_bson(detail::input_adapter&&, const bool, const bool) for - the related BSON format - - @since version 2.0.9; parameter @a start_index since 2.1.1; changed to - consume input adapters, removed start_index parameter, and added - @a strict parameter since 3.0.0; added @a allow_exceptions parameter - since 3.2.0 - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_msgpack(InputType&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::forward(i)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - /*! - @copydoc from_msgpack(detail::input_adapter&&, const bool, const bool) - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_msgpack(IteratorType first, IteratorType last, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::move(first), std::move(last)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - - template - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_msgpack(ptr, ptr + len)) - static basic_json from_msgpack(const T* ptr, std::size_t len, - const bool strict = true, - const bool allow_exceptions = true) - { - return from_msgpack(ptr, ptr + len, strict, allow_exceptions); - } - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_msgpack(ptr, ptr + len)) - static basic_json from_msgpack(detail::span_input_adapter&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = i.get(); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - - /*! - @brief create a JSON value from an input in UBJSON format - - Deserializes a given input @a i to a JSON value using the UBJSON (Universal - Binary JSON) serialization format. - - The library maps UBJSON types to JSON value types as follows: - - UBJSON type | JSON value type | marker - ----------- | --------------------------------------- | ------ - no-op | *no value, next value is read* | `N` - null | `null` | `Z` - false | `false` | `F` - true | `true` | `T` - float32 | number_float | `d` - float64 | number_float | `D` - uint8 | number_unsigned | `U` - int8 | number_integer | `i` - int16 | number_integer | `I` - int32 | number_integer | `l` - int64 | number_integer | `L` - high-precision number | number_integer, number_unsigned, or number_float - depends on number string | 'H' - string | string | `S` - char | string | `C` - array | array (optimized values are supported) | `[` - object | object (optimized values are supported) | `{` - - @note The mapping is **complete** in the sense that any UBJSON value can - be converted to a JSON value. - - @param[in] i an input in UBJSON format convertible to an input adapter - @param[in] strict whether to expect the input to be consumed until EOF - (true by default) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.110 if the given input ends prematurely or the end of - file was not reached when @a strict was set to true - @throw parse_error.112 if a parse error occurs - @throw parse_error.113 if a string could not be parsed successfully - - @complexity Linear in the size of the input @a i. - - @liveexample{The example shows the deserialization of a byte vector in - UBJSON format to a JSON value.,from_ubjson} - - @sa http://ubjson.org - @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the - analogous serialization - @sa @ref from_cbor(detail::input_adapter&&, const bool, const bool, const cbor_tag_handler_t) for the - related CBOR format - @sa @ref from_msgpack(detail::input_adapter&&, const bool, const bool) for - the related MessagePack format - @sa @ref from_bson(detail::input_adapter&&, const bool, const bool) for - the related BSON format - - @since version 3.1.0; added @a allow_exceptions parameter since 3.2.0 - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_ubjson(InputType&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::forward(i)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - /*! - @copydoc from_ubjson(detail::input_adapter&&, const bool, const bool) - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_ubjson(IteratorType first, IteratorType last, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::move(first), std::move(last)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - template - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_ubjson(ptr, ptr + len)) - static basic_json from_ubjson(const T* ptr, std::size_t len, - const bool strict = true, - const bool allow_exceptions = true) - { - return from_ubjson(ptr, ptr + len, strict, allow_exceptions); - } - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_ubjson(ptr, ptr + len)) - static basic_json from_ubjson(detail::span_input_adapter&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = i.get(); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - - /*! - @brief Create a JSON value from an input in BSON format - - Deserializes a given input @a i to a JSON value using the BSON (Binary JSON) - serialization format. - - The library maps BSON record types to JSON value types as follows: - - BSON type | BSON marker byte | JSON value type - --------------- | ---------------- | --------------------------- - double | 0x01 | number_float - string | 0x02 | string - document | 0x03 | object - array | 0x04 | array - binary | 0x05 | still unsupported - undefined | 0x06 | still unsupported - ObjectId | 0x07 | still unsupported - boolean | 0x08 | boolean - UTC Date-Time | 0x09 | still unsupported - null | 0x0A | null - Regular Expr. | 0x0B | still unsupported - DB Pointer | 0x0C | still unsupported - JavaScript Code | 0x0D | still unsupported - Symbol | 0x0E | still unsupported - JavaScript Code | 0x0F | still unsupported - int32 | 0x10 | number_integer - Timestamp | 0x11 | still unsupported - 128-bit decimal float | 0x13 | still unsupported - Max Key | 0x7F | still unsupported - Min Key | 0xFF | still unsupported - - @warning The mapping is **incomplete**. The unsupported mappings - are indicated in the table above. - - @param[in] i an input in BSON format convertible to an input adapter - @param[in] strict whether to expect the input to be consumed until EOF - (true by default) - @param[in] allow_exceptions whether to throw exceptions in case of a - parse error (optional, true by default) - - @return deserialized JSON value; in case of a parse error and - @a allow_exceptions set to `false`, the return value will be - value_t::discarded. - - @throw parse_error.114 if an unsupported BSON record type is encountered - - @complexity Linear in the size of the input @a i. - - @liveexample{The example shows the deserialization of a byte vector in - BSON format to a JSON value.,from_bson} - - @sa http://bsonspec.org/spec.html - @sa @ref to_bson(const basic_json&) for the analogous serialization - @sa @ref from_cbor(detail::input_adapter&&, const bool, const bool, const cbor_tag_handler_t) for the - related CBOR format - @sa @ref from_msgpack(detail::input_adapter&&, const bool, const bool) for - the related MessagePack format - @sa @ref from_ubjson(detail::input_adapter&&, const bool, const bool) for the - related UBJSON format - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_bson(InputType&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::forward(i)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - /*! - @copydoc from_bson(detail::input_adapter&&, const bool, const bool) - */ - template - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json from_bson(IteratorType first, IteratorType last, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = detail::input_adapter(std::move(first), std::move(last)); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - - template - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_bson(ptr, ptr + len)) - static basic_json from_bson(const T* ptr, std::size_t len, - const bool strict = true, - const bool allow_exceptions = true) - { - return from_bson(ptr, ptr + len, strict, allow_exceptions); - } - - JSON_HEDLEY_WARN_UNUSED_RESULT - JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_bson(ptr, ptr + len)) - static basic_json from_bson(detail::span_input_adapter&& i, - const bool strict = true, - const bool allow_exceptions = true) - { - basic_json result; - detail::json_sax_dom_parser sdp(result, allow_exceptions); - auto ia = i.get(); - const bool res = binary_reader(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict); - return res ? result : basic_json(value_t::discarded); - } - /// @} - - ////////////////////////// - // JSON Pointer support // - ////////////////////////// - - /// @name JSON Pointer functions - /// @{ - - /*! - @brief access specified element via JSON Pointer - - Uses a JSON pointer to retrieve a reference to the respective JSON value. - No bound checking is performed. Similar to @ref operator[](const typename - object_t::key_type&), `null` values are created in arrays and objects if - necessary. - - In particular: - - If the JSON pointer points to an object key that does not exist, it - is created an filled with a `null` value before a reference to it - is returned. - - If the JSON pointer points to an array index that does not exist, it - is created an filled with a `null` value before a reference to it - is returned. All indices between the current maximum and the given - index are also filled with `null`. - - The special value `-` is treated as a synonym for the index past the - end. - - @param[in] ptr a JSON pointer - - @return reference to the element pointed to by @a ptr - - @complexity Constant. - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.404 if the JSON pointer can not be resolved - - @liveexample{The behavior is shown in the example.,operatorjson_pointer} - - @since version 2.0.0 - */ - reference operator[](const json_pointer& ptr) - { - return ptr.get_unchecked(this); - } - - /*! - @brief access specified element via JSON Pointer - - Uses a JSON pointer to retrieve a reference to the respective JSON value. - No bound checking is performed. The function does not change the JSON - value; no `null` values are created. In particular, the special value - `-` yields an exception. - - @param[in] ptr JSON pointer to the desired element - - @return const reference to the element pointed to by @a ptr - - @complexity Constant. - - @throw parse_error.106 if an array index begins with '0' - @throw parse_error.109 if an array index was not a number - @throw out_of_range.402 if the array index '-' is used - @throw out_of_range.404 if the JSON pointer can not be resolved - - @liveexample{The behavior is shown in the example.,operatorjson_pointer_const} - - @since version 2.0.0 - */ - const_reference operator[](const json_pointer& ptr) const - { - return ptr.get_unchecked(this); - } - - /*! - @brief access specified element via JSON Pointer - - Returns a reference to the element at with specified JSON pointer @a ptr, - with bounds checking. - - @param[in] ptr JSON pointer to the desired element - - @return reference to the element pointed to by @a ptr - - @throw parse_error.106 if an array index in the passed JSON pointer @a ptr - begins with '0'. See example below. - - @throw parse_error.109 if an array index in the passed JSON pointer @a ptr - is not a number. See example below. - - @throw out_of_range.401 if an array index in the passed JSON pointer @a ptr - is out of range. See example below. - - @throw out_of_range.402 if the array index '-' is used in the passed JSON - pointer @a ptr. As `at` provides checked access (and no elements are - implicitly inserted), the index '-' is always invalid. See example below. - - @throw out_of_range.403 if the JSON pointer describes a key of an object - which cannot be found. See example below. - - @throw out_of_range.404 if the JSON pointer @a ptr can not be resolved. - See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @since version 2.0.0 - - @liveexample{The behavior is shown in the example.,at_json_pointer} - */ - reference at(const json_pointer& ptr) - { - return ptr.get_checked(this); - } - - /*! - @brief access specified element via JSON Pointer - - Returns a const reference to the element at with specified JSON pointer @a - ptr, with bounds checking. - - @param[in] ptr JSON pointer to the desired element - - @return reference to the element pointed to by @a ptr - - @throw parse_error.106 if an array index in the passed JSON pointer @a ptr - begins with '0'. See example below. - - @throw parse_error.109 if an array index in the passed JSON pointer @a ptr - is not a number. See example below. - - @throw out_of_range.401 if an array index in the passed JSON pointer @a ptr - is out of range. See example below. - - @throw out_of_range.402 if the array index '-' is used in the passed JSON - pointer @a ptr. As `at` provides checked access (and no elements are - implicitly inserted), the index '-' is always invalid. See example below. - - @throw out_of_range.403 if the JSON pointer describes a key of an object - which cannot be found. See example below. - - @throw out_of_range.404 if the JSON pointer @a ptr can not be resolved. - See example below. - - @exceptionsafety Strong guarantee: if an exception is thrown, there are no - changes in the JSON value. - - @complexity Constant. - - @since version 2.0.0 - - @liveexample{The behavior is shown in the example.,at_json_pointer_const} - */ - const_reference at(const json_pointer& ptr) const - { - return ptr.get_checked(this); - } - - /*! - @brief return flattened JSON value - - The function creates a JSON object whose keys are JSON pointers (see [RFC - 6901](https://tools.ietf.org/html/rfc6901)) and whose values are all - primitive. The original JSON value can be restored using the @ref - unflatten() function. - - @return an object that maps JSON pointers to primitive values - - @note Empty objects and arrays are flattened to `null` and will not be - reconstructed correctly by the @ref unflatten() function. - - @complexity Linear in the size the JSON value. - - @liveexample{The following code shows how a JSON object is flattened to an - object whose keys consist of JSON pointers.,flatten} - - @sa @ref unflatten() for the reverse function - - @since version 2.0.0 - */ - basic_json flatten() const - { - basic_json result(value_t::object); - json_pointer::flatten("", *this, result); - return result; - } - - /*! - @brief unflatten a previously flattened JSON value - - The function restores the arbitrary nesting of a JSON value that has been - flattened before using the @ref flatten() function. The JSON value must - meet certain constraints: - 1. The value must be an object. - 2. The keys must be JSON pointers (see - [RFC 6901](https://tools.ietf.org/html/rfc6901)) - 3. The mapped values must be primitive JSON types. - - @return the original JSON from a flattened version - - @note Empty objects and arrays are flattened by @ref flatten() to `null` - values and can not unflattened to their original type. Apart from - this example, for a JSON value `j`, the following is always true: - `j == j.flatten().unflatten()`. - - @complexity Linear in the size the JSON value. - - @throw type_error.314 if value is not an object - @throw type_error.315 if object values are not primitive - - @liveexample{The following code shows how a flattened JSON object is - unflattened into the original nested JSON object.,unflatten} - - @sa @ref flatten() for the reverse function - - @since version 2.0.0 - */ - basic_json unflatten() const - { - return json_pointer::unflatten(*this); - } - - /// @} - - ////////////////////////// - // JSON Patch functions // - ////////////////////////// - - /// @name JSON Patch functions - /// @{ - - /*! - @brief applies a JSON patch - - [JSON Patch](http://jsonpatch.com) defines a JSON document structure for - expressing a sequence of operations to apply to a JSON) document. With - this function, a JSON Patch is applied to the current JSON value by - executing all operations from the patch. - - @param[in] json_patch JSON patch document - @return patched document - - @note The application of a patch is atomic: Either all operations succeed - and the patched document is returned or an exception is thrown. In - any case, the original value is not changed: the patch is applied - to a copy of the value. - - @throw parse_error.104 if the JSON patch does not consist of an array of - objects - - @throw parse_error.105 if the JSON patch is malformed (e.g., mandatory - attributes are missing); example: `"operation add must have member path"` - - @throw out_of_range.401 if an array index is out of range. - - @throw out_of_range.403 if a JSON pointer inside the patch could not be - resolved successfully in the current JSON value; example: `"key baz not - found"` - - @throw out_of_range.405 if JSON pointer has no parent ("add", "remove", - "move") - - @throw other_error.501 if "test" operation was unsuccessful - - @complexity Linear in the size of the JSON value and the length of the - JSON patch. As usually only a fraction of the JSON value is affected by - the patch, the complexity can usually be neglected. - - @liveexample{The following code shows how a JSON patch is applied to a - value.,patch} - - @sa @ref diff -- create a JSON patch by comparing two JSON values - - @sa [RFC 6902 (JSON Patch)](https://tools.ietf.org/html/rfc6902) - @sa [RFC 6901 (JSON Pointer)](https://tools.ietf.org/html/rfc6901) - - @since version 2.0.0 - */ - basic_json patch(const basic_json& json_patch) const - { - // make a working copy to apply the patch to - basic_json result = *this; - - // the valid JSON Patch operations - enum class patch_operations {add, remove, replace, move, copy, test, invalid}; - - const auto get_op = [](const std::string & op) - { - if (op == "add") - { - return patch_operations::add; - } - if (op == "remove") - { - return patch_operations::remove; - } - if (op == "replace") - { - return patch_operations::replace; - } - if (op == "move") - { - return patch_operations::move; - } - if (op == "copy") - { - return patch_operations::copy; - } - if (op == "test") - { - return patch_operations::test; - } - - return patch_operations::invalid; - }; - - // wrapper for "add" operation; add value at ptr - const auto operation_add = [&result](json_pointer & ptr, basic_json val) - { - // adding to the root of the target document means replacing it - if (ptr.empty()) - { - result = val; - return; - } - - // make sure the top element of the pointer exists - json_pointer top_pointer = ptr.top(); - if (top_pointer != ptr) - { - result.at(top_pointer); - } - - // get reference to parent of JSON pointer ptr - const auto last_path = ptr.back(); - ptr.pop_back(); - basic_json& parent = result[ptr]; - - switch (parent.m_type) - { - case value_t::null: - case value_t::object: - { - // use operator[] to add value - parent[last_path] = val; - break; - } - - case value_t::array: - { - if (last_path == "-") - { - // special case: append to back - parent.push_back(val); - } - else - { - const auto idx = json_pointer::array_index(last_path); - if (JSON_HEDLEY_UNLIKELY(idx > parent.size())) - { - // avoid undefined behavior - JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range")); - } - - // default case: insert add offset - parent.insert(parent.begin() + static_cast(idx), val); - } - break; - } - - // if there exists a parent it cannot be primitive - default: // LCOV_EXCL_LINE - JSON_ASSERT(false); // LCOV_EXCL_LINE - } - }; - - // wrapper for "remove" operation; remove value at ptr - const auto operation_remove = [&result](json_pointer & ptr) - { - // get reference to parent of JSON pointer ptr - const auto last_path = ptr.back(); - ptr.pop_back(); - basic_json& parent = result.at(ptr); - - // remove child - if (parent.is_object()) - { - // perform range check - auto it = parent.find(last_path); - if (JSON_HEDLEY_LIKELY(it != parent.end())) - { - parent.erase(it); - } - else - { - JSON_THROW(out_of_range::create(403, "key '" + last_path + "' not found")); - } - } - else if (parent.is_array()) - { - // note erase performs range check - parent.erase(json_pointer::array_index(last_path)); - } - }; - - // type check: top level value must be an array - if (JSON_HEDLEY_UNLIKELY(!json_patch.is_array())) - { - JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects")); - } - - // iterate and apply the operations - for (const auto& val : json_patch) - { - // wrapper to get a value for an operation - const auto get_value = [&val](const std::string & op, - const std::string & member, - bool string_type) -> basic_json & - { - // find value - auto it = val.m_value.object->find(member); - - // context-sensitive error message - const auto error_msg = (op == "op") ? "operation" : "operation '" + op + "'"; - - // check if desired value is present - if (JSON_HEDLEY_UNLIKELY(it == val.m_value.object->end())) - { - JSON_THROW(parse_error::create(105, 0, error_msg + " must have member '" + member + "'")); - } - - // check if result is of type string - if (JSON_HEDLEY_UNLIKELY(string_type && !it->second.is_string())) - { - JSON_THROW(parse_error::create(105, 0, error_msg + " must have string member '" + member + "'")); - } - - // no error: return value - return it->second; - }; - - // type check: every element of the array must be an object - if (JSON_HEDLEY_UNLIKELY(!val.is_object())) - { - JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects")); - } - - // collect mandatory members - const auto op = get_value("op", "op", true).template get(); - const auto path = get_value(op, "path", true).template get(); - json_pointer ptr(path); - - switch (get_op(op)) - { - case patch_operations::add: - { - operation_add(ptr, get_value("add", "value", false)); - break; - } - - case patch_operations::remove: - { - operation_remove(ptr); - break; - } - - case patch_operations::replace: - { - // the "path" location must exist - use at() - result.at(ptr) = get_value("replace", "value", false); - break; - } - - case patch_operations::move: - { - const auto from_path = get_value("move", "from", true).template get(); - json_pointer from_ptr(from_path); - - // the "from" location must exist - use at() - basic_json v = result.at(from_ptr); - - // The move operation is functionally identical to a - // "remove" operation on the "from" location, followed - // immediately by an "add" operation at the target - // location with the value that was just removed. - operation_remove(from_ptr); - operation_add(ptr, v); - break; - } - - case patch_operations::copy: - { - const auto from_path = get_value("copy", "from", true).template get(); - const json_pointer from_ptr(from_path); - - // the "from" location must exist - use at() - basic_json v = result.at(from_ptr); - - // The copy is functionally identical to an "add" - // operation at the target location using the value - // specified in the "from" member. - operation_add(ptr, v); - break; - } - - case patch_operations::test: - { - bool success = false; - JSON_TRY - { - // check if "value" matches the one at "path" - // the "path" location must exist - use at() - success = (result.at(ptr) == get_value("test", "value", false)); - } - JSON_INTERNAL_CATCH (out_of_range&) - { - // ignore out of range errors: success remains false - } - - // throw an exception if test fails - if (JSON_HEDLEY_UNLIKELY(!success)) - { - JSON_THROW(other_error::create(501, "unsuccessful: " + val.dump())); - } - - break; - } - - default: - { - // op must be "add", "remove", "replace", "move", "copy", or - // "test" - JSON_THROW(parse_error::create(105, 0, "operation value '" + op + "' is invalid")); - } - } - } - - return result; - } - - /*! - @brief creates a diff as a JSON patch - - Creates a [JSON Patch](http://jsonpatch.com) so that value @a source can - be changed into the value @a target by calling @ref patch function. - - @invariant For two JSON values @a source and @a target, the following code - yields always `true`: - @code {.cpp} - source.patch(diff(source, target)) == target; - @endcode - - @note Currently, only `remove`, `add`, and `replace` operations are - generated. - - @param[in] source JSON value to compare from - @param[in] target JSON value to compare against - @param[in] path helper value to create JSON pointers - - @return a JSON patch to convert the @a source to @a target - - @complexity Linear in the lengths of @a source and @a target. - - @liveexample{The following code shows how a JSON patch is created as a - diff for two JSON values.,diff} - - @sa @ref patch -- apply a JSON patch - @sa @ref merge_patch -- apply a JSON Merge Patch - - @sa [RFC 6902 (JSON Patch)](https://tools.ietf.org/html/rfc6902) - - @since version 2.0.0 - */ - JSON_HEDLEY_WARN_UNUSED_RESULT - static basic_json diff(const basic_json& source, const basic_json& target, - const std::string& path = "") - { - // the patch - basic_json result(value_t::array); - - // if the values are the same, return empty patch - if (source == target) - { - return result; - } - - if (source.type() != target.type()) - { - // different types: replace value - result.push_back( - { - {"op", "replace"}, {"path", path}, {"value", target} - }); - return result; - } - - switch (source.type()) - { - case value_t::array: - { - // first pass: traverse common elements - std::size_t i = 0; - while (i < source.size() && i < target.size()) - { - // recursive call to compare array values at index i - auto temp_diff = diff(source[i], target[i], path + "/" + std::to_string(i)); - result.insert(result.end(), temp_diff.begin(), temp_diff.end()); - ++i; - } - - // i now reached the end of at least one array - // in a second pass, traverse the remaining elements - - // remove my remaining elements - const auto end_index = static_cast(result.size()); - while (i < source.size()) - { - // add operations in reverse order to avoid invalid - // indices - result.insert(result.begin() + end_index, object( - { - {"op", "remove"}, - {"path", path + "/" + std::to_string(i)} - })); - ++i; - } - - // add other remaining elements - while (i < target.size()) - { - result.push_back( - { - {"op", "add"}, - {"path", path + "/-"}, - {"value", target[i]} - }); - ++i; - } - - break; - } - - case value_t::object: - { - // first pass: traverse this object's elements - for (auto it = source.cbegin(); it != source.cend(); ++it) - { - // escape the key name to be used in a JSON patch - const auto key = json_pointer::escape(it.key()); - - if (target.find(it.key()) != target.end()) - { - // recursive call to compare object values at key it - auto temp_diff = diff(it.value(), target[it.key()], path + "/" + key); - result.insert(result.end(), temp_diff.begin(), temp_diff.end()); - } - else - { - // found a key that is not in o -> remove it - result.push_back(object( - { - {"op", "remove"}, {"path", path + "/" + key} - })); - } - } - - // second pass: traverse other object's elements - for (auto it = target.cbegin(); it != target.cend(); ++it) - { - if (source.find(it.key()) == source.end()) - { - // found a key that is not in this -> add it - const auto key = json_pointer::escape(it.key()); - result.push_back( - { - {"op", "add"}, {"path", path + "/" + key}, - {"value", it.value()} - }); - } - } - - break; - } - - default: - { - // both primitive type: replace value - result.push_back( - { - {"op", "replace"}, {"path", path}, {"value", target} - }); - break; - } - } - - return result; - } - - /// @} - - //////////////////////////////// - // JSON Merge Patch functions // - //////////////////////////////// - - /// @name JSON Merge Patch functions - /// @{ - - /*! - @brief applies a JSON Merge Patch - - The merge patch format is primarily intended for use with the HTTP PATCH - method as a means of describing a set of modifications to a target - resource's content. This function applies a merge patch to the current - JSON value. - - The function implements the following algorithm from Section 2 of - [RFC 7396 (JSON Merge Patch)](https://tools.ietf.org/html/rfc7396): - - ``` - define MergePatch(Target, Patch): - if Patch is an Object: - if Target is not an Object: - Target = {} // Ignore the contents and set it to an empty Object - for each Name/Value pair in Patch: - if Value is null: - if Name exists in Target: - remove the Name/Value pair from Target - else: - Target[Name] = MergePatch(Target[Name], Value) - return Target - else: - return Patch - ``` - - Thereby, `Target` is the current object; that is, the patch is applied to - the current value. - - @param[in] apply_patch the patch to apply - - @complexity Linear in the lengths of @a patch. - - @liveexample{The following code shows how a JSON Merge Patch is applied to - a JSON document.,merge_patch} - - @sa @ref patch -- apply a JSON patch - @sa [RFC 7396 (JSON Merge Patch)](https://tools.ietf.org/html/rfc7396) - - @since version 3.0.0 - */ - void merge_patch(const basic_json& apply_patch) - { - if (apply_patch.is_object()) - { - if (!is_object()) - { - *this = object(); - } - for (auto it = apply_patch.begin(); it != apply_patch.end(); ++it) - { - if (it.value().is_null()) - { - erase(it.key()); - } - else - { - operator[](it.key()).merge_patch(it.value()); - } - } - } - else - { - *this = apply_patch; - } - } - - /// @} -}; - -/*! -@brief user-defined to_string function for JSON values - -This function implements a user-defined to_string for JSON objects. - -@param[in] j a JSON object -@return a std::string object -*/ - -NLOHMANN_BASIC_JSON_TPL_DECLARATION -std::string to_string(const NLOHMANN_BASIC_JSON_TPL& j) -{ - return j.dump(); -} -} // namespace nlohmann - -/////////////////////// -// nonmember support // -/////////////////////// - -// specialization of std::swap, and std::hash -namespace std -{ - -/// hash value for JSON objects -template<> -struct hash -{ - /*! - @brief return a hash value for a JSON object - - @since version 1.0.0 - */ - std::size_t operator()(const nlohmann::json& j) const - { - return nlohmann::detail::hash(j); - } -}; - -/// specialization for std::less -/// @note: do not remove the space after '<', -/// see https://github.com/nlohmann/json/pull/679 -template<> -struct less<::nlohmann::detail::value_t> -{ - /*! - @brief compare two value_t enum values - @since version 3.0.0 - */ - bool operator()(nlohmann::detail::value_t lhs, - nlohmann::detail::value_t rhs) const noexcept - { - return nlohmann::detail::operator<(lhs, rhs); - } -}; - -// C++20 prohibit function specialization in the std namespace. -#ifndef JSON_HAS_CPP_20 - -/*! -@brief exchanges the values of two JSON objects - -@since version 1.0.0 -*/ -template<> -inline void swap(nlohmann::json& j1, nlohmann::json& j2) noexcept( - is_nothrow_move_constructible::value&& - is_nothrow_move_assignable::value -) -{ - j1.swap(j2); -} - -#endif - -} // namespace std - -/*! -@brief user-defined string literal for JSON values - -This operator implements a user-defined string literal for JSON objects. It -can be used by adding `"_json"` to a string literal and returns a JSON object -if no parse error occurred. - -@param[in] s a string representation of a JSON object -@param[in] n the length of string @a s -@return a JSON object - -@since version 1.0.0 -*/ -JSON_HEDLEY_NON_NULL(1) -inline nlohmann::json operator "" _json(const char* s, std::size_t n) -{ - return nlohmann::json::parse(s, s + n); -} - -/*! -@brief user-defined string literal for JSON pointer - -This operator implements a user-defined string literal for JSON Pointers. It -can be used by adding `"_json_pointer"` to a string literal and returns a JSON pointer -object if no parse error occurred. - -@param[in] s a string representation of a JSON Pointer -@param[in] n the length of string @a s -@return a JSON pointer object - -@since version 2.0.0 -*/ -JSON_HEDLEY_NON_NULL(1) -inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std::size_t n) -{ - return nlohmann::json::json_pointer(std::string(s, n)); -} - -#include - -#endif // INCLUDE_NLOHMANN_JSON_HPP_ diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json_fwd.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json_fwd.hpp deleted file mode 100644 index 332227c1ba2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/json_fwd.hpp +++ /dev/null @@ -1,78 +0,0 @@ -#ifndef INCLUDE_NLOHMANN_JSON_FWD_HPP_ -#define INCLUDE_NLOHMANN_JSON_FWD_HPP_ - -#include // int64_t, uint64_t -#include // map -#include // allocator -#include // string -#include // vector - -/*! -@brief namespace for Niels Lohmann -@see https://github.com/nlohmann -@since version 1.0.0 -*/ -namespace nlohmann -{ -/*! -@brief default JSONSerializer template argument - -This serializer ignores the template arguments and uses ADL -([argument-dependent lookup](https://en.cppreference.com/w/cpp/language/adl)) -for serialization. -*/ -template -struct adl_serializer; - -template class ObjectType = - std::map, - template class ArrayType = std::vector, - class StringType = std::string, class BooleanType = bool, - class NumberIntegerType = std::int64_t, - class NumberUnsignedType = std::uint64_t, - class NumberFloatType = double, - template class AllocatorType = std::allocator, - template class JSONSerializer = - adl_serializer, - class BinaryType = std::vector> -class basic_json; - -/*! -@brief JSON Pointer - -A JSON pointer defines a string syntax for identifying a specific value -within a JSON document. It can be used with functions `at` and -`operator[]`. Furthermore, JSON pointers are the base for JSON patches. - -@sa [RFC 6901](https://tools.ietf.org/html/rfc6901) - -@since version 2.0.0 -*/ -template -class json_pointer; - -/*! -@brief default JSON class - -This type is the default specialization of the @ref basic_json class which -uses the standard template types. - -@since version 1.0.0 -*/ -using json = basic_json<>; - -template -struct ordered_map; - -/*! -@brief ordered JSON class - -This type preserves the insertion order of object keys. - -@since version 3.9.0 -*/ -using ordered_json = basic_json; - -} // namespace nlohmann - -#endif // INCLUDE_NLOHMANN_JSON_FWD_HPP_ diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/ordered_map.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/ordered_map.hpp deleted file mode 100644 index 57679091523..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/ordered_map.hpp +++ /dev/null @@ -1,171 +0,0 @@ -#pragma once - -#include // less -#include // allocator -#include // pair -#include // vector - -namespace nlohmann -{ - -/// ordered_map: a minimal map-like container that preserves insertion order -/// for use within nlohmann::basic_json -template , - class Allocator = std::allocator>> - struct ordered_map : std::vector, Allocator> -{ - using key_type = Key; - using mapped_type = T; - using Container = std::vector, Allocator>; - using typename Container::iterator; - using typename Container::const_iterator; - using typename Container::size_type; - using typename Container::value_type; - - // Explicit constructors instead of `using Container::Container` - // otherwise older compilers choke on it (GCC <= 5.5, xcode <= 9.4) - ordered_map(const Allocator& alloc = Allocator()) : Container{alloc} {} - template - ordered_map(It first, It last, const Allocator& alloc = Allocator()) - : Container{first, last, alloc} {} - ordered_map(std::initializer_list init, const Allocator& alloc = Allocator() ) - : Container{init, alloc} {} - - std::pair emplace(const key_type& key, T&& t) - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return {it, false}; - } - } - Container::emplace_back(key, t); - return {--this->end(), true}; - } - - T& operator[](const Key& key) - { - return emplace(key, T{}).first->second; - } - - const T& operator[](const Key& key) const - { - return at(key); - } - - T& at(const Key& key) - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return it->second; - } - } - - throw std::out_of_range("key not found"); - } - - const T& at(const Key& key) const - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return it->second; - } - } - - throw std::out_of_range("key not found"); - } - - size_type erase(const Key& key) - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - // Since we cannot move const Keys, re-construct them in place - for (auto next = it; ++next != this->end(); ++it) - { - it->~value_type(); // Destroy but keep allocation - new (&*it) value_type{std::move(*next)}; - } - Container::pop_back(); - return 1; - } - } - return 0; - } - - iterator erase(iterator pos) - { - auto it = pos; - - // Since we cannot move const Keys, re-construct them in place - for (auto next = it; ++next != this->end(); ++it) - { - it->~value_type(); // Destroy but keep allocation - new (&*it) value_type{std::move(*next)}; - } - Container::pop_back(); - return pos; - } - - size_type count(const Key& key) const - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return 1; - } - } - return 0; - } - - iterator find(const Key& key) - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return it; - } - } - return Container::end(); - } - - const_iterator find(const Key& key) const - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == key) - { - return it; - } - } - return Container::end(); - } - - std::pair insert( value_type&& value ) - { - return emplace(value.first, std::move(value.second)); - } - - std::pair insert( const value_type& value ) - { - for (auto it = this->begin(); it != this->end(); ++it) - { - if (it->first == value.first) - { - return {it, false}; - } - } - Container::push_back(value); - return {--this->end(), true}; - } -}; - -} // namespace nlohmann diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley.hpp deleted file mode 100644 index 521c78f1a4c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley.hpp +++ /dev/null @@ -1,1911 +0,0 @@ -/* Hedley - https://nemequ.github.io/hedley - * Created by Evan Nemerson - * - * To the extent possible under law, the author(s) have dedicated all - * copyright and related and neighboring rights to this software to - * the public domain worldwide. This software is distributed without - * any warranty. - * - * For details, see . - * SPDX-License-Identifier: CC0-1.0 - */ - -#if !defined(JSON_HEDLEY_VERSION) || (JSON_HEDLEY_VERSION < 13) -#if defined(JSON_HEDLEY_VERSION) - #undef JSON_HEDLEY_VERSION -#endif -#define JSON_HEDLEY_VERSION 13 - -#if defined(JSON_HEDLEY_STRINGIFY_EX) - #undef JSON_HEDLEY_STRINGIFY_EX -#endif -#define JSON_HEDLEY_STRINGIFY_EX(x) #x - -#if defined(JSON_HEDLEY_STRINGIFY) - #undef JSON_HEDLEY_STRINGIFY -#endif -#define JSON_HEDLEY_STRINGIFY(x) JSON_HEDLEY_STRINGIFY_EX(x) - -#if defined(JSON_HEDLEY_CONCAT_EX) - #undef JSON_HEDLEY_CONCAT_EX -#endif -#define JSON_HEDLEY_CONCAT_EX(a,b) a##b - -#if defined(JSON_HEDLEY_CONCAT) - #undef JSON_HEDLEY_CONCAT -#endif -#define JSON_HEDLEY_CONCAT(a,b) JSON_HEDLEY_CONCAT_EX(a,b) - -#if defined(JSON_HEDLEY_CONCAT3_EX) - #undef JSON_HEDLEY_CONCAT3_EX -#endif -#define JSON_HEDLEY_CONCAT3_EX(a,b,c) a##b##c - -#if defined(JSON_HEDLEY_CONCAT3) - #undef JSON_HEDLEY_CONCAT3 -#endif -#define JSON_HEDLEY_CONCAT3(a,b,c) JSON_HEDLEY_CONCAT3_EX(a,b,c) - -#if defined(JSON_HEDLEY_VERSION_ENCODE) - #undef JSON_HEDLEY_VERSION_ENCODE -#endif -#define JSON_HEDLEY_VERSION_ENCODE(major,minor,revision) (((major) * 1000000) + ((minor) * 1000) + (revision)) - -#if defined(JSON_HEDLEY_VERSION_DECODE_MAJOR) - #undef JSON_HEDLEY_VERSION_DECODE_MAJOR -#endif -#define JSON_HEDLEY_VERSION_DECODE_MAJOR(version) ((version) / 1000000) - -#if defined(JSON_HEDLEY_VERSION_DECODE_MINOR) - #undef JSON_HEDLEY_VERSION_DECODE_MINOR -#endif -#define JSON_HEDLEY_VERSION_DECODE_MINOR(version) (((version) % 1000000) / 1000) - -#if defined(JSON_HEDLEY_VERSION_DECODE_REVISION) - #undef JSON_HEDLEY_VERSION_DECODE_REVISION -#endif -#define JSON_HEDLEY_VERSION_DECODE_REVISION(version) ((version) % 1000) - -#if defined(JSON_HEDLEY_GNUC_VERSION) - #undef JSON_HEDLEY_GNUC_VERSION -#endif -#if defined(__GNUC__) && defined(__GNUC_PATCHLEVEL__) - #define JSON_HEDLEY_GNUC_VERSION JSON_HEDLEY_VERSION_ENCODE(__GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__) -#elif defined(__GNUC__) - #define JSON_HEDLEY_GNUC_VERSION JSON_HEDLEY_VERSION_ENCODE(__GNUC__, __GNUC_MINOR__, 0) -#endif - -#if defined(JSON_HEDLEY_GNUC_VERSION_CHECK) - #undef JSON_HEDLEY_GNUC_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_GNUC_VERSION) - #define JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_GNUC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_MSVC_VERSION) - #undef JSON_HEDLEY_MSVC_VERSION -#endif -#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 140000000) - #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 10000000, (_MSC_FULL_VER % 10000000) / 100000, (_MSC_FULL_VER % 100000) / 100) -#elif defined(_MSC_FULL_VER) - #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 1000000, (_MSC_FULL_VER % 1000000) / 10000, (_MSC_FULL_VER % 10000) / 10) -#elif defined(_MSC_VER) - #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_VER / 100, _MSC_VER % 100, 0) -#endif - -#if defined(JSON_HEDLEY_MSVC_VERSION_CHECK) - #undef JSON_HEDLEY_MSVC_VERSION_CHECK -#endif -#if !defined(_MSC_VER) - #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (0) -#elif defined(_MSC_VER) && (_MSC_VER >= 1400) - #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_FULL_VER >= ((major * 10000000) + (minor * 100000) + (patch))) -#elif defined(_MSC_VER) && (_MSC_VER >= 1200) - #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_FULL_VER >= ((major * 1000000) + (minor * 10000) + (patch))) -#else - #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_VER >= ((major * 100) + (minor))) -#endif - -#if defined(JSON_HEDLEY_INTEL_VERSION) - #undef JSON_HEDLEY_INTEL_VERSION -#endif -#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE) - #define JSON_HEDLEY_INTEL_VERSION JSON_HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, __INTEL_COMPILER_UPDATE) -#elif defined(__INTEL_COMPILER) - #define JSON_HEDLEY_INTEL_VERSION JSON_HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, 0) -#endif - -#if defined(JSON_HEDLEY_INTEL_VERSION_CHECK) - #undef JSON_HEDLEY_INTEL_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_INTEL_VERSION) - #define JSON_HEDLEY_INTEL_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_INTEL_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_INTEL_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_PGI_VERSION) - #undef JSON_HEDLEY_PGI_VERSION -#endif -#if defined(__PGI) && defined(__PGIC__) && defined(__PGIC_MINOR__) && defined(__PGIC_PATCHLEVEL__) - #define JSON_HEDLEY_PGI_VERSION JSON_HEDLEY_VERSION_ENCODE(__PGIC__, __PGIC_MINOR__, __PGIC_PATCHLEVEL__) -#endif - -#if defined(JSON_HEDLEY_PGI_VERSION_CHECK) - #undef JSON_HEDLEY_PGI_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_PGI_VERSION) - #define JSON_HEDLEY_PGI_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_PGI_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_PGI_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_SUNPRO_VERSION) - #undef JSON_HEDLEY_SUNPRO_VERSION -#endif -#if defined(__SUNPRO_C) && (__SUNPRO_C > 0x1000) - #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((((__SUNPRO_C >> 16) & 0xf) * 10) + ((__SUNPRO_C >> 12) & 0xf), (((__SUNPRO_C >> 8) & 0xf) * 10) + ((__SUNPRO_C >> 4) & 0xf), (__SUNPRO_C & 0xf) * 10) -#elif defined(__SUNPRO_C) - #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((__SUNPRO_C >> 8) & 0xf, (__SUNPRO_C >> 4) & 0xf, (__SUNPRO_C) & 0xf) -#elif defined(__SUNPRO_CC) && (__SUNPRO_CC > 0x1000) - #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((((__SUNPRO_CC >> 16) & 0xf) * 10) + ((__SUNPRO_CC >> 12) & 0xf), (((__SUNPRO_CC >> 8) & 0xf) * 10) + ((__SUNPRO_CC >> 4) & 0xf), (__SUNPRO_CC & 0xf) * 10) -#elif defined(__SUNPRO_CC) - #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((__SUNPRO_CC >> 8) & 0xf, (__SUNPRO_CC >> 4) & 0xf, (__SUNPRO_CC) & 0xf) -#endif - -#if defined(JSON_HEDLEY_SUNPRO_VERSION_CHECK) - #undef JSON_HEDLEY_SUNPRO_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_SUNPRO_VERSION) - #define JSON_HEDLEY_SUNPRO_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_SUNPRO_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_SUNPRO_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION) - #undef JSON_HEDLEY_EMSCRIPTEN_VERSION -#endif -#if defined(__EMSCRIPTEN__) - #define JSON_HEDLEY_EMSCRIPTEN_VERSION JSON_HEDLEY_VERSION_ENCODE(__EMSCRIPTEN_major__, __EMSCRIPTEN_minor__, __EMSCRIPTEN_tiny__) -#endif - -#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK) - #undef JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION) - #define JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_EMSCRIPTEN_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_ARM_VERSION) - #undef JSON_HEDLEY_ARM_VERSION -#endif -#if defined(__CC_ARM) && defined(__ARMCOMPILER_VERSION) - #define JSON_HEDLEY_ARM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ARMCOMPILER_VERSION / 1000000, (__ARMCOMPILER_VERSION % 1000000) / 10000, (__ARMCOMPILER_VERSION % 10000) / 100) -#elif defined(__CC_ARM) && defined(__ARMCC_VERSION) - #define JSON_HEDLEY_ARM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ARMCC_VERSION / 1000000, (__ARMCC_VERSION % 1000000) / 10000, (__ARMCC_VERSION % 10000) / 100) -#endif - -#if defined(JSON_HEDLEY_ARM_VERSION_CHECK) - #undef JSON_HEDLEY_ARM_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_ARM_VERSION) - #define JSON_HEDLEY_ARM_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_ARM_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_ARM_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_IBM_VERSION) - #undef JSON_HEDLEY_IBM_VERSION -#endif -#if defined(__ibmxl__) - #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ibmxl_version__, __ibmxl_release__, __ibmxl_modification__) -#elif defined(__xlC__) && defined(__xlC_ver__) - #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__xlC__ >> 8, __xlC__ & 0xff, (__xlC_ver__ >> 8) & 0xff) -#elif defined(__xlC__) - #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__xlC__ >> 8, __xlC__ & 0xff, 0) -#endif - -#if defined(JSON_HEDLEY_IBM_VERSION_CHECK) - #undef JSON_HEDLEY_IBM_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_IBM_VERSION) - #define JSON_HEDLEY_IBM_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_IBM_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_IBM_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_VERSION) - #undef JSON_HEDLEY_TI_VERSION -#endif -#if \ - defined(__TI_COMPILER_VERSION__) && \ - ( \ - defined(__TMS470__) || defined(__TI_ARM__) || \ - defined(__MSP430__) || \ - defined(__TMS320C2000__) \ - ) -#if (__TI_COMPILER_VERSION__ >= 16000000) - #define JSON_HEDLEY_TI_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif -#endif - -#if defined(JSON_HEDLEY_TI_VERSION_CHECK) - #undef JSON_HEDLEY_TI_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_VERSION) - #define JSON_HEDLEY_TI_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_CL2000_VERSION) - #undef JSON_HEDLEY_TI_CL2000_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && defined(__TMS320C2000__) - #define JSON_HEDLEY_TI_CL2000_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_CL2000_VERSION_CHECK) - #undef JSON_HEDLEY_TI_CL2000_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_CL2000_VERSION) - #define JSON_HEDLEY_TI_CL2000_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL2000_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_CL2000_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_CL430_VERSION) - #undef JSON_HEDLEY_TI_CL430_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && defined(__MSP430__) - #define JSON_HEDLEY_TI_CL430_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_CL430_VERSION_CHECK) - #undef JSON_HEDLEY_TI_CL430_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_CL430_VERSION) - #define JSON_HEDLEY_TI_CL430_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL430_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_CL430_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_ARMCL_VERSION) - #undef JSON_HEDLEY_TI_ARMCL_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && (defined(__TMS470__) || defined(__TI_ARM__)) - #define JSON_HEDLEY_TI_ARMCL_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_ARMCL_VERSION_CHECK) - #undef JSON_HEDLEY_TI_ARMCL_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_ARMCL_VERSION) - #define JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_ARMCL_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_CL6X_VERSION) - #undef JSON_HEDLEY_TI_CL6X_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && defined(__TMS320C6X__) - #define JSON_HEDLEY_TI_CL6X_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_CL6X_VERSION_CHECK) - #undef JSON_HEDLEY_TI_CL6X_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_CL6X_VERSION) - #define JSON_HEDLEY_TI_CL6X_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL6X_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_CL6X_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_CL7X_VERSION) - #undef JSON_HEDLEY_TI_CL7X_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && defined(__C7000__) - #define JSON_HEDLEY_TI_CL7X_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_CL7X_VERSION_CHECK) - #undef JSON_HEDLEY_TI_CL7X_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_CL7X_VERSION) - #define JSON_HEDLEY_TI_CL7X_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL7X_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_CL7X_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TI_CLPRU_VERSION) - #undef JSON_HEDLEY_TI_CLPRU_VERSION -#endif -#if defined(__TI_COMPILER_VERSION__) && defined(__PRU__) - #define JSON_HEDLEY_TI_CLPRU_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000)) -#endif - -#if defined(JSON_HEDLEY_TI_CLPRU_VERSION_CHECK) - #undef JSON_HEDLEY_TI_CLPRU_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TI_CLPRU_VERSION) - #define JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CLPRU_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_CRAY_VERSION) - #undef JSON_HEDLEY_CRAY_VERSION -#endif -#if defined(_CRAYC) - #if defined(_RELEASE_PATCHLEVEL) - #define JSON_HEDLEY_CRAY_VERSION JSON_HEDLEY_VERSION_ENCODE(_RELEASE_MAJOR, _RELEASE_MINOR, _RELEASE_PATCHLEVEL) - #else - #define JSON_HEDLEY_CRAY_VERSION JSON_HEDLEY_VERSION_ENCODE(_RELEASE_MAJOR, _RELEASE_MINOR, 0) - #endif -#endif - -#if defined(JSON_HEDLEY_CRAY_VERSION_CHECK) - #undef JSON_HEDLEY_CRAY_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_CRAY_VERSION) - #define JSON_HEDLEY_CRAY_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_CRAY_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_CRAY_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_IAR_VERSION) - #undef JSON_HEDLEY_IAR_VERSION -#endif -#if defined(__IAR_SYSTEMS_ICC__) - #if __VER__ > 1000 - #define JSON_HEDLEY_IAR_VERSION JSON_HEDLEY_VERSION_ENCODE((__VER__ / 1000000), ((__VER__ / 1000) % 1000), (__VER__ % 1000)) - #else - #define JSON_HEDLEY_IAR_VERSION JSON_HEDLEY_VERSION_ENCODE(VER / 100, __VER__ % 100, 0) - #endif -#endif - -#if defined(JSON_HEDLEY_IAR_VERSION_CHECK) - #undef JSON_HEDLEY_IAR_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_IAR_VERSION) - #define JSON_HEDLEY_IAR_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_IAR_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_IAR_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_TINYC_VERSION) - #undef JSON_HEDLEY_TINYC_VERSION -#endif -#if defined(__TINYC__) - #define JSON_HEDLEY_TINYC_VERSION JSON_HEDLEY_VERSION_ENCODE(__TINYC__ / 1000, (__TINYC__ / 100) % 10, __TINYC__ % 100) -#endif - -#if defined(JSON_HEDLEY_TINYC_VERSION_CHECK) - #undef JSON_HEDLEY_TINYC_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_TINYC_VERSION) - #define JSON_HEDLEY_TINYC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TINYC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_TINYC_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_DMC_VERSION) - #undef JSON_HEDLEY_DMC_VERSION -#endif -#if defined(__DMC__) - #define JSON_HEDLEY_DMC_VERSION JSON_HEDLEY_VERSION_ENCODE(__DMC__ >> 8, (__DMC__ >> 4) & 0xf, __DMC__ & 0xf) -#endif - -#if defined(JSON_HEDLEY_DMC_VERSION_CHECK) - #undef JSON_HEDLEY_DMC_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_DMC_VERSION) - #define JSON_HEDLEY_DMC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_DMC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_DMC_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_COMPCERT_VERSION) - #undef JSON_HEDLEY_COMPCERT_VERSION -#endif -#if defined(__COMPCERT_VERSION__) - #define JSON_HEDLEY_COMPCERT_VERSION JSON_HEDLEY_VERSION_ENCODE(__COMPCERT_VERSION__ / 10000, (__COMPCERT_VERSION__ / 100) % 100, __COMPCERT_VERSION__ % 100) -#endif - -#if defined(JSON_HEDLEY_COMPCERT_VERSION_CHECK) - #undef JSON_HEDLEY_COMPCERT_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_COMPCERT_VERSION) - #define JSON_HEDLEY_COMPCERT_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_COMPCERT_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_COMPCERT_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_PELLES_VERSION) - #undef JSON_HEDLEY_PELLES_VERSION -#endif -#if defined(__POCC__) - #define JSON_HEDLEY_PELLES_VERSION JSON_HEDLEY_VERSION_ENCODE(__POCC__ / 100, __POCC__ % 100, 0) -#endif - -#if defined(JSON_HEDLEY_PELLES_VERSION_CHECK) - #undef JSON_HEDLEY_PELLES_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_PELLES_VERSION) - #define JSON_HEDLEY_PELLES_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_PELLES_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_PELLES_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_GCC_VERSION) - #undef JSON_HEDLEY_GCC_VERSION -#endif -#if \ - defined(JSON_HEDLEY_GNUC_VERSION) && \ - !defined(__clang__) && \ - !defined(JSON_HEDLEY_INTEL_VERSION) && \ - !defined(JSON_HEDLEY_PGI_VERSION) && \ - !defined(JSON_HEDLEY_ARM_VERSION) && \ - !defined(JSON_HEDLEY_TI_VERSION) && \ - !defined(JSON_HEDLEY_TI_ARMCL_VERSION) && \ - !defined(JSON_HEDLEY_TI_CL430_VERSION) && \ - !defined(JSON_HEDLEY_TI_CL2000_VERSION) && \ - !defined(JSON_HEDLEY_TI_CL6X_VERSION) && \ - !defined(JSON_HEDLEY_TI_CL7X_VERSION) && \ - !defined(JSON_HEDLEY_TI_CLPRU_VERSION) && \ - !defined(__COMPCERT__) - #define JSON_HEDLEY_GCC_VERSION JSON_HEDLEY_GNUC_VERSION -#endif - -#if defined(JSON_HEDLEY_GCC_VERSION_CHECK) - #undef JSON_HEDLEY_GCC_VERSION_CHECK -#endif -#if defined(JSON_HEDLEY_GCC_VERSION) - #define JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_GCC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch)) -#else - #define JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) (0) -#endif - -#if defined(JSON_HEDLEY_HAS_ATTRIBUTE) - #undef JSON_HEDLEY_HAS_ATTRIBUTE -#endif -#if defined(__has_attribute) - #define JSON_HEDLEY_HAS_ATTRIBUTE(attribute) __has_attribute(attribute) -#else - #define JSON_HEDLEY_HAS_ATTRIBUTE(attribute) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_ATTRIBUTE) - #undef JSON_HEDLEY_GNUC_HAS_ATTRIBUTE -#endif -#if defined(__has_attribute) - #define JSON_HEDLEY_GNUC_HAS_ATTRIBUTE(attribute,major,minor,patch) __has_attribute(attribute) -#else - #define JSON_HEDLEY_GNUC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_ATTRIBUTE) - #undef JSON_HEDLEY_GCC_HAS_ATTRIBUTE -#endif -#if defined(__has_attribute) - #define JSON_HEDLEY_GCC_HAS_ATTRIBUTE(attribute,major,minor,patch) __has_attribute(attribute) -#else - #define JSON_HEDLEY_GCC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_CPP_ATTRIBUTE) - #undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE -#endif -#if \ - defined(__has_cpp_attribute) && \ - defined(__cplusplus) && \ - (!defined(JSON_HEDLEY_SUNPRO_VERSION) || JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0)) - #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute) __has_cpp_attribute(attribute) -#else - #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute) (0) -#endif - -#if defined(JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS) - #undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS -#endif -#if !defined(__cplusplus) || !defined(__has_cpp_attribute) - #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) (0) -#elif \ - !defined(JSON_HEDLEY_PGI_VERSION) && \ - !defined(JSON_HEDLEY_IAR_VERSION) && \ - (!defined(JSON_HEDLEY_SUNPRO_VERSION) || JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0)) && \ - (!defined(JSON_HEDLEY_MSVC_VERSION) || JSON_HEDLEY_MSVC_VERSION_CHECK(19,20,0)) - #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) JSON_HEDLEY_HAS_CPP_ATTRIBUTE(ns::attribute) -#else - #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE) - #undef JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE -#endif -#if defined(__has_cpp_attribute) && defined(__cplusplus) - #define JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) __has_cpp_attribute(attribute) -#else - #define JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE) - #undef JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE -#endif -#if defined(__has_cpp_attribute) && defined(__cplusplus) - #define JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) __has_cpp_attribute(attribute) -#else - #define JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_BUILTIN) - #undef JSON_HEDLEY_HAS_BUILTIN -#endif -#if defined(__has_builtin) - #define JSON_HEDLEY_HAS_BUILTIN(builtin) __has_builtin(builtin) -#else - #define JSON_HEDLEY_HAS_BUILTIN(builtin) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_BUILTIN) - #undef JSON_HEDLEY_GNUC_HAS_BUILTIN -#endif -#if defined(__has_builtin) - #define JSON_HEDLEY_GNUC_HAS_BUILTIN(builtin,major,minor,patch) __has_builtin(builtin) -#else - #define JSON_HEDLEY_GNUC_HAS_BUILTIN(builtin,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_BUILTIN) - #undef JSON_HEDLEY_GCC_HAS_BUILTIN -#endif -#if defined(__has_builtin) - #define JSON_HEDLEY_GCC_HAS_BUILTIN(builtin,major,minor,patch) __has_builtin(builtin) -#else - #define JSON_HEDLEY_GCC_HAS_BUILTIN(builtin,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_FEATURE) - #undef JSON_HEDLEY_HAS_FEATURE -#endif -#if defined(__has_feature) - #define JSON_HEDLEY_HAS_FEATURE(feature) __has_feature(feature) -#else - #define JSON_HEDLEY_HAS_FEATURE(feature) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_FEATURE) - #undef JSON_HEDLEY_GNUC_HAS_FEATURE -#endif -#if defined(__has_feature) - #define JSON_HEDLEY_GNUC_HAS_FEATURE(feature,major,minor,patch) __has_feature(feature) -#else - #define JSON_HEDLEY_GNUC_HAS_FEATURE(feature,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_FEATURE) - #undef JSON_HEDLEY_GCC_HAS_FEATURE -#endif -#if defined(__has_feature) - #define JSON_HEDLEY_GCC_HAS_FEATURE(feature,major,minor,patch) __has_feature(feature) -#else - #define JSON_HEDLEY_GCC_HAS_FEATURE(feature,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_EXTENSION) - #undef JSON_HEDLEY_HAS_EXTENSION -#endif -#if defined(__has_extension) - #define JSON_HEDLEY_HAS_EXTENSION(extension) __has_extension(extension) -#else - #define JSON_HEDLEY_HAS_EXTENSION(extension) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_EXTENSION) - #undef JSON_HEDLEY_GNUC_HAS_EXTENSION -#endif -#if defined(__has_extension) - #define JSON_HEDLEY_GNUC_HAS_EXTENSION(extension,major,minor,patch) __has_extension(extension) -#else - #define JSON_HEDLEY_GNUC_HAS_EXTENSION(extension,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_EXTENSION) - #undef JSON_HEDLEY_GCC_HAS_EXTENSION -#endif -#if defined(__has_extension) - #define JSON_HEDLEY_GCC_HAS_EXTENSION(extension,major,minor,patch) __has_extension(extension) -#else - #define JSON_HEDLEY_GCC_HAS_EXTENSION(extension,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE) - #undef JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE -#endif -#if defined(__has_declspec_attribute) - #define JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute) __has_declspec_attribute(attribute) -#else - #define JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE) - #undef JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE -#endif -#if defined(__has_declspec_attribute) - #define JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) __has_declspec_attribute(attribute) -#else - #define JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE) - #undef JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE -#endif -#if defined(__has_declspec_attribute) - #define JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) __has_declspec_attribute(attribute) -#else - #define JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_HAS_WARNING) - #undef JSON_HEDLEY_HAS_WARNING -#endif -#if defined(__has_warning) - #define JSON_HEDLEY_HAS_WARNING(warning) __has_warning(warning) -#else - #define JSON_HEDLEY_HAS_WARNING(warning) (0) -#endif - -#if defined(JSON_HEDLEY_GNUC_HAS_WARNING) - #undef JSON_HEDLEY_GNUC_HAS_WARNING -#endif -#if defined(__has_warning) - #define JSON_HEDLEY_GNUC_HAS_WARNING(warning,major,minor,patch) __has_warning(warning) -#else - #define JSON_HEDLEY_GNUC_HAS_WARNING(warning,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_GCC_HAS_WARNING) - #undef JSON_HEDLEY_GCC_HAS_WARNING -#endif -#if defined(__has_warning) - #define JSON_HEDLEY_GCC_HAS_WARNING(warning,major,minor,patch) __has_warning(warning) -#else - #define JSON_HEDLEY_GCC_HAS_WARNING(warning,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -/* JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_ is for - HEDLEY INTERNAL USE ONLY. API subject to change without notice. */ -#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_) - #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_ -#endif -#if defined(__cplusplus) -# if JSON_HEDLEY_HAS_WARNING("-Wc++98-compat") -# if JSON_HEDLEY_HAS_WARNING("-Wc++17-extensions") -# define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \ - _Pragma("clang diagnostic ignored \"-Wc++17-extensions\"") \ - xpr \ - JSON_HEDLEY_DIAGNOSTIC_POP -# else -# define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \ - xpr \ - JSON_HEDLEY_DIAGNOSTIC_POP -# endif -# endif -#endif -#if !defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(x) x -#endif - -#if defined(JSON_HEDLEY_CONST_CAST) - #undef JSON_HEDLEY_CONST_CAST -#endif -#if defined(__cplusplus) -# define JSON_HEDLEY_CONST_CAST(T, expr) (const_cast(expr)) -#elif \ - JSON_HEDLEY_HAS_WARNING("-Wcast-qual") || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) -# define JSON_HEDLEY_CONST_CAST(T, expr) (__extension__ ({ \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL \ - ((T) (expr)); \ - JSON_HEDLEY_DIAGNOSTIC_POP \ - })) -#else -# define JSON_HEDLEY_CONST_CAST(T, expr) ((T) (expr)) -#endif - -#if defined(JSON_HEDLEY_REINTERPRET_CAST) - #undef JSON_HEDLEY_REINTERPRET_CAST -#endif -#if defined(__cplusplus) - #define JSON_HEDLEY_REINTERPRET_CAST(T, expr) (reinterpret_cast(expr)) -#else - #define JSON_HEDLEY_REINTERPRET_CAST(T, expr) ((T) (expr)) -#endif - -#if defined(JSON_HEDLEY_STATIC_CAST) - #undef JSON_HEDLEY_STATIC_CAST -#endif -#if defined(__cplusplus) - #define JSON_HEDLEY_STATIC_CAST(T, expr) (static_cast(expr)) -#else - #define JSON_HEDLEY_STATIC_CAST(T, expr) ((T) (expr)) -#endif - -#if defined(JSON_HEDLEY_CPP_CAST) - #undef JSON_HEDLEY_CPP_CAST -#endif -#if defined(__cplusplus) -# if JSON_HEDLEY_HAS_WARNING("-Wold-style-cast") -# define JSON_HEDLEY_CPP_CAST(T, expr) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("clang diagnostic ignored \"-Wold-style-cast\"") \ - ((T) (expr)) \ - JSON_HEDLEY_DIAGNOSTIC_POP -# elif JSON_HEDLEY_IAR_VERSION_CHECK(8,3,0) -# define JSON_HEDLEY_CPP_CAST(T, expr) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("diag_suppress=Pe137") \ - JSON_HEDLEY_DIAGNOSTIC_POP \ -# else -# define JSON_HEDLEY_CPP_CAST(T, expr) ((T) (expr)) -# endif -#else -# define JSON_HEDLEY_CPP_CAST(T, expr) (expr) -#endif - -#if \ - (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \ - defined(__clang__) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(18,4,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,7,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(2,0,1) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,1,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,0,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \ - JSON_HEDLEY_CRAY_VERSION_CHECK(5,0,0) || \ - JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,17) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(8,0,0) || \ - (JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) && defined(__C99_PRAGMA_OPERATOR)) - #define JSON_HEDLEY_PRAGMA(value) _Pragma(#value) -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) - #define JSON_HEDLEY_PRAGMA(value) __pragma(value) -#else - #define JSON_HEDLEY_PRAGMA(value) -#endif - -#if defined(JSON_HEDLEY_DIAGNOSTIC_PUSH) - #undef JSON_HEDLEY_DIAGNOSTIC_PUSH -#endif -#if defined(JSON_HEDLEY_DIAGNOSTIC_POP) - #undef JSON_HEDLEY_DIAGNOSTIC_POP -#endif -#if defined(__clang__) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("clang diagnostic push") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("clang diagnostic pop") -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)") -#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("GCC diagnostic push") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("GCC diagnostic pop") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH __pragma(warning(push)) - #define JSON_HEDLEY_DIAGNOSTIC_POP __pragma(warning(pop)) -#elif JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("push") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("pop") -#elif \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,4,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,1,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("diag_push") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("diag_pop") -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,90,0) - #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)") - #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)") -#else - #define JSON_HEDLEY_DIAGNOSTIC_PUSH - #define JSON_HEDLEY_DIAGNOSTIC_POP -#endif - -#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED) - #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wdeprecated-declarations") - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("clang diagnostic ignored \"-Wdeprecated-declarations\"") -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("warning(disable:1478 1786)") -#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1215,1444") -#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED __pragma(warning(disable:4996)) -#elif \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1291,1718") -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) && !defined(__cplusplus) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("error_messages(off,E_DEPRECATED_ATT,E_DEPRECATED_ATT_MESS)") -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) && defined(__cplusplus) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("error_messages(off,symdeprecated,symdeprecated2)") -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress=Pe1444,Pe1215") -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,90,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("warn(disable:2241)") -#else - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED -#endif - -#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS) - #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas") - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("clang diagnostic ignored \"-Wunknown-pragmas\"") -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("warning(disable:161)") -#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 1675") -#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("GCC diagnostic ignored \"-Wunknown-pragmas\"") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS __pragma(warning(disable:4068)) -#elif \ - JSON_HEDLEY_TI_VERSION_CHECK(16,9,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,3,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 163") -#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 163") -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress=Pe161") -#else - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS -#endif - -#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES) - #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wunknown-attributes") - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("clang diagnostic ignored \"-Wunknown-attributes\"") -#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"") -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(17,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("warning(disable:1292)") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(19,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES __pragma(warning(disable:5030)) -#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1097") -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,14,0) && defined(__cplusplus) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("error_messages(off,attrskipunsup)") -#elif \ - JSON_HEDLEY_TI_VERSION_CHECK(18,1,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,3,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1173") -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress=Pe1097") -#else - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES -#endif - -#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL) - #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wcast-qual") - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("clang diagnostic ignored \"-Wcast-qual\"") -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("warning(disable:2203 2331)") -#elif JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0) - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("GCC diagnostic ignored \"-Wcast-qual\"") -#else - #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL -#endif - -#if defined(JSON_HEDLEY_DEPRECATED) - #undef JSON_HEDLEY_DEPRECATED -#endif -#if defined(JSON_HEDLEY_DEPRECATED_FOR) - #undef JSON_HEDLEY_DEPRECATED_FOR -#endif -#if JSON_HEDLEY_MSVC_VERSION_CHECK(14,0,0) - #define JSON_HEDLEY_DEPRECATED(since) __declspec(deprecated("Since " # since)) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __declspec(deprecated("Since " #since "; use " #replacement)) -#elif defined(__cplusplus) && (__cplusplus >= 201402L) - #define JSON_HEDLEY_DEPRECATED(since) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[deprecated("Since " #since)]]) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[deprecated("Since " #since "; use " #replacement)]]) -#elif \ - JSON_HEDLEY_HAS_EXTENSION(attribute_deprecated_with_message) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,5,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(18,1,0) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(18,1,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,3,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,3,0) - #define JSON_HEDLEY_DEPRECATED(since) __attribute__((__deprecated__("Since " #since))) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __attribute__((__deprecated__("Since " #since "; use " #replacement))) -#elif \ - JSON_HEDLEY_HAS_ATTRIBUTE(deprecated) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_DEPRECATED(since) __attribute__((__deprecated__)) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __attribute__((__deprecated__)) -#elif \ - JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \ - JSON_HEDLEY_PELLES_VERSION_CHECK(6,50,0) - #define JSON_HEDLEY_DEPRECATED(since) __declspec(deprecated) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __declspec(deprecated) -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_DEPRECATED(since) _Pragma("deprecated") - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) _Pragma("deprecated") -#else - #define JSON_HEDLEY_DEPRECATED(since) - #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) -#endif - -#if defined(JSON_HEDLEY_UNAVAILABLE) - #undef JSON_HEDLEY_UNAVAILABLE -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(warning) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_UNAVAILABLE(available_since) __attribute__((__warning__("Not available until " #available_since))) -#else - #define JSON_HEDLEY_UNAVAILABLE(available_since) -#endif - -#if defined(JSON_HEDLEY_WARN_UNUSED_RESULT) - #undef JSON_HEDLEY_WARN_UNUSED_RESULT -#endif -#if defined(JSON_HEDLEY_WARN_UNUSED_RESULT_MSG) - #undef JSON_HEDLEY_WARN_UNUSED_RESULT_MSG -#endif -#if (JSON_HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard) >= 201907L) - #define JSON_HEDLEY_WARN_UNUSED_RESULT JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]]) - #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard(msg)]]) -#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard) - #define JSON_HEDLEY_WARN_UNUSED_RESULT JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]]) - #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]]) -#elif \ - JSON_HEDLEY_HAS_ATTRIBUTE(warn_unused_result) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \ - (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0) && defined(__cplusplus)) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) - #define JSON_HEDLEY_WARN_UNUSED_RESULT __attribute__((__warn_unused_result__)) - #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) __attribute__((__warn_unused_result__)) -#elif defined(_Check_return_) /* SAL */ - #define JSON_HEDLEY_WARN_UNUSED_RESULT _Check_return_ - #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) _Check_return_ -#else - #define JSON_HEDLEY_WARN_UNUSED_RESULT - #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) -#endif - -#if defined(JSON_HEDLEY_SENTINEL) - #undef JSON_HEDLEY_SENTINEL -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(sentinel) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(5,4,0) - #define JSON_HEDLEY_SENTINEL(position) __attribute__((__sentinel__(position))) -#else - #define JSON_HEDLEY_SENTINEL(position) -#endif - -#if defined(JSON_HEDLEY_NO_RETURN) - #undef JSON_HEDLEY_NO_RETURN -#endif -#if JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_NO_RETURN __noreturn -#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_NO_RETURN __attribute__((__noreturn__)) -#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L - #define JSON_HEDLEY_NO_RETURN _Noreturn -#elif defined(__cplusplus) && (__cplusplus >= 201103L) - #define JSON_HEDLEY_NO_RETURN JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[noreturn]]) -#elif \ - JSON_HEDLEY_HAS_ATTRIBUTE(noreturn) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,2,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_NO_RETURN __attribute__((__noreturn__)) -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) - #define JSON_HEDLEY_NO_RETURN _Pragma("does_not_return") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) - #define JSON_HEDLEY_NO_RETURN __declspec(noreturn) -#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,0,0) && defined(__cplusplus) - #define JSON_HEDLEY_NO_RETURN _Pragma("FUNC_NEVER_RETURNS;") -#elif JSON_HEDLEY_COMPCERT_VERSION_CHECK(3,2,0) - #define JSON_HEDLEY_NO_RETURN __attribute((noreturn)) -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(9,0,0) - #define JSON_HEDLEY_NO_RETURN __declspec(noreturn) -#else - #define JSON_HEDLEY_NO_RETURN -#endif - -#if defined(JSON_HEDLEY_NO_ESCAPE) - #undef JSON_HEDLEY_NO_ESCAPE -#endif -#if JSON_HEDLEY_HAS_ATTRIBUTE(noescape) - #define JSON_HEDLEY_NO_ESCAPE __attribute__((__noescape__)) -#else - #define JSON_HEDLEY_NO_ESCAPE -#endif - -#if defined(JSON_HEDLEY_UNREACHABLE) - #undef JSON_HEDLEY_UNREACHABLE -#endif -#if defined(JSON_HEDLEY_UNREACHABLE_RETURN) - #undef JSON_HEDLEY_UNREACHABLE_RETURN -#endif -#if defined(JSON_HEDLEY_ASSUME) - #undef JSON_HEDLEY_ASSUME -#endif -#if \ - JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_ASSUME(expr) __assume(expr) -#elif JSON_HEDLEY_HAS_BUILTIN(__builtin_assume) - #define JSON_HEDLEY_ASSUME(expr) __builtin_assume(expr) -#elif \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0) - #if defined(__cplusplus) - #define JSON_HEDLEY_ASSUME(expr) std::_nassert(expr) - #else - #define JSON_HEDLEY_ASSUME(expr) _nassert(expr) - #endif -#endif -#if \ - (JSON_HEDLEY_HAS_BUILTIN(__builtin_unreachable) && (!defined(JSON_HEDLEY_ARM_VERSION))) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,5,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(18,10,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(13,1,5) - #define JSON_HEDLEY_UNREACHABLE() __builtin_unreachable() -#elif defined(JSON_HEDLEY_ASSUME) - #define JSON_HEDLEY_UNREACHABLE() JSON_HEDLEY_ASSUME(0) -#endif -#if !defined(JSON_HEDLEY_ASSUME) - #if defined(JSON_HEDLEY_UNREACHABLE) - #define JSON_HEDLEY_ASSUME(expr) JSON_HEDLEY_STATIC_CAST(void, ((expr) ? 1 : (JSON_HEDLEY_UNREACHABLE(), 1))) - #else - #define JSON_HEDLEY_ASSUME(expr) JSON_HEDLEY_STATIC_CAST(void, expr) - #endif -#endif -#if defined(JSON_HEDLEY_UNREACHABLE) - #if \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0) - #define JSON_HEDLEY_UNREACHABLE_RETURN(value) return (JSON_HEDLEY_STATIC_CAST(void, JSON_HEDLEY_ASSUME(0)), (value)) - #else - #define JSON_HEDLEY_UNREACHABLE_RETURN(value) JSON_HEDLEY_UNREACHABLE() - #endif -#else - #define JSON_HEDLEY_UNREACHABLE_RETURN(value) return (value) -#endif -#if !defined(JSON_HEDLEY_UNREACHABLE) - #define JSON_HEDLEY_UNREACHABLE() JSON_HEDLEY_ASSUME(0) -#endif - -JSON_HEDLEY_DIAGNOSTIC_PUSH -#if JSON_HEDLEY_HAS_WARNING("-Wpedantic") - #pragma clang diagnostic ignored "-Wpedantic" -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wc++98-compat-pedantic") && defined(__cplusplus) - #pragma clang diagnostic ignored "-Wc++98-compat-pedantic" -#endif -#if JSON_HEDLEY_GCC_HAS_WARNING("-Wvariadic-macros",4,0,0) - #if defined(__clang__) - #pragma clang diagnostic ignored "-Wvariadic-macros" - #elif defined(JSON_HEDLEY_GCC_VERSION) - #pragma GCC diagnostic ignored "-Wvariadic-macros" - #endif -#endif -#if defined(JSON_HEDLEY_NON_NULL) - #undef JSON_HEDLEY_NON_NULL -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(nonnull) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) - #define JSON_HEDLEY_NON_NULL(...) __attribute__((__nonnull__(__VA_ARGS__))) -#else - #define JSON_HEDLEY_NON_NULL(...) -#endif -JSON_HEDLEY_DIAGNOSTIC_POP - -#if defined(JSON_HEDLEY_PRINTF_FORMAT) - #undef JSON_HEDLEY_PRINTF_FORMAT -#endif -#if defined(__MINGW32__) && JSON_HEDLEY_GCC_HAS_ATTRIBUTE(format,4,4,0) && !defined(__USE_MINGW_ANSI_STDIO) - #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(ms_printf, string_idx, first_to_check))) -#elif defined(__MINGW32__) && JSON_HEDLEY_GCC_HAS_ATTRIBUTE(format,4,4,0) && defined(__USE_MINGW_ANSI_STDIO) - #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(gnu_printf, string_idx, first_to_check))) -#elif \ - JSON_HEDLEY_HAS_ATTRIBUTE(format) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(__printf__, string_idx, first_to_check))) -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(6,0,0) - #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __declspec(vaformat(printf,string_idx,first_to_check)) -#else - #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) -#endif - -#if defined(JSON_HEDLEY_CONSTEXPR) - #undef JSON_HEDLEY_CONSTEXPR -#endif -#if defined(__cplusplus) - #if __cplusplus >= 201103L - #define JSON_HEDLEY_CONSTEXPR JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(constexpr) - #endif -#endif -#if !defined(JSON_HEDLEY_CONSTEXPR) - #define JSON_HEDLEY_CONSTEXPR -#endif - -#if defined(JSON_HEDLEY_PREDICT) - #undef JSON_HEDLEY_PREDICT -#endif -#if defined(JSON_HEDLEY_LIKELY) - #undef JSON_HEDLEY_LIKELY -#endif -#if defined(JSON_HEDLEY_UNLIKELY) - #undef JSON_HEDLEY_UNLIKELY -#endif -#if defined(JSON_HEDLEY_UNPREDICTABLE) - #undef JSON_HEDLEY_UNPREDICTABLE -#endif -#if JSON_HEDLEY_HAS_BUILTIN(__builtin_unpredictable) - #define JSON_HEDLEY_UNPREDICTABLE(expr) __builtin_unpredictable((expr)) -#endif -#if \ - JSON_HEDLEY_HAS_BUILTIN(__builtin_expect_with_probability) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(9,0,0) -# define JSON_HEDLEY_PREDICT(expr, value, probability) __builtin_expect_with_probability( (expr), (value), (probability)) -# define JSON_HEDLEY_PREDICT_TRUE(expr, probability) __builtin_expect_with_probability(!!(expr), 1 , (probability)) -# define JSON_HEDLEY_PREDICT_FALSE(expr, probability) __builtin_expect_with_probability(!!(expr), 0 , (probability)) -# define JSON_HEDLEY_LIKELY(expr) __builtin_expect (!!(expr), 1 ) -# define JSON_HEDLEY_UNLIKELY(expr) __builtin_expect (!!(expr), 0 ) -#elif \ - JSON_HEDLEY_HAS_BUILTIN(__builtin_expect) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0) && defined(__cplusplus)) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,7,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,1,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \ - JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,27) || \ - JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) -# define JSON_HEDLEY_PREDICT(expr, expected, probability) \ - (((probability) >= 0.9) ? __builtin_expect((expr), (expected)) : (JSON_HEDLEY_STATIC_CAST(void, expected), (expr))) -# define JSON_HEDLEY_PREDICT_TRUE(expr, probability) \ - (__extension__ ({ \ - double hedley_probability_ = (probability); \ - ((hedley_probability_ >= 0.9) ? __builtin_expect(!!(expr), 1) : ((hedley_probability_ <= 0.1) ? __builtin_expect(!!(expr), 0) : !!(expr))); \ - })) -# define JSON_HEDLEY_PREDICT_FALSE(expr, probability) \ - (__extension__ ({ \ - double hedley_probability_ = (probability); \ - ((hedley_probability_ >= 0.9) ? __builtin_expect(!!(expr), 0) : ((hedley_probability_ <= 0.1) ? __builtin_expect(!!(expr), 1) : !!(expr))); \ - })) -# define JSON_HEDLEY_LIKELY(expr) __builtin_expect(!!(expr), 1) -# define JSON_HEDLEY_UNLIKELY(expr) __builtin_expect(!!(expr), 0) -#else -# define JSON_HEDLEY_PREDICT(expr, expected, probability) (JSON_HEDLEY_STATIC_CAST(void, expected), (expr)) -# define JSON_HEDLEY_PREDICT_TRUE(expr, probability) (!!(expr)) -# define JSON_HEDLEY_PREDICT_FALSE(expr, probability) (!!(expr)) -# define JSON_HEDLEY_LIKELY(expr) (!!(expr)) -# define JSON_HEDLEY_UNLIKELY(expr) (!!(expr)) -#endif -#if !defined(JSON_HEDLEY_UNPREDICTABLE) - #define JSON_HEDLEY_UNPREDICTABLE(expr) JSON_HEDLEY_PREDICT(expr, 1, 0.5) -#endif - -#if defined(JSON_HEDLEY_MALLOC) - #undef JSON_HEDLEY_MALLOC -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(malloc) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(12,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_MALLOC __attribute__((__malloc__)) -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) - #define JSON_HEDLEY_MALLOC _Pragma("returns_new_memory") -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(14, 0, 0) - #define JSON_HEDLEY_MALLOC __declspec(restrict) -#else - #define JSON_HEDLEY_MALLOC -#endif - -#if defined(JSON_HEDLEY_PURE) - #undef JSON_HEDLEY_PURE -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(pure) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(2,96,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) -# define JSON_HEDLEY_PURE __attribute__((__pure__)) -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) -# define JSON_HEDLEY_PURE _Pragma("does_not_write_global_data") -#elif defined(__cplusplus) && \ - ( \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(2,0,1) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) \ - ) -# define JSON_HEDLEY_PURE _Pragma("FUNC_IS_PURE;") -#else -# define JSON_HEDLEY_PURE -#endif - -#if defined(JSON_HEDLEY_CONST) - #undef JSON_HEDLEY_CONST -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(const) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(2,5,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) - #define JSON_HEDLEY_CONST __attribute__((__const__)) -#elif \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) - #define JSON_HEDLEY_CONST _Pragma("no_side_effect") -#else - #define JSON_HEDLEY_CONST JSON_HEDLEY_PURE -#endif - -#if defined(JSON_HEDLEY_RESTRICT) - #undef JSON_HEDLEY_RESTRICT -#endif -#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) && !defined(__cplusplus) - #define JSON_HEDLEY_RESTRICT restrict -#elif \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_MSVC_VERSION_CHECK(14,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,4) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,1,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,14,0) && defined(__cplusplus)) || \ - JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) || \ - defined(__clang__) - #define JSON_HEDLEY_RESTRICT __restrict -#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,3,0) && !defined(__cplusplus) - #define JSON_HEDLEY_RESTRICT _Restrict -#else - #define JSON_HEDLEY_RESTRICT -#endif - -#if defined(JSON_HEDLEY_INLINE) - #undef JSON_HEDLEY_INLINE -#endif -#if \ - (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \ - (defined(__cplusplus) && (__cplusplus >= 199711L)) - #define JSON_HEDLEY_INLINE inline -#elif \ - defined(JSON_HEDLEY_GCC_VERSION) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(6,2,0) - #define JSON_HEDLEY_INLINE __inline__ -#elif \ - JSON_HEDLEY_MSVC_VERSION_CHECK(12,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,1,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(3,1,0) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_INLINE __inline -#else - #define JSON_HEDLEY_INLINE -#endif - -#if defined(JSON_HEDLEY_ALWAYS_INLINE) - #undef JSON_HEDLEY_ALWAYS_INLINE -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(always_inline) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) -# define JSON_HEDLEY_ALWAYS_INLINE __attribute__((__always_inline__)) JSON_HEDLEY_INLINE -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(12,0,0) -# define JSON_HEDLEY_ALWAYS_INLINE __forceinline -#elif defined(__cplusplus) && \ - ( \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) \ - ) -# define JSON_HEDLEY_ALWAYS_INLINE _Pragma("FUNC_ALWAYS_INLINE;") -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) -# define JSON_HEDLEY_ALWAYS_INLINE _Pragma("inline=forced") -#else -# define JSON_HEDLEY_ALWAYS_INLINE JSON_HEDLEY_INLINE -#endif - -#if defined(JSON_HEDLEY_NEVER_INLINE) - #undef JSON_HEDLEY_NEVER_INLINE -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(noinline) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \ - JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \ - (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \ - (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \ - (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \ - JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \ - JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) - #define JSON_HEDLEY_NEVER_INLINE __attribute__((__noinline__)) -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) - #define JSON_HEDLEY_NEVER_INLINE __declspec(noinline) -#elif JSON_HEDLEY_PGI_VERSION_CHECK(10,2,0) - #define JSON_HEDLEY_NEVER_INLINE _Pragma("noinline") -#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,0,0) && defined(__cplusplus) - #define JSON_HEDLEY_NEVER_INLINE _Pragma("FUNC_CANNOT_INLINE;") -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) - #define JSON_HEDLEY_NEVER_INLINE _Pragma("inline=never") -#elif JSON_HEDLEY_COMPCERT_VERSION_CHECK(3,2,0) - #define JSON_HEDLEY_NEVER_INLINE __attribute((noinline)) -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(9,0,0) - #define JSON_HEDLEY_NEVER_INLINE __declspec(noinline) -#else - #define JSON_HEDLEY_NEVER_INLINE -#endif - -#if defined(JSON_HEDLEY_PRIVATE) - #undef JSON_HEDLEY_PRIVATE -#endif -#if defined(JSON_HEDLEY_PUBLIC) - #undef JSON_HEDLEY_PUBLIC -#endif -#if defined(JSON_HEDLEY_IMPORT) - #undef JSON_HEDLEY_IMPORT -#endif -#if defined(_WIN32) || defined(__CYGWIN__) -# define JSON_HEDLEY_PRIVATE -# define JSON_HEDLEY_PUBLIC __declspec(dllexport) -# define JSON_HEDLEY_IMPORT __declspec(dllimport) -#else -# if \ - JSON_HEDLEY_HAS_ATTRIBUTE(visibility) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \ - JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \ - ( \ - defined(__TI_EABI__) && \ - ( \ - (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) \ - ) \ - ) -# define JSON_HEDLEY_PRIVATE __attribute__((__visibility__("hidden"))) -# define JSON_HEDLEY_PUBLIC __attribute__((__visibility__("default"))) -# else -# define JSON_HEDLEY_PRIVATE -# define JSON_HEDLEY_PUBLIC -# endif -# define JSON_HEDLEY_IMPORT extern -#endif - -#if defined(JSON_HEDLEY_NO_THROW) - #undef JSON_HEDLEY_NO_THROW -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(nothrow) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) - #define JSON_HEDLEY_NO_THROW __attribute__((__nothrow__)) -#elif \ - JSON_HEDLEY_MSVC_VERSION_CHECK(13,1,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) - #define JSON_HEDLEY_NO_THROW __declspec(nothrow) -#else - #define JSON_HEDLEY_NO_THROW -#endif - -#if defined(JSON_HEDLEY_FALL_THROUGH) - #undef JSON_HEDLEY_FALL_THROUGH -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(fallthrough) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(7,0,0) - #define JSON_HEDLEY_FALL_THROUGH __attribute__((__fallthrough__)) -#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(clang,fallthrough) - #define JSON_HEDLEY_FALL_THROUGH JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[clang::fallthrough]]) -#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE(fallthrough) - #define JSON_HEDLEY_FALL_THROUGH JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[fallthrough]]) -#elif defined(__fallthrough) /* SAL */ - #define JSON_HEDLEY_FALL_THROUGH __fallthrough -#else - #define JSON_HEDLEY_FALL_THROUGH -#endif - -#if defined(JSON_HEDLEY_RETURNS_NON_NULL) - #undef JSON_HEDLEY_RETURNS_NON_NULL -#endif -#if \ - JSON_HEDLEY_HAS_ATTRIBUTE(returns_nonnull) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,9,0) - #define JSON_HEDLEY_RETURNS_NON_NULL __attribute__((__returns_nonnull__)) -#elif defined(_Ret_notnull_) /* SAL */ - #define JSON_HEDLEY_RETURNS_NON_NULL _Ret_notnull_ -#else - #define JSON_HEDLEY_RETURNS_NON_NULL -#endif - -#if defined(JSON_HEDLEY_ARRAY_PARAM) - #undef JSON_HEDLEY_ARRAY_PARAM -#endif -#if \ - defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) && \ - !defined(__STDC_NO_VLA__) && \ - !defined(__cplusplus) && \ - !defined(JSON_HEDLEY_PGI_VERSION) && \ - !defined(JSON_HEDLEY_TINYC_VERSION) - #define JSON_HEDLEY_ARRAY_PARAM(name) (name) -#else - #define JSON_HEDLEY_ARRAY_PARAM(name) -#endif - -#if defined(JSON_HEDLEY_IS_CONSTANT) - #undef JSON_HEDLEY_IS_CONSTANT -#endif -#if defined(JSON_HEDLEY_REQUIRE_CONSTEXPR) - #undef JSON_HEDLEY_REQUIRE_CONSTEXPR -#endif -/* JSON_HEDLEY_IS_CONSTEXPR_ is for - HEDLEY INTERNAL USE ONLY. API subject to change without notice. */ -#if defined(JSON_HEDLEY_IS_CONSTEXPR_) - #undef JSON_HEDLEY_IS_CONSTEXPR_ -#endif -#if \ - JSON_HEDLEY_HAS_BUILTIN(__builtin_constant_p) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,19) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \ - JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \ - (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) && !defined(__cplusplus)) || \ - JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) - #define JSON_HEDLEY_IS_CONSTANT(expr) __builtin_constant_p(expr) -#endif -#if !defined(__cplusplus) -# if \ - JSON_HEDLEY_HAS_BUILTIN(__builtin_types_compatible_p) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \ - JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(5,4,0) || \ - JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,24) -#if defined(__INTPTR_TYPE__) - #define JSON_HEDLEY_IS_CONSTEXPR_(expr) __builtin_types_compatible_p(__typeof__((1 ? (void*) ((__INTPTR_TYPE__) ((expr) * 0)) : (int*) 0)), int*) -#else - #include - #define JSON_HEDLEY_IS_CONSTEXPR_(expr) __builtin_types_compatible_p(__typeof__((1 ? (void*) ((intptr_t) ((expr) * 0)) : (int*) 0)), int*) -#endif -# elif \ - ( \ - defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) && \ - !defined(JSON_HEDLEY_SUNPRO_VERSION) && \ - !defined(JSON_HEDLEY_PGI_VERSION) && \ - !defined(JSON_HEDLEY_IAR_VERSION)) || \ - JSON_HEDLEY_HAS_EXTENSION(c_generic_selections) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,9,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(17,0,0) || \ - JSON_HEDLEY_IBM_VERSION_CHECK(12,1,0) || \ - JSON_HEDLEY_ARM_VERSION_CHECK(5,3,0) -#if defined(__INTPTR_TYPE__) - #define JSON_HEDLEY_IS_CONSTEXPR_(expr) _Generic((1 ? (void*) ((__INTPTR_TYPE__) ((expr) * 0)) : (int*) 0), int*: 1, void*: 0) -#else - #include - #define JSON_HEDLEY_IS_CONSTEXPR_(expr) _Generic((1 ? (void*) ((intptr_t) * 0) : (int*) 0), int*: 1, void*: 0) -#endif -# elif \ - defined(JSON_HEDLEY_GCC_VERSION) || \ - defined(JSON_HEDLEY_INTEL_VERSION) || \ - defined(JSON_HEDLEY_TINYC_VERSION) || \ - defined(JSON_HEDLEY_TI_ARMCL_VERSION) || \ - JSON_HEDLEY_TI_CL430_VERSION_CHECK(18,12,0) || \ - defined(JSON_HEDLEY_TI_CL2000_VERSION) || \ - defined(JSON_HEDLEY_TI_CL6X_VERSION) || \ - defined(JSON_HEDLEY_TI_CL7X_VERSION) || \ - defined(JSON_HEDLEY_TI_CLPRU_VERSION) || \ - defined(__clang__) -# define JSON_HEDLEY_IS_CONSTEXPR_(expr) ( \ - sizeof(void) != \ - sizeof(*( \ - 1 ? \ - ((void*) ((expr) * 0L) ) : \ -((struct { char v[sizeof(void) * 2]; } *) 1) \ - ) \ - ) \ - ) -# endif -#endif -#if defined(JSON_HEDLEY_IS_CONSTEXPR_) - #if !defined(JSON_HEDLEY_IS_CONSTANT) - #define JSON_HEDLEY_IS_CONSTANT(expr) JSON_HEDLEY_IS_CONSTEXPR_(expr) - #endif - #define JSON_HEDLEY_REQUIRE_CONSTEXPR(expr) (JSON_HEDLEY_IS_CONSTEXPR_(expr) ? (expr) : (-1)) -#else - #if !defined(JSON_HEDLEY_IS_CONSTANT) - #define JSON_HEDLEY_IS_CONSTANT(expr) (0) - #endif - #define JSON_HEDLEY_REQUIRE_CONSTEXPR(expr) (expr) -#endif - -#if defined(JSON_HEDLEY_BEGIN_C_DECLS) - #undef JSON_HEDLEY_BEGIN_C_DECLS -#endif -#if defined(JSON_HEDLEY_END_C_DECLS) - #undef JSON_HEDLEY_END_C_DECLS -#endif -#if defined(JSON_HEDLEY_C_DECL) - #undef JSON_HEDLEY_C_DECL -#endif -#if defined(__cplusplus) - #define JSON_HEDLEY_BEGIN_C_DECLS extern "C" { - #define JSON_HEDLEY_END_C_DECLS } - #define JSON_HEDLEY_C_DECL extern "C" -#else - #define JSON_HEDLEY_BEGIN_C_DECLS - #define JSON_HEDLEY_END_C_DECLS - #define JSON_HEDLEY_C_DECL -#endif - -#if defined(JSON_HEDLEY_STATIC_ASSERT) - #undef JSON_HEDLEY_STATIC_ASSERT -#endif -#if \ - !defined(__cplusplus) && ( \ - (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \ - JSON_HEDLEY_HAS_FEATURE(c_static_assert) || \ - JSON_HEDLEY_GCC_VERSION_CHECK(6,0,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \ - defined(_Static_assert) \ - ) -# define JSON_HEDLEY_STATIC_ASSERT(expr, message) _Static_assert(expr, message) -#elif \ - (defined(__cplusplus) && (__cplusplus >= 201103L)) || \ - JSON_HEDLEY_MSVC_VERSION_CHECK(16,0,0) -# define JSON_HEDLEY_STATIC_ASSERT(expr, message) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(static_assert(expr, message)) -#else -# define JSON_HEDLEY_STATIC_ASSERT(expr, message) -#endif - -#if defined(JSON_HEDLEY_NULL) - #undef JSON_HEDLEY_NULL -#endif -#if defined(__cplusplus) - #if __cplusplus >= 201103L - #define JSON_HEDLEY_NULL JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(nullptr) - #elif defined(NULL) - #define JSON_HEDLEY_NULL NULL - #else - #define JSON_HEDLEY_NULL JSON_HEDLEY_STATIC_CAST(void*, 0) - #endif -#elif defined(NULL) - #define JSON_HEDLEY_NULL NULL -#else - #define JSON_HEDLEY_NULL ((void*) 0) -#endif - -#if defined(JSON_HEDLEY_MESSAGE) - #undef JSON_HEDLEY_MESSAGE -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas") -# define JSON_HEDLEY_MESSAGE(msg) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \ - JSON_HEDLEY_PRAGMA(message msg) \ - JSON_HEDLEY_DIAGNOSTIC_POP -#elif \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,4,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) -# define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message msg) -#elif JSON_HEDLEY_CRAY_VERSION_CHECK(5,0,0) -# define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(_CRI message msg) -#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) -# define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message(msg)) -#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,0,0) -# define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message(msg)) -#else -# define JSON_HEDLEY_MESSAGE(msg) -#endif - -#if defined(JSON_HEDLEY_WARNING) - #undef JSON_HEDLEY_WARNING -#endif -#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas") -# define JSON_HEDLEY_WARNING(msg) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \ - JSON_HEDLEY_PRAGMA(clang warning msg) \ - JSON_HEDLEY_DIAGNOSTIC_POP -#elif \ - JSON_HEDLEY_GCC_VERSION_CHECK(4,8,0) || \ - JSON_HEDLEY_PGI_VERSION_CHECK(18,4,0) || \ - JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) -# define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_PRAGMA(GCC warning msg) -#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) -# define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_PRAGMA(message(msg)) -#else -# define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_MESSAGE(msg) -#endif - -#if defined(JSON_HEDLEY_REQUIRE) - #undef JSON_HEDLEY_REQUIRE -#endif -#if defined(JSON_HEDLEY_REQUIRE_MSG) - #undef JSON_HEDLEY_REQUIRE_MSG -#endif -#if JSON_HEDLEY_HAS_ATTRIBUTE(diagnose_if) -# if JSON_HEDLEY_HAS_WARNING("-Wgcc-compat") -# define JSON_HEDLEY_REQUIRE(expr) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("clang diagnostic ignored \"-Wgcc-compat\"") \ - __attribute__((diagnose_if(!(expr), #expr, "error"))) \ - JSON_HEDLEY_DIAGNOSTIC_POP -# define JSON_HEDLEY_REQUIRE_MSG(expr,msg) \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("clang diagnostic ignored \"-Wgcc-compat\"") \ - __attribute__((diagnose_if(!(expr), msg, "error"))) \ - JSON_HEDLEY_DIAGNOSTIC_POP -# else -# define JSON_HEDLEY_REQUIRE(expr) __attribute__((diagnose_if(!(expr), #expr, "error"))) -# define JSON_HEDLEY_REQUIRE_MSG(expr,msg) __attribute__((diagnose_if(!(expr), msg, "error"))) -# endif -#else -# define JSON_HEDLEY_REQUIRE(expr) -# define JSON_HEDLEY_REQUIRE_MSG(expr,msg) -#endif - -#if defined(JSON_HEDLEY_FLAGS) - #undef JSON_HEDLEY_FLAGS -#endif -#if JSON_HEDLEY_HAS_ATTRIBUTE(flag_enum) - #define JSON_HEDLEY_FLAGS __attribute__((__flag_enum__)) -#endif - -#if defined(JSON_HEDLEY_FLAGS_CAST) - #undef JSON_HEDLEY_FLAGS_CAST -#endif -#if JSON_HEDLEY_INTEL_VERSION_CHECK(19,0,0) -# define JSON_HEDLEY_FLAGS_CAST(T, expr) (__extension__ ({ \ - JSON_HEDLEY_DIAGNOSTIC_PUSH \ - _Pragma("warning(disable:188)") \ - ((T) (expr)); \ - JSON_HEDLEY_DIAGNOSTIC_POP \ - })) -#else -# define JSON_HEDLEY_FLAGS_CAST(T, expr) JSON_HEDLEY_STATIC_CAST(T, expr) -#endif - -#if defined(JSON_HEDLEY_EMPTY_BASES) - #undef JSON_HEDLEY_EMPTY_BASES -#endif -#if JSON_HEDLEY_MSVC_VERSION_CHECK(19,0,23918) && !JSON_HEDLEY_MSVC_VERSION_CHECK(20,0,0) - #define JSON_HEDLEY_EMPTY_BASES __declspec(empty_bases) -#else - #define JSON_HEDLEY_EMPTY_BASES -#endif - -/* Remaining macros are deprecated. */ - -#if defined(JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK) - #undef JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK -#endif -#if defined(__clang__) - #define JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK(major,minor,patch) (0) -#else - #define JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK(major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) -#endif - -#if defined(JSON_HEDLEY_CLANG_HAS_ATTRIBUTE) - #undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE -#endif -#define JSON_HEDLEY_CLANG_HAS_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_ATTRIBUTE(attribute) - -#if defined(JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE) - #undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE -#endif -#define JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute) - -#if defined(JSON_HEDLEY_CLANG_HAS_BUILTIN) - #undef JSON_HEDLEY_CLANG_HAS_BUILTIN -#endif -#define JSON_HEDLEY_CLANG_HAS_BUILTIN(builtin) JSON_HEDLEY_HAS_BUILTIN(builtin) - -#if defined(JSON_HEDLEY_CLANG_HAS_FEATURE) - #undef JSON_HEDLEY_CLANG_HAS_FEATURE -#endif -#define JSON_HEDLEY_CLANG_HAS_FEATURE(feature) JSON_HEDLEY_HAS_FEATURE(feature) - -#if defined(JSON_HEDLEY_CLANG_HAS_EXTENSION) - #undef JSON_HEDLEY_CLANG_HAS_EXTENSION -#endif -#define JSON_HEDLEY_CLANG_HAS_EXTENSION(extension) JSON_HEDLEY_HAS_EXTENSION(extension) - -#if defined(JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE) - #undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE -#endif -#define JSON_HEDLEY_CLANG_HAS_DECLSPEC_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute) - -#if defined(JSON_HEDLEY_CLANG_HAS_WARNING) - #undef JSON_HEDLEY_CLANG_HAS_WARNING -#endif -#define JSON_HEDLEY_CLANG_HAS_WARNING(warning) JSON_HEDLEY_HAS_WARNING(warning) - -#endif /* !defined(JSON_HEDLEY_VERSION) || (JSON_HEDLEY_VERSION < X) */ diff --git a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley_undef.hpp b/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley_undef.hpp deleted file mode 100644 index 5fc2e31bb99..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/nlohmann-3.9.1/nlohmann/thirdparty/hedley/hedley_undef.hpp +++ /dev/null @@ -1,143 +0,0 @@ -#undef JSON_HEDLEY_ALWAYS_INLINE -#undef JSON_HEDLEY_ARM_VERSION -#undef JSON_HEDLEY_ARM_VERSION_CHECK -#undef JSON_HEDLEY_ARRAY_PARAM -#undef JSON_HEDLEY_ASSUME -#undef JSON_HEDLEY_BEGIN_C_DECLS -#undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE -#undef JSON_HEDLEY_CLANG_HAS_BUILTIN -#undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE -#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE -#undef JSON_HEDLEY_CLANG_HAS_EXTENSION -#undef JSON_HEDLEY_CLANG_HAS_FEATURE -#undef JSON_HEDLEY_CLANG_HAS_WARNING -#undef JSON_HEDLEY_COMPCERT_VERSION -#undef JSON_HEDLEY_COMPCERT_VERSION_CHECK -#undef JSON_HEDLEY_CONCAT -#undef JSON_HEDLEY_CONCAT3 -#undef JSON_HEDLEY_CONCAT3_EX -#undef JSON_HEDLEY_CONCAT_EX -#undef JSON_HEDLEY_CONST -#undef JSON_HEDLEY_CONSTEXPR -#undef JSON_HEDLEY_CONST_CAST -#undef JSON_HEDLEY_CPP_CAST -#undef JSON_HEDLEY_CRAY_VERSION -#undef JSON_HEDLEY_CRAY_VERSION_CHECK -#undef JSON_HEDLEY_C_DECL -#undef JSON_HEDLEY_DEPRECATED -#undef JSON_HEDLEY_DEPRECATED_FOR -#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL -#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_ -#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED -#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES -#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS -#undef JSON_HEDLEY_DIAGNOSTIC_POP -#undef JSON_HEDLEY_DIAGNOSTIC_PUSH -#undef JSON_HEDLEY_DMC_VERSION -#undef JSON_HEDLEY_DMC_VERSION_CHECK -#undef JSON_HEDLEY_EMPTY_BASES -#undef JSON_HEDLEY_EMSCRIPTEN_VERSION -#undef JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK -#undef JSON_HEDLEY_END_C_DECLS -#undef JSON_HEDLEY_FLAGS -#undef JSON_HEDLEY_FLAGS_CAST -#undef JSON_HEDLEY_GCC_HAS_ATTRIBUTE -#undef JSON_HEDLEY_GCC_HAS_BUILTIN -#undef JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE -#undef JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE -#undef JSON_HEDLEY_GCC_HAS_EXTENSION -#undef JSON_HEDLEY_GCC_HAS_FEATURE -#undef JSON_HEDLEY_GCC_HAS_WARNING -#undef JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK -#undef JSON_HEDLEY_GCC_VERSION -#undef JSON_HEDLEY_GCC_VERSION_CHECK -#undef JSON_HEDLEY_GNUC_HAS_ATTRIBUTE -#undef JSON_HEDLEY_GNUC_HAS_BUILTIN -#undef JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE -#undef JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE -#undef JSON_HEDLEY_GNUC_HAS_EXTENSION -#undef JSON_HEDLEY_GNUC_HAS_FEATURE -#undef JSON_HEDLEY_GNUC_HAS_WARNING -#undef JSON_HEDLEY_GNUC_VERSION -#undef JSON_HEDLEY_GNUC_VERSION_CHECK -#undef JSON_HEDLEY_HAS_ATTRIBUTE -#undef JSON_HEDLEY_HAS_BUILTIN -#undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE -#undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS -#undef JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE -#undef JSON_HEDLEY_HAS_EXTENSION -#undef JSON_HEDLEY_HAS_FEATURE -#undef JSON_HEDLEY_HAS_WARNING -#undef JSON_HEDLEY_IAR_VERSION -#undef JSON_HEDLEY_IAR_VERSION_CHECK -#undef JSON_HEDLEY_IBM_VERSION -#undef JSON_HEDLEY_IBM_VERSION_CHECK -#undef JSON_HEDLEY_IMPORT -#undef JSON_HEDLEY_INLINE -#undef JSON_HEDLEY_INTEL_VERSION -#undef JSON_HEDLEY_INTEL_VERSION_CHECK -#undef JSON_HEDLEY_IS_CONSTANT -#undef JSON_HEDLEY_IS_CONSTEXPR_ -#undef JSON_HEDLEY_LIKELY -#undef JSON_HEDLEY_MALLOC -#undef JSON_HEDLEY_MESSAGE -#undef JSON_HEDLEY_MSVC_VERSION -#undef JSON_HEDLEY_MSVC_VERSION_CHECK -#undef JSON_HEDLEY_NEVER_INLINE -#undef JSON_HEDLEY_NON_NULL -#undef JSON_HEDLEY_NO_ESCAPE -#undef JSON_HEDLEY_NO_RETURN -#undef JSON_HEDLEY_NO_THROW -#undef JSON_HEDLEY_NULL -#undef JSON_HEDLEY_PELLES_VERSION -#undef JSON_HEDLEY_PELLES_VERSION_CHECK -#undef JSON_HEDLEY_PGI_VERSION -#undef JSON_HEDLEY_PGI_VERSION_CHECK -#undef JSON_HEDLEY_PREDICT -#undef JSON_HEDLEY_PRINTF_FORMAT -#undef JSON_HEDLEY_PRIVATE -#undef JSON_HEDLEY_PUBLIC -#undef JSON_HEDLEY_PURE -#undef JSON_HEDLEY_REINTERPRET_CAST -#undef JSON_HEDLEY_REQUIRE -#undef JSON_HEDLEY_REQUIRE_CONSTEXPR -#undef JSON_HEDLEY_REQUIRE_MSG -#undef JSON_HEDLEY_RESTRICT -#undef JSON_HEDLEY_RETURNS_NON_NULL -#undef JSON_HEDLEY_SENTINEL -#undef JSON_HEDLEY_STATIC_ASSERT -#undef JSON_HEDLEY_STATIC_CAST -#undef JSON_HEDLEY_STRINGIFY -#undef JSON_HEDLEY_STRINGIFY_EX -#undef JSON_HEDLEY_SUNPRO_VERSION -#undef JSON_HEDLEY_SUNPRO_VERSION_CHECK -#undef JSON_HEDLEY_TINYC_VERSION -#undef JSON_HEDLEY_TINYC_VERSION_CHECK -#undef JSON_HEDLEY_TI_ARMCL_VERSION -#undef JSON_HEDLEY_TI_ARMCL_VERSION_CHECK -#undef JSON_HEDLEY_TI_CL2000_VERSION -#undef JSON_HEDLEY_TI_CL2000_VERSION_CHECK -#undef JSON_HEDLEY_TI_CL430_VERSION -#undef JSON_HEDLEY_TI_CL430_VERSION_CHECK -#undef JSON_HEDLEY_TI_CL6X_VERSION -#undef JSON_HEDLEY_TI_CL6X_VERSION_CHECK -#undef JSON_HEDLEY_TI_CL7X_VERSION -#undef JSON_HEDLEY_TI_CL7X_VERSION_CHECK -#undef JSON_HEDLEY_TI_CLPRU_VERSION -#undef JSON_HEDLEY_TI_CLPRU_VERSION_CHECK -#undef JSON_HEDLEY_TI_VERSION -#undef JSON_HEDLEY_TI_VERSION_CHECK -#undef JSON_HEDLEY_UNAVAILABLE -#undef JSON_HEDLEY_UNLIKELY -#undef JSON_HEDLEY_UNPREDICTABLE -#undef JSON_HEDLEY_UNREACHABLE -#undef JSON_HEDLEY_UNREACHABLE_RETURN -#undef JSON_HEDLEY_VERSION -#undef JSON_HEDLEY_VERSION_DECODE_MAJOR -#undef JSON_HEDLEY_VERSION_DECODE_MINOR -#undef JSON_HEDLEY_VERSION_DECODE_REVISION -#undef JSON_HEDLEY_VERSION_ENCODE -#undef JSON_HEDLEY_WARNING -#undef JSON_HEDLEY_WARN_UNUSED_RESULT -#undef JSON_HEDLEY_WARN_UNUSED_RESULT_MSG -#undef JSON_HEDLEY_FALL_THROUGH diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/.gitignore b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/.gitignore deleted file mode 100644 index b8bd0267bdf..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/.gitignore +++ /dev/null @@ -1,28 +0,0 @@ -# Compiled Object files -*.slo -*.lo -*.o -*.obj - -# Precompiled Headers -*.gch -*.pch - -# Compiled Dynamic libraries -*.so -*.dylib -*.dll - -# Fortran module files -*.mod - -# Compiled Static libraries -*.lai -*.la -*.a -*.lib - -# Executables -*.exe -*.out -*.app diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/CMakeLists.txt b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/CMakeLists.txt deleted file mode 100644 index 7d9a43b8a9c..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/CMakeLists.txt +++ /dev/null @@ -1,3 +0,0 @@ -cmake_minimum_required(VERSION 2.4) -add_definitions(-std=c++11) -add_executable(qdt-pi examples/qdt-pi.cc) diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/LICENSE.txt b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/LICENSE.txt deleted file mode 100644 index 94a9ed024d3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/LICENSE.txt +++ /dev/null @@ -1,674 +0,0 @@ - GNU GENERAL PUBLIC LICENSE - Version 3, 29 June 2007 - - Copyright (C) 2007 Free Software Foundation, Inc. - Everyone is permitted to copy and distribute verbatim copies - of this license document, but changing it is not allowed. - - Preamble - - The GNU General Public License is a free, copyleft license for -software and other kinds of works. - - The licenses for most software and other practical works are designed -to take away your freedom to share and change the works. By contrast, -the GNU General Public License is intended to guarantee your freedom to -share and change all versions of a program--to make sure it remains free -software for all its users. We, the Free Software Foundation, use the -GNU General Public License for most of our software; it applies also to -any other work released this way by its authors. You can apply it to -your programs, too. - - When we speak of free software, we are referring to freedom, not -price. Our General Public Licenses are designed to make sure that you -have the freedom to distribute copies of free software (and charge for -them if you wish), that you receive source code or can get it if you -want it, that you can change the software or use pieces of it in new -free programs, and that you know you can do these things. - - To protect your rights, we need to prevent others from denying you -these rights or asking you to surrender the rights. Therefore, you have -certain responsibilities if you distribute copies of the software, or if -you modify it: responsibilities to respect the freedom of others. - - For example, if you distribute copies of such a program, whether -gratis or for a fee, you must pass on to the recipients the same -freedoms that you received. You must make sure that they, too, receive -or can get the source code. And you must show them these terms so they -know their rights. - - Developers that use the GNU GPL protect your rights with two steps: -(1) assert copyright on the software, and (2) offer you this License -giving you legal permission to copy, distribute and/or modify it. - - For the developers' and authors' protection, the GPL clearly explains -that there is no warranty for this free software. For both users' and -authors' sake, the GPL requires that modified versions be marked as -changed, so that their problems will not be attributed erroneously to -authors of previous versions. - - Some devices are designed to deny users access to install or run -modified versions of the software inside them, although the manufacturer -can do so. This is fundamentally incompatible with the aim of -protecting users' freedom to change the software. The systematic -pattern of such abuse occurs in the area of products for individuals to -use, which is precisely where it is most unacceptable. Therefore, we -have designed this version of the GPL to prohibit the practice for those -products. If such problems arise substantially in other domains, we -stand ready to extend this provision to those domains in future versions -of the GPL, as needed to protect the freedom of users. - - Finally, every program is threatened constantly by software patents. -States should not allow patents to restrict development and use of -software on general-purpose computers, but in those that do, we wish to -avoid the special danger that patents applied to a free program could -make it effectively proprietary. To prevent this, the GPL assures that -patents cannot be used to render the program non-free. - - The precise terms and conditions for copying, distribution and -modification follow. - - TERMS AND CONDITIONS - - 0. Definitions. - - "This License" refers to version 3 of the GNU General Public License. - - "Copyright" also means copyright-like laws that apply to other kinds of -works, such as semiconductor masks. - - "The Program" refers to any copyrightable work licensed under this -License. Each licensee is addressed as "you". "Licensees" and -"recipients" may be individuals or organizations. - - To "modify" a work means to copy from or adapt all or part of the work -in a fashion requiring copyright permission, other than the making of an -exact copy. The resulting work is called a "modified version" of the -earlier work or a work "based on" the earlier work. - - A "covered work" means either the unmodified Program or a work based -on the Program. - - To "propagate" a work means to do anything with it that, without -permission, would make you directly or secondarily liable for -infringement under applicable copyright law, except executing it on a -computer or modifying a private copy. Propagation includes copying, -distribution (with or without modification), making available to the -public, and in some countries other activities as well. - - To "convey" a work means any kind of propagation that enables other -parties to make or receive copies. Mere interaction with a user through -a computer network, with no transfer of a copy, is not conveying. - - An interactive user interface displays "Appropriate Legal Notices" -to the extent that it includes a convenient and prominently visible -feature that (1) displays an appropriate copyright notice, and (2) -tells the user that there is no warranty for the work (except to the -extent that warranties are provided), that licensees may convey the -work under this License, and how to view a copy of this License. If -the interface presents a list of user commands or options, such as a -menu, a prominent item in the list meets this criterion. - - 1. Source Code. - - The "source code" for a work means the preferred form of the work -for making modifications to it. "Object code" means any non-source -form of a work. - - A "Standard Interface" means an interface that either is an official -standard defined by a recognized standards body, or, in the case of -interfaces specified for a particular programming language, one that -is widely used among developers working in that language. - - The "System Libraries" of an executable work include anything, other -than the work as a whole, that (a) is included in the normal form of -packaging a Major Component, but which is not part of that Major -Component, and (b) serves only to enable use of the work with that -Major Component, or to implement a Standard Interface for which an -implementation is available to the public in source code form. A -"Major Component", in this context, means a major essential component -(kernel, window system, and so on) of the specific operating system -(if any) on which the executable work runs, or a compiler used to -produce the work, or an object code interpreter used to run it. - - The "Corresponding Source" for a work in object code form means all -the source code needed to generate, install, and (for an executable -work) run the object code and to modify the work, including scripts to -control those activities. However, it does not include the work's -System Libraries, or general-purpose tools or generally available free -programs which are used unmodified in performing those activities but -which are not part of the work. For example, Corresponding Source -includes interface definition files associated with source files for -the work, and the source code for shared libraries and dynamically -linked subprograms that the work is specifically designed to require, -such as by intimate data communication or control flow between those -subprograms and other parts of the work. - - The Corresponding Source need not include anything that users -can regenerate automatically from other parts of the Corresponding -Source. - - The Corresponding Source for a work in source code form is that -same work. - - 2. Basic Permissions. - - All rights granted under this License are granted for the term of -copyright on the Program, and are irrevocable provided the stated -conditions are met. This License explicitly affirms your unlimited -permission to run the unmodified Program. The output from running a -covered work is covered by this License only if the output, given its -content, constitutes a covered work. This License acknowledges your -rights of fair use or other equivalent, as provided by copyright law. - - You may make, run and propagate covered works that you do not -convey, without conditions so long as your license otherwise remains -in force. You may convey covered works to others for the sole purpose -of having them make modifications exclusively for you, or provide you -with facilities for running those works, provided that you comply with -the terms of this License in conveying all material for which you do -not control copyright. Those thus making or running the covered works -for you must do so exclusively on your behalf, under your direction -and control, on terms that prohibit them from making any copies of -your copyrighted material outside their relationship with you. - - Conveying under any other circumstances is permitted solely under -the conditions stated below. Sublicensing is not allowed; section 10 -makes it unnecessary. - - 3. Protecting Users' Legal Rights From Anti-Circumvention Law. - - No covered work shall be deemed part of an effective technological -measure under any applicable law fulfilling obligations under article -11 of the WIPO copyright treaty adopted on 20 December 1996, or -similar laws prohibiting or restricting circumvention of such -measures. - - When you convey a covered work, you waive any legal power to forbid -circumvention of technological measures to the extent such circumvention -is effected by exercising rights under this License with respect to -the covered work, and you disclaim any intention to limit operation or -modification of the work as a means of enforcing, against the work's -users, your or third parties' legal rights to forbid circumvention of -technological measures. - - 4. Conveying Verbatim Copies. - - You may convey verbatim copies of the Program's source code as you -receive it, in any medium, provided that you conspicuously and -appropriately publish on each copy an appropriate copyright notice; -keep intact all notices stating that this License and any -non-permissive terms added in accord with section 7 apply to the code; -keep intact all notices of the absence of any warranty; and give all -recipients a copy of this License along with the Program. - - You may charge any price or no price for each copy that you convey, -and you may offer support or warranty protection for a fee. - - 5. Conveying Modified Source Versions. - - You may convey a work based on the Program, or the modifications to -produce it from the Program, in the form of source code under the -terms of section 4, provided that you also meet all of these conditions: - - a) The work must carry prominent notices stating that you modified - it, and giving a relevant date. - - b) The work must carry prominent notices stating that it is - released under this License and any conditions added under section - 7. This requirement modifies the requirement in section 4 to - "keep intact all notices". - - c) You must license the entire work, as a whole, under this - License to anyone who comes into possession of a copy. This - License will therefore apply, along with any applicable section 7 - additional terms, to the whole of the work, and all its parts, - regardless of how they are packaged. This License gives no - permission to license the work in any other way, but it does not - invalidate such permission if you have separately received it. - - d) If the work has interactive user interfaces, each must display - Appropriate Legal Notices; however, if the Program has interactive - interfaces that do not display Appropriate Legal Notices, your - work need not make them do so. - - A compilation of a covered work with other separate and independent -works, which are not by their nature extensions of the covered work, -and which are not combined with it such as to form a larger program, -in or on a volume of a storage or distribution medium, is called an -"aggregate" if the compilation and its resulting copyright are not -used to limit the access or legal rights of the compilation's users -beyond what the individual works permit. Inclusion of a covered work -in an aggregate does not cause this License to apply to the other -parts of the aggregate. - - 6. Conveying Non-Source Forms. - - You may convey a covered work in object code form under the terms -of sections 4 and 5, provided that you also convey the -machine-readable Corresponding Source under the terms of this License, -in one of these ways: - - a) Convey the object code in, or embodied in, a physical product - (including a physical distribution medium), accompanied by the - Corresponding Source fixed on a durable physical medium - customarily used for software interchange. - - b) Convey the object code in, or embodied in, a physical product - (including a physical distribution medium), accompanied by a - written offer, valid for at least three years and valid for as - long as you offer spare parts or customer support for that product - model, to give anyone who possesses the object code either (1) a - copy of the Corresponding Source for all the software in the - product that is covered by this License, on a durable physical - medium customarily used for software interchange, for a price no - more than your reasonable cost of physically performing this - conveying of source, or (2) access to copy the - Corresponding Source from a network server at no charge. - - c) Convey individual copies of the object code with a copy of the - written offer to provide the Corresponding Source. This - alternative is allowed only occasionally and noncommercially, and - only if you received the object code with such an offer, in accord - with subsection 6b. - - d) Convey the object code by offering access from a designated - place (gratis or for a charge), and offer equivalent access to the - Corresponding Source in the same way through the same place at no - further charge. You need not require recipients to copy the - Corresponding Source along with the object code. If the place to - copy the object code is a network server, the Corresponding Source - may be on a different server (operated by you or a third party) - that supports equivalent copying facilities, provided you maintain - clear directions next to the object code saying where to find the - Corresponding Source. Regardless of what server hosts the - Corresponding Source, you remain obligated to ensure that it is - available for as long as needed to satisfy these requirements. - - e) Convey the object code using peer-to-peer transmission, provided - you inform other peers where the object code and Corresponding - Source of the work are being offered to the general public at no - charge under subsection 6d. - - A separable portion of the object code, whose source code is excluded -from the Corresponding Source as a System Library, need not be -included in conveying the object code work. - - A "User Product" is either (1) a "consumer product", which means any -tangible personal property which is normally used for personal, family, -or household purposes, or (2) anything designed or sold for incorporation -into a dwelling. In determining whether a product is a consumer product, -doubtful cases shall be resolved in favor of coverage. For a particular -product received by a particular user, "normally used" refers to a -typical or common use of that class of product, regardless of the status -of the particular user or of the way in which the particular user -actually uses, or expects or is expected to use, the product. A product -is a consumer product regardless of whether the product has substantial -commercial, industrial or non-consumer uses, unless such uses represent -the only significant mode of use of the product. - - "Installation Information" for a User Product means any methods, -procedures, authorization keys, or other information required to install -and execute modified versions of a covered work in that User Product from -a modified version of its Corresponding Source. The information must -suffice to ensure that the continued functioning of the modified object -code is in no case prevented or interfered with solely because -modification has been made. - - If you convey an object code work under this section in, or with, or -specifically for use in, a User Product, and the conveying occurs as -part of a transaction in which the right of possession and use of the -User Product is transferred to the recipient in perpetuity or for a -fixed term (regardless of how the transaction is characterized), the -Corresponding Source conveyed under this section must be accompanied -by the Installation Information. But this requirement does not apply -if neither you nor any third party retains the ability to install -modified object code on the User Product (for example, the work has -been installed in ROM). - - The requirement to provide Installation Information does not include a -requirement to continue to provide support service, warranty, or updates -for a work that has been modified or installed by the recipient, or for -the User Product in which it has been modified or installed. Access to a -network may be denied when the modification itself materially and -adversely affects the operation of the network or violates the rules and -protocols for communication across the network. - - Corresponding Source conveyed, and Installation Information provided, -in accord with this section must be in a format that is publicly -documented (and with an implementation available to the public in -source code form), and must require no special password or key for -unpacking, reading or copying. - - 7. Additional Terms. - - "Additional permissions" are terms that supplement the terms of this -License by making exceptions from one or more of its conditions. -Additional permissions that are applicable to the entire Program shall -be treated as though they were included in this License, to the extent -that they are valid under applicable law. If additional permissions -apply only to part of the Program, that part may be used separately -under those permissions, but the entire Program remains governed by -this License without regard to the additional permissions. - - When you convey a copy of a covered work, you may at your option -remove any additional permissions from that copy, or from any part of -it. (Additional permissions may be written to require their own -removal in certain cases when you modify the work.) You may place -additional permissions on material, added by you to a covered work, -for which you have or can give appropriate copyright permission. - - Notwithstanding any other provision of this License, for material you -add to a covered work, you may (if authorized by the copyright holders of -that material) supplement the terms of this License with terms: - - a) Disclaiming warranty or limiting liability differently from the - terms of sections 15 and 16 of this License; or - - b) Requiring preservation of specified reasonable legal notices or - author attributions in that material or in the Appropriate Legal - Notices displayed by works containing it; or - - c) Prohibiting misrepresentation of the origin of that material, or - requiring that modified versions of such material be marked in - reasonable ways as different from the original version; or - - d) Limiting the use for publicity purposes of names of licensors or - authors of the material; or - - e) Declining to grant rights under trademark law for use of some - trade names, trademarks, or service marks; or - - f) Requiring indemnification of licensors and authors of that - material by anyone who conveys the material (or modified versions of - it) with contractual assumptions of liability to the recipient, for - any liability that these contractual assumptions directly impose on - those licensors and authors. - - All other non-permissive additional terms are considered "further -restrictions" within the meaning of section 10. If the Program as you -received it, or any part of it, contains a notice stating that it is -governed by this License along with a term that is a further -restriction, you may remove that term. If a license document contains -a further restriction but permits relicensing or conveying under this -License, you may add to a covered work material governed by the terms -of that license document, provided that the further restriction does -not survive such relicensing or conveying. - - If you add terms to a covered work in accord with this section, you -must place, in the relevant source files, a statement of the -additional terms that apply to those files, or a notice indicating -where to find the applicable terms. - - Additional terms, permissive or non-permissive, may be stated in the -form of a separately written license, or stated as exceptions; -the above requirements apply either way. - - 8. Termination. - - You may not propagate or modify a covered work except as expressly -provided under this License. Any attempt otherwise to propagate or -modify it is void, and will automatically terminate your rights under -this License (including any patent licenses granted under the third -paragraph of section 11). - - However, if you cease all violation of this License, then your -license from a particular copyright holder is reinstated (a) -provisionally, unless and until the copyright holder explicitly and -finally terminates your license, and (b) permanently, if the copyright -holder fails to notify you of the violation by some reasonable means -prior to 60 days after the cessation. - - Moreover, your license from a particular copyright holder is -reinstated permanently if the copyright holder notifies you of the -violation by some reasonable means, this is the first time you have -received notice of violation of this License (for any work) from that -copyright holder, and you cure the violation prior to 30 days after -your receipt of the notice. - - Termination of your rights under this section does not terminate the -licenses of parties who have received copies or rights from you under -this License. If your rights have been terminated and not permanently -reinstated, you do not qualify to receive new licenses for the same -material under section 10. - - 9. Acceptance Not Required for Having Copies. - - You are not required to accept this License in order to receive or -run a copy of the Program. Ancillary propagation of a covered work -occurring solely as a consequence of using peer-to-peer transmission -to receive a copy likewise does not require acceptance. However, -nothing other than this License grants you permission to propagate or -modify any covered work. These actions infringe copyright if you do -not accept this License. Therefore, by modifying or propagating a -covered work, you indicate your acceptance of this License to do so. - - 10. Automatic Licensing of Downstream Recipients. - - Each time you convey a covered work, the recipient automatically -receives a license from the original licensors, to run, modify and -propagate that work, subject to this License. You are not responsible -for enforcing compliance by third parties with this License. - - An "entity transaction" is a transaction transferring control of an -organization, or substantially all assets of one, or subdividing an -organization, or merging organizations. If propagation of a covered -work results from an entity transaction, each party to that -transaction who receives a copy of the work also receives whatever -licenses to the work the party's predecessor in interest had or could -give under the previous paragraph, plus a right to possession of the -Corresponding Source of the work from the predecessor in interest, if -the predecessor has it or can get it with reasonable efforts. - - You may not impose any further restrictions on the exercise of the -rights granted or affirmed under this License. For example, you may -not impose a license fee, royalty, or other charge for exercise of -rights granted under this License, and you may not initiate litigation -(including a cross-claim or counterclaim in a lawsuit) alleging that -any patent claim is infringed by making, using, selling, offering for -sale, or importing the Program or any portion of it. - - 11. Patents. - - A "contributor" is a copyright holder who authorizes use under this -License of the Program or a work on which the Program is based. The -work thus licensed is called the contributor's "contributor version". - - A contributor's "essential patent claims" are all patent claims -owned or controlled by the contributor, whether already acquired or -hereafter acquired, that would be infringed by some manner, permitted -by this License, of making, using, or selling its contributor version, -but do not include claims that would be infringed only as a -consequence of further modification of the contributor version. For -purposes of this definition, "control" includes the right to grant -patent sublicenses in a manner consistent with the requirements of -this License. - - Each contributor grants you a non-exclusive, worldwide, royalty-free -patent license under the contributor's essential patent claims, to -make, use, sell, offer for sale, import and otherwise run, modify and -propagate the contents of its contributor version. - - In the following three paragraphs, a "patent license" is any express -agreement or commitment, however denominated, not to enforce a patent -(such as an express permission to practice a patent or covenant not to -sue for patent infringement). To "grant" such a patent license to a -party means to make such an agreement or commitment not to enforce a -patent against the party. - - If you convey a covered work, knowingly relying on a patent license, -and the Corresponding Source of the work is not available for anyone -to copy, free of charge and under the terms of this License, through a -publicly available network server or other readily accessible means, -then you must either (1) cause the Corresponding Source to be so -available, or (2) arrange to deprive yourself of the benefit of the -patent license for this particular work, or (3) arrange, in a manner -consistent with the requirements of this License, to extend the patent -license to downstream recipients. "Knowingly relying" means you have -actual knowledge that, but for the patent license, your conveying the -covered work in a country, or your recipient's use of the covered work -in a country, would infringe one or more identifiable patents in that -country that you have reason to believe are valid. - - If, pursuant to or in connection with a single transaction or -arrangement, you convey, or propagate by procuring conveyance of, a -covered work, and grant a patent license to some of the parties -receiving the covered work authorizing them to use, propagate, modify -or convey a specific copy of the covered work, then the patent license -you grant is automatically extended to all recipients of the covered -work and works based on it. - - A patent license is "discriminatory" if it does not include within -the scope of its coverage, prohibits the exercise of, or is -conditioned on the non-exercise of one or more of the rights that are -specifically granted under this License. You may not convey a covered -work if you are a party to an arrangement with a third party that is -in the business of distributing software, under which you make payment -to the third party based on the extent of your activity of conveying -the work, and under which the third party grants, to any of the -parties who would receive the covered work from you, a discriminatory -patent license (a) in connection with copies of the covered work -conveyed by you (or copies made from those copies), or (b) primarily -for and in connection with specific products or compilations that -contain the covered work, unless you entered into that arrangement, -or that patent license was granted, prior to 28 March 2007. - - Nothing in this License shall be construed as excluding or limiting -any implied license or other defenses to infringement that may -otherwise be available to you under applicable patent law. - - 12. No Surrender of Others' Freedom. - - If conditions are imposed on you (whether by court order, agreement or -otherwise) that contradict the conditions of this License, they do not -excuse you from the conditions of this License. If you cannot convey a -covered work so as to satisfy simultaneously your obligations under this -License and any other pertinent obligations, then as a consequence you may -not convey it at all. For example, if you agree to terms that obligate you -to collect a royalty for further conveying from those to whom you convey -the Program, the only way you could satisfy both those terms and this -License would be to refrain entirely from conveying the Program. - - 13. Use with the GNU Affero General Public License. - - Notwithstanding any other provision of this License, you have -permission to link or combine any covered work with a work licensed -under version 3 of the GNU Affero General Public License into a single -combined work, and to convey the resulting work. The terms of this -License will continue to apply to the part which is the covered work, -but the special requirements of the GNU Affero General Public License, -section 13, concerning interaction through a network will apply to the -combination as such. - - 14. Revised Versions of this License. - - The Free Software Foundation may publish revised and/or new versions of -the GNU General Public License from time to time. Such new versions will -be similar in spirit to the present version, but may differ in detail to -address new problems or concerns. - - Each version is given a distinguishing version number. If the -Program specifies that a certain numbered version of the GNU General -Public License "or any later version" applies to it, you have the -option of following the terms and conditions either of that numbered -version or of any later version published by the Free Software -Foundation. If the Program does not specify a version number of the -GNU General Public License, you may choose any version ever published -by the Free Software Foundation. - - If the Program specifies that a proxy can decide which future -versions of the GNU General Public License can be used, that proxy's -public statement of acceptance of a version permanently authorizes you -to choose that version for the Program. - - Later license versions may give you additional or different -permissions. However, no additional obligations are imposed on any -author or copyright holder as a result of your choosing to follow a -later version. - - 15. Disclaimer of Warranty. - - THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY -APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT -HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY -OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, -THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR -PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM -IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF -ALL NECESSARY SERVICING, REPAIR OR CORRECTION. - - 16. Limitation of Liability. - - IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING -WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS -THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY -GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE -USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF -DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD -PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), -EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF -SUCH DAMAGES. - - 17. Interpretation of Sections 15 and 16. - - If the disclaimer of warranty and limitation of liability provided -above cannot be given local legal effect according to their terms, -reviewing courts shall apply local law that most closely approximates -an absolute waiver of all civil liability in connection with the -Program, unless a warranty or assumption of liability accompanies a -copy of the Program in return for a fee. - - END OF TERMS AND CONDITIONS - - How to Apply These Terms to Your New Programs - - If you develop a new program, and you want it to be of the greatest -possible use to the public, the best way to achieve this is to make it -free software which everyone can redistribute and change under these terms. - - To do so, attach the following notices to the program. It is safest -to attach them to the start of each source file to most effectively -state the exclusion of warranty; and each file should have at least -the "copyright" line and a pointer to where the full notice is found. - - - Copyright (C) - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . - -Also add information on how to contact you by electronic and paper mail. - - If the program does terminal interaction, make it output a short -notice like this when it starts in an interactive mode: - - Copyright (C) - This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. - This is free software, and you are welcome to redistribute it - under certain conditions; type `show c' for details. - -The hypothetical commands `show w' and `show c' should show the appropriate -parts of the General Public License. Of course, your program's commands -might be different; for a GUI interface, you would use an "about box". - - You should also get your employer (if you work as a programmer) or school, -if any, to sign a "copyright disclaimer" for the program, if necessary. -For more information on this, and how to apply and follow the GNU GPL, see -. - - The GNU General Public License does not permit incorporating your program -into proprietary programs. If your program is a subroutine library, you -may consider it more useful to permit linking proprietary applications with -the library. If this is what you want to do, use the GNU Lesser General -Public License instead of this License. But first, please read -. diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/README.md b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/README.md deleted file mode 100644 index 947a2224d2e..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/README.md +++ /dev/null @@ -1,67 +0,0 @@ -# qdt -A C++11 header-based library for quadrature integrators. - -This library provides a set of quadrature methods for numerically integrating one-dimensional functions (that return real or vector values). This code is part of my research on rendering participating media, which has been published as the paper [Higher Order Ray Marching](http://giga.cps.unizar.es/~amunoz/projects/CGF2014_higherorder/). - -It supports the following quadrature methods: - -Method | Code -----------------------------------------------------------------|----------------------------------------------------- -[Rectangle rule](http://en.wikipedia.org/wiki/Rectangle_method) | `qdt::rectangle()` -[Trapezoid rule](http://en.wikipedia.org/wiki/Trapezoidal_rule) | `qdt::trapezoid()` -[Simpson rule](http://en.wikipedia.org/wiki/Simpson%27s_rule) | `qdt::simpson()` -[Monte Carlo](http://en.wikipedia.org/wiki/Monte_Carlo_integration) | `qdt::monte_carlo(/*Number of samples*/)` -[Romberg](http://en.wikipedia.org/wiki/Romberg%27s_method) | `qdt::romberg(/*Order*/)` -[Gauss-Kronrod](http://en.wikipedia.org/wiki/Gauss%E2%80%93Kronrod_quadrature_formula) (order 7-15) | `qdt::gauss_kronrod()` - -Each of those quadrature rules can be applied in different ways: - -Usage | Code -----------------------------------------------------------------|----------------------------------------------------- -As is | None -Subdividing the integration range into steps | `qdt::constant_step(/*Number of steps*/, /*Base method*/)` -Adaptatively with a specified tolerance parameter | `qdt::adaptive(/*Tolerance*/, /*Base method*/)` - -Note that either the stepping or the adaptativeness of methods is reasonably smart. For instance, the stepping of the Trapezoid rule does not calculate twice the same points for two consecutive steps. For adaptive methods similar rules are followed: adaptive Simpson´s rule does not re-evaluate the function in previously evaluated points. Furthermore, nested methods (Gauss-Kronrod) take advantage of such (do not reevaluate points for calculating the error) and multi-order methods (Romberg) increase their order for getting more accurate (instead of subdividing the step). - - -## Compiling the example -There is an [example](https://github.com/adolfomunoz/qdt/blob/master/examples/qdt-pi.cc) that computes number pi as the arc tangent integral over the infinite range using different methods. It can be compiled with any C++11 compiler. We also provide a very simple `CMakeLists.txt` for CMake lovers. - -You can clone the repository and compile the example (git + gcc) as follows: -``` -git clone https://github.com/adolfomunoz/qdt.git -g++ --std=c++11 qdt/examples/qdt-pi.cc -o qdt-pi -``` - -## Usage -You could first get inspired by the [example](https://github.com/adolfomunoz/qdt/blob/master/examples/qdt-pi.cc). Foryour code, you only need to include [qdt.h](https://github.com/adolfomunoz/qdt/blob/master/qdt.h). I pesonally have this in a `qdt` directory, so I just type `#include ` and that's it. There is no linkage, this is a header-only library. - -All quadrature algorithms have a method `integrate` that integrates the provided function along the specified range. For instance: - -``` -auto method = qdt::constant_step(10, qdt::trapezoid()); -std::cout << method.integrate([] (float x) { return x*x; }, -1.0f, 1.0f)< -class Method -{ -public: - static const bool uses_middle = false; - static const bool uses_boundaries = false; - static const bool is_nested = false; - static const bool is_multi_level = false; - - - - template - auto integrate(const Function& f, real a, real b) const -> decltype(f(a)) - { - return static_cast(*this).integrate_real(f,a,b); - } - - template - auto integrate_from_minus_inf(const Function& f, real b) const -> decltype(f(b)) - { - return static_cast(*this).integrate_real - ([&](real t){ return (real(1.0)/(t*t))*f(b - (real(1.0) - t)/t); }, - real(0.0), real(1.0)); - } - - template - auto integrate_to_inf(const Function& f, real a) const -> decltype(f(a)) - { - return static_cast(*this).integrate_real - ([&](real t){return (real(1.0)/((real(1.0) - t)*(real(1.0) - t)))*f(a + (t/(real(1.0) - t)));}, - real(0.0), real(1.0)); - } - - template - auto integrate_from_minus_inf_to_inf(const Function& f) const -> decltype(f(real(0.0))) - { - return static_cast(*this).integrate_real - ([&](real t){real t2=t*t; return ((real(1.0)+t2)/((real(1.0) - t2)*(real(1.0) - t2)))*f(t/(real(1.0) - t2));}, - real(-1.0), real(1.0)); - } - - template - auto integrate(const Function& f, const MinusInfinity& a, real b) const -> decltype(f(b)) - { - return static_cast(*this).integrate_from_minus_inf(f,b); - } - - template - auto integrate(const Function& f, real a, const Infinity& b) const -> decltype(f(a)) - { - return static_cast(*this).integrate_to_inf(f,a); - } - - template - auto integrate(const Function& f, const MinusInfinity& a, const Infinity& b) const -> decltype(f(real(0.0))) - { - return static_cast(*this).integrate_from_minus_inf_to_inf(f); - } - - template - auto integrate_change_of_variable(const Function& f, real a, real b, - const VChange& vc, const InvVChange& vc_inv, const DVChange& dvc) const -> decltype(f(vc(a))) - { - return integrate( - [&] (real t) { return f(vc(t))*dvc(t); }, - vc_inv(a),vc_inv(b)); - } -}; - -template -class MethodWithBoundaries : public Method > -{ -public: - static const bool uses_boundaries = true; - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - return static_cast(*this).integrate_with_boundaries(f,a,f(a),b,f(b)); - } - - template - auto integrate_from_minus_inf(const Function& f, real b) const -> decltype(f(b)) - { - return static_cast(*this).integrate_with_boundaries - ([&](real t){ return (real(1.0)/(t*t))*f(b - (real(1.0) - t)/t); }, - real(0.0), f(b - real(1.e5))*real(1.e10), - real(1.0), f(b)); - } - - template - auto integrate_to_inf(const Function& f, real a) const -> decltype(f(a)) - { - return static_cast(*this).integrate_with_boundaries - ([&](real t){return (real(1.0)/((real(1.0) - t)*(real(1.0) - t)))*f(a + (t/(real(1.0) - t)));}, - real(0.0), f(a), - real(1.0), f(a + real(1.e5))*real(1.e10)); - } - - template - auto integrate_from_minus_inf_to_inf(const Function& f) const -> decltype(f(real(0.0))) - { - return static_cast(*this).integrate_with_boundaries - ([&](real t){ real t2=t*t; return ((real(1.0)+t2)/((real(1.0) - t2)*(real(1.0) - t2)))*f(t/(real(1.0) - t2));}, - real(-1.0), real(0.5*1.e10)*f(-1.e5), - real(1.0), real(0.5*1.e10)*f(1.e5)); - } -}; - -template -class MethodWithBoundariesMiddle : public MethodWithBoundaries > -{ -public: - static const bool uses_middle = true; - - template - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - return static_cast(*this).integrate_with_boundaries_and_middle(f,a,fa,b,fb,f(0.5*(a+b))); - } -}; - -template -class MethodNested : public Method > -{ -public: - static const bool is_nested = true; - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - decltype(f(a)) that; - return static_cast(*this).integrate_nested(f,a,b,that); - } -}; - -template -class MethodMultiLevel : public Method > -{ -public: - static const bool is_multi_level = true; - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - auto data = static_cast(*this).first_level_data(f,a,b); - decltype(f(a)) sol = static_cast(*this).integrate_level(f,a,b,data); - while(!static_cast(*this).max_level(data)) - { - sol = static_cast(*this).integrate_level(f,a,b,data); - } - return sol; - } -}; - -template -class MethodMultiLevelWithBoundaries : public MethodWithBoundaries > -{ -public: - static const bool is_multi_level = true; - - template - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - auto data = static_cast(*this).first_level_data(f,a,fa,b,fb); - decltype(f(a)) sol = static_cast(*this).integrate_level(f,a,fa,b,fb,data); - while(!static_cast(*this).max_level(data)) - { - sol = static_cast(*this).integrate_level(f,a,fa,b,fb,data); - } - return sol; - } -}; - -template -class MethodMultiLevelWithBoundariesMiddle : public MethodWithBoundariesMiddle > -{ -public: - static const bool is_multi_level = true; - - template - YType integrate_with_boundaries_and_middle(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& fmiddle) const - { - auto data = static_cast(*this).first_level_data(f,a,fa,b,fb,fmiddle); - decltype(f(a)) sol = static_cast(*this).integrate_level(f,a,fa,b,fb,fmiddle,data); - while(!static_cast(*this).max_level(data)) - { - sol = static_cast(*this).integrate_level(f,a,fa,b,fb,fmiddle,data); - } - return sol; - } -}; - - - -}; - -#endif - diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/examples/qdt-pi.cc b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/examples/qdt-pi.cc deleted file mode 100644 index 92e6c63e296..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/examples/qdt-pi.cc +++ /dev/null @@ -1,28 +0,0 @@ -#include -#include -#include "../qdt.h" - -int main(int argc, char** argv) { - std::cout<<"Calculating PI as an integral over an infinite range"< -#include -#include -#include -#include -#include - -namespace qdt { - -template -std::ostream& operator<<(std::ostream& os, const MonteCarlo& method) -{ os<<"MonteCarlo-"< -std::ostream& operator<<(std::ostream& os, const Adaptive& method) -{ os<<"Adapt-"< -std::ostream& operator<<(std::ostream& os, const ConstantStep& method) -{ os<<"Steps-"< -std::string traits_string(const M& method) -{ - std::stringstream ss; - ss<<(M::uses_boundaries?"[B]":" "); - ss<<(M::uses_middle?"[M]":" "); - ss<<(M::is_nested?"[N]":" "); - ss<<(M::is_multi_level?"[L]":" "); - return ss.str(); -} - -template -class TestResult -{ - float time; - unsigned int nevals; - T sol, approx; -public: - TestResult(float _time, unsigned int _nevals, const T& _sol, const T& _approx) : - time(_time), nevals(_nevals), sol(_sol), approx(_approx) { } - - float absolute_error() const { return fabs(sol - approx); } - float relative_error() const { return fabs(sol - approx)/std::max(fabs(sol),fabs(approx)); } - float seconds() const { return time; } - unsigned int evaluations() const { return nevals; } - T solution() const { return sol; } - T approximation() const { return approx; } -}; - -template -std::ostream& operator<<(std::ostream& os, const TestResult& tr) -{ - os< -class EvaluationCounter -{ - F f; - mutable long int n; -public: - EvaluationCounter(const F& _f) : f(_f), n(0) { } - float operator()(float t) const { n++; return f(t); } - long int count() const { return n; } -}; - -std::ostream& operator<<(std::ostream& os,const MinusInfinity& t) { os<<"-inf"; return os; } -std::ostream& operator<<(std::ostream& os,const Infinity& t) { os<<"inf"; return os; } - -template -TestResult test(const Function& f, const R& sol, const Method& m, const A& a, const B& b) -{ - EvaluationCounter fc(f); - unsigned int n; - R approx = m.integrate(fc,a,b); - - std::chrono::time_point start = std::chrono::system_clock::now(); - m.integrate(f,a,b); n=1; - std::chrono::duration elapsed_seconds = std::chrono::system_clock::now() - start; - while (elapsed_seconds.count()<=0.01) - { - m.integrate(f,a,b); n++; - elapsed_seconds = std::chrono::system_clock::now() - start; - } - return TestResult(elapsed_seconds.count()/float(n),fc.count(),sol,approx); -} - -template -class EvaluationSamples -{ - F f; - mutable std::list _samples; -public: - EvaluationSamples(const F& _f) : f(_f) { } - float operator()(real t) const { _samples.push_back(t); return f(t); } - std::list samples() const { return _samples; } -}; - -template -std::list sampled_values(const Function& f, const Method& m, const A& a, const B& b) -{ - EvaluationSamples fs(f); - m.integrate(fs,a,b); - std::list samples = fs.samples(); - return samples; -} - - -} - -#endif - diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/adaptive.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/adaptive.h deleted file mode 100644 index 282161ebef3..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/adaptive.h +++ /dev/null @@ -1,395 +0,0 @@ -#ifndef _QDT_ADAPTIVE_H_ -#define _QDT_ADAPTIVE_H_ - -#include "../common/method.h" -#include -#include - -namespace qdt -{ - -class ErrorEstimator -{ -public: - template - T estimate_error(const T& e1, const T& e2); - -private: - template - typename T::value_type estimate_error_vector(const T& e1, const U& e2) const - { -// typename std::remove_const::type>::type sol -// = decltype(*(e1.begin()))(0.0); - auto sol = (typename T::value_type)(0.0); - auto i1 = e1.begin(); - auto i2 = e2.begin(); - for(; (i1 != e1.end()) && (i2!=e2.end());i1++, i2++) - { - auto err = estimate_error(*i1, *i2); - if (err>sol) sol = err; - } - return sol; - } - - template - T estimate_error_scalar(const T& e1, const T& e2) const - { - if (std::max(std::abs(e1),std::abs(e2)) < 1.e-6) return std::abs(e1 - e2); - else return std::abs(e1-e2)/std::max(std::abs(e1),std::abs(e2)); - } - -public: - template - auto estimate_error(const T& e1, const U& e2) const -> decltype(estimate_error_vector(e1,e2)) - { return estimate_error_vector(e1,e2); }; - - int estimate_error(int e1, int e2) const { return estimate_error_scalar(e1,e2); } - float estimate_error(float e1, float e2) const { return estimate_error_scalar(e1,e2); } - double estimate_error(double e1, double e2) const { return estimate_error_scalar(e1,e2); } -}; - -template -class Adaptive : public Method > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - YType integrate_real_tol(const Function& f, const real& a, const real& b, const YType& whole, const real& tol) const - { - if (std::abs(a-b) - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - return integrate_real_tol(f,a,b,base_method().integrate_real(f,a,b),real(tolerance())); - } -}; - -template -class Adaptive : - public MethodWithBoundaries > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - YType integrate_with_boundaries_tol(const Function& f, real a, const YType& fa, real b, const YType& fb, - const YType& whole, const real& tol) const - { - if (std::abs(a-b) - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - return integrate_with_boundaries_tol(f,a,fa,b,fb,base_method().integrate_with_boundaries(f,a,fa,b,fb),real(tolerance())); - } -}; - -template -class Adaptive : - public MethodWithBoundariesMiddle > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - YType integrate_with_boundaries_and_middle_tol(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& f_half, - const YType& whole, const real& tol) const - { - if (std::abs(a-b) - YType integrate_with_boundaries_and_middle(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& f_half) const - { - return integrate_with_boundaries_and_middle_tol(f,a,fa,b,fb,f_half, - base_method().integrate_with_boundaries_and_middle(f,a,fa,b,fb,f_half),real(tolerance())); - } -}; - - -template -class Adaptive : public Method > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - auto integrate_real_tol(const Function& f, const real& a, const real& b, const real& tol) const -> decltype(f(a)) - { - decltype(f(a)) prev; - decltype(f(a)) sol = base_method().integrate_nested(f,a,b,prev); - real error = estimator.estimate_error(sol, prev); - - if ((std::abs(a-b) - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - return integrate_real_tol(f,a,b,real(tolerance())); - } -}; - -template -class Adaptive : public Method > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - auto integrate_real_tol(const Function& f, const real& a, const real& b, const real& tol) const -> decltype(f(a)) - { - auto data = base_method().first_level_data(f,a,b); - decltype(f(a)) sol = base_method().integrate_level(f,a,b,data); - real error = tol + real(1.0); - while ( (!base_method().max_level(data)) && (error >= tol) ) - { - decltype(f(a)) prev = sol; - sol = base_method().integrate_level(f,a,b,data); - error = estimator.estimate_error(sol, prev); - } - - if ((std::abs(a-b) - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - return integrate_real_tol(f,a,b,real(tolerance())); - } -}; - -template -class Adaptive : - public MethodWithBoundaries > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - YType integrate_with_boundaries_tol(const Function& f, - const real& a, const YType& fa, const real& b, const YType& fb, const real& tol) const - { - auto data = base_method().first_level_data(f,a,fa,b,fb); - decltype(f(a)) sol = base_method().integrate_level(f,a,fa,b,fb,data); - real error = tol + real(1.0); - while ( (!base_method().max_level(data)) && (error >= tol) ) - { - decltype(f(a)) prev = sol; - sol = base_method().integrate_level(f,a,fa,b,fb,data); - error = estimator.estimate_error(sol, prev); - } - - if ((std::abs(a-b) - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - return integrate_with_boundaries_tol(f,a,fa,b,fb,real(tolerance())); - } -}; - -template -class Adaptive : - public MethodWithBoundariesMiddle > -{ - BaseMethod _base_method; - Estimator estimator; - float _tolerance; - float min_step; -public: - Adaptive(float tol = 1.e-5, float _min_step = 1.e-5): _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), _tolerance(tol), min_step(_min_step) { } - Adaptive(const BaseMethod& base_method, const Estimator& _estimator, float tol = 1.e-5, float _min_step = 1.e-5): - _base_method(base_method), estimator(_estimator), _tolerance(tol), min_step(_min_step) { } - - const BaseMethod& base_method() const { return _base_method; } - float tolerance() const { return _tolerance; } - -private: - template - YType integrate_with_boundaries_and_middle_tol(const Function& f, - const real& a, const YType& fa, const real& b, const YType& fb, const YType& fmiddle, const real& tol) const - { - auto data = base_method().first_level_data(f,a,fa,b,fb,fmiddle); - decltype(f(a)) sol = base_method().integrate_level(f,a,fa,b,fb,fmiddle,data); - real error = tol + real(1.0); - while ( (!base_method().max_level(data)) && (error >= tol) ) - { - decltype(f(a)) prev = sol; - sol = base_method().integrate_level(f,a,fa,b,fb,fmiddle,data); - error = estimator.estimate_error(sol, prev); - } - - if ((std::abs(a-b) - YType integrate_with_boundaries_and_middle(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& fmiddle) const - { - return integrate_with_boundaries_and_middle_tol(f,a,fa,b,fb,fmiddle,real(tolerance())); - } -}; - -template -Adaptive adaptive(const BM& bm) -{ - return Adaptive(bm, 1.e-5, 1.e-10); -} - -template -Adaptive adaptive(float tol, const BM& bm) -{ - return Adaptive(bm, tol, 1.e-10); -} - - - - - - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/constant-step.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/constant-step.h deleted file mode 100644 index 88c5cd2fdc9..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/constant-step.h +++ /dev/null @@ -1,102 +0,0 @@ -#ifndef _QDT_CONSTANT_STEP_H_ -#define _QDT_CONSTANT_STEP_H_ - -#include "../common/method.h" - -namespace qdt -{ - -template -class ConstantStep : public Method > -{ - float st; - unsigned int nsteps; - BaseMethod m; -public: - ConstantStep(const BaseMethod& _m = BaseMethod()): st(1.0f), nsteps(1), m(_m) { } - ConstantStep(float s, const BaseMethod& _m) : st(s),nsteps(0), m(_m) { } - ConstantStep(int ns, const BaseMethod& _m) : st(1.0f),nsteps(ns), m(_m) { } - - template - ConstantStep(float s, Params... constructor_parameters) : st(s),nsteps(0), m(constructor_parameters...) { } - template - ConstantStep(int ns, Params... constructor_parameters) : st(1.0f),nsteps(ns), m(constructor_parameters...) { } - - const BaseMethod& base_method() const { return m; } - int steps() const { return nsteps>0?nsteps:int(1.0f/st); } - - template - real step(const real& total) const { return (nsteps>0?(total/real(nsteps)):real(st)); } - - template - static bool stop(real x, real b, real h) { return h>0?x<(b-0.5*h):x>(b+0.5*h); } - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - real h = step(b-a); - real x_prev = a; - real x = a+h; - decltype(f(a)) sol(0.0); - - for (;stop(x,b,h);x+=h, x_prev+=h) sol+=m.integrate(f,x_prev,x); - sol+=m.integrate(f,x_prev,b); - - return sol; - } -}; - - -template -class ConstantStep : public MethodWithBoundaries > -{ - float st; - unsigned int nsteps; - BaseMethod m; -public: - ConstantStep(const BaseMethod& _m = BaseMethod()): st(-1.0), nsteps(1), m(_m) { } - ConstantStep(float s, const BaseMethod& _m = BaseMethod()) : st(s),nsteps(0), m(_m) { } - ConstantStep(int ns, const BaseMethod& _m = BaseMethod()) : st(-1.0),nsteps(ns), m(_m) { } - - template - ConstantStep(float s, Params... constructor_parameters) : st(s),nsteps(0), m(constructor_parameters...) { } - template - ConstantStep(int ns, Params... constructor_parameters) : st(1.0f),nsteps(ns), m(constructor_parameters...) { } - - const BaseMethod& base_method() const { return m; } - int steps() const { return nsteps>0?nsteps:int(1.0f/st); } - - template - real step(const real& total) const { return (nsteps>0?(total/real(nsteps)):real(st)); } - - template - static bool stop(real x, real b, real h) { return h>0?x<(b-0.5*h):x>(b+0.5*h); } - - template - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - real h = step(b-a); - real x_prev = a; - YType fx_prev = fa; - real x = a+h; - YType sol(0.0); - - for (;stop(x,b,h);x+=h, x_prev+=h) - { - YType fx = f(x); - sol+=m.integrate_with_boundaries(f,x_prev,fx_prev,x,fx); - fx_prev=fx; - } - sol+=m.integrate_with_boundaries(f,x_prev,fx_prev,b,fb); - return sol; - } -}; - -template -ConstantStep constant_step(int nsteps, const BM& bm) -{ return ConstantStep(nsteps,bm); } - - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/gauss-kronrod.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/gauss-kronrod.h deleted file mode 100644 index cd006f3b019..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/gauss-kronrod.h +++ /dev/null @@ -1,110 +0,0 @@ -#ifndef _QDT_GAUSS_KRONROD_H_ -#define _QDT_GAUSS_KRONROD_H_ - -#include -#include -#include "../common/method.h" - -namespace qdt -{ - -class GaussKronrod : public MethodNested -{ -private: - template - YType integrate_gauss_kronrod(const Function& f, YType& gauss) const - { - YType kronrod(0.0); gauss = YType(0.0); - auto fi = f(0.0); - gauss += 0.417959183673469*fi; kronrod += 0.209482141084728*fi; - fi = f(0.405845151377397) + f(-0.405845151377397); - gauss += 0.381830050505119*fi; kronrod += 0.190350578064785*fi; - fi = f(0.741531185599394) + f(-0.741531185599394); - gauss += 0.279705391489277*fi; kronrod += 0.140653259715525*fi; - fi = f(0.949107912342759) + f(-0.949107912342759); - gauss += 0.129484966168870*fi; kronrod += 0.063092092629979*fi; - - kronrod += 0.204432940075298*( f(0.207784955007898) + f(-0.207784955007898) ) - + 0.169004726639267*( f(0.586087235467691) + f(-0.586087235467691) ) - + 0.104790010322250*( f(0.864864423359769) + f(-0.864864423359769) ) - + 0.022935322010529*( f(0.991455371120813) + f(-0.991455371120813) ); - - return kronrod; - } - -public: - template - YType integrate_nested(const Function& f, real a, real b, YType& prev) const - { - return integrate_gauss_kronrod( - [&](real t) {return real(0.5)*(b-a)*f( (t + real(1.0))*real(0.5)*(b-a) + a); }, prev); - } - -}; - -class GaussKronrodML : public MethodMultiLevel -{ -public: - GaussKronrodML() { } - -private: - - template - YType integrate_gauss_kronrod(const Function& f, std::vector& data) const - { - //If we are in the gauss pass - if (data.size()==1) - { - data.resize(4); - data[0] = f(0.0); - data[1] = f(0.405845151377397) + f(-0.405845151377397); - data[2] = f(0.741531185599394) + f(-0.741531185599394); - data[3] = f(0.949107912342759) + f(-0.949107912342759); - return 0.417959183673469*data[0] - + 0.381830050505119*data[1] - + 0.279705391489277*data[2] - + 0.129484966168870*data[3]; - - } //We have the date from the gauss pass, compute Kronrod - else if (data.size()==4) - { - YType kronrod = 0.204432940075298*( f(0.207784955007898) + f(-0.207784955007898) ) - + 0.169004726639267*( f(0.586087235467691) + f(-0.586087235467691) ) - + 0.104790010322250*( f(0.864864423359769) + f(-0.864864423359769) ) - + 0.022935322010529*( f(0.991455371120813) + f(-0.991455371120813) ) - + 0.209482141084728*data[0] - + 0.190350578064785*data[1] - + 0.140653259715525*data[2] - + 0.063092092629979*data[3]; - data.clear(); - return kronrod; - } else return YType(0.0); - } - -public: - template - auto first_level_data(const Function& f, real a, real b) const -> std::vector - { - //We mark the start (gauss pass) with a vector of size 1 (could be anything else, but this makes sense and is easier to do - return std::vector(1); - } - - template - bool max_level(const std::vector& data) const { return data.empty(); } - - - //This nested method just has two levels - template - YType integrate_level(const Function& f, real a, real b, std::vector& data) const - { - return integrate_gauss_kronrod( - [&](real t) {return real(0.5)*(b-a)*f( (t + real(1.0))*real(0.5)*(b-a) + a); }, data); - } -}; - -GaussKronrod gauss_kronrod() -{ return GaussKronrod(); } - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/monte-carlo.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/monte-carlo.h deleted file mode 100644 index 48f1298c877..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/monte-carlo.h +++ /dev/null @@ -1,41 +0,0 @@ -#ifndef _QDT_MONTE_CARLO_H_ -#define _QDT_MONTE_CARLO_H_ - -#include -#include -#include "../common/method.h" - -namespace qdt -{ - -template -class MonteCarlo : public Method > -{ - unsigned int nsamples_; - mutable RNG rng; -public: - MonteCarlo(unsigned int nsamples = 1, - typename RNG::result_type seed = std::random_device()()): nsamples_(nsamples), rng(seed) { } - - unsigned int nsamples() const { return nsamples_; } - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - std::uniform_real_distribution dist(a, b); - decltype(f(a)) sol = f(dist(rng)); - for (unsigned int i = 1;i monte_carlo(unsigned int nsamples = 1) -{ - return MonteCarlo<>(nsamples); -} - - - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/rectangle.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/rectangle.h deleted file mode 100644 index 38753848de8..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/rectangle.h +++ /dev/null @@ -1,26 +0,0 @@ -#ifndef _QDT_RECTANGLE_H_ -#define _QDT_RECTANGLE_H_ - -#include "../common/method.h" - -namespace qdt -{ - -class Rectangle : public Method -{ -public: - Rectangle() { } - - template - auto integrate_real(const Function& f, real a, real b) const -> decltype(f(a)) - { - return (b-a)*f(0.5*(b+a)); - } -}; - -Rectangle rectangle() -{ return Rectangle(); } - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/romberg.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/romberg.h deleted file mode 100644 index faac7f8a0c0..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/romberg.h +++ /dev/null @@ -1,87 +0,0 @@ -#ifndef _QDT_ROMBERG_H_ -#define _QDT_ROMBERG_H_ - -#include -#include -#include "../common/method.h" - -namespace qdt -{ - -class Romberg : public MethodMultiLevelWithBoundariesMiddle -{ - unsigned int max_levels; -public: - /** - * \brief Constructs the Romberg numerical method. - * - * \param _max_levels is the maximum level of subdivision of the method, which corresponds to the maximum order of the method. - * This parameter should be smaller than 16 in order to avoid extreme subdivision and numerical errors. - */ - Romberg(unsigned int _max_levels = 16) : max_levels(_max_levels) { } - - //We reserve memory in a vector for all the levels, but we include a number which indicates in which level we are. - template - class LevelData - { - friend class Romberg; - unsigned int level; - real hn; unsigned int twonminus1; - std::vector Rm; - - LevelData(unsigned int max_levels, real a, real b) : level(0), hn(b-a), twonminus1(1),Rm(max_levels) { } - void add_level() { level++; hn*=0.5; if (level>1) twonminus1*=2; } - YType return_value() { return Rm[level-1]; } - }; - -public: - template - LevelData first_level_data(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& fmiddle) const - { - return LevelData(this->max_levels, a, b); - } - - template - bool max_level(const LevelData& data) const { return data.level >= this->max_levels; } - - unsigned int levels() const { return max_levels; } - - //This nested method just has two levels - template - YType integrate_level(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& fmiddle, LevelData& data) const - { - if (data.level == 0) - data.Rm[0] = 0.5*(b-a)*(fa + fb); - else - { - YType prev = data.Rm[0]; - if (data.level == 1) - data.Rm[0] = 0.5*data.Rm[0] + data.hn*fmiddle; - else - { - YType sum(0.0); - for (unsigned int k = 1; k<=data.twonminus1; k++) sum+=f(a + real(2*k - 1)*data.hn); - data.Rm[0] = 0.5*data.Rm[0] + data.hn*sum; - } - for (unsigned int m = 1, fourm = 4; m<=data.level; m++, fourm*=4) - { - YType actual = data.Rm[m]; - data.Rm[m] = (real(1.0)/real(fourm - 1))*(real(fourm)*data.Rm[m-1] - prev); - prev = actual; - } - } - - data.add_level(); - return data.return_value(); - } -}; - - -Romberg romberg(unsigned int order) -{ return Romberg(order); } - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/simpson.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/simpson.h deleted file mode 100644 index 68027976039..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/simpson.h +++ /dev/null @@ -1,27 +0,0 @@ -#ifndef _QDT_SIMPSON_H_ -#define _QDT_SIMPSON_H_ - -#include "../common/method.h" - -namespace qdt -{ - -class Simpson: public MethodWithBoundariesMiddle -{ -public: - Simpson() { } - - template - YType integrate_with_boundaries_and_middle(const Function& f, - real a, const YType& fa, real b, const YType& fb, const YType& fmiddle) const - { - return (b-a)*(fa + fb + 4.0*fmiddle)/6.0; - } -}; - -Simpson simpson() -{ return Simpson(); } - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/trapezoid.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/trapezoid.h deleted file mode 100644 index efcde94d6aa..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/methods/trapezoid.h +++ /dev/null @@ -1,26 +0,0 @@ -#ifndef _QDT_TRAPEZOID_H_ -#define _QDT_TRAPEZOID_H_ - -#include "../common/method.h" - -namespace qdt -{ - -class Trapezoid: public MethodWithBoundaries -{ -public: - Trapezoid() { } - - template - YType integrate_with_boundaries(const Function& f, real a, const YType& fa, real b, const YType& fb) const - { - return (b-a)*0.5*(fa + fb); - } -}; - -Trapezoid trapezoid() -{ return Trapezoid(); } - -}; - -#endif diff --git a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/qdt.h b/third_party/cpgfunctionEP-0.1/third_party/qdt-master/qdt.h deleted file mode 100644 index 323cc981dc2..00000000000 --- a/third_party/cpgfunctionEP-0.1/third_party/qdt-master/qdt.h +++ /dev/null @@ -1,8 +0,0 @@ -#include "methods/adaptive.h" -#include "methods/constant-step.h" -#include "methods/gauss-kronrod.h" -#include "methods/monte-carlo.h" -#include "methods/rectangle.h" -#include "methods/romberg.h" -#include "methods/simpson.h" -#include "methods/trapezoid.h" From 7d73957241f712792dc13fe9e225c5386771021f Mon Sep 17 00:00:00 2001 From: j_c_cook Date: Tue, 20 Jul 2021 14:37:04 -0500 Subject: [PATCH 26/77] Add cpgfunction-v0.2 --- third_party/cpgfunctionEP-0.2/.gitignore | 7 + third_party/cpgfunctionEP-0.2/CHANGELOG.md | 149 ++ third_party/cpgfunctionEP-0.2/CMakeLists.txt | 82 + third_party/cpgfunctionEP-0.2/CONTRIBUTING.md | 50 + third_party/cpgfunctionEP-0.2/LICENSE | 29 + third_party/cpgfunctionEP-0.2/README.md | 88 + .../cpgfunctionEP-0.2/include/blas/blas.h | 26 + .../include/cpgfunction/boreholes.h | 96 + .../include/cpgfunction/coordinates.h | 35 + .../include/cpgfunction/gfunction.h | 63 + .../include/cpgfunction/heat_transfer.h | 52 + .../include/cpgfunction/interpolation.h | 24 + .../include/cpgfunction/statistics.h | 20 + .../include/cpgfunction/utilities.h | 23 + third_party/cpgfunctionEP-0.2/src/blas.cpp | 43 + .../cpgfunctionEP-0.2/src/boreholes.cpp | 311 +++ .../cpgfunctionEP-0.2/src/coordinates.cpp | 192 ++ .../cpgfunctionEP-0.2/src/gfunction.cpp | 507 ++++ .../cpgfunctionEP-0.2/src/heat_transfer.cpp | 270 +++ .../cpgfunctionEP-0.2/src/interpolation.cpp | 91 + .../cpgfunctionEP-0.2/src/statistics.cpp | 25 + .../cpgfunctionEP-0.2/src/utilities.cpp | 85 + .../third_party/eigen-3.3.9/.gitignore | 37 + .../third_party/eigen-3.3.9/.hgeol | 11 + .../eigen-3.3.9/Eigen/CMakeLists.txt | 19 + .../third_party/eigen-3.3.9/Eigen/Core | 542 +++++ .../third_party/eigen-3.3.9/Eigen/Dense | 3 + .../third_party/eigen-3.3.9/Eigen/Householder | 30 + .../third_party/eigen-3.3.9/Eigen/Jacobi | 33 + .../third_party/eigen-3.3.9/Eigen/LU | 50 + .../eigen-3.3.9/Eigen/src/Core/Array.h | 329 +++ .../eigen-3.3.9/Eigen/src/Core/ArrayBase.h | 226 ++ .../eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h | 209 ++ .../eigen-3.3.9/Eigen/src/Core/Assign.h | 90 + .../Eigen/src/Core/AssignEvaluator.h | 935 +++++++ .../eigen-3.3.9/Eigen/src/Core/BandMatrix.h | 353 +++ .../eigen-3.3.9/Eigen/src/Core/Block.h | 452 ++++ .../eigen-3.3.9/Eigen/src/Core/BooleanRedux.h | 164 ++ .../Eigen/src/Core/CommaInitializer.h | 160 ++ .../Eigen/src/Core/ConditionEstimator.h | 175 ++ .../Eigen/src/Core/CoreEvaluators.h | 1688 +++++++++++++ .../Eigen/src/Core/CoreIterators.h | 127 + .../Eigen/src/Core/CwiseBinaryOp.h | 184 ++ .../Eigen/src/Core/CwiseNullaryOp.h | 866 +++++++ .../Eigen/src/Core/CwiseTernaryOp.h | 197 ++ .../eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h | 103 + .../Eigen/src/Core/CwiseUnaryView.h | 130 + .../eigen-3.3.9/Eigen/src/Core/DenseBase.h | 612 +++++ .../Eigen/src/Core/DenseCoeffsBase.h | 681 ++++++ .../eigen-3.3.9/Eigen/src/Core/DenseStorage.h | 570 +++++ .../eigen-3.3.9/Eigen/src/Core/Diagonal.h | 260 ++ .../Eigen/src/Core/DiagonalMatrix.h | 343 +++ .../Eigen/src/Core/DiagonalProduct.h | 28 + .../eigen-3.3.9/Eigen/src/Core/Dot.h | 318 +++ .../eigen-3.3.9/Eigen/src/Core/EigenBase.h | 159 ++ .../eigen-3.3.9/Eigen/src/Core/Fuzzy.h | 155 ++ .../Eigen/src/Core/GeneralProduct.h | 455 ++++ .../Eigen/src/Core/GenericPacketMath.h | 590 +++++ .../Eigen/src/Core/GlobalFunctions.h | 187 ++ .../eigen-3.3.9/Eigen/src/Core/IO.h | 225 ++ .../eigen-3.3.9/Eigen/src/Core/Inverse.h | 118 + .../eigen-3.3.9/Eigen/src/Core/Map.h | 171 ++ .../eigen-3.3.9/Eigen/src/Core/MapBase.h | 308 +++ .../Eigen/src/Core/MathFunctions.h | 1421 +++++++++++ .../Eigen/src/Core/MathFunctionsImpl.h | 101 + .../eigen-3.3.9/Eigen/src/Core/Matrix.h | 459 ++++ .../eigen-3.3.9/Eigen/src/Core/MatrixBase.h | 530 ++++ .../eigen-3.3.9/Eigen/src/Core/NestByValue.h | 110 + .../eigen-3.3.9/Eigen/src/Core/NoAlias.h | 108 + .../eigen-3.3.9/Eigen/src/Core/NumTraits.h | 248 ++ .../Eigen/src/Core/PermutationMatrix.h | 605 +++++ .../Eigen/src/Core/PlainObjectBase.h | 1037 ++++++++ .../eigen-3.3.9/Eigen/src/Core/Product.h | 186 ++ .../Eigen/src/Core/ProductEvaluators.h | 1138 +++++++++ .../eigen-3.3.9/Eigen/src/Core/Random.h | 182 ++ .../eigen-3.3.9/Eigen/src/Core/Redux.h | 505 ++++ .../eigen-3.3.9/Eigen/src/Core/Ref.h | 284 +++ .../eigen-3.3.9/Eigen/src/Core/Replicate.h | 142 ++ .../Eigen/src/Core/ReturnByValue.h | 117 + .../eigen-3.3.9/Eigen/src/Core/Reverse.h | 211 ++ .../eigen-3.3.9/Eigen/src/Core/Select.h | 162 ++ .../Eigen/src/Core/SelfAdjointView.h | 352 +++ .../Eigen/src/Core/SelfCwiseBinaryOp.h | 47 + .../eigen-3.3.9/Eigen/src/Core/Solve.h | 188 ++ .../Eigen/src/Core/SolveTriangular.h | 235 ++ .../eigen-3.3.9/Eigen/src/Core/SolverBase.h | 130 + .../eigen-3.3.9/Eigen/src/Core/StableNorm.h | 221 ++ .../eigen-3.3.9/Eigen/src/Core/Stride.h | 111 + .../eigen-3.3.9/Eigen/src/Core/Swap.h | 67 + .../eigen-3.3.9/Eigen/src/Core/Transpose.h | 405 ++++ .../Eigen/src/Core/Transpositions.h | 368 +++ .../Eigen/src/Core/TriangularMatrix.h | 985 ++++++++ .../eigen-3.3.9/Eigen/src/Core/VectorBlock.h | 96 + .../eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h | 695 ++++++ .../eigen-3.3.9/Eigen/src/Core/Visitor.h | 273 +++ .../Eigen/src/Core/arch/CUDA/Complex.h | 103 + .../Eigen/src/Core/arch/CUDA/Half.h | 675 ++++++ .../Eigen/src/Core/arch/CUDA/PacketMathHalf.h | 1124 +++++++++ .../Eigen/src/Core/arch/CUDA/TypeCasting.h | 212 ++ .../Eigen/src/Core/arch/Default/ConjHelper.h | 29 + .../Eigen/src/Core/arch/Default/Settings.h | 49 + .../Eigen/src/Core/arch/SSE/Complex.h | 471 ++++ .../Eigen/src/Core/arch/SSE/MathFunctions.h | 562 +++++ .../Eigen/src/Core/arch/SSE/PacketMath.h | 895 +++++++ .../Eigen/src/Core/arch/SSE/TypeCasting.h | 77 + .../src/Core/functors/AssignmentFunctors.h | 168 ++ .../Eigen/src/Core/functors/BinaryFunctors.h | 475 ++++ .../Eigen/src/Core/functors/NullaryFunctors.h | 188 ++ .../Eigen/src/Core/functors/StlFunctors.h | 136 ++ .../Eigen/src/Core/functors/TernaryFunctors.h | 25 + .../Eigen/src/Core/functors/UnaryFunctors.h | 792 ++++++ .../Core/products/GeneralBlockPanelKernel.h | 2157 +++++++++++++++++ .../src/Core/products/GeneralMatrixMatrix.h | 495 ++++ .../products/GeneralMatrixMatrixTriangular.h | 317 +++ .../GeneralMatrixMatrixTriangular_BLAS.h | 145 ++ .../Core/products/GeneralMatrixMatrix_BLAS.h | 124 + .../src/Core/products/GeneralMatrixVector.h | 619 +++++ .../Core/products/GeneralMatrixVector_BLAS.h | 136 ++ .../Eigen/src/Core/products/Parallelizer.h | 166 ++ .../Core/products/SelfadjointMatrixMatrix.h | 527 ++++ .../products/SelfadjointMatrixMatrix_BLAS.h | 295 +++ .../Core/products/SelfadjointMatrixVector.h | 260 ++ .../products/SelfadjointMatrixVector_BLAS.h | 118 + .../src/Core/products/SelfadjointProduct.h | 133 + .../Core/products/SelfadjointRank2Update.h | 93 + .../Core/products/TriangularMatrixMatrix.h | 472 ++++ .../products/TriangularMatrixMatrix_BLAS.h | 317 +++ .../Core/products/TriangularMatrixVector.h | 350 +++ .../products/TriangularMatrixVector_BLAS.h | 255 ++ .../Core/products/TriangularSolverMatrix.h | 335 +++ .../products/TriangularSolverMatrix_BLAS.h | 167 ++ .../Core/products/TriangularSolverVector.h | 145 ++ .../Eigen/src/Core/util/BlasUtil.h | 499 ++++ .../Eigen/src/Core/util/Constants.h | 547 +++++ .../src/Core/util/DisableStupidWarnings.h | 94 + .../Eigen/src/Core/util/ForwardDeclarations.h | 298 +++ .../Eigen/src/Core/util/MKL_support.h | 130 + .../eigen-3.3.9/Eigen/src/Core/util/Macros.h | 1053 ++++++++ .../eigen-3.3.9/Eigen/src/Core/util/Memory.h | 993 ++++++++ .../eigen-3.3.9/Eigen/src/Core/util/Meta.h | 568 +++++ .../src/Core/util/ReenableStupidWarnings.h | 31 + .../Eigen/src/Core/util/StaticAssert.h | 218 ++ .../Eigen/src/Core/util/XprHelper.h | 838 +++++++ .../Eigen/src/Householder/BlockHouseholder.h | 103 + .../Eigen/src/Householder/Householder.h | 172 ++ .../src/Householder/HouseholderSequence.h | 470 ++++ .../eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h | 462 ++++ .../eigen-3.3.9/Eigen/src/LU/Determinant.h | 101 + .../eigen-3.3.9/Eigen/src/LU/FullPivLU.h | 891 +++++++ .../eigen-3.3.9/Eigen/src/LU/InverseImpl.h | 415 ++++ .../eigen-3.3.9/Eigen/src/LU/PartialPivLU.h | 614 +++++ .../Eigen/src/LU/PartialPivLU_LAPACKE.h | 83 + .../Eigen/src/LU/arch/Inverse_SSE.h | 338 +++ .../eigen-3.3.9/Eigen/src/misc/Image.h | 82 + .../eigen-3.3.9/Eigen/src/misc/Kernel.h | 79 + .../Eigen/src/plugins/ArrayCwiseBinaryOps.h | 332 +++ .../Eigen/src/plugins/ArrayCwiseUnaryOps.h | 552 +++++ .../Eigen/src/plugins/BlockMethods.h | 1058 ++++++++ .../Eigen/src/plugins/CommonCwiseBinaryOps.h | 115 + .../Eigen/src/plugins/CommonCwiseUnaryOps.h | 163 ++ .../Eigen/src/plugins/MatrixCwiseBinaryOps.h | 152 ++ .../Eigen/src/plugins/MatrixCwiseUnaryOps.h | 85 + .../third_party/eigen-3.3.9/README.md | 3 + .../third_party/qdt-master/.gitignore | 28 + .../third_party/qdt-master/LICENSE.txt | 674 +++++ .../third_party/qdt-master/README.md | 67 + .../third_party/qdt-master/common/method.h | 211 ++ .../third_party/qdt-master/methods/adaptive.h | 395 +++ .../qdt-master/methods/gauss-kronrod.h | 110 + .../third_party/qdt-master/qdt.h | 2 + 170 files changed, 52913 insertions(+) create mode 100644 third_party/cpgfunctionEP-0.2/.gitignore create mode 100644 third_party/cpgfunctionEP-0.2/CHANGELOG.md create mode 100644 third_party/cpgfunctionEP-0.2/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.2/CONTRIBUTING.md create mode 100644 third_party/cpgfunctionEP-0.2/LICENSE create mode 100644 third_party/cpgfunctionEP-0.2/README.md create mode 100644 third_party/cpgfunctionEP-0.2/include/blas/blas.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/boreholes.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/coordinates.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/gfunction.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/heat_transfer.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/interpolation.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/statistics.h create mode 100644 third_party/cpgfunctionEP-0.2/include/cpgfunction/utilities.h create mode 100644 third_party/cpgfunctionEP-0.2/src/blas.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/boreholes.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/coordinates.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/gfunction.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/heat_transfer.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/interpolation.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/statistics.cpp create mode 100644 third_party/cpgfunctionEP-0.2/src/utilities.cpp create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/.gitignore create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/.hgeol create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Core create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Dense create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Householder create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Jacobi create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/LU create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Array.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Block.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/IO.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Map.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Product.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Random.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Select.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/misc/Image.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/README.md create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/.gitignore create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/LICENSE.txt create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/README.md create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/common/method.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/methods/adaptive.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/methods/gauss-kronrod.h create mode 100644 third_party/cpgfunctionEP-0.2/third_party/qdt-master/qdt.h diff --git a/third_party/cpgfunctionEP-0.2/.gitignore b/third_party/cpgfunctionEP-0.2/.gitignore new file mode 100644 index 00000000000..a4122fb56eb --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/.gitignore @@ -0,0 +1,7 @@ +cmake-build-debug/ +.idea/ +cmake-build-release/ +build/ +Create_Branch.sh +package-lock.json +Shell_Speed/* diff --git a/third_party/cpgfunctionEP-0.2/CHANGELOG.md b/third_party/cpgfunctionEP-0.2/CHANGELOG.md new file mode 100644 index 00000000000..51e8dd3153c --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/CHANGELOG.md @@ -0,0 +1,149 @@ +# History of changes - cpgfunctionEP + +## Version 0.2 (2021-07-20) + +### Removes + +* [Issue 19](https://github.com/j-c-cook/cpgfunctionEP/issues/19) - The + references to nlohmann json are removed from inside of the library. The + EnergyPlus team does not want the dependency, and there will never be a + reason for EnergyPlus to use the coordinate input/output functions. The + tests are still dependent on nlohmann when built. + +### Fixes + +* [Issue 12](https://github.com/j-c-cook/cpgfunctionEP/issues/12) - The use + similarities false option now computes and fills the packed segment response + matrix. + +### Enhancements + +* [Issue 14](https://github.com/j-c-cook/cpgfunctionEP/issues/14) - The size of + the borehole fields being run in the regression tests are reduced so that less + time is spent running tests (that occur at each push). + +* [Issue 13](https://github.com/j-c-cook/cpgfunctionEP/issues/13) - The tests + are only configured and ran if the main project being worked on is + cpgfunctionEP. This is done because EnergyPlus will use this as a third party + subdirectory, and will not want to run the tests located in this library. + +### Maintenance + +* [Issue 26](https://github.com/j-c-cook/cpgfunctionEP/issues/26) - The size + (number of files) of the Eigen dependency is minimized by only including the + necesary files to execute the LU decomposition. + +* [Issue 28](https://github.com/j-c-cook/cpgfunctionEP/issues/28) - The third + party library `nlohmann` json now only makes use of the single json.hpp + header include. This has been done to reduce the number of files in this + project. + +### Changes + +* [Issue 8](https://github.com/j-c-cook/cpgfunctionEP/issues/8) - The uniform + borehole wall temperature g-function calculation has an API change. The + boolean argument for multi-threading is taken away. The number of threads is + still an argument, and the default is all available threads on the processor. + +## Version 0.1 (2021-07-08) + +### Maintenance + +* [Issue 9](https://github.com/j-c-cook/cpgfunctionEP/issues/9) - The project builds on Linux, MacOS and Windows upon + commit [558c63c](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/558c63c9b5c4041282fdb473327463b8fe183fbd). + +* [Issue 6](https://github.com/j-c-cook/cpgfunctionEP/issues/6) - Dependencies are placed into the `third_party/` + folder. The dependencies source code is pasted in, rather than using `git subtree` or `git submodule`. + +* [Issue 5](https://github.com/j-c-cook/cpgfunctionEP/issues/5) - Blas and lapack have been removed, thus deprecating + libraries dependency on Fortran. The following lists what the Fortran function calls were replaced with by commit: + - Commit [a20d3ea](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/a20d3eacec67d5994b65d6716774c50404e26428) - + Replaces the blas copy function with the native C++ algorithm standard copy function (`std::copy`) + - Commits [ec35baf](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/ec35baf21025bb61a105ac9edd06e1e348676702) + and [70ab158](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/70ab1585d0efe5dd53dca90f3cf5685585030ef9) - + Replaces the `blas::axpy` Fortran code with a C++ version. + - Commit [383bf08](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/383bf08f31d020af2cabe63d0d8b86eb4bbb8191) - + Replaces the symmetric packed matrix Fortran call with C++ code. + - Commit [231e35e](https://github.com/j-c-cook/cpgfunctionEP/pull/1/commits/231e35ea416fe0943514cc82dc79238f3f2b79dd) - The + LU decomposition of lapack is replaced by an Eigen function call. + +* [Issue 3](https://github.com/j-c-cook/cpgfunctionEP/issues/3) - Boost is removed by replacing the thread pools with + `OpenMP` calls, and the Gauss-Kronrod integration is replaced with a header only file by the name of + [qdt](https://github.com/j-c-cook/cpgfunctionEP/tree/MilestoneV0.1/third_party/qdt-master). Qdt is a library + written by Adolfo Munoz who released the code alongside a paper by the name of "Higher Order Ray Marching". + The Kronrod integral of Munoz appears much faster than boost's. + + +### Fixes + +* [Issue 10](https://github.com/j-c-cook/cpgfunctionEP/issues/10) - Vector over flow error associated with number of + time steps being less than 6 is fixed. + Commit [558c63c](https://github.com/j-c-cook/cpgfunctionEP/commit/558c63c9b5c4041282fdb473327463b8fe183fbd) + fixes a vector overflow error that occured when the number of time steps described were less than 6 (`nt<6`). Commit + [94e222b](https://github.com/j-c-cook/cpgfunctionEP/commit/94e222bb0cf670f1a4f3058bd1b7b6c7f8bf8233) now allows the + function time_geometric to have less than 6 time steps. + + +# History of changes - cpgfunction + +## Version 2.0.0 (2021-05-23) + +### Enhancements + +* [Issue 25](https://github.com/j-c-cook/cpgfunction/issues/25) - Removes all references to the 3D `h_ij` + segment response matrix. See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +* [Issue 32](https://github.com/j-c-cook/cpgfunction/issues/32) - The multi-dimensional matrices, + `q_reconstructed` and `h_ij`, are made one dimensional prior to passage into the temporal superposition + function so that `BLAS` routines can be heavily depended on and the loops completely unraveled. + See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +* [Issue 33](https://github.com/j-c-cook/cpgfunction/issues/33) - It is found that the + packed segment resopnse matrix can be directly made us of in `BLAS spmv`, and that addition + greatly optimizes the temporal superposition function. For now the assumption is made that all + segments in the field are of equivalent length, which is true and fine, but at some point in the + future unequal segment lengths should be made possible again. + See [PR 30](https://github.com/j-c-cook/cpgfunction/pull/30). + +### New features + +* [Issue 28](https://github.com/j-c-cook/cpgfunction/issues/28) - + The third party library LinearAlgebra (`jcc:la`) is included and made use of for `LU` + factorization in `gfunction.cpp` + +* [Issue 12](https://github.com/j-c-cook/cpgfunction/issues/12) - + A boolean toggle option is added for multi-threading for computing the + uniform borehole wall temperature (UBHWT) g-function + +### API Changes + +* [Issue 16](https://github.com/j-c-cook/cpgfunction/issues/16) - The `uniform borehole wall temperature` + g-function definition is defined for planned use in EnergyPlus with all arguments. Not all the arguments + currently have a purpose, the adaptive discretization and number of thread arguments are place holders. + +## Version 1.0.0 (2021-05-12) + +### New features + +* [Issue 20](https://github.com/j-c-cook/cpgfunction/issues/20) - + Added OpenBlas as the basic linear algebra subprogram (BLAS) vendor to CMakeLists.txt + +* [Issue 18](https://github.com/j-c-cook/cpgfunction/issues/18) - + Added new borefield interface with API access to typical borehole configurations + +* [Issue 13](https://github.com/j-c-cook/cpgfunction/issues/13) - + Implemented g-function accuracy tests via CMakeLists.txt for a Rectangle, Open Rectangle, U shape, + L shape and a custom (Poisson disk) configuration + +* [Commit 45141fa](https://github.com/j-c-cook/cpgfunction/pull/14/commits/45141fa745d92ac8a08eea2a06801d7a01fac367) - + Create new uniform borehole wall temperature API to consider the new borefield and time API's + +* [Commit f8863ad](https://github.com/j-c-cook/cpgfunction/pull/14/commits/f8863ad6879bdcb43d8bbed48ab1be1701eb56f5) - + Added time vector API and associated test + +* [Commit 654160f](https://github.com/j-c-cook/cpgfunction/pull/14/commits/654160f9b508f57b917fc0630437cff726dc8440) - + Modify API for creating a vector of boreholes (borefield) + + + + diff --git a/third_party/cpgfunctionEP-0.2/CMakeLists.txt b/third_party/cpgfunctionEP-0.2/CMakeLists.txt new file mode 100644 index 00000000000..d46d5c777b4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/CMakeLists.txt @@ -0,0 +1,82 @@ +cmake_minimum_required(VERSION 3.6) +project(cpgfunctionEP) + +set(CMAKE_CXX_STANDARD 17) +if(CMAKE_CXX_COMPILER_ID MATCHES "GNU|Clang") + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} --std=c++17") +else() + set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON) + add_compile_definitions(_USE_MATH_DEFINES) +endif() + +# Everything will need to access this +include_directories(include) +# Third party includes +include_directories(third_party/eigen-3.3.9) # Eigen for LU decomposition +include_directories(third_party/qdt-master) + +# Create a DLL with an API +add_library(cpgfunctionEP SHARED + src/boreholes.cpp + src/gfunction.cpp + src/heat_transfer.cpp + src/interpolation.cpp + src/utilities.cpp + src/coordinates.cpp + src/statistics.cpp + src/blas.cpp) + +# It depends on threads +find_package(Threads REQUIRED) +target_link_libraries(cpgfunctionEP ${CMAKE_THREAD_LIBS_INIT}) + +OPTION (USE_OpenMP "Use OpenMP" ON) +IF(USE_OpenMP) + FIND_PACKAGE(OpenMP) + IF(OPENMP_FOUND) + SET(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}") + SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}") + ENDIF() +ENDIF() + +# if we are only building cpgfnctionEP +if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME) + include_directories(third_party/nlohmann-3.9.1) + # Turn on testing + enable_testing() + + # Create a test binary that depends on the DLL + add_executable(gFunction_minimal test/gFunction_minimal.cpp) + add_executable(interpolation test/interpolation.cpp) + # add_executable(test3 test/configurations.cpp) + add_executable(borefield_definition test/borefield_definition.cpp) + add_executable(time_definition test/time_definition.cpp) + add_executable(gFunction_tests test/gFunction_tests.cpp) + add_executable(packed test/packed.cpp) + + target_link_libraries(gFunction_minimal cpgfunctionEP) + target_link_libraries(interpolation cpgfunctionEP) + # target_link_libraries(test3 cpgfunctionEP) + target_link_libraries(borefield_definition cpgfunctionEP) + target_link_libraries(time_definition cpgfunctionEP) + target_link_libraries(gFunction_tests cpgfunctionEP) + target_link_libraries(packed cpgfunctionEP) + + # Copy validation files to build directory so tests can open + file(GLOB JsonValidation test/validation/*.json) + file(GLOB CustomLayout test/custom_configurations/*.json) + + file(COPY ${JsonValidation} DESTINATION ${CMAKE_BINARY_DIR}) + file(COPY ${CustomLayout} DESTINATION ${CMAKE_BINARY_DIR}) + message(STATUS "The binary directory: ${CMAKE_BINARY_DIR}") + + # Test the test binary itself + add_test(NAME RunTest1 COMMAND "${CMAKE_BINARY_DIR}/gFunction_minimal") + add_test(NAME RunTest2 COMMAND "${CMAKE_BINARY_DIR}/interpolation") + # add_test(NAME RunTest3 COMMAND "${CMAKE_BINARY_DIR}/test3") + add_test(NAME RunTest4 COMMAND "${CMAKE_BINARY_DIR}/borefield_definition") + add_test(NAME RunTest5 COMMAND "${CMAKE_BINARY_DIR}/time_definition") + # Pass variable path into test 6 for json files + add_test(NAME RunTest6 COMMAND ${CMAKE_BINARY_DIR}/gFunction_tests) + add_test(NAME RunTest7 COMMAND ${CMAKE_BINARY_DIR}/packed) +endif() diff --git a/third_party/cpgfunctionEP-0.2/CONTRIBUTING.md b/third_party/cpgfunctionEP-0.2/CONTRIBUTING.md new file mode 100644 index 00000000000..fe63ee28656 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/CONTRIBUTING.md @@ -0,0 +1,50 @@ +# Contributing to cpgfunction + +## Branching + +To contribute code please first start an issue if one if not already started. +Then: +- Branch from the most current `master` branch: +``` +git checkout -b # checkout a branch locally from master +git push -u origin # push to your remote that you have a new branch +git branch --set-upstream-to=/ +``` +- Commit your changes and push them up +``` +git add +git commit -m "Type 50 character message here" +git push origin issuexx_ShortDescription +``` +- Create a pull request + +- Document the changelog for your issue + +The branch will then be merged into master once it has been reviewed. After the +branch is merged (which will subsequently close the pull request and related +issue(s)): +- Delete your remote and local branches (proceed with caution) +``` +git push -d +git branch -d +``` + +## Changelog + +The cpgfunction library keeps a +[changelog](https://github.com/j-c-cook/cpgfunction/blob/master/CHANGELOG.md) +so that changes upon each release are transparent and easily understood. Prior +to a pull request being accepted, all changes must be marked in the changelog. +The changes should fall under one of the following categories: + +- New features - for new features +- Enhancements - for improvements made to code performance and functionality +- Maintenance - for tidying code +- Changes - for changes in functionality of the code +- Deprecates - for soon-to-be removed features +- Removes - for removed features +- Fixes - for any bug fixes + +## Versioning + +This library makes use of [Semantic Versioning](https://semver.org/). diff --git a/third_party/cpgfunctionEP-0.2/LICENSE b/third_party/cpgfunctionEP-0.2/LICENSE new file mode 100644 index 00000000000..1c1a20e0de4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/LICENSE @@ -0,0 +1,29 @@ +BSD 3-Clause License + +Copyright (c) 2020, Jack C. Cook +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are met: + +1. Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + +2. Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + +3. Neither the name of the copyright holder nor the names of its + contributors may be used to endorse or promote products derived from + this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/third_party/cpgfunctionEP-0.2/README.md b/third_party/cpgfunctionEP-0.2/README.md new file mode 100644 index 00000000000..b1040e32bc8 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/README.md @@ -0,0 +1,88 @@ +# cpgfunctionEP +An open source low level (C++) implementation of [Massimo Cimmino][1]'s +g-function methodology. The g-function calculation has to do with the distribution +of heat in a ground heat exchanger (GHE). The g-function, after computed, +can be used to simulate a +ground source heat pump to determine the heat pump +exiting fluid temperatures after a +period of time. It is of paramount importance to accurately predict +the thermal response of the borefield. An improper prediction of the +ground response could result in a system too large (waste of money) or a +system too small (resulting in failure). The g-function is currently the only +known methodology to accurately predict the thermal response of the ground. + +Computing the g-function is a computationally demanding procedure. Historically, +g-functions are pre-computed and stored in libraries (databases) which can be +accessed in GHE design tools. This is done so that the ground source heat pump +can be sized within a few seconds, rather than waiting anywhere from minutes +to hours for a g-function to be computed. This code base was implemented to help +solve a major limitation (problem) associated with +computing large sets of g-functions for borefields containing anywhere from 1 +to 1024 boreholes. Massimo Cimmino's open source implementation of the g-function +calculation +in Python ([pygfunction][2]) has a limitation; at times the memory consumption is +inordinate. +This library was implemented so that computing g-functions would require +significantly less memory to increase throughput on high performance clusters +which contain many "low memory" nodes, and only a few "high memory" nodes. + +The g-function is greatly dependent on the boundary condition used. The following +is a checklist of boundary conditions contained in this library: + +- [ ] Uniform heat flux (UHF) +- [x] Uniform borehole wall temperature (UBHWT) +- [ ] Uniform inlet fluid temperature (UIFT) + +[1]: https://github.com/MassimoCimmino +[2]: https://github.com/MassimoCimmino/pygfunction + +# Build + +Create a build with cmake: + +``` +cd /path/to/repo +mkdir build +cd build +ccmake .. # set settings as needed, shouldn't be any of interest, configure then generate +``` + +On a Visual Studio generator this will create a solution you can launch into Visual Studio. +On makefile style generators, this will create a makefile. From the build directory, just make the project: + +``` +make -j 4 # number of processors to use to build +``` + +Then you can run the test: + +``` +ctest +``` + +On my platform this resulted in: +``` +Test project /home/edwin/Projects/cpgfunction/cmake-build-debug + Start 1: RunTest1 +1/1 Test #1: RunTest1 ......................... Passed 0.79 sec +``` + +This ran what used to just be the main.cpp built file. + +# Acknowledgments +The initial funding of this research project was provided by Dr. Jeffrey D. +Spitler of Oklahoma State University. Further development of the library was +funded by the National Renewable Energy Laboratory (NREL), which is owned by +the U.S. Department of Energy (DOE). The NREL contract requests delivery of +cpgfunction for use in EnergyPlus as a third party tool to calculate fast and +accurate g-functions. Initial integration into EnergyPlus makes use of the +UBHWT boundary condition with the adaptive discretization scheme to approximate +UIFT g-functions. + +The original g-function theory appeared in a Per Eskilon's PhD thesis, +his advisor who helped develop that theory was Prof. Johan Claesson of Lund +University in Sweden. Massimo Cimmino, Assistant Professor at Polytechnique +Montreal an Canada, has built on the work of Eskilson, Claesson +and others. Cimmino has developed a methodology for g-function calculations, +written programs to compute the g-function and then made the program open source. +The openness of Cimmino's g-function calculation is monumental. diff --git a/third_party/cpgfunctionEP-0.2/include/blas/blas.h b/third_party/cpgfunctionEP-0.2/include/blas/blas.h new file mode 100644 index 00000000000..5dc19d916cd --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/blas/blas.h @@ -0,0 +1,26 @@ +// +// Created by jackcook on 6/29/21. +// + +// BLAS functions that have been written in C++ + +#include + +using namespace std; + +#ifndef CPGFUNCTION_BLAS_H +#define CPGFUNCTION_BLAS_H + +namespace jcc { namespace blas { + + void axpy(int &n, double &a, vector &x, vector &y, + int &start, int &n_threads); + + void spmv(int &n, double &alpha, vector &A, vector &x, + double &beta, vector &y, int &start, int &n_threads); + + } // namespace blas +} // namespace jcc + + +#endif //CPGFUNCTION_BLAS_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/boreholes.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/boreholes.h new file mode 100644 index 00000000000..3becd4f33ab --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/boreholes.h @@ -0,0 +1,96 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_BOREHOLES_H +#define CPPGFUNCTION_BOREHOLES_H + +namespace gt { + + double Distance_Formula(double x1, double y1, double x2, double y2); + + namespace boreholes { + struct Borehole { + // Destructor + virtual ~Borehole() { + } + + double H; // height or length of the borehole (meters) + double D; // borehole burial depth (meters) + double r_b; // borehole radius (meters) + double x; // position (meters) of the center of the borehole along the x-axis + double y; // position (meters) of the center of the borehole along the y-axis + + Borehole(double H=0.0, + double D=0.0, + double r_b=0.0, + double x=0.0, + double y=0.0) : H(H), D(D), r_b(r_b), x(x), y(y) { + } + + double distance(Borehole target); + tuple position(); + }; + + vectorboreField( + const vector > &coordinates, + const double &r_b, const double &H, const double &D); + + struct SimilaritiesType { + ~SimilaritiesType() {} // destructor + + int nSim = 0; + vector< vector > > Sim; + vector > HSim; + vector > DSim; + vector disSim; + + SimilaritiesType() {} // constructor + }; + + struct Similarity { + ~Similarity() {} // destructor + + // delcare variables + // positive similarities + vector nSimPos; // number of positive similarities + vector > > simPos; // number of sim positions + vector disSimPos; // the distances between each position + vector > HSimPos; // the heights of each borehole + vector > DSimPos; // the burial depth of each borehole + // negative similarities + vector nSimNeg; + vector > > simNeg; + vector disSimNeg; + vector > HSimNeg; + vector > DSimNeg; + + Similarity() {} // constructor + + + void similarities(SimilaritiesType &SimReal, SimilaritiesType &SimImage, + vector &boreSegments, + bool splitRealAndImage = true, double disTol = 0.1, double tol = 1.0e-6); + + void _similarities_group_by_distance(vector &boreSegments, + vector > > &Pairs, + vector &nPairs, vector &disPairs, int &nDis, + double disTol = 0.1); + + void _similarities_one_distance(SimilaritiesType &SimT, vector > &pairs, + vector &boreSegments, const string &kind, + double tol = 1.0e-6); + }; + + } // boreholes name space + +} // gt namespace + +#endif //CPPGFUNCTION_BOREHOLES_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/coordinates.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/coordinates.h new file mode 100644 index 00000000000..6118bf95f3d --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/coordinates.h @@ -0,0 +1,35 @@ +// +// Created by jackcook on 5/6/21. +// + +#include +#include +#include +#include + +using namespace std; + +#ifndef CPGFUNCTION_COORDINATES_H +#define CPGFUNCTION_COORDINATES_H + +namespace gt::coordinates { + + vector> configuration(const string& shape, int Nx, + int Ny, double Bx, double By); +// vector> configuration(const string& shape, +// const string& input_path); + + vector> rectangle(int Nx, int Ny, double Bx, + double By); + vector> Open_rectangle(int Nx, int Ny, double Bx, + double By); + vector> U_shape(int Nx, int Ny, double Bx, double By); + vector> L_shape(int Nx, int Ny, double Bx, double By); + +// vector> import_coordinates_from_file(const string& input_path); +// void export_coordinates_to_file(const vector> &coordinates, +// const string& output_path); + +} + +#endif //CPGFUNCTION_COORDINATES_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/gfunction.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/gfunction.h new file mode 100644 index 00000000000..6b97671055c --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/gfunction.h @@ -0,0 +1,63 @@ +// -*- lsst-c++ -*- + +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_GFUNCTION_H +#define CPPGFUNCTION_GFUNCTION_H + +/** The functions will be listed here in the order which they appear in the + * associated cpp file **/ + +namespace gt::gfunction { + /** + * Uniform borehole wall temperature (UBWHT) g-function calculation method + * + * This function superimposes the finite line source (FLS) solution to + * estimate the g-function of a geothermal bore field. Each borehole is + * modeled as a series of finite line source segments, as proposed + * in [CITE: CimminoBernier2014]. + * + * @param gfunction + * @param boreholes + * @param time + * @param alpha + * @param nSegments + * @param use_similarities + * @param disp + */ + vector uniform_borehole_wall_temperature( + vector &boreField, + vector &time, double alpha, int nSegments=12, + bool use_similarities=true, bool adaptive=true, + int n_Threads=int(thread::hardware_concurrency()), + bool display=false); + + void _borehole_segments(vector& boreSegments, + vector& boreholes, + int nSegments); + void load_history_reconstruction(vector& q_reconstructed, + vector& time, + vector& _time, + vector >& Q, + vector& dt, int p); + void _temporal_superposition(vector& Tb_0, + gt::heat_transfer::SegmentResponse &SegRes, + vector &h_ij, + vector &q_reconstructed, int p, + int &nSources); + void _solve_eqn(vector& x, vector>& A, + vector& b); + +} // namespace gt::gfunction + +#endif //CPPGFUNCTION_GFUNCTION_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/heat_transfer.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/heat_transfer.h new file mode 100644 index 00000000000..57c0ba9f6f2 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/heat_transfer.h @@ -0,0 +1,52 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include + +using namespace std; +using namespace gt; + +#ifndef CPPGFUNCTION_HEAT_TRANSFER_H +#define CPPGFUNCTION_HEAT_TRANSFER_H + +namespace gt::heat_transfer { + + struct SegmentResponse { + ~SegmentResponse() {} // destructor + + int nSources; + int nSum; + vector < vector < double > > h_ij; + vector boreSegments; + + SegmentResponse(int nSources, + int nSum, + int nt) : + nSources(nSources), + boreSegments(nSources), + h_ij(nSum, vector(nt, 0)), + nSum(nSum) + {} // constructor + + // storage_mode = 1 is the reduced segment response vector + int storage_mode = 1; + +// void ReSizeContainers(int n, int nt); + void get_h_value(double &h, int i, int j, int k); + void get_index_value(int &index, int i, int j); + }; // struct SegmentResponse(); + + double finite_line_source(double time_, double alpha, + boreholes::Borehole& b1, boreholes::Borehole& b2, + bool reaSource=true, bool imgSource=true); + void thermal_response_factors(SegmentResponse &SegRes, + vector& time, double alpha, + bool use_similaries, bool disp=false); + +} // namespace gt::heat_transfer + +#endif //CPPGFUNCTION_HEAT_TRANSFER_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/interpolation.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/interpolation.h new file mode 100644 index 00000000000..019a271534b --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/interpolation.h @@ -0,0 +1,24 @@ +// +// Created by jackcook on 7/15/20. +// + +#include +#include +#include + +using namespace std; + +#ifndef CPPGFUNCTION_INTERPOLATION_H +#define CPPGFUNCTION_INTERPOLATION_H + +namespace jcc { namespace interpolation { + + double linterp(double xp, double x0, double y0, double x1, double y1); + void interp1d(vector& xp, vector& yp, vector& x, vector& y); + void interp1d(double &xp, double &yp, vector& x, vector& y); + void interp1d(double &xp, double &yp, vector &time, + gt::heat_transfer::SegmentResponse &SegRes, int &i, int &j, int &k); + +} } // jcc::interpolation + +#endif //CPPGFUNCTION_INTERPOLATION_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/statistics.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/statistics.h new file mode 100644 index 00000000000..b9dc497bf67 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/statistics.h @@ -0,0 +1,20 @@ +// +// Created by jackcook on 5/10/21. +// + +#include +#include +#include + +#ifndef CPGFUNCTION_STATISTICS_H +#define CPGFUNCTION_STATISTICS_H + +namespace gt { + namespace statistics { + + double root_mean_square_error(std::vector actual, std::vector predicted); + + } // namespace statistics +} // namespace gt + +#endif //CPGFUNCTION_STATISTICS_H diff --git a/third_party/cpgfunctionEP-0.2/include/cpgfunction/utilities.h b/third_party/cpgfunctionEP-0.2/include/cpgfunction/utilities.h new file mode 100644 index 00000000000..8b7cbc28d86 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/include/cpgfunction/utilities.h @@ -0,0 +1,23 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include + +#ifndef CPPGFUNCTION_UTILITIES_H +#define CPPGFUNCTION_UTILITIES_H + +namespace gt { + namespace utilities { + std::vector time_geometric(double dt, double tmax, int Nt); + std::vector Eskilson_original_points(); + std::vector time_Eskilson(const double &H, const double &alpha); + std::vector convert_time(std::vector &logtime, const double &H, const double &alpha); + void cook_spitler_time (std::vector &logtime); + void convert_time(std::vector &logtime, std::vector &time, double H, double alpha); + } // namespace utilities +} // namespace gt + +#endif //CPPGFUNCTION_UTILITIES_H diff --git a/third_party/cpgfunctionEP-0.2/src/blas.cpp b/third_party/cpgfunctionEP-0.2/src/blas.cpp new file mode 100644 index 00000000000..f3806d57b9e --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/blas.cpp @@ -0,0 +1,43 @@ +// +// Created by jackcook on 6/29/21. +// + +#include + +namespace jcc::blas { + + void axpy(int &n, double &a, vector &x, vector &y, + int &start, int &n_threads){ + // y = a*x + y + #pragma omp parallel for num_threads(n_threads) + for (int i=0; i &A, vector &x, + double &beta, vector &y, int &start, int &n_threads) { + // TODO: implement upper + + // LOWER + double zero = 0.; + double temp1; + double temp2; + int kk = 0; + int k; + for (int j=0; j + +using namespace std; + +namespace gt { + + double Distance_Formula(double x1, double y1, double x2, double y2) { + return sqrt(pow((x1 - x2), 2) + pow((y1 - y2), 2)); + } + +} // namespace gt + +namespace gt::boreholes { + + double Borehole::distance(Borehole target) { + double x1 = x; + double y1 = y; + double x2 = target.x; + double y2 = target.y; + double dist = Distance_Formula(x1, y1, x2, y2); + return max(r_b, double(dist)); // max needs doubles + } + + tuple Borehole::position() { + tuple t (x, y); + return t; + }; + + vector boreField(const vector> &coordinates, + const double &r_b, const double &H, + const double &D){ + vector bores(coordinates.size()); + + double x; + double y; + + for (size_t i = 0; i < coordinates.size(); i++) { + x = get<0>(coordinates[i]); + y = get<1>(coordinates[i]); + bores[i] = Borehole(H, D, r_b, x, y); + } // next i + + return bores; + } // boreField(); + + void Similarity::similarities(SimilaritiesType &SimReal, + SimilaritiesType &SimImage, + vector &boreSegments, + bool splitRealAndImage, double disTol, + double tol) { + // TODO: fork a pool + + // declare the variables local to this function + int nDis; + vector disPairs; + vector nPairs; + vector< vector < tuple > > Pairs; + _similarities_group_by_distance(boreSegments, Pairs, nPairs, + disPairs, nDis); + + vector RealSimT(Pairs.size()); + vector ImageSimT; + + // if real and image parts of the FLS are split, + // evaluate real and image similarities seperately: + if (splitRealAndImage) { + ImageSimT.resize(Pairs.size()); + for (size_t i=0; i &SimFrom, + SimilaritiesType &SimTo) { + int nSimtmp; + // reserve and insert + auto _res_and_ins = [](vector &SimFrom, + SimilaritiesType &SimTo, const int &i) { + // Sim positions + SimTo.Sim.reserve(SimTo.Sim.size() + + distance(SimFrom[i].Sim.begin(), SimFrom[i].Sim.end())); + SimTo.Sim.insert(SimTo.Sim.end(), SimFrom[i].Sim.begin(), + SimFrom[i].Sim.end()); + // H values + SimTo.HSim.reserve(SimTo.HSim.size() + + distance(SimFrom[i].HSim.begin(), SimFrom[i].HSim.end())); + SimTo.HSim.insert(SimTo.HSim.end(), SimFrom[i].HSim.begin(), + SimFrom[i].HSim.end()); + // D Values + SimTo.DSim.reserve(SimTo.DSim.size() + + distance(SimFrom[i].DSim.begin(), SimFrom[i].DSim.end())); + SimTo.DSim.insert(SimTo.DSim.end(), SimFrom[i].DSim.begin(), + SimFrom[i].DSim.end()); + }; + for (int i=0; i > > tmp; + tmp = SimFrom[i].Sim; + _res_and_ins(SimFrom, SimTo, i); + } // next i + }; // auto _aggregate(); + // Sim positions + _aggregate(RealSimT, SimReal); + if (splitRealAndImage) { + _aggregate(ImageSimT, SimImage); + } // fi + } // Similarity::similarities + + void Similarity::_similarities_group_by_distance( + vector &boreSegments, + vector< vector < tuple > > &Pairs, vector &nPairs, + vector &disPairs, int &nDis, double disTol) { + // initialize lists + nPairs.push_back(1); + vector< tuple > vect_w_tup(1); + vect_w_tup[0] = tuple (0, 0); + Pairs.push_back(vect_w_tup); + disPairs.push_back(boreSegments[0].r_b); + nDis = 1; + + int nb = boreSegments.size(); + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + int i2; + double dis; + double rTol; + double diff; + for (int i=0; i (i, j)); + nPairs[k]++; + break; + } // fi disPairs[k] - dis < rTol + // add symmetry to the list if no match was found + if (k == nDis-1) { + nDis++; + disPairs.push_back(dis); + vect_w_tup[0] = tuple (i, j); + Pairs.push_back(vect_w_tup); + nPairs.push_back(1); + break; + } + } // next k + } // for j in range(i2, nb) + } // for i in range(nb) + int a = 1; + } // Similarity::_similarities_group_by_distance + + void Similarity::_similarities_one_distance( + SimilaritiesType & SimT, vector > &pairs, + vector &boreSegments, const string& kind, + double tol) { + // Condition for equivalence of the real part of the FLS solution + auto compare_real_segments = [](const double &H1a, const double &H1b, + const double &H2a, const double &H2b, const double &D1a, + const double &D1b, const double &D2a, const double &D2b, + const double &tol){ + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && + abs((H2a - H2b) / H2a) < tol && + abs(((D2a - D1a) - (D2b - D1b)) / (D2a - D1a + 1e-30)) < tol; + return similarity; + }; + // Condition for equivalence of the image part of the FLS solution + auto compare_image_segments = [](const double &H1a, const double &H1b, + const double &H2a, const double &H2b, const double &D1a, + const double &D1b, const double &D2a, const double &D2b, + const double &tol) { + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && + abs((H2a - H2b) / H2a) < tol && + abs(((D2a + D1a) - (D2b + D1b)) / (D2a + D1a + 1e-30)) < tol; + return similarity; + }; + // Condition for equivalence of the full FLS solution + auto compare_realandimage_segments = [](const double &H1a, + const double &H1b, const double &H2a, const double &H2b, + const double &D1a, const double &D1b, const double &D2a, + const double &D2b, const double &tol) { + bool similarity; + similarity = abs((H1a - H1b) / H1a) < tol && + abs((H2a - H2b) / H2a) < tol && + abs((D1a - D1b) / (D1a + 1e-30)) < tol && + abs((D2a - D2b) / (D2a + 1e-30)) < tol; + return similarity; + }; + + string real("real"); + string image("image"); + string realandimage("realandimage"); + + // compare segments is a pointer to one of the lambda functions + bool(*compare_segments)(const double&, const double&, const double&, + const double&, const double&, const double&, const double&, + const double&, const double&); + + if (real.compare(kind) == 0) { + compare_segments = compare_real_segments; + } else if (image.compare(kind) == 0) { + compare_segments = compare_image_segments; + } else if (realandimage.compare(kind) == 0) { + compare_segments = compare_realandimage_segments; + } else { + throw invalid_argument("Error kind not implemented yet."); + } + + SimT.nSim = 1; + tuple doub_tup_temp_H; + tuple doub_tup_temp_D; + tuple int_tup_temp_sim; + vector< tuple > vect_w_tup(1); + vect_w_tup[0] = pairs[0]; +// tuple pair0 = pairs[0]; + int i0 = get<0>(pairs[0]); + int j0 = get<1>(pairs[0]); + SimT.Sim.push_back(vect_w_tup); + doub_tup_temp_H = make_tuple(boreSegments[i0].H, boreSegments[j0].H); + SimT.HSim.push_back(doub_tup_temp_H); + doub_tup_temp_D = make_tuple(boreSegments[i0].D, boreSegments[j0].D); + SimT.DSim.push_back(doub_tup_temp_D); + + // values used in loops + int ibor; + int jbor; + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + double H1; + double H2; + double D1; + double D2; + + // Cycle through all pairs of boreholes for the given distance + for (size_t i=1; i(pairs[i]); + jbor = get<1>(pairs[i]); + if (ibor > jbor) { + swap(ibor, jbor); + } + b1 = boreSegments[ibor]; + b2 = boreSegments[jbor]; + // Verify if the current pair should be included in the previously identified symmetries + for (int j=0; j(SimT.HSim[j]); + H2 = get<1>(SimT.HSim[j]); + D1 = get<0>(SimT.DSim[j]); + D2 = get<1>(SimT.DSim[j]); + if (compare_segments(H1, b1.H, H2, b2.H, D1, b1.D, D2, b2.D, + tol)) { + int_tup_temp_sim = make_tuple(ibor, jbor); + SimT.Sim[j].push_back(int_tup_temp_sim); + break; + } else if (compare_segments(H1, b2.H, H2, b1.H, D1, b2.D, D2, + b1.D, tol)) { + int_tup_temp_sim = make_tuple(jbor, ibor); + SimT.Sim[j].push_back(int_tup_temp_sim); + break; + } else if (j == SimT.nSim-1) { + SimT.nSim++; + int_tup_temp_sim = make_tuple(ibor, jbor); + vect_w_tup[0] = int_tup_temp_sim; + SimT.Sim.push_back(vect_w_tup); + doub_tup_temp_H = make_tuple(b1.H, b2.H); + SimT.HSim.push_back(doub_tup_temp_H); + doub_tup_temp_D = make_tuple(b1.D, b2.D); + SimT.DSim.push_back(doub_tup_temp_D); + break; + } + } // next j + } // next i + } // Similarity::_similarities_one_distance + +} // namespace gt::boreholes diff --git a/third_party/cpgfunctionEP-0.2/src/coordinates.cpp b/third_party/cpgfunctionEP-0.2/src/coordinates.cpp new file mode 100644 index 00000000000..1f0e496f7e6 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/coordinates.cpp @@ -0,0 +1,192 @@ +// +// Created by jackcook on 5/6/21. +// + +#include +#include +#include +#include + +using namespace std; + +namespace gt::coordinates { + + vector> configuration(const string& shape, int Nx, + int Ny, double Bx, double By){ + // define acceptable inputs + vector acceptable_arguments{"Rectangle", + "OpenRectangle", + "U", + "L"}; + // check if the input string is acceptable + bool acceptable = (find(acceptable_arguments.begin(), + acceptable_arguments.end(), + shape) != acceptable_arguments.end()); + + if (!acceptable) { + throw std::invalid_argument("The shape described (" + + shape + ") is not an available input for " + "gt::coordinates::configuration()."); + } + + std::vector> coordinates; + + if (shape == "Rectangle") { + coordinates = rectangle(Nx, Ny, Bx, By); + } else if (shape == "OpenRectangle") { + coordinates = Open_rectangle(Nx, Ny, Bx, By); + } else if (shape == "U") { + coordinates = U_shape(Nx, Ny, Bx, By); + } else if (shape == "L") { + coordinates = L_shape(Nx, Ny, Bx, By); + } + + return coordinates; + } // configuration(); + +// vector> configuration(const string& shape, +// const string& input_path){ +// // define acceptable inputs +// std::vector acceptable_arguments{"custom"}; +// // check if the input string is acceptable +// bool acceptable = (std::find(acceptable_arguments.begin(), +// acceptable_arguments.end(), shape) != +// acceptable_arguments.end()); +// +// if (!acceptable) { +// throw invalid_argument("The shape described (" + +// shape + ") is not an available input for " +// "gt::coordinates::configuration()."); +// } +// +// vector> custom = +// import_coordinates_from_file(input_path); +// +// return custom; +// } // configuration(); + + vector> rectangle(int Nx, int Ny, double Bx, + double By) { + std::vector> r; // rectangular coordinates + + int nbh = Nx * Ny; + r.reserve(nbh); + for (int i = 0; i < Nx; i++) { + for (int j = 0; j < Ny; j++) { + r.emplace_back(i * Bx, j * By); + } // next j + } // next i + + return r; + } // rectangle(); + + vector> Open_rectangle(int Nx, int Ny, double Bx, + double By){ + vector> open_r; // open rectangle coordinates + + int nbh; + if (Nx > 2 && Ny > 2) { + nbh = Ny * 2 + (Nx - 2) * 2; + open_r.reserve(nbh); + for (int i = 0; i < Nx; i++) { + open_r.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny - 1; j++) { + open_r.emplace_back(0., j * By); + open_r.emplace_back((Nx-1) * Bx, j * By); + } // next j + for (int i = 0; i < Nx; i++) { + open_r.emplace_back(i * Bx, (Ny-1)*By); + } // next i + } // if (Nx > 2 && Ny > 2) + else { + nbh = Nx * Ny; + open_r.reserve(nbh); + open_r = rectangle(Nx, Ny, Bx, By); + } // else() + + return open_r; + } // Open_rectangle(); + + vector> U_shape(int Nx, int Ny, double Bx, double By){ + std::vector> U; // U-shape coordinates + + int nbh; + if (Nx > 2 && Ny > 1) { + nbh = 2 * Ny + (Nx - 2); + U.reserve(nbh); + for (int i = 0; i < Nx; i++){ + U.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny; j++) { + U.emplace_back(0., j * By); + U.emplace_back((Nx-1) * Bx, j * By); + } // next j + } // if (Nx > 2 && Ny > 1) + else { + nbh = Nx * Ny; + U.reserve(nbh); + U = rectangle(Nx, Ny, Bx, By); + } // else() + + return U; + } // U_shape(); + + vector> L_shape(int Nx, int Ny, double Bx, double By){ + std::vector> L; // L-shape coordinates + + int nbh = Nx + Ny - 1; + L.reserve(nbh); // reserve for "emplace_back" + for (int i = 0; i < Nx; i++){ + L.emplace_back(i * Bx, 0.); + } // next i + for (int j = 1; j < Ny; j++){ + L.emplace_back(0., j * By); + } // next j + + return L; + } // L_shape(); + +// vector> import_coordinates_from_file( +// const string& input_path){ +// vector> custom; // custom-shape from .json file +// +// // nlohmann json input +// std::ifstream in(input_path); +// nlohmann::json js; +// in >> js; +// +// std::vector x = js["x"]; +// std::vector y = js["y"]; +// +// custom.reserve(x.size()); +// for (int i = 0; i < x.size(); i++){ +// custom.emplace_back(x[i], y[i]); +// } +// +// return custom; +// } // import_coordinates_from_file(); +// +// void export_coordinates_to_file( +// const vector> &coordinates, +// const std::string& output_path){ +// // Use nlohmann json to export the coordinates to a path +// +// std::vector x_values(coordinates.size()); +// std::vector y_values(coordinates.size()); +// +// for (int i=0; i(coordinates[i]); +// y_values[i] = std::get<1>(coordinates[i]); +// } +// +// std::ofstream o(output_path); +// +// nlohmann::json j; +// +// j["x"] = x_values; +// j["y"] = y_values; +// +// o << std::setw(4) << j << std::endl; +// } // export_coordinates_to_file(); +} // namespace gt::coordinates \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.2/src/gfunction.cpp b/third_party/cpgfunctionEP-0.2/src/gfunction.cpp new file mode 100644 index 00000000000..a64d432a9c9 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/gfunction.cpp @@ -0,0 +1,507 @@ +// -*- lsst-c++ -*- + +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include +#include +#include + + +using namespace std; // lots of vectors, only namespace to be used + +namespace gt { namespace gfunction { + // The uniform borehole wall temperature (UBWHT) g-function calculation. + // Originally presented in Cimmino and Bernier (2014) and a later paper on + // speed improvements by Cimmino (2018) + vector uniform_borehole_wall_temperature( + vector &boreField, vector &time, + double alpha, int nSegments, bool use_similarities, bool adaptive, + int n_Threads, bool display){ + vector gFunction(time.size()); + + if (display) { + std::cout << "---------------------------------------------------" + "---------" << std::endl; + std::cout << "Calculating g-function for uniform borehole wall " + "temperature" << std::endl; + std::cout << "---------------------------------------------------" + "---------" << std::endl; + } + auto startall = std::chrono::steady_clock::now(); + + if (display) { + cout << "\tMaking use of " << n_Threads << " threads." << endl; + } + + // Number of boreholes + int nbh = boreField.size(); + // Total number of line sources + int nSources = nSegments * nbh; + // Number of time values + int nt = time.size(); + + auto sum_to_n = [](const int n) { + return n * (n + 1) / 2; + }; + int nSum = sum_to_n(nSources); + + // Segment Response struct + gt::heat_transfer::SegmentResponse SegRes(nSources, nSum, nt); + + // Split boreholes into segments + _borehole_segments(SegRes.boreSegments, boreField, nSegments); + + // Calculate segment to segment thermal response factors + auto start = std::chrono::steady_clock::now(); + gt::heat_transfer::thermal_response_factors(SegRes, + time, + alpha, + use_similarities, + display); + auto end = std::chrono::steady_clock::now(); + + if (display) { + std::cout << "Building and solving system of equations ..." + << std::endl; + } + // --------------------------------------------------------------------- + // Build a system of equation [A]*[X] = [B] for the evaluation of the + // g-function. [A] is a coefficient matrix, [X] = [Qb,Tb] is a state + // space vector of the borehole heat extraction rates and borehole wall + // temperature (equal for all segments), [B] is a coefficient vector. + // --------------------------------------------------------------------- + + // -------- timings for debug + double milli = 0; + double segment_length_time = 0; + double time_vector_time = 0; + double segment_h_values_time = 0; + double fill_A_time = 0; + double load_history_reconstruction_time = 0; + double temporal_superposition_time = 0; + double fill_gsl_matrices_time = 0; + double LU_decomposition_time = 0; + + auto start2 = std::chrono::steady_clock::now(); + + // ------ Segment lengths ------- + start = std::chrono::steady_clock::now(); + std::vector Hb(nSources); + # pragma omp parallel for num_threads(n_Threads) + for (int b=0; b + (end - start).count(); + segment_length_time += milli; + + // ------ time vectors --------- + start = std::chrono::steady_clock::now(); + // create new time vector that starts at 0 + std::vector _time_untouched(time.size()+1); + std::vector _time(time.size()+1); + std::vector dt(time.size()+1); + +// auto _fill_time = [&_time, &time, &dt, &_time_untouched]() { +// for (int i=0; i<_time.size(); i++) { +// if (i==0) { +// _time[0] = 0; +// _time_untouched[0] = 0; +// dt[i] = time[i]; +// } else { +// _time[i] = time[i-1]; +// _time_untouched[i] = time[i-1]; +// dt[i] = time[i] - time[i-1]; +// } // fi +// } // next i +// }; // auto _fill_time +// if (multi_thread) { +// boost::asio::post(pool, [&_fill_time]{ _fill_time() ;}); +// } else { +// _fill_time(); +// } // if (multi_thread); + # pragma omp parallel for num_threads(n_Threads) + for (int i=0; i<_time.size(); i++) { + if (i==0) { + _time[0] = 0; + _time_untouched[0] = 0; + dt[i] = time[i]; + } else { + _time[i] = time[i-1]; + _time_untouched[i] = time[i-1]; + if (i < time.size()) { + dt[i] = time[i] - time[i-1]; + } else if (i == time.size()) { + dt[i] = time[i-1] - time[i-2]; + } + } // fi + } // next i + + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + time_vector_time += milli; + + // ---------- segment h values ------------- + /** Starting up pool2 here **/ + // Launch the pool with n threads. + auto tic = std::chrono::steady_clock::now(); + auto toc = std::chrono::steady_clock::now(); + if (display) { + double milli = chrono::duration_cast + (tic - toc).count(); + double seconds = milli; + std::cout << "Time to open a pool : " + << seconds + << " sec" << std::endl; + } + + start = std::chrono::steady_clock::now(); + + end = chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + segment_h_values_time += milli; + + // after interpolation scheme, get rid of h_ij first + // Initialize segment heat extraction rates + vector > Q(nSources, vector (nt)); + + // Define A and b for utitilizing Ax=b + /** + * A = [ [ ], + * [hb[0:len(hb), 0] + * ] + * b = [ [ ], + * [sum(hb)] + * ] + * **/ + + int SIZE = nSources + 1; + Eigen::MatrixXd A(SIZE, SIZE); + Eigen::MatrixXd B(SIZE, 1); + Eigen::MatrixXd X(SIZE, 1); + + // Fill A + int n = SIZE - 1; + double Hb_sum=0; + for (auto & _hb : Hb) { + Hb_sum += _hb; + } + + // Build and solve the system of equations at all times + + // the loop p=n depends on what occured at p=n-1, so this will be be in + // series however threading will be interspersed throughout to make use + // of as many threads as possible + std::vector Tb_0 (nSources); + // Restructured load history + // create interpolation object for accumulated heat extraction + std::vector> + q_reconstructed (nSources, std::vector (nt)); + std::vector q_r(nSources * nt, 0); + + int gauss_sum = nSources * (nSources + 1) / 2; + std::vector H_ij(gauss_sum * nt, 0); // 1D nSources x nt + int idx; + for (int i=0; i + (end - start).count(); + fill_A_time += milli; + + // ----- load history reconstruction ------- + start = std::chrono::steady_clock::now(); + load_history_reconstruction(q_r,time, _time, Q, dt, p); + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + load_history_reconstruction_time += milli; + + // ----- temporal superposition + start = chrono::steady_clock::now(); + _temporal_superposition(Tb_0, + SegRes, + H_ij, + q_r, + p, + nSources); + // fill b with -Tb + B(SIZE-1, 0) = Hb_sum; + for (int i=0; i + (end - start).count(); + temporal_superposition_time += milli; + + int m = SIZE; + int n = SIZE; + vector x(SIZE); +// _solve_eqn(x, A, b); + /** was _solve_eqn **/ + + // ---- fill gsl matrix A and b ----- + start = std::chrono::steady_clock::now(); + + end = std::chrono::steady_clock::now(); + milli = chrono::duration_cast + (end - start).count(); + fill_gsl_matrices_time += milli; + + // ----- LU decomposition ----- + start = std::chrono::steady_clock::now(); + X = A.lu().solve(B); + + for (int i=0; i + (end - start).count(); + LU_decomposition_time += milli; + + // ---- Save Q's for next p --- + for (int j=0; j + (end2 - start2).count(); + double seconds1 = milli1 / 1000; + double milli2 = chrono::duration_cast + (end2 - startall).count(); + double seconds2 = milli2 / 1000; + std::cout << "Elapsed time in seconds : " + << seconds1 + << " sec" << std::endl; + std::cout << "Total time for g-function evaluation : " + << seconds2 + << " sec" << std::endl; + } + + return gFunction; + } // uniform_borehole_wall_temperature(); + + void _borehole_segments(std::vector& boreSegments, + std::vector& boreholes, const int nSegments) { + double H; + double D; + int count = 0; + // Split boreholes into segments + for(auto& b : boreholes) { + // TODO: maybe thread this later on + for (int i=0; i& q_reconstructed, + vector& time, vector& _time, + vector >& Q, vector& dt, const int p) { + // for s in range p+1 + int nSources = Q.size(); + + // Inverted time steps + std::vector dt_reconstructed (p+1); + for (int i=p; i>=0; i--) { + dt_reconstructed[p-i] = dt[i]; // reverse the dt + } + // t_restructured is [0, cumsum(dt_reversed)] + std::vector t_reconstructed(p+2); // will start at 0 + for (int i=1; i<=p+1; i++) { + t_reconstructed[i] = dt_reconstructed[i-1]; + } + for (int i=1; i<=p+1; i++) { + t_reconstructed[i] = t_reconstructed[i] + t_reconstructed[i-1]; + } + // local time vector + std::vector t(p+3); + for (int i=0; i> Q_dt (nSources, + std::vector (t.size())); + auto _Q_dot_dt = [&Q_dt, &Q, &dt, &p, &_tsize, &t](const int i) { + for (int j = 1; j<_tsize; j++) { + if (j>=p+1) { + Q_dt[i][j] = Q_dt[i][j-1]; + } else { + Q_dt[i][j] = Q[i][j-1] * dt[j-1] + Q_dt[i][j-1]; + } // fi + } // next j + }; + for (int i=0; i y(n); + for (int j=0; j yp(n2); + jcc::interpolation::interp1d(t_reconstructed, yp, t, y); + + int idx; + for (int j=0; j& Tb_0, + gt::heat_transfer::SegmentResponse &SegRes, + vector &h_ij, + vector &q_reconstructed, + const int p, int &nSources) + { + // This function performs equation (37) of Cimmino (2017) + std::fill(Tb_0.begin(), Tb_0.end(), 0); + // Number of time steps + int nt = p + 1; + + const auto processor_count = thread::hardware_concurrency(); + int n_threads = int(processor_count); + + int gauss_sum = nSources * (nSources + 1) / 2; // Number of positions in packed symmetric matrix + // Storage of h_ij differences + std::vector dh_ij(gauss_sum, 0); + int begin_1; // integer declarations for where the linear algebra will begin + int begin_2; + int begin_q; // time for q_reconstructed to begin + + double alpha = 1; + double alpha_n = -1; + + std::vector::iterator begin_it_1; + std::vector::iterator end_it_1; + + for (int k = 0; k < nt; k++) { + begin_1 = k * gauss_sum; + begin_it_1 = h_ij.begin() + begin_1; + end_it_1 = h_ij.begin() + begin_1 + gauss_sum; + if (k==0){ + // dh_ij = h(k) + std::copy(begin_it_1, end_it_1, dh_ij.begin()); + } else { + begin_2 = (k-1) * gauss_sum; + // h_1 -> dh_ij + std::copy(begin_it_1, end_it_1, dh_ij.begin()); + // dh_ij = -1 * h(k) + h(k-1) + jcc::blas::axpy(gauss_sum, alpha_n, h_ij, dh_ij, begin_2, + n_threads); + } + // q_reconstructed(t_k - t_k') + begin_q = (nt - k - 1) * nSources; + // dh_ij is a lower triangular packed matrix + // Tb_0 = 1 * dh_ij * q(t_k-t_k') + 1 * Tb_0 + jcc::blas::spmv(nSources, alpha, dh_ij, q_reconstructed, alpha, + Tb_0, begin_q, n_threads); + } // next k + } // _temporal_superposition(); +} } // namespace gt::gfunction diff --git a/third_party/cpgfunctionEP-0.2/src/heat_transfer.cpp b/third_party/cpgfunctionEP-0.2/src/heat_transfer.cpp new file mode 100644 index 00000000000..9813b040acb --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/heat_transfer.cpp @@ -0,0 +1,270 @@ +// +// Created by jackcook on 7/11/20. +// + +#include +#include +#include +#include +#include +#include + +using namespace gt; +using namespace std; + +namespace gt::heat_transfer { + + double finite_line_source(const double time_, const double alpha, + boreholes::Borehole &b1, boreholes::Borehole &b2, + bool reaSource, bool imgSource) { + + auto _Ils = [&b1, &b2, reaSource, imgSource](const double s) { + auto _erfint = [](const double x) { + return x * std::erf(x) - (1 / sqrt(M_PI)) * (1 - exp(-pow(x, 2))); + }; + double r = b1.distance(b2); + double func = 0.; + // function to integrate + if (reaSource) { + // Real part of the FLS solution + func += _erfint(double(b2.D - b1.D + b2.H) * s); + func += -_erfint(double(b2.D - b1.D) * s); + func += _erfint(double(b2.D - b1.D - b1.H) * s); + func += -_erfint(double(b2.D - b1.D + b2.H - b1.H) * s); + } // fi reaSource + if (imgSource) { + // Image part of the FLS solution + func += _erfint(double(b2.D + b1.D + b2.H) * s); + func += -_erfint(double(b2.D + b1.D) * s); + func += _erfint(double(b2.D + b1.D + b1.H) * s); + func += -_erfint(double(b2.D + b1.D + b2.H + b1.H) * s); + } // fi imgSource + double a = 0.5 / (b2.H * pow(s, 2)) * func * exp(-pow(r, 2) * pow(s, 2)); + return a; + }; // auto _Ils + + // lower bound of integration + double a = double(1.) / sqrt(double(4.) * alpha * time_); + // Evaluate the integral using Gauss-Kronrod + double result; + auto method = qdt::adaptive(qdt::gauss_kronrod()); + result = method.integrate(_Ils, a, qdt::INF); + + return result; + } // void finite_line_source + + void thermal_response_factors(SegmentResponse &SegRes, + vector &time, const double alpha, + bool use_similaries, bool disp) { + // total number of line sources + int nSources = SegRes.boreSegments.size(); + // number of time values + int nt = time.size(); + + // Open up processes here + // Create a vector of threads + //may return 0 when not able to detect + const auto processor_count = thread::hardware_concurrency(); + if (disp) { + cout << "\tDetected " << processor_count + << " as the number of available threads" << endl; + } + + gt::boreholes::SimilaritiesType SimReal; // positive + gt::boreholes::SimilaritiesType SimImage; // negative + + auto sum_to_n = [](const int n) { + return n * (n + 1) / 2; + }; + if (use_similaries) { + auto start = std::chrono::steady_clock::now(); + // Calculations with similarities + if (disp) { + cout << "Identifying similarities..." << endl; + } + bool splitRealAndImage = true; + double disTol = 0.1; + double tol = 1.0e-6; + gt::boreholes::Similarity sim; + sim.similarities(SimReal, SimImage, SegRes.boreSegments, + splitRealAndImage, disTol, tol); + + // --- + // Adaptive hashing scheme if statement + // Determine the Segment Response storing mode here + int Ntot = sum_to_n(nSources); + + // lambda function for calculating h at each time step + auto _calculate_h = [&SegRes, &splitRealAndImage, &time, &alpha, + &nt](boreholes::SimilaritiesType &SimReal, + int s, bool reaSource, + bool imgSource) { + // begin function + int n1; + int n2; + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + // begin thread + n1 = get<0>(SimReal.Sim[s][0]); + n2 = get<1>(SimReal.Sim[s][0]); + b1 = SegRes.boreSegments[n1]; + b2 = SegRes.boreSegments[n2]; + vector hPos(nt); + if (splitRealAndImage) { + for (int k=0; k(SimReal.Sim[s][k]); + j = get<1>(SimReal.Sim[s][k]); + for (std::size_t t=0; t(end - start).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " << seconds + << " sec" << std::endl; + std::cout << "Calculating segment to segment response " + "factors ..." << std::endl; + } // end if + + // inputs + bool reaSource; + bool imgSource; + # pragma omp parallel for num_threads(processor_count) + for (int s=0; s(end2 - end).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " + << seconds + << " sec" << std::endl; + } + } else { + if (disp) { + std::cout << "Calculating segment to segment response factors ..." << std::endl; + } // end if + auto start = std::chrono::steady_clock::now(); + bool sameSegment; + bool otherSegment; + + auto _fill_line = [&SegRes, &time](const int i, const int j, + const double alpha, bool sameSegment, bool otherSegment) { + double h; + int index; + gt::boreholes::Borehole b1; + gt::boreholes::Borehole b2; + b2 = SegRes.boreSegments[i]; + for (std::size_t k = 0; k < time.size(); k++) { + double t = time[k]; + if (!otherSegment){ + if (sameSegment) { + b1 = SegRes.boreSegments[i]; + h = finite_line_source(t, alpha, b2, b2); + } + } else if (otherSegment && !sameSegment) { + b1 = SegRes.boreSegments[j]; + h = finite_line_source(t, alpha, b1, b2); + } else { + throw std::invalid_argument( "sameSegment and otherSegment cannot both be true" ); + } // end if + SegRes.get_index_value(index, i, j); + SegRes.h_ij[index][k] = h; + }; // end for + }; // auto _fill_line + #pragma omp parallel for num_threads(processor_count) + for (int i = 0; i < nSources; i++) { + // Segment to same-segment thermal response factor + // FLS solution for combined real and image sources + sameSegment = true; + otherSegment = false; + _fill_line(i, i, alpha, sameSegment, otherSegment); + // Segment to other segment thermal response factor + for (int j = i + 1; j(end - start).count(); + double seconds = double(milli) / 1000; + std::cout << "Elapsed time in seconds : " + << seconds + << " sec" << std::endl; + } + // Iterate over the thread vector + } // fi similarity + } // void thermal_response_factors + + void SegmentResponse::get_h_value(double &h, const int i, const int j, const int k) { + int index; + switch (storage_mode) { + case 0 : + cout << "Case 0 not written yet" << endl; + break; + case 1 : + if (i <= j) { + get_index_value(index, i, j); + h = h_ij[index][k]; + } else { + get_index_value(index, j, i); + h = boreSegments[j].H/boreSegments[i].H * h_ij[index][k]; + } + break; + default: + throw invalid_argument("The case selected is not currently implemented."); + } // switch(); + } // SegmentResponse::get_h_value(); + + void SegmentResponse::get_index_value(int &index, const int i, const int j) { + index = i * (2*nSources - i - 1) / 2 + j; + } // SegmentResponse::get_index_value(); + +} // namespace gt::heat_transfer \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.2/src/interpolation.cpp b/third_party/cpgfunctionEP-0.2/src/interpolation.cpp new file mode 100644 index 00000000000..f48cf914976 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/interpolation.cpp @@ -0,0 +1,91 @@ +// +// Created by jackcook on 7/15/20. +// + +#include +#include + +using namespace std; + +namespace jcc { namespace interpolation { + + double linterp(double xp, double x0, double y0, double x1, double y1) { + double yp; + yp = y0 + ((y1-y0) / (x1-x0)) * (xp-x0); + return yp; + } // interp + + void interp1d(vector& xp, vector& yp, vector& x, vector& y) { + int counter = 0; + for (int i=0; i x[x.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + for (int j = counter; j= x[j] && xp[i] <= x[j+1]) { + yp[i] = linterp(xp[i], x[j], y[j], x[j + 1], y[j + 1]); + break; + } else { + counter++; + } // fi + } // next j + } // next i + } // interp1d + + void interp1d(double &xp, double &yp, vector& x, vector& y) { + int counter = 0; + + if (xp < x[0] || xp > x[x.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + for (int j = counter; j= x[j] && xp <= x[j+1]) { + yp = linterp(xp, x[j], y[j], x[j + 1], y[j + 1]); + return; + } else { + counter++; + } // fi + } // next j + } // interp1d + + void interp1d(double &xp, double &yp, vector &time, + gt::heat_transfer::SegmentResponse &SegRes, int &i, int &j, int &k) { + // if the x point is out of bounds, then tell the user that extrapolation is not possible + if (xp < 0 || xp > time[time.size()-1]) { + throw invalid_argument("Need to add extrapolation"); + } + // if the time value falls in between 0 and the first time value, then interpolate 0 to t1 + if (0 < xp && xp < time[0]) { + double h; + SegRes.get_h_value(h, i, j, k); + yp = linterp(xp, 0, 0, time[0], h); + return; + } + // loop until the value for interpolation is found + int counter=0; + double h1; + SegRes.get_h_value(h1, i, j, 0); + double h2; + SegRes.get_h_value(h2, i, j, 1); + for (int k=counter; k=time[k] && xp <=time[k+1]) { + yp = linterp(xp, time[k], h1, time[k+1], h2); + return; + } else { + counter++; + h1 = h2; + SegRes.get_h_value(h2, i, j, counter+1); + } // else() + + } // next k + } // interp1d(); + +} } // jcc::interpolation + + diff --git a/third_party/cpgfunctionEP-0.2/src/statistics.cpp b/third_party/cpgfunctionEP-0.2/src/statistics.cpp new file mode 100644 index 00000000000..96320c13f71 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/statistics.cpp @@ -0,0 +1,25 @@ +// +// Created by jackcook on 5/10/21. +// + +#include + +namespace gt { + namespace statistics { + + double root_mean_square_error(std::vector actual, std::vector predicted) { + // Return the root mean square error between two g-function curves + + double square_error = 0; + + for (int i = 0; i < actual.size(); i++) { + square_error += std::pow(((actual[i] - predicted[i]) / actual[i]), 2); + } // next i + + double rmse = std::sqrt(square_error / double(actual.size())); + + return rmse; + } + + } // namespace statistics +} // namespace gt \ No newline at end of file diff --git a/third_party/cpgfunctionEP-0.2/src/utilities.cpp b/third_party/cpgfunctionEP-0.2/src/utilities.cpp new file mode 100644 index 00000000000..d331ad83bd9 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/src/utilities.cpp @@ -0,0 +1,85 @@ +// +// Created by jackcook on 7/11/20. +// + +#include + +namespace gt { + namespace utilities { + std::vector time_geometric(double dt, double tmax, int Nt) { + std::vector time(Nt); // create a time vector of size Nt + + double value; + double tmax_calc = double(Nt) * double(dt); + if (tmax > tmax_calc) { + double dr = 1.0e99; + double r = 2.; + while (std::abs(dr) > 1.0e-10) { + dr = std::pow(1+tmax/double(dt)*(r-1), 1/double(Nt)) - r; + r += dr; + } // end while + for (int j=0; j < time.size(); j++) { + value = 1 - std::pow(r, double(j+1)); + value = value * double(dt) / (1 - r); + time[j] = value; + } // end for + } else { + for (int j=0; j < time.size(); j++) { + value = double(dt) * double(j + 1); + time[j] = value; + } // end for + } // end if + return time; + } // vector time_geometric + + std::vector Eskilson_original_points() { + // Eskilsons original 27 time steps + std::vector logtime = {-8.5, -7.8, -7.2, -6.5, -5.9, -5.2, -4.5, -3.963, -3.27, -2.864,-2.577, + -2.171, -1.884, -1.191, -0.497, -0.274, -0.051, 0.196, 0.419, 0.642, 0.873, + 1.112, 1.335, 1.679, 2.028, 2.275, 3.003}; + return logtime; + + } + + std::vector time_Eskilson(const double &H, const double &alpha){ + std::vector logtime = Eskilson_original_points(); + std::vector time = convert_time(logtime, H, alpha); + + return time; + } // time_Eskilson(); + + std::vector convert_time(std::vector &logtime, const double &H, const double &alpha) { + int nt = logtime.size(); + std::vector time(nt); + + double ts = pow(H, 2) / (9 * alpha); + for (int i=0; i &logtime){ + int np = 31; // 31 total points + if (logtime.size() != np) { + logtime.resize(np); + } + for (int i=1; i &logtime, std::vector &time, const double H, const double alpha) { + int nt = logtime.size(); + if (time.size() != nt) { + time.resize(nt); + } + double ts = pow(H, 2) / (9 * alpha); + for (int i=0; i +// Copyright (C) 2007-2011 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CORE_H +#define EIGEN_CORE_H + +// first thing Eigen does: stop the compiler from committing suicide +#include "src/Core/util/DisableStupidWarnings.h" + +#if defined(__CUDACC__) && !defined(EIGEN_NO_CUDA) + #define EIGEN_CUDACC __CUDACC__ +#endif + +#if defined(__CUDA_ARCH__) && !defined(EIGEN_NO_CUDA) + #define EIGEN_CUDA_ARCH __CUDA_ARCH__ +#endif + +#if defined(__CUDACC_VER_MAJOR__) && (__CUDACC_VER_MAJOR__ >= 9) +#define EIGEN_CUDACC_VER ((__CUDACC_VER_MAJOR__ * 10000) + (__CUDACC_VER_MINOR__ * 100)) +#elif defined(__CUDACC_VER__) +#define EIGEN_CUDACC_VER __CUDACC_VER__ +#else +#define EIGEN_CUDACC_VER 0 +#endif + +// Handle NVCC/CUDA/SYCL +#if defined(__CUDACC__) || defined(__SYCL_DEVICE_ONLY__) + // Do not try asserts on CUDA and SYCL! + #ifndef EIGEN_NO_DEBUG + #define EIGEN_NO_DEBUG + #endif + + #ifdef EIGEN_INTERNAL_DEBUGGING + #undef EIGEN_INTERNAL_DEBUGGING + #endif + + #ifdef EIGEN_EXCEPTIONS + #undef EIGEN_EXCEPTIONS + #endif + + // All functions callable from CUDA code must be qualified with __device__ + #ifdef __CUDACC__ + // Do not try to vectorize on CUDA and SYCL! + #ifndef EIGEN_DONT_VECTORIZE + #define EIGEN_DONT_VECTORIZE + #endif + + #define EIGEN_DEVICE_FUNC __host__ __device__ + // We need cuda_runtime.h to ensure that that EIGEN_USING_STD_MATH macro + // works properly on the device side + #include + #else + #define EIGEN_DEVICE_FUNC + #endif + +#else + #define EIGEN_DEVICE_FUNC + +#endif + +// When compiling CUDA device code with NVCC, pull in math functions from the +// global namespace. In host mode, and when device doee with clang, use the +// std versions. +#if defined(__CUDA_ARCH__) && defined(__NVCC__) + #define EIGEN_USING_STD_MATH(FUNC) using ::FUNC; +#else + #define EIGEN_USING_STD_MATH(FUNC) using std::FUNC; +#endif + +#if (defined(_CPPUNWIND) || defined(__EXCEPTIONS)) && !defined(__CUDA_ARCH__) && !defined(EIGEN_EXCEPTIONS) && !defined(EIGEN_USE_SYCL) + #define EIGEN_EXCEPTIONS +#endif + +#ifdef EIGEN_EXCEPTIONS + #include +#endif + +// then include this file where all our macros are defined. It's really important to do it first because +// it's where we do all the alignment settings (platform detection and honoring the user's will if he +// defined e.g. EIGEN_DONT_ALIGN) so it needs to be done before we do anything with vectorization. +#include "src/Core/util/Macros.h" + +// Disable the ipa-cp-clone optimization flag with MinGW 6.x or newer (enabled by default with -O3) +// See http://eigen.tuxfamily.org/bz/show_bug.cgi?id=556 for details. +#if EIGEN_COMP_MINGW && EIGEN_GNUC_AT_LEAST(4,6) + #pragma GCC optimize ("-fno-ipa-cp-clone") +#endif + +#include + +// this include file manages BLAS and MKL related macros +// and inclusion of their respective header files +#include "src/Core/util/MKL_support.h" + +// if alignment is disabled, then disable vectorization. Note: EIGEN_MAX_ALIGN_BYTES is the proper check, it takes into +// account both the user's will (EIGEN_MAX_ALIGN_BYTES,EIGEN_DONT_ALIGN) and our own platform checks +#if EIGEN_MAX_ALIGN_BYTES==0 + #ifndef EIGEN_DONT_VECTORIZE + #define EIGEN_DONT_VECTORIZE + #endif +#endif + +#if EIGEN_COMP_MSVC + #include // for _aligned_malloc -- need it regardless of whether vectorization is enabled + #if (EIGEN_COMP_MSVC >= 1500) // 2008 or later + // Remember that usage of defined() in a #define is undefined by the standard. + // a user reported that in 64-bit mode, MSVC doesn't care to define _M_IX86_FP. + #if (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || EIGEN_ARCH_x86_64 + #define EIGEN_SSE2_ON_MSVC_2008_OR_LATER + #endif + #endif +#else + // Remember that usage of defined() in a #define is undefined by the standard + #if (defined __SSE2__) && ( (!EIGEN_COMP_GNUC) || EIGEN_COMP_ICC || EIGEN_GNUC_AT_LEAST(4,2) ) + #define EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC + #endif +#endif + +#ifndef EIGEN_DONT_VECTORIZE + + #if defined (EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC) || defined(EIGEN_SSE2_ON_MSVC_2008_OR_LATER) + + // Defines symbols for compile-time detection of which instructions are + // used. + // EIGEN_VECTORIZE_YY is defined if and only if the instruction set YY is used + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_SSE + #define EIGEN_VECTORIZE_SSE2 + + // Detect sse3/ssse3/sse4: + // gcc and icc defines __SSE3__, ... + // there is no way to know about this on msvc. You can define EIGEN_VECTORIZE_SSE* if you + // want to force the use of those instructions with msvc. + #ifdef __SSE3__ + #define EIGEN_VECTORIZE_SSE3 + #endif + #ifdef __SSSE3__ + #define EIGEN_VECTORIZE_SSSE3 + #endif + #ifdef __SSE4_1__ + #define EIGEN_VECTORIZE_SSE4_1 + #endif + #ifdef __SSE4_2__ + #define EIGEN_VECTORIZE_SSE4_2 + #endif + #ifdef __AVX__ + #define EIGEN_VECTORIZE_AVX + #define EIGEN_VECTORIZE_SSE3 + #define EIGEN_VECTORIZE_SSSE3 + #define EIGEN_VECTORIZE_SSE4_1 + #define EIGEN_VECTORIZE_SSE4_2 + #endif + #ifdef __AVX2__ + #define EIGEN_VECTORIZE_AVX2 + #endif + #ifdef __FMA__ + #define EIGEN_VECTORIZE_FMA + #endif + #if defined(__AVX512F__) && defined(EIGEN_ENABLE_AVX512) + #define EIGEN_VECTORIZE_AVX512 + #define EIGEN_VECTORIZE_AVX2 + #define EIGEN_VECTORIZE_AVX + #define EIGEN_VECTORIZE_FMA + #ifdef __AVX512DQ__ + #define EIGEN_VECTORIZE_AVX512DQ + #endif + #ifdef __AVX512ER__ + #define EIGEN_VECTORIZE_AVX512ER + #endif + #endif + + // include files + + // This extern "C" works around a MINGW-w64 compilation issue + // https://sourceforge.net/tracker/index.php?func=detail&aid=3018394&group_id=202880&atid=983354 + // In essence, intrin.h is included by windows.h and also declares intrinsics (just as emmintrin.h etc. below do). + // However, intrin.h uses an extern "C" declaration, and g++ thus complains of duplicate declarations + // with conflicting linkage. The linkage for intrinsics doesn't matter, but at that stage the compiler doesn't know; + // so, to avoid compile errors when windows.h is included after Eigen/Core, ensure intrinsics are extern "C" here too. + // notice that since these are C headers, the extern "C" is theoretically needed anyways. + extern "C" { + // In theory we should only include immintrin.h and not the other *mmintrin.h header files directly. + // Doing so triggers some issues with ICC. However old gcc versions seems to not have this file, thus: + #if EIGEN_COMP_ICC >= 1110 + #include + #else + #include + #include + #include + #ifdef EIGEN_VECTORIZE_SSE3 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSSE3 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSE4_1 + #include + #endif + #ifdef EIGEN_VECTORIZE_SSE4_2 + #include + #endif + #if defined(EIGEN_VECTORIZE_AVX) || defined(EIGEN_VECTORIZE_AVX512) + #include + #endif + #endif + } // end extern "C" + #elif defined __VSX__ + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_VSX + #include + // We need to #undef all these ugly tokens defined in + // => use __vector instead of vector + #undef bool + #undef vector + #undef pixel + #elif defined __ALTIVEC__ + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_ALTIVEC + #include + // We need to #undef all these ugly tokens defined in + // => use __vector instead of vector + #undef bool + #undef vector + #undef pixel + #elif (defined __ARM_NEON) || (defined __ARM_NEON__) + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_NEON + #include + #elif (defined __s390x__ && defined __VEC__) + #define EIGEN_VECTORIZE + #define EIGEN_VECTORIZE_ZVECTOR + #include + #endif +#endif + +#if defined(__F16C__) && !defined(EIGEN_COMP_CLANG) + // We can use the optimized fp16 to float and float to fp16 conversion routines + #define EIGEN_HAS_FP16_C +#endif + +#if defined __CUDACC__ + #define EIGEN_VECTORIZE_CUDA + #include + #if EIGEN_CUDACC_VER >= 70500 + #define EIGEN_HAS_CUDA_FP16 + #endif +#endif + +#if defined EIGEN_HAS_CUDA_FP16 + #include + #include +#endif + +#if (defined _OPENMP) && (!defined EIGEN_DONT_PARALLELIZE) + #define EIGEN_HAS_OPENMP +#endif + +#ifdef EIGEN_HAS_OPENMP +#include +#endif + +// MSVC for windows mobile does not have the errno.h file +#if !(EIGEN_COMP_MSVC && EIGEN_OS_WINCE) && !EIGEN_COMP_ARM +#define EIGEN_HAS_ERRNO +#endif + +#ifdef EIGEN_HAS_ERRNO +#include +#endif +#include +#include +#include +#include +#include +#include +#ifndef EIGEN_NO_IO + #include +#endif +#include +#include +#include +#include // for CHAR_BIT +// for min/max: +#include + +// for std::is_nothrow_move_assignable +#ifdef EIGEN_INCLUDE_TYPE_TRAITS +#include +#endif + +// for outputting debug info +#ifdef EIGEN_DEBUG_ASSIGN +#include +#endif + +// required for __cpuid, needs to be included after cmath +#if EIGEN_COMP_MSVC && EIGEN_ARCH_i386_OR_x86_64 && !EIGEN_OS_WINCE + #include +#endif + +/** \brief Namespace containing all symbols from the %Eigen library. */ +namespace Eigen { + +inline static const char *SimdInstructionSetsInUse(void) { +#if defined(EIGEN_VECTORIZE_AVX512) + return "AVX512, FMA, AVX2, AVX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_AVX) + return "AVX SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_SSE4_2) + return "SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2"; +#elif defined(EIGEN_VECTORIZE_SSE4_1) + return "SSE, SSE2, SSE3, SSSE3, SSE4.1"; +#elif defined(EIGEN_VECTORIZE_SSSE3) + return "SSE, SSE2, SSE3, SSSE3"; +#elif defined(EIGEN_VECTORIZE_SSE3) + return "SSE, SSE2, SSE3"; +#elif defined(EIGEN_VECTORIZE_SSE2) + return "SSE, SSE2"; +#elif defined(EIGEN_VECTORIZE_ALTIVEC) + return "AltiVec"; +#elif defined(EIGEN_VECTORIZE_VSX) + return "VSX"; +#elif defined(EIGEN_VECTORIZE_NEON) + return "ARM NEON"; +#elif defined(EIGEN_VECTORIZE_ZVECTOR) + return "S390X ZVECTOR"; +#else + return "None"; +#endif +} + +} // end namespace Eigen + +#if defined EIGEN2_SUPPORT_STAGE40_FULL_EIGEN3_STRICTNESS || defined EIGEN2_SUPPORT_STAGE30_FULL_EIGEN3_API || defined EIGEN2_SUPPORT_STAGE20_RESOLVE_API_CONFLICTS || defined EIGEN2_SUPPORT_STAGE10_FULL_EIGEN2_API || defined EIGEN2_SUPPORT +// This will generate an error message: +#error Eigen2-support is only available up to version 3.2. Please go to "http://eigen.tuxfamily.org/index.php?title=Eigen2" for further information +#endif + +namespace Eigen { + +// we use size_t frequently and we'll never remember to prepend it with std:: everytime just to +// ensure QNX/QCC support +using std::size_t; +// gcc 4.6.0 wants std:: for ptrdiff_t +using std::ptrdiff_t; + +} + +/** \defgroup Core_Module Core module + * This is the main module of Eigen providing dense matrix and vector support + * (both fixed and dynamic size) with all the features corresponding to a BLAS library + * and much more... + * + * \code + * #include + * \endcode + */ + +#include "src/Core/util/Constants.h" +#include "src/Core/util/Meta.h" +#include "src/Core/util/ForwardDeclarations.h" +#include "src/Core/util/StaticAssert.h" +#include "src/Core/util/XprHelper.h" +#include "src/Core/util/Memory.h" + +#include "src/Core/NumTraits.h" +#include "src/Core/MathFunctions.h" +#include "src/Core/GenericPacketMath.h" +#include "src/Core/MathFunctionsImpl.h" +#include "src/Core/arch/Default/ConjHelper.h" + +#if defined EIGEN_VECTORIZE_AVX512 + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/AVX/PacketMath.h" + #include "src/Core/arch/AVX/MathFunctions.h" + #include "src/Core/arch/AVX512/PacketMath.h" + #include "src/Core/arch/AVX512/MathFunctions.h" +#elif defined EIGEN_VECTORIZE_AVX + // Use AVX for floats and doubles, SSE for integers + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/Complex.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/AVX/PacketMath.h" + #include "src/Core/arch/AVX/MathFunctions.h" + #include "src/Core/arch/AVX/Complex.h" + #include "src/Core/arch/AVX/TypeCasting.h" + #include "src/Core/arch/SSE/TypeCasting.h" +#elif defined EIGEN_VECTORIZE_SSE + #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/MathFunctions.h" + #include "src/Core/arch/SSE/Complex.h" + #include "src/Core/arch/SSE/TypeCasting.h" +#elif defined(EIGEN_VECTORIZE_ALTIVEC) || defined(EIGEN_VECTORIZE_VSX) + #include "src/Core/arch/AltiVec/PacketMath.h" + #include "src/Core/arch/AltiVec/MathFunctions.h" + #include "src/Core/arch/AltiVec/Complex.h" +#elif defined EIGEN_VECTORIZE_NEON + #include "src/Core/arch/NEON/PacketMath.h" + #include "src/Core/arch/NEON/MathFunctions.h" + #include "src/Core/arch/NEON/Complex.h" +#elif defined EIGEN_VECTORIZE_ZVECTOR + #include "src/Core/arch/ZVector/PacketMath.h" + #include "src/Core/arch/ZVector/MathFunctions.h" + #include "src/Core/arch/ZVector/Complex.h" +#endif + +// Half float support +#include "src/Core/arch/CUDA/Half.h" +#include "src/Core/arch/CUDA/PacketMathHalf.h" +#include "src/Core/arch/CUDA/TypeCasting.h" + +#if defined EIGEN_VECTORIZE_CUDA + #include "src/Core/arch/CUDA/PacketMath.h" + #include "src/Core/arch/CUDA/MathFunctions.h" +#endif + +#include "src/Core/arch/Default/Settings.h" + +#include "src/Core/functors/TernaryFunctors.h" +#include "src/Core/functors/BinaryFunctors.h" +#include "src/Core/functors/UnaryFunctors.h" +#include "src/Core/functors/NullaryFunctors.h" +#include "src/Core/functors/StlFunctors.h" +#include "src/Core/functors/AssignmentFunctors.h" + +// Specialized functors to enable the processing of complex numbers +// on CUDA devices +#include "src/Core/arch/CUDA/Complex.h" + +#include "src/Core/IO.h" +#include "src/Core/DenseCoeffsBase.h" +#include "src/Core/DenseBase.h" +#include "src/Core/MatrixBase.h" +#include "src/Core/EigenBase.h" + +#include "src/Core/Product.h" +#include "src/Core/CoreEvaluators.h" +#include "src/Core/AssignEvaluator.h" + +#ifndef EIGEN_PARSED_BY_DOXYGEN // work around Doxygen bug triggered by Assign.h r814874 + // at least confirmed with Doxygen 1.5.5 and 1.5.6 + #include "src/Core/Assign.h" +#endif + +#include "src/Core/ArrayBase.h" +#include "src/Core/util/BlasUtil.h" +#include "src/Core/DenseStorage.h" +#include "src/Core/NestByValue.h" + +// #include "src/Core/ForceAlignedAccess.h" + +#include "src/Core/ReturnByValue.h" +#include "src/Core/NoAlias.h" +#include "src/Core/PlainObjectBase.h" +#include "src/Core/Matrix.h" +#include "src/Core/Array.h" +#include "src/Core/CwiseTernaryOp.h" +#include "src/Core/CwiseBinaryOp.h" +#include "src/Core/CwiseUnaryOp.h" +#include "src/Core/CwiseNullaryOp.h" +#include "src/Core/CwiseUnaryView.h" +#include "src/Core/SelfCwiseBinaryOp.h" +#include "src/Core/Dot.h" +#include "src/Core/StableNorm.h" +#include "src/Core/Stride.h" +#include "src/Core/MapBase.h" +#include "src/Core/Map.h" +#include "src/Core/Ref.h" +#include "src/Core/Block.h" +#include "src/Core/VectorBlock.h" +#include "src/Core/Transpose.h" +#include "src/Core/DiagonalMatrix.h" +#include "src/Core/Diagonal.h" +#include "src/Core/DiagonalProduct.h" +#include "src/Core/Redux.h" +#include "src/Core/Visitor.h" +#include "src/Core/Fuzzy.h" +#include "src/Core/Swap.h" +#include "src/Core/CommaInitializer.h" +#include "src/Core/GeneralProduct.h" +#include "src/Core/Solve.h" +#include "src/Core/Inverse.h" +#include "src/Core/SolverBase.h" +#include "src/Core/PermutationMatrix.h" +#include "src/Core/Transpositions.h" +#include "src/Core/TriangularMatrix.h" +#include "src/Core/SelfAdjointView.h" +#include "src/Core/products/GeneralBlockPanelKernel.h" +#include "src/Core/products/Parallelizer.h" +#include "src/Core/ProductEvaluators.h" +#include "src/Core/products/GeneralMatrixVector.h" +#include "src/Core/products/GeneralMatrixMatrix.h" +#include "src/Core/SolveTriangular.h" +#include "src/Core/products/GeneralMatrixMatrixTriangular.h" +#include "src/Core/products/SelfadjointMatrixVector.h" +#include "src/Core/products/SelfadjointMatrixMatrix.h" +#include "src/Core/products/SelfadjointProduct.h" +#include "src/Core/products/SelfadjointRank2Update.h" +#include "src/Core/products/TriangularMatrixVector.h" +#include "src/Core/products/TriangularMatrixMatrix.h" +#include "src/Core/products/TriangularSolverMatrix.h" +#include "src/Core/products/TriangularSolverVector.h" +#include "src/Core/BandMatrix.h" +#include "src/Core/CoreIterators.h" +#include "src/Core/ConditionEstimator.h" + +#include "src/Core/BooleanRedux.h" +#include "src/Core/Select.h" +#include "src/Core/VectorwiseOp.h" +#include "src/Core/Random.h" +#include "src/Core/Replicate.h" +#include "src/Core/Reverse.h" +#include "src/Core/ArrayWrapper.h" + +#ifdef EIGEN_USE_BLAS +#include "src/Core/products/GeneralMatrixMatrix_BLAS.h" +#include "src/Core/products/GeneralMatrixVector_BLAS.h" +#include "src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h" +#include "src/Core/products/SelfadjointMatrixMatrix_BLAS.h" +#include "src/Core/products/SelfadjointMatrixVector_BLAS.h" +#include "src/Core/products/TriangularMatrixMatrix_BLAS.h" +#include "src/Core/products/TriangularMatrixVector_BLAS.h" +#include "src/Core/products/TriangularSolverMatrix_BLAS.h" +#endif // EIGEN_USE_BLAS + +#ifdef EIGEN_USE_MKL_VML +#include "src/Core/Assign_MKL.h" +#endif + +#include "src/Core/GlobalFunctions.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_CORE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Dense b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Dense new file mode 100644 index 00000000000..f98b4b94951 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Dense @@ -0,0 +1,3 @@ +#include "Core" +#include "LU" + diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Householder b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Householder new file mode 100644 index 00000000000..89cd81b1afb --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Householder @@ -0,0 +1,30 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_HOUSEHOLDER_MODULE_H +#define EIGEN_HOUSEHOLDER_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Householder_Module Householder module + * This module provides Householder transformations. + * + * \code + * #include + * \endcode + */ + +#include "src/Householder/Householder.h" +#include "src/Householder/HouseholderSequence.h" +#include "src/Householder/BlockHouseholder.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_HOUSEHOLDER_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Jacobi b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Jacobi new file mode 100644 index 00000000000..17c1d785a16 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/Jacobi @@ -0,0 +1,33 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_JACOBI_MODULE_H +#define EIGEN_JACOBI_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup Jacobi_Module Jacobi module + * This module provides Jacobi and Givens rotations. + * + * \code + * #include + * \endcode + * + * In addition to listed classes, it defines the two following MatrixBase methods to apply a Jacobi or Givens rotation: + * - MatrixBase::applyOnTheLeft() + * - MatrixBase::applyOnTheRight(). + */ + +#include "src/Jacobi/Jacobi.h" + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_JACOBI_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ + diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/LU b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/LU new file mode 100644 index 00000000000..6418a86e192 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/LU @@ -0,0 +1,50 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_LU_MODULE_H +#define EIGEN_LU_MODULE_H + +#include "Core" + +#include "src/Core/util/DisableStupidWarnings.h" + +/** \defgroup LU_Module LU module + * This module includes %LU decomposition and related notions such as matrix inversion and determinant. + * This module defines the following MatrixBase methods: + * - MatrixBase::inverse() + * - MatrixBase::determinant() + * + * \code + * #include + * \endcode + */ + +#include "src/misc/Kernel.h" +#include "src/misc/Image.h" +#include "src/LU/FullPivLU.h" +#include "src/LU/PartialPivLU.h" +#ifdef EIGEN_USE_LAPACKE +#ifdef EIGEN_USE_MKL +#include "mkl_lapacke.h" +#else +#include "src/misc/lapacke.h" +#endif +#include "src/LU/PartialPivLU_LAPACKE.h" +#endif +#include "src/LU/Determinant.h" +#include "src/LU/InverseImpl.h" + +// Use the SSE optimized version whenever possible. At the moment the +// SSE version doesn't compile when AVX is enabled +#if defined EIGEN_VECTORIZE_SSE && !defined EIGEN_VECTORIZE_AVX + #include "src/LU/arch/Inverse_SSE.h" +#endif + +#include "src/Core/util/ReenableStupidWarnings.h" + +#endif // EIGEN_LU_MODULE_H +/* vim: set filetype=cpp et sw=2 ts=2 ai: */ diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Array.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Array.h new file mode 100644 index 00000000000..16770fc7b3a --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Array.h @@ -0,0 +1,329 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAY_H +#define EIGEN_ARRAY_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits > +{ + typedef ArrayXpr XprKind; + typedef ArrayBase > XprBase; +}; +} + +/** \class Array + * \ingroup Core_Module + * + * \brief General-purpose arrays with easy API for coefficient-wise operations + * + * The %Array class is very similar to the Matrix class. It provides + * general-purpose one- and two-dimensional arrays. The difference between the + * %Array and the %Matrix class is primarily in the API: the API for the + * %Array class provides easy access to coefficient-wise operations, while the + * API for the %Matrix class provides easy access to linear-algebra + * operations. + * + * See documentation of class Matrix for detailed information on the template parameters + * storage layout. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN. + * + * \sa \blank \ref TutorialArrayClass, \ref TopicClassHierarchy + */ +template +class Array + : public PlainObjectBase > +{ + public: + + typedef PlainObjectBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Array) + + enum { Options = _Options }; + typedef typename Base::PlainObject PlainObject; + + protected: + template + friend struct internal::conservative_resize_like_impl; + + using Base::m_storage; + + public: + + using Base::base; + using Base::coeff; + using Base::coeffRef; + + /** + * The usage of + * using Base::operator=; + * fails on MSVC. Since the code below is working with GCC and MSVC, we skipped + * the usage of 'using'. This should be done only for operator=. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const EigenBase &other) + { + return Base::operator=(other); + } + + /** Set all the entries to \a value. + * \sa DenseBase::setConstant(), DenseBase::fill() + */ + /* This overload is needed because the usage of + * using Base::operator=; + * fails on MSVC. Since the code below is working with GCC and MSVC, we skipped + * the usage of 'using'. This should be done only for operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const Scalar &value) + { + Base::setConstant(value); + return *this; + } + + /** Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const DenseBase& other) + { + return Base::_set(other); + } + + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array& operator=(const Array& other) + { + return Base::_set(other); + } + + /** Default constructor. + * + * For fixed-size matrices, does nothing. + * + * For dynamic-size matrices, creates an empty matrix of size 0. Does not allocate any array. Such a matrix + * is called a null matrix. This constructor is the unique way to create null matrices: resizing + * a matrix to 0 is not supported. + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array() : Base() + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + // FIXME is it still needed ?? + /** \internal */ + EIGEN_DEVICE_FUNC + Array(internal::constructor_without_unaligned_array_assert) + : Base(internal::constructor_without_unaligned_array_assert()) + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } +#endif + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + Array(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible::value) + : Base(std::move(other)) + { + Base::_check_template_params(); + } + EIGEN_DEVICE_FUNC + Array& operator=(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable::value) + { + other.swap(*this); + return *this; + } +#endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Array(const T& x) + { + Base::_check_template_params(); + Base::template _init1(x); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const T0& val0, const T1& val1) + { + Base::_check_template_params(); + this->template _init2(val0, val1); + } + #else + /** \brief Constructs a fixed-sized array initialized with coefficients starting at \a data */ + EIGEN_DEVICE_FUNC explicit Array(const Scalar *data); + /** Constructs a vector or row-vector with given dimension. \only_for_vectors + * + * Note that this is only useful for dynamic-size vectors. For fixed-size vectors, + * it is redundant to pass the dimension here, so it makes more sense to use the default + * constructor Array() instead. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Array(Index dim); + /** constructs an initialized 1x1 Array with the given coefficient */ + Array(const Scalar& value); + /** constructs an uninitialized array with \a rows rows and \a cols columns. + * + * This is useful for dynamic-size arrays. For fixed-size arrays, + * it is redundant to pass these parameters, so one should use the default constructor + * Array() instead. */ + Array(Index rows, Index cols); + /** constructs an initialized 2D vector with given coefficients */ + Array(const Scalar& val0, const Scalar& val1); + #endif + + /** constructs an initialized 3D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 3) + m_storage.data()[0] = val0; + m_storage.data()[1] = val1; + m_storage.data()[2] = val2; + } + /** constructs an initialized 4D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2, const Scalar& val3) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 4) + m_storage.data()[0] = val0; + m_storage.data()[1] = val1; + m_storage.data()[2] = val2; + m_storage.data()[3] = val3; + } + + /** Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const Array& other) + : Base(other) + { } + + private: + struct PrivateType {}; + public: + + /** \sa MatrixBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Array(const EigenBase &other, + typename internal::enable_if::value, + PrivateType>::type = PrivateType()) + : Base(other.derived()) + { } + + EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); } + + #ifdef EIGEN_ARRAY_PLUGIN + #include EIGEN_ARRAY_PLUGIN + #endif + + private: + + template + friend struct internal::matrix_swap_impl; +}; + +/** \defgroup arraytypedefs Global array typedefs + * \ingroup Core_Module + * + * Eigen defines several typedef shortcuts for most common 1D and 2D array types. + * + * The general patterns are the following: + * + * \c ArrayRowsColsType where \c Rows and \c Cols can be \c 2,\c 3,\c 4 for fixed size square matrices or \c X for dynamic size, + * and where \c Type can be \c i for integer, \c f for float, \c d for double, \c cf for complex float, \c cd + * for complex double. + * + * For example, \c Array33d is a fixed-size 3x3 array type of doubles, and \c ArrayXXf is a dynamic-size matrix of floats. + * + * There are also \c ArraySizeType which are self-explanatory. For example, \c Array4cf is + * a fixed-size 1D array of 4 complex floats. + * + * \sa class Array + */ + +#define EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, Size, SizeSuffix) \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##SizeSuffix##SizeSuffix##TypeSuffix; \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##SizeSuffix##TypeSuffix; + +#define EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, Size) \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##Size##X##TypeSuffix; \ +/** \ingroup arraytypedefs */ \ +typedef Array Array##X##Size##TypeSuffix; + +#define EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(Type, TypeSuffix) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 2, 2) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 3, 3) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, 4, 4) \ +EIGEN_MAKE_ARRAY_TYPEDEFS(Type, TypeSuffix, Dynamic, X) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 2) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 3) \ +EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS(Type, TypeSuffix, 4) + +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(int, i) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(float, f) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(double, d) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(std::complex, cf) +EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES(std::complex, cd) + +#undef EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES +#undef EIGEN_MAKE_ARRAY_TYPEDEFS + +#undef EIGEN_MAKE_ARRAY_TYPEDEFS_LARGE + +#define EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, SizeSuffix) \ +using Eigen::Matrix##SizeSuffix##TypeSuffix; \ +using Eigen::Vector##SizeSuffix##TypeSuffix; \ +using Eigen::RowVector##SizeSuffix##TypeSuffix; + +#define EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(TypeSuffix) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 2) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 3) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, 4) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE_AND_SIZE(TypeSuffix, X) \ + +#define EIGEN_USING_ARRAY_TYPEDEFS \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(i) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(f) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(d) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(cf) \ +EIGEN_USING_ARRAY_TYPEDEFS_FOR_TYPE(cd) + +} // end namespace Eigen + +#endif // EIGEN_ARRAY_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h new file mode 100644 index 00000000000..33f644e2132 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayBase.h @@ -0,0 +1,226 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAYBASE_H +#define EIGEN_ARRAYBASE_H + +namespace Eigen { + +template class MatrixWrapper; + +/** \class ArrayBase + * \ingroup Core_Module + * + * \brief Base class for all 1D and 2D array, and related expressions + * + * An array is similar to a dense vector or matrix. While matrices are mathematical + * objects with well defined linear algebra operators, an array is just a collection + * of scalar values arranged in a one or two dimensionnal fashion. As the main consequence, + * all operations applied to an array are performed coefficient wise. Furthermore, + * arrays support scalar math functions of the c++ standard library (e.g., std::sin(x)), and convenient + * constructors allowing to easily write generic code working for both scalar values + * and arrays. + * + * This class is the base that is inherited by all array expression types. + * + * \tparam Derived is the derived type, e.g., an array or an expression type. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN. + * + * \sa class MatrixBase, \ref TopicClassHierarchy + */ +template class ArrayBase + : public DenseBase +{ + public: +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** The base class for a given storage type. */ + typedef ArrayBase StorageBaseType; + + typedef ArrayBase Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + typedef DenseBase Base; + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::operator=; + using Base::operator+=; + using Base::operator-=; + using Base::operator*=; + using Base::operator/=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Base::PlainObject PlainObject; + + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::ArrayBase +#define EIGEN_DOC_UNARY_ADDONS(X,Y) +# include "../plugins/CommonCwiseUnaryOps.h" +# include "../plugins/MatrixCwiseUnaryOps.h" +# include "../plugins/ArrayCwiseUnaryOps.h" +# include "../plugins/CommonCwiseBinaryOps.h" +# include "../plugins/MatrixCwiseBinaryOps.h" +# include "../plugins/ArrayCwiseBinaryOps.h" +# ifdef EIGEN_ARRAYBASE_PLUGIN +# include EIGEN_ARRAYBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_UNARY_ADDONS + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const ArrayBase& other) + { + internal::call_assignment(derived(), other.derived()); + return derived(); + } + + /** Set all the entries to \a value. + * \sa DenseBase::setConstant(), DenseBase::fill() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const Scalar &value) + { Base::setConstant(value); return derived(); } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const Scalar& scalar); + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const Scalar& scalar); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const ArrayBase& other); + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const ArrayBase& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator*=(const ArrayBase& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator/=(const ArrayBase& other); + + public: + EIGEN_DEVICE_FUNC + ArrayBase& array() { return *this; } + EIGEN_DEVICE_FUNC + const ArrayBase& array() const { return *this; } + + /** \returns an \link Eigen::MatrixBase Matrix \endlink expression of this array + * \sa MatrixBase::array() */ + EIGEN_DEVICE_FUNC + MatrixWrapper matrix() { return MatrixWrapper(derived()); } + EIGEN_DEVICE_FUNC + const MatrixWrapper matrix() const { return MatrixWrapper(derived()); } + +// template +// inline void evalTo(Dest& dst) const { dst = matrix(); } + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(ArrayBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(ArrayBase) + + private: + explicit ArrayBase(Index); + ArrayBase(Index,Index); + template explicit ArrayBase(const ArrayBase&); + protected: + // mixing arrays and matrices is not legal + template Derived& operator+=(const MatrixBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} + // mixing arrays and matrices is not legal + template Derived& operator-=(const MatrixBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} +}; + +/** replaces \c *this by \c *this - \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator-=(const ArrayBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this + \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator+=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this * \a other coefficient wise. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator*=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::mul_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this / \a other coefficient wise. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived & +ArrayBase::operator/=(const ArrayBase& other) +{ + call_assignment(derived(), other.derived(), internal::div_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_ARRAYBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h new file mode 100644 index 00000000000..688aadd6260 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ArrayWrapper.h @@ -0,0 +1,209 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ARRAYWRAPPER_H +#define EIGEN_ARRAYWRAPPER_H + +namespace Eigen { + +/** \class ArrayWrapper + * \ingroup Core_Module + * + * \brief Expression of a mathematical vector or matrix as an array object + * + * This class is the return type of MatrixBase::array(), and most of the time + * this is the only way it is use. + * + * \sa MatrixBase::array(), class MatrixWrapper + */ + +namespace internal { +template +struct traits > + : public traits::type > +{ + typedef ArrayXpr XprKind; + // Let's remove NestByRefBit + enum { + Flags0 = traits::type >::Flags, + LvalueBitFlag = is_lvalue::value ? LvalueBit : 0, + Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag + }; +}; +} + +template +class ArrayWrapper : public ArrayBase > +{ + public: + typedef ArrayBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(ArrayWrapper) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(ArrayWrapper) + typedef typename internal::remove_all::type NestedExpression; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + typedef typename internal::ref_selector::non_const_type NestedExpressionType; + + using Base::coeffRef; + + EIGEN_DEVICE_FUNC + explicit EIGEN_STRONG_INLINE ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC + inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC + inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return m_expression.data(); } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_expression.coeffRef(rowId, colId); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_expression.coeffRef(index); + } + + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const { dst = m_expression; } + + const typename internal::remove_all::type& + EIGEN_DEVICE_FUNC + nestedExpression() const + { + return m_expression; + } + + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index) */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) { m_expression.resize(newSize); } + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index,Index)*/ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) { m_expression.resize(rows,cols); } + + protected: + NestedExpressionType m_expression; +}; + +/** \class MatrixWrapper + * \ingroup Core_Module + * + * \brief Expression of an array as a mathematical vector or matrix + * + * This class is the return type of ArrayBase::matrix(), and most of the time + * this is the only way it is use. + * + * \sa MatrixBase::matrix(), class ArrayWrapper + */ + +namespace internal { +template +struct traits > + : public traits::type > +{ + typedef MatrixXpr XprKind; + // Let's remove NestByRefBit + enum { + Flags0 = traits::type >::Flags, + LvalueBitFlag = is_lvalue::value ? LvalueBit : 0, + Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag + }; +}; +} + +template +class MatrixWrapper : public MatrixBase > +{ + public: + typedef MatrixBase > Base; + EIGEN_DENSE_PUBLIC_INTERFACE(MatrixWrapper) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(MatrixWrapper) + typedef typename internal::remove_all::type NestedExpression; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + typedef typename internal::ref_selector::non_const_type NestedExpressionType; + + using Base::coeffRef; + + EIGEN_DEVICE_FUNC + explicit inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC + inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC + inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return m_expression.data(); } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_expression.derived().coeffRef(rowId, colId); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_expression.coeffRef(index); + } + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& + nestedExpression() const + { + return m_expression; + } + + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index) */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) { m_expression.resize(newSize); } + /** Forwards the resizing request to the nested expression + * \sa DenseBase::resize(Index,Index)*/ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) { m_expression.resize(rows,cols); } + + protected: + NestedExpressionType m_expression; +}; + +} // end namespace Eigen + +#endif // EIGEN_ARRAYWRAPPER_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h new file mode 100644 index 00000000000..53806ba33c4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Assign.h @@ -0,0 +1,90 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007 Michael Olbrich +// Copyright (C) 2006-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ASSIGN_H +#define EIGEN_ASSIGN_H + +namespace Eigen { + +template +template +EIGEN_STRONG_INLINE Derived& DenseBase + ::lazyAssign(const DenseBase& other) +{ + enum{ + SameType = internal::is_same::value + }; + + EIGEN_STATIC_ASSERT_LVALUE(Derived) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Derived,OtherDerived) + EIGEN_STATIC_ASSERT(SameType,YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) + + eigen_assert(rows() == other.rows() && cols() == other.cols()); + internal::call_assignment_no_alias(derived(),other.derived()); + + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& DenseBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& DenseBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const MatrixBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const DenseBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const EigenBase& other) +{ + internal::call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE Derived& MatrixBase::operator=(const ReturnByValue& other) +{ + other.derived().evalTo(derived()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_ASSIGN_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h new file mode 100644 index 00000000000..dbe435d86b3 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/AssignEvaluator.h @@ -0,0 +1,935 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2011 Benoit Jacob +// Copyright (C) 2011-2014 Gael Guennebaud +// Copyright (C) 2011-2012 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ASSIGN_EVALUATOR_H +#define EIGEN_ASSIGN_EVALUATOR_H + +namespace Eigen { + +// This implementation is based on Assign.h + +namespace internal { + +/*************************************************************************** +* Part 1 : the logic deciding a strategy for traversal and unrolling * +***************************************************************************/ + +// copy_using_evaluator_traits is based on assign_traits + +template +struct copy_using_evaluator_traits +{ + typedef typename DstEvaluator::XprType Dst; + typedef typename Dst::Scalar DstScalar; + + enum { + DstFlags = DstEvaluator::Flags, + SrcFlags = SrcEvaluator::Flags + }; + +public: + enum { + DstAlignment = DstEvaluator::Alignment, + SrcAlignment = SrcEvaluator::Alignment, + DstHasDirectAccess = (DstFlags & DirectAccessBit) == DirectAccessBit, + JointAlignment = EIGEN_PLAIN_ENUM_MIN(DstAlignment,SrcAlignment) + }; + +private: + enum { + InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime) + : int(DstFlags)&RowMajorBit ? int(Dst::ColsAtCompileTime) + : int(Dst::RowsAtCompileTime), + InnerMaxSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::MaxSizeAtCompileTime) + : int(DstFlags)&RowMajorBit ? int(Dst::MaxColsAtCompileTime) + : int(Dst::MaxRowsAtCompileTime), + OuterStride = int(outer_stride_at_compile_time::ret), + MaxSizeAtCompileTime = Dst::SizeAtCompileTime + }; + + // TODO distinguish between linear traversal and inner-traversals + typedef typename find_best_packet::type LinearPacketType; + typedef typename find_best_packet::type InnerPacketType; + + enum { + LinearPacketSize = unpacket_traits::size, + InnerPacketSize = unpacket_traits::size + }; + +public: + enum { + LinearRequiredAlignment = unpacket_traits::alignment, + InnerRequiredAlignment = unpacket_traits::alignment + }; + +private: + enum { + DstIsRowMajor = DstFlags&RowMajorBit, + SrcIsRowMajor = SrcFlags&RowMajorBit, + StorageOrdersAgree = (int(DstIsRowMajor) == int(SrcIsRowMajor)), + MightVectorize = bool(StorageOrdersAgree) + && (int(DstFlags) & int(SrcFlags) & ActualPacketAccessBit) + && bool(functor_traits::PacketAccess), + MayInnerVectorize = MightVectorize + && int(InnerSize)!=Dynamic && int(InnerSize)%int(InnerPacketSize)==0 + && int(OuterStride)!=Dynamic && int(OuterStride)%int(InnerPacketSize)==0 + && (EIGEN_UNALIGNED_VECTORIZE || int(JointAlignment)>=int(InnerRequiredAlignment)), + MayLinearize = bool(StorageOrdersAgree) && (int(DstFlags) & int(SrcFlags) & LinearAccessBit), + MayLinearVectorize = bool(MightVectorize) && bool(MayLinearize) && bool(DstHasDirectAccess) + && (EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment)) || MaxSizeAtCompileTime == Dynamic), + /* If the destination isn't aligned, we have to do runtime checks and we don't unroll, + so it's only good for large enough sizes. */ + MaySliceVectorize = bool(MightVectorize) && bool(DstHasDirectAccess) + && (int(InnerMaxSize)==Dynamic || int(InnerMaxSize)>=(EIGEN_UNALIGNED_VECTORIZE?InnerPacketSize:(3*InnerPacketSize))) + /* slice vectorization can be slow, so we only want it if the slices are big, which is + indicated by InnerMaxSize rather than InnerSize, think of the case of a dynamic block + in a fixed-size matrix + However, with EIGEN_UNALIGNED_VECTORIZE and unrolling, slice vectorization is still worth it */ + }; + +public: + enum { + Traversal = int(MayLinearVectorize) && (LinearPacketSize>InnerPacketSize) ? int(LinearVectorizedTraversal) + : int(MayInnerVectorize) ? int(InnerVectorizedTraversal) + : int(MayLinearVectorize) ? int(LinearVectorizedTraversal) + : int(MaySliceVectorize) ? int(SliceVectorizedTraversal) + : int(MayLinearize) ? int(LinearTraversal) + : int(DefaultTraversal), + Vectorized = int(Traversal) == InnerVectorizedTraversal + || int(Traversal) == LinearVectorizedTraversal + || int(Traversal) == SliceVectorizedTraversal + }; + + typedef typename conditional::type PacketType; + +private: + enum { + ActualPacketSize = int(Traversal)==LinearVectorizedTraversal ? LinearPacketSize + : Vectorized ? InnerPacketSize + : 1, + UnrollingLimit = EIGEN_UNROLLING_LIMIT * ActualPacketSize, + MayUnrollCompletely = int(Dst::SizeAtCompileTime) != Dynamic + && int(Dst::SizeAtCompileTime) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit), + MayUnrollInner = int(InnerSize) != Dynamic + && int(InnerSize) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit) + }; + +public: + enum { + Unrolling = (int(Traversal) == int(InnerVectorizedTraversal) || int(Traversal) == int(DefaultTraversal)) + ? ( + int(MayUnrollCompletely) ? int(CompleteUnrolling) + : int(MayUnrollInner) ? int(InnerUnrolling) + : int(NoUnrolling) + ) + : int(Traversal) == int(LinearVectorizedTraversal) + ? ( bool(MayUnrollCompletely) && ( EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment))) + ? int(CompleteUnrolling) + : int(NoUnrolling) ) + : int(Traversal) == int(LinearTraversal) + ? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling) + : int(NoUnrolling) ) +#if EIGEN_UNALIGNED_VECTORIZE + : int(Traversal) == int(SliceVectorizedTraversal) + ? ( bool(MayUnrollInner) ? int(InnerUnrolling) + : int(NoUnrolling) ) +#endif + : int(NoUnrolling) + }; + +#ifdef EIGEN_DEBUG_ASSIGN + static void debug() + { + std::cerr << "DstXpr: " << typeid(typename DstEvaluator::XprType).name() << std::endl; + std::cerr << "SrcXpr: " << typeid(typename SrcEvaluator::XprType).name() << std::endl; + std::cerr.setf(std::ios::hex, std::ios::basefield); + std::cerr << "DstFlags" << " = " << DstFlags << " (" << demangle_flags(DstFlags) << " )" << std::endl; + std::cerr << "SrcFlags" << " = " << SrcFlags << " (" << demangle_flags(SrcFlags) << " )" << std::endl; + std::cerr.unsetf(std::ios::hex); + EIGEN_DEBUG_VAR(DstAlignment) + EIGEN_DEBUG_VAR(SrcAlignment) + EIGEN_DEBUG_VAR(LinearRequiredAlignment) + EIGEN_DEBUG_VAR(InnerRequiredAlignment) + EIGEN_DEBUG_VAR(JointAlignment) + EIGEN_DEBUG_VAR(InnerSize) + EIGEN_DEBUG_VAR(InnerMaxSize) + EIGEN_DEBUG_VAR(LinearPacketSize) + EIGEN_DEBUG_VAR(InnerPacketSize) + EIGEN_DEBUG_VAR(ActualPacketSize) + EIGEN_DEBUG_VAR(StorageOrdersAgree) + EIGEN_DEBUG_VAR(MightVectorize) + EIGEN_DEBUG_VAR(MayLinearize) + EIGEN_DEBUG_VAR(MayInnerVectorize) + EIGEN_DEBUG_VAR(MayLinearVectorize) + EIGEN_DEBUG_VAR(MaySliceVectorize) + std::cerr << "Traversal" << " = " << Traversal << " (" << demangle_traversal(Traversal) << ")" << std::endl; + EIGEN_DEBUG_VAR(SrcEvaluator::CoeffReadCost) + EIGEN_DEBUG_VAR(UnrollingLimit) + EIGEN_DEBUG_VAR(MayUnrollCompletely) + EIGEN_DEBUG_VAR(MayUnrollInner) + std::cerr << "Unrolling" << " = " << Unrolling << " (" << demangle_unrolling(Unrolling) << ")" << std::endl; + std::cerr << std::endl; + } +#endif +}; + +/*************************************************************************** +* Part 2 : meta-unrollers +***************************************************************************/ + +/************************ +*** Default traversal *** +************************/ + +template +struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling +{ + // FIXME: this is not very clean, perhaps this information should be provided by the kernel? + typedef typename Kernel::DstEvaluatorType DstEvaluatorType; + typedef typename DstEvaluatorType::XprType DstXprType; + + enum { + outer = Index / DstXprType::InnerSizeAtCompileTime, + inner = Index % DstXprType::InnerSizeAtCompileTime + }; + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + kernel.assignCoeffByOuterInner(outer, inner); + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +template +struct copy_using_evaluator_DefaultTraversal_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer) + { + kernel.assignCoeffByOuterInner(outer, Index_); + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } +}; + +template +struct copy_using_evaluator_DefaultTraversal_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index) { } +}; + +/*********************** +*** Linear traversal *** +***********************/ + +template +struct copy_using_evaluator_LinearTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel& kernel) + { + kernel.assignCoeff(Index); + copy_using_evaluator_LinearTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_LinearTraversal_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +/************************** +*** Inner vectorization *** +**************************/ + +template +struct copy_using_evaluator_innervec_CompleteUnrolling +{ + // FIXME: this is not very clean, perhaps this information should be provided by the kernel? + typedef typename Kernel::DstEvaluatorType DstEvaluatorType; + typedef typename DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { + outer = Index / DstXprType::InnerSizeAtCompileTime, + inner = Index % DstXprType::InnerSizeAtCompileTime, + SrcAlignment = Kernel::AssignmentTraits::SrcAlignment, + DstAlignment = Kernel::AssignmentTraits::DstAlignment + }; + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + kernel.template assignPacketByOuterInner(outer, inner); + enum { NextIndex = Index + unpacket_traits::size }; + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + } +}; + +template +struct copy_using_evaluator_innervec_CompleteUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { } +}; + +template +struct copy_using_evaluator_innervec_InnerUnrolling +{ + typedef typename Kernel::PacketType PacketType; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer) + { + kernel.template assignPacketByOuterInner(outer, Index_); + enum { NextIndex = Index_ + unpacket_traits::size }; + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + } +}; + +template +struct copy_using_evaluator_innervec_InnerUnrolling +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &, Index) { } +}; + +/*************************************************************************** +* Part 3 : implementation of all cases +***************************************************************************/ + +// dense_assignment_loop is based on assign_impl + +template +struct dense_assignment_loop; + +/************************ +*** Default traversal *** +************************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static void EIGEN_STRONG_INLINE run(Kernel &kernel) + { + for(Index outer = 0; outer < kernel.outerSize(); ++outer) { + for(Index inner = 0; inner < kernel.innerSize(); ++inner) { + kernel.assignCoeffByOuterInner(outer, inner); + } + } + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + + const Index outerSize = kernel.outerSize(); + for(Index outer = 0; outer < outerSize; ++outer) + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } +}; + +/*************************** +*** Linear vectorization *** +***************************/ + + +// The goal of unaligned_dense_assignment_loop is simply to factorize the handling +// of the non vectorizable beginning and ending parts + +template +struct unaligned_dense_assignment_loop +{ + // if IsAligned = true, then do nothing + template + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index, Index) {} +}; + +template <> +struct unaligned_dense_assignment_loop +{ + // MSVC must not inline this functions. If it does, it fails to optimize the + // packet access path. + // FIXME check which version exhibits this issue +#if EIGEN_COMP_MSVC + template + static EIGEN_DONT_INLINE void run(Kernel &kernel, + Index start, + Index end) +#else + template + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, + Index start, + Index end) +#endif + { + for (Index index = start; index < end; ++index) + kernel.assignCoeff(index); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index size = kernel.size(); + typedef typename Kernel::Scalar Scalar; + typedef typename Kernel::PacketType PacketType; + enum { + requestedAlignment = Kernel::AssignmentTraits::LinearRequiredAlignment, + packetSize = unpacket_traits::size, + dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment), + dstAlignment = packet_traits::AlignedOnScalar ? int(requestedAlignment) + : int(Kernel::AssignmentTraits::DstAlignment), + srcAlignment = Kernel::AssignmentTraits::JointAlignment + }; + const Index alignedStart = dstIsAligned ? 0 : internal::first_aligned(kernel.dstDataPtr(), size); + const Index alignedEnd = alignedStart + ((size-alignedStart)/packetSize)*packetSize; + + unaligned_dense_assignment_loop::run(kernel, 0, alignedStart); + + for(Index index = alignedStart; index < alignedEnd; index += packetSize) + kernel.template assignPacket(index); + + unaligned_dense_assignment_loop<>::run(kernel, alignedEnd, size); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { size = DstXprType::SizeAtCompileTime, + packetSize =unpacket_traits::size, + alignedSize = (size/packetSize)*packetSize }; + + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + copy_using_evaluator_DefaultTraversal_CompleteUnrolling::run(kernel); + } +}; + +/************************** +*** Inner vectorization *** +**************************/ + +template +struct dense_assignment_loop +{ + typedef typename Kernel::PacketType PacketType; + enum { + SrcAlignment = Kernel::AssignmentTraits::SrcAlignment, + DstAlignment = Kernel::AssignmentTraits::DstAlignment + }; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index innerSize = kernel.innerSize(); + const Index outerSize = kernel.outerSize(); + const Index packetSize = unpacket_traits::size; + for(Index outer = 0; outer < outerSize; ++outer) + for(Index inner = 0; inner < innerSize; inner+=packetSize) + kernel.template assignPacketByOuterInner(outer, inner); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_innervec_CompleteUnrolling::run(kernel); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::AssignmentTraits Traits; + const Index outerSize = kernel.outerSize(); + for(Index outer = 0; outer < outerSize; ++outer) + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + } +}; + +/*********************** +*** Linear traversal *** +***********************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + const Index size = kernel.size(); + for(Index i = 0; i < size; ++i) + kernel.assignCoeff(i); + } +}; + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + copy_using_evaluator_LinearTraversal_CompleteUnrolling::run(kernel); + } +}; + +/************************** +*** Slice vectorization *** +***************************/ + +template +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::Scalar Scalar; + typedef typename Kernel::PacketType PacketType; + enum { + packetSize = unpacket_traits::size, + requestedAlignment = int(Kernel::AssignmentTraits::InnerRequiredAlignment), + alignable = packet_traits::AlignedOnScalar || int(Kernel::AssignmentTraits::DstAlignment)>=sizeof(Scalar), + dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment), + dstAlignment = alignable ? int(requestedAlignment) + : int(Kernel::AssignmentTraits::DstAlignment) + }; + const Scalar *dst_ptr = kernel.dstDataPtr(); + if((!bool(dstIsAligned)) && (UIntPtr(dst_ptr) % sizeof(Scalar))>0) + { + // the pointer is not aligend-on scalar, so alignment is not possible + return dense_assignment_loop::run(kernel); + } + const Index packetAlignedMask = packetSize - 1; + const Index innerSize = kernel.innerSize(); + const Index outerSize = kernel.outerSize(); + const Index alignedStep = alignable ? (packetSize - kernel.outerStride() % packetSize) & packetAlignedMask : 0; + Index alignedStart = ((!alignable) || bool(dstIsAligned)) ? 0 : internal::first_aligned(dst_ptr, innerSize); + + for(Index outer = 0; outer < outerSize; ++outer) + { + const Index alignedEnd = alignedStart + ((innerSize-alignedStart) & ~packetAlignedMask); + // do the non-vectorizable part of the assignment + for(Index inner = 0; inner(outer, inner); + + // do the non-vectorizable part of the assignment + for(Index inner = alignedEnd; inner +struct dense_assignment_loop +{ + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel) + { + typedef typename Kernel::DstEvaluatorType::XprType DstXprType; + typedef typename Kernel::PacketType PacketType; + + enum { size = DstXprType::InnerSizeAtCompileTime, + packetSize =unpacket_traits::size, + vectorizableSize = (size/packetSize)*packetSize }; + + for(Index outer = 0; outer < kernel.outerSize(); ++outer) + { + copy_using_evaluator_innervec_InnerUnrolling::run(kernel, outer); + copy_using_evaluator_DefaultTraversal_InnerUnrolling::run(kernel, outer); + } + } +}; +#endif + + +/*************************************************************************** +* Part 4 : Generic dense assignment kernel +***************************************************************************/ + +// This class generalize the assignment of a coefficient (or packet) from one dense evaluator +// to another dense writable evaluator. +// It is parametrized by the two evaluators, and the actual assignment functor. +// This abstraction level permits to keep the evaluation loops as simple and as generic as possible. +// One can customize the assignment using this generic dense_assignment_kernel with different +// functors, or by completely overloading it, by-passing a functor. +template +class generic_dense_assignment_kernel +{ +protected: + typedef typename DstEvaluatorTypeT::XprType DstXprType; + typedef typename SrcEvaluatorTypeT::XprType SrcXprType; +public: + + typedef DstEvaluatorTypeT DstEvaluatorType; + typedef SrcEvaluatorTypeT SrcEvaluatorType; + typedef typename DstEvaluatorType::Scalar Scalar; + typedef copy_using_evaluator_traits AssignmentTraits; + typedef typename AssignmentTraits::PacketType PacketType; + + + EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) + : m_dst(dst), m_src(src), m_functor(func), m_dstExpr(dstExpr) + { + #ifdef EIGEN_DEBUG_ASSIGN + AssignmentTraits::debug(); + #endif + } + + EIGEN_DEVICE_FUNC Index size() const { return m_dstExpr.size(); } + EIGEN_DEVICE_FUNC Index innerSize() const { return m_dstExpr.innerSize(); } + EIGEN_DEVICE_FUNC Index outerSize() const { return m_dstExpr.outerSize(); } + EIGEN_DEVICE_FUNC Index rows() const { return m_dstExpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_dstExpr.cols(); } + EIGEN_DEVICE_FUNC Index outerStride() const { return m_dstExpr.outerStride(); } + + EIGEN_DEVICE_FUNC DstEvaluatorType& dstEvaluator() { return m_dst; } + EIGEN_DEVICE_FUNC const SrcEvaluatorType& srcEvaluator() const { return m_src; } + + /// Assign src(row,col) to dst(row,col) through the assignment functor. + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index row, Index col) + { + m_functor.assignCoeff(m_dst.coeffRef(row,col), m_src.coeff(row,col)); + } + + /// \sa assignCoeff(Index,Index) + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index index) + { + m_functor.assignCoeff(m_dst.coeffRef(index), m_src.coeff(index)); + } + + /// \sa assignCoeff(Index,Index) + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeffByOuterInner(Index outer, Index inner) + { + Index row = rowIndexByOuterInner(outer, inner); + Index col = colIndexByOuterInner(outer, inner); + assignCoeff(row, col); + } + + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index row, Index col) + { + m_functor.template assignPacket(&m_dst.coeffRef(row,col), m_src.template packet(row,col)); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index index) + { + m_functor.template assignPacket(&m_dst.coeffRef(index), m_src.template packet(index)); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacketByOuterInner(Index outer, Index inner) + { + Index row = rowIndexByOuterInner(outer, inner); + Index col = colIndexByOuterInner(outer, inner); + assignPacket(row, col); + } + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner) + { + typedef typename DstEvaluatorType::ExpressionTraits Traits; + return int(Traits::RowsAtCompileTime) == 1 ? 0 + : int(Traits::ColsAtCompileTime) == 1 ? inner + : int(DstEvaluatorType::Flags)&RowMajorBit ? outer + : inner; + } + + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner) + { + typedef typename DstEvaluatorType::ExpressionTraits Traits; + return int(Traits::ColsAtCompileTime) == 1 ? 0 + : int(Traits::RowsAtCompileTime) == 1 ? inner + : int(DstEvaluatorType::Flags)&RowMajorBit ? inner + : outer; + } + + EIGEN_DEVICE_FUNC const Scalar* dstDataPtr() const + { + return m_dstExpr.data(); + } + +protected: + DstEvaluatorType& m_dst; + const SrcEvaluatorType& m_src; + const Functor &m_functor; + // TODO find a way to avoid the needs of the original expression + DstXprType& m_dstExpr; +}; + +/*************************************************************************** +* Part 5 : Entry point for dense rectangular assignment +***************************************************************************/ + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const Functor &/*func*/) +{ + EIGEN_ONLY_USED_FOR_DEBUG(dst); + EIGEN_ONLY_USED_FOR_DEBUG(src); + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const internal::assign_op &/*func*/) +{ + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if(((dst.rows()!=dstRows) || (dst.cols()!=dstCols))) + dst.resize(dstRows, dstCols); + eigen_assert(dst.rows() == dstRows && dst.cols() == dstCols); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src, const Functor &func) +{ + typedef evaluator DstEvaluatorType; + typedef evaluator SrcEvaluatorType; + + SrcEvaluatorType srcEvaluator(src); + + // NOTE To properly handle A = (A*A.transpose())/s with A rectangular, + // we need to resize the destination after the source evaluator has been created. + resize_if_allowed(dst, src, func); + + DstEvaluatorType dstEvaluator(dst); + + typedef generic_dense_assignment_kernel Kernel; + Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived()); + + dense_assignment_loop::run(kernel); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src) +{ + call_dense_assignment_loop(dst, src, internal::assign_op()); +} + +/*************************************************************************** +* Part 6 : Generic assignment +***************************************************************************/ + +// Based on the respective shapes of the destination and source, +// the class AssignmentKind determine the kind of assignment mechanism. +// AssignmentKind must define a Kind typedef. +template struct AssignmentKind; + +// Assignement kind defined in this file: +struct Dense2Dense {}; +struct EigenBase2EigenBase {}; + +template struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; +template<> struct AssignmentKind { typedef Dense2Dense Kind; }; + +// This is the main assignment class +template< typename DstXprType, typename SrcXprType, typename Functor, + typename Kind = typename AssignmentKind< typename evaluator_traits::Shape , typename evaluator_traits::Shape >::Kind, + typename EnableIf = void> +struct Assignment; + + +// The only purpose of this call_assignment() function is to deal with noalias() / "assume-aliasing" and automatic transposition. +// Indeed, I (Gael) think that this concept of "assume-aliasing" was a mistake, and it makes thing quite complicated. +// So this intermediate function removes everything related to "assume-aliasing" such that Assignment +// does not has to bother about these annoying details. + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src) +{ + call_assignment(dst, src, internal::assign_op()); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(const Dst& dst, const Src& src) +{ + call_assignment(dst, src, internal::assign_op()); +} + +// Deal with "assume-aliasing" +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if< evaluator_assume_aliasing::value, void*>::type = 0) +{ + typename plain_matrix_type::type tmp(src); + call_assignment_no_alias(dst, tmp, func); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if::value, void*>::type = 0) +{ + call_assignment_no_alias(dst, src, func); +} + +// by-pass "assume-aliasing" +// When there is no aliasing, we require that 'dst' has been properly resized +template class StorageBase, typename Src, typename Func> +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment(NoAlias& dst, const Src& src, const Func& func) +{ + call_assignment_no_alias(dst.expression(), src, func); +} + + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias(Dst& dst, const Src& src, const Func& func) +{ + enum { + NeedToTranspose = ( (int(Dst::RowsAtCompileTime) == 1 && int(Src::ColsAtCompileTime) == 1) + || (int(Dst::ColsAtCompileTime) == 1 && int(Src::RowsAtCompileTime) == 1) + ) && int(Dst::SizeAtCompileTime) != 1 + }; + + typedef typename internal::conditional, Dst>::type ActualDstTypeCleaned; + typedef typename internal::conditional, Dst&>::type ActualDstType; + ActualDstType actualDst(dst); + + // TODO check whether this is the right place to perform these checks: + EIGEN_STATIC_ASSERT_LVALUE(Dst) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(ActualDstTypeCleaned,Src) + EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename ActualDstTypeCleaned::Scalar,typename Src::Scalar); + + Assignment::run(actualDst, src, func); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias(Dst& dst, const Src& src) +{ + call_assignment_no_alias(dst, src, internal::assign_op()); +} + +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src, const Func& func) +{ + // TODO check whether this is the right place to perform these checks: + EIGEN_STATIC_ASSERT_LVALUE(Dst) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Dst,Src) + EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename Dst::Scalar,typename Src::Scalar); + + Assignment::run(dst, src, func); +} +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE +void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src) +{ + call_assignment_no_alias_no_transpose(dst, src, internal::assign_op()); +} + +// forward declaration +template void check_for_aliasing(const Dst &dst, const Src &src); + +// Generic Dense to Dense assignment +// Note that the last template argument "Weak" is needed to make it possible to perform +// both partial specialization+SFINAE without ambiguous specialization +template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak> +struct Assignment +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const Functor &func) + { +#ifndef EIGEN_NO_DEBUG + internal::check_for_aliasing(dst, src); +#endif + + call_dense_assignment_loop(dst, src, func); + } +}; + +// Generic assignment through evalTo. +// TODO: not sure we have to keep that one, but it helps porting current code to new evaluator mechanism. +// Note that the last template argument "Weak" is needed to make it possible to perform +// both partial specialization+SFINAE without ambiguous specialization +template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak> +struct Assignment +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.evalTo(dst); + } + + // NOTE The following two functions are templated to avoid their instanciation if not needed + // This is needed because some expressions supports evalTo only and/or have 'void' as scalar type. + template + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.addTo(dst); + } + + template + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + src.subTo(dst); + } +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_ASSIGN_EVALUATOR_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h new file mode 100644 index 00000000000..4978c914057 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BandMatrix.h @@ -0,0 +1,353 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_BANDMATRIX_H +#define EIGEN_BANDMATRIX_H + +namespace Eigen { + +namespace internal { + +template +class BandMatrixBase : public EigenBase +{ + public: + + enum { + Flags = internal::traits::Flags, + CoeffReadCost = internal::traits::CoeffReadCost, + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, + Supers = internal::traits::Supers, + Subs = internal::traits::Subs, + Options = internal::traits::Options + }; + typedef typename internal::traits::Scalar Scalar; + typedef Matrix DenseMatrixType; + typedef typename DenseMatrixType::StorageIndex StorageIndex; + typedef typename internal::traits::CoefficientsType CoefficientsType; + typedef EigenBase Base; + + protected: + enum { + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) + ? 1 + Supers + Subs + : Dynamic, + SizeAtCompileTime = EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime,ColsAtCompileTime) + }; + + public: + + using Base::derived; + using Base::rows; + using Base::cols; + + /** \returns the number of super diagonals */ + inline Index supers() const { return derived().supers(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return derived().subs(); } + + /** \returns an expression of the underlying coefficient matrix */ + inline const CoefficientsType& coeffs() const { return derived().coeffs(); } + + /** \returns an expression of the underlying coefficient matrix */ + inline CoefficientsType& coeffs() { return derived().coeffs(); } + + /** \returns a vector expression of the \a i -th column, + * only the meaningful part is returned. + * \warning the internal storage must be column major. */ + inline Block col(Index i) + { + EIGEN_STATIC_ASSERT((Options&RowMajor)==0,THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES); + Index start = 0; + Index len = coeffs().rows(); + if (i<=supers()) + { + start = supers()-i; + len = (std::min)(rows(),std::max(0,coeffs().rows() - (supers()-i))); + } + else if (i>=rows()-subs()) + len = std::max(0,coeffs().rows() - (i + 1 - rows() + subs())); + return Block(coeffs(), start, i, len, 1); + } + + /** \returns a vector expression of the main diagonal */ + inline Block diagonal() + { return Block(coeffs(),supers(),0,1,(std::min)(rows(),cols())); } + + /** \returns a vector expression of the main diagonal (const version) */ + inline const Block diagonal() const + { return Block(coeffs(),supers(),0,1,(std::min)(rows(),cols())); } + + template struct DiagonalIntReturnType { + enum { + ReturnOpposite = (Options&SelfAdjoint) && (((Index)>0 && Supers==0) || ((Index)<0 && Subs==0)), + Conjugate = ReturnOpposite && NumTraits::IsComplex, + ActualIndex = ReturnOpposite ? -Index : Index, + DiagonalSize = (RowsAtCompileTime==Dynamic || ColsAtCompileTime==Dynamic) + ? Dynamic + : (ActualIndex<0 + ? EIGEN_SIZE_MIN_PREFER_DYNAMIC(ColsAtCompileTime, RowsAtCompileTime + ActualIndex) + : EIGEN_SIZE_MIN_PREFER_DYNAMIC(RowsAtCompileTime, ColsAtCompileTime - ActualIndex)) + }; + typedef Block BuildType; + typedef typename internal::conditional,BuildType >, + BuildType>::type Type; + }; + + /** \returns a vector expression of the \a N -th sub or super diagonal */ + template inline typename DiagonalIntReturnType::Type diagonal() + { + return typename DiagonalIntReturnType::BuildType(coeffs(), supers()-N, (std::max)(0,N), 1, diagonalLength(N)); + } + + /** \returns a vector expression of the \a N -th sub or super diagonal */ + template inline const typename DiagonalIntReturnType::Type diagonal() const + { + return typename DiagonalIntReturnType::BuildType(coeffs(), supers()-N, (std::max)(0,N), 1, diagonalLength(N)); + } + + /** \returns a vector expression of the \a i -th sub or super diagonal */ + inline Block diagonal(Index i) + { + eigen_assert((i<0 && -i<=subs()) || (i>=0 && i<=supers())); + return Block(coeffs(), supers()-i, std::max(0,i), 1, diagonalLength(i)); + } + + /** \returns a vector expression of the \a i -th sub or super diagonal */ + inline const Block diagonal(Index i) const + { + eigen_assert((i<0 && -i<=subs()) || (i>=0 && i<=supers())); + return Block(coeffs(), supers()-i, std::max(0,i), 1, diagonalLength(i)); + } + + template inline void evalTo(Dest& dst) const + { + dst.resize(rows(),cols()); + dst.setZero(); + dst.diagonal() = diagonal(); + for (Index i=1; i<=supers();++i) + dst.diagonal(i) = diagonal(i); + for (Index i=1; i<=subs();++i) + dst.diagonal(-i) = diagonal(-i); + } + + DenseMatrixType toDenseMatrix() const + { + DenseMatrixType res(rows(),cols()); + evalTo(res); + return res; + } + + protected: + + inline Index diagonalLength(Index i) const + { return i<0 ? (std::min)(cols(),rows()+i) : (std::min)(rows(),cols()-i); } +}; + +/** + * \class BandMatrix + * \ingroup Core_Module + * + * \brief Represents a rectangular matrix with a banded storage + * + * \tparam _Scalar Numeric type, i.e. float, double, int + * \tparam _Rows Number of rows, or \b Dynamic + * \tparam _Cols Number of columns, or \b Dynamic + * \tparam _Supers Number of super diagonal + * \tparam _Subs Number of sub diagonal + * \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of \b #SelfAdjoint + * The former controls \ref TopicStorageOrders "storage order", and defaults to + * column-major. The latter controls whether the matrix represents a selfadjoint + * matrix in which case either Supers of Subs have to be null. + * + * \sa class TridiagonalMatrix + */ + +template +struct traits > +{ + typedef _Scalar Scalar; + typedef Dense StorageKind; + typedef Eigen::Index StorageIndex; + enum { + CoeffReadCost = NumTraits::ReadCost, + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _Rows, + MaxColsAtCompileTime = _Cols, + Flags = LvalueBit, + Supers = _Supers, + Subs = _Subs, + Options = _Options, + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) ? 1 + Supers + Subs : Dynamic + }; + typedef Matrix CoefficientsType; +}; + +template +class BandMatrix : public BandMatrixBase > +{ + public: + + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::traits::StorageIndex StorageIndex; + typedef typename internal::traits::CoefficientsType CoefficientsType; + + explicit inline BandMatrix(Index rows=Rows, Index cols=Cols, Index supers=Supers, Index subs=Subs) + : m_coeffs(1+supers+subs,cols), + m_rows(rows), m_supers(supers), m_subs(subs) + { + } + + /** \returns the number of columns */ + inline Index rows() const { return m_rows.value(); } + + /** \returns the number of rows */ + inline Index cols() const { return m_coeffs.cols(); } + + /** \returns the number of super diagonals */ + inline Index supers() const { return m_supers.value(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return m_subs.value(); } + + inline const CoefficientsType& coeffs() const { return m_coeffs; } + inline CoefficientsType& coeffs() { return m_coeffs; } + + protected: + + CoefficientsType m_coeffs; + internal::variable_if_dynamic m_rows; + internal::variable_if_dynamic m_supers; + internal::variable_if_dynamic m_subs; +}; + +template +class BandMatrixWrapper; + +template +struct traits > +{ + typedef typename _CoefficientsType::Scalar Scalar; + typedef typename _CoefficientsType::StorageKind StorageKind; + typedef typename _CoefficientsType::StorageIndex StorageIndex; + enum { + CoeffReadCost = internal::traits<_CoefficientsType>::CoeffReadCost, + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _Rows, + MaxColsAtCompileTime = _Cols, + Flags = LvalueBit, + Supers = _Supers, + Subs = _Subs, + Options = _Options, + DataRowsAtCompileTime = ((Supers!=Dynamic) && (Subs!=Dynamic)) ? 1 + Supers + Subs : Dynamic + }; + typedef _CoefficientsType CoefficientsType; +}; + +template +class BandMatrixWrapper : public BandMatrixBase > +{ + public: + + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::traits::CoefficientsType CoefficientsType; + typedef typename internal::traits::StorageIndex StorageIndex; + + explicit inline BandMatrixWrapper(const CoefficientsType& coeffs, Index rows=_Rows, Index cols=_Cols, Index supers=_Supers, Index subs=_Subs) + : m_coeffs(coeffs), + m_rows(rows), m_supers(supers), m_subs(subs) + { + EIGEN_UNUSED_VARIABLE(cols); + //internal::assert(coeffs.cols()==cols() && (supers()+subs()+1)==coeffs.rows()); + } + + /** \returns the number of columns */ + inline Index rows() const { return m_rows.value(); } + + /** \returns the number of rows */ + inline Index cols() const { return m_coeffs.cols(); } + + /** \returns the number of super diagonals */ + inline Index supers() const { return m_supers.value(); } + + /** \returns the number of sub diagonals */ + inline Index subs() const { return m_subs.value(); } + + inline const CoefficientsType& coeffs() const { return m_coeffs; } + + protected: + + const CoefficientsType& m_coeffs; + internal::variable_if_dynamic m_rows; + internal::variable_if_dynamic m_supers; + internal::variable_if_dynamic m_subs; +}; + +/** + * \class TridiagonalMatrix + * \ingroup Core_Module + * + * \brief Represents a tridiagonal matrix with a compact banded storage + * + * \tparam Scalar Numeric type, i.e. float, double, int + * \tparam Size Number of rows and cols, or \b Dynamic + * \tparam Options Can be 0 or \b SelfAdjoint + * + * \sa class BandMatrix + */ +template +class TridiagonalMatrix : public BandMatrix +{ + typedef BandMatrix Base; + typedef typename Base::StorageIndex StorageIndex; + public: + explicit TridiagonalMatrix(Index size = Size) : Base(size,size,Options&SelfAdjoint?0:1,1) {} + + inline typename Base::template DiagonalIntReturnType<1>::Type super() + { return Base::template diagonal<1>(); } + inline const typename Base::template DiagonalIntReturnType<1>::Type super() const + { return Base::template diagonal<1>(); } + inline typename Base::template DiagonalIntReturnType<-1>::Type sub() + { return Base::template diagonal<-1>(); } + inline const typename Base::template DiagonalIntReturnType<-1>::Type sub() const + { return Base::template diagonal<-1>(); } + protected: +}; + + +struct BandShape {}; + +template +struct evaluator_traits > + : public evaluator_traits_base > +{ + typedef BandShape Shape; +}; + +template +struct evaluator_traits > + : public evaluator_traits_base > +{ + typedef BandShape Shape; +}; + +template<> struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_BANDMATRIX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Block.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Block.h new file mode 100644 index 00000000000..11de45c2ecb --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Block.h @@ -0,0 +1,452 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_BLOCK_H +#define EIGEN_BLOCK_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits +{ + typedef typename traits::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type XprTypeNested; + typedef typename remove_reference::type _XprTypeNested; + enum{ + MatrixRows = traits::RowsAtCompileTime, + MatrixCols = traits::ColsAtCompileTime, + RowsAtCompileTime = MatrixRows == 0 ? 0 : BlockRows, + ColsAtCompileTime = MatrixCols == 0 ? 0 : BlockCols, + MaxRowsAtCompileTime = BlockRows==0 ? 0 + : RowsAtCompileTime != Dynamic ? int(RowsAtCompileTime) + : int(traits::MaxRowsAtCompileTime), + MaxColsAtCompileTime = BlockCols==0 ? 0 + : ColsAtCompileTime != Dynamic ? int(ColsAtCompileTime) + : int(traits::MaxColsAtCompileTime), + + XprTypeIsRowMajor = (int(traits::Flags)&RowMajorBit) != 0, + IsRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0 + : XprTypeIsRowMajor, + HasSameStorageOrderAsXprType = (IsRowMajor == XprTypeIsRowMajor), + InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + InnerStrideAtCompileTime = HasSameStorageOrderAsXprType + ? int(inner_stride_at_compile_time::ret) + : int(outer_stride_at_compile_time::ret), + OuterStrideAtCompileTime = HasSameStorageOrderAsXprType + ? int(outer_stride_at_compile_time::ret) + : int(inner_stride_at_compile_time::ret), + + // FIXME, this traits is rather specialized for dense object and it needs to be cleaned further + FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, + FlagsRowMajorBit = IsRowMajor ? RowMajorBit : 0, + Flags = (traits::Flags & (DirectAccessBit | (InnerPanel?CompressedAccessBit:0))) | FlagsLvalueBit | FlagsRowMajorBit, + // FIXME DirectAccessBit should not be handled by expressions + // + // Alignment is needed by MapBase's assertions + // We can sefely set it to false here. Internal alignment errors will be detected by an eigen_internal_assert in the respective evaluator + Alignment = 0 + }; +}; + +template::ret> class BlockImpl_dense; + +} // end namespace internal + +template class BlockImpl; + +/** \class Block + * \ingroup Core_Module + * + * \brief Expression of a fixed-size or dynamic-size block + * + * \tparam XprType the type of the expression in which we are taking a block + * \tparam BlockRows the number of rows of the block we are taking at compile time (optional) + * \tparam BlockCols the number of columns of the block we are taking at compile time (optional) + * \tparam InnerPanel is true, if the block maps to a set of rows of a row major matrix or + * to set of columns of a column major matrix (optional). The parameter allows to determine + * at compile time whether aligned access is possible on the block expression. + * + * This class represents an expression of either a fixed-size or dynamic-size block. It is the return + * type of DenseBase::block(Index,Index,Index,Index) and DenseBase::block(Index,Index) and + * most of the time this is the only way it is used. + * + * However, if you want to directly maniputate block expressions, + * for instance if you want to write a function returning such an expression, you + * will need to use this class. + * + * Here is an example illustrating the dynamic case: + * \include class_Block.cpp + * Output: \verbinclude class_Block.out + * + * \note Even though this expression has dynamic size, in the case where \a XprType + * has fixed size, this expression inherits a fixed maximal size which means that evaluating + * it does not cause a dynamic memory allocation. + * + * Here is an example illustrating the fixed-size case: + * \include class_FixedBlock.cpp + * Output: \verbinclude class_FixedBlock.out + * + * \sa DenseBase::block(Index,Index,Index,Index), DenseBase::block(Index,Index), class VectorBlock + */ +template class Block + : public BlockImpl::StorageKind> +{ + typedef BlockImpl::StorageKind> Impl; + public: + //typedef typename Impl::Base Base; + typedef Impl Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(Block) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block) + + typedef typename internal::remove_all::type NestedExpression; + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline Block(XprType& xpr, Index i) : Impl(xpr,i) + { + eigen_assert( (i>=0) && ( + ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && i= 0 && BlockRows >= 0 && startRow + BlockRows <= xpr.rows() + && startCol >= 0 && BlockCols >= 0 && startCol + BlockCols <= xpr.cols()); + } + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline Block(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : Impl(xpr, startRow, startCol, blockRows, blockCols) + { + eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows) + && (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols)); + eigen_assert(startRow >= 0 && blockRows >= 0 && startRow <= xpr.rows() - blockRows + && startCol >= 0 && blockCols >= 0 && startCol <= xpr.cols() - blockCols); + } +}; + +// The generic default implementation for dense block simplu forward to the internal::BlockImpl_dense +// that must be specialized for direct and non-direct access... +template +class BlockImpl + : public internal::BlockImpl_dense +{ + typedef internal::BlockImpl_dense Impl; + typedef typename XprType::StorageIndex StorageIndex; + public: + typedef Impl Base; + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl) + EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index i) : Impl(xpr,i) {} + EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index startRow, Index startCol) : Impl(xpr, startRow, startCol) {} + EIGEN_DEVICE_FUNC + inline BlockImpl(XprType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols) + : Impl(xpr, startRow, startCol, blockRows, blockCols) {} +}; + +namespace internal { + +/** \internal Internal implementation of dense Blocks in the general case. */ +template class BlockImpl_dense + : public internal::dense_xpr_base >::type +{ + typedef Block BlockType; + typedef typename internal::ref_selector::non_const_type XprTypeNested; + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(BlockType) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense) + + // class InnerIterator; // FIXME apparently never used + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index i) + : m_xpr(xpr), + // It is a row if and only if BlockRows==1 and BlockCols==XprType::ColsAtCompileTime, + // and it is a column if and only if BlockRows==XprType::RowsAtCompileTime and BlockCols==1, + // all other cases are invalid. + // The case a 1x1 matrix seems ambiguous, but the result is the same anyway. + m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0), + m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0), + m_blockRows(BlockRows==1 ? 1 : xpr.rows()), + m_blockCols(BlockCols==1 ? 1 : xpr.cols()) + {} + + /** Fixed-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol) + : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol), + m_blockRows(BlockRows), m_blockCols(BlockCols) + {} + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : m_xpr(xpr), m_startRow(startRow), m_startCol(startCol), + m_blockRows(blockRows), m_blockCols(blockCols) + {} + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_blockRows.value(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_blockCols.value(); } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index rowId, Index colId) + { + EIGEN_STATIC_ASSERT_LVALUE(XprType) + return m_xpr.coeffRef(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return m_xpr.derived().coeffRef(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index rowId, Index colId) const + { + return m_xpr.coeff(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index index) + { + EIGEN_STATIC_ASSERT_LVALUE(XprType) + return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + EIGEN_DEVICE_FUNC + inline const CoeffReturnType coeff(Index index) const + { + return m_xpr.coeff(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + template + inline PacketScalar packet(Index rowId, Index colId) const + { + return m_xpr.template packet(rowId + m_startRow.value(), colId + m_startCol.value()); + } + + template + inline void writePacket(Index rowId, Index colId, const PacketScalar& val) + { + m_xpr.template writePacket(rowId + m_startRow.value(), colId + m_startCol.value(), val); + } + + template + inline PacketScalar packet(Index index) const + { + return m_xpr.template packet + (m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0)); + } + + template + inline void writePacket(Index index, const PacketScalar& val) + { + m_xpr.template writePacket + (m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index), + m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0), val); + } + + #ifdef EIGEN_PARSED_BY_DOXYGEN + /** \sa MapBase::data() */ + EIGEN_DEVICE_FUNC inline const Scalar* data() const; + EIGEN_DEVICE_FUNC inline Index innerStride() const; + EIGEN_DEVICE_FUNC inline Index outerStride() const; + #endif + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& nestedExpression() const + { + return m_xpr; + } + + EIGEN_DEVICE_FUNC + XprType& nestedExpression() { return m_xpr; } + + EIGEN_DEVICE_FUNC + StorageIndex startRow() const + { + return m_startRow.value(); + } + + EIGEN_DEVICE_FUNC + StorageIndex startCol() const + { + return m_startCol.value(); + } + + protected: + + XprTypeNested m_xpr; + const internal::variable_if_dynamic m_startRow; + const internal::variable_if_dynamic m_startCol; + const internal::variable_if_dynamic m_blockRows; + const internal::variable_if_dynamic m_blockCols; +}; + +/** \internal Internal implementation of dense Blocks in the direct access case.*/ +template +class BlockImpl_dense + : public MapBase > +{ + typedef Block BlockType; + typedef typename internal::ref_selector::non_const_type XprTypeNested; + enum { + XprTypeIsRowMajor = (int(traits::Flags)&RowMajorBit) != 0 + }; + public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(BlockType) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense) + + /** Column or Row constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index i) + : Base(xpr.data() + i * ( ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && (!XprTypeIsRowMajor)) + || ((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && ( XprTypeIsRowMajor)) ? xpr.innerStride() : xpr.outerStride()), + BlockRows==1 ? 1 : xpr.rows(), + BlockCols==1 ? 1 : xpr.cols()), + m_xpr(xpr), + m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0), + m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0) + { + init(); + } + + /** Fixed-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol) + : Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol)), + m_xpr(xpr), m_startRow(startRow), m_startCol(startCol) + { + init(); + } + + /** Dynamic-size constructor + */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, + Index startRow, Index startCol, + Index blockRows, Index blockCols) + : Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol), blockRows, blockCols), + m_xpr(xpr), m_startRow(startRow), m_startCol(startCol) + { + init(); + } + + EIGEN_DEVICE_FUNC + const typename internal::remove_all::type& nestedExpression() const + { + return m_xpr; + } + + EIGEN_DEVICE_FUNC + XprType& nestedExpression() { return m_xpr; } + + /** \sa MapBase::innerStride() */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return internal::traits::HasSameStorageOrderAsXprType + ? m_xpr.innerStride() + : m_xpr.outerStride(); + } + + /** \sa MapBase::outerStride() */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return m_outerStride; + } + + EIGEN_DEVICE_FUNC + StorageIndex startRow() const + { + return m_startRow.value(); + } + + EIGEN_DEVICE_FUNC + StorageIndex startCol() const + { + return m_startCol.value(); + } + + #ifndef __SUNPRO_CC + // FIXME sunstudio is not friendly with the above friend... + // META-FIXME there is no 'friend' keyword around here. Is this obsolete? + protected: + #endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal used by allowAligned() */ + EIGEN_DEVICE_FUNC + inline BlockImpl_dense(XprType& xpr, const Scalar* data, Index blockRows, Index blockCols) + : Base(data, blockRows, blockCols), m_xpr(xpr) + { + init(); + } + #endif + + protected: + EIGEN_DEVICE_FUNC + void init() + { + m_outerStride = internal::traits::HasSameStorageOrderAsXprType + ? m_xpr.outerStride() + : m_xpr.innerStride(); + } + + XprTypeNested m_xpr; + const internal::variable_if_dynamic m_startRow; + const internal::variable_if_dynamic m_startCol; + Index m_outerStride; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_BLOCK_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h new file mode 100644 index 00000000000..8409d8749ad --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/BooleanRedux.h @@ -0,0 +1,164 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_ALLANDANY_H +#define EIGEN_ALLANDANY_H + +namespace Eigen { + +namespace internal { + +template +struct all_unroller +{ + typedef typename Derived::ExpressionTraits Traits; + enum { + col = (UnrollCount-1) / Traits::RowsAtCompileTime, + row = (UnrollCount-1) % Traits::RowsAtCompileTime + }; + + static inline bool run(const Derived &mat) + { + return all_unroller::run(mat) && mat.coeff(row, col); + } +}; + +template +struct all_unroller +{ + static inline bool run(const Derived &/*mat*/) { return true; } +}; + +template +struct all_unroller +{ + static inline bool run(const Derived &) { return false; } +}; + +template +struct any_unroller +{ + typedef typename Derived::ExpressionTraits Traits; + enum { + col = (UnrollCount-1) / Traits::RowsAtCompileTime, + row = (UnrollCount-1) % Traits::RowsAtCompileTime + }; + + static inline bool run(const Derived &mat) + { + return any_unroller::run(mat) || mat.coeff(row, col); + } +}; + +template +struct any_unroller +{ + static inline bool run(const Derived & /*mat*/) { return false; } +}; + +template +struct any_unroller +{ + static inline bool run(const Derived &) { return false; } +}; + +} // end namespace internal + +/** \returns true if all coefficients are true + * + * Example: \include MatrixBase_all.cpp + * Output: \verbinclude MatrixBase_all.out + * + * \sa any(), Cwise::operator<() + */ +template +inline bool DenseBase::all() const +{ + typedef internal::evaluator Evaluator; + enum { + unroll = SizeAtCompileTime != Dynamic + && SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits::AddCost) <= EIGEN_UNROLLING_LIMIT + }; + Evaluator evaluator(derived()); + if(unroll) + return internal::all_unroller::run(evaluator); + else + { + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if (!evaluator.coeff(i, j)) return false; + return true; + } +} + +/** \returns true if at least one coefficient is true + * + * \sa all() + */ +template +inline bool DenseBase::any() const +{ + typedef internal::evaluator Evaluator; + enum { + unroll = SizeAtCompileTime != Dynamic + && SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits::AddCost) <= EIGEN_UNROLLING_LIMIT + }; + Evaluator evaluator(derived()); + if(unroll) + return internal::any_unroller::run(evaluator); + else + { + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if (evaluator.coeff(i, j)) return true; + return false; + } +} + +/** \returns the number of coefficients which evaluate to true + * + * \sa all(), any() + */ +template +inline Eigen::Index DenseBase::count() const +{ + return derived().template cast().template cast().sum(); +} + +/** \returns true is \c *this contains at least one Not A Number (NaN). + * + * \sa allFinite() + */ +template +inline bool DenseBase::hasNaN() const +{ +#if EIGEN_COMP_MSVC || (defined __FAST_MATH__) + return derived().array().isNaN().any(); +#else + return !((derived().array()==derived().array()).all()); +#endif +} + +/** \returns true if \c *this contains only finite numbers, i.e., no NaN and no +/-INF values. + * + * \sa hasNaN() + */ +template +inline bool DenseBase::allFinite() const +{ +#if EIGEN_COMP_MSVC || (defined __FAST_MATH__) + return derived().array().isFinite().all(); +#else + return !((derived()-derived()).hasNaN()); +#endif +} + +} // end namespace Eigen + +#endif // EIGEN_ALLANDANY_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h new file mode 100644 index 00000000000..d218e98143f --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CommaInitializer.h @@ -0,0 +1,160 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_COMMAINITIALIZER_H +#define EIGEN_COMMAINITIALIZER_H + +namespace Eigen { + +/** \class CommaInitializer + * \ingroup Core_Module + * + * \brief Helper class used by the comma initializer operator + * + * This class is internally used to implement the comma initializer feature. It is + * the return type of MatrixBase::operator<<, and most of the time this is the only + * way it is used. + * + * \sa \blank \ref MatrixBaseCommaInitRef "MatrixBase::operator<<", CommaInitializer::finished() + */ +template +struct CommaInitializer +{ + typedef typename XprType::Scalar Scalar; + + EIGEN_DEVICE_FUNC + inline CommaInitializer(XprType& xpr, const Scalar& s) + : m_xpr(xpr), m_row(0), m_col(1), m_currentBlockRows(1) + { + m_xpr.coeffRef(0,0) = s; + } + + template + EIGEN_DEVICE_FUNC + inline CommaInitializer(XprType& xpr, const DenseBase& other) + : m_xpr(xpr), m_row(0), m_col(other.cols()), m_currentBlockRows(other.rows()) + { + m_xpr.block(0, 0, other.rows(), other.cols()) = other; + } + + /* Copy/Move constructor which transfers ownership. This is crucial in + * absence of return value optimization to avoid assertions during destruction. */ + // FIXME in C++11 mode this could be replaced by a proper RValue constructor + EIGEN_DEVICE_FUNC + inline CommaInitializer(const CommaInitializer& o) + : m_xpr(o.m_xpr), m_row(o.m_row), m_col(o.m_col), m_currentBlockRows(o.m_currentBlockRows) { + // Mark original object as finished. In absence of R-value references we need to const_cast: + const_cast(o).m_row = m_xpr.rows(); + const_cast(o).m_col = m_xpr.cols(); + const_cast(o).m_currentBlockRows = 0; + } + + /* inserts a scalar value in the target matrix */ + EIGEN_DEVICE_FUNC + CommaInitializer& operator,(const Scalar& s) + { + if (m_col==m_xpr.cols()) + { + m_row+=m_currentBlockRows; + m_col = 0; + m_currentBlockRows = 1; + eigen_assert(m_row + EIGEN_DEVICE_FUNC + CommaInitializer& operator,(const DenseBase& other) + { + if (m_col==m_xpr.cols() && (other.cols()!=0 || other.rows()!=m_currentBlockRows)) + { + m_row+=m_currentBlockRows; + m_col = 0; + m_currentBlockRows = other.rows(); + eigen_assert(m_row+m_currentBlockRows<=m_xpr.rows() + && "Too many rows passed to comma initializer (operator<<)"); + } + eigen_assert((m_col + other.cols() <= m_xpr.cols()) + && "Too many coefficients passed to comma initializer (operator<<)"); + eigen_assert(m_currentBlockRows==other.rows()); + m_xpr.template block + (m_row, m_col, other.rows(), other.cols()) = other; + m_col += other.cols(); + return *this; + } + + EIGEN_DEVICE_FUNC + inline ~CommaInitializer() +#if defined VERIFY_RAISES_ASSERT && (!defined EIGEN_NO_ASSERTION_CHECKING) && defined EIGEN_EXCEPTIONS + EIGEN_EXCEPTION_SPEC(Eigen::eigen_assert_exception) +#endif + { + finished(); + } + + /** \returns the built matrix once all its coefficients have been set. + * Calling finished is 100% optional. Its purpose is to write expressions + * like this: + * \code + * quaternion.fromRotationMatrix((Matrix3f() << axis0, axis1, axis2).finished()); + * \endcode + */ + EIGEN_DEVICE_FUNC + inline XprType& finished() { + eigen_assert(((m_row+m_currentBlockRows) == m_xpr.rows() || m_xpr.cols() == 0) + && m_col == m_xpr.cols() + && "Too few coefficients passed to comma initializer (operator<<)"); + return m_xpr; + } + + XprType& m_xpr; // target expression + Index m_row; // current row id + Index m_col; // current col id + Index m_currentBlockRows; // current block height +}; + +/** \anchor MatrixBaseCommaInitRef + * Convenient operator to set the coefficients of a matrix. + * + * The coefficients must be provided in a row major order and exactly match + * the size of the matrix. Otherwise an assertion is raised. + * + * Example: \include MatrixBase_set.cpp + * Output: \verbinclude MatrixBase_set.out + * + * \note According the c++ standard, the argument expressions of this comma initializer are evaluated in arbitrary order. + * + * \sa CommaInitializer::finished(), class CommaInitializer + */ +template +inline CommaInitializer DenseBase::operator<< (const Scalar& s) +{ + return CommaInitializer(*static_cast(this), s); +} + +/** \sa operator<<(const Scalar&) */ +template +template +inline CommaInitializer +DenseBase::operator<<(const DenseBase& other) +{ + return CommaInitializer(*static_cast(this), other); +} + +} // end namespace Eigen + +#endif // EIGEN_COMMAINITIALIZER_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h new file mode 100644 index 00000000000..51a2e5f1b6f --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ConditionEstimator.h @@ -0,0 +1,175 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2016 Rasmus Munk Larsen (rmlarsen@google.com) +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CONDITIONESTIMATOR_H +#define EIGEN_CONDITIONESTIMATOR_H + +namespace Eigen { + +namespace internal { + +template +struct rcond_compute_sign { + static inline Vector run(const Vector& v) { + const RealVector v_abs = v.cwiseAbs(); + return (v_abs.array() == static_cast(0)) + .select(Vector::Ones(v.size()), v.cwiseQuotient(v_abs)); + } +}; + +// Partial specialization to avoid elementwise division for real vectors. +template +struct rcond_compute_sign { + static inline Vector run(const Vector& v) { + return (v.array() < static_cast(0)) + .select(-Vector::Ones(v.size()), Vector::Ones(v.size())); + } +}; + +/** + * \returns an estimate of ||inv(matrix)||_1 given a decomposition of + * \a matrix that implements .solve() and .adjoint().solve() methods. + * + * This function implements Algorithms 4.1 and 5.1 from + * http://www.maths.manchester.ac.uk/~higham/narep/narep135.pdf + * which also forms the basis for the condition number estimators in + * LAPACK. Since at most 10 calls to the solve method of dec are + * performed, the total cost is O(dims^2), as opposed to O(dims^3) + * needed to compute the inverse matrix explicitly. + * + * The most common usage is in estimating the condition number + * ||matrix||_1 * ||inv(matrix)||_1. The first term ||matrix||_1 can be + * computed directly in O(n^2) operations. + * + * Supports the following decompositions: FullPivLU, PartialPivLU, LDLT, and + * LLT. + * + * \sa FullPivLU, PartialPivLU, LDLT, LLT. + */ +template +typename Decomposition::RealScalar rcond_invmatrix_L1_norm_estimate(const Decomposition& dec) +{ + typedef typename Decomposition::MatrixType MatrixType; + typedef typename Decomposition::Scalar Scalar; + typedef typename Decomposition::RealScalar RealScalar; + typedef typename internal::plain_col_type::type Vector; + typedef typename internal::plain_col_type::type RealVector; + const bool is_complex = (NumTraits::IsComplex != 0); + + eigen_assert(dec.rows() == dec.cols()); + const Index n = dec.rows(); + if (n == 0) + return 0; + + // Disable Index to float conversion warning +#ifdef __INTEL_COMPILER + #pragma warning push + #pragma warning ( disable : 2259 ) +#endif + Vector v = dec.solve(Vector::Ones(n) / Scalar(n)); +#ifdef __INTEL_COMPILER + #pragma warning pop +#endif + + // lower_bound is a lower bound on + // ||inv(matrix)||_1 = sup_v ||inv(matrix) v||_1 / ||v||_1 + // and is the objective maximized by the ("super-") gradient ascent + // algorithm below. + RealScalar lower_bound = v.template lpNorm<1>(); + if (n == 1) + return lower_bound; + + // Gradient ascent algorithm follows: We know that the optimum is achieved at + // one of the simplices v = e_i, so in each iteration we follow a + // super-gradient to move towards the optimal one. + RealScalar old_lower_bound = lower_bound; + Vector sign_vector(n); + Vector old_sign_vector; + Index v_max_abs_index = -1; + Index old_v_max_abs_index = v_max_abs_index; + for (int k = 0; k < 4; ++k) + { + sign_vector = internal::rcond_compute_sign::run(v); + if (k > 0 && !is_complex && sign_vector == old_sign_vector) { + // Break if the solution stagnated. + break; + } + // v_max_abs_index = argmax |real( inv(matrix)^T * sign_vector )| + v = dec.adjoint().solve(sign_vector); + v.real().cwiseAbs().maxCoeff(&v_max_abs_index); + if (v_max_abs_index == old_v_max_abs_index) { + // Break if the solution stagnated. + break; + } + // Move to the new simplex e_j, where j = v_max_abs_index. + v = dec.solve(Vector::Unit(n, v_max_abs_index)); // v = inv(matrix) * e_j. + lower_bound = v.template lpNorm<1>(); + if (lower_bound <= old_lower_bound) { + // Break if the gradient step did not increase the lower_bound. + break; + } + if (!is_complex) { + old_sign_vector = sign_vector; + } + old_v_max_abs_index = v_max_abs_index; + old_lower_bound = lower_bound; + } + // The following calculates an independent estimate of ||matrix||_1 by + // multiplying matrix by a vector with entries of slowly increasing + // magnitude and alternating sign: + // v_i = (-1)^{i} (1 + (i / (dim-1))), i = 0,...,dim-1. + // This improvement to Hager's algorithm above is due to Higham. It was + // added to make the algorithm more robust in certain corner cases where + // large elements in the matrix might otherwise escape detection due to + // exact cancellation (especially when op and op_adjoint correspond to a + // sequence of backsubstitutions and permutations), which could cause + // Hager's algorithm to vastly underestimate ||matrix||_1. + Scalar alternating_sign(RealScalar(1)); + for (Index i = 0; i < n; ++i) { + // The static_cast is needed when Scalar is a complex and RealScalar implements expression templates + v[i] = alternating_sign * static_cast(RealScalar(1) + (RealScalar(i) / (RealScalar(n - 1)))); + alternating_sign = -alternating_sign; + } + v = dec.solve(v); + const RealScalar alternate_lower_bound = (2 * v.template lpNorm<1>()) / (3 * RealScalar(n)); + return numext::maxi(lower_bound, alternate_lower_bound); +} + +/** \brief Reciprocal condition number estimator. + * + * Computing a decomposition of a dense matrix takes O(n^3) operations, while + * this method estimates the condition number quickly and reliably in O(n^2) + * operations. + * + * \returns an estimate of the reciprocal condition number + * (1 / (||matrix||_1 * ||inv(matrix)||_1)) of matrix, given ||matrix||_1 and + * its decomposition. Supports the following decompositions: FullPivLU, + * PartialPivLU, LDLT, and LLT. + * + * \sa FullPivLU, PartialPivLU, LDLT, LLT. + */ +template +typename Decomposition::RealScalar +rcond_estimate_helper(typename Decomposition::RealScalar matrix_norm, const Decomposition& dec) +{ + typedef typename Decomposition::RealScalar RealScalar; + eigen_assert(dec.rows() == dec.cols()); + if (dec.rows() == 0) return NumTraits::infinity(); + if (matrix_norm == RealScalar(0)) return RealScalar(0); + if (dec.rows() == 1) return RealScalar(1); + const RealScalar inverse_matrix_norm = rcond_invmatrix_L1_norm_estimate(dec); + return (inverse_matrix_norm == RealScalar(0) ? RealScalar(0) + : (RealScalar(1) / inverse_matrix_norm) / matrix_norm); +} + +} // namespace internal + +} // namespace Eigen + +#endif diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h new file mode 100644 index 00000000000..910889efa70 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreEvaluators.h @@ -0,0 +1,1688 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2011 Benoit Jacob +// Copyright (C) 2011-2014 Gael Guennebaud +// Copyright (C) 2011-2012 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + + +#ifndef EIGEN_COREEVALUATORS_H +#define EIGEN_COREEVALUATORS_H + +namespace Eigen { + +namespace internal { + +// This class returns the evaluator kind from the expression storage kind. +// Default assumes index based accessors +template +struct storage_kind_to_evaluator_kind { + typedef IndexBased Kind; +}; + +// This class returns the evaluator shape from the expression storage kind. +// It can be Dense, Sparse, Triangular, Diagonal, SelfAdjoint, Band, etc. +template struct storage_kind_to_shape; + +template<> struct storage_kind_to_shape { typedef DenseShape Shape; }; +template<> struct storage_kind_to_shape { typedef SolverShape Shape; }; +template<> struct storage_kind_to_shape { typedef PermutationShape Shape; }; +template<> struct storage_kind_to_shape { typedef TranspositionsShape Shape; }; + +// Evaluators have to be specialized with respect to various criteria such as: +// - storage/structure/shape +// - scalar type +// - etc. +// Therefore, we need specialization of evaluator providing additional template arguments for each kind of evaluators. +// We currently distinguish the following kind of evaluators: +// - unary_evaluator for expressions taking only one arguments (CwiseUnaryOp, CwiseUnaryView, Transpose, MatrixWrapper, ArrayWrapper, Reverse, Replicate) +// - binary_evaluator for expression taking two arguments (CwiseBinaryOp) +// - ternary_evaluator for expression taking three arguments (CwiseTernaryOp) +// - product_evaluator for linear algebra products (Product); special case of binary_evaluator because it requires additional tags for dispatching. +// - mapbase_evaluator for Map, Block, Ref +// - block_evaluator for Block (special dispatching to a mapbase_evaluator or unary_evaluator) + +template< typename T, + typename Arg1Kind = typename evaluator_traits::Kind, + typename Arg2Kind = typename evaluator_traits::Kind, + typename Arg3Kind = typename evaluator_traits::Kind, + typename Arg1Scalar = typename traits::Scalar, + typename Arg2Scalar = typename traits::Scalar, + typename Arg3Scalar = typename traits::Scalar> struct ternary_evaluator; + +template< typename T, + typename LhsKind = typename evaluator_traits::Kind, + typename RhsKind = typename evaluator_traits::Kind, + typename LhsScalar = typename traits::Scalar, + typename RhsScalar = typename traits::Scalar> struct binary_evaluator; + +template< typename T, + typename Kind = typename evaluator_traits::Kind, + typename Scalar = typename T::Scalar> struct unary_evaluator; + +// evaluator_traits contains traits for evaluator + +template +struct evaluator_traits_base +{ + // by default, get evaluator kind and shape from storage + typedef typename storage_kind_to_evaluator_kind::StorageKind>::Kind Kind; + typedef typename storage_kind_to_shape::StorageKind>::Shape Shape; +}; + +// Default evaluator traits +template +struct evaluator_traits : public evaluator_traits_base +{ +}; + +template::Shape > +struct evaluator_assume_aliasing { + static const bool value = false; +}; + +// By default, we assume a unary expression: +template +struct evaluator : public unary_evaluator +{ + typedef unary_evaluator Base; + EIGEN_DEVICE_FUNC explicit evaluator(const T& xpr) : Base(xpr) {} +}; + + +// TODO: Think about const-correctness +template +struct evaluator + : evaluator +{ + EIGEN_DEVICE_FUNC + explicit evaluator(const T& xpr) : evaluator(xpr) {} +}; + +// ---------- base class for all evaluators ---------- + +template +struct evaluator_base : public noncopyable +{ + // TODO that's not very nice to have to propagate all these traits. They are currently only needed to handle outer,inner indices. + typedef traits ExpressionTraits; + + enum { + Alignment = 0 + }; +}; + +// -------------------- Matrix and Array -------------------- +// +// evaluator is a common base class for the +// Matrix and Array evaluators. +// Here we directly specialize evaluator. This is not really a unary expression, and it is, by definition, dense, +// so no need for more sophisticated dispatching. + +template +struct evaluator > + : evaluator_base +{ + typedef PlainObjectBase PlainObjectType; + typedef typename PlainObjectType::Scalar Scalar; + typedef typename PlainObjectType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = PlainObjectType::IsRowMajor, + IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime, + RowsAtCompileTime = PlainObjectType::RowsAtCompileTime, + ColsAtCompileTime = PlainObjectType::ColsAtCompileTime, + + CoeffReadCost = NumTraits::ReadCost, + Flags = traits::EvaluatorFlags, + Alignment = traits::Alignment + }; + + EIGEN_DEVICE_FUNC evaluator() + : m_data(0), + m_outerStride(IsVectorAtCompileTime ? 0 + : int(IsRowMajor) ? ColsAtCompileTime + : RowsAtCompileTime) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC explicit evaluator(const PlainObjectType& m) + : m_data(m.data()), m_outerStride(IsVectorAtCompileTime ? 0 : m.outerStride()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + if (IsRowMajor) + return m_data[row * m_outerStride.value() + col]; + else + return m_data[row + col * m_outerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_data[index]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + if (IsRowMajor) + return const_cast(m_data)[row * m_outerStride.value() + col]; + else + return const_cast(m_data)[row + col * m_outerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return const_cast(m_data)[index]; + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + if (IsRowMajor) + return ploadt(m_data + row * m_outerStride.value() + col); + else + return ploadt(m_data + row + col * m_outerStride.value()); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return ploadt(m_data + index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + if (IsRowMajor) + return pstoret + (const_cast(m_data) + row * m_outerStride.value() + col, x); + else + return pstoret + (const_cast(m_data) + row + col * m_outerStride.value(), x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + return pstoret(const_cast(m_data) + index, x); + } + +protected: + const Scalar *m_data; + + // We do not need to know the outer stride for vectors + variable_if_dynamic m_outerStride; +}; + +template +struct evaluator > + : evaluator > > +{ + typedef Matrix XprType; + + EIGEN_DEVICE_FUNC evaluator() {} + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& m) + : evaluator >(m) + { } +}; + +template +struct evaluator > + : evaluator > > +{ + typedef Array XprType; + + EIGEN_DEVICE_FUNC evaluator() {} + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& m) + : evaluator >(m) + { } +}; + +// -------------------- Transpose -------------------- + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef Transpose XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + Flags = evaluator::Flags ^ RowMajorBit, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& t) : m_argImpl(t.nestedExpression()) {} + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(col, row); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(col, row); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + typename XprType::Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(col, row); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_argImpl.template packet(index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + m_argImpl.template writePacket(col, row, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + m_argImpl.template writePacket(index, x); + } + +protected: + evaluator m_argImpl; +}; + +// -------------------- CwiseNullaryOp -------------------- +// Like Matrix and Array, this is not really a unary expression, so we directly specialize evaluator. +// Likewise, there is not need to more sophisticated dispatching here. + +template::value, + bool has_unary = has_unary_operator::value, + bool has_binary = has_binary_operator::value> +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { return op(i,j); } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { return op(i); } + + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { return op.template packetOp(i,j); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { return op.template packetOp(i); } +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType=0, IndexType=0) const { return op(); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType=0, IndexType=0) const { return op.template packetOp(); } +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j=0) const { return op(i,j); } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j=0) const { return op.template packetOp(i,j); } +}; + +// We need the following specialization for vector-only functors assigned to a runtime vector, +// for instance, using linspace and assigning a RowVectorXd to a MatrixXd or even a row of a MatrixXd. +// In this case, i==0 and j is used for the actual iteration. +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { + eigen_assert(i==0 || j==0); + return op(i+j); + } + template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { + eigen_assert(i==0 || j==0); + return op.template packetOp(i+j); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { return op(i); } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { return op.template packetOp(i); } +}; + +template +struct nullary_wrapper {}; + +#if 0 && EIGEN_COMP_MSVC>0 +// Disable this ugly workaround. This is now handled in traits::match, +// but this piece of code might still become handly if some other weird compilation +// erros pop up again. + +// MSVC exhibits a weird compilation error when +// compiling: +// Eigen::MatrixXf A = MatrixXf::Random(3,3); +// Ref R = 2.f*A; +// and that has_*ary_operator> have not been instantiated yet. +// The "problem" is that evaluator<2.f*A> is instantiated by traits::match<2.f*A> +// and at that time has_*ary_operator returns true regardless of T. +// Then nullary_wrapper is badly instantiated as nullary_wrapper<.,.,true,true,true>. +// The trick is thus to defer the proper instantiation of nullary_wrapper when coeff(), +// and packet() are really instantiated as implemented below: + +// This is a simple wrapper around Index to enforce the re-instantiation of +// has_*ary_operator when needed. +template struct nullary_wrapper_workaround_msvc { + nullary_wrapper_workaround_msvc(const T&); + operator T()const; +}; + +template +struct nullary_wrapper +{ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i, IndexType j) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().operator()(op,i,j); + } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar operator()(const NullaryOp& op, IndexType i) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().operator()(op,i); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i, IndexType j) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().template packetOp(op,i,j); + } + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T packetOp(const NullaryOp& op, IndexType i) const { + return nullary_wrapper >::value, + has_unary_operator >::value, + has_binary_operator >::value>().template packetOp(op,i); + } +}; +#endif // MSVC workaround + +template +struct evaluator > + : evaluator_base > +{ + typedef CwiseNullaryOp XprType; + typedef typename internal::remove_all::type PlainObjectTypeCleaned; + + enum { + CoeffReadCost = internal::functor_traits::Cost, + + Flags = (evaluator::Flags + & ( HereditaryBits + | (functor_has_linear_access::ret ? LinearAccessBit : 0) + | (functor_traits::PacketAccess ? PacketAccessBit : 0))) + | (functor_traits::IsRepeatable ? 0 : EvalBeforeNestingBit), + Alignment = AlignedMax + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& n) + : m_functor(n.functor()), m_wrapper() + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(IndexType row, IndexType col) const + { + return m_wrapper(m_functor, row, col); + } + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(IndexType index) const + { + return m_wrapper(m_functor,index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(IndexType row, IndexType col) const + { + return m_wrapper.template packetOp(m_functor, row, col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(IndexType index) const + { + return m_wrapper.template packetOp(m_functor, index); + } + +protected: + const NullaryOp m_functor; + const internal::nullary_wrapper m_wrapper; +}; + +// -------------------- CwiseUnaryOp -------------------- + +template +struct unary_evaluator, IndexBased > + : evaluator_base > +{ + typedef CwiseUnaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, + + Flags = evaluator::Flags + & (HereditaryBits | LinearAccessBit | (functor_traits::PacketAccess ? PacketAccessBit : 0)), + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit unary_evaluator(const XprType& op) + : m_functor(op.functor()), + m_argImpl(op.nestedExpression()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_argImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_argImpl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_argImpl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_argImpl.template packet(index)); + } + +protected: + const UnaryOp m_functor; + evaluator m_argImpl; +}; + +// -------------------- CwiseTernaryOp -------------------- + +// this is a ternary expression +template +struct evaluator > + : public ternary_evaluator > +{ + typedef CwiseTernaryOp XprType; + typedef ternary_evaluator > Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +template +struct ternary_evaluator, IndexBased, IndexBased> + : evaluator_base > +{ + typedef CwiseTernaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, + + Arg1Flags = evaluator::Flags, + Arg2Flags = evaluator::Flags, + Arg3Flags = evaluator::Flags, + SameType = is_same::value && is_same::value, + StorageOrdersAgree = (int(Arg1Flags)&RowMajorBit)==(int(Arg2Flags)&RowMajorBit) && (int(Arg1Flags)&RowMajorBit)==(int(Arg3Flags)&RowMajorBit), + Flags0 = (int(Arg1Flags) | int(Arg2Flags) | int(Arg3Flags)) & ( + HereditaryBits + | (int(Arg1Flags) & int(Arg2Flags) & int(Arg3Flags) & + ( (StorageOrdersAgree ? LinearAccessBit : 0) + | (functor_traits::PacketAccess && StorageOrdersAgree && SameType ? PacketAccessBit : 0) + ) + ) + ), + Flags = (Flags0 & ~RowMajorBit) | (Arg1Flags & RowMajorBit), + Alignment = EIGEN_PLAIN_ENUM_MIN( + EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment), + evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit ternary_evaluator(const XprType& xpr) + : m_functor(xpr.functor()), + m_arg1Impl(xpr.arg1()), + m_arg2Impl(xpr.arg2()), + m_arg3Impl(xpr.arg3()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_arg1Impl.coeff(row, col), m_arg2Impl.coeff(row, col), m_arg3Impl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_arg1Impl.coeff(index), m_arg2Impl.coeff(index), m_arg3Impl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_arg1Impl.template packet(row, col), + m_arg2Impl.template packet(row, col), + m_arg3Impl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_arg1Impl.template packet(index), + m_arg2Impl.template packet(index), + m_arg3Impl.template packet(index)); + } + +protected: + const TernaryOp m_functor; + evaluator m_arg1Impl; + evaluator m_arg2Impl; + evaluator m_arg3Impl; +}; + +// -------------------- CwiseBinaryOp -------------------- + +// this is a binary expression +template +struct evaluator > + : public binary_evaluator > +{ + typedef CwiseBinaryOp XprType; + typedef binary_evaluator > Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +template +struct binary_evaluator, IndexBased, IndexBased> + : evaluator_base > +{ + typedef CwiseBinaryOp XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + evaluator::CoeffReadCost + functor_traits::Cost, + + LhsFlags = evaluator::Flags, + RhsFlags = evaluator::Flags, + SameType = is_same::value, + StorageOrdersAgree = (int(LhsFlags)&RowMajorBit)==(int(RhsFlags)&RowMajorBit), + Flags0 = (int(LhsFlags) | int(RhsFlags)) & ( + HereditaryBits + | (int(LhsFlags) & int(RhsFlags) & + ( (StorageOrdersAgree ? LinearAccessBit : 0) + | (functor_traits::PacketAccess && StorageOrdersAgree && SameType ? PacketAccessBit : 0) + ) + ) + ), + Flags = (Flags0 & ~RowMajorBit) | (LhsFlags & RowMajorBit), + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment,evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit binary_evaluator(const XprType& xpr) + : m_functor(xpr.functor()), + m_lhsImpl(xpr.lhs()), + m_rhsImpl(xpr.rhs()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_functor(m_lhsImpl.coeff(row, col), m_rhsImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_functor(m_lhsImpl.coeff(index), m_rhsImpl.coeff(index)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_functor.packetOp(m_lhsImpl.template packet(row, col), + m_rhsImpl.template packet(row, col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_functor.packetOp(m_lhsImpl.template packet(index), + m_rhsImpl.template packet(index)); + } + +protected: + const BinaryOp m_functor; + evaluator m_lhsImpl; + evaluator m_rhsImpl; +}; + +// -------------------- CwiseUnaryView -------------------- + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef CwiseUnaryView XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost + functor_traits::Cost, + + Flags = (evaluator::Flags & (HereditaryBits | LinearAccessBit | DirectAccessBit)), + + Alignment = 0 // FIXME it is not very clear why alignment is necessarily lost... + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& op) + : m_unaryOp(op.functor()), + m_argImpl(op.nestedExpression()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(functor_traits::Cost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_unaryOp(m_argImpl.coeff(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_unaryOp(m_argImpl.coeff(index)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_unaryOp(m_argImpl.coeffRef(row, col)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_unaryOp(m_argImpl.coeffRef(index)); + } + +protected: + const UnaryOp m_unaryOp; + evaluator m_argImpl; +}; + +// -------------------- Map -------------------- + +// FIXME perhaps the PlainObjectType could be provided by Derived::PlainObject ? +// but that might complicate template specialization +template +struct mapbase_evaluator; + +template +struct mapbase_evaluator : evaluator_base +{ + typedef Derived XprType; + typedef typename XprType::PointerType PointerType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = XprType::RowsAtCompileTime, + ColsAtCompileTime = XprType::ColsAtCompileTime, + CoeffReadCost = NumTraits::ReadCost + }; + + EIGEN_DEVICE_FUNC explicit mapbase_evaluator(const XprType& map) + : m_data(const_cast(map.data())), + m_innerStride(map.innerStride()), + m_outerStride(map.outerStride()) + { + EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(evaluator::Flags&PacketAccessBit, internal::inner_stride_at_compile_time::ret==1), + PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_data[index * m_innerStride.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_data[index * m_innerStride.value()]; + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + PointerType ptr = m_data + row * rowStride() + col * colStride(); + return internal::ploadt(ptr); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return internal::ploadt(m_data + index * m_innerStride.value()); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + PointerType ptr = m_data + row * rowStride() + col * colStride(); + return internal::pstoret(ptr, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + internal::pstoret(m_data + index * m_innerStride.value(), x); + } +protected: + EIGEN_DEVICE_FUNC + inline Index rowStride() const { return XprType::IsRowMajor ? m_outerStride.value() : m_innerStride.value(); } + EIGEN_DEVICE_FUNC + inline Index colStride() const { return XprType::IsRowMajor ? m_innerStride.value() : m_outerStride.value(); } + + PointerType m_data; + const internal::variable_if_dynamic m_innerStride; + const internal::variable_if_dynamic m_outerStride; +}; + +template +struct evaluator > + : public mapbase_evaluator, PlainObjectType> +{ + typedef Map XprType; + typedef typename XprType::Scalar Scalar; + // TODO: should check for smaller packet types once we can handle multi-sized packet types + typedef typename packet_traits::type PacketScalar; + + enum { + InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0 + ? int(PlainObjectType::InnerStrideAtCompileTime) + : int(StrideType::InnerStrideAtCompileTime), + OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0 + ? int(PlainObjectType::OuterStrideAtCompileTime) + : int(StrideType::OuterStrideAtCompileTime), + HasNoInnerStride = InnerStrideAtCompileTime == 1, + HasNoOuterStride = StrideType::OuterStrideAtCompileTime == 0, + HasNoStride = HasNoInnerStride && HasNoOuterStride, + IsDynamicSize = PlainObjectType::SizeAtCompileTime==Dynamic, + + PacketAccessMask = bool(HasNoInnerStride) ? ~int(0) : ~int(PacketAccessBit), + LinearAccessMask = bool(HasNoStride) || bool(PlainObjectType::IsVectorAtCompileTime) ? ~int(0) : ~int(LinearAccessBit), + Flags = int( evaluator::Flags) & (LinearAccessMask&PacketAccessMask), + + Alignment = int(MapOptions)&int(AlignedMask) + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& map) + : mapbase_evaluator(map) + { } +}; + +// -------------------- Ref -------------------- + +template +struct evaluator > + : public mapbase_evaluator, PlainObjectType> +{ + typedef Ref XprType; + + enum { + Flags = evaluator >::Flags, + Alignment = evaluator >::Alignment + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& ref) + : mapbase_evaluator(ref) + { } +}; + +// -------------------- Block -------------------- + +template::ret> struct block_evaluator; + +template +struct evaluator > + : block_evaluator +{ + typedef Block XprType; + typedef typename XprType::Scalar Scalar; + // TODO: should check for smaller packet types once we can handle multi-sized packet types + typedef typename packet_traits::type PacketScalar; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime, + + ArgTypeIsRowMajor = (int(evaluator::Flags)&RowMajorBit) != 0, + IsRowMajor = (MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1) ? 0 + : ArgTypeIsRowMajor, + HasSameStorageOrderAsArgType = (IsRowMajor == ArgTypeIsRowMajor), + InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + InnerStrideAtCompileTime = HasSameStorageOrderAsArgType + ? int(inner_stride_at_compile_time::ret) + : int(outer_stride_at_compile_time::ret), + OuterStrideAtCompileTime = HasSameStorageOrderAsArgType + ? int(outer_stride_at_compile_time::ret) + : int(inner_stride_at_compile_time::ret), + MaskPacketAccessBit = (InnerStrideAtCompileTime == 1 || HasSameStorageOrderAsArgType) ? PacketAccessBit : 0, + + FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1 || (InnerPanel && (evaluator::Flags&LinearAccessBit))) ? LinearAccessBit : 0, + FlagsRowMajorBit = XprType::Flags&RowMajorBit, + Flags0 = evaluator::Flags & ( (HereditaryBits & ~RowMajorBit) | + DirectAccessBit | + MaskPacketAccessBit), + Flags = Flags0 | FlagsLinearAccessBit | FlagsRowMajorBit, + + PacketAlignment = unpacket_traits::alignment, + Alignment0 = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) + && (OuterStrideAtCompileTime!=0) + && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % int(PacketAlignment)) == 0)) ? int(PacketAlignment) : 0, + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, Alignment0) + }; + typedef block_evaluator block_evaluator_type; + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& block) : block_evaluator_type(block) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } +}; + +// no direct-access => dispatch to a unary evaluator +template +struct block_evaluator + : unary_evaluator > +{ + typedef Block XprType; + + EIGEN_DEVICE_FUNC explicit block_evaluator(const XprType& block) + : unary_evaluator(block) + {} +}; + +template +struct unary_evaluator, IndexBased> + : evaluator_base > +{ + typedef Block XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& block) + : m_argImpl(block.nestedExpression()), + m_startRow(block.startRow()), + m_startCol(block.startCol()), + m_linear_offset(InnerPanel?(XprType::IsRowMajor ? block.startRow()*block.cols() : block.startCol()*block.rows()):0) + { } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + RowsAtCompileTime = XprType::RowsAtCompileTime, + ForwardLinearAccess = InnerPanel && bool(evaluator::Flags&LinearAccessBit) + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(m_startRow.value() + row, m_startCol.value() + col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + if (ForwardLinearAccess) + return m_argImpl.coeff(m_linear_offset.value() + index); + else + return coeff(RowsAtCompileTime == 1 ? 0 : index, RowsAtCompileTime == 1 ? index : 0); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(m_startRow.value() + row, m_startCol.value() + col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + if (ForwardLinearAccess) + return m_argImpl.coeffRef(m_linear_offset.value() + index); + else + return coeffRef(RowsAtCompileTime == 1 ? 0 : index, RowsAtCompileTime == 1 ? index : 0); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(m_startRow.value() + row, m_startCol.value() + col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + if (ForwardLinearAccess) + return m_argImpl.template packet(m_linear_offset.value() + index); + else + return packet(RowsAtCompileTime == 1 ? 0 : index, + RowsAtCompileTime == 1 ? index : 0); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + return m_argImpl.template writePacket(m_startRow.value() + row, m_startCol.value() + col, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + if (ForwardLinearAccess) + return m_argImpl.template writePacket(m_linear_offset.value() + index, x); + else + return writePacket(RowsAtCompileTime == 1 ? 0 : index, + RowsAtCompileTime == 1 ? index : 0, + x); + } + +protected: + evaluator m_argImpl; + const variable_if_dynamic m_startRow; + const variable_if_dynamic m_startCol; + const variable_if_dynamic m_linear_offset; +}; + +// TODO: This evaluator does not actually use the child evaluator; +// all action is via the data() as returned by the Block expression. + +template +struct block_evaluator + : mapbase_evaluator, + typename Block::PlainObject> +{ + typedef Block XprType; + typedef typename XprType::Scalar Scalar; + + EIGEN_DEVICE_FUNC explicit block_evaluator(const XprType& block) + : mapbase_evaluator(block) + { + // TODO: for the 3.3 release, this should be turned to an internal assertion, but let's keep it as is for the beta lifetime + eigen_assert(((internal::UIntPtr(block.data()) % EIGEN_PLAIN_ENUM_MAX(1,evaluator::Alignment)) == 0) && "data is not aligned"); + } +}; + + +// -------------------- Select -------------------- +// NOTE shall we introduce a ternary_evaluator? + +// TODO enable vectorization for Select +template +struct evaluator > + : evaluator_base > +{ + typedef Select XprType; + enum { + CoeffReadCost = evaluator::CoeffReadCost + + EIGEN_PLAIN_ENUM_MAX(evaluator::CoeffReadCost, + evaluator::CoeffReadCost), + + Flags = (unsigned int)evaluator::Flags & evaluator::Flags & HereditaryBits, + + Alignment = EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment) + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& select) + : m_conditionImpl(select.conditionMatrix()), + m_thenImpl(select.thenMatrix()), + m_elseImpl(select.elseMatrix()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + if (m_conditionImpl.coeff(row, col)) + return m_thenImpl.coeff(row, col); + else + return m_elseImpl.coeff(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + if (m_conditionImpl.coeff(index)) + return m_thenImpl.coeff(index); + else + return m_elseImpl.coeff(index); + } + +protected: + evaluator m_conditionImpl; + evaluator m_thenImpl; + evaluator m_elseImpl; +}; + + +// -------------------- Replicate -------------------- + +template +struct unary_evaluator > + : evaluator_base > +{ + typedef Replicate XprType; + typedef typename XprType::CoeffReturnType CoeffReturnType; + enum { + Factor = (RowFactor==Dynamic || ColFactor==Dynamic) ? Dynamic : RowFactor*ColFactor + }; + typedef typename internal::nested_eval::type ArgTypeNested; + typedef typename internal::remove_all::type ArgTypeNestedCleaned; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + LinearAccessMask = XprType::IsVectorAtCompileTime ? LinearAccessBit : 0, + Flags = (evaluator::Flags & (HereditaryBits|LinearAccessMask) & ~RowMajorBit) | (traits::Flags & RowMajorBit), + + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& replicate) + : m_arg(replicate.nestedExpression()), + m_argImpl(m_arg), + m_rows(replicate.nestedExpression().rows()), + m_cols(replicate.nestedExpression().cols()) + {} + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + // try to avoid using modulo; this is a pure optimization strategy + const Index actual_row = internal::traits::RowsAtCompileTime==1 ? 0 + : RowFactor==1 ? row + : row % m_rows.value(); + const Index actual_col = internal::traits::ColsAtCompileTime==1 ? 0 + : ColFactor==1 ? col + : col % m_cols.value(); + + return m_argImpl.coeff(actual_row, actual_col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + // try to avoid using modulo; this is a pure optimization strategy + const Index actual_index = internal::traits::RowsAtCompileTime==1 + ? (ColFactor==1 ? index : index%m_cols.value()) + : (RowFactor==1 ? index : index%m_rows.value()); + + return m_argImpl.coeff(actual_index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + const Index actual_row = internal::traits::RowsAtCompileTime==1 ? 0 + : RowFactor==1 ? row + : row % m_rows.value(); + const Index actual_col = internal::traits::ColsAtCompileTime==1 ? 0 + : ColFactor==1 ? col + : col % m_cols.value(); + + return m_argImpl.template packet(actual_row, actual_col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + const Index actual_index = internal::traits::RowsAtCompileTime==1 + ? (ColFactor==1 ? index : index%m_cols.value()) + : (RowFactor==1 ? index : index%m_rows.value()); + + return m_argImpl.template packet(actual_index); + } + +protected: + const ArgTypeNested m_arg; + evaluator m_argImpl; + const variable_if_dynamic m_rows; + const variable_if_dynamic m_cols; +}; + + +// -------------------- PartialReduxExpr -------------------- + +template< typename ArgType, typename MemberOp, int Direction> +struct evaluator > + : evaluator_base > +{ + typedef PartialReduxExpr XprType; + typedef typename internal::nested_eval::type ArgTypeNested; + typedef typename internal::remove_all::type ArgTypeNestedCleaned; + typedef typename ArgType::Scalar InputScalar; + typedef typename XprType::Scalar Scalar; + enum { + TraversalSize = Direction==int(Vertical) ? int(ArgType::RowsAtCompileTime) : int(ArgType::ColsAtCompileTime) + }; + typedef typename MemberOp::template Cost CostOpType; + enum { + CoeffReadCost = TraversalSize==Dynamic ? HugeCost + : TraversalSize * evaluator::CoeffReadCost + int(CostOpType::value), + + Flags = (traits::Flags&RowMajorBit) | (evaluator::Flags&(HereditaryBits&(~RowMajorBit))) | LinearAccessBit, + + Alignment = 0 // FIXME this will need to be improved once PartialReduxExpr is vectorized + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType xpr) + : m_arg(xpr.nestedExpression()), m_functor(xpr.functor()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(TraversalSize==Dynamic ? HugeCost : int(CostOpType::value)); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const Scalar coeff(Index i, Index j) const + { + if (Direction==Vertical) + return m_functor(m_arg.col(j)); + else + return m_functor(m_arg.row(i)); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const Scalar coeff(Index index) const + { + if (Direction==Vertical) + return m_functor(m_arg.col(index)); + else + return m_functor(m_arg.row(index)); + } + +protected: + typename internal::add_const_on_value_type::type m_arg; + const MemberOp m_functor; +}; + + +// -------------------- MatrixWrapper and ArrayWrapper -------------------- +// +// evaluator_wrapper_base is a common base class for the +// MatrixWrapper and ArrayWrapper evaluators. + +template +struct evaluator_wrapper_base + : evaluator_base +{ + typedef typename remove_all::type ArgType; + enum { + CoeffReadCost = evaluator::CoeffReadCost, + Flags = evaluator::Flags, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC explicit evaluator_wrapper_base(const ArgType& arg) : m_argImpl(arg) {} + + typedef typename ArgType::Scalar Scalar; + typedef typename ArgType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(row, col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + return m_argImpl.template packet(row, col); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + return m_argImpl.template packet(index); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + m_argImpl.template writePacket(row, col, x); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + m_argImpl.template writePacket(index, x); + } + +protected: + evaluator m_argImpl; +}; + +template +struct unary_evaluator > + : evaluator_wrapper_base > +{ + typedef MatrixWrapper XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& wrapper) + : evaluator_wrapper_base >(wrapper.nestedExpression()) + { } +}; + +template +struct unary_evaluator > + : evaluator_wrapper_base > +{ + typedef ArrayWrapper XprType; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& wrapper) + : evaluator_wrapper_base >(wrapper.nestedExpression()) + { } +}; + + +// -------------------- Reverse -------------------- + +// defined in Reverse.h: +template struct reverse_packet_cond; + +template +struct unary_evaluator > + : evaluator_base > +{ + typedef Reverse XprType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + enum { + IsRowMajor = XprType::IsRowMajor, + IsColMajor = !IsRowMajor, + ReverseRow = (Direction == Vertical) || (Direction == BothDirections), + ReverseCol = (Direction == Horizontal) || (Direction == BothDirections), + ReversePacket = (Direction == BothDirections) + || ((Direction == Vertical) && IsColMajor) + || ((Direction == Horizontal) && IsRowMajor), + + CoeffReadCost = evaluator::CoeffReadCost, + + // let's enable LinearAccess only with vectorization because of the product overhead + // FIXME enable DirectAccess with negative strides? + Flags0 = evaluator::Flags, + LinearAccess = ( (Direction==BothDirections) && (int(Flags0)&PacketAccessBit) ) + || ((ReverseRow && XprType::ColsAtCompileTime==1) || (ReverseCol && XprType::RowsAtCompileTime==1)) + ? LinearAccessBit : 0, + + Flags = int(Flags0) & (HereditaryBits | PacketAccessBit | LinearAccess), + + Alignment = 0 // FIXME in some rare cases, Alignment could be preserved, like a Vector4f. + }; + + EIGEN_DEVICE_FUNC explicit unary_evaluator(const XprType& reverse) + : m_argImpl(reverse.nestedExpression()), + m_rows(ReverseRow ? reverse.nestedExpression().rows() : 1), + m_cols(ReverseCol ? reverse.nestedExpression().cols() : 1) + { } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index col) const + { + return m_argImpl.coeff(ReverseRow ? m_rows.value() - row - 1 : row, + ReverseCol ? m_cols.value() - col - 1 : col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(m_rows.value() * m_cols.value() - index - 1); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index col) + { + return m_argImpl.coeffRef(ReverseRow ? m_rows.value() - row - 1 : row, + ReverseCol ? m_cols.value() - col - 1 : col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(m_rows.value() * m_cols.value() - index - 1); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index row, Index col) const + { + enum { + PacketSize = unpacket_traits::size, + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1 + }; + typedef internal::reverse_packet_cond reverse_packet; + return reverse_packet::run(m_argImpl.template packet( + ReverseRow ? m_rows.value() - row - OffsetRow : row, + ReverseCol ? m_cols.value() - col - OffsetCol : col)); + } + + template + EIGEN_STRONG_INLINE + PacketType packet(Index index) const + { + enum { PacketSize = unpacket_traits::size }; + return preverse(m_argImpl.template packet(m_rows.value() * m_cols.value() - index - PacketSize)); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index row, Index col, const PacketType& x) + { + // FIXME we could factorize some code with packet(i,j) + enum { + PacketSize = unpacket_traits::size, + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1 + }; + typedef internal::reverse_packet_cond reverse_packet; + m_argImpl.template writePacket( + ReverseRow ? m_rows.value() - row - OffsetRow : row, + ReverseCol ? m_cols.value() - col - OffsetCol : col, + reverse_packet::run(x)); + } + + template + EIGEN_STRONG_INLINE + void writePacket(Index index, const PacketType& x) + { + enum { PacketSize = unpacket_traits::size }; + m_argImpl.template writePacket + (m_rows.value() * m_cols.value() - index - PacketSize, preverse(x)); + } + +protected: + evaluator m_argImpl; + + // If we do not reverse rows, then we do not need to know the number of rows; same for columns + // Nonetheless, in this case it is important to set to 1 such that the coeff(index) method works fine for vectors. + const variable_if_dynamic m_rows; + const variable_if_dynamic m_cols; +}; + + +// -------------------- Diagonal -------------------- + +template +struct evaluator > + : evaluator_base > +{ + typedef Diagonal XprType; + + enum { + CoeffReadCost = evaluator::CoeffReadCost, + + Flags = (unsigned int)(evaluator::Flags & (HereditaryBits | DirectAccessBit) & ~RowMajorBit) | LinearAccessBit, + + Alignment = 0 + }; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& diagonal) + : m_argImpl(diagonal.nestedExpression()), + m_index(diagonal.index()) + { } + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index row, Index) const + { + return m_argImpl.coeff(row + rowOffset(), row + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + CoeffReturnType coeff(Index index) const + { + return m_argImpl.coeff(index + rowOffset(), index + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index row, Index) + { + return m_argImpl.coeffRef(row + rowOffset(), row + colOffset()); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Scalar& coeffRef(Index index) + { + return m_argImpl.coeffRef(index + rowOffset(), index + colOffset()); + } + +protected: + evaluator m_argImpl; + const internal::variable_if_dynamicindex m_index; + +private: + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index rowOffset() const { return m_index.value() > 0 ? 0 : -m_index.value(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index colOffset() const { return m_index.value() > 0 ? m_index.value() : 0; } +}; + + +//---------------------------------------------------------------------- +// deprecated code +//---------------------------------------------------------------------- + +// -------------------- EvalToTemp -------------------- + +// expression class for evaluating nested expression to a temporary + +template class EvalToTemp; + +template +struct traits > + : public traits +{ }; + +template +class EvalToTemp + : public dense_xpr_base >::type +{ + public: + + typedef typename dense_xpr_base::type Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(EvalToTemp) + + explicit EvalToTemp(const ArgType& arg) + : m_arg(arg) + { } + + const ArgType& arg() const + { + return m_arg; + } + + Index rows() const + { + return m_arg.rows(); + } + + Index cols() const + { + return m_arg.cols(); + } + + private: + const ArgType& m_arg; +}; + +template +struct evaluator > + : public evaluator +{ + typedef EvalToTemp XprType; + typedef typename ArgType::PlainObject PlainObject; + typedef evaluator Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) + : m_result(xpr.arg()) + { + ::new (static_cast(this)) Base(m_result); + } + + // This constructor is used when nesting an EvalTo evaluator in another evaluator + EIGEN_DEVICE_FUNC evaluator(const ArgType& arg) + : m_result(arg) + { + ::new (static_cast(this)) Base(m_result); + } + +protected: + PlainObject m_result; +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_COREEVALUATORS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h new file mode 100644 index 00000000000..4eb42b93af1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CoreIterators.h @@ -0,0 +1,127 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_COREITERATORS_H +#define EIGEN_COREITERATORS_H + +namespace Eigen { + +/* This file contains the respective InnerIterator definition of the expressions defined in Eigen/Core + */ + +namespace internal { + +template +class inner_iterator_selector; + +} + +/** \class InnerIterator + * \brief An InnerIterator allows to loop over the element of any matrix expression. + * + * \warning To be used with care because an evaluator is constructed every time an InnerIterator iterator is constructed. + * + * TODO: add a usage example + */ +template +class InnerIterator +{ +protected: + typedef internal::inner_iterator_selector::Kind> IteratorType; + typedef internal::evaluator EvaluatorType; + typedef typename internal::traits::Scalar Scalar; +public: + /** Construct an iterator over the \a outerId -th row or column of \a xpr */ + InnerIterator(const XprType &xpr, const Index &outerId) + : m_eval(xpr), m_iter(m_eval, outerId, xpr.innerSize()) + {} + + /// \returns the value of the current coefficient. + EIGEN_STRONG_INLINE Scalar value() const { return m_iter.value(); } + /** Increment the iterator \c *this to the next non-zero coefficient. + * Explicit zeros are not skipped over. To skip explicit zeros, see class SparseView + */ + EIGEN_STRONG_INLINE InnerIterator& operator++() { m_iter.operator++(); return *this; } + /// \returns the column or row index of the current coefficient. + EIGEN_STRONG_INLINE Index index() const { return m_iter.index(); } + /// \returns the row index of the current coefficient. + EIGEN_STRONG_INLINE Index row() const { return m_iter.row(); } + /// \returns the column index of the current coefficient. + EIGEN_STRONG_INLINE Index col() const { return m_iter.col(); } + /// \returns \c true if the iterator \c *this still references a valid coefficient. + EIGEN_STRONG_INLINE operator bool() const { return m_iter; } + +protected: + EvaluatorType m_eval; + IteratorType m_iter; +private: + // If you get here, then you're not using the right InnerIterator type, e.g.: + // SparseMatrix A; + // SparseMatrix::InnerIterator it(A,0); + template InnerIterator(const EigenBase&,Index outer); +}; + +namespace internal { + +// Generic inner iterator implementation for dense objects +template +class inner_iterator_selector +{ +protected: + typedef evaluator EvaluatorType; + typedef typename traits::Scalar Scalar; + enum { IsRowMajor = (XprType::Flags&RowMajorBit)==RowMajorBit }; + +public: + EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &innerSize) + : m_eval(eval), m_inner(0), m_outer(outerId), m_end(innerSize) + {} + + EIGEN_STRONG_INLINE Scalar value() const + { + return (IsRowMajor) ? m_eval.coeff(m_outer, m_inner) + : m_eval.coeff(m_inner, m_outer); + } + + EIGEN_STRONG_INLINE inner_iterator_selector& operator++() { m_inner++; return *this; } + + EIGEN_STRONG_INLINE Index index() const { return m_inner; } + inline Index row() const { return IsRowMajor ? m_outer : index(); } + inline Index col() const { return IsRowMajor ? index() : m_outer; } + + EIGEN_STRONG_INLINE operator bool() const { return m_inner < m_end && m_inner>=0; } + +protected: + const EvaluatorType& m_eval; + Index m_inner; + const Index m_outer; + const Index m_end; +}; + +// For iterator-based evaluator, inner-iterator is already implemented as +// evaluator<>::InnerIterator +template +class inner_iterator_selector + : public evaluator::InnerIterator +{ +protected: + typedef typename evaluator::InnerIterator Base; + typedef evaluator EvaluatorType; + +public: + EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &/*innerSize*/) + : Base(eval, outerId) + {} +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_COREITERATORS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h new file mode 100644 index 00000000000..a36765e396b --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseBinaryOp.h @@ -0,0 +1,184 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_BINARY_OP_H +#define EIGEN_CWISE_BINARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > +{ + // we must not inherit from traits since it has + // the potential to cause problems with MSVC + typedef typename remove_all::type Ancestor; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime + }; + + // even though we require Lhs and Rhs to have the same scalar type (see CwiseBinaryOp constructor), + // we still want to handle the case when the result type is different. + typedef typename result_of< + BinaryOp( + const typename Lhs::Scalar&, + const typename Rhs::Scalar& + ) + >::type Scalar; + typedef typename cwise_promote_storage_type::StorageKind, + typename traits::StorageKind, + BinaryOp>::ret StorageKind; + typedef typename promote_index_type::StorageIndex, + typename traits::StorageIndex>::type StorageIndex; + typedef typename Lhs::Nested LhsNested; + typedef typename Rhs::Nested RhsNested; + typedef typename remove_reference::type _LhsNested; + typedef typename remove_reference::type _RhsNested; + enum { + Flags = cwise_promote_storage_order::StorageKind,typename traits::StorageKind,_LhsNested::Flags & RowMajorBit,_RhsNested::Flags & RowMajorBit>::value + }; +}; +} // end namespace internal + +template +class CwiseBinaryOpImpl; + +/** \class CwiseBinaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise binary operator is applied to two expressions + * + * \tparam BinaryOp template functor implementing the operator + * \tparam LhsType the type of the left-hand side + * \tparam RhsType the type of the right-hand side + * + * This class represents an expression where a coefficient-wise binary operator is applied to two expressions. + * It is the return type of binary operators, by which we mean only those binary operators where + * both the left-hand side and the right-hand side are Eigen expressions. + * For example, the return type of matrix1+matrix2 is a CwiseBinaryOp. + * + * Most of the time, this is the only way that it is used, so you typically don't have to name + * CwiseBinaryOp types explicitly. + * + * \sa MatrixBase::binaryExpr(const MatrixBase &,const CustomBinaryOp &) const, class CwiseUnaryOp, class CwiseNullaryOp + */ +template +class CwiseBinaryOp : + public CwiseBinaryOpImpl< + BinaryOp, LhsType, RhsType, + typename internal::cwise_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + BinaryOp>::ret>, + internal::no_assignment_operator +{ + public: + + typedef typename internal::remove_all::type Functor; + typedef typename internal::remove_all::type Lhs; + typedef typename internal::remove_all::type Rhs; + + typedef typename CwiseBinaryOpImpl< + BinaryOp, LhsType, RhsType, + typename internal::cwise_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + BinaryOp>::ret>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseBinaryOp) + + typedef typename internal::ref_selector::type LhsNested; + typedef typename internal::ref_selector::type RhsNested; + typedef typename internal::remove_reference::type _LhsNested; + typedef typename internal::remove_reference::type _RhsNested; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CwiseBinaryOp(const Lhs& aLhs, const Rhs& aRhs, const BinaryOp& func = BinaryOp()) + : m_lhs(aLhs), m_rhs(aRhs), m_functor(func) + { + EIGEN_CHECK_BINARY_COMPATIBILIY(BinaryOp,typename Lhs::Scalar,typename Rhs::Scalar); + // require the sizes to match + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Lhs, Rhs) + eigen_assert(aLhs.rows() == aRhs.rows() && aLhs.cols() == aRhs.cols()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { + // return the fixed size type if available to enable compile time optimizations + if (internal::traits::type>::RowsAtCompileTime==Dynamic) + return m_rhs.rows(); + else + return m_lhs.rows(); + } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { + // return the fixed size type if available to enable compile time optimizations + if (internal::traits::type>::ColsAtCompileTime==Dynamic) + return m_rhs.cols(); + else + return m_lhs.cols(); + } + + /** \returns the left hand side nested expression */ + EIGEN_DEVICE_FUNC + const _LhsNested& lhs() const { return m_lhs; } + /** \returns the right hand side nested expression */ + EIGEN_DEVICE_FUNC + const _RhsNested& rhs() const { return m_rhs; } + /** \returns the functor representing the binary operation */ + EIGEN_DEVICE_FUNC + const BinaryOp& functor() const { return m_functor; } + + protected: + LhsNested m_lhs; + RhsNested m_rhs; + const BinaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseBinaryOpImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +/** replaces \c *this by \c *this - \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_STRONG_INLINE Derived & +MatrixBase::operator-=(const MatrixBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +/** replaces \c *this by \c *this + \a other. + * + * \returns a reference to \c *this + */ +template +template +EIGEN_STRONG_INLINE Derived & +MatrixBase::operator+=(const MatrixBase& other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_CWISE_BINARY_OP_H + diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h new file mode 100644 index 00000000000..ddd607e3832 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseNullaryOp.h @@ -0,0 +1,866 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_NULLARY_OP_H +#define EIGEN_CWISE_NULLARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > : traits +{ + enum { + Flags = traits::Flags & RowMajorBit + }; +}; + +} // namespace internal + +/** \class CwiseNullaryOp + * \ingroup Core_Module + * + * \brief Generic expression of a matrix where all coefficients are defined by a functor + * + * \tparam NullaryOp template functor implementing the operator + * \tparam PlainObjectType the underlying plain matrix/array type + * + * This class represents an expression of a generic nullary operator. + * It is the return type of the Ones(), Zero(), Constant(), Identity() and Random() methods, + * and most of the time this is the only way it is used. + * + * However, if you want to write a function returning such an expression, you + * will need to use this class. + * + * The functor NullaryOp must expose one of the following method: + + + + +
\c operator()() if the procedural generation does not depend on the coefficient entries (e.g., random numbers)
\c operator()(Index i)if the procedural generation makes sense for vectors only and that it depends on the coefficient index \c i (e.g., linspace)
\c operator()(Index i,Index j)if the procedural generation depends on the matrix coordinates \c i, \c j (e.g., to generate a checkerboard with 0 and 1)
+ * It is also possible to expose the last two operators if the generation makes sense for matrices but can be optimized for vectors. + * + * See DenseBase::NullaryExpr(Index,const CustomNullaryOp&) for an example binding + * C++11 random number generators. + * + * A nullary expression can also be used to implement custom sophisticated matrix manipulations + * that cannot be covered by the existing set of natively supported matrix manipulations. + * See this \ref TopicCustomizing_NullaryExpr "page" for some examples and additional explanations + * on the behavior of CwiseNullaryOp. + * + * \sa class CwiseUnaryOp, class CwiseBinaryOp, DenseBase::NullaryExpr + */ +template +class CwiseNullaryOp : public internal::dense_xpr_base< CwiseNullaryOp >::type, internal::no_assignment_operator +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(CwiseNullaryOp) + + EIGEN_DEVICE_FUNC + CwiseNullaryOp(Index rows, Index cols, const NullaryOp& func = NullaryOp()) + : m_rows(rows), m_cols(cols), m_functor(func) + { + eigen_assert(rows >= 0 + && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows) + && cols >= 0 + && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols)); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { return m_rows.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { return m_cols.value(); } + + /** \returns the functor representing the nullary operation */ + EIGEN_DEVICE_FUNC + const NullaryOp& functor() const { return m_functor; } + + protected: + const internal::variable_if_dynamic m_rows; + const internal::variable_if_dynamic m_cols; + const NullaryOp m_functor; +}; + + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func) +{ + return CwiseNullaryOp(rows, cols, func); +} + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * Here is an example with C++11 random generators: \include random_cpp11.cpp + * Output: \verbinclude random_cpp11.out + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(Index size, const CustomNullaryOp& func) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + if(RowsAtCompileTime == 1) return CwiseNullaryOp(1, size, func); + else return CwiseNullaryOp(size, 1, func); +} + +/** \returns an expression of a matrix defined by a custom functor \a func + * + * This variant is only for fixed-size DenseBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp::PlainObject> +DenseBase::NullaryExpr(const CustomNullaryOp& func) +{ + return CwiseNullaryOp(RowsAtCompileTime, ColsAtCompileTime, func); +} + +/** \returns an expression of a constant matrix of value \a value + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this DenseBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(Index rows, Index cols, const Scalar& value) +{ + return DenseBase::NullaryExpr(rows, cols, internal::scalar_constant_op(value)); +} + +/** \returns an expression of a constant matrix of value \a value + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this DenseBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(Index size, const Scalar& value) +{ + return DenseBase::NullaryExpr(size, internal::scalar_constant_op(value)); +} + +/** \returns an expression of a constant matrix of value \a value + * + * This variant is only for fixed-size DenseBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * The template parameter \a CustomNullaryOp is the type of the functor. + * + * \sa class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Constant(const Scalar& value) +{ + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_constant_op(value)); +} + +/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(Index,const Scalar&,const Scalar&) + * + * \sa LinSpaced(Index,Scalar,Scalar), setLinSpaced(Index,const Scalar&,const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return DenseBase::NullaryExpr(size, internal::linspaced_op(low,high,size)); +} + +/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(const Scalar&,const Scalar&) + * + * \sa LinSpaced(Scalar,Scalar) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Sequential_t, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op(low,high,Derived::SizeAtCompileTime)); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function generates 'size' equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * Example: \include DenseBase_LinSpaced.cpp + * Output: \verbinclude DenseBase_LinSpaced.out + * + * For integer scalar types, an even spacing is possible if and only if the length of the range, + * i.e., \c high-low is a scalar multiple of \c size-1, or if \c size is a scalar multiple of the + * number of values \c high-low+1 (meaning each value can be repeated the same number of time). + * If one of these two considions is not satisfied, then \c high is lowered to the largest value + * satisfying one of this constraint. + * Here are some examples: + * + * Example: \include DenseBase_LinSpacedInt.cpp + * Output: \verbinclude DenseBase_LinSpacedInt.out + * + * \sa setLinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(Index size, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return DenseBase::NullaryExpr(size, internal::linspaced_op(low,high,size)); +} + +/** + * \copydoc DenseBase::LinSpaced(Index, const Scalar&, const Scalar&) + * Special version for fixed size types which does not require the size parameter. + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::RandomAccessLinSpacedReturnType +DenseBase::LinSpaced(const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return DenseBase::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op(low,high,Derived::SizeAtCompileTime)); +} + +/** \returns true if all coefficients in this matrix are approximately equal to \a val, to within precision \a prec */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isApproxToConstant +(const Scalar& val, const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if(!internal::isApprox(self.coeff(i, j), val, prec)) + return false; + return true; +} + +/** This is just an alias for isApproxToConstant(). + * + * \returns true if all coefficients in this matrix are approximately equal to \a value, to within precision \a prec */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isConstant +(const Scalar& val, const RealScalar& prec) const +{ + return isApproxToConstant(val, prec); +} + +/** Alias for setConstant(): sets all coefficients in this expression to \a val. + * + * \sa setConstant(), Constant(), class CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void DenseBase::fill(const Scalar& val) +{ + setConstant(val); +} + +/** Sets all coefficients in this expression to value \a val. + * + * \sa fill(), setConstant(Index,const Scalar&), setConstant(Index,Index,const Scalar&), setZero(), setOnes(), Constant(), class CwiseNullaryOp, setZero(), setOnes() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setConstant(const Scalar& val) +{ + return derived() = Constant(rows(), cols(), val); +} + +/** Resizes to the given \a size, and sets all coefficients in this expression to the given value \a val. + * + * \only_for_vectors + * + * Example: \include Matrix_setConstant_int.cpp + * Output: \verbinclude Matrix_setConstant_int.out + * + * \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setConstant(Index size, const Scalar& val) +{ + resize(size); + return setConstant(val); +} + +/** Resizes to the given size, and sets all coefficients in this expression to the given value \a val. + * + * \param rows the new number of rows + * \param cols the new number of columns + * \param val the value to which all coefficients are set + * + * Example: \include Matrix_setConstant_int_int.cpp + * Output: \verbinclude Matrix_setConstant_int_int.out + * + * \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setConstant(Index rows, Index cols, const Scalar& val) +{ + resize(rows, cols); + return setConstant(val); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function generates 'size' equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * Example: \include DenseBase_setLinSpaced.cpp + * Output: \verbinclude DenseBase_setLinSpaced.out + * + * For integer scalar types, do not miss the explanations on the definition + * of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink. + * + * \sa LinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setLinSpaced(Index newSize, const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return derived() = Derived::NullaryExpr(newSize, internal::linspaced_op(low,high,newSize)); +} + +/** + * \brief Sets a linearly spaced vector. + * + * The function fills \c *this with equally spaced values in the closed interval [low,high]. + * When size is set to 1, a vector of length 1 containing 'high' is returned. + * + * \only_for_vectors + * + * For integer scalar types, do not miss the explanations on the definition + * of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink. + * + * \sa LinSpaced(Index,const Scalar&,const Scalar&), setLinSpaced(Index, const Scalar&, const Scalar&), CwiseNullaryOp + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setLinSpaced(const Scalar& low, const Scalar& high) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return setLinSpaced(size(), low, high); +} + +// zero: + +/** \returns an expression of a zero matrix. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used + * instead. + * + * Example: \include MatrixBase_zero_int_int.cpp + * Output: \verbinclude MatrixBase_zero_int_int.out + * + * \sa Zero(), Zero(Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero(Index rows, Index cols) +{ + return Constant(rows, cols, Scalar(0)); +} + +/** \returns an expression of a zero vector. + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Zero() should be used + * instead. + * + * Example: \include MatrixBase_zero_int.cpp + * Output: \verbinclude MatrixBase_zero_int.out + * + * \sa Zero(), Zero(Index,Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero(Index size) +{ + return Constant(size, Scalar(0)); +} + +/** \returns an expression of a fixed-size zero matrix or vector. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_zero.cpp + * Output: \verbinclude MatrixBase_zero.out + * + * \sa Zero(Index), Zero(Index,Index) + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Zero() +{ + return Constant(Scalar(0)); +} + +/** \returns true if *this is approximately equal to the zero matrix, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isZero.cpp + * Output: \verbinclude MatrixBase_isZero.out + * + * \sa class CwiseNullaryOp, Zero() + */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isZero(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < rows(); ++i) + if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast(1), prec)) + return false; + return true; +} + +/** Sets all coefficients in this expression to zero. + * + * Example: \include MatrixBase_setZero.cpp + * Output: \verbinclude MatrixBase_setZero.out + * + * \sa class CwiseNullaryOp, Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setZero() +{ + return setConstant(Scalar(0)); +} + +/** Resizes to the given \a size, and sets all coefficients in this expression to zero. + * + * \only_for_vectors + * + * Example: \include Matrix_setZero_int.cpp + * Output: \verbinclude Matrix_setZero_int.out + * + * \sa DenseBase::setZero(), setZero(Index,Index), class CwiseNullaryOp, DenseBase::Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setZero(Index newSize) +{ + resize(newSize); + return setConstant(Scalar(0)); +} + +/** Resizes to the given size, and sets all coefficients in this expression to zero. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setZero_int_int.cpp + * Output: \verbinclude Matrix_setZero_int_int.out + * + * \sa DenseBase::setZero(), setZero(Index), class CwiseNullaryOp, DenseBase::Zero() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setZero(Index rows, Index cols) +{ + resize(rows, cols); + return setConstant(Scalar(0)); +} + +// ones: + +/** \returns an expression of a matrix where all coefficients equal one. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Ones() should be used + * instead. + * + * Example: \include MatrixBase_ones_int_int.cpp + * Output: \verbinclude MatrixBase_ones_int_int.out + * + * \sa Ones(), Ones(Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones(Index rows, Index cols) +{ + return Constant(rows, cols, Scalar(1)); +} + +/** \returns an expression of a vector where all coefficients equal one. + * + * The parameter \a newSize is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Ones() should be used + * instead. + * + * Example: \include MatrixBase_ones_int.cpp + * Output: \verbinclude MatrixBase_ones_int.out + * + * \sa Ones(), Ones(Index,Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones(Index newSize) +{ + return Constant(newSize, Scalar(1)); +} + +/** \returns an expression of a fixed-size matrix or vector where all coefficients equal one. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_ones.cpp + * Output: \verbinclude MatrixBase_ones.out + * + * \sa Ones(Index), Ones(Index,Index), isOnes(), class Ones + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase::ConstantReturnType +DenseBase::Ones() +{ + return Constant(Scalar(1)); +} + +/** \returns true if *this is approximately equal to the matrix where all coefficients + * are equal to 1, within the precision given by \a prec. + * + * Example: \include MatrixBase_isOnes.cpp + * Output: \verbinclude MatrixBase_isOnes.out + * + * \sa class CwiseNullaryOp, Ones() + */ +template +EIGEN_DEVICE_FUNC bool DenseBase::isOnes +(const RealScalar& prec) const +{ + return isApproxToConstant(Scalar(1), prec); +} + +/** Sets all coefficients in this expression to one. + * + * Example: \include MatrixBase_setOnes.cpp + * Output: \verbinclude MatrixBase_setOnes.out + * + * \sa class CwiseNullaryOp, Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::setOnes() +{ + return setConstant(Scalar(1)); +} + +/** Resizes to the given \a newSize, and sets all coefficients in this expression to one. + * + * \only_for_vectors + * + * Example: \include Matrix_setOnes_int.cpp + * Output: \verbinclude Matrix_setOnes_int.out + * + * \sa MatrixBase::setOnes(), setOnes(Index,Index), class CwiseNullaryOp, MatrixBase::Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setOnes(Index newSize) +{ + resize(newSize); + return setConstant(Scalar(1)); +} + +/** Resizes to the given size, and sets all coefficients in this expression to one. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setOnes_int_int.cpp + * Output: \verbinclude Matrix_setOnes_int_int.out + * + * \sa MatrixBase::setOnes(), setOnes(Index), class CwiseNullaryOp, MatrixBase::Ones() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setOnes(Index rows, Index cols) +{ + resize(rows, cols); + return setConstant(Scalar(1)); +} + +// Identity: + +/** \returns an expression of the identity matrix (not necessarily square). + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Identity() should be used + * instead. + * + * Example: \include MatrixBase_identity_int_int.cpp + * Output: \verbinclude MatrixBase_identity_int_int.out + * + * \sa Identity(), setIdentity(), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::IdentityReturnType +MatrixBase::Identity(Index rows, Index cols) +{ + return DenseBase::NullaryExpr(rows, cols, internal::scalar_identity_op()); +} + +/** \returns an expression of the identity matrix (not necessarily square). + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variant taking size arguments. + * + * Example: \include MatrixBase_identity.cpp + * Output: \verbinclude MatrixBase_identity.out + * + * \sa Identity(Index,Index), setIdentity(), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::IdentityReturnType +MatrixBase::Identity() +{ + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + return MatrixBase::NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_identity_op()); +} + +/** \returns true if *this is approximately equal to the identity matrix + * (not necessarily square), + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isIdentity.cpp + * Output: \verbinclude MatrixBase_isIdentity.out + * + * \sa class CwiseNullaryOp, Identity(), Identity(Index,Index), setIdentity() + */ +template +bool MatrixBase::isIdentity +(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index j = 0; j < cols(); ++j) + { + for(Index i = 0; i < rows(); ++i) + { + if(i == j) + { + if(!internal::isApprox(self.coeff(i, j), static_cast(1), prec)) + return false; + } + else + { + if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast(1), prec)) + return false; + } + } + } + return true; +} + +namespace internal { + +template=16)> +struct setIdentity_impl +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Derived& run(Derived& m) + { + return m = Derived::Identity(m.rows(), m.cols()); + } +}; + +template +struct setIdentity_impl +{ + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Derived& run(Derived& m) + { + m.setZero(); + const Index size = numext::mini(m.rows(), m.cols()); + for(Index i = 0; i < size; ++i) m.coeffRef(i,i) = typename Derived::Scalar(1); + return m; + } +}; + +} // end namespace internal + +/** Writes the identity expression (not necessarily square) into *this. + * + * Example: \include MatrixBase_setIdentity.cpp + * Output: \verbinclude MatrixBase_setIdentity.out + * + * \sa class CwiseNullaryOp, Identity(), Identity(Index,Index), isIdentity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase::setIdentity() +{ + return internal::setIdentity_impl::run(derived()); +} + +/** \brief Resizes to the given size, and writes the identity expression (not necessarily square) into *this. + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setIdentity_int_int.cpp + * Output: \verbinclude Matrix_setIdentity_int_int.out + * + * \sa MatrixBase::setIdentity(), class CwiseNullaryOp, MatrixBase::Identity() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase::setIdentity(Index rows, Index cols) +{ + derived().resize(rows, cols); + return setIdentity(); +} + +/** \returns an expression of the i-th unit (basis) vector. + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::Unit(Index newSize, Index i) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return BasisReturnType(SquareMatrixType::Identity(newSize,newSize), i); +} + +/** \returns an expression of the i-th unit (basis) vector. + * + * \only_for_vectors + * + * This variant is for fixed-size vector only. + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::Unit(Index i) +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + return BasisReturnType(SquareMatrixType::Identity(),i); +} + +/** \returns an expression of the X axis unit vector (1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitX() +{ return Derived::Unit(0); } + +/** \returns an expression of the Y axis unit vector (0,1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitY() +{ return Derived::Unit(1); } + +/** \returns an expression of the Z axis unit vector (0,0,1{,0}^*) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitZ() +{ return Derived::Unit(2); } + +/** \returns an expression of the W axis unit vector (0,0,0,1) + * + * \only_for_vectors + * + * \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW() + */ +template +EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase::BasisReturnType MatrixBase::UnitW() +{ return Derived::Unit(3); } + +} // end namespace Eigen + +#endif // EIGEN_CWISE_NULLARY_OP_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h new file mode 100644 index 00000000000..9f3576fecef --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseTernaryOp.h @@ -0,0 +1,197 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2016 Eugene Brevdo +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_TERNARY_OP_H +#define EIGEN_CWISE_TERNARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > { + // we must not inherit from traits since it has + // the potential to cause problems with MSVC + typedef typename remove_all::type Ancestor; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = traits::RowsAtCompileTime, + ColsAtCompileTime = traits::ColsAtCompileTime, + MaxRowsAtCompileTime = traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = traits::MaxColsAtCompileTime + }; + + // even though we require Arg1, Arg2, and Arg3 to have the same scalar type + // (see CwiseTernaryOp constructor), + // we still want to handle the case when the result type is different. + typedef typename result_of::type Scalar; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + + typedef typename Arg1::Nested Arg1Nested; + typedef typename Arg2::Nested Arg2Nested; + typedef typename Arg3::Nested Arg3Nested; + typedef typename remove_reference::type _Arg1Nested; + typedef typename remove_reference::type _Arg2Nested; + typedef typename remove_reference::type _Arg3Nested; + enum { Flags = _Arg1Nested::Flags & RowMajorBit }; +}; +} // end namespace internal + +template +class CwiseTernaryOpImpl; + +/** \class CwiseTernaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise ternary operator is + * applied to two expressions + * + * \tparam TernaryOp template functor implementing the operator + * \tparam Arg1Type the type of the first argument + * \tparam Arg2Type the type of the second argument + * \tparam Arg3Type the type of the third argument + * + * This class represents an expression where a coefficient-wise ternary + * operator is applied to three expressions. + * It is the return type of ternary operators, by which we mean only those + * ternary operators where + * all three arguments are Eigen expressions. + * For example, the return type of betainc(matrix1, matrix2, matrix3) is a + * CwiseTernaryOp. + * + * Most of the time, this is the only way that it is used, so you typically + * don't have to name + * CwiseTernaryOp types explicitly. + * + * \sa MatrixBase::ternaryExpr(const MatrixBase &, const + * MatrixBase &, const CustomTernaryOp &) const, class CwiseBinaryOp, + * class CwiseUnaryOp, class CwiseNullaryOp + */ +template +class CwiseTernaryOp : public CwiseTernaryOpImpl< + TernaryOp, Arg1Type, Arg2Type, Arg3Type, + typename internal::traits::StorageKind>, + internal::no_assignment_operator +{ + public: + typedef typename internal::remove_all::type Arg1; + typedef typename internal::remove_all::type Arg2; + typedef typename internal::remove_all::type Arg3; + + typedef typename CwiseTernaryOpImpl< + TernaryOp, Arg1Type, Arg2Type, Arg3Type, + typename internal::traits::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseTernaryOp) + + typedef typename internal::ref_selector::type Arg1Nested; + typedef typename internal::ref_selector::type Arg2Nested; + typedef typename internal::ref_selector::type Arg3Nested; + typedef typename internal::remove_reference::type _Arg1Nested; + typedef typename internal::remove_reference::type _Arg2Nested; + typedef typename internal::remove_reference::type _Arg3Nested; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CwiseTernaryOp(const Arg1& a1, const Arg2& a2, + const Arg3& a3, + const TernaryOp& func = TernaryOp()) + : m_arg1(a1), m_arg2(a2), m_arg3(a3), m_functor(func) { + // require the sizes to match + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg2) + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg3) + + // The index types should match + EIGEN_STATIC_ASSERT((internal::is_same< + typename internal::traits::StorageKind, + typename internal::traits::StorageKind>::value), + STORAGE_KIND_MUST_MATCH) + EIGEN_STATIC_ASSERT((internal::is_same< + typename internal::traits::StorageKind, + typename internal::traits::StorageKind>::value), + STORAGE_KIND_MUST_MATCH) + + eigen_assert(a1.rows() == a2.rows() && a1.cols() == a2.cols() && + a1.rows() == a3.rows() && a1.cols() == a3.cols()); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { + // return the fixed size type if available to enable compile time + // optimizations + if (internal::traits::type>:: + RowsAtCompileTime == Dynamic && + internal::traits::type>:: + RowsAtCompileTime == Dynamic) + return m_arg3.rows(); + else if (internal::traits::type>:: + RowsAtCompileTime == Dynamic && + internal::traits::type>:: + RowsAtCompileTime == Dynamic) + return m_arg2.rows(); + else + return m_arg1.rows(); + } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { + // return the fixed size type if available to enable compile time + // optimizations + if (internal::traits::type>:: + ColsAtCompileTime == Dynamic && + internal::traits::type>:: + ColsAtCompileTime == Dynamic) + return m_arg3.cols(); + else if (internal::traits::type>:: + ColsAtCompileTime == Dynamic && + internal::traits::type>:: + ColsAtCompileTime == Dynamic) + return m_arg2.cols(); + else + return m_arg1.cols(); + } + + /** \returns the first argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg1Nested& arg1() const { return m_arg1; } + /** \returns the first argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg2Nested& arg2() const { return m_arg2; } + /** \returns the third argument nested expression */ + EIGEN_DEVICE_FUNC + const _Arg3Nested& arg3() const { return m_arg3; } + /** \returns the functor representing the ternary operation */ + EIGEN_DEVICE_FUNC + const TernaryOp& functor() const { return m_functor; } + + protected: + Arg1Nested m_arg1; + Arg2Nested m_arg2; + Arg3Nested m_arg3; + const TernaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseTernaryOpImpl + : public internal::generic_xpr_base< + CwiseTernaryOp >::type { + public: + typedef typename internal::generic_xpr_base< + CwiseTernaryOp >::type Base; +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_TERNARY_OP_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h new file mode 100644 index 00000000000..1d2dd19f2b1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryOp.h @@ -0,0 +1,103 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2014 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_UNARY_OP_H +#define EIGEN_CWISE_UNARY_OP_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename result_of< + UnaryOp(const typename XprType::Scalar&) + >::type Scalar; + typedef typename XprType::Nested XprTypeNested; + typedef typename remove_reference::type _XprTypeNested; + enum { + Flags = _XprTypeNested::Flags & RowMajorBit + }; +}; +} + +template +class CwiseUnaryOpImpl; + +/** \class CwiseUnaryOp + * \ingroup Core_Module + * + * \brief Generic expression where a coefficient-wise unary operator is applied to an expression + * + * \tparam UnaryOp template functor implementing the operator + * \tparam XprType the type of the expression to which we are applying the unary operator + * + * This class represents an expression where a unary operator is applied to an expression. + * It is the return type of all operations taking exactly 1 input expression, regardless of the + * presence of other inputs such as scalars. For example, the operator* in the expression 3*matrix + * is considered unary, because only the right-hand side is an expression, and its + * return type is a specialization of CwiseUnaryOp. + * + * Most of the time, this is the only way that it is used, so you typically don't have to name + * CwiseUnaryOp types explicitly. + * + * \sa MatrixBase::unaryExpr(const CustomUnaryOp &) const, class CwiseBinaryOp, class CwiseNullaryOp + */ +template +class CwiseUnaryOp : public CwiseUnaryOpImpl::StorageKind>, internal::no_assignment_operator +{ + public: + + typedef typename CwiseUnaryOpImpl::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryOp) + typedef typename internal::ref_selector::type XprTypeNested; + typedef typename internal::remove_all::type NestedExpression; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit CwiseUnaryOp(const XprType& xpr, const UnaryOp& func = UnaryOp()) + : m_xpr(xpr), m_functor(func) {} + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Index cols() const { return m_xpr.cols(); } + + /** \returns the functor representing the unary operation */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const UnaryOp& functor() const { return m_functor; } + + /** \returns the nested expression */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + const typename internal::remove_all::type& + nestedExpression() const { return m_xpr; } + + /** \returns the nested expression */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + typename internal::remove_all::type& + nestedExpression() { return m_xpr; } + + protected: + XprTypeNested m_xpr; + const UnaryOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseUnaryOpImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_UNARY_OP_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h new file mode 100644 index 00000000000..5a30fa8df18 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/CwiseUnaryView.h @@ -0,0 +1,130 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_CWISE_UNARY_VIEW_H +#define EIGEN_CWISE_UNARY_VIEW_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename result_of< + ViewOp(const typename traits::Scalar&) + >::type Scalar; + typedef typename MatrixType::Nested MatrixTypeNested; + typedef typename remove_all::type _MatrixTypeNested; + enum { + FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, + Flags = traits<_MatrixTypeNested>::Flags & (RowMajorBit | FlagsLvalueBit | DirectAccessBit), // FIXME DirectAccessBit should not be handled by expressions + MatrixTypeInnerStride = inner_stride_at_compile_time::ret, + // need to cast the sizeof's from size_t to int explicitly, otherwise: + // "error: no integral type can represent all of the enumerator values + InnerStrideAtCompileTime = MatrixTypeInnerStride == Dynamic + ? int(Dynamic) + : int(MatrixTypeInnerStride) * int(sizeof(typename traits::Scalar) / sizeof(Scalar)), + OuterStrideAtCompileTime = outer_stride_at_compile_time::ret == Dynamic + ? int(Dynamic) + : outer_stride_at_compile_time::ret * int(sizeof(typename traits::Scalar) / sizeof(Scalar)) + }; +}; +} + +template +class CwiseUnaryViewImpl; + +/** \class CwiseUnaryView + * \ingroup Core_Module + * + * \brief Generic lvalue expression of a coefficient-wise unary operator of a matrix or a vector + * + * \tparam ViewOp template functor implementing the view + * \tparam MatrixType the type of the matrix we are applying the unary operator + * + * This class represents a lvalue expression of a generic unary view operator of a matrix or a vector. + * It is the return type of real() and imag(), and most of the time this is the only way it is used. + * + * \sa MatrixBase::unaryViewExpr(const CustomUnaryOp &) const, class CwiseUnaryOp + */ +template +class CwiseUnaryView : public CwiseUnaryViewImpl::StorageKind> +{ + public: + + typedef typename CwiseUnaryViewImpl::StorageKind>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryView) + typedef typename internal::ref_selector::non_const_type MatrixTypeNested; + typedef typename internal::remove_all::type NestedExpression; + + explicit inline CwiseUnaryView(MatrixType& mat, const ViewOp& func = ViewOp()) + : m_matrix(mat), m_functor(func) {} + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryView) + + EIGEN_STRONG_INLINE Index rows() const { return m_matrix.rows(); } + EIGEN_STRONG_INLINE Index cols() const { return m_matrix.cols(); } + + /** \returns the functor representing unary operation */ + const ViewOp& functor() const { return m_functor; } + + /** \returns the nested expression */ + const typename internal::remove_all::type& + nestedExpression() const { return m_matrix; } + + /** \returns the nested expression */ + typename internal::remove_reference::type& + nestedExpression() { return m_matrix.const_cast_derived(); } + + protected: + MatrixTypeNested m_matrix; + ViewOp m_functor; +}; + +// Generic API dispatcher +template +class CwiseUnaryViewImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; +}; + +template +class CwiseUnaryViewImpl + : public internal::dense_xpr_base< CwiseUnaryView >::type +{ + public: + + typedef CwiseUnaryView Derived; + typedef typename internal::dense_xpr_base< CwiseUnaryView >::type Base; + + EIGEN_DENSE_PUBLIC_INTERFACE(Derived) + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryViewImpl) + + EIGEN_DEVICE_FUNC inline Scalar* data() { return &(this->coeffRef(0)); } + EIGEN_DEVICE_FUNC inline const Scalar* data() const { return &(this->coeff(0)); } + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return derived().nestedExpression().innerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); + } + + EIGEN_DEVICE_FUNC inline Index outerStride() const + { + return derived().nestedExpression().outerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); + } + protected: + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(CwiseUnaryViewImpl) +}; + +} // end namespace Eigen + +#endif // EIGEN_CWISE_UNARY_VIEW_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h new file mode 100644 index 00000000000..c55a68230cd --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseBase.h @@ -0,0 +1,612 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSEBASE_H +#define EIGEN_DENSEBASE_H + +namespace Eigen { + +namespace internal { + +// The index type defined by EIGEN_DEFAULT_DENSE_INDEX_TYPE must be a signed type. +// This dummy function simply aims at checking that at compile time. +static inline void check_DenseIndex_is_signed() { + EIGEN_STATIC_ASSERT(NumTraits::IsSigned,THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE); +} + +} // end namespace internal + +/** \class DenseBase + * \ingroup Core_Module + * + * \brief Base class for all dense matrices, vectors, and arrays + * + * This class is the base that is inherited by all dense objects (matrix, vector, arrays, + * and related expression types). The common Eigen API for dense objects is contained in this class. + * + * \tparam Derived is the derived type, e.g., a matrix type or an expression. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_DENSEBASE_PLUGIN. + * + * \sa \blank \ref TopicClassHierarchy + */ +template class DenseBase +#ifndef EIGEN_PARSED_BY_DOXYGEN + : public DenseCoeffsBase::value> +#else + : public DenseCoeffsBase +#endif // not EIGEN_PARSED_BY_DOXYGEN +{ + public: + + /** Inner iterator type to iterate over the coefficients of a row or column. + * \sa class InnerIterator + */ + typedef Eigen::InnerIterator InnerIterator; + + typedef typename internal::traits::StorageKind StorageKind; + + /** + * \brief The type used to store indices + * \details This typedef is relevant for types that store multiple indices such as + * PermutationMatrix or Transpositions, otherwise it defaults to Eigen::Index + * \sa \blank \ref TopicPreprocessorDirectives, Eigen::Index, SparseMatrixBase. + */ + typedef typename internal::traits::StorageIndex StorageIndex; + + /** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex, etc. */ + typedef typename internal::traits::Scalar Scalar; + + /** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex, etc. + * + * It is an alias for the Scalar type */ + typedef Scalar value_type; + + typedef typename NumTraits::Real RealScalar; + typedef DenseCoeffsBase::value> Base; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::rowIndexByOuterInner; + using Base::colIndexByOuterInner; + using Base::coeff; + using Base::coeffByOuterInner; + using Base::operator(); + using Base::operator[]; + using Base::x; + using Base::y; + using Base::z; + using Base::w; + using Base::stride; + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + typedef typename Base::CoeffReturnType CoeffReturnType; + + enum { + + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + /**< The number of rows at compile-time. This is just a copy of the value provided + * by the \a Derived type. If a value is not known at compile-time, + * it is set to the \a Dynamic constant. + * \sa MatrixBase::rows(), MatrixBase::cols(), ColsAtCompileTime, SizeAtCompileTime */ + + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + /**< The number of columns at compile-time. This is just a copy of the value provided + * by the \a Derived type. If a value is not known at compile-time, + * it is set to the \a Dynamic constant. + * \sa MatrixBase::rows(), MatrixBase::cols(), RowsAtCompileTime, SizeAtCompileTime */ + + + SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, + internal::traits::ColsAtCompileTime>::ret), + /**< This is equal to the number of coefficients, i.e. the number of + * rows times the number of columns, or to \a Dynamic if this is not + * known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */ + + MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, + /**< This value is equal to the maximum possible number of rows that this expression + * might have. If this expression might have an arbitrarily high number of rows, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa RowsAtCompileTime, MaxColsAtCompileTime, MaxSizeAtCompileTime + */ + + MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, + /**< This value is equal to the maximum possible number of columns that this expression + * might have. If this expression might have an arbitrarily high number of columns, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa ColsAtCompileTime, MaxRowsAtCompileTime, MaxSizeAtCompileTime + */ + + MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime>::ret), + /**< This value is equal to the maximum possible number of coefficients that this expression + * might have. If this expression might have an arbitrarily high number of coefficients, + * this value is set to \a Dynamic. + * + * This value is useful to know when evaluating an expression, in order to determine + * whether it is possible to avoid doing a dynamic memory allocation. + * + * \sa SizeAtCompileTime, MaxRowsAtCompileTime, MaxColsAtCompileTime + */ + + IsVectorAtCompileTime = internal::traits::MaxRowsAtCompileTime == 1 + || internal::traits::MaxColsAtCompileTime == 1, + /**< This is set to true if either the number of rows or the number of + * columns is known at compile-time to be equal to 1. Indeed, in that case, + * we are dealing with a column-vector (if there is only one column) or with + * a row-vector (if there is only one row). */ + + Flags = internal::traits::Flags, + /**< This stores expression \ref flags flags which may or may not be inherited by new expressions + * constructed from this one. See the \ref flags "list of flags". + */ + + IsRowMajor = int(Flags) & RowMajorBit, /**< True if this expression has row-major storage order. */ + + InnerSizeAtCompileTime = int(IsVectorAtCompileTime) ? int(SizeAtCompileTime) + : int(IsRowMajor) ? int(ColsAtCompileTime) : int(RowsAtCompileTime), + + InnerStrideAtCompileTime = internal::inner_stride_at_compile_time::ret, + OuterStrideAtCompileTime = internal::outer_stride_at_compile_time::ret + }; + + typedef typename internal::find_best_packet::type PacketScalar; + + enum { IsPlainObjectBase = 0 }; + + /** The plain matrix type corresponding to this expression. + * \sa PlainObject */ + typedef Matrix::Scalar, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime, + AutoAlign | (internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor), + internal::traits::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime + > PlainMatrix; + + /** The plain array type corresponding to this expression. + * \sa PlainObject */ + typedef Array::Scalar, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime, + AutoAlign | (internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor), + internal::traits::MaxRowsAtCompileTime, + internal::traits::MaxColsAtCompileTime + > PlainArray; + + /** \brief The plain matrix or array type corresponding to this expression. + * + * This is not necessarily exactly the return type of eval(). In the case of plain matrices, + * the return type of eval() is a const reference to a matrix, not a matrix! It is however guaranteed + * that the return type of eval() is either PlainObject or const PlainObject&. + */ + typedef typename internal::conditional::XprKind,MatrixXpr >::value, + PlainMatrix, PlainArray>::type PlainObject; + + /** \returns the number of nonzero coefficients which is in practice the number + * of stored coefficients. */ + EIGEN_DEVICE_FUNC + inline Index nonZeros() const { return size(); } + + /** \returns the outer size. + * + * \note For a vector, this returns just 1. For a matrix (non-vector), this is the major dimension + * with respect to the \ref TopicStorageOrders "storage order", i.e., the number of columns for a + * column-major matrix, and the number of rows for a row-major matrix. */ + EIGEN_DEVICE_FUNC + Index outerSize() const + { + return IsVectorAtCompileTime ? 1 + : int(IsRowMajor) ? this->rows() : this->cols(); + } + + /** \returns the inner size. + * + * \note For a vector, this is just the size. For a matrix (non-vector), this is the minor dimension + * with respect to the \ref TopicStorageOrders "storage order", i.e., the number of rows for a + * column-major matrix, and the number of columns for a row-major matrix. */ + EIGEN_DEVICE_FUNC + Index innerSize() const + { + return IsVectorAtCompileTime ? this->size() + : int(IsRowMajor) ? this->cols() : this->rows(); + } + + /** Only plain matrices/arrays, not expressions, may be resized; therefore the only useful resize methods are + * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does + * nothing else. + */ + EIGEN_DEVICE_FUNC + void resize(Index newSize) + { + EIGEN_ONLY_USED_FOR_DEBUG(newSize); + eigen_assert(newSize == this->size() + && "DenseBase::resize() does not actually allow to resize."); + } + /** Only plain matrices/arrays, not expressions, may be resized; therefore the only useful resize methods are + * Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does + * nothing else. + */ + EIGEN_DEVICE_FUNC + void resize(Index rows, Index cols) + { + EIGEN_ONLY_USED_FOR_DEBUG(rows); + EIGEN_ONLY_USED_FOR_DEBUG(cols); + eigen_assert(rows == this->rows() && cols == this->cols() + && "DenseBase::resize() does not actually allow to resize."); + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; + /** \internal \deprecated Represents a vector with linearly spaced coefficients that allows sequential access only. */ + typedef CwiseNullaryOp,PlainObject> SequentialLinSpacedReturnType; + /** \internal Represents a vector with linearly spaced coefficients that allows random access. */ + typedef CwiseNullaryOp,PlainObject> RandomAccessLinSpacedReturnType; + /** \internal the return type of MatrixBase::eigenvalues() */ + typedef Matrix::Scalar>::Real, internal::traits::ColsAtCompileTime, 1> EigenvaluesReturnType; + +#endif // not EIGEN_PARSED_BY_DOXYGEN + + /** Copies \a other into *this. \returns a reference to *this. */ + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator+=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator-=(const EigenBase &other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const ReturnByValue& func); + + /** \internal + * Copies \a other into *this without evaluating other. \returns a reference to *this. + * \deprecated */ + template + EIGEN_DEVICE_FUNC + Derived& lazyAssign(const DenseBase& other); + + EIGEN_DEVICE_FUNC + CommaInitializer operator<< (const Scalar& s); + + /** \deprecated it now returns \c *this */ + template + EIGEN_DEPRECATED + const Derived& flagged() const + { return derived(); } + + template + EIGEN_DEVICE_FUNC + CommaInitializer operator<< (const DenseBase& other); + + typedef Transpose TransposeReturnType; + EIGEN_DEVICE_FUNC + TransposeReturnType transpose(); + typedef typename internal::add_const >::type ConstTransposeReturnType; + EIGEN_DEVICE_FUNC + ConstTransposeReturnType transpose() const; + EIGEN_DEVICE_FUNC + void transposeInPlace(); + + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(Index rows, Index cols, const Scalar& value); + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(Index size, const Scalar& value); + EIGEN_DEVICE_FUNC static const ConstantReturnType + Constant(const Scalar& value); + + EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType + LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType + LinSpaced(Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType + LinSpaced(Sequential_t, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType + LinSpaced(const Scalar& low, const Scalar& high); + + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func); + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(Index size, const CustomNullaryOp& func); + template EIGEN_DEVICE_FUNC + static const CwiseNullaryOp + NullaryExpr(const CustomNullaryOp& func); + + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index size); + EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index size); + EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(); + + EIGEN_DEVICE_FUNC void fill(const Scalar& value); + EIGEN_DEVICE_FUNC Derived& setConstant(const Scalar& value); + EIGEN_DEVICE_FUNC Derived& setLinSpaced(Index size, const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC Derived& setLinSpaced(const Scalar& low, const Scalar& high); + EIGEN_DEVICE_FUNC Derived& setZero(); + EIGEN_DEVICE_FUNC Derived& setOnes(); + EIGEN_DEVICE_FUNC Derived& setRandom(); + + template EIGEN_DEVICE_FUNC + bool isApprox(const DenseBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC + bool isMuchSmallerThan(const RealScalar& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + template EIGEN_DEVICE_FUNC + bool isMuchSmallerThan(const DenseBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + + EIGEN_DEVICE_FUNC bool isApproxToConstant(const Scalar& value, const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isConstant(const Scalar& value, const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isZero(const RealScalar& prec = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC bool isOnes(const RealScalar& prec = NumTraits::dummy_precision()) const; + + inline bool hasNaN() const; + inline bool allFinite() const; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator*=(const Scalar& other); + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator/=(const Scalar& other); + + typedef typename internal::add_const_on_value_type::type>::type EvalReturnType; + /** \returns the matrix or vector obtained by evaluating this expression. + * + * Notice that in the case of a plain matrix or vector (not an expression) this function just returns + * a const reference, in order to avoid a useless copy. + * + * \warning Be carefull with eval() and the auto C++ keyword, as detailed in this \link TopicPitfalls_auto_keyword page \endlink. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE EvalReturnType eval() const + { + // Even though MSVC does not honor strong inlining when the return type + // is a dynamic matrix, we desperately need strong inlining for fixed + // size types on MSVC. + return typename internal::eval::type(derived()); + } + + /** swaps *this with the expression \a other. + * + */ + template + EIGEN_DEVICE_FUNC + void swap(const DenseBase& other) + { + EIGEN_STATIC_ASSERT(!OtherDerived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + eigen_assert(rows()==other.rows() && cols()==other.cols()); + call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); + } + + /** swaps *this with the matrix or array \a other. + * + */ + template + EIGEN_DEVICE_FUNC + void swap(PlainObjectBase& other) + { + eigen_assert(rows()==other.rows() && cols()==other.cols()); + call_assignment(derived(), other.derived(), internal::swap_assign_op()); + } + + EIGEN_DEVICE_FUNC inline const NestByValue nestByValue() const; + EIGEN_DEVICE_FUNC inline const ForceAlignedAccess forceAlignedAccess() const; + EIGEN_DEVICE_FUNC inline ForceAlignedAccess forceAlignedAccess(); + template EIGEN_DEVICE_FUNC + inline const typename internal::conditional,Derived&>::type forceAlignedAccessIf() const; + template EIGEN_DEVICE_FUNC + inline typename internal::conditional,Derived&>::type forceAlignedAccessIf(); + + EIGEN_DEVICE_FUNC Scalar sum() const; + EIGEN_DEVICE_FUNC Scalar mean() const; + EIGEN_DEVICE_FUNC Scalar trace() const; + + EIGEN_DEVICE_FUNC Scalar prod() const; + + EIGEN_DEVICE_FUNC typename internal::traits::Scalar minCoeff() const; + EIGEN_DEVICE_FUNC typename internal::traits::Scalar maxCoeff() const; + + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar minCoeff(IndexType* row, IndexType* col) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar maxCoeff(IndexType* row, IndexType* col) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar minCoeff(IndexType* index) const; + template EIGEN_DEVICE_FUNC + typename internal::traits::Scalar maxCoeff(IndexType* index) const; + + template + EIGEN_DEVICE_FUNC + Scalar redux(const BinaryOp& func) const; + + template + EIGEN_DEVICE_FUNC + void visit(Visitor& func) const; + + /** \returns a WithFormat proxy object allowing to print a matrix the with given + * format \a fmt. + * + * See class IOFormat for some examples. + * + * \sa class IOFormat, class WithFormat + */ + inline const WithFormat format(const IOFormat& fmt) const + { + return WithFormat(derived(), fmt); + } + + /** \returns the unique coefficient of a 1x1 expression */ + EIGEN_DEVICE_FUNC + CoeffReturnType value() const + { + EIGEN_STATIC_ASSERT_SIZE_1x1(Derived) + eigen_assert(this->rows() == 1 && this->cols() == 1); + return derived().coeff(0,0); + } + + EIGEN_DEVICE_FUNC bool all() const; + EIGEN_DEVICE_FUNC bool any() const; + EIGEN_DEVICE_FUNC Index count() const; + + typedef VectorwiseOp RowwiseReturnType; + typedef const VectorwiseOp ConstRowwiseReturnType; + typedef VectorwiseOp ColwiseReturnType; + typedef const VectorwiseOp ConstColwiseReturnType; + + /** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations + * + * Example: \include MatrixBase_rowwise.cpp + * Output: \verbinclude MatrixBase_rowwise.out + * + * \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting + */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC inline ConstRowwiseReturnType rowwise() const { + return ConstRowwiseReturnType(derived()); + } + EIGEN_DEVICE_FUNC RowwiseReturnType rowwise(); + + /** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations + * + * Example: \include MatrixBase_colwise.cpp + * Output: \verbinclude MatrixBase_colwise.out + * + * \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting + */ + EIGEN_DEVICE_FUNC inline ConstColwiseReturnType colwise() const { + return ConstColwiseReturnType(derived()); + } + EIGEN_DEVICE_FUNC ColwiseReturnType colwise(); + + typedef CwiseNullaryOp,PlainObject> RandomReturnType; + static const RandomReturnType Random(Index rows, Index cols); + static const RandomReturnType Random(Index size); + static const RandomReturnType Random(); + + template + const Select + select(const DenseBase& thenMatrix, + const DenseBase& elseMatrix) const; + + template + inline const Select + select(const DenseBase& thenMatrix, const typename ThenDerived::Scalar& elseScalar) const; + + template + inline const Select + select(const typename ElseDerived::Scalar& thenScalar, const DenseBase& elseMatrix) const; + + template RealScalar lpNorm() const; + + template + EIGEN_DEVICE_FUNC + const Replicate replicate() const; + /** + * \return an expression of the replication of \c *this + * + * Example: \include MatrixBase_replicate_int_int.cpp + * Output: \verbinclude MatrixBase_replicate_int_int.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(), class Replicate + */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC + const Replicate replicate(Index rowFactor, Index colFactor) const + { + return Replicate(derived(), rowFactor, colFactor); + } + + typedef Reverse ReverseReturnType; + typedef const Reverse ConstReverseReturnType; + EIGEN_DEVICE_FUNC ReverseReturnType reverse(); + /** This is the const version of reverse(). */ + //Code moved here due to a CUDA compiler bug + EIGEN_DEVICE_FUNC ConstReverseReturnType reverse() const + { + return ConstReverseReturnType(derived()); + } + EIGEN_DEVICE_FUNC void reverseInPlace(); + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::DenseBase +#define EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL +#define EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(COND) +# include "../plugins/BlockMethods.h" +# ifdef EIGEN_DENSEBASE_PLUGIN +# include EIGEN_DENSEBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL +#undef EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF + + // disable the use of evalTo for dense objects with a nice compilation error + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& ) const + { + EIGEN_STATIC_ASSERT((internal::is_same::value),THE_EVAL_EVALTO_FUNCTION_SHOULD_NEVER_BE_CALLED_FOR_DENSE_OBJECTS); + } + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(DenseBase) + /** Default constructor. Do nothing. */ + EIGEN_DEVICE_FUNC DenseBase() + { + /* Just checks for self-consistency of the flags. + * Only do it when debugging Eigen, as this borders on paranoia and could slow compilation down + */ +#ifdef EIGEN_INTERNAL_DEBUGGING + EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, int(IsRowMajor)) + && EIGEN_IMPLIES(MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1, int(!IsRowMajor))), + INVALID_STORAGE_ORDER_FOR_THIS_VECTOR_EXPRESSION) +#endif + } + + private: + EIGEN_DEVICE_FUNC explicit DenseBase(int); + EIGEN_DEVICE_FUNC DenseBase(int,int); + template EIGEN_DEVICE_FUNC explicit DenseBase(const DenseBase&); +}; + +} // end namespace Eigen + +#endif // EIGEN_DENSEBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h new file mode 100644 index 00000000000..c4af48ab699 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseCoeffsBase.h @@ -0,0 +1,681 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSECOEFFSBASE_H +#define EIGEN_DENSECOEFFSBASE_H + +namespace Eigen { + +namespace internal { +template struct add_const_on_value_type_if_arithmetic +{ + typedef typename conditional::value, T, typename add_const_on_value_type::type>::type type; +}; +} + +/** \brief Base class providing read-only coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #ReadOnlyAccessors Constant indicating read-only access + * + * This class defines the \c operator() \c const function and friends, which can be used to read specific + * entries of a matrix or array. + * + * \sa DenseCoeffsBase, DenseCoeffsBase, + * \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public EigenBase +{ + public: + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + + // Explanation for this CoeffReturnType typedef. + // - This is the return type of the coeff() method. + // - The LvalueBit means exactly that we can offer a coeffRef() method, which means exactly that we can get references + // to coeffs, which means exactly that we can have coeff() return a const reference (as opposed to returning a value). + // - The is_artihmetic check is required since "const int", "const double", etc. will cause warnings on some systems + // while the declaration of "const T", where T is a non arithmetic type does not. Always returning "const Scalar&" is + // not possible, since the underlying expressions might not offer a valid address the reference could be referring to. + typedef typename internal::conditional::Flags&LvalueBit), + const Scalar&, + typename internal::conditional::value, Scalar, const Scalar>::type + >::type CoeffReturnType; + + typedef typename internal::add_const_on_value_type_if_arithmetic< + typename internal::packet_traits::type + >::type PacketReturnType; + + typedef EigenBase Base; + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner) const + { + return int(Derived::RowsAtCompileTime) == 1 ? 0 + : int(Derived::ColsAtCompileTime) == 1 ? inner + : int(Derived::Flags)&RowMajorBit ? outer + : inner; + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner) const + { + return int(Derived::ColsAtCompileTime) == 1 ? 0 + : int(Derived::RowsAtCompileTime) == 1 ? inner + : int(Derived::Flags)&RowMajorBit ? inner + : outer; + } + + /** Short version: don't use this function, use + * \link operator()(Index,Index) const \endlink instead. + * + * Long version: this function is similar to + * \link operator()(Index,Index) const \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator()(Index,Index) const \endlink. + * + * \sa operator()(Index,Index) const, coeffRef(Index,Index), coeff(Index) const + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType coeff(Index row, Index col) const + { + eigen_internal_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return internal::evaluator(derived()).coeff(row,col); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType coeffByOuterInner(Index outer, Index inner) const + { + return coeff(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \returns the coefficient at given the given row and column. + * + * \sa operator()(Index,Index), operator[](Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType operator()(Index row, Index col) const + { + eigen_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return coeff(row, col); + } + + /** Short version: don't use this function, use + * \link operator[](Index) const \endlink instead. + * + * Long version: this function is similar to + * \link operator[](Index) const \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameter \a index is in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator[](Index) const \endlink. + * + * \sa operator[](Index) const, coeffRef(Index), coeff(Index,Index) const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + coeff(Index index) const + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).coeff(index); + } + + + /** \returns the coefficient at given index. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index), operator()(Index,Index) const, x() const, y() const, + * z() const, w() const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + operator[](Index index) const + { + EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime, + THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD) + eigen_assert(index >= 0 && index < size()); + return coeff(index); + } + + /** \returns the coefficient at given index. + * + * This is synonymous to operator[](Index) const. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index), operator()(Index,Index) const, x() const, y() const, + * z() const, w() const + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + operator()(Index index) const + { + eigen_assert(index >= 0 && index < size()); + return coeff(index); + } + + /** equivalent to operator[](0). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + x() const { return (*this)[0]; } + + /** equivalent to operator[](1). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + y() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS); + return (*this)[1]; + } + + /** equivalent to operator[](2). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + z() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS); + return (*this)[2]; + } + + /** equivalent to operator[](3). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE CoeffReturnType + w() const + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS); + return (*this)[3]; + } + + /** \internal + * \returns the packet of coefficients starting at the given row and column. It is your responsibility + * to ensure that a packet really starts there. This method is only available on expressions having the + * PacketAccessBit. + * + * The \a LoadMode parameter may have the value \a #Aligned or \a #Unaligned. Its effect is to select + * the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets + * starting at an address which is a multiple of the packet size. + */ + + template + EIGEN_STRONG_INLINE PacketReturnType packet(Index row, Index col) const + { + typedef typename internal::packet_traits::type DefaultPacketType; + eigen_internal_assert(row >= 0 && row < rows() && col >= 0 && col < cols()); + return internal::evaluator(derived()).template packet(row,col); + } + + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketReturnType packetByOuterInner(Index outer, Index inner) const + { + return packet(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \internal + * \returns the packet of coefficients starting at the given index. It is your responsibility + * to ensure that a packet really starts there. This method is only available on expressions having the + * PacketAccessBit and the LinearAccessBit. + * + * The \a LoadMode parameter may have the value \a #Aligned or \a #Unaligned. Its effect is to select + * the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets + * starting at an address which is a multiple of the packet size. + */ + + template + EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + typedef typename internal::packet_traits::type DefaultPacketType; + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).template packet(index); + } + + protected: + // explanation: DenseBase is doing "using ..." on the methods from DenseCoeffsBase. + // But some methods are only available in the DirectAccess case. + // So we add dummy methods here with these names, so that "using... " doesn't fail. + // It's not private so that the child class DenseBase can access them, and it's not public + // either since it's an implementation detail, so has to be protected. + void coeffRef(); + void coeffRefByOuterInner(); + void writePacket(); + void writePacketByOuterInner(); + void copyCoeff(); + void copyCoeffByOuterInner(); + void copyPacket(); + void copyPacketByOuterInner(); + void stride(); + void innerStride(); + void outerStride(); + void rowStride(); + void colStride(); +}; + +/** \brief Base class providing read/write coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #WriteAccessors Constant indicating read/write access + * + * This class defines the non-const \c operator() function and friends, which can be used to write specific + * entries of a matrix or array. This class inherits DenseCoeffsBase which + * defines the const variant for reading specific entries. + * + * \sa DenseCoeffsBase, \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + using Base::coeff; + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + using Base::rowIndexByOuterInner; + using Base::colIndexByOuterInner; + using Base::operator[]; + using Base::operator(); + using Base::x; + using Base::y; + using Base::z; + using Base::w; + + /** Short version: don't use this function, use + * \link operator()(Index,Index) \endlink instead. + * + * Long version: this function is similar to + * \link operator()(Index,Index) \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator()(Index,Index) \endlink. + * + * \sa operator()(Index,Index), coeff(Index, Index) const, coeffRef(Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index row, Index col) + { + eigen_internal_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return internal::evaluator(derived()).coeffRef(row,col); + } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + coeffRefByOuterInner(Index outer, Index inner) + { + return coeffRef(rowIndexByOuterInner(outer, inner), + colIndexByOuterInner(outer, inner)); + } + + /** \returns a reference to the coefficient at given the given row and column. + * + * \sa operator[](Index) + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator()(Index row, Index col) + { + eigen_assert(row >= 0 && row < rows() + && col >= 0 && col < cols()); + return coeffRef(row, col); + } + + + /** Short version: don't use this function, use + * \link operator[](Index) \endlink instead. + * + * Long version: this function is similar to + * \link operator[](Index) \endlink, but without the assertion. + * Use this for limiting the performance cost of debugging code when doing + * repeated coefficient access. Only use this when it is guaranteed that the + * parameters \a row and \a col are in range. + * + * If EIGEN_INTERNAL_DEBUGGING is defined, an assertion will be made, making this + * function equivalent to \link operator[](Index) \endlink. + * + * \sa operator[](Index), coeff(Index) const, coeffRef(Index,Index) + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + coeffRef(Index index) + { + EIGEN_STATIC_ASSERT(internal::evaluator::Flags & LinearAccessBit, + THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS) + eigen_internal_assert(index >= 0 && index < size()); + return internal::evaluator(derived()).coeffRef(index); + } + + /** \returns a reference to the coefficient at given index. + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w() + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator[](Index index) + { + EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime, + THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD) + eigen_assert(index >= 0 && index < size()); + return coeffRef(index); + } + + /** \returns a reference to the coefficient at given index. + * + * This is synonymous to operator[](Index). + * + * This method is allowed only for vector expressions, and for matrix expressions having the LinearAccessBit. + * + * \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w() + */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + operator()(Index index) + { + eigen_assert(index >= 0 && index < size()); + return coeffRef(index); + } + + /** equivalent to operator[](0). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + x() { return (*this)[0]; } + + /** equivalent to operator[](1). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + y() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS); + return (*this)[1]; + } + + /** equivalent to operator[](2). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + z() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS); + return (*this)[2]; + } + + /** equivalent to operator[](3). */ + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& + w() + { + EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS); + return (*this)[3]; + } +}; + +/** \brief Base class providing direct read-only coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #DirectAccessors Constant indicating direct access + * + * This class defines functions to work with strides which can be used to access entries directly. This class + * inherits DenseCoeffsBase which defines functions to access entries read-only using + * \c operator() . + * + * \sa \blank \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + typedef typename internal::traits::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + /** \returns the pointer increment between two consecutive elements within a slice in the inner direction. + * + * \sa outerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return derived().innerStride(); + } + + /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns + * in a column-major matrix). + * + * \sa innerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return derived().outerStride(); + } + + // FIXME shall we remove it ? + inline Index stride() const + { + return Derived::IsVectorAtCompileTime ? innerStride() : outerStride(); + } + + /** \returns the pointer increment between two consecutive rows. + * + * \sa innerStride(), outerStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index rowStride() const + { + return Derived::IsRowMajor ? outerStride() : innerStride(); + } + + /** \returns the pointer increment between two consecutive columns. + * + * \sa innerStride(), outerStride(), rowStride() + */ + EIGEN_DEVICE_FUNC + inline Index colStride() const + { + return Derived::IsRowMajor ? innerStride() : outerStride(); + } +}; + +/** \brief Base class providing direct read/write coefficient access to matrices and arrays. + * \ingroup Core_Module + * \tparam Derived Type of the derived class + * \tparam #DirectWriteAccessors Constant indicating direct access + * + * This class defines functions to work with strides which can be used to access entries directly. This class + * inherits DenseCoeffsBase which defines functions to access entries read/write using + * \c operator(). + * + * \sa \blank \ref TopicClassHierarchy + */ +template +class DenseCoeffsBase + : public DenseCoeffsBase +{ + public: + + typedef DenseCoeffsBase Base; + typedef typename internal::traits::Scalar Scalar; + typedef typename NumTraits::Real RealScalar; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::derived; + + /** \returns the pointer increment between two consecutive elements within a slice in the inner direction. + * + * \sa outerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return derived().innerStride(); + } + + /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns + * in a column-major matrix). + * + * \sa innerStride(), rowStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return derived().outerStride(); + } + + // FIXME shall we remove it ? + inline Index stride() const + { + return Derived::IsVectorAtCompileTime ? innerStride() : outerStride(); + } + + /** \returns the pointer increment between two consecutive rows. + * + * \sa innerStride(), outerStride(), colStride() + */ + EIGEN_DEVICE_FUNC + inline Index rowStride() const + { + return Derived::IsRowMajor ? outerStride() : innerStride(); + } + + /** \returns the pointer increment between two consecutive columns. + * + * \sa innerStride(), outerStride(), rowStride() + */ + EIGEN_DEVICE_FUNC + inline Index colStride() const + { + return Derived::IsRowMajor ? innerStride() : outerStride(); + } +}; + +namespace internal { + +template +struct first_aligned_impl +{ + static inline Index run(const Derived&) + { return 0; } +}; + +template +struct first_aligned_impl +{ + static inline Index run(const Derived& m) + { + return internal::first_aligned(m.data(), m.size()); + } +}; + +/** \internal \returns the index of the first element of the array stored by \a m that is properly aligned with respect to \a Alignment for vectorization. + * + * \tparam Alignment requested alignment in Bytes. + * + * There is also the variant first_aligned(const Scalar*, Integer) defined in Memory.h. See it for more + * documentation. + */ +template +static inline Index first_aligned(const DenseBase& m) +{ + enum { ReturnZero = (int(evaluator::Alignment) >= Alignment) || !(Derived::Flags & DirectAccessBit) }; + return first_aligned_impl::run(m.derived()); +} + +template +static inline Index first_default_aligned(const DenseBase& m) +{ + typedef typename Derived::Scalar Scalar; + typedef typename packet_traits::type DefaultPacketType; + return internal::first_aligned::alignment),Derived>(m); +} + +template::ret> +struct inner_stride_at_compile_time +{ + enum { ret = traits::InnerStrideAtCompileTime }; +}; + +template +struct inner_stride_at_compile_time +{ + enum { ret = 0 }; +}; + +template::ret> +struct outer_stride_at_compile_time +{ + enum { ret = traits::OuterStrideAtCompileTime }; +}; + +template +struct outer_stride_at_compile_time +{ + enum { ret = 0 }; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DENSECOEFFSBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h new file mode 100644 index 00000000000..7d6d4e66d4b --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DenseStorage.h @@ -0,0 +1,570 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2009 Benoit Jacob +// Copyright (C) 2010-2013 Hauke Heibel +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIXSTORAGE_H +#define EIGEN_MATRIXSTORAGE_H + +#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) X; EIGEN_DENSE_STORAGE_CTOR_PLUGIN; +#else + #define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) +#endif + +namespace Eigen { + +namespace internal { + +struct constructor_without_unaligned_array_assert {}; + +template +EIGEN_DEVICE_FUNC +void check_static_allocation_size() +{ + // if EIGEN_STACK_ALLOCATION_LIMIT is defined to 0, then no limit + #if EIGEN_STACK_ALLOCATION_LIMIT + EIGEN_STATIC_ASSERT(Size * sizeof(T) <= EIGEN_STACK_ALLOCATION_LIMIT, OBJECT_ALLOCATED_ON_STACK_IS_TOO_BIG); + #endif +} + +/** \internal + * Static array. If the MatrixOrArrayOptions require auto-alignment, the array will be automatically aligned: + * to 16 bytes boundary if the total size is a multiple of 16 bytes. + */ +template ::value > +struct plain_array +{ + T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +#if defined(EIGEN_DISABLE_UNALIGNED_ARRAY_ASSERT) + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) +#elif EIGEN_GNUC_AT_LEAST(4,7) + // GCC 4.7 is too aggressive in its optimizations and remove the alignement test based on the fact the array is declared to be aligned. + // See this bug report: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53900 + // Hiding the origin of the array pointer behind a function argument seems to do the trick even if the function is inlined: + template + EIGEN_ALWAYS_INLINE PtrType eigen_unaligned_array_assert_workaround_gcc47(PtrType array) { return array; } + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \ + eigen_assert((internal::UIntPtr(eigen_unaligned_array_assert_workaround_gcc47(array)) & (sizemask)) == 0 \ + && "this assertion is explained here: " \ + "http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \ + " **** READ THIS WEB PAGE !!! ****"); +#else + #define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \ + eigen_assert((internal::UIntPtr(array) & (sizemask)) == 0 \ + && "this assertion is explained here: " \ + "http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \ + " **** READ THIS WEB PAGE !!! ****"); +#endif + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(8) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(7); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(16) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(15); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(32) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(31); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + EIGEN_ALIGN_TO_BOUNDARY(64) T array[Size]; + + EIGEN_DEVICE_FUNC + plain_array() + { + EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(63); + check_static_allocation_size(); + } + + EIGEN_DEVICE_FUNC + plain_array(constructor_without_unaligned_array_assert) + { + check_static_allocation_size(); + } +}; + +template +struct plain_array +{ + T array[1]; + EIGEN_DEVICE_FUNC plain_array() {} + EIGEN_DEVICE_FUNC plain_array(constructor_without_unaligned_array_assert) {} +}; + +} // end namespace internal + +/** \internal + * + * \class DenseStorage + * \ingroup Core_Module + * + * \brief Stores the data of a matrix + * + * This class stores the data of fixed-size, dynamic-size or mixed matrices + * in a way as compact as possible. + * + * \sa Matrix + */ +template class DenseStorage; + +// purely fixed-size matrix +template class DenseStorage +{ + internal::plain_array m_data; + public: + EIGEN_DEVICE_FUNC DenseStorage() { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size) + } + EIGEN_DEVICE_FUNC + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()) {} + EIGEN_DEVICE_FUNC + DenseStorage(const DenseStorage& other) : m_data(other.m_data) { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size) + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) m_data = other.m_data; + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows==_Rows && cols==_Cols); + EIGEN_UNUSED_VARIABLE(size); + EIGEN_UNUSED_VARIABLE(rows); + EIGEN_UNUSED_VARIABLE(cols); + } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); } + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// null matrix +template class DenseStorage +{ + public: + EIGEN_DEVICE_FUNC DenseStorage() {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage&) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage&) { return *this; } + EIGEN_DEVICE_FUNC DenseStorage(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& ) {} + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {} + EIGEN_DEVICE_FUNC const T *data() const { return 0; } + EIGEN_DEVICE_FUNC T *data() { return 0; } +}; + +// more specializations for null matrices; these are necessary to resolve ambiguities +template class DenseStorage +: public DenseStorage { }; + +template class DenseStorage +: public DenseStorage { }; + +template class DenseStorage +: public DenseStorage { }; + +// dynamic-size matrix with fixed-size storage +template class DenseStorage +{ + internal::plain_array m_data; + Index m_rows; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows), m_cols(other.m_cols) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_rows = other.m_rows; + m_cols = other.m_cols; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index cols) : m_rows(rows), m_cols(cols) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) + { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows() const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols() const {return m_cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; } + EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// dynamic-size matrix with fixed-size storage and fixed width +template class DenseStorage +{ + internal::plain_array m_data; + Index m_rows; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_rows = other.m_rows; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index) : m_rows(rows) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return _Cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index) { m_rows = rows; } + EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index) { m_rows = rows; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// dynamic-size matrix with fixed-size storage and fixed height +template class DenseStorage +{ + internal::plain_array m_data; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(internal::constructor_without_unaligned_array_assert()), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_cols(other.m_cols) {} + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + m_data = other.m_data; + m_cols = other.m_cols; + } + return *this; + } + EIGEN_DEVICE_FUNC DenseStorage(Index, Index, Index cols) : m_cols(cols) {} + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows(void) const {return _Rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + void conservativeResize(Index, Index, Index cols) { m_cols = cols; } + void resize(Index, Index, Index cols) { m_cols = cols; } + EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; } + EIGEN_DEVICE_FUNC T *data() { return m_data.array; } +}; + +// purely dynamic matrix. +template class DenseStorage +{ + T *m_data; + Index m_rows; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) + : m_data(0), m_rows(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) + : m_data(internal::conditional_aligned_new_auto(size)), m_rows(rows), m_cols(cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows>=0 && cols >=0); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(other.m_rows*other.m_cols)) + , m_rows(other.m_rows) + , m_cols(other.m_cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*m_cols) + internal::smart_copy(other.m_data, other.m_data+other.m_rows*other.m_cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_rows(std::move(other.m_rows)) + , m_cols(std::move(other.m_cols)) + { + other.m_data = nullptr; + other.m_rows = 0; + other.m_cols = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_rows, other.m_rows); + swap(m_cols, other.m_cols); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, m_rows*m_cols); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) + { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + void conservativeResize(Index size, Index rows, Index cols) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, m_rows*m_cols); + m_rows = rows; + m_cols = cols; + } + EIGEN_DEVICE_FUNC void resize(Index size, Index rows, Index cols) + { + if(size != m_rows*m_cols) + { + internal::conditional_aligned_delete_auto(m_data, m_rows*m_cols); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_rows = rows; + m_cols = cols; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +// matrix with dynamic width and fixed height (so that matrix has dynamic size). +template class DenseStorage +{ + T *m_data; + Index m_cols; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_cols(0) {} + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto(size)), m_cols(cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows==_Rows && cols >=0); + EIGEN_UNUSED_VARIABLE(rows); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(_Rows*other.m_cols)) + , m_cols(other.m_cols) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_cols*_Rows) + internal::smart_copy(other.m_data, other.m_data+_Rows*m_cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_cols(std::move(other.m_cols)) + { + other.m_data = nullptr; + other.m_cols = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_cols, other.m_cols); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, _Rows*m_cols); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); } + EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;} + EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;} + EIGEN_DEVICE_FUNC void conservativeResize(Index size, Index, Index cols) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, _Rows*m_cols); + m_cols = cols; + } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index, Index cols) + { + if(size != _Rows*m_cols) + { + internal::conditional_aligned_delete_auto(m_data, _Rows*m_cols); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_cols = cols; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +// matrix with dynamic height and fixed width (so that matrix has dynamic size). +template class DenseStorage +{ + T *m_data; + Index m_rows; + public: + EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0) {} + explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {} + EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto(size)), m_rows(rows) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + eigen_internal_assert(size==rows*cols && rows>=0 && cols == _Cols); + EIGEN_UNUSED_VARIABLE(cols); + } + EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) + : m_data(internal::conditional_aligned_new_auto(other.m_rows*_Cols)) + , m_rows(other.m_rows) + { + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*_Cols) + internal::smart_copy(other.m_data, other.m_data+other.m_rows*_Cols, m_data); + } + EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other) + { + if (this != &other) + { + DenseStorage tmp(other); + this->swap(tmp); + } + return *this; + } +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT + : m_data(std::move(other.m_data)) + , m_rows(std::move(other.m_rows)) + { + other.m_data = nullptr; + other.m_rows = 0; + } + EIGEN_DEVICE_FUNC + DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_data, other.m_data); + swap(m_rows, other.m_rows); + return *this; + } +#endif + EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto(m_data, _Cols*m_rows); } + EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); } + EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;} + EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;} + void conservativeResize(Index size, Index rows, Index) + { + m_data = internal::conditional_aligned_realloc_new_auto(m_data, size, m_rows*_Cols); + m_rows = rows; + } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index rows, Index) + { + if(size != m_rows*_Cols) + { + internal::conditional_aligned_delete_auto(m_data, _Cols*m_rows); + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + m_data = internal::conditional_aligned_new_auto(size); + else + m_data = 0; + EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({}) + } + m_rows = rows; + } + EIGEN_DEVICE_FUNC const T *data() const { return m_data; } + EIGEN_DEVICE_FUNC T *data() { return m_data; } +}; + +} // end namespace Eigen + +#endif // EIGEN_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h new file mode 100644 index 00000000000..afcaf357567 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Diagonal.h @@ -0,0 +1,260 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2009 Benoit Jacob +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONAL_H +#define EIGEN_DIAGONAL_H + +namespace Eigen { + +/** \class Diagonal + * \ingroup Core_Module + * + * \brief Expression of a diagonal/subdiagonal/superdiagonal in a matrix + * + * \param MatrixType the type of the object in which we are taking a sub/main/super diagonal + * \param DiagIndex the index of the sub/super diagonal. The default is 0 and it means the main diagonal. + * A positive value means a superdiagonal, a negative value means a subdiagonal. + * You can also use DynamicIndex so the index can be set at runtime. + * + * The matrix is not required to be square. + * + * This class represents an expression of the main diagonal, or any sub/super diagonal + * of a square matrix. It is the return type of MatrixBase::diagonal() and MatrixBase::diagonal(Index) and most of the + * time this is the only way it is used. + * + * \sa MatrixBase::diagonal(), MatrixBase::diagonal(Index) + */ + +namespace internal { +template +struct traits > + : traits +{ + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + typedef typename MatrixType::StorageKind StorageKind; + enum { + RowsAtCompileTime = (int(DiagIndex) == DynamicIndex || int(MatrixType::SizeAtCompileTime) == Dynamic) ? Dynamic + : (EIGEN_PLAIN_ENUM_MIN(MatrixType::RowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0), + MatrixType::ColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))), + ColsAtCompileTime = 1, + MaxRowsAtCompileTime = int(MatrixType::MaxSizeAtCompileTime) == Dynamic ? Dynamic + : DiagIndex == DynamicIndex ? EIGEN_SIZE_MIN_PREFER_FIXED(MatrixType::MaxRowsAtCompileTime, + MatrixType::MaxColsAtCompileTime) + : (EIGEN_PLAIN_ENUM_MIN(MatrixType::MaxRowsAtCompileTime - EIGEN_PLAIN_ENUM_MAX(-DiagIndex, 0), + MatrixType::MaxColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))), + MaxColsAtCompileTime = 1, + MaskLvalueBit = is_lvalue::value ? LvalueBit : 0, + Flags = (unsigned int)_MatrixTypeNested::Flags & (RowMajorBit | MaskLvalueBit | DirectAccessBit) & ~RowMajorBit, // FIXME DirectAccessBit should not be handled by expressions + MatrixTypeOuterStride = outer_stride_at_compile_time::ret, + InnerStrideAtCompileTime = MatrixTypeOuterStride == Dynamic ? Dynamic : MatrixTypeOuterStride+1, + OuterStrideAtCompileTime = 0 + }; +}; +} + +template class Diagonal + : public internal::dense_xpr_base< Diagonal >::type +{ + public: + + enum { DiagIndex = _DiagIndex }; + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Diagonal) + + EIGEN_DEVICE_FUNC + explicit inline Diagonal(MatrixType& matrix, Index a_index = DiagIndex) : m_matrix(matrix), m_index(a_index) + { + eigen_assert( a_index <= m_matrix.cols() && -a_index <= m_matrix.rows() ); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Diagonal) + + EIGEN_DEVICE_FUNC + inline Index rows() const + { + return m_index.value()<0 ? numext::mini(m_matrix.cols(),m_matrix.rows()+m_index.value()) + : numext::mini(m_matrix.rows(),m_matrix.cols()-m_index.value()); + } + + EIGEN_DEVICE_FUNC + inline Index cols() const { return 1; } + + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return m_matrix.outerStride() + 1; + } + + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return 0; + } + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return &(m_matrix.coeffRef(rowOffset(), colOffset())); } + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return &(m_matrix.coeffRef(rowOffset(), colOffset())); } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index row, Index) + { + EIGEN_STATIC_ASSERT_LVALUE(MatrixType) + return m_matrix.coeffRef(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index row, Index) const + { + return m_matrix.coeffRef(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline CoeffReturnType coeff(Index row, Index) const + { + return m_matrix.coeff(row+rowOffset(), row+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline Scalar& coeffRef(Index idx) + { + EIGEN_STATIC_ASSERT_LVALUE(MatrixType) + return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index idx) const + { + return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline CoeffReturnType coeff(Index idx) const + { + return m_matrix.coeff(idx+rowOffset(), idx+colOffset()); + } + + EIGEN_DEVICE_FUNC + inline const typename internal::remove_all::type& + nestedExpression() const + { + return m_matrix; + } + + EIGEN_DEVICE_FUNC + inline Index index() const + { + return m_index.value(); + } + + protected: + typename internal::ref_selector::non_const_type m_matrix; + const internal::variable_if_dynamicindex m_index; + + private: + // some compilers may fail to optimize std::max etc in case of compile-time constants... + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index absDiagIndex() const { return m_index.value()>0 ? m_index.value() : -m_index.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rowOffset() const { return m_index.value()>0 ? 0 : -m_index.value(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index colOffset() const { return m_index.value()>0 ? m_index.value() : 0; } + // trigger a compile-time error if someone try to call packet + template typename MatrixType::PacketReturnType packet(Index) const; + template typename MatrixType::PacketReturnType packet(Index,Index) const; +}; + +/** \returns an expression of the main diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * Example: \include MatrixBase_diagonal.cpp + * Output: \verbinclude MatrixBase_diagonal.out + * + * \sa class Diagonal */ +template +inline typename MatrixBase::DiagonalReturnType +MatrixBase::diagonal() +{ + return DiagonalReturnType(derived()); +} + +/** This is the const version of diagonal(). */ +template +inline typename MatrixBase::ConstDiagonalReturnType +MatrixBase::diagonal() const +{ + return ConstDiagonalReturnType(derived()); +} + +/** \returns an expression of the \a DiagIndex-th sub or super diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * The template parameter \a DiagIndex represent a super diagonal if \a DiagIndex > 0 + * and a sub diagonal otherwise. \a DiagIndex == 0 is equivalent to the main diagonal. + * + * Example: \include MatrixBase_diagonal_int.cpp + * Output: \verbinclude MatrixBase_diagonal_int.out + * + * \sa MatrixBase::diagonal(), class Diagonal */ +template +inline typename MatrixBase::DiagonalDynamicIndexReturnType +MatrixBase::diagonal(Index index) +{ + return DiagonalDynamicIndexReturnType(derived(), index); +} + +/** This is the const version of diagonal(Index). */ +template +inline typename MatrixBase::ConstDiagonalDynamicIndexReturnType +MatrixBase::diagonal(Index index) const +{ + return ConstDiagonalDynamicIndexReturnType(derived(), index); +} + +/** \returns an expression of the \a DiagIndex-th sub or super diagonal of the matrix \c *this + * + * \c *this is not required to be square. + * + * The template parameter \a DiagIndex represent a super diagonal if \a DiagIndex > 0 + * and a sub diagonal otherwise. \a DiagIndex == 0 is equivalent to the main diagonal. + * + * Example: \include MatrixBase_diagonal_template_int.cpp + * Output: \verbinclude MatrixBase_diagonal_template_int.out + * + * \sa MatrixBase::diagonal(), class Diagonal */ +template +template +inline typename MatrixBase::template DiagonalIndexReturnType::Type +MatrixBase::diagonal() +{ + return typename DiagonalIndexReturnType::Type(derived()); +} + +/** This is the const version of diagonal(). */ +template +template +inline typename MatrixBase::template ConstDiagonalIndexReturnType::Type +MatrixBase::diagonal() const +{ + return typename ConstDiagonalIndexReturnType::Type(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_DIAGONAL_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h new file mode 100644 index 00000000000..ecfdce8efa1 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalMatrix.h @@ -0,0 +1,343 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// Copyright (C) 2007-2009 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONALMATRIX_H +#define EIGEN_DIAGONALMATRIX_H + +namespace Eigen { + +#ifndef EIGEN_PARSED_BY_DOXYGEN +template +class DiagonalBase : public EigenBase +{ + public: + typedef typename internal::traits::DiagonalVectorType DiagonalVectorType; + typedef typename DiagonalVectorType::Scalar Scalar; + typedef typename DiagonalVectorType::RealScalar RealScalar; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + + enum { + RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + ColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + IsVectorAtCompileTime = 0, + Flags = NoPreferredStorageOrderBit + }; + + typedef Matrix DenseMatrixType; + typedef DenseMatrixType DenseType; + typedef DiagonalMatrix PlainObject; + + EIGEN_DEVICE_FUNC + inline const Derived& derived() const { return *static_cast(this); } + EIGEN_DEVICE_FUNC + inline Derived& derived() { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + DenseMatrixType toDenseMatrix() const { return derived(); } + + EIGEN_DEVICE_FUNC + inline const DiagonalVectorType& diagonal() const { return derived().diagonal(); } + EIGEN_DEVICE_FUNC + inline DiagonalVectorType& diagonal() { return derived().diagonal(); } + + EIGEN_DEVICE_FUNC + inline Index rows() const { return diagonal().size(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return diagonal().size(); } + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const MatrixBase &matrix) const + { + return Product(derived(),matrix.derived()); + } + + typedef DiagonalWrapper, const DiagonalVectorType> > InverseReturnType; + EIGEN_DEVICE_FUNC + inline const InverseReturnType + inverse() const + { + return InverseReturnType(diagonal().cwiseInverse()); + } + + EIGEN_DEVICE_FUNC + inline const DiagonalWrapper + operator*(const Scalar& scalar) const + { + return DiagonalWrapper(diagonal() * scalar); + } + EIGEN_DEVICE_FUNC + friend inline const DiagonalWrapper + operator*(const Scalar& scalar, const DiagonalBase& other) + { + return DiagonalWrapper(scalar * other.diagonal()); + } +}; + +#endif + +/** \class DiagonalMatrix + * \ingroup Core_Module + * + * \brief Represents a diagonal matrix with its storage + * + * \param _Scalar the type of coefficients + * \param SizeAtCompileTime the dimension of the matrix, or Dynamic + * \param MaxSizeAtCompileTime the dimension of the matrix, or Dynamic. This parameter is optional and defaults + * to SizeAtCompileTime. Most of the time, you do not need to specify it. + * + * \sa class DiagonalWrapper + */ + +namespace internal { +template +struct traits > + : traits > +{ + typedef Matrix<_Scalar,SizeAtCompileTime,1,0,MaxSizeAtCompileTime,1> DiagonalVectorType; + typedef DiagonalShape StorageKind; + enum { + Flags = LvalueBit | NoPreferredStorageOrderBit + }; +}; +} +template +class DiagonalMatrix + : public DiagonalBase > +{ + public: + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename internal::traits::DiagonalVectorType DiagonalVectorType; + typedef const DiagonalMatrix& Nested; + typedef _Scalar Scalar; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + #endif + + protected: + + DiagonalVectorType m_diagonal; + + public: + + /** const version of diagonal(). */ + EIGEN_DEVICE_FUNC + inline const DiagonalVectorType& diagonal() const { return m_diagonal; } + /** \returns a reference to the stored vector of diagonal coefficients. */ + EIGEN_DEVICE_FUNC + inline DiagonalVectorType& diagonal() { return m_diagonal; } + + /** Default constructor without initialization */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix() {} + + /** Constructs a diagonal matrix with given dimension */ + EIGEN_DEVICE_FUNC + explicit inline DiagonalMatrix(Index dim) : m_diagonal(dim) {} + + /** 2D constructor. */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const Scalar& x, const Scalar& y) : m_diagonal(x,y) {} + + /** 3D constructor. */ + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const Scalar& x, const Scalar& y, const Scalar& z) : m_diagonal(x,y,z) {} + + /** Copy constructor. */ + template + EIGEN_DEVICE_FUNC + inline DiagonalMatrix(const DiagonalBase& other) : m_diagonal(other.diagonal()) {} + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** copy constructor. prevent a default copy constructor from hiding the other templated constructor */ + inline DiagonalMatrix(const DiagonalMatrix& other) : m_diagonal(other.diagonal()) {} + #endif + + /** generic constructor from expression of the diagonal coefficients */ + template + EIGEN_DEVICE_FUNC + explicit inline DiagonalMatrix(const MatrixBase& other) : m_diagonal(other) + {} + + /** Copy operator. */ + template + EIGEN_DEVICE_FUNC + DiagonalMatrix& operator=(const DiagonalBase& other) + { + m_diagonal = other.diagonal(); + return *this; + } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + DiagonalMatrix& operator=(const DiagonalMatrix& other) + { + m_diagonal = other.diagonal(); + return *this; + } + #endif + + /** Resizes to given size. */ + EIGEN_DEVICE_FUNC + inline void resize(Index size) { m_diagonal.resize(size); } + /** Sets all coefficients to zero. */ + EIGEN_DEVICE_FUNC + inline void setZero() { m_diagonal.setZero(); } + /** Resizes and sets all coefficients to zero. */ + EIGEN_DEVICE_FUNC + inline void setZero(Index size) { m_diagonal.setZero(size); } + /** Sets this matrix to be the identity matrix of the current size. */ + EIGEN_DEVICE_FUNC + inline void setIdentity() { m_diagonal.setOnes(); } + /** Sets this matrix to be the identity matrix of the given size. */ + EIGEN_DEVICE_FUNC + inline void setIdentity(Index size) { m_diagonal.setOnes(size); } +}; + +/** \class DiagonalWrapper + * \ingroup Core_Module + * + * \brief Expression of a diagonal matrix + * + * \param _DiagonalVectorType the type of the vector of diagonal coefficients + * + * This class is an expression of a diagonal matrix, but not storing its own vector of diagonal coefficients, + * instead wrapping an existing vector expression. It is the return type of MatrixBase::asDiagonal() + * and most of the time this is the only way that it is used. + * + * \sa class DiagonalMatrix, class DiagonalBase, MatrixBase::asDiagonal() + */ + +namespace internal { +template +struct traits > +{ + typedef _DiagonalVectorType DiagonalVectorType; + typedef typename DiagonalVectorType::Scalar Scalar; + typedef typename DiagonalVectorType::StorageIndex StorageIndex; + typedef DiagonalShape StorageKind; + typedef typename traits::XprKind XprKind; + enum { + RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + ColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime, + MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime, + Flags = (traits::Flags & LvalueBit) | NoPreferredStorageOrderBit + }; +}; +} + +template +class DiagonalWrapper + : public DiagonalBase >, internal::no_assignment_operator +{ + public: + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef _DiagonalVectorType DiagonalVectorType; + typedef DiagonalWrapper Nested; + #endif + + /** Constructor from expression of diagonal coefficients to wrap. */ + EIGEN_DEVICE_FUNC + explicit inline DiagonalWrapper(DiagonalVectorType& a_diagonal) : m_diagonal(a_diagonal) {} + + /** \returns a const reference to the wrapped expression of diagonal coefficients. */ + EIGEN_DEVICE_FUNC + const DiagonalVectorType& diagonal() const { return m_diagonal; } + + protected: + typename DiagonalVectorType::Nested m_diagonal; +}; + +/** \returns a pseudo-expression of a diagonal matrix with *this as vector of diagonal coefficients + * + * \only_for_vectors + * + * Example: \include MatrixBase_asDiagonal.cpp + * Output: \verbinclude MatrixBase_asDiagonal.out + * + * \sa class DiagonalWrapper, class DiagonalMatrix, diagonal(), isDiagonal() + **/ +template +inline const DiagonalWrapper +MatrixBase::asDiagonal() const +{ + return DiagonalWrapper(derived()); +} + +/** \returns true if *this is approximately equal to a diagonal matrix, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isDiagonal.cpp + * Output: \verbinclude MatrixBase_isDiagonal.out + * + * \sa asDiagonal() + */ +template +bool MatrixBase::isDiagonal(const RealScalar& prec) const +{ + if(cols() != rows()) return false; + RealScalar maxAbsOnDiagonal = static_cast(-1); + for(Index j = 0; j < cols(); ++j) + { + RealScalar absOnDiagonal = numext::abs(coeff(j,j)); + if(absOnDiagonal > maxAbsOnDiagonal) maxAbsOnDiagonal = absOnDiagonal; + } + for(Index j = 0; j < cols(); ++j) + for(Index i = 0; i < j; ++i) + { + if(!internal::isMuchSmallerThan(coeff(i, j), maxAbsOnDiagonal, prec)) return false; + if(!internal::isMuchSmallerThan(coeff(j, i), maxAbsOnDiagonal, prec)) return false; + } + return true; +} + +namespace internal { + +template<> struct storage_kind_to_shape { typedef DiagonalShape Shape; }; + +struct Diagonal2Dense {}; + +template<> struct AssignmentKind { typedef Diagonal2Dense Kind; }; + +// Diagonal matrix to Dense assignment +template< typename DstXprType, typename SrcXprType, typename Functor> +struct Assignment +{ + static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &/*func*/) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + + dst.setZero(); + dst.diagonal() = src.diagonal(); + } + + static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &/*func*/) + { dst.diagonal() += src.diagonal(); } + + static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &/*func*/) + { dst.diagonal() -= src.diagonal(); } +}; + +} // namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DIAGONALMATRIX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h new file mode 100644 index 00000000000..d372b938f65 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/DiagonalProduct.h @@ -0,0 +1,28 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2007-2009 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DIAGONALPRODUCT_H +#define EIGEN_DIAGONALPRODUCT_H + +namespace Eigen { + +/** \returns the diagonal matrix product of \c *this by the diagonal matrix \a diagonal. + */ +template +template +inline const Product +MatrixBase::operator*(const DiagonalBase &a_diagonal) const +{ + return Product(derived(),a_diagonal.derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_DIAGONALPRODUCT_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h new file mode 100644 index 00000000000..1fe7a84a48d --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Dot.h @@ -0,0 +1,318 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008, 2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DOT_H +#define EIGEN_DOT_H + +namespace Eigen { + +namespace internal { + +// helper function for dot(). The problem is that if we put that in the body of dot(), then upon calling dot +// with mismatched types, the compiler emits errors about failing to instantiate cwiseProduct BEFORE +// looking at the static assertions. Thus this is a trick to get better compile errors. +template +struct dot_nocheck +{ + typedef scalar_conj_product_op::Scalar,typename traits::Scalar> conj_prod; + typedef typename conj_prod::result_type ResScalar; + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE + static ResScalar run(const MatrixBase& a, const MatrixBase& b) + { + return a.template binaryExpr(b).sum(); + } +}; + +template +struct dot_nocheck +{ + typedef scalar_conj_product_op::Scalar,typename traits::Scalar> conj_prod; + typedef typename conj_prod::result_type ResScalar; + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE + static ResScalar run(const MatrixBase& a, const MatrixBase& b) + { + return a.transpose().template binaryExpr(b).sum(); + } +}; + +} // end namespace internal + +/** \fn MatrixBase::dot + * \returns the dot product of *this with other. + * + * \only_for_vectors + * + * \note If the scalar type is complex numbers, then this function returns the hermitian + * (sesquilinear) dot product, conjugate-linear in the first variable and linear in the + * second variable. + * + * \sa squaredNorm(), norm() + */ +template +template +EIGEN_DEVICE_FUNC +EIGEN_STRONG_INLINE +typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType +MatrixBase::dot(const MatrixBase& other) const +{ + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) + EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived) +#if !(defined(EIGEN_NO_STATIC_ASSERT) && defined(EIGEN_NO_DEBUG)) + typedef internal::scalar_conj_product_op func; + EIGEN_CHECK_BINARY_COMPATIBILIY(func,Scalar,typename OtherDerived::Scalar); +#endif + + eigen_assert(size() == other.size()); + + return internal::dot_nocheck::run(*this, other); +} + +//---------- implementation of L2 norm and related functions ---------- + +/** \returns, for vectors, the squared \em l2 norm of \c *this, and for matrices the Frobenius norm. + * In both cases, it consists in the sum of the square of all the matrix entries. + * For vectors, this is also equals to the dot product of \c *this with itself. + * + * \sa dot(), norm(), lpNorm() + */ +template +EIGEN_STRONG_INLINE typename NumTraits::Scalar>::Real MatrixBase::squaredNorm() const +{ + return numext::real((*this).cwiseAbs2().sum()); +} + +/** \returns, for vectors, the \em l2 norm of \c *this, and for matrices the Frobenius norm. + * In both cases, it consists in the square root of the sum of the square of all the matrix entries. + * For vectors, this is also equals to the square root of the dot product of \c *this with itself. + * + * \sa lpNorm(), dot(), squaredNorm() + */ +template +EIGEN_STRONG_INLINE typename NumTraits::Scalar>::Real MatrixBase::norm() const +{ + return numext::sqrt(squaredNorm()); +} + +/** \returns an expression of the quotient of \c *this by its own norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), + * then this function returns a copy of the input. + * + * \only_for_vectors + * + * \sa norm(), normalize() + */ +template +EIGEN_STRONG_INLINE const typename MatrixBase::PlainObject +MatrixBase::normalized() const +{ + typedef typename internal::nested_eval::type _Nested; + _Nested n(derived()); + RealScalar z = n.squaredNorm(); + // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU + if(z>RealScalar(0)) + return n / numext::sqrt(z); + else + return n; +} + +/** Normalizes the vector, i.e. divides it by its own norm. + * + * \only_for_vectors + * + * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged. + * + * \sa norm(), normalized() + */ +template +EIGEN_STRONG_INLINE void MatrixBase::normalize() +{ + RealScalar z = squaredNorm(); + // NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU + if(z>RealScalar(0)) + derived() /= numext::sqrt(z); +} + +/** \returns an expression of the quotient of \c *this by its own norm while avoiding underflow and overflow. + * + * \only_for_vectors + * + * This method is analogue to the normalized() method, but it reduces the risk of + * underflow and overflow when computing the norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), + * then this function returns a copy of the input. + * + * \sa stableNorm(), stableNormalize(), normalized() + */ +template +EIGEN_STRONG_INLINE const typename MatrixBase::PlainObject +MatrixBase::stableNormalized() const +{ + typedef typename internal::nested_eval::type _Nested; + _Nested n(derived()); + RealScalar w = n.cwiseAbs().maxCoeff(); + RealScalar z = (n/w).squaredNorm(); + if(z>RealScalar(0)) + return n / (numext::sqrt(z)*w); + else + return n; +} + +/** Normalizes the vector while avoid underflow and overflow + * + * \only_for_vectors + * + * This method is analogue to the normalize() method, but it reduces the risk of + * underflow and overflow when computing the norm. + * + * \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged. + * + * \sa stableNorm(), stableNormalized(), normalize() + */ +template +EIGEN_STRONG_INLINE void MatrixBase::stableNormalize() +{ + RealScalar w = cwiseAbs().maxCoeff(); + RealScalar z = (derived()/w).squaredNorm(); + if(z>RealScalar(0)) + derived() /= numext::sqrt(z)*w; +} + +//---------- implementation of other norms ---------- + +namespace internal { + +template +struct lpNorm_selector +{ + typedef typename NumTraits::Scalar>::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const MatrixBase& m) + { + EIGEN_USING_STD_MATH(pow) + return pow(m.cwiseAbs().array().pow(p).sum(), RealScalar(1)/p); + } +}; + +template +struct lpNorm_selector +{ + EIGEN_DEVICE_FUNC + static inline typename NumTraits::Scalar>::Real run(const MatrixBase& m) + { + return m.cwiseAbs().sum(); + } +}; + +template +struct lpNorm_selector +{ + EIGEN_DEVICE_FUNC + static inline typename NumTraits::Scalar>::Real run(const MatrixBase& m) + { + return m.norm(); + } +}; + +template +struct lpNorm_selector +{ + typedef typename NumTraits::Scalar>::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const MatrixBase& m) + { + if(Derived::SizeAtCompileTime==0 || (Derived::SizeAtCompileTime==Dynamic && m.size()==0)) + return RealScalar(0); + return m.cwiseAbs().maxCoeff(); + } +}; + +} // end namespace internal + +/** \returns the \b coefficient-wise \f$ \ell^p \f$ norm of \c *this, that is, returns the p-th root of the sum of the p-th powers of the absolute values + * of the coefficients of \c *this. If \a p is the special value \a Eigen::Infinity, this function returns the \f$ \ell^\infty \f$ + * norm, that is the maximum of the absolute values of the coefficients of \c *this. + * + * In all cases, if \c *this is empty, then the value 0 is returned. + * + * \note For matrices, this function does not compute the operator-norm. That is, if \c *this is a matrix, then its coefficients are interpreted as a 1D vector. Nonetheless, you can easily compute the 1-norm and \f$\infty\f$-norm matrix operator norms using \link TutorialReductionsVisitorsBroadcastingReductionsNorm partial reductions \endlink. + * + * \sa norm() + */ +template +template +#ifndef EIGEN_PARSED_BY_DOXYGEN +inline typename NumTraits::Scalar>::Real +#else +MatrixBase::RealScalar +#endif +MatrixBase::lpNorm() const +{ + return internal::lpNorm_selector::run(*this); +} + +//---------- implementation of isOrthogonal / isUnitary ---------- + +/** \returns true if *this is approximately orthogonal to \a other, + * within the precision given by \a prec. + * + * Example: \include MatrixBase_isOrthogonal.cpp + * Output: \verbinclude MatrixBase_isOrthogonal.out + */ +template +template +bool MatrixBase::isOrthogonal +(const MatrixBase& other, const RealScalar& prec) const +{ + typename internal::nested_eval::type nested(derived()); + typename internal::nested_eval::type otherNested(other.derived()); + return numext::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm(); +} + +/** \returns true if *this is approximately an unitary matrix, + * within the precision given by \a prec. In the case where the \a Scalar + * type is real numbers, a unitary matrix is an orthogonal matrix, whence the name. + * + * \note This can be used to check whether a family of vectors forms an orthonormal basis. + * Indeed, \c m.isUnitary() returns true if and only if the columns (equivalently, the rows) of m form an + * orthonormal basis. + * + * Example: \include MatrixBase_isUnitary.cpp + * Output: \verbinclude MatrixBase_isUnitary.out + */ +template +bool MatrixBase::isUnitary(const RealScalar& prec) const +{ + typename internal::nested_eval::type self(derived()); + for(Index i = 0; i < cols(); ++i) + { + if(!internal::isApprox(self.col(i).squaredNorm(), static_cast(1), prec)) + return false; + for(Index j = 0; j < i; ++j) + if(!internal::isMuchSmallerThan(self.col(i).dot(self.col(j)), static_cast(1), prec)) + return false; + } + return true; +} + +} // end namespace Eigen + +#endif // EIGEN_DOT_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h new file mode 100644 index 00000000000..b195506a91e --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/EigenBase.h @@ -0,0 +1,159 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Benoit Jacob +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_EIGENBASE_H +#define EIGEN_EIGENBASE_H + +namespace Eigen { + +/** \class EigenBase + * \ingroup Core_Module + * + * Common base class for all classes T such that MatrixBase has an operator=(T) and a constructor MatrixBase(T). + * + * In other words, an EigenBase object is an object that can be copied into a MatrixBase. + * + * Besides MatrixBase-derived classes, this also includes special matrix classes such as diagonal matrices, etc. + * + * Notice that this class is trivial, it is only used to disambiguate overloaded functions. + * + * \sa \blank \ref TopicClassHierarchy + */ +template struct EigenBase +{ +// typedef typename internal::plain_matrix_type::type PlainObject; + + /** \brief The interface type of indices + * \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE. + * \deprecated Since Eigen 3.3, its usage is deprecated. Use Eigen::Index instead. + * \sa StorageIndex, \ref TopicPreprocessorDirectives. + */ + typedef Eigen::Index Index; + + // FIXME is it needed? + typedef typename internal::traits::StorageKind StorageKind; + + /** \returns a reference to the derived object */ + EIGEN_DEVICE_FUNC + Derived& derived() { return *static_cast(this); } + /** \returns a const reference to the derived object */ + EIGEN_DEVICE_FUNC + const Derived& derived() const { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + inline Derived& const_cast_derived() const + { return *static_cast(const_cast(this)); } + EIGEN_DEVICE_FUNC + inline const Derived& const_derived() const + { return *static_cast(this); } + + /** \returns the number of rows. \sa cols(), RowsAtCompileTime */ + EIGEN_DEVICE_FUNC + inline Index rows() const { return derived().rows(); } + /** \returns the number of columns. \sa rows(), ColsAtCompileTime*/ + EIGEN_DEVICE_FUNC + inline Index cols() const { return derived().cols(); } + /** \returns the number of coefficients, which is rows()*cols(). + * \sa rows(), cols(), SizeAtCompileTime. */ + EIGEN_DEVICE_FUNC + inline Index size() const { return rows() * cols(); } + + /** \internal Don't use it, but do the equivalent: \code dst = *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const + { derived().evalTo(dst); } + + /** \internal Don't use it, but do the equivalent: \code dst += *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void addTo(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + typename Dest::PlainObject res(rows(),cols()); + evalTo(res); + dst += res; + } + + /** \internal Don't use it, but do the equivalent: \code dst -= *this; \endcode */ + template + EIGEN_DEVICE_FUNC + inline void subTo(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + typename Dest::PlainObject res(rows(),cols()); + evalTo(res); + dst -= res; + } + + /** \internal Don't use it, but do the equivalent: \code dst.applyOnTheRight(*this); \endcode */ + template + EIGEN_DEVICE_FUNC inline void applyThisOnTheRight(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + dst = dst * this->derived(); + } + + /** \internal Don't use it, but do the equivalent: \code dst.applyOnTheLeft(*this); \endcode */ + template + EIGEN_DEVICE_FUNC inline void applyThisOnTheLeft(Dest& dst) const + { + // This is the default implementation, + // derived class can reimplement it in a more optimized way. + dst = this->derived() * dst; + } + +}; + +/*************************************************************************** +* Implementation of matrix base methods +***************************************************************************/ + +/** \brief Copies the generic expression \a other into *this. + * + * \details The expression must provide a (templated) evalTo(Derived& dst) const + * function which does the actual job. In practice, this allows any user to write + * its own special matrix without having to modify MatrixBase + * + * \returns a reference to *this. + */ +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator=(const EigenBase &other) +{ + call_assignment(derived(), other.derived()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator+=(const EigenBase &other) +{ + call_assignment(derived(), other.derived(), internal::add_assign_op()); + return derived(); +} + +template +template +EIGEN_DEVICE_FUNC +Derived& DenseBase::operator-=(const EigenBase &other) +{ + call_assignment(derived(), other.derived(), internal::sub_assign_op()); + return derived(); +} + +} // end namespace Eigen + +#endif // EIGEN_EIGENBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h new file mode 100644 index 00000000000..3e403a09d92 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Fuzzy.h @@ -0,0 +1,155 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_FUZZY_H +#define EIGEN_FUZZY_H + +namespace Eigen { + +namespace internal +{ + +template::IsInteger> +struct isApprox_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec) + { + typename internal::nested_eval::type nested(x); + typename internal::nested_eval::type otherNested(y); + return (nested - otherNested).cwiseAbs2().sum() <= prec * prec * numext::mini(nested.cwiseAbs2().sum(), otherNested.cwiseAbs2().sum()); + } +}; + +template +struct isApprox_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar&) + { + return x.matrix() == y.matrix(); + } +}; + +template::IsInteger> +struct isMuchSmallerThan_object_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec) + { + return x.cwiseAbs2().sum() <= numext::abs2(prec) * y.cwiseAbs2().sum(); + } +}; + +template +struct isMuchSmallerThan_object_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const OtherDerived&, const typename Derived::RealScalar&) + { + return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix(); + } +}; + +template::IsInteger> +struct isMuchSmallerThan_scalar_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const typename Derived::RealScalar& y, const typename Derived::RealScalar& prec) + { + return x.cwiseAbs2().sum() <= numext::abs2(prec * y); + } +}; + +template +struct isMuchSmallerThan_scalar_selector +{ + EIGEN_DEVICE_FUNC + static bool run(const Derived& x, const typename Derived::RealScalar&, const typename Derived::RealScalar&) + { + return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix(); + } +}; + +} // end namespace internal + + +/** \returns \c true if \c *this is approximately equal to \a other, within the precision + * determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. Two vectors \f$ v \f$ and \f$ w \f$ + * are considered to be approximately equal within precision \f$ p \f$ if + * \f[ \Vert v - w \Vert \leqslant p\,\min(\Vert v\Vert, \Vert w\Vert). \f] + * For matrices, the comparison is done using the Hilbert-Schmidt norm (aka Frobenius norm + * L2 norm). + * + * \note Because of the multiplicativeness of this comparison, one can't use this function + * to check whether \c *this is approximately equal to the zero matrix or vector. + * Indeed, \c isApprox(zero) returns false unless \c *this itself is exactly the zero matrix + * or vector. If you want to test whether \c *this is zero, use internal::isMuchSmallerThan(const + * RealScalar&, RealScalar) instead. + * + * \sa internal::isMuchSmallerThan(const RealScalar&, RealScalar) const + */ +template +template +bool DenseBase::isApprox( + const DenseBase& other, + const RealScalar& prec +) const +{ + return internal::isApprox_selector::run(derived(), other.derived(), prec); +} + +/** \returns \c true if the norm of \c *this is much smaller than \a other, + * within the precision determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. A vector \f$ v \f$ is + * considered to be much smaller than \f$ x \f$ within precision \f$ p \f$ if + * \f[ \Vert v \Vert \leqslant p\,\vert x\vert. \f] + * + * For matrices, the comparison is done using the Hilbert-Schmidt norm. For this reason, + * the value of the reference scalar \a other should come from the Hilbert-Schmidt norm + * of a reference matrix of same dimensions. + * + * \sa isApprox(), isMuchSmallerThan(const DenseBase&, RealScalar) const + */ +template +bool DenseBase::isMuchSmallerThan( + const typename NumTraits::Real& other, + const RealScalar& prec +) const +{ + return internal::isMuchSmallerThan_scalar_selector::run(derived(), other, prec); +} + +/** \returns \c true if the norm of \c *this is much smaller than the norm of \a other, + * within the precision determined by \a prec. + * + * \note The fuzzy compares are done multiplicatively. A vector \f$ v \f$ is + * considered to be much smaller than a vector \f$ w \f$ within precision \f$ p \f$ if + * \f[ \Vert v \Vert \leqslant p\,\Vert w\Vert. \f] + * For matrices, the comparison is done using the Hilbert-Schmidt norm. + * + * \sa isApprox(), isMuchSmallerThan(const RealScalar&, RealScalar) const + */ +template +template +bool DenseBase::isMuchSmallerThan( + const DenseBase& other, + const RealScalar& prec +) const +{ + return internal::isMuchSmallerThan_object_selector::run(derived(), other.derived(), prec); +} + +} // end namespace Eigen + +#endif // EIGEN_FUZZY_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h new file mode 100644 index 00000000000..6f0cc80e940 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GeneralProduct.h @@ -0,0 +1,455 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008-2011 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GENERAL_PRODUCT_H +#define EIGEN_GENERAL_PRODUCT_H + +namespace Eigen { + +enum { + Large = 2, + Small = 3 +}; + +namespace internal { + +template struct product_type_selector; + +template struct product_size_category +{ + enum { + #ifndef EIGEN_CUDA_ARCH + is_large = MaxSize == Dynamic || + Size >= EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD || + (Size==Dynamic && MaxSize>=EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD), + #else + is_large = 0, + #endif + value = is_large ? Large + : Size == 1 ? 1 + : Small + }; +}; + +template struct product_type +{ + typedef typename remove_all::type _Lhs; + typedef typename remove_all::type _Rhs; + enum { + MaxRows = traits<_Lhs>::MaxRowsAtCompileTime, + Rows = traits<_Lhs>::RowsAtCompileTime, + MaxCols = traits<_Rhs>::MaxColsAtCompileTime, + Cols = traits<_Rhs>::ColsAtCompileTime, + MaxDepth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::MaxColsAtCompileTime, + traits<_Rhs>::MaxRowsAtCompileTime), + Depth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::ColsAtCompileTime, + traits<_Rhs>::RowsAtCompileTime) + }; + + // the splitting into different lines of code here, introducing the _select enums and the typedef below, + // is to work around an internal compiler error with gcc 4.1 and 4.2. +private: + enum { + rows_select = product_size_category::value, + cols_select = product_size_category::value, + depth_select = product_size_category::value + }; + typedef product_type_selector selector; + +public: + enum { + value = selector::ret, + ret = selector::ret + }; +#ifdef EIGEN_DEBUG_PRODUCT + static void debug() + { + EIGEN_DEBUG_VAR(Rows); + EIGEN_DEBUG_VAR(Cols); + EIGEN_DEBUG_VAR(Depth); + EIGEN_DEBUG_VAR(rows_select); + EIGEN_DEBUG_VAR(cols_select); + EIGEN_DEBUG_VAR(depth_select); + EIGEN_DEBUG_VAR(value); + } +#endif +}; + +/* The following allows to select the kind of product at compile time + * based on the three dimensions of the product. + * This is a compile time mapping from {1,Small,Large}^3 -> {product types} */ +// FIXME I'm not sure the current mapping is the ideal one. +template struct product_type_selector { enum { ret = OuterProduct }; }; +template struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template struct product_type_selector<1, N, 1> { enum { ret = LazyCoeffBasedProductMode }; }; +template struct product_type_selector<1, 1, Depth> { enum { ret = InnerProduct }; }; +template<> struct product_type_selector<1, 1, 1> { enum { ret = InnerProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Small,Small> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = LazyCoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Large,Small> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector<1, Large,Large> { enum { ret = GemvProduct }; }; +template<> struct product_type_selector<1, Small,Large> { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemvProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = CoeffBasedProductMode }; }; +template<> struct product_type_selector { enum { ret = GemmProduct }; }; + +} // end namespace internal + +/*********************************************************************** +* Implementation of Inner Vector Vector Product +***********************************************************************/ + +// FIXME : maybe the "inner product" could return a Scalar +// instead of a 1x1 matrix ?? +// Pro: more natural for the user +// Cons: this could be a problem if in a meta unrolled algorithm a matrix-matrix +// product ends up to a row-vector times col-vector product... To tackle this use +// case, we could have a specialization for Block with: operator=(Scalar x); + +/*********************************************************************** +* Implementation of Outer Vector Vector Product +***********************************************************************/ + +/*********************************************************************** +* Implementation of General Matrix Vector Product +***********************************************************************/ + +/* According to the shape/flags of the matrix we have to distinghish 3 different cases: + * 1 - the matrix is col-major, BLAS compatible and M is large => call fast BLAS-like colmajor routine + * 2 - the matrix is row-major, BLAS compatible and N is large => call fast BLAS-like rowmajor routine + * 3 - all other cases are handled using a simple loop along the outer-storage direction. + * Therefore we need a lower level meta selector. + * Furthermore, if the matrix is the rhs, then the product has to be transposed. + */ +namespace internal { + +template +struct gemv_dense_selector; + +} // end namespace internal + +namespace internal { + +template struct gemv_static_vector_if; + +template +struct gemv_static_vector_if +{ + EIGEN_STRONG_INLINE Scalar* data() { eigen_internal_assert(false && "should never be called"); return 0; } +}; + +template +struct gemv_static_vector_if +{ + EIGEN_STRONG_INLINE Scalar* data() { return 0; } +}; + +template +struct gemv_static_vector_if +{ + enum { + ForceAlignment = internal::packet_traits::Vectorizable, + PacketSize = internal::packet_traits::size + }; + #if EIGEN_MAX_STATIC_ALIGN_BYTES!=0 + internal::plain_array m_data; + EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; } + #else + // Some architectures cannot align on the stack, + // => let's manually enforce alignment by allocating more data and return the address of the first aligned element. + internal::plain_array m_data; + EIGEN_STRONG_INLINE Scalar* data() { + return ForceAlignment + ? reinterpret_cast((internal::UIntPtr(m_data.array) & ~(std::size_t(EIGEN_MAX_ALIGN_BYTES-1))) + EIGEN_MAX_ALIGN_BYTES) + : m_data.array; + } + #endif +}; + +// The vector is on the left => transposition +template +struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + Transpose destT(dest); + enum { OtherStorageOrder = StorageOrder == RowMajor ? ColMajor : RowMajor }; + gemv_dense_selector + ::run(rhs.transpose(), lhs.transpose(), destT, alpha); + } +}; + +template<> struct gemv_dense_selector +{ + template + static inline void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + typedef typename Lhs::Scalar LhsScalar; + typedef typename Rhs::Scalar RhsScalar; + typedef typename Dest::Scalar ResScalar; + typedef typename Dest::RealScalar RealScalar; + + typedef internal::blas_traits LhsBlasTraits; + typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; + typedef internal::blas_traits RhsBlasTraits; + typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; + + typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; + + ActualLhsType actualLhs = LhsBlasTraits::extract(lhs); + ActualRhsType actualRhs = RhsBlasTraits::extract(rhs); + + ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs) + * RhsBlasTraits::extractScalarFactor(rhs); + + // make sure Dest is a compile-time vector type (bug 1166) + typedef typename conditional::type ActualDest; + + enum { + // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 + // on, the other hand it is good for the cache to pack the vector anyways... + EvalToDestAtCompileTime = (ActualDest::InnerStrideAtCompileTime==1), + ComplexByReal = (NumTraits::IsComplex) && (!NumTraits::IsComplex), + MightCannotUseDest = (!EvalToDestAtCompileTime) || ComplexByReal + }; + + typedef const_blas_data_mapper LhsMapper; + typedef const_blas_data_mapper RhsMapper; + RhsScalar compatibleAlpha = get_factor::run(actualAlpha); + + if(!MightCannotUseDest) + { + // shortcut if we are sure to be able to use dest directly, + // this ease the compiler to generate cleaner and more optimzized code for most common cases + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhs.data(), actualRhs.innerStride()), + dest.data(), 1, + compatibleAlpha); + } + else + { + gemv_static_vector_if static_dest; + + const bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0)); + const bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible; + + ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), + evalToDest ? dest.data() : static_dest.data()); + + if(!evalToDest) + { + #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + Index size = dest.size(); + EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #endif + if(!alphaIsCompatible) + { + MappedDest(actualDestPtr, dest.size()).setZero(); + compatibleAlpha = RhsScalar(1); + } + else + MappedDest(actualDestPtr, dest.size()) = dest; + } + + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhs.data(), actualRhs.innerStride()), + actualDestPtr, 1, + compatibleAlpha); + + if (!evalToDest) + { + if(!alphaIsCompatible) + dest.matrix() += actualAlpha * MappedDest(actualDestPtr, dest.size()); + else + dest = MappedDest(actualDestPtr, dest.size()); + } + } + } +}; + +template<> struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + typedef typename Lhs::Scalar LhsScalar; + typedef typename Rhs::Scalar RhsScalar; + typedef typename Dest::Scalar ResScalar; + + typedef internal::blas_traits LhsBlasTraits; + typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; + typedef internal::blas_traits RhsBlasTraits; + typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; + typedef typename internal::remove_all::type ActualRhsTypeCleaned; + + typename add_const::type actualLhs = LhsBlasTraits::extract(lhs); + typename add_const::type actualRhs = RhsBlasTraits::extract(rhs); + + ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs) + * RhsBlasTraits::extractScalarFactor(rhs); + + enum { + // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 + // on, the other hand it is good for the cache to pack the vector anyways... + DirectlyUseRhs = ActualRhsTypeCleaned::InnerStrideAtCompileTime==1 + }; + + gemv_static_vector_if static_rhs; + + ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,actualRhs.size(), + DirectlyUseRhs ? const_cast(actualRhs.data()) : static_rhs.data()); + + if(!DirectlyUseRhs) + { + #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN + Index size = actualRhs.size(); + EIGEN_DENSE_STORAGE_CTOR_PLUGIN + #endif + Map(actualRhsPtr, actualRhs.size()) = actualRhs; + } + + typedef const_blas_data_mapper LhsMapper; + typedef const_blas_data_mapper RhsMapper; + general_matrix_vector_product + ::run( + actualLhs.rows(), actualLhs.cols(), + LhsMapper(actualLhs.data(), actualLhs.outerStride()), + RhsMapper(actualRhsPtr, 1), + dest.data(), dest.col(0).innerStride(), //NOTE if dest is not a vector at compile-time, then dest.innerStride() might be wrong. (bug 1166) + actualAlpha); + } +}; + +template<> struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + EIGEN_STATIC_ASSERT((!nested_eval::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE); + // TODO if rhs is large enough it might be beneficial to make sure that dest is sequentially stored in memory, otherwise use a temp + typename nested_eval::type actual_rhs(rhs); + const Index size = rhs.rows(); + for(Index k=0; k struct gemv_dense_selector +{ + template + static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) + { + EIGEN_STATIC_ASSERT((!nested_eval::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE); + typename nested_eval::type actual_rhs(rhs); + const Index rows = dest.rows(); + for(Index i=0; i +template +inline const Product +MatrixBase::operator*(const MatrixBase &other) const +{ + // A note regarding the function declaration: In MSVC, this function will sometimes + // not be inlined since DenseStorage is an unwindable object for dynamic + // matrices and product types are holding a member to store the result. + // Thus it does not help tagging this function with EIGEN_STRONG_INLINE. + enum { + ProductIsValid = Derived::ColsAtCompileTime==Dynamic + || OtherDerived::RowsAtCompileTime==Dynamic + || int(Derived::ColsAtCompileTime)==int(OtherDerived::RowsAtCompileTime), + AreVectors = Derived::IsVectorAtCompileTime && OtherDerived::IsVectorAtCompileTime, + SameSizes = EIGEN_PREDICATE_SAME_MATRIX_SIZE(Derived,OtherDerived) + }; + // note to the lost user: + // * for a dot product use: v1.dot(v2) + // * for a coeff-wise product use: v1.cwiseProduct(v2) + EIGEN_STATIC_ASSERT(ProductIsValid || !(AreVectors && SameSizes), + INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS) + EIGEN_STATIC_ASSERT(ProductIsValid || !(SameSizes && !AreVectors), + INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION) + EIGEN_STATIC_ASSERT(ProductIsValid || SameSizes, INVALID_MATRIX_PRODUCT) +#ifdef EIGEN_DEBUG_PRODUCT + internal::product_type::debug(); +#endif + + return Product(derived(), other.derived()); +} + +/** \returns an expression of the matrix product of \c *this and \a other without implicit evaluation. + * + * The returned product will behave like any other expressions: the coefficients of the product will be + * computed once at a time as requested. This might be useful in some extremely rare cases when only + * a small and no coherent fraction of the result's coefficients have to be computed. + * + * \warning This version of the matrix product can be much much slower. So use it only if you know + * what you are doing and that you measured a true speed improvement. + * + * \sa operator*(const MatrixBase&) + */ +template +template +const Product +MatrixBase::lazyProduct(const MatrixBase &other) const +{ + enum { + ProductIsValid = Derived::ColsAtCompileTime==Dynamic + || OtherDerived::RowsAtCompileTime==Dynamic + || int(Derived::ColsAtCompileTime)==int(OtherDerived::RowsAtCompileTime), + AreVectors = Derived::IsVectorAtCompileTime && OtherDerived::IsVectorAtCompileTime, + SameSizes = EIGEN_PREDICATE_SAME_MATRIX_SIZE(Derived,OtherDerived) + }; + // note to the lost user: + // * for a dot product use: v1.dot(v2) + // * for a coeff-wise product use: v1.cwiseProduct(v2) + EIGEN_STATIC_ASSERT(ProductIsValid || !(AreVectors && SameSizes), + INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS) + EIGEN_STATIC_ASSERT(ProductIsValid || !(SameSizes && !AreVectors), + INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION) + EIGEN_STATIC_ASSERT(ProductIsValid || SameSizes, INVALID_MATRIX_PRODUCT) + + return Product(derived(), other.derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h new file mode 100644 index 00000000000..e594437791b --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GenericPacketMath.h @@ -0,0 +1,590 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GENERIC_PACKET_MATH_H +#define EIGEN_GENERIC_PACKET_MATH_H + +namespace Eigen { + +namespace internal { + +/** \internal + * \file GenericPacketMath.h + * + * Default implementation for types not supported by the vectorization. + * In practice these functions are provided to make easier the writing + * of generic vectorized code. + */ + +#ifndef EIGEN_DEBUG_ALIGNED_LOAD +#define EIGEN_DEBUG_ALIGNED_LOAD +#endif + +#ifndef EIGEN_DEBUG_UNALIGNED_LOAD +#define EIGEN_DEBUG_UNALIGNED_LOAD +#endif + +#ifndef EIGEN_DEBUG_ALIGNED_STORE +#define EIGEN_DEBUG_ALIGNED_STORE +#endif + +#ifndef EIGEN_DEBUG_UNALIGNED_STORE +#define EIGEN_DEBUG_UNALIGNED_STORE +#endif + +struct default_packet_traits +{ + enum { + HasHalfPacket = 0, + + HasAdd = 1, + HasSub = 1, + HasMul = 1, + HasNegate = 1, + HasAbs = 1, + HasArg = 0, + HasAbs2 = 1, + HasMin = 1, + HasMax = 1, + HasConj = 1, + HasSetLinear = 1, + HasBlend = 0, + + HasDiv = 0, + HasSqrt = 0, + HasRsqrt = 0, + HasExp = 0, + HasLog = 0, + HasLog1p = 0, + HasLog10 = 0, + HasPow = 0, + + HasSin = 0, + HasCos = 0, + HasTan = 0, + HasASin = 0, + HasACos = 0, + HasATan = 0, + HasSinh = 0, + HasCosh = 0, + HasTanh = 0, + HasLGamma = 0, + HasDiGamma = 0, + HasZeta = 0, + HasPolygamma = 0, + HasErf = 0, + HasErfc = 0, + HasIGamma = 0, + HasIGammac = 0, + HasBetaInc = 0, + + HasRound = 0, + HasFloor = 0, + HasCeil = 0, + + HasSign = 0 + }; +}; + +template struct packet_traits : default_packet_traits +{ + typedef T type; + typedef T half; + enum { + Vectorizable = 0, + size = 1, + AlignedOnScalar = 0, + HasHalfPacket = 0 + }; + enum { + HasAdd = 0, + HasSub = 0, + HasMul = 0, + HasNegate = 0, + HasAbs = 0, + HasAbs2 = 0, + HasMin = 0, + HasMax = 0, + HasConj = 0, + HasSetLinear = 0 + }; +}; + +template struct packet_traits : packet_traits { }; + +template struct type_casting_traits { + enum { + VectorizedCast = 0, + SrcCoeffRatio = 1, + TgtCoeffRatio = 1 + }; +}; + + +/** \internal \returns static_cast(a) (coeff-wise) */ +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a) { + return static_cast(a); +} +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a, const SrcPacket& /*b*/) { + return static_cast(a); +} + +template +EIGEN_DEVICE_FUNC inline TgtPacket +pcast(const SrcPacket& a, const SrcPacket& /*b*/, const SrcPacket& /*c*/, const SrcPacket& /*d*/) { + return static_cast(a); +} + +/** \internal \returns a + b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +padd(const Packet& a, + const Packet& b) { return a+b; } + +/** \internal \returns a - b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +psub(const Packet& a, + const Packet& b) { return a-b; } + +/** \internal \returns -a (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pnegate(const Packet& a) { return -a; } + +/** \internal \returns conj(a) (coeff-wise) */ + +template EIGEN_DEVICE_FUNC inline Packet +pconj(const Packet& a) { return numext::conj(a); } + +/** \internal \returns a * b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmul(const Packet& a, + const Packet& b) { return a*b; } + +/** \internal \returns a / b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pdiv(const Packet& a, + const Packet& b) { return a/b; } + +/** \internal \returns the min of \a a and \a b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmin(const Packet& a, + const Packet& b) { return numext::mini(a, b); } + +/** \internal \returns the max of \a a and \a b (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmax(const Packet& a, + const Packet& b) { return numext::maxi(a, b); } + +/** \internal \returns the absolute value of \a a */ +template EIGEN_DEVICE_FUNC inline Packet +pabs(const Packet& a) { using std::abs; return abs(a); } + +/** \internal \returns the phase angle of \a a */ +template EIGEN_DEVICE_FUNC inline Packet +parg(const Packet& a) { using numext::arg; return arg(a); } + +/** \internal \returns the bitwise and of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pand(const Packet& a, const Packet& b) { return a & b; } + +/** \internal \returns the bitwise or of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +por(const Packet& a, const Packet& b) { return a | b; } + +/** \internal \returns the bitwise xor of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pxor(const Packet& a, const Packet& b) { return a ^ b; } + +/** \internal \returns the bitwise andnot of \a a and \a b */ +template EIGEN_DEVICE_FUNC inline Packet +pandnot(const Packet& a, const Packet& b) { return a & (!b); } + +/** \internal \returns a packet version of \a *from, from must be 16 bytes aligned */ +template EIGEN_DEVICE_FUNC inline Packet +pload(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet version of \a *from, (un-aligned load) */ +template EIGEN_DEVICE_FUNC inline Packet +ploadu(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet with constant coefficients \a a, e.g.: (a,a,a,a) */ +template EIGEN_DEVICE_FUNC inline Packet +pset1(const typename unpacket_traits::type& a) { return a; } + +/** \internal \returns a packet with constant coefficients \a a[0], e.g.: (a[0],a[0],a[0],a[0]) */ +template EIGEN_DEVICE_FUNC inline Packet +pload1(const typename unpacket_traits::type *a) { return pset1(*a); } + +/** \internal \returns a packet with elements of \a *from duplicated. + * For instance, for a packet of 8 elements, 4 scalars will be read from \a *from and + * duplicated to form: {from[0],from[0],from[1],from[1],from[2],from[2],from[3],from[3]} + * Currently, this function is only used for scalar * complex products. + */ +template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet +ploaddup(const typename unpacket_traits::type* from) { return *from; } + +/** \internal \returns a packet with elements of \a *from quadrupled. + * For instance, for a packet of 8 elements, 2 scalars will be read from \a *from and + * replicated to form: {from[0],from[0],from[0],from[0],from[1],from[1],from[1],from[1]} + * Currently, this function is only used in matrix products. + * For packet-size smaller or equal to 4, this function is equivalent to pload1 + */ +template EIGEN_DEVICE_FUNC inline Packet +ploadquad(const typename unpacket_traits::type* from) +{ return pload1(from); } + +/** \internal equivalent to + * \code + * a0 = pload1(a+0); + * a1 = pload1(a+1); + * a2 = pload1(a+2); + * a3 = pload1(a+3); + * \endcode + * \sa pset1, pload1, ploaddup, pbroadcast2 + */ +template EIGEN_DEVICE_FUNC +inline void pbroadcast4(const typename unpacket_traits::type *a, + Packet& a0, Packet& a1, Packet& a2, Packet& a3) +{ + a0 = pload1(a+0); + a1 = pload1(a+1); + a2 = pload1(a+2); + a3 = pload1(a+3); +} + +/** \internal equivalent to + * \code + * a0 = pload1(a+0); + * a1 = pload1(a+1); + * \endcode + * \sa pset1, pload1, ploaddup, pbroadcast4 + */ +template EIGEN_DEVICE_FUNC +inline void pbroadcast2(const typename unpacket_traits::type *a, + Packet& a0, Packet& a1) +{ + a0 = pload1(a+0); + a1 = pload1(a+1); +} + +/** \internal \brief Returns a packet with coefficients (a,a+1,...,a+packet_size-1). */ +template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet +plset(const typename unpacket_traits::type& a) { return a; } + +/** \internal copy the packet \a from to \a *to, \a to must be 16 bytes aligned */ +template EIGEN_DEVICE_FUNC inline void pstore(Scalar* to, const Packet& from) +{ (*to) = from; } + +/** \internal copy the packet \a from to \a *to, (un-aligned store) */ +template EIGEN_DEVICE_FUNC inline void pstoreu(Scalar* to, const Packet& from) +{ (*to) = from; } + + template EIGEN_DEVICE_FUNC inline Packet pgather(const Scalar* from, Index /*stride*/) + { return ploadu(from); } + + template EIGEN_DEVICE_FUNC inline void pscatter(Scalar* to, const Packet& from, Index /*stride*/) + { pstore(to, from); } + +/** \internal tries to do cache prefetching of \a addr */ +template EIGEN_DEVICE_FUNC inline void prefetch(const Scalar* addr) +{ +#ifdef __CUDA_ARCH__ +#if defined(__LP64__) + // 64-bit pointer operand constraint for inlined asm + asm(" prefetch.L1 [ %1 ];" : "=l"(addr) : "l"(addr)); +#else + // 32-bit pointer operand constraint for inlined asm + asm(" prefetch.L1 [ %1 ];" : "=r"(addr) : "r"(addr)); +#endif +#elif (!EIGEN_COMP_MSVC) && (EIGEN_COMP_GNUC || EIGEN_COMP_CLANG || EIGEN_COMP_ICC) + __builtin_prefetch(addr); +#endif +} + +/** \internal \returns the first element of a packet */ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type pfirst(const Packet& a) +{ return a; } + +/** \internal \returns a packet where the element i contains the sum of the packet of \a vec[i] */ +template EIGEN_DEVICE_FUNC inline Packet +preduxp(const Packet* vecs) { return vecs[0]; } + +/** \internal \returns the sum of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux(const Packet& a) +{ return a; } + +/** \internal \returns the sum of the elements of \a a by block of 4 elements. + * For a packet {a0, a1, a2, a3, a4, a5, a6, a7}, it returns a half packet {a0+a4, a1+a5, a2+a6, a3+a7} + * For packet-size smaller or equal to 4, this boils down to a noop. + */ +template EIGEN_DEVICE_FUNC inline +typename conditional<(unpacket_traits::size%8)==0,typename unpacket_traits::half,Packet>::type +predux_downto4(const Packet& a) +{ return a; } + +/** \internal \returns the product of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_mul(const Packet& a) +{ return a; } + +/** \internal \returns the min of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_min(const Packet& a) +{ return a; } + +/** \internal \returns the max of the elements of \a a*/ +template EIGEN_DEVICE_FUNC inline typename unpacket_traits::type predux_max(const Packet& a) +{ return a; } + +/** \internal \returns the reversed elements of \a a*/ +template EIGEN_DEVICE_FUNC inline Packet preverse(const Packet& a) +{ return a; } + +/** \internal \returns \a a with real and imaginary part flipped (for complex type only) */ +template EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet& a) +{ + return Packet(a.imag(),a.real()); +} + +/************************** +* Special math functions +***************************/ + +/** \internal \returns the sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psin(const Packet& a) { using std::sin; return sin(a); } + +/** \internal \returns the cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pcos(const Packet& a) { using std::cos; return cos(a); } + +/** \internal \returns the tan of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet ptan(const Packet& a) { using std::tan; return tan(a); } + +/** \internal \returns the arc sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pasin(const Packet& a) { using std::asin; return asin(a); } + +/** \internal \returns the arc cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pacos(const Packet& a) { using std::acos; return acos(a); } + +/** \internal \returns the arc tangent of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet patan(const Packet& a) { using std::atan; return atan(a); } + +/** \internal \returns the hyperbolic sine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psinh(const Packet& a) { using std::sinh; return sinh(a); } + +/** \internal \returns the hyperbolic cosine of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pcosh(const Packet& a) { using std::cosh; return cosh(a); } + +/** \internal \returns the hyperbolic tan of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet ptanh(const Packet& a) { using std::tanh; return tanh(a); } + +/** \internal \returns the exp of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pexp(const Packet& a) { using std::exp; return exp(a); } + +/** \internal \returns the log of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog(const Packet& a) { using std::log; return log(a); } + +/** \internal \returns the log1p of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog1p(const Packet& a) { return numext::log1p(a); } + +/** \internal \returns the log10 of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet plog10(const Packet& a) { using std::log10; return log10(a); } + +/** \internal \returns the square-root of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); } + +/** \internal \returns the reciprocal square-root of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet prsqrt(const Packet& a) { + return pdiv(pset1(1), psqrt(a)); +} + +/** \internal \returns the rounded value of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pround(const Packet& a) { using numext::round; return round(a); } + +/** \internal \returns the floor of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pfloor(const Packet& a) { using numext::floor; return floor(a); } + +/** \internal \returns the ceil of \a a (coeff-wise) */ +template EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS +Packet pceil(const Packet& a) { using numext::ceil; return ceil(a); } + +/*************************************************************************** +* The following functions might not have to be overwritten for vectorized types +***************************************************************************/ + +/** \internal copy a packet with constant coeficient \a a (e.g., [a,a,a,a]) to \a *to. \a to must be 16 bytes aligned */ +// NOTE: this function must really be templated on the packet type (think about different packet types for the same scalar type) +template +inline void pstore1(typename unpacket_traits::type* to, const typename unpacket_traits::type& a) +{ + pstore(to, pset1(a)); +} + +/** \internal \returns a * b + c (coeff-wise) */ +template EIGEN_DEVICE_FUNC inline Packet +pmadd(const Packet& a, + const Packet& b, + const Packet& c) +{ return padd(pmul(a, b),c); } + +/** \internal \returns a packet version of \a *from. + * The pointer \a from must be aligned on a \a Alignment bytes boundary. */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt(const typename unpacket_traits::type* from) +{ + if(Alignment >= unpacket_traits::alignment) + return pload(from); + else + return ploadu(from); +} + +/** \internal copy the packet \a from to \a *to. + * The pointer \a from must be aligned on a \a Alignment bytes boundary. */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void pstoret(Scalar* to, const Packet& from) +{ + if(Alignment >= unpacket_traits::alignment) + pstore(to, from); + else + pstoreu(to, from); +} + +/** \internal \returns a packet version of \a *from. + * Unlike ploadt, ploadt_ro takes advantage of the read-only memory path on the + * hardware if available to speedup the loading of data that won't be modified + * by the current computation. + */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt_ro(const typename unpacket_traits::type* from) +{ + return ploadt(from); +} + +/** \internal default implementation of palign() allowing partial specialization */ +template +struct palign_impl +{ + // by default data are aligned, so there is nothing to be done :) + static inline void run(PacketType&, const PacketType&) {} +}; + +/** \internal update \a first using the concatenation of the packet_size minus \a Offset last elements + * of \a first and \a Offset first elements of \a second. + * + * This function is currently only used to optimize matrix-vector products on unligned matrices. + * It takes 2 packets that represent a contiguous memory array, and returns a packet starting + * at the position \a Offset. For instance, for packets of 4 elements, we have: + * Input: + * - first = {f0,f1,f2,f3} + * - second = {s0,s1,s2,s3} + * Output: + * - if Offset==0 then {f0,f1,f2,f3} + * - if Offset==1 then {f1,f2,f3,s0} + * - if Offset==2 then {f2,f3,s0,s1} + * - if Offset==3 then {f3,s0,s1,s3} + */ +template +inline void palign(PacketType& first, const PacketType& second) +{ + palign_impl::run(first,second); +} + +/*************************************************************************** +* Fast complex products (GCC generates a function call which is very slow) +***************************************************************************/ + +// Eigen+CUDA does not support complexes. +#ifndef __CUDACC__ + +template<> inline std::complex pmul(const std::complex& a, const std::complex& b) +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } + +template<> inline std::complex pmul(const std::complex& a, const std::complex& b) +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } + +#endif + + +/*************************************************************************** + * PacketBlock, that is a collection of N packets where the number of words + * in the packet is a multiple of N. +***************************************************************************/ +template ::size> struct PacketBlock { + Packet packet[N]; +}; + +template EIGEN_DEVICE_FUNC inline void +ptranspose(PacketBlock& /*kernel*/) { + // Nothing to do in the scalar case, i.e. a 1x1 matrix. +} + +/*************************************************************************** + * Selector, i.e. vector of N boolean values used to select (i.e. blend) + * words from 2 packets. +***************************************************************************/ +template struct Selector { + bool select[N]; +}; + +template EIGEN_DEVICE_FUNC inline Packet +pblend(const Selector::size>& ifPacket, const Packet& thenPacket, const Packet& elsePacket) { + return ifPacket.select[0] ? thenPacket : elsePacket; +} + +/** \internal \returns \a a with the first coefficient replaced by the scalar b */ +template EIGEN_DEVICE_FUNC inline Packet +pinsertfirst(const Packet& a, typename unpacket_traits::type b) +{ + // Default implementation based on pblend. + // It must be specialized for higher performance. + Selector::size> mask; + mask.select[0] = true; + // This for loop should be optimized away by the compiler. + for(Index i=1; i::size; ++i) + mask.select[i] = false; + return pblend(mask, pset1(b), a); +} + +/** \internal \returns \a a with the last coefficient replaced by the scalar b */ +template EIGEN_DEVICE_FUNC inline Packet +pinsertlast(const Packet& a, typename unpacket_traits::type b) +{ + // Default implementation based on pblend. + // It must be specialized for higher performance. + Selector::size> mask; + // This for loop should be optimized away by the compiler. + for(Index i=0; i::size-1; ++i) + mask.select[i] = false; + mask.select[unpacket_traits::size-1] = true; + return pblend(mask, pset1(b), a); +} + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_GENERIC_PACKET_MATH_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h new file mode 100644 index 00000000000..769dc255c20 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/GlobalFunctions.h @@ -0,0 +1,187 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2010-2016 Gael Guennebaud +// Copyright (C) 2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_GLOBAL_FUNCTIONS_H +#define EIGEN_GLOBAL_FUNCTIONS_H + +#ifdef EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \ + /** \returns an expression of the coefficient-wise DOC_OP of \a x + + DOC_DETAILS + + \sa Math functions, class CwiseUnaryOp + */ \ + template \ + inline const Eigen::CwiseUnaryOp, const Derived> \ + NAME(const Eigen::ArrayBase& x); + +#else + +#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \ + template \ + inline const Eigen::CwiseUnaryOp, const Derived> \ + (NAME)(const Eigen::ArrayBase& x) { \ + return Eigen::CwiseUnaryOp, const Derived>(x.derived()); \ + } + +#endif // EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(NAME,FUNCTOR) \ + \ + template \ + struct NAME##_retval > \ + { \ + typedef const Eigen::CwiseUnaryOp, const Derived> type; \ + }; \ + template \ + struct NAME##_impl > \ + { \ + static inline typename NAME##_retval >::type run(const Eigen::ArrayBase& x) \ + { \ + return typename NAME##_retval >::type(x.derived()); \ + } \ + }; + +namespace Eigen +{ + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(real,scalar_real_op,real part,\sa ArrayBase::real) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(imag,scalar_imag_op,imaginary part,\sa ArrayBase::imag) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(conj,scalar_conjugate_op,complex conjugate,\sa ArrayBase::conjugate) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(inverse,scalar_inverse_op,inverse,\sa ArrayBase::inverse) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sin,scalar_sin_op,sine,\sa ArrayBase::sin) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cos,scalar_cos_op,cosine,\sa ArrayBase::cos) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tan,scalar_tan_op,tangent,\sa ArrayBase::tan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(atan,scalar_atan_op,arc-tangent,\sa ArrayBase::atan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(asin,scalar_asin_op,arc-sine,\sa ArrayBase::asin) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(acos,scalar_acos_op,arc-consine,\sa ArrayBase::acos) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sinh,scalar_sinh_op,hyperbolic sine,\sa ArrayBase::sinh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cosh,scalar_cosh_op,hyperbolic cosine,\sa ArrayBase::cosh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tanh,scalar_tanh_op,hyperbolic tangent,\sa ArrayBase::tanh) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(lgamma,scalar_lgamma_op,natural logarithm of the gamma function,\sa ArrayBase::lgamma) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(digamma,scalar_digamma_op,derivative of lgamma,\sa ArrayBase::digamma) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erf,scalar_erf_op,error function,\sa ArrayBase::erf) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erfc,scalar_erfc_op,complement error function,\sa ArrayBase::erfc) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(exp,scalar_exp_op,exponential,\sa ArrayBase::exp) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log,scalar_log_op,natural logarithm,\sa Eigen::log10 DOXCOMMA ArrayBase::log) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log1p,scalar_log1p_op,natural logarithm of 1 plus the value,\sa ArrayBase::log1p) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log10,scalar_log10_op,base 10 logarithm,\sa Eigen::log DOXCOMMA ArrayBase::log) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs,scalar_abs_op,absolute value,\sa ArrayBase::abs DOXCOMMA MatrixBase::cwiseAbs) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs2,scalar_abs2_op,squared absolute value,\sa ArrayBase::abs2 DOXCOMMA MatrixBase::cwiseAbs2) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(arg,scalar_arg_op,complex argument,\sa ArrayBase::arg) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sqrt,scalar_sqrt_op,square root,\sa ArrayBase::sqrt DOXCOMMA MatrixBase::cwiseSqrt) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(rsqrt,scalar_rsqrt_op,reciprocal square root,\sa ArrayBase::rsqrt) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(square,scalar_square_op,square (power 2),\sa Eigen::abs2 DOXCOMMA Eigen::pow DOXCOMMA ArrayBase::square) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cube,scalar_cube_op,cube (power 3),\sa Eigen::pow DOXCOMMA ArrayBase::cube) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(round,scalar_round_op,nearest integer,\sa Eigen::floor DOXCOMMA Eigen::ceil DOXCOMMA ArrayBase::round) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(floor,scalar_floor_op,nearest integer not greater than the giben value,\sa Eigen::ceil DOXCOMMA ArrayBase::floor) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(ceil,scalar_ceil_op,nearest integer not less than the giben value,\sa Eigen::floor DOXCOMMA ArrayBase::ceil) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isnan,scalar_isnan_op,not-a-number test,\sa Eigen::isinf DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isnan) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isinf,scalar_isinf_op,infinite value test,\sa Eigen::isnan DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isinf) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isfinite,scalar_isfinite_op,finite value test,\sa Eigen::isinf DOXCOMMA Eigen::isnan DOXCOMMA ArrayBase::isfinite) + EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sign,scalar_sign_op,sign (or 0),\sa ArrayBase::sign) + + /** \returns an expression of the coefficient-wise power of \a x to the given constant \a exponent. + * + * \tparam ScalarExponent is the scalar type of \a exponent. It must be compatible with the scalar type of the given expression (\c Derived::Scalar). + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ +#ifdef EIGEN_PARSED_BY_DOXYGEN + template + inline const CwiseBinaryOp,Derived,Constant > + pow(const Eigen::ArrayBase& x, const ScalarExponent& exponent); +#else + template + inline typename internal::enable_if< !(internal::is_same::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,typename Derived::Scalar,ScalarExponent), + const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,ScalarExponent,pow) >::type + pow(const Eigen::ArrayBase& x, const ScalarExponent& exponent) { + return x.derived().pow(exponent); + } + + template + inline const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename Derived::Scalar,pow) + pow(const Eigen::ArrayBase& x, const typename Derived::Scalar& exponent) { + return x.derived().pow(exponent); + } +#endif + + /** \returns an expression of the coefficient-wise power of \a x to the given array of \a exponents. + * + * This function computes the coefficient-wise power. + * + * Example: \include Cwise_array_power_array.cpp + * Output: \verbinclude Cwise_array_power_array.out + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ + template + inline const Eigen::CwiseBinaryOp, const Derived, const ExponentDerived> + pow(const Eigen::ArrayBase& x, const Eigen::ArrayBase& exponents) + { + return Eigen::CwiseBinaryOp, const Derived, const ExponentDerived>( + x.derived(), + exponents.derived() + ); + } + + /** \returns an expression of the coefficient-wise power of the scalar \a x to the given array of \a exponents. + * + * This function computes the coefficient-wise power between a scalar and an array of exponents. + * + * \tparam Scalar is the scalar type of \a x. It must be compatible with the scalar type of the given array expression (\c Derived::Scalar). + * + * Example: \include Cwise_scalar_power_array.cpp + * Output: \verbinclude Cwise_scalar_power_array.out + * + * \sa ArrayBase::pow() + * + * \relates ArrayBase + */ +#ifdef EIGEN_PARSED_BY_DOXYGEN + template + inline const CwiseBinaryOp,Constant,Derived> + pow(const Scalar& x,const Eigen::ArrayBase& x); +#else + template + inline typename internal::enable_if< !(internal::is_same::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,Scalar,typename Derived::Scalar), + const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow) >::type + pow(const Scalar& x, const Eigen::ArrayBase& exponents) + { + return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow)( + typename internal::plain_constant_type::type(exponents.rows(), exponents.cols(), x), exponents.derived() ); + } + + template + inline const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow) + pow(const typename Derived::Scalar& x, const Eigen::ArrayBase& exponents) + { + return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow)( + typename internal::plain_constant_type::type(exponents.rows(), exponents.cols(), x), exponents.derived() ); + } +#endif + + + namespace internal + { + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(real,scalar_real_op) + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(imag,scalar_imag_op) + EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(abs2,scalar_abs2_op) + } +} + +// TODO: cleanly disable those functions that are not supported on Array (numext::real_ref, internal::random, internal::isApprox...) + +#endif // EIGEN_GLOBAL_FUNCTIONS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/IO.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/IO.h new file mode 100644 index 00000000000..da7fd6cce2e --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/IO.h @@ -0,0 +1,225 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_IO_H +#define EIGEN_IO_H + +namespace Eigen { + +enum { DontAlignCols = 1 }; +enum { StreamPrecision = -1, + FullPrecision = -2 }; + +namespace internal { +template +std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat& fmt); +} + +/** \class IOFormat + * \ingroup Core_Module + * + * \brief Stores a set of parameters controlling the way matrices are printed + * + * List of available parameters: + * - \b precision number of digits for floating point values, or one of the special constants \c StreamPrecision and \c FullPrecision. + * The default is the special value \c StreamPrecision which means to use the + * stream's own precision setting, as set for instance using \c cout.precision(3). The other special value + * \c FullPrecision means that the number of digits will be computed to match the full precision of each floating-point + * type. + * - \b flags an OR-ed combination of flags, the default value is 0, the only currently available flag is \c DontAlignCols which + * allows to disable the alignment of columns, resulting in faster code. + * - \b coeffSeparator string printed between two coefficients of the same row + * - \b rowSeparator string printed between two rows + * - \b rowPrefix string printed at the beginning of each row + * - \b rowSuffix string printed at the end of each row + * - \b matPrefix string printed at the beginning of the matrix + * - \b matSuffix string printed at the end of the matrix + * + * Example: \include IOFormat.cpp + * Output: \verbinclude IOFormat.out + * + * \sa DenseBase::format(), class WithFormat + */ +struct IOFormat +{ + /** Default constructor, see class IOFormat for the meaning of the parameters */ + IOFormat(int _precision = StreamPrecision, int _flags = 0, + const std::string& _coeffSeparator = " ", + const std::string& _rowSeparator = "\n", const std::string& _rowPrefix="", const std::string& _rowSuffix="", + const std::string& _matPrefix="", const std::string& _matSuffix="") + : matPrefix(_matPrefix), matSuffix(_matSuffix), rowPrefix(_rowPrefix), rowSuffix(_rowSuffix), rowSeparator(_rowSeparator), + rowSpacer(""), coeffSeparator(_coeffSeparator), precision(_precision), flags(_flags) + { + // TODO check if rowPrefix, rowSuffix or rowSeparator contains a newline + // don't add rowSpacer if columns are not to be aligned + if((flags & DontAlignCols)) + return; + int i = int(matSuffix.length())-1; + while (i>=0 && matSuffix[i]!='\n') + { + rowSpacer += ' '; + i--; + } + } + std::string matPrefix, matSuffix; + std::string rowPrefix, rowSuffix, rowSeparator, rowSpacer; + std::string coeffSeparator; + int precision; + int flags; +}; + +/** \class WithFormat + * \ingroup Core_Module + * + * \brief Pseudo expression providing matrix output with given format + * + * \tparam ExpressionType the type of the object on which IO stream operations are performed + * + * This class represents an expression with stream operators controlled by a given IOFormat. + * It is the return type of DenseBase::format() + * and most of the time this is the only way it is used. + * + * See class IOFormat for some examples. + * + * \sa DenseBase::format(), class IOFormat + */ +template +class WithFormat +{ + public: + + WithFormat(const ExpressionType& matrix, const IOFormat& format) + : m_matrix(matrix), m_format(format) + {} + + friend std::ostream & operator << (std::ostream & s, const WithFormat& wf) + { + return internal::print_matrix(s, wf.m_matrix.eval(), wf.m_format); + } + + protected: + typename ExpressionType::Nested m_matrix; + IOFormat m_format; +}; + +namespace internal { + +// NOTE: This helper is kept for backward compatibility with previous code specializing +// this internal::significant_decimals_impl structure. In the future we should directly +// call digits10() which has been introduced in July 2016 in 3.3. +template +struct significant_decimals_impl +{ + static inline int run() + { + return NumTraits::digits10(); + } +}; + +/** \internal + * print the matrix \a _m to the output stream \a s using the output format \a fmt */ +template +std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat& fmt) +{ + if(_m.size() == 0) + { + s << fmt.matPrefix << fmt.matSuffix; + return s; + } + + typename Derived::Nested m = _m; + typedef typename Derived::Scalar Scalar; + + Index width = 0; + + std::streamsize explicit_precision; + if(fmt.precision == StreamPrecision) + { + explicit_precision = 0; + } + else if(fmt.precision == FullPrecision) + { + if (NumTraits::IsInteger) + { + explicit_precision = 0; + } + else + { + explicit_precision = significant_decimals_impl::run(); + } + } + else + { + explicit_precision = fmt.precision; + } + + std::streamsize old_precision = 0; + if(explicit_precision) old_precision = s.precision(explicit_precision); + + bool align_cols = !(fmt.flags & DontAlignCols); + if(align_cols) + { + // compute the largest width + for(Index j = 0; j < m.cols(); ++j) + for(Index i = 0; i < m.rows(); ++i) + { + std::stringstream sstr; + sstr.copyfmt(s); + sstr << m.coeff(i,j); + width = std::max(width, Index(sstr.str().length())); + } + } + s << fmt.matPrefix; + for(Index i = 0; i < m.rows(); ++i) + { + if (i) + s << fmt.rowSpacer; + s << fmt.rowPrefix; + if(width) s.width(width); + s << m.coeff(i, 0); + for(Index j = 1; j < m.cols(); ++j) + { + s << fmt.coeffSeparator; + if (width) s.width(width); + s << m.coeff(i, j); + } + s << fmt.rowSuffix; + if( i < m.rows() - 1) + s << fmt.rowSeparator; + } + s << fmt.matSuffix; + if(explicit_precision) s.precision(old_precision); + return s; +} + +} // end namespace internal + +/** \relates DenseBase + * + * Outputs the matrix, to the given stream. + * + * If you wish to print the matrix with a format different than the default, use DenseBase::format(). + * + * It is also possible to change the default format by defining EIGEN_DEFAULT_IO_FORMAT before including Eigen headers. + * If not defined, this will automatically be defined to Eigen::IOFormat(), that is the Eigen::IOFormat with default parameters. + * + * \sa DenseBase::format() + */ +template +std::ostream & operator << +(std::ostream & s, + const DenseBase & m) +{ + return internal::print_matrix(s, m.eval(), EIGEN_DEFAULT_IO_FORMAT); +} + +} // end namespace Eigen + +#endif // EIGEN_IO_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h new file mode 100644 index 00000000000..b76f0439d80 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Inverse.h @@ -0,0 +1,118 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2014 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_INVERSE_H +#define EIGEN_INVERSE_H + +namespace Eigen { + +template class InverseImpl; + +namespace internal { + +template +struct traits > + : traits +{ + typedef typename XprType::PlainObject PlainObject; + typedef traits BaseTraits; + enum { + Flags = BaseTraits::Flags & RowMajorBit + }; +}; + +} // end namespace internal + +/** \class Inverse + * + * \brief Expression of the inverse of another expression + * + * \tparam XprType the type of the expression we are taking the inverse + * + * This class represents an abstract expression of A.inverse() + * and most of the time this is the only way it is used. + * + */ +template +class Inverse : public InverseImpl::StorageKind> +{ +public: + typedef typename XprType::StorageIndex StorageIndex; + typedef typename XprType::PlainObject PlainObject; + typedef typename XprType::Scalar Scalar; + typedef typename internal::ref_selector::type XprTypeNested; + typedef typename internal::remove_all::type XprTypeNestedCleaned; + typedef typename internal::ref_selector::type Nested; + typedef typename internal::remove_all::type NestedExpression; + + explicit EIGEN_DEVICE_FUNC Inverse(const XprType &xpr) + : m_xpr(xpr) + {} + + EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } + + EIGEN_DEVICE_FUNC const XprTypeNestedCleaned& nestedExpression() const { return m_xpr; } + +protected: + XprTypeNested m_xpr; +}; + +// Generic API dispatcher +template +class InverseImpl + : public internal::generic_xpr_base >::type +{ +public: + typedef typename internal::generic_xpr_base >::type Base; + typedef typename XprType::Scalar Scalar; +private: + + Scalar coeff(Index row, Index col) const; + Scalar coeff(Index i) const; +}; + +namespace internal { + +/** \internal + * \brief Default evaluator for Inverse expression. + * + * This default evaluator for Inverse expression simply evaluate the inverse into a temporary + * by a call to internal::call_assignment_no_alias. + * Therefore, inverse implementers only have to specialize Assignment, ...> for + * there own nested expression. + * + * \sa class Inverse + */ +template +struct unary_evaluator > + : public evaluator::PlainObject> +{ + typedef Inverse InverseType; + typedef typename InverseType::PlainObject PlainObject; + typedef evaluator Base; + + enum { Flags = Base::Flags | EvalBeforeNestingBit }; + + unary_evaluator(const InverseType& inv_xpr) + : m_result(inv_xpr.rows(), inv_xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + internal::call_assignment_no_alias(m_result, inv_xpr); + } + +protected: + PlainObject m_result; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_INVERSE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Map.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Map.h new file mode 100644 index 00000000000..548bf9a2d55 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Map.h @@ -0,0 +1,171 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MAP_H +#define EIGEN_MAP_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : public traits +{ + typedef traits TraitsBase; + enum { + PlainObjectTypeInnerSize = ((traits::Flags&RowMajorBit)==RowMajorBit) + ? PlainObjectType::ColsAtCompileTime + : PlainObjectType::RowsAtCompileTime, + + InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0 + ? int(PlainObjectType::InnerStrideAtCompileTime) + : int(StrideType::InnerStrideAtCompileTime), + OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0 + ? (InnerStrideAtCompileTime==Dynamic || PlainObjectTypeInnerSize==Dynamic + ? Dynamic + : int(InnerStrideAtCompileTime) * int(PlainObjectTypeInnerSize)) + : int(StrideType::OuterStrideAtCompileTime), + Alignment = int(MapOptions)&int(AlignedMask), + Flags0 = TraitsBase::Flags & (~NestByRefBit), + Flags = is_lvalue::value ? int(Flags0) : (int(Flags0) & ~LvalueBit) + }; +private: + enum { Options }; // Expressions don't have Options +}; +} + +/** \class Map + * \ingroup Core_Module + * + * \brief A matrix or vector expression mapping an existing array of data. + * + * \tparam PlainObjectType the equivalent matrix type of the mapped data + * \tparam MapOptions specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned. + * The default is \c #Unaligned. + * \tparam StrideType optionally specifies strides. By default, Map assumes the memory layout + * of an ordinary, contiguous array. This can be overridden by specifying strides. + * The type passed here must be a specialization of the Stride template, see examples below. + * + * This class represents a matrix or vector expression mapping an existing array of data. + * It can be used to let Eigen interface without any overhead with non-Eigen data structures, + * such as plain C arrays or structures from other libraries. By default, it assumes that the + * data is laid out contiguously in memory. You can however override this by explicitly specifying + * inner and outer strides. + * + * Here's an example of simply mapping a contiguous array as a \ref TopicStorageOrders "column-major" matrix: + * \include Map_simple.cpp + * Output: \verbinclude Map_simple.out + * + * If you need to map non-contiguous arrays, you can do so by specifying strides: + * + * Here's an example of mapping an array as a vector, specifying an inner stride, that is, the pointer + * increment between two consecutive coefficients. Here, we're specifying the inner stride as a compile-time + * fixed value. + * \include Map_inner_stride.cpp + * Output: \verbinclude Map_inner_stride.out + * + * Here's an example of mapping an array while specifying an outer stride. Here, since we're mapping + * as a column-major matrix, 'outer stride' means the pointer increment between two consecutive columns. + * Here, we're specifying the outer stride as a runtime parameter. Note that here \c OuterStride<> is + * a short version of \c OuterStride because the default template parameter of OuterStride + * is \c Dynamic + * \include Map_outer_stride.cpp + * Output: \verbinclude Map_outer_stride.out + * + * For more details and for an example of specifying both an inner and an outer stride, see class Stride. + * + * \b Tip: to change the array of data mapped by a Map object, you can use the C++ + * placement new syntax: + * + * Example: \include Map_placement_new.cpp + * Output: \verbinclude Map_placement_new.out + * + * This class is the return type of PlainObjectBase::Map() but can also be used directly. + * + * \sa PlainObjectBase::Map(), \ref TopicStorageOrders + */ +template class Map + : public MapBase > +{ + public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Map) + + typedef typename Base::PointerType PointerType; + typedef PointerType PointerArgType; + EIGEN_DEVICE_FUNC + inline PointerType cast_to_pointer_type(PointerArgType ptr) { return ptr; } + + EIGEN_DEVICE_FUNC + inline Index innerStride() const + { + return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1; + } + + EIGEN_DEVICE_FUNC + inline Index outerStride() const + { + return int(StrideType::OuterStrideAtCompileTime) != 0 ? m_stride.outer() + : int(internal::traits::OuterStrideAtCompileTime) != Dynamic ? Index(internal::traits::OuterStrideAtCompileTime) + : IsVectorAtCompileTime ? (this->size() * innerStride()) + : (int(Flags)&RowMajorBit) ? (this->cols() * innerStride()) + : (this->rows() * innerStride()); + } + + /** Constructor in the fixed-size case. + * + * \param dataPtr pointer to the array to map + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + explicit inline Map(PointerArgType dataPtr, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr)), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + /** Constructor in the dynamic-size vector case. + * + * \param dataPtr pointer to the array to map + * \param size the size of the vector expression + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + inline Map(PointerArgType dataPtr, Index size, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr), size), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + /** Constructor in the dynamic-size matrix case. + * + * \param dataPtr pointer to the array to map + * \param rows the number of rows of the matrix expression + * \param cols the number of columns of the matrix expression + * \param stride optional Stride object, passing the strides. + */ + EIGEN_DEVICE_FUNC + inline Map(PointerArgType dataPtr, Index rows, Index cols, const StrideType& stride = StrideType()) + : Base(cast_to_pointer_type(dataPtr), rows, cols), m_stride(stride) + { + PlainObjectType::Base::_check_template_params(); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Map) + + protected: + StrideType m_stride; +}; + + +} // end namespace Eigen + +#endif // EIGEN_MAP_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h new file mode 100644 index 00000000000..92c3b2818a4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MapBase.h @@ -0,0 +1,308 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2007-2010 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MAPBASE_H +#define EIGEN_MAPBASE_H + +#define EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) \ + EIGEN_STATIC_ASSERT((int(internal::evaluator::Flags) & LinearAccessBit) || Derived::IsVectorAtCompileTime, \ + YOU_ARE_TRYING_TO_USE_AN_INDEX_BASED_ACCESSOR_ON_AN_EXPRESSION_THAT_DOES_NOT_SUPPORT_THAT) + +namespace Eigen { + +/** \ingroup Core_Module + * + * \brief Base class for dense Map and Block expression with direct access + * + * This base class provides the const low-level accessors (e.g. coeff, coeffRef) of dense + * Map and Block objects with direct access. + * Typical users do not have to directly deal with this class. + * + * This class can be extended by through the macro plugin \c EIGEN_MAPBASE_PLUGIN. + * See \link TopicCustomizing_Plugins customizing Eigen \endlink for details. + * + * The \c Derived class has to provide the following two methods describing the memory layout: + * \code Index innerStride() const; \endcode + * \code Index outerStride() const; \endcode + * + * \sa class Map, class Block + */ +template class MapBase + : public internal::dense_xpr_base::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + enum { + RowsAtCompileTime = internal::traits::RowsAtCompileTime, + ColsAtCompileTime = internal::traits::ColsAtCompileTime, + InnerStrideAtCompileTime = internal::traits::InnerStrideAtCompileTime, + SizeAtCompileTime = Base::SizeAtCompileTime + }; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + typedef typename internal::conditional< + bool(internal::is_lvalue::value), + Scalar *, + const Scalar *>::type + PointerType; + + using Base::derived; +// using Base::RowsAtCompileTime; +// using Base::ColsAtCompileTime; +// using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + using Base::IsRowMajor; + + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::eval; + + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + + // bug 217 - compile error on ICC 11.1 + using Base::operator=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + + /** \copydoc DenseBase::rows() */ + EIGEN_DEVICE_FUNC inline Index rows() const { return m_rows.value(); } + /** \copydoc DenseBase::cols() */ + EIGEN_DEVICE_FUNC inline Index cols() const { return m_cols.value(); } + + /** Returns a pointer to the first coefficient of the matrix or vector. + * + * \note When addressing this data, make sure to honor the strides returned by innerStride() and outerStride(). + * + * \sa innerStride(), outerStride() + */ + EIGEN_DEVICE_FUNC inline const Scalar* data() const { return m_data; } + + /** \copydoc PlainObjectBase::coeff(Index,Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeff(Index rowId, Index colId) const + { + return m_data[colId * colStride() + rowId * rowStride()]; + } + + /** \copydoc PlainObjectBase::coeff(Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeff(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return m_data[index * innerStride()]; + } + + /** \copydoc PlainObjectBase::coeffRef(Index,Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index rowId, Index colId) const + { + return this->m_data[colId * colStride() + rowId * rowStride()]; + } + + /** \copydoc PlainObjectBase::coeffRef(Index) const */ + EIGEN_DEVICE_FUNC + inline const Scalar& coeffRef(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return this->m_data[index * innerStride()]; + } + + /** \internal */ + template + inline PacketScalar packet(Index rowId, Index colId) const + { + return internal::ploadt + (m_data + (colId * colStride() + rowId * rowStride())); + } + + /** \internal */ + template + inline PacketScalar packet(Index index) const + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return internal::ploadt(m_data + index * innerStride()); + } + + /** \internal Constructor for fixed size matrices or vectors */ + EIGEN_DEVICE_FUNC + explicit inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime) + { + EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) + checkSanity(); + } + + /** \internal Constructor for dynamically sized vectors */ + EIGEN_DEVICE_FUNC + inline MapBase(PointerType dataPtr, Index vecSize) + : m_data(dataPtr), + m_rows(RowsAtCompileTime == Dynamic ? vecSize : Index(RowsAtCompileTime)), + m_cols(ColsAtCompileTime == Dynamic ? vecSize : Index(ColsAtCompileTime)) + { + EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) + eigen_assert(vecSize >= 0); + eigen_assert(dataPtr == 0 || SizeAtCompileTime == Dynamic || SizeAtCompileTime == vecSize); + checkSanity(); + } + + /** \internal Constructor for dynamically sized matrices */ + EIGEN_DEVICE_FUNC + inline MapBase(PointerType dataPtr, Index rows, Index cols) + : m_data(dataPtr), m_rows(rows), m_cols(cols) + { + eigen_assert( (dataPtr == 0) + || ( rows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows) + && cols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols))); + checkSanity(); + } + + #ifdef EIGEN_MAPBASE_PLUGIN + #include EIGEN_MAPBASE_PLUGIN + #endif + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) + + template + EIGEN_DEVICE_FUNC + void checkSanity(typename internal::enable_if<(internal::traits::Alignment>0),void*>::type = 0) const + { +#if EIGEN_MAX_ALIGN_BYTES>0 + // innerStride() is not set yet when this function is called, so we optimistically assume the lowest plausible value: + const Index minInnerStride = InnerStrideAtCompileTime == Dynamic ? 1 : Index(InnerStrideAtCompileTime); + EIGEN_ONLY_USED_FOR_DEBUG(minInnerStride); + eigen_assert(( ((internal::UIntPtr(m_data) % internal::traits::Alignment) == 0) + || (cols() * rows() * minInnerStride * sizeof(Scalar)) < internal::traits::Alignment ) && "data is not aligned"); +#endif + } + + template + EIGEN_DEVICE_FUNC + void checkSanity(typename internal::enable_if::Alignment==0,void*>::type = 0) const + {} + + PointerType m_data; + const internal::variable_if_dynamic m_rows; + const internal::variable_if_dynamic m_cols; +}; + +/** \ingroup Core_Module + * + * \brief Base class for non-const dense Map and Block expression with direct access + * + * This base class provides the non-const low-level accessors (e.g. coeff and coeffRef) of + * dense Map and Block objects with direct access. + * It inherits MapBase which defines the const variant for reading specific entries. + * + * \sa class Map, class Block + */ +template class MapBase + : public MapBase +{ + typedef MapBase ReadOnlyMapBase; + public: + + typedef MapBase Base; + + typedef typename Base::Scalar Scalar; + typedef typename Base::PacketScalar PacketScalar; + typedef typename Base::StorageIndex StorageIndex; + typedef typename Base::PointerType PointerType; + + using Base::derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + + using Base::innerStride; + using Base::outerStride; + using Base::rowStride; + using Base::colStride; + + typedef typename internal::conditional< + internal::is_lvalue::value, + Scalar, + const Scalar + >::type ScalarWithConstIfNotLvalue; + + EIGEN_DEVICE_FUNC + inline const Scalar* data() const { return this->m_data; } + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue* data() { return this->m_data; } // no const-cast here so non-const-correct code will give a compile error + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue& coeffRef(Index row, Index col) + { + return this->m_data[col * colStride() + row * rowStride()]; + } + + EIGEN_DEVICE_FUNC + inline ScalarWithConstIfNotLvalue& coeffRef(Index index) + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + return this->m_data[index * innerStride()]; + } + + template + inline void writePacket(Index row, Index col, const PacketScalar& val) + { + internal::pstoret + (this->m_data + (col * colStride() + row * rowStride()), val); + } + + template + inline void writePacket(Index index, const PacketScalar& val) + { + EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) + internal::pstoret + (this->m_data + index * innerStride(), val); + } + + EIGEN_DEVICE_FUNC explicit inline MapBase(PointerType dataPtr) : Base(dataPtr) {} + EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index vecSize) : Base(dataPtr, vecSize) {} + EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index rows, Index cols) : Base(dataPtr, rows, cols) {} + + EIGEN_DEVICE_FUNC + Derived& operator=(const MapBase& other) + { + ReadOnlyMapBase::Base::operator=(other); + return derived(); + } + + // In theory we could simply refer to Base:Base::operator=, but MSVC does not like Base::Base, + // see bugs 821 and 920. + using ReadOnlyMapBase::Base::operator=; + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) +}; + +#undef EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS + +} // end namespace Eigen + +#endif // EIGEN_MAPBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h new file mode 100644 index 00000000000..01736c2a060 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctions.h @@ -0,0 +1,1421 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATHFUNCTIONS_H +#define EIGEN_MATHFUNCTIONS_H + +// source: http://www.geom.uiuc.edu/~huberty/math5337/groupe/digits.html +// TODO this should better be moved to NumTraits +#define EIGEN_PI 3.141592653589793238462643383279502884197169399375105820974944592307816406L + + +namespace Eigen { + +// On WINCE, std::abs is defined for int only, so let's defined our own overloads: +// This issue has been confirmed with MSVC 2008 only, but the issue might exist for more recent versions too. +#if EIGEN_OS_WINCE && EIGEN_COMP_MSVC && EIGEN_COMP_MSVC<=1500 +long abs(long x) { return (labs(x)); } +double abs(double x) { return (fabs(x)); } +float abs(float x) { return (fabsf(x)); } +long double abs(long double x) { return (fabsl(x)); } +#endif + +namespace internal { + +/** \internal \class global_math_functions_filtering_base + * + * What it does: + * Defines a typedef 'type' as follows: + * - if type T has a member typedef Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl, then + * global_math_functions_filtering_base::type is a typedef for it. + * - otherwise, global_math_functions_filtering_base::type is a typedef for T. + * + * How it's used: + * To allow to defined the global math functions (like sin...) in certain cases, like the Array expressions. + * When you do sin(array1+array2), the object array1+array2 has a complicated expression type, all what you want to know + * is that it inherits ArrayBase. So we implement a partial specialization of sin_impl for ArrayBase. + * So we must make sure to use sin_impl > and not sin_impl, otherwise our partial specialization + * won't be used. How does sin know that? That's exactly what global_math_functions_filtering_base tells it. + * + * How it's implemented: + * SFINAE in the style of enable_if. Highly susceptible of breaking compilers. With GCC, it sure does work, but if you replace + * the typename dummy by an integer template parameter, it doesn't work anymore! + */ + +template +struct global_math_functions_filtering_base +{ + typedef T type; +}; + +template struct always_void { typedef void type; }; + +template +struct global_math_functions_filtering_base + ::type + > +{ + typedef typename T::Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl type; +}; + +#define EIGEN_MATHFUNC_IMPL(func, scalar) Eigen::internal::func##_impl::type> +#define EIGEN_MATHFUNC_RETVAL(func, scalar) typename Eigen::internal::func##_retval::type>::type + +/**************************************************************************** +* Implementation of real * +****************************************************************************/ + +template::IsComplex> +struct real_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x; + } +}; + +template +struct real_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + using std::real; + return real(x); + } +}; + +template struct real_impl : real_default_impl {}; + +#ifdef __CUDA_ARCH__ +template +struct real_impl > +{ + typedef T RealScalar; + EIGEN_DEVICE_FUNC + static inline T run(const std::complex& x) + { + return x.real(); + } +}; +#endif + +template +struct real_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of imag * +****************************************************************************/ + +template::IsComplex> +struct imag_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar&) + { + return RealScalar(0); + } +}; + +template +struct imag_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + using std::imag; + return imag(x); + } +}; + +template struct imag_impl : imag_default_impl {}; + +#ifdef __CUDA_ARCH__ +template +struct imag_impl > +{ + typedef T RealScalar; + EIGEN_DEVICE_FUNC + static inline T run(const std::complex& x) + { + return x.imag(); + } +}; +#endif + +template +struct imag_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of real_ref * +****************************************************************************/ + +template +struct real_ref_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar& run(Scalar& x) + { + return reinterpret_cast(&x)[0]; + } + EIGEN_DEVICE_FUNC + static inline const RealScalar& run(const Scalar& x) + { + return reinterpret_cast(&x)[0]; + } +}; + +template +struct real_ref_retval +{ + typedef typename NumTraits::Real & type; +}; + +/**************************************************************************** +* Implementation of imag_ref * +****************************************************************************/ + +template +struct imag_ref_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar& run(Scalar& x) + { + return reinterpret_cast(&x)[1]; + } + EIGEN_DEVICE_FUNC + static inline const RealScalar& run(const Scalar& x) + { + return reinterpret_cast(&x)[1]; + } +}; + +template +struct imag_ref_default_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(Scalar&) + { + return Scalar(0); + } + EIGEN_DEVICE_FUNC + static inline const Scalar run(const Scalar&) + { + return Scalar(0); + } +}; + +template +struct imag_ref_impl : imag_ref_default_impl::IsComplex> {}; + +template +struct imag_ref_retval +{ + typedef typename NumTraits::Real & type; +}; + +/**************************************************************************** +* Implementation of conj * +****************************************************************************/ + +template::IsComplex> +struct conj_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + return x; + } +}; + +template +struct conj_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + using std::conj; + return conj(x); + } +}; + +template +struct conj_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of abs2 * +****************************************************************************/ + +template +struct abs2_impl_default +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x*x; + } +}; + +template +struct abs2_impl_default // IsComplex +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return x.real()*x.real() + x.imag()*x.imag(); + } +}; + +template +struct abs2_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return abs2_impl_default::IsComplex>::run(x); + } +}; + +template +struct abs2_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of norm1 * +****************************************************************************/ + +template +struct norm1_default_impl; + +template +struct norm1_default_impl +{ + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(abs); + return abs(x.real()) + abs(x.imag()); + } +}; + +template +struct norm1_default_impl +{ + EIGEN_DEVICE_FUNC + static inline Scalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(abs); + return abs(x); + } +}; + +template +struct norm1_impl : norm1_default_impl::IsComplex> {}; + +template +struct norm1_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of hypot * +****************************************************************************/ + +template struct hypot_impl; + +template +struct hypot_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of cast * +****************************************************************************/ + +template +struct cast_impl +{ + EIGEN_DEVICE_FUNC + static inline NewType run(const OldType& x) + { + return static_cast(x); + } +}; + +// here, for once, we're plainly returning NewType: we don't want cast to do weird things. + +template +EIGEN_DEVICE_FUNC +inline NewType cast(const OldType& x) +{ + return cast_impl::run(x); +} + +/**************************************************************************** +* Implementation of round * +****************************************************************************/ + +#if EIGEN_HAS_CXX11_MATH + template + struct round_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT((!NumTraits::IsComplex), NUMERIC_TYPE_MUST_BE_REAL) + using std::round; + return round(x); + } + }; +#else + template + struct round_impl + { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT((!NumTraits::IsComplex), NUMERIC_TYPE_MUST_BE_REAL) + EIGEN_USING_STD_MATH(floor); + EIGEN_USING_STD_MATH(ceil); + return (x > Scalar(0)) ? floor(x + Scalar(0.5)) : ceil(x - Scalar(0.5)); + } + }; +#endif + +template +struct round_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of arg * +****************************************************************************/ + +#if EIGEN_HAS_CXX11_MATH + template + struct arg_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(arg); + return arg(x); + } + }; +#else + template::IsComplex> + struct arg_default_impl + { + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + return (x < Scalar(0)) ? Scalar(EIGEN_PI) : Scalar(0); } + }; + + template + struct arg_default_impl + { + typedef typename NumTraits::Real RealScalar; + EIGEN_DEVICE_FUNC + static inline RealScalar run(const Scalar& x) + { + EIGEN_USING_STD_MATH(arg); + return arg(x); + } + }; + + template struct arg_impl : arg_default_impl {}; +#endif + +template +struct arg_retval +{ + typedef typename NumTraits::Real type; +}; + +/**************************************************************************** +* Implementation of log1p * +****************************************************************************/ + +namespace std_fallback { + // fallback log1p implementation in case there is no log1p(Scalar) function in namespace of Scalar, + // or that there is no suitable std::log1p function available + template + EIGEN_DEVICE_FUNC inline Scalar log1p(const Scalar& x) { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar) + typedef typename NumTraits::Real RealScalar; + EIGEN_USING_STD_MATH(log); + Scalar x1p = RealScalar(1) + x; + return numext::equal_strict(x1p, Scalar(1)) ? x : x * ( log(x1p) / (x1p - RealScalar(1)) ); + } +} + +template +struct log1p_impl { + static inline Scalar run(const Scalar& x) + { + EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar) + #if EIGEN_HAS_CXX11_MATH + using std::log1p; + #endif + using std_fallback::log1p; + return log1p(x); + } +}; + + +template +struct log1p_retval +{ + typedef Scalar type; +}; + +/**************************************************************************** +* Implementation of pow * +****************************************************************************/ + +template::IsInteger&&NumTraits::IsInteger> +struct pow_impl +{ + //typedef Scalar retval; + typedef typename ScalarBinaryOpTraits >::ReturnType result_type; + static EIGEN_DEVICE_FUNC inline result_type run(const ScalarX& x, const ScalarY& y) + { + EIGEN_USING_STD_MATH(pow); + return pow(x, y); + } +}; + +template +struct pow_impl +{ + typedef ScalarX result_type; + static EIGEN_DEVICE_FUNC inline ScalarX run(ScalarX x, ScalarY y) + { + ScalarX res(1); + eigen_assert(!NumTraits::IsSigned || y >= 0); + if(y & 1) res *= x; + y >>= 1; + while(y) + { + x *= x; + if(y&1) res *= x; + y >>= 1; + } + return res; + } +}; + +/**************************************************************************** +* Implementation of random * +****************************************************************************/ + +template +struct random_default_impl {}; + +template +struct random_impl : random_default_impl::IsComplex, NumTraits::IsInteger> {}; + +template +struct random_retval +{ + typedef Scalar type; +}; + +template inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(const Scalar& x, const Scalar& y); +template inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(); + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + return x + (y-x) * Scalar(std::rand()) / Scalar(RAND_MAX); + } + static inline Scalar run() + { + return run(Scalar(NumTraits::IsSigned ? -1 : 0), Scalar(1)); + } +}; + +enum { + meta_floor_log2_terminate, + meta_floor_log2_move_up, + meta_floor_log2_move_down, + meta_floor_log2_bogus +}; + +template struct meta_floor_log2_selector +{ + enum { middle = (lower + upper) / 2, + value = (upper <= lower + 1) ? int(meta_floor_log2_terminate) + : (n < (1 << middle)) ? int(meta_floor_log2_move_down) + : (n==0) ? int(meta_floor_log2_bogus) + : int(meta_floor_log2_move_up) + }; +}; + +template::value> +struct meta_floor_log2 {}; + +template +struct meta_floor_log2 +{ + enum { value = meta_floor_log2::middle>::value }; +}; + +template +struct meta_floor_log2 +{ + enum { value = meta_floor_log2::middle, upper>::value }; +}; + +template +struct meta_floor_log2 +{ + enum { value = (n >= ((unsigned int)(1) << (lower+1))) ? lower+1 : lower }; +}; + +template +struct meta_floor_log2 +{ + // no value, error at compile time +}; + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + if (y <= x) + return x; + // ScalarU is the unsigned counterpart of Scalar, possibly Scalar itself. + typedef typename make_unsigned::type ScalarU; + // ScalarX is the widest of ScalarU and unsigned int. + // We'll deal only with ScalarX and unsigned int below thus avoiding signed + // types and arithmetic and signed overflows (which are undefined behavior). + typedef typename conditional<(ScalarU(-1) > unsigned(-1)), ScalarU, unsigned>::type ScalarX; + // The following difference doesn't overflow, provided our integer types are two's + // complement and have the same number of padding bits in signed and unsigned variants. + // This is the case in most modern implementations of C++. + ScalarX range = ScalarX(y) - ScalarX(x); + ScalarX offset = 0; + ScalarX divisor = 1; + ScalarX multiplier = 1; + const unsigned rand_max = RAND_MAX; + if (range <= rand_max) divisor = (rand_max + 1) / (range + 1); + else multiplier = 1 + range / (rand_max + 1); + // Rejection sampling. + do { + offset = (unsigned(std::rand()) * multiplier) / divisor; + } while (offset > range); + return Scalar(ScalarX(x) + offset); + } + + static inline Scalar run() + { +#ifdef EIGEN_MAKING_DOCS + return run(Scalar(NumTraits::IsSigned ? -10 : 0), Scalar(10)); +#else + enum { rand_bits = meta_floor_log2<(unsigned int)(RAND_MAX)+1>::value, + scalar_bits = sizeof(Scalar) * CHAR_BIT, + shift = EIGEN_PLAIN_ENUM_MAX(0, int(rand_bits) - int(scalar_bits)), + offset = NumTraits::IsSigned ? (1 << (EIGEN_PLAIN_ENUM_MIN(rand_bits,scalar_bits)-1)) : 0 + }; + return Scalar((std::rand() >> shift) - offset); +#endif + } +}; + +template +struct random_default_impl +{ + static inline Scalar run(const Scalar& x, const Scalar& y) + { + return Scalar(random(x.real(), y.real()), + random(x.imag(), y.imag())); + } + static inline Scalar run() + { + typedef typename NumTraits::Real RealScalar; + return Scalar(random(), random()); + } +}; + +template +inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random(const Scalar& x, const Scalar& y) +{ + return EIGEN_MATHFUNC_IMPL(random, Scalar)::run(x, y); +} + +template +inline EIGEN_MATHFUNC_RETVAL(random, Scalar) random() +{ + return EIGEN_MATHFUNC_IMPL(random, Scalar)::run(); +} + +// Implementatin of is* functions + +// std::is* do not work with fast-math and gcc, std::is* are available on MSVC 2013 and newer, as well as in clang. +#if (EIGEN_HAS_CXX11_MATH && !(EIGEN_COMP_GNUC_STRICT && __FINITE_MATH_ONLY__)) || (EIGEN_COMP_MSVC>=1800) || (EIGEN_COMP_CLANG) +#define EIGEN_USE_STD_FPCLASSIFY 1 +#else +#define EIGEN_USE_STD_FPCLASSIFY 0 +#endif + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isnan_impl(const T&) { return false; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isinf_impl(const T&) { return false; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if::value,bool>::type +isfinite_impl(const T&) { return true; } + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isfinite_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isfinite)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isfinite; + return isfinite EIGEN_NOT_A_MACRO (x); + #else + return x<=NumTraits::highest() && x>=NumTraits::lowest(); + #endif +} + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isinf_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isinf)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isinf; + return isinf EIGEN_NOT_A_MACRO (x); + #else + return x>NumTraits::highest() || x::lowest(); + #endif +} + +template +EIGEN_DEVICE_FUNC +typename internal::enable_if<(!internal::is_integral::value)&&(!NumTraits::IsComplex),bool>::type +isnan_impl(const T& x) +{ + #ifdef __CUDA_ARCH__ + return (::isnan)(x); + #elif EIGEN_USE_STD_FPCLASSIFY + using std::isnan; + return isnan EIGEN_NOT_A_MACRO (x); + #else + return x != x; + #endif +} + +#if (!EIGEN_USE_STD_FPCLASSIFY) + +#if EIGEN_COMP_MSVC + +template EIGEN_DEVICE_FUNC bool isinf_msvc_helper(T x) +{ + return _fpclass(x)==_FPCLASS_NINF || _fpclass(x)==_FPCLASS_PINF; +} + +//MSVC defines a _isnan builtin function, but for double only +EIGEN_DEVICE_FUNC inline bool isnan_impl(const long double& x) { return _isnan(x)!=0; } +EIGEN_DEVICE_FUNC inline bool isnan_impl(const double& x) { return _isnan(x)!=0; } +EIGEN_DEVICE_FUNC inline bool isnan_impl(const float& x) { return _isnan(x)!=0; } + +EIGEN_DEVICE_FUNC inline bool isinf_impl(const long double& x) { return isinf_msvc_helper(x); } +EIGEN_DEVICE_FUNC inline bool isinf_impl(const double& x) { return isinf_msvc_helper(x); } +EIGEN_DEVICE_FUNC inline bool isinf_impl(const float& x) { return isinf_msvc_helper(x); } + +#elif (defined __FINITE_MATH_ONLY__ && __FINITE_MATH_ONLY__ && EIGEN_COMP_GNUC) + +#if EIGEN_GNUC_AT_LEAST(5,0) + #define EIGEN_TMP_NOOPT_ATTRIB EIGEN_DEVICE_FUNC inline __attribute__((optimize("no-finite-math-only"))) +#else + // NOTE the inline qualifier and noinline attribute are both needed: the former is to avoid linking issue (duplicate symbol), + // while the second prevent too aggressive optimizations in fast-math mode: + #define EIGEN_TMP_NOOPT_ATTRIB EIGEN_DEVICE_FUNC inline __attribute__((noinline,optimize("no-finite-math-only"))) +#endif + +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const long double& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const double& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isnan_impl(const float& x) { return __builtin_isnan(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const double& x) { return __builtin_isinf(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const float& x) { return __builtin_isinf(x); } +template<> EIGEN_TMP_NOOPT_ATTRIB bool isinf_impl(const long double& x) { return __builtin_isinf(x); } + +#undef EIGEN_TMP_NOOPT_ATTRIB + +#endif + +#endif + +// The following overload are defined at the end of this file +template EIGEN_DEVICE_FUNC bool isfinite_impl(const std::complex& x); +template EIGEN_DEVICE_FUNC bool isnan_impl(const std::complex& x); +template EIGEN_DEVICE_FUNC bool isinf_impl(const std::complex& x); + +template T generic_fast_tanh_float(const T& a_x); + +} // end namespace internal + +/**************************************************************************** +* Generic math functions * +****************************************************************************/ + +namespace numext { + +#ifndef __CUDA_ARCH__ +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T mini(const T& x, const T& y) +{ + EIGEN_USING_STD_MATH(min); + return min EIGEN_NOT_A_MACRO (x,y); +} + +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T maxi(const T& x, const T& y) +{ + EIGEN_USING_STD_MATH(max); + return max EIGEN_NOT_A_MACRO (x,y); +} +#else +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T mini(const T& x, const T& y) +{ + return y < x ? y : x; +} +template<> +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE float mini(const float& x, const float& y) +{ + return fminf(x, y); +} +template +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE T maxi(const T& x, const T& y) +{ + return x < y ? y : x; +} +template<> +EIGEN_DEVICE_FUNC +EIGEN_ALWAYS_INLINE float maxi(const float& x, const float& y) +{ + return fmaxf(x, y); +} +#endif + + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(real, Scalar) real(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(real, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) >::type real_ref(const Scalar& x) +{ + return internal::real_ref_impl::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(real_ref, Scalar) real_ref(Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(real_ref, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(imag, Scalar) imag(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(imag, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(arg, Scalar) arg(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(arg, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline typename internal::add_const_on_value_type< EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) >::type imag_ref(const Scalar& x) +{ + return internal::imag_ref_impl::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(imag_ref, Scalar) imag_ref(Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(imag_ref, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(conj, Scalar) conj(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(conj, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x); +} + +EIGEN_DEVICE_FUNC +inline bool abs2(bool x) { return x; } + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(norm1, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(hypot, Scalar) hypot(const Scalar& x, const Scalar& y) +{ + return EIGEN_MATHFUNC_IMPL(hypot, Scalar)::run(x, y); +} + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(log1p, Scalar) log1p(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(log1p, Scalar)::run(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float log1p(const float &x) { return ::log1pf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double log1p(const double &x) { return ::log1p(x); } +#endif + +template +EIGEN_DEVICE_FUNC +inline typename internal::pow_impl::result_type pow(const ScalarX& x, const ScalarY& y) +{ + return internal::pow_impl::run(x, y); +} + +template EIGEN_DEVICE_FUNC bool (isnan) (const T &x) { return internal::isnan_impl(x); } +template EIGEN_DEVICE_FUNC bool (isinf) (const T &x) { return internal::isinf_impl(x); } +template EIGEN_DEVICE_FUNC bool (isfinite)(const T &x) { return internal::isfinite_impl(x); } + +template +EIGEN_DEVICE_FUNC +inline EIGEN_MATHFUNC_RETVAL(round, Scalar) round(const Scalar& x) +{ + return EIGEN_MATHFUNC_IMPL(round, Scalar)::run(x); +} + +template +EIGEN_DEVICE_FUNC +T (floor)(const T& x) +{ + EIGEN_USING_STD_MATH(floor); + return floor(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float floor(const float &x) { return ::floorf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double floor(const double &x) { return ::floor(x); } +#endif + +template +EIGEN_DEVICE_FUNC +T (ceil)(const T& x) +{ + EIGEN_USING_STD_MATH(ceil); + return ceil(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float ceil(const float &x) { return ::ceilf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double ceil(const double &x) { return ::ceil(x); } +#endif + + +/** Log base 2 for 32 bits positive integers. + * Conveniently returns 0 for x==0. */ +inline int log2(int x) +{ + eigen_assert(x>=0); + unsigned int v(x); + static const int table[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 }; + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + return table[(v * 0x07C4ACDDU) >> 27]; +} + +/** \returns the square root of \a x. + * + * It is essentially equivalent to + * \code using std::sqrt; return sqrt(x); \endcode + * but slightly faster for float/double and some compilers (e.g., gcc), thanks to + * specializations when SSE is enabled. + * + * It's usage is justified in performance critical functions, like norm/normalize. + */ +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sqrt(const T &x) +{ + EIGEN_USING_STD_MATH(sqrt); + return sqrt(x); +} + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T log(const T &x) { + EIGEN_USING_STD_MATH(log); + return log(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float log(const float &x) { return ::logf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double log(const double &x) { return ::log(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +typename internal::enable_if::IsSigned || NumTraits::IsComplex,typename NumTraits::Real>::type +abs(const T &x) { + EIGEN_USING_STD_MATH(abs); + return abs(x); +} + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +typename internal::enable_if::IsSigned || NumTraits::IsComplex),typename NumTraits::Real>::type +abs(const T &x) { + return x; +} + +#if defined(__SYCL_DEVICE_ONLY__) +EIGEN_ALWAYS_INLINE float abs(float x) { return cl::sycl::fabs(x); } +EIGEN_ALWAYS_INLINE double abs(double x) { return cl::sycl::fabs(x); } +#endif // defined(__SYCL_DEVICE_ONLY__) + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float abs(const float &x) { return ::fabsf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double abs(const double &x) { return ::fabs(x); } + +template <> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float abs(const std::complex& x) { + return ::hypotf(x.real(), x.imag()); +} + +template <> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double abs(const std::complex& x) { + return ::hypot(x.real(), x.imag()); +} +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T exp(const T &x) { + EIGEN_USING_STD_MATH(exp); + return exp(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float exp(const float &x) { return ::expf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double exp(const double &x) { return ::exp(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T cos(const T &x) { + EIGEN_USING_STD_MATH(cos); + return cos(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float cos(const float &x) { return ::cosf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double cos(const double &x) { return ::cos(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sin(const T &x) { + EIGEN_USING_STD_MATH(sin); + return sin(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float sin(const float &x) { return ::sinf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double sin(const double &x) { return ::sin(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T tan(const T &x) { + EIGEN_USING_STD_MATH(tan); + return tan(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tan(const float &x) { return ::tanf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double tan(const double &x) { return ::tan(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T acos(const T &x) { + EIGEN_USING_STD_MATH(acos); + return acos(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float acos(const float &x) { return ::acosf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double acos(const double &x) { return ::acos(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T asin(const T &x) { + EIGEN_USING_STD_MATH(asin); + return asin(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float asin(const float &x) { return ::asinf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double asin(const double &x) { return ::asin(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T atan(const T &x) { + EIGEN_USING_STD_MATH(atan); + return atan(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float atan(const float &x) { return ::atanf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double atan(const double &x) { return ::atan(x); } +#endif + + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T cosh(const T &x) { + EIGEN_USING_STD_MATH(cosh); + return cosh(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float cosh(const float &x) { return ::coshf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double cosh(const double &x) { return ::cosh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T sinh(const T &x) { + EIGEN_USING_STD_MATH(sinh); + return sinh(x); +} + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float sinh(const float &x) { return ::sinhf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double sinh(const double &x) { return ::sinh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T tanh(const T &x) { + EIGEN_USING_STD_MATH(tanh); + return tanh(x); +} + +#if (!defined(__CUDACC__)) && EIGEN_FAST_MATH +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tanh(float x) { return internal::generic_fast_tanh_float(x); } +#endif + +#ifdef __CUDACC__ +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float tanh(const float &x) { return ::tanhf(x); } + +template<> EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double tanh(const double &x) { return ::tanh(x); } +#endif + +template +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +T fmod(const T& a, const T& b) { + EIGEN_USING_STD_MATH(fmod); + return fmod(a, b); +} + +#ifdef __CUDACC__ +template <> +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +float fmod(const float& a, const float& b) { + return ::fmodf(a, b); +} + +template <> +EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE +double fmod(const double& a, const double& b) { + return ::fmod(a, b); +} +#endif + +} // end namespace numext + +namespace internal { + +template +EIGEN_DEVICE_FUNC bool isfinite_impl(const std::complex& x) +{ + return (numext::isfinite)(numext::real(x)) && (numext::isfinite)(numext::imag(x)); +} + +template +EIGEN_DEVICE_FUNC bool isnan_impl(const std::complex& x) +{ + return (numext::isnan)(numext::real(x)) || (numext::isnan)(numext::imag(x)); +} + +template +EIGEN_DEVICE_FUNC bool isinf_impl(const std::complex& x) +{ + return ((numext::isinf)(numext::real(x)) || (numext::isinf)(numext::imag(x))) && (!(numext::isnan)(x)); +} + +/**************************************************************************** +* Implementation of fuzzy comparisons * +****************************************************************************/ + +template +struct scalar_fuzzy_default_impl {}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec) + { + return numext::abs(x) <= numext::abs(y) * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return numext::abs(x - y) <= numext::mini(numext::abs(x), numext::abs(y)) * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return x <= y || isApprox(x, y, prec); + } +}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const Scalar&, const RealScalar&) + { + return x == Scalar(0); + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar&) + { + return x == y; + } + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, const RealScalar&) + { + return x <= y; + } +}; + +template +struct scalar_fuzzy_default_impl +{ + typedef typename NumTraits::Real RealScalar; + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, const RealScalar& prec) + { + return numext::abs2(x) <= numext::abs2(y) * prec * prec; + } + EIGEN_DEVICE_FUNC + static inline bool isApprox(const Scalar& x, const Scalar& y, const RealScalar& prec) + { + return numext::abs2(x - y) <= numext::mini(numext::abs2(x), numext::abs2(y)) * prec * prec; + } +}; + +template +struct scalar_fuzzy_impl : scalar_fuzzy_default_impl::IsComplex, NumTraits::IsInteger> {}; + +template EIGEN_DEVICE_FUNC +inline bool isMuchSmallerThan(const Scalar& x, const OtherScalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::template isMuchSmallerThan(x, y, precision); +} + +template EIGEN_DEVICE_FUNC +inline bool isApprox(const Scalar& x, const Scalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::isApprox(x, y, precision); +} + +template EIGEN_DEVICE_FUNC +inline bool isApproxOrLessThan(const Scalar& x, const Scalar& y, + const typename NumTraits::Real &precision = NumTraits::dummy_precision()) +{ + return scalar_fuzzy_impl::isApproxOrLessThan(x, y, precision); +} + +/****************************************** +*** The special case of the bool type *** +******************************************/ + +template<> struct random_impl +{ + static inline bool run() + { + return random(0,1)==0 ? false : true; + } +}; + +template<> struct scalar_fuzzy_impl +{ + typedef bool RealScalar; + + template EIGEN_DEVICE_FUNC + static inline bool isMuchSmallerThan(const bool& x, const bool&, const bool&) + { + return !x; + } + + EIGEN_DEVICE_FUNC + static inline bool isApprox(bool x, bool y, bool) + { + return x == y; + } + + EIGEN_DEVICE_FUNC + static inline bool isApproxOrLessThan(const bool& x, const bool& y, const bool&) + { + return (!x) || y; + } + +}; + + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_MATHFUNCTIONS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h new file mode 100644 index 00000000000..9c1ceb0eb0f --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MathFunctionsImpl.h @@ -0,0 +1,101 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2014 Pedro Gonnet (pedro.gonnet@gmail.com) +// Copyright (C) 2016 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATHFUNCTIONSIMPL_H +#define EIGEN_MATHFUNCTIONSIMPL_H + +namespace Eigen { + +namespace internal { + +/** \internal \returns the hyperbolic tan of \a a (coeff-wise) + Doesn't do anything fancy, just a 13/6-degree rational interpolant which + is accurate up to a couple of ulp in the range [-9, 9], outside of which + the tanh(x) = +/-1. + + This implementation works on both scalars and packets. +*/ +template +T generic_fast_tanh_float(const T& a_x) +{ + // Clamp the inputs to the range [-9, 9] since anything outside + // this range is +/-1.0f in single-precision. + const T plus_9 = pset1(9.f); + const T minus_9 = pset1(-9.f); + // NOTE GCC prior to 6.3 might improperly optimize this max/min + // step such that if a_x is nan, x will be either 9 or -9, + // and tanh will return 1 or -1 instead of nan. + // This is supposed to be fixed in gcc6.3, + // see: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72867 + const T x = pmax(minus_9,pmin(plus_9,a_x)); + // The monomial coefficients of the numerator polynomial (odd). + const T alpha_1 = pset1(4.89352455891786e-03f); + const T alpha_3 = pset1(6.37261928875436e-04f); + const T alpha_5 = pset1(1.48572235717979e-05f); + const T alpha_7 = pset1(5.12229709037114e-08f); + const T alpha_9 = pset1(-8.60467152213735e-11f); + const T alpha_11 = pset1(2.00018790482477e-13f); + const T alpha_13 = pset1(-2.76076847742355e-16f); + + // The monomial coefficients of the denominator polynomial (even). + const T beta_0 = pset1(4.89352518554385e-03f); + const T beta_2 = pset1(2.26843463243900e-03f); + const T beta_4 = pset1(1.18534705686654e-04f); + const T beta_6 = pset1(1.19825839466702e-06f); + + // Since the polynomials are odd/even, we need x^2. + const T x2 = pmul(x, x); + + // Evaluate the numerator polynomial p. + T p = pmadd(x2, alpha_13, alpha_11); + p = pmadd(x2, p, alpha_9); + p = pmadd(x2, p, alpha_7); + p = pmadd(x2, p, alpha_5); + p = pmadd(x2, p, alpha_3); + p = pmadd(x2, p, alpha_1); + p = pmul(x, p); + + // Evaluate the denominator polynomial p. + T q = pmadd(x2, beta_6, beta_4); + q = pmadd(x2, q, beta_2); + q = pmadd(x2, q, beta_0); + + // Divide the numerator by the denominator. + return pdiv(p, q); +} + +template +EIGEN_STRONG_INLINE +RealScalar positive_real_hypot(const RealScalar& x, const RealScalar& y) +{ + EIGEN_USING_STD_MATH(sqrt); + RealScalar p, qp; + p = numext::maxi(x,y); + if(p==RealScalar(0)) return RealScalar(0); + qp = numext::mini(y,x) / p; + return p * sqrt(RealScalar(1) + qp*qp); +} + +template +struct hypot_impl +{ + typedef typename NumTraits::Real RealScalar; + static inline RealScalar run(const Scalar& x, const Scalar& y) + { + EIGEN_USING_STD_MATH(abs); + return positive_real_hypot(abs(x), abs(y)); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_MATHFUNCTIONSIMPL_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h new file mode 100644 index 00000000000..7f4a7af93c6 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Matrix.h @@ -0,0 +1,459 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// Copyright (C) 2008-2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIX_H +#define EIGEN_MATRIX_H + +namespace Eigen { + +namespace internal { +template +struct traits > +{ +private: + enum { size = internal::size_at_compile_time<_Rows,_Cols>::ret }; + typedef typename find_best_packet<_Scalar,size>::type PacketScalar; + enum { + row_major_bit = _Options&RowMajor ? RowMajorBit : 0, + is_dynamic_size_storage = _MaxRows==Dynamic || _MaxCols==Dynamic, + max_size = is_dynamic_size_storage ? Dynamic : _MaxRows*_MaxCols, + default_alignment = compute_default_alignment<_Scalar,max_size>::value, + actual_alignment = ((_Options&DontAlign)==0) ? default_alignment : 0, + required_alignment = unpacket_traits::alignment, + packet_access_bit = (packet_traits<_Scalar>::Vectorizable && (EIGEN_UNALIGNED_VECTORIZE || (actual_alignment>=required_alignment))) ? PacketAccessBit : 0 + }; + +public: + typedef _Scalar Scalar; + typedef Dense StorageKind; + typedef Eigen::Index StorageIndex; + typedef MatrixXpr XprKind; + enum { + RowsAtCompileTime = _Rows, + ColsAtCompileTime = _Cols, + MaxRowsAtCompileTime = _MaxRows, + MaxColsAtCompileTime = _MaxCols, + Flags = compute_matrix_flags<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::ret, + Options = _Options, + InnerStrideAtCompileTime = 1, + OuterStrideAtCompileTime = (Options&RowMajor) ? ColsAtCompileTime : RowsAtCompileTime, + + // FIXME, the following flag in only used to define NeedsToAlign in PlainObjectBase + EvaluatorFlags = LinearAccessBit | DirectAccessBit | packet_access_bit | row_major_bit, + Alignment = actual_alignment + }; +}; +} + +/** \class Matrix + * \ingroup Core_Module + * + * \brief The matrix class, also used for vectors and row-vectors + * + * The %Matrix class is the work-horse for all \em dense (\ref dense "note") matrices and vectors within Eigen. + * Vectors are matrices with one column, and row-vectors are matrices with one row. + * + * The %Matrix class encompasses \em both fixed-size and dynamic-size objects (\ref fixedsize "note"). + * + * The first three template parameters are required: + * \tparam _Scalar Numeric type, e.g. float, double, int or std::complex. + * User defined scalar types are supported as well (see \ref user_defined_scalars "here"). + * \tparam _Rows Number of rows, or \b Dynamic + * \tparam _Cols Number of columns, or \b Dynamic + * + * The remaining template parameters are optional -- in most cases you don't have to worry about them. + * \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of either + * \b #AutoAlign or \b #DontAlign. + * The former controls \ref TopicStorageOrders "storage order", and defaults to column-major. The latter controls alignment, which is required + * for vectorization. It defaults to aligning matrices except for fixed sizes that aren't a multiple of the packet size. + * \tparam _MaxRows Maximum number of rows. Defaults to \a _Rows (\ref maxrows "note"). + * \tparam _MaxCols Maximum number of columns. Defaults to \a _Cols (\ref maxrows "note"). + * + * Eigen provides a number of typedefs covering the usual cases. Here are some examples: + * + * \li \c Matrix2d is a 2x2 square matrix of doubles (\c Matrix) + * \li \c Vector4f is a vector of 4 floats (\c Matrix) + * \li \c RowVector3i is a row-vector of 3 ints (\c Matrix) + * + * \li \c MatrixXf is a dynamic-size matrix of floats (\c Matrix) + * \li \c VectorXf is a dynamic-size vector of floats (\c Matrix) + * + * \li \c Matrix2Xf is a partially fixed-size (dynamic-size) matrix of floats (\c Matrix) + * \li \c MatrixX3d is a partially dynamic-size (fixed-size) matrix of double (\c Matrix) + * + * See \link matrixtypedefs this page \endlink for a complete list of predefined \em %Matrix and \em Vector typedefs. + * + * You can access elements of vectors and matrices using normal subscripting: + * + * \code + * Eigen::VectorXd v(10); + * v[0] = 0.1; + * v[1] = 0.2; + * v(0) = 0.3; + * v(1) = 0.4; + * + * Eigen::MatrixXi m(10, 10); + * m(0, 1) = 1; + * m(0, 2) = 2; + * m(0, 3) = 3; + * \endcode + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIX_PLUGIN. + * + * Some notes: + * + *
+ *
\anchor dense Dense versus sparse:
+ *
This %Matrix class handles dense, not sparse matrices and vectors. For sparse matrices and vectors, see the Sparse module. + * + * Dense matrices and vectors are plain usual arrays of coefficients. All the coefficients are stored, in an ordinary contiguous array. + * This is unlike Sparse matrices and vectors where the coefficients are stored as a list of nonzero coefficients.
+ * + *
\anchor fixedsize Fixed-size versus dynamic-size:
+ *
Fixed-size means that the numbers of rows and columns are known are compile-time. In this case, Eigen allocates the array + * of coefficients as a fixed-size array, as a class member. This makes sense for very small matrices, typically up to 4x4, sometimes up + * to 16x16. Larger matrices should be declared as dynamic-size even if one happens to know their size at compile-time. + * + * Dynamic-size means that the numbers of rows or columns are not necessarily known at compile-time. In this case they are runtime + * variables, and the array of coefficients is allocated dynamically on the heap. + * + * Note that \em dense matrices, be they Fixed-size or Dynamic-size, do not expand dynamically in the sense of a std::map. + * If you want this behavior, see the Sparse module.
+ * + *
\anchor maxrows _MaxRows and _MaxCols:
+ *
In most cases, one just leaves these parameters to the default values. + * These parameters mean the maximum size of rows and columns that the matrix may have. They are useful in cases + * when the exact numbers of rows and columns are not known are compile-time, but it is known at compile-time that they cannot + * exceed a certain value. This happens when taking dynamic-size blocks inside fixed-size matrices: in this case _MaxRows and _MaxCols + * are the dimensions of the original matrix, while _Rows and _Cols are Dynamic.
+ *
+ * + * ABI and storage layout + * + * The table below summarizes the ABI of some possible Matrix instances which is fixed thorough the lifetime of Eigen 3. + * + * + * + * + * + * + *
Matrix typeEquivalent C structure
\code Matrix \endcode\code + * struct { + * T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0 + * Eigen::Index rows, cols; + * }; + * \endcode
\code + * Matrix + * Matrix \endcode\code + * struct { + * T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0 + * Eigen::Index size; + * }; + * \endcode
\code Matrix \endcode\code + * struct { + * T data[Rows*Cols]; // with (size_t(data)%A(Rows*Cols*sizeof(T)))==0 + * }; + * \endcode
\code Matrix \endcode\code + * struct { + * T data[MaxRows*MaxCols]; // with (size_t(data)%A(MaxRows*MaxCols*sizeof(T)))==0 + * Eigen::Index rows, cols; + * }; + * \endcode
+ * Note that in this table Rows, Cols, MaxRows and MaxCols are all positive integers. A(S) is defined to the largest possible power-of-two + * smaller to EIGEN_MAX_STATIC_ALIGN_BYTES. + * + * \see MatrixBase for the majority of the API methods for matrices, \ref TopicClassHierarchy, + * \ref TopicStorageOrders + */ + +template +class Matrix + : public PlainObjectBase > +{ + public: + + /** \brief Base class typedef. + * \sa PlainObjectBase + */ + typedef PlainObjectBase Base; + + enum { Options = _Options }; + + EIGEN_DENSE_PUBLIC_INTERFACE(Matrix) + + typedef typename Base::PlainObject PlainObject; + + using Base::base; + using Base::coeffRef; + + /** + * \brief Assigns matrices to each other. + * + * \note This is a special case of the templated operator=. Its purpose is + * to prevent a default operator= from hiding the templated operator=. + * + * \callgraph + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const Matrix& other) + { + return Base::_set(other); + } + + /** \internal + * \brief Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const DenseBase& other) + { + return Base::_set(other); + } + + /* Here, doxygen failed to copy the brief information when using \copydoc */ + + /** + * \brief Copies the generic expression \a other into *this. + * \copydetails DenseBase::operator=(const EigenBase &other) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const EigenBase &other) + { + return Base::operator=(other); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix& operator=(const ReturnByValue& func) + { + return Base::operator=(func); + } + + /** \brief Default constructor. + * + * For fixed-size matrices, does nothing. + * + * For dynamic-size matrices, creates an empty matrix of size 0. Does not allocate any array. Such a matrix + * is called a null matrix. This constructor is the unique way to create null matrices: resizing + * a matrix to 0 is not supported. + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix() : Base() + { + Base::_check_template_params(); + EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + + // FIXME is it still needed + EIGEN_DEVICE_FUNC + explicit Matrix(internal::constructor_without_unaligned_array_assert) + : Base(internal::constructor_without_unaligned_array_assert()) + { Base::_check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED } + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + Matrix(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible::value) + : Base(std::move(other)) + { + Base::_check_template_params(); + } + EIGEN_DEVICE_FUNC + Matrix& operator=(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable::value) + { + other.swap(*this); + return *this; + } +#endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + + // This constructor is for both 1x1 matrices and dynamic vectors + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE explicit Matrix(const T& x) + { + Base::_check_template_params(); + Base::template _init1(x); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const T0& x, const T1& y) + { + Base::_check_template_params(); + Base::template _init2(x, y); + } + #else + /** \brief Constructs a fixed-sized matrix initialized with coefficients starting at \a data */ + EIGEN_DEVICE_FUNC + explicit Matrix(const Scalar *data); + + /** \brief Constructs a vector or row-vector with given dimension. \only_for_vectors + * + * This is useful for dynamic-size vectors. For fixed-size vectors, + * it is redundant to pass these parameters, so one should use the default constructor + * Matrix() instead. + * + * \warning This constructor is disabled for fixed-size \c 1x1 matrices. For instance, + * calling Matrix(1) will call the initialization constructor: Matrix(const Scalar&). + * For fixed-size \c 1x1 matrices it is therefore recommended to use the default + * constructor Matrix() instead, especially when using one of the non standard + * \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives). + */ + EIGEN_STRONG_INLINE explicit Matrix(Index dim); + /** \brief Constructs an initialized 1x1 matrix with the given coefficient */ + Matrix(const Scalar& x); + /** \brief Constructs an uninitialized matrix with \a rows rows and \a cols columns. + * + * This is useful for dynamic-size matrices. For fixed-size matrices, + * it is redundant to pass these parameters, so one should use the default constructor + * Matrix() instead. + * + * \warning This constructor is disabled for fixed-size \c 1x2 and \c 2x1 vectors. For instance, + * calling Matrix2f(2,1) will call the initialization constructor: Matrix(const Scalar& x, const Scalar& y). + * For fixed-size \c 1x2 or \c 2x1 vectors it is therefore recommended to use the default + * constructor Matrix() instead, especially when using one of the non standard + * \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives). + */ + EIGEN_DEVICE_FUNC + Matrix(Index rows, Index cols); + + /** \brief Constructs an initialized 2D vector with given coefficients */ + Matrix(const Scalar& x, const Scalar& y); + #endif + + /** \brief Constructs an initialized 3D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Matrix, 3) + m_storage.data()[0] = x; + m_storage.data()[1] = y; + m_storage.data()[2] = z; + } + /** \brief Constructs an initialized 4D vector with given coefficients */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z, const Scalar& w) + { + Base::_check_template_params(); + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Matrix, 4) + m_storage.data()[0] = x; + m_storage.data()[1] = y; + m_storage.data()[2] = z; + m_storage.data()[3] = w; + } + + + /** \brief Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const Matrix& other) : Base(other) + { } + + /** \brief Copy constructor for generic expressions. + * \sa MatrixBase::operator=(const EigenBase&) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Matrix(const EigenBase &other) + : Base(other.derived()) + { } + + EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); } + + /////////// Geometry module /////////// + + template + EIGEN_DEVICE_FUNC + explicit Matrix(const RotationBase& r); + template + EIGEN_DEVICE_FUNC + Matrix& operator=(const RotationBase& r); + + // allow to extend Matrix outside Eigen + #ifdef EIGEN_MATRIX_PLUGIN + #include EIGEN_MATRIX_PLUGIN + #endif + + protected: + template + friend struct internal::conservative_resize_like_impl; + + using Base::m_storage; +}; + +/** \defgroup matrixtypedefs Global matrix typedefs + * + * \ingroup Core_Module + * + * Eigen defines several typedef shortcuts for most common matrix and vector types. + * + * The general patterns are the following: + * + * \c MatrixSizeType where \c Size can be \c 2,\c 3,\c 4 for fixed size square matrices or \c X for dynamic size, + * and where \c Type can be \c i for integer, \c f for float, \c d for double, \c cf for complex float, \c cd + * for complex double. + * + * For example, \c Matrix3d is a fixed-size 3x3 matrix type of doubles, and \c MatrixXf is a dynamic-size matrix of floats. + * + * There are also \c VectorSizeType and \c RowVectorSizeType which are self-explanatory. For example, \c Vector4cf is + * a fixed-size vector of 4 complex floats. + * + * \sa class Matrix + */ + +#define EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Size, SizeSuffix) \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##SizeSuffix##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Vector##SizeSuffix##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix RowVector##SizeSuffix##TypeSuffix; + +#define EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, Size) \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##Size##X##TypeSuffix; \ +/** \ingroup matrixtypedefs */ \ +typedef Matrix Matrix##X##Size##TypeSuffix; + +#define EIGEN_MAKE_TYPEDEFS_ALL_SIZES(Type, TypeSuffix) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 2, 2) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 3, 3) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, 4, 4) \ +EIGEN_MAKE_TYPEDEFS(Type, TypeSuffix, Dynamic, X) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 2) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 3) \ +EIGEN_MAKE_FIXED_TYPEDEFS(Type, TypeSuffix, 4) + +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(int, i) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(float, f) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(double, d) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(std::complex, cf) +EIGEN_MAKE_TYPEDEFS_ALL_SIZES(std::complex, cd) + +#undef EIGEN_MAKE_TYPEDEFS_ALL_SIZES +#undef EIGEN_MAKE_TYPEDEFS +#undef EIGEN_MAKE_FIXED_TYPEDEFS + +} // end namespace Eigen + +#endif // EIGEN_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h new file mode 100644 index 00000000000..f8bcc8c6f56 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/MatrixBase.h @@ -0,0 +1,530 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2009 Benoit Jacob +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_MATRIXBASE_H +#define EIGEN_MATRIXBASE_H + +namespace Eigen { + +/** \class MatrixBase + * \ingroup Core_Module + * + * \brief Base class for all dense matrices, vectors, and expressions + * + * This class is the base that is inherited by all matrix, vector, and related expression + * types. Most of the Eigen API is contained in this class, and its base classes. Other important + * classes for the Eigen API are Matrix, and VectorwiseOp. + * + * Note that some methods are defined in other modules such as the \ref LU_Module LU module + * for all functions related to matrix inversions. + * + * \tparam Derived is the derived type, e.g. a matrix type, or an expression, etc. + * + * When writing a function taking Eigen objects as argument, if you want your function + * to take as argument any matrix, vector, or expression, just let it take a + * MatrixBase argument. As an example, here is a function printFirstRow which, given + * a matrix, vector, or expression \a x, prints the first row of \a x. + * + * \code + template + void printFirstRow(const Eigen::MatrixBase& x) + { + cout << x.row(0) << endl; + } + * \endcode + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIXBASE_PLUGIN. + * + * \sa \blank \ref TopicClassHierarchy + */ +template class MatrixBase + : public DenseBase +{ + public: +#ifndef EIGEN_PARSED_BY_DOXYGEN + typedef MatrixBase StorageBaseType; + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::StorageIndex StorageIndex; + typedef typename internal::traits::Scalar Scalar; + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + + typedef DenseBase Base; + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + using Base::derived; + using Base::const_cast_derived; + using Base::rows; + using Base::cols; + using Base::size; + using Base::coeff; + using Base::coeffRef; + using Base::lazyAssign; + using Base::eval; + using Base::operator+=; + using Base::operator-=; + using Base::operator*=; + using Base::operator/=; + + typedef typename Base::CoeffReturnType CoeffReturnType; + typedef typename Base::ConstTransposeReturnType ConstTransposeReturnType; + typedef typename Base::RowXpr RowXpr; + typedef typename Base::ColXpr ColXpr; +#endif // not EIGEN_PARSED_BY_DOXYGEN + + + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** type of the equivalent square matrix */ + typedef Matrix SquareMatrixType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + + /** \returns the size of the main diagonal, which is min(rows(),cols()). + * \sa rows(), cols(), SizeAtCompileTime. */ + EIGEN_DEVICE_FUNC + inline Index diagonalSize() const { return (numext::mini)(rows(),cols()); } + + typedef typename Base::PlainObject PlainObject; + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal Represents a matrix with all coefficients equal to one another*/ + typedef CwiseNullaryOp,PlainObject> ConstantReturnType; + /** \internal the return type of MatrixBase::adjoint() */ + typedef typename internal::conditional::IsComplex, + CwiseUnaryOp, ConstTransposeReturnType>, + ConstTransposeReturnType + >::type AdjointReturnType; + /** \internal Return type of eigenvalues() */ + typedef Matrix, internal::traits::ColsAtCompileTime, 1, ColMajor> EigenvaluesReturnType; + /** \internal the return type of identity */ + typedef CwiseNullaryOp,PlainObject> IdentityReturnType; + /** \internal the return type of unit vectors */ + typedef Block, SquareMatrixType>, + internal::traits::RowsAtCompileTime, + internal::traits::ColsAtCompileTime> BasisReturnType; +#endif // not EIGEN_PARSED_BY_DOXYGEN + +#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::MatrixBase +#define EIGEN_DOC_UNARY_ADDONS(X,Y) +# include "../plugins/CommonCwiseUnaryOps.h" +# include "../plugins/CommonCwiseBinaryOps.h" +# include "../plugins/MatrixCwiseUnaryOps.h" +# include "../plugins/MatrixCwiseBinaryOps.h" +# ifdef EIGEN_MATRIXBASE_PLUGIN +# include EIGEN_MATRIXBASE_PLUGIN +# endif +#undef EIGEN_CURRENT_STORAGE_BASE_CLASS +#undef EIGEN_DOC_UNARY_ADDONS + + /** Special case of the template operator=, in order to prevent the compiler + * from generating a default operator= (issue hit with g++ 4.1) + */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const MatrixBase& other); + + // We cannot inherit here via Base::operator= since it is causing + // trouble with MSVC. + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator=(const DenseBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const EigenBase& other); + + template + EIGEN_DEVICE_FUNC + Derived& operator=(const ReturnByValue& other); + + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator+=(const MatrixBase& other); + template + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + Derived& operator-=(const MatrixBase& other); + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const MatrixBase &other) const; + + template + EIGEN_DEVICE_FUNC + const Product + lazyProduct(const MatrixBase &other) const; + + template + Derived& operator*=(const EigenBase& other); + + template + void applyOnTheLeft(const EigenBase& other); + + template + void applyOnTheRight(const EigenBase& other); + + template + EIGEN_DEVICE_FUNC + const Product + operator*(const DiagonalBase &diagonal) const; + + template + EIGEN_DEVICE_FUNC + typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType + dot(const MatrixBase& other) const; + + EIGEN_DEVICE_FUNC RealScalar squaredNorm() const; + EIGEN_DEVICE_FUNC RealScalar norm() const; + RealScalar stableNorm() const; + RealScalar blueNorm() const; + RealScalar hypotNorm() const; + EIGEN_DEVICE_FUNC const PlainObject normalized() const; + EIGEN_DEVICE_FUNC const PlainObject stableNormalized() const; + EIGEN_DEVICE_FUNC void normalize(); + EIGEN_DEVICE_FUNC void stableNormalize(); + + EIGEN_DEVICE_FUNC const AdjointReturnType adjoint() const; + EIGEN_DEVICE_FUNC void adjointInPlace(); + + typedef Diagonal DiagonalReturnType; + EIGEN_DEVICE_FUNC + DiagonalReturnType diagonal(); + + typedef typename internal::add_const >::type ConstDiagonalReturnType; + EIGEN_DEVICE_FUNC + ConstDiagonalReturnType diagonal() const; + + template struct DiagonalIndexReturnType { typedef Diagonal Type; }; + template struct ConstDiagonalIndexReturnType { typedef const Diagonal Type; }; + + template + EIGEN_DEVICE_FUNC + typename DiagonalIndexReturnType::Type diagonal(); + + template + EIGEN_DEVICE_FUNC + typename ConstDiagonalIndexReturnType::Type diagonal() const; + + typedef Diagonal DiagonalDynamicIndexReturnType; + typedef typename internal::add_const >::type ConstDiagonalDynamicIndexReturnType; + + EIGEN_DEVICE_FUNC + DiagonalDynamicIndexReturnType diagonal(Index index); + EIGEN_DEVICE_FUNC + ConstDiagonalDynamicIndexReturnType diagonal(Index index) const; + + template struct TriangularViewReturnType { typedef TriangularView Type; }; + template struct ConstTriangularViewReturnType { typedef const TriangularView Type; }; + + template + EIGEN_DEVICE_FUNC + typename TriangularViewReturnType::Type triangularView(); + template + EIGEN_DEVICE_FUNC + typename ConstTriangularViewReturnType::Type triangularView() const; + + template struct SelfAdjointViewReturnType { typedef SelfAdjointView Type; }; + template struct ConstSelfAdjointViewReturnType { typedef const SelfAdjointView Type; }; + + template + EIGEN_DEVICE_FUNC + typename SelfAdjointViewReturnType::Type selfadjointView(); + template + EIGEN_DEVICE_FUNC + typename ConstSelfAdjointViewReturnType::Type selfadjointView() const; + + const SparseView sparseView(const Scalar& m_reference = Scalar(0), + const typename NumTraits::Real& m_epsilon = NumTraits::dummy_precision()) const; + EIGEN_DEVICE_FUNC static const IdentityReturnType Identity(); + EIGEN_DEVICE_FUNC static const IdentityReturnType Identity(Index rows, Index cols); + EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index size, Index i); + EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index i); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitX(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitY(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitZ(); + EIGEN_DEVICE_FUNC static const BasisReturnType UnitW(); + + EIGEN_DEVICE_FUNC + const DiagonalWrapper asDiagonal() const; + const PermutationWrapper asPermutation() const; + + EIGEN_DEVICE_FUNC + Derived& setIdentity(); + EIGEN_DEVICE_FUNC + Derived& setIdentity(Index rows, Index cols); + + bool isIdentity(const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isDiagonal(const RealScalar& prec = NumTraits::dummy_precision()) const; + + bool isUpperTriangular(const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isLowerTriangular(const RealScalar& prec = NumTraits::dummy_precision()) const; + + template + bool isOrthogonal(const MatrixBase& other, + const RealScalar& prec = NumTraits::dummy_precision()) const; + bool isUnitary(const RealScalar& prec = NumTraits::dummy_precision()) const; + + /** \returns true if each coefficients of \c *this and \a other are all exactly equal. + * \warning When using floating point scalar values you probably should rather use a + * fuzzy comparison such as isApprox() + * \sa isApprox(), operator!= */ + template + EIGEN_DEVICE_FUNC inline bool operator==(const MatrixBase& other) const + { return cwiseEqual(other).all(); } + + /** \returns true if at least one pair of coefficients of \c *this and \a other are not exactly equal to each other. + * \warning When using floating point scalar values you probably should rather use a + * fuzzy comparison such as isApprox() + * \sa isApprox(), operator== */ + template + EIGEN_DEVICE_FUNC inline bool operator!=(const MatrixBase& other) const + { return cwiseNotEqual(other).any(); } + + NoAlias noalias(); + + // TODO forceAlignedAccess is temporarily disabled + // Need to find a nicer workaround. + inline const Derived& forceAlignedAccess() const { return derived(); } + inline Derived& forceAlignedAccess() { return derived(); } + template inline const Derived& forceAlignedAccessIf() const { return derived(); } + template inline Derived& forceAlignedAccessIf() { return derived(); } + + EIGEN_DEVICE_FUNC Scalar trace() const; + + template EIGEN_DEVICE_FUNC RealScalar lpNorm() const; + + EIGEN_DEVICE_FUNC MatrixBase& matrix() { return *this; } + EIGEN_DEVICE_FUNC const MatrixBase& matrix() const { return *this; } + + /** \returns an \link Eigen::ArrayBase Array \endlink expression of this matrix + * \sa ArrayBase::matrix() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE ArrayWrapper array() { return ArrayWrapper(derived()); } + /** \returns a const \link Eigen::ArrayBase Array \endlink expression of this matrix + * \sa ArrayBase::matrix() */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const ArrayWrapper array() const { return ArrayWrapper(derived()); } + +/////////// LU module /////////// + + inline const FullPivLU fullPivLu() const; + inline const PartialPivLU partialPivLu() const; + + inline const PartialPivLU lu() const; + + inline const Inverse inverse() const; + + template + inline void computeInverseAndDetWithCheck( + ResultType& inverse, + typename ResultType::Scalar& determinant, + bool& invertible, + const RealScalar& absDeterminantThreshold = NumTraits::dummy_precision() + ) const; + template + inline void computeInverseWithCheck( + ResultType& inverse, + bool& invertible, + const RealScalar& absDeterminantThreshold = NumTraits::dummy_precision() + ) const; + Scalar determinant() const; + +/////////// Cholesky module /////////// + + inline const LLT llt() const; + inline const LDLT ldlt() const; + +/////////// QR module /////////// + + inline const HouseholderQR householderQr() const; + inline const ColPivHouseholderQR colPivHouseholderQr() const; + inline const FullPivHouseholderQR fullPivHouseholderQr() const; + inline const CompleteOrthogonalDecomposition completeOrthogonalDecomposition() const; + +/////////// Eigenvalues module /////////// + + inline EigenvaluesReturnType eigenvalues() const; + inline RealScalar operatorNorm() const; + +/////////// SVD module /////////// + + inline JacobiSVD jacobiSvd(unsigned int computationOptions = 0) const; + inline BDCSVD bdcSvd(unsigned int computationOptions = 0) const; + +/////////// Geometry module /////////// + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /// \internal helper struct to form the return type of the cross product + template struct cross_product_return_type { + typedef typename ScalarBinaryOpTraits::Scalar,typename internal::traits::Scalar>::ReturnType Scalar; + typedef Matrix type; + }; + #endif // EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC +#ifndef EIGEN_PARSED_BY_DOXYGEN + inline typename cross_product_return_type::type +#else + inline PlainObject +#endif + cross(const MatrixBase& other) const; + + template + EIGEN_DEVICE_FUNC + inline PlainObject cross3(const MatrixBase& other) const; + + EIGEN_DEVICE_FUNC + inline PlainObject unitOrthogonal(void) const; + + EIGEN_DEVICE_FUNC + inline Matrix eulerAngles(Index a0, Index a1, Index a2) const; + + // put this as separate enum value to work around possible GCC 4.3 bug (?) + enum { HomogeneousReturnTypeDirection = ColsAtCompileTime==1&&RowsAtCompileTime==1 ? ((internal::traits::Flags&RowMajorBit)==RowMajorBit ? Horizontal : Vertical) + : ColsAtCompileTime==1 ? Vertical : Horizontal }; + typedef Homogeneous HomogeneousReturnType; + EIGEN_DEVICE_FUNC + inline HomogeneousReturnType homogeneous() const; + + enum { + SizeMinusOne = SizeAtCompileTime==Dynamic ? Dynamic : SizeAtCompileTime-1 + }; + typedef Block::ColsAtCompileTime==1 ? SizeMinusOne : 1, + internal::traits::ColsAtCompileTime==1 ? 1 : SizeMinusOne> ConstStartMinusOne; + typedef EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(ConstStartMinusOne,Scalar,quotient) HNormalizedReturnType; + EIGEN_DEVICE_FUNC + inline const HNormalizedReturnType hnormalized() const; + +////////// Householder module /////////// + + void makeHouseholderInPlace(Scalar& tau, RealScalar& beta); + template + void makeHouseholder(EssentialPart& essential, + Scalar& tau, RealScalar& beta) const; + template + void applyHouseholderOnTheLeft(const EssentialPart& essential, + const Scalar& tau, + Scalar* workspace); + template + void applyHouseholderOnTheRight(const EssentialPart& essential, + const Scalar& tau, + Scalar* workspace); + +///////// Jacobi module ///////// + + template + void applyOnTheLeft(Index p, Index q, const JacobiRotation& j); + template + void applyOnTheRight(Index p, Index q, const JacobiRotation& j); + +///////// SparseCore module ///////// + + template + EIGEN_STRONG_INLINE const typename SparseMatrixBase::template CwiseProductDenseReturnType::Type + cwiseProduct(const SparseMatrixBase &other) const + { + return other.cwiseProduct(derived()); + } + +///////// MatrixFunctions module ///////// + + typedef typename internal::stem_function::type StemFunction; +#define EIGEN_MATRIX_FUNCTION(ReturnType, Name, Description) \ + /** \returns an expression of the matrix Description of \c *this. \brief This function requires the unsupported MatrixFunctions module. To compute the coefficient-wise Description use ArrayBase::##Name . */ \ + const ReturnType Name() const; +#define EIGEN_MATRIX_FUNCTION_1(ReturnType, Name, Description, Argument) \ + /** \returns an expression of the matrix Description of \c *this. \brief This function requires the unsupported MatrixFunctions module. To compute the coefficient-wise Description use ArrayBase::##Name . */ \ + const ReturnType Name(Argument) const; + + EIGEN_MATRIX_FUNCTION(MatrixExponentialReturnValue, exp, exponential) + /** \brief Helper function for the unsupported MatrixFunctions module.*/ + const MatrixFunctionReturnValue matrixFunction(StemFunction f) const; + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cosh, hyperbolic cosine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sinh, hyperbolic sine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cos, cosine) + EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sin, sine) + EIGEN_MATRIX_FUNCTION(MatrixSquareRootReturnValue, sqrt, square root) + EIGEN_MATRIX_FUNCTION(MatrixLogarithmReturnValue, log, logarithm) + EIGEN_MATRIX_FUNCTION_1(MatrixPowerReturnValue, pow, power to \c p, const RealScalar& p) + EIGEN_MATRIX_FUNCTION_1(MatrixComplexPowerReturnValue, pow, power to \c p, const std::complex& p) + + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MatrixBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MatrixBase) + + private: + EIGEN_DEVICE_FUNC explicit MatrixBase(int); + EIGEN_DEVICE_FUNC MatrixBase(int,int); + template EIGEN_DEVICE_FUNC explicit MatrixBase(const MatrixBase&); + protected: + // mixing arrays and matrices is not legal + template Derived& operator+=(const ArrayBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} + // mixing arrays and matrices is not legal + template Derived& operator-=(const ArrayBase& ) + {EIGEN_STATIC_ASSERT(std::ptrdiff_t(sizeof(typename OtherDerived::Scalar))==-1,YOU_CANNOT_MIX_ARRAYS_AND_MATRICES); return *this;} +}; + + +/*************************************************************************** +* Implementation of matrix base methods +***************************************************************************/ + +/** replaces \c *this by \c *this * \a other. + * + * \returns a reference to \c *this + * + * Example: \include MatrixBase_applyOnTheRight.cpp + * Output: \verbinclude MatrixBase_applyOnTheRight.out + */ +template +template +inline Derived& +MatrixBase::operator*=(const EigenBase &other) +{ + other.derived().applyThisOnTheRight(derived()); + return derived(); +} + +/** replaces \c *this by \c *this * \a other. It is equivalent to MatrixBase::operator*=(). + * + * Example: \include MatrixBase_applyOnTheRight.cpp + * Output: \verbinclude MatrixBase_applyOnTheRight.out + */ +template +template +inline void MatrixBase::applyOnTheRight(const EigenBase &other) +{ + other.derived().applyThisOnTheRight(derived()); +} + +/** replaces \c *this by \a other * \c *this. + * + * Example: \include MatrixBase_applyOnTheLeft.cpp + * Output: \verbinclude MatrixBase_applyOnTheLeft.out + */ +template +template +inline void MatrixBase::applyOnTheLeft(const EigenBase &other) +{ + other.derived().applyThisOnTheLeft(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_MATRIXBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h new file mode 100644 index 00000000000..13adf070e89 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NestByValue.h @@ -0,0 +1,110 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NESTBYVALUE_H +#define EIGEN_NESTBYVALUE_H + +namespace Eigen { + +namespace internal { +template +struct traits > : public traits +{}; +} + +/** \class NestByValue + * \ingroup Core_Module + * + * \brief Expression which must be nested by value + * + * \tparam ExpressionType the type of the object of which we are requiring nesting-by-value + * + * This class is the return type of MatrixBase::nestByValue() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::nestByValue() + */ +template class NestByValue + : public internal::dense_xpr_base< NestByValue >::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(NestByValue) + + EIGEN_DEVICE_FUNC explicit inline NestByValue(const ExpressionType& matrix) : m_expression(matrix) {} + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_expression.rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_expression.cols(); } + EIGEN_DEVICE_FUNC inline Index outerStride() const { return m_expression.outerStride(); } + EIGEN_DEVICE_FUNC inline Index innerStride() const { return m_expression.innerStride(); } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index row, Index col) const + { + return m_expression.coeff(row, col); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index row, Index col) + { + return m_expression.const_cast_derived().coeffRef(row, col); + } + + EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index index) const + { + return m_expression.coeff(index); + } + + EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index index) + { + return m_expression.const_cast_derived().coeffRef(index); + } + + template + inline const PacketScalar packet(Index row, Index col) const + { + return m_expression.template packet(row, col); + } + + template + inline void writePacket(Index row, Index col, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(row, col, x); + } + + template + inline const PacketScalar packet(Index index) const + { + return m_expression.template packet(index); + } + + template + inline void writePacket(Index index, const PacketScalar& x) + { + m_expression.const_cast_derived().template writePacket(index, x); + } + + EIGEN_DEVICE_FUNC operator const ExpressionType&() const { return m_expression; } + + protected: + const ExpressionType m_expression; +}; + +/** \returns an expression of the temporary version of *this. + */ +template +inline const NestByValue +DenseBase::nestByValue() const +{ + return NestByValue(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_NESTBYVALUE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h new file mode 100644 index 00000000000..33908010b4f --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NoAlias.h @@ -0,0 +1,108 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NOALIAS_H +#define EIGEN_NOALIAS_H + +namespace Eigen { + +/** \class NoAlias + * \ingroup Core_Module + * + * \brief Pseudo expression providing an operator = assuming no aliasing + * + * \tparam ExpressionType the type of the object on which to do the lazy assignment + * + * This class represents an expression with special assignment operators + * assuming no aliasing between the target expression and the source expression. + * More precisely it alloas to bypass the EvalBeforeAssignBit flag of the source expression. + * It is the return type of MatrixBase::noalias() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::noalias() + */ +template class StorageBase> +class NoAlias +{ + public: + typedef typename ExpressionType::Scalar Scalar; + + explicit NoAlias(ExpressionType& expression) : m_expression(expression) {} + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::assign_op()); + return m_expression; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator+=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::add_assign_op()); + return m_expression; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE ExpressionType& operator-=(const StorageBase& other) + { + call_assignment_no_alias(m_expression, other.derived(), internal::sub_assign_op()); + return m_expression; + } + + EIGEN_DEVICE_FUNC + ExpressionType& expression() const + { + return m_expression; + } + + protected: + ExpressionType& m_expression; +}; + +/** \returns a pseudo expression of \c *this with an operator= assuming + * no aliasing between \c *this and the source expression. + * + * More precisely, noalias() allows to bypass the EvalBeforeAssignBit flag. + * Currently, even though several expressions may alias, only product + * expressions have this flag. Therefore, noalias() is only usefull when + * the source expression contains a matrix product. + * + * Here are some examples where noalias is usefull: + * \code + * D.noalias() = A * B; + * D.noalias() += A.transpose() * B; + * D.noalias() -= 2 * A * B.adjoint(); + * \endcode + * + * On the other hand the following example will lead to a \b wrong result: + * \code + * A.noalias() = A * B; + * \endcode + * because the result matrix A is also an operand of the matrix product. Therefore, + * there is no alternative than evaluating A * B in a temporary, that is the default + * behavior when you write: + * \code + * A = A * B; + * \endcode + * + * \sa class NoAlias + */ +template +NoAlias MatrixBase::noalias() +{ + return NoAlias(derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_NOALIAS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h new file mode 100644 index 00000000000..daf48987898 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/NumTraits.h @@ -0,0 +1,248 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_NUMTRAITS_H +#define EIGEN_NUMTRAITS_H + +namespace Eigen { + +namespace internal { + +// default implementation of digits10(), based on numeric_limits if specialized, +// 0 for integer types, and log10(epsilon()) otherwise. +template< typename T, + bool use_numeric_limits = std::numeric_limits::is_specialized, + bool is_integer = NumTraits::IsInteger> +struct default_digits10_impl +{ + static int run() { return std::numeric_limits::digits10; } +}; + +template +struct default_digits10_impl // Floating point +{ + static int run() { + using std::log10; + using std::ceil; + typedef typename NumTraits::Real Real; + return int(ceil(-log10(NumTraits::epsilon()))); + } +}; + +template +struct default_digits10_impl // Integer +{ + static int run() { return 0; } +}; + +} // end namespace internal + +/** \class NumTraits + * \ingroup Core_Module + * + * \brief Holds information about the various numeric (i.e. scalar) types allowed by Eigen. + * + * \tparam T the numeric type at hand + * + * This class stores enums, typedefs and static methods giving information about a numeric type. + * + * The provided data consists of: + * \li A typedef \c Real, giving the "real part" type of \a T. If \a T is already real, + * then \c Real is just a typedef to \a T. If \a T is \c std::complex then \c Real + * is a typedef to \a U. + * \li A typedef \c NonInteger, giving the type that should be used for operations producing non-integral values, + * such as quotients, square roots, etc. If \a T is a floating-point type, then this typedef just gives + * \a T again. Note however that many Eigen functions such as internal::sqrt simply refuse to + * take integers. Outside of a few cases, Eigen doesn't do automatic type promotion. Thus, this typedef is + * only intended as a helper for code that needs to explicitly promote types. + * \li A typedef \c Literal giving the type to use for numeric literals such as "2" or "0.5". For instance, for \c std::complex, Literal is defined as \c U. + * Of course, this type must be fully compatible with \a T. In doubt, just use \a T here. + * \li A typedef \a Nested giving the type to use to nest a value inside of the expression tree. If you don't know what + * this means, just use \a T here. + * \li An enum value \a IsComplex. It is equal to 1 if \a T is a \c std::complex + * type, and to 0 otherwise. + * \li An enum value \a IsInteger. It is equal to \c 1 if \a T is an integer type such as \c int, + * and to \c 0 otherwise. + * \li Enum values ReadCost, AddCost and MulCost representing a rough estimate of the number of CPU cycles needed + * to by move / add / mul instructions respectively, assuming the data is already stored in CPU registers. + * Stay vague here. No need to do architecture-specific stuff. + * \li An enum value \a IsSigned. It is equal to \c 1 if \a T is a signed type and to 0 if \a T is unsigned. + * \li An enum value \a RequireInitialization. It is equal to \c 1 if the constructor of the numeric type \a T must + * be called, and to 0 if it is safe not to call it. Default is 0 if \a T is an arithmetic type, and 1 otherwise. + * \li An epsilon() function which, unlike std::numeric_limits::epsilon(), + * it returns a \a Real instead of a \a T. + * \li A dummy_precision() function returning a weak epsilon value. It is mainly used as a default + * value by the fuzzy comparison operators. + * \li highest() and lowest() functions returning the highest and lowest possible values respectively. + * \li digits10() function returning the number of decimal digits that can be represented without change. This is + * the analogue of std::numeric_limits::digits10 + * which is used as the default implementation if specialized. + */ + +template struct GenericNumTraits +{ + enum { + IsInteger = std::numeric_limits::is_integer, + IsSigned = std::numeric_limits::is_signed, + IsComplex = 0, + RequireInitialization = internal::is_arithmetic::value ? 0 : 1, + ReadCost = 1, + AddCost = 1, + MulCost = 1 + }; + + typedef T Real; + typedef typename internal::conditional< + IsInteger, + typename internal::conditional::type, + T + >::type NonInteger; + typedef T Nested; + typedef T Literal; + + EIGEN_DEVICE_FUNC + static inline Real epsilon() + { + return numext::numeric_limits::epsilon(); + } + + EIGEN_DEVICE_FUNC + static inline int digits10() + { + return internal::default_digits10_impl::run(); + } + + EIGEN_DEVICE_FUNC + static inline Real dummy_precision() + { + // make sure to override this for floating-point types + return Real(0); + } + + + EIGEN_DEVICE_FUNC + static inline T highest() { + return (numext::numeric_limits::max)(); + } + + EIGEN_DEVICE_FUNC + static inline T lowest() { + return IsInteger ? (numext::numeric_limits::min)() : (-(numext::numeric_limits::max)()); + } + + EIGEN_DEVICE_FUNC + static inline T infinity() { + return numext::numeric_limits::infinity(); + } + + EIGEN_DEVICE_FUNC + static inline T quiet_NaN() { + return numext::numeric_limits::quiet_NaN(); + } +}; + +template struct NumTraits : GenericNumTraits +{}; + +template<> struct NumTraits + : GenericNumTraits +{ + EIGEN_DEVICE_FUNC + static inline float dummy_precision() { return 1e-5f; } +}; + +template<> struct NumTraits : GenericNumTraits +{ + EIGEN_DEVICE_FUNC + static inline double dummy_precision() { return 1e-12; } +}; + +template<> struct NumTraits + : GenericNumTraits +{ + static inline long double dummy_precision() { return 1e-15l; } +}; + +template struct NumTraits > + : GenericNumTraits > +{ + typedef _Real Real; + typedef typename NumTraits<_Real>::Literal Literal; + enum { + IsComplex = 1, + RequireInitialization = NumTraits<_Real>::RequireInitialization, + ReadCost = 2 * NumTraits<_Real>::ReadCost, + AddCost = 2 * NumTraits::AddCost, + MulCost = 4 * NumTraits::MulCost + 2 * NumTraits::AddCost + }; + + EIGEN_DEVICE_FUNC + static inline Real epsilon() { return NumTraits::epsilon(); } + EIGEN_DEVICE_FUNC + static inline Real dummy_precision() { return NumTraits::dummy_precision(); } + EIGEN_DEVICE_FUNC + static inline int digits10() { return NumTraits::digits10(); } +}; + +template +struct NumTraits > +{ + typedef Array ArrayType; + typedef typename NumTraits::Real RealScalar; + typedef Array Real; + typedef typename NumTraits::NonInteger NonIntegerScalar; + typedef Array NonInteger; + typedef ArrayType & Nested; + typedef typename NumTraits::Literal Literal; + + enum { + IsComplex = NumTraits::IsComplex, + IsInteger = NumTraits::IsInteger, + IsSigned = NumTraits::IsSigned, + RequireInitialization = 1, + ReadCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::ReadCost, + AddCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::AddCost, + MulCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits::MulCost + }; + + EIGEN_DEVICE_FUNC + static inline RealScalar epsilon() { return NumTraits::epsilon(); } + EIGEN_DEVICE_FUNC + static inline RealScalar dummy_precision() { return NumTraits::dummy_precision(); } + + static inline int digits10() { return NumTraits::digits10(); } +}; + +template<> struct NumTraits + : GenericNumTraits +{ + enum { + RequireInitialization = 1, + ReadCost = HugeCost, + AddCost = HugeCost, + MulCost = HugeCost + }; + + static inline int digits10() { return 0; } + +private: + static inline std::string epsilon(); + static inline std::string dummy_precision(); + static inline std::string lowest(); + static inline std::string highest(); + static inline std::string infinity(); + static inline std::string quiet_NaN(); +}; + +// Empty specialization for void to allow template specialization based on NumTraits::Real with T==void and SFINAE. +template<> struct NumTraits {}; + +} // end namespace Eigen + +#endif // EIGEN_NUMTRAITS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h new file mode 100644 index 00000000000..47c06ba7706 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PermutationMatrix.h @@ -0,0 +1,605 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009 Benoit Jacob +// Copyright (C) 2009-2015 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PERMUTATIONMATRIX_H +#define EIGEN_PERMUTATIONMATRIX_H + +namespace Eigen { + +namespace internal { + +enum PermPermProduct_t {PermPermProduct}; + +} // end namespace internal + +/** \class PermutationBase + * \ingroup Core_Module + * + * \brief Base class for permutations + * + * \tparam Derived the derived class + * + * This class is the base class for all expressions representing a permutation matrix, + * internally stored as a vector of integers. + * The convention followed here is that if \f$ \sigma \f$ is a permutation, the corresponding permutation matrix + * \f$ P_\sigma \f$ is such that if \f$ (e_1,\ldots,e_p) \f$ is the canonical basis, we have: + * \f[ P_\sigma(e_i) = e_{\sigma(i)}. \f] + * This convention ensures that for any two permutations \f$ \sigma, \tau \f$, we have: + * \f[ P_{\sigma\circ\tau} = P_\sigma P_\tau. \f] + * + * Permutation matrices are square and invertible. + * + * Notice that in addition to the member functions and operators listed here, there also are non-member + * operator* to multiply any kind of permutation object with any kind of matrix expression (MatrixBase) + * on either side. + * + * \sa class PermutationMatrix, class PermutationWrapper + */ +template +class PermutationBase : public EigenBase +{ + typedef internal::traits Traits; + typedef EigenBase Base; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + enum { + Flags = Traits::Flags, + RowsAtCompileTime = Traits::RowsAtCompileTime, + ColsAtCompileTime = Traits::ColsAtCompileTime, + MaxRowsAtCompileTime = Traits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = Traits::MaxColsAtCompileTime + }; + typedef typename Traits::StorageIndex StorageIndex; + typedef Matrix + DenseMatrixType; + typedef PermutationMatrix + PlainPermutationType; + typedef PlainPermutationType PlainObject; + using Base::derived; + typedef Inverse InverseReturnType; + typedef void Scalar; + #endif + + /** Copies the other permutation into *this */ + template + Derived& operator=(const PermutationBase& other) + { + indices() = other.indices(); + return derived(); + } + + /** Assignment from the Transpositions \a tr */ + template + Derived& operator=(const TranspositionsBase& tr) + { + setIdentity(tr.size()); + for(Index k=size()-1; k>=0; --k) + applyTranspositionOnTheRight(k,tr.coeff(k)); + return derived(); + } + + /** \returns the number of rows */ + inline Index rows() const { return Index(indices().size()); } + + /** \returns the number of columns */ + inline Index cols() const { return Index(indices().size()); } + + /** \returns the size of a side of the respective square matrix, i.e., the number of indices */ + inline Index size() const { return Index(indices().size()); } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + void evalTo(MatrixBase& other) const + { + other.setZero(); + for (Index i=0; i=0 && j>=0 && i=0 && j>=0 && i + void assignTranspose(const PermutationBase& other) + { + for (Index i=0; i + void assignProduct(const Lhs& lhs, const Rhs& rhs) + { + eigen_assert(lhs.cols() == rhs.rows()); + for (Index i=0; i + inline PlainPermutationType operator*(const PermutationBase& other) const + { return PlainPermutationType(internal::PermPermProduct, derived(), other.derived()); } + + /** \returns the product of a permutation with another inverse permutation. + * + * \note \blank \note_try_to_help_rvo + */ + template + inline PlainPermutationType operator*(const InverseImpl& other) const + { return PlainPermutationType(internal::PermPermProduct, *this, other.eval()); } + + /** \returns the product of an inverse permutation with another permutation. + * + * \note \blank \note_try_to_help_rvo + */ + template friend + inline PlainPermutationType operator*(const InverseImpl& other, const PermutationBase& perm) + { return PlainPermutationType(internal::PermPermProduct, other.eval(), perm); } + + /** \returns the determinant of the permutation matrix, which is either 1 or -1 depending on the parity of the permutation. + * + * This function is O(\c n) procedure allocating a buffer of \c n booleans. + */ + Index determinant() const + { + Index res = 1; + Index n = size(); + Matrix mask(n); + mask.fill(false); + Index r = 0; + while(r < n) + { + // search for the next seed + while(r=n) + break; + // we got one, let's follow it until we are back to the seed + Index k0 = r++; + mask.coeffRef(k0) = true; + for(Index k=indices().coeff(k0); k!=k0; k=indices().coeff(k)) + { + mask.coeffRef(k) = true; + res = -res; + } + } + return res; + } + + protected: + +}; + +namespace internal { +template +struct traits > + : traits > +{ + typedef PermutationStorage StorageKind; + typedef Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType; + typedef _StorageIndex StorageIndex; + typedef void Scalar; +}; +} + +/** \class PermutationMatrix + * \ingroup Core_Module + * + * \brief Permutation matrix + * + * \tparam SizeAtCompileTime the number of rows/cols, or Dynamic + * \tparam MaxSizeAtCompileTime the maximum number of rows/cols, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it. + * \tparam _StorageIndex the integer type of the indices + * + * This class represents a permutation matrix, internally stored as a vector of integers. + * + * \sa class PermutationBase, class PermutationWrapper, class DiagonalMatrix + */ +template +class PermutationMatrix : public PermutationBase > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + typedef const PermutationMatrix& Nested; + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + typedef typename Traits::StorageIndex StorageIndex; + #endif + + inline PermutationMatrix() + {} + + /** Constructs an uninitialized permutation matrix of given size. + */ + explicit inline PermutationMatrix(Index size) : m_indices(size) + { + eigen_internal_assert(size <= NumTraits::highest()); + } + + /** Copy constructor. */ + template + inline PermutationMatrix(const PermutationBase& other) + : m_indices(other.indices()) {} + + /** Generic constructor from expression of the indices. The indices + * array has the meaning that the permutations sends each integer i to indices[i]. + * + * \warning It is your responsibility to check that the indices array that you passes actually + * describes a permutation, i.e., each value between 0 and n-1 occurs exactly once, where n is the + * array's size. + */ + template + explicit inline PermutationMatrix(const MatrixBase& indices) : m_indices(indices) + {} + + /** Convert the Transpositions \a tr to a permutation matrix */ + template + explicit PermutationMatrix(const TranspositionsBase& tr) + : m_indices(tr.size()) + { + *this = tr; + } + + /** Copies the other permutation into *this */ + template + PermutationMatrix& operator=(const PermutationBase& other) + { + m_indices = other.indices(); + return *this; + } + + /** Assignment from the Transpositions \a tr */ + template + PermutationMatrix& operator=(const TranspositionsBase& tr) + { + return Base::operator=(tr.derived()); + } + + /** const version of indices(). */ + const IndicesType& indices() const { return m_indices; } + /** \returns a reference to the stored array representing the permutation. */ + IndicesType& indices() { return m_indices; } + + + /**** multiplication helpers to hopefully get RVO ****/ + +#ifndef EIGEN_PARSED_BY_DOXYGEN + template + PermutationMatrix(const InverseImpl& other) + : m_indices(other.derived().nestedExpression().size()) + { + eigen_internal_assert(m_indices.size() <= NumTraits::highest()); + StorageIndex end = StorageIndex(m_indices.size()); + for (StorageIndex i=0; i + PermutationMatrix(internal::PermPermProduct_t, const Lhs& lhs, const Rhs& rhs) + : m_indices(lhs.indices().size()) + { + Base::assignProduct(lhs,rhs); + } +#endif + + protected: + + IndicesType m_indices; +}; + + +namespace internal { +template +struct traits,_PacketAccess> > + : traits > +{ + typedef PermutationStorage StorageKind; + typedef Map, _PacketAccess> IndicesType; + typedef _StorageIndex StorageIndex; + typedef void Scalar; +}; +} + +template +class Map,_PacketAccess> + : public PermutationBase,_PacketAccess> > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + typedef typename IndicesType::Scalar StorageIndex; + #endif + + inline Map(const StorageIndex* indicesPtr) + : m_indices(indicesPtr) + {} + + inline Map(const StorageIndex* indicesPtr, Index size) + : m_indices(indicesPtr,size) + {} + + /** Copies the other permutation into *this */ + template + Map& operator=(const PermutationBase& other) + { return Base::operator=(other.derived()); } + + /** Assignment from the Transpositions \a tr */ + template + Map& operator=(const TranspositionsBase& tr) + { return Base::operator=(tr.derived()); } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + Map& operator=(const Map& other) + { + m_indices = other.m_indices; + return *this; + } + #endif + + /** const version of indices(). */ + const IndicesType& indices() const { return m_indices; } + /** \returns a reference to the stored array representing the permutation. */ + IndicesType& indices() { return m_indices; } + + protected: + + IndicesType m_indices; +}; + +template class TranspositionsWrapper; +namespace internal { +template +struct traits > +{ + typedef PermutationStorage StorageKind; + typedef void Scalar; + typedef typename _IndicesType::Scalar StorageIndex; + typedef _IndicesType IndicesType; + enum { + RowsAtCompileTime = _IndicesType::SizeAtCompileTime, + ColsAtCompileTime = _IndicesType::SizeAtCompileTime, + MaxRowsAtCompileTime = IndicesType::MaxSizeAtCompileTime, + MaxColsAtCompileTime = IndicesType::MaxSizeAtCompileTime, + Flags = 0 + }; +}; +} + +/** \class PermutationWrapper + * \ingroup Core_Module + * + * \brief Class to view a vector of integers as a permutation matrix + * + * \tparam _IndicesType the type of the vector of integer (can be any compatible expression) + * + * This class allows to view any vector expression of integers as a permutation matrix. + * + * \sa class PermutationBase, class PermutationMatrix + */ +template +class PermutationWrapper : public PermutationBase > +{ + typedef PermutationBase Base; + typedef internal::traits Traits; + public: + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename Traits::IndicesType IndicesType; + #endif + + inline PermutationWrapper(const IndicesType& indices) + : m_indices(indices) + {} + + /** const version of indices(). */ + const typename internal::remove_all::type& + indices() const { return m_indices; } + + protected: + + typename IndicesType::Nested m_indices; +}; + + +/** \returns the matrix with the permutation applied to the columns. + */ +template +EIGEN_DEVICE_FUNC +const Product +operator*(const MatrixBase &matrix, + const PermutationBase& permutation) +{ + return Product + (matrix.derived(), permutation.derived()); +} + +/** \returns the matrix with the permutation applied to the rows. + */ +template +EIGEN_DEVICE_FUNC +const Product +operator*(const PermutationBase &permutation, + const MatrixBase& matrix) +{ + return Product + (permutation.derived(), matrix.derived()); +} + + +template +class InverseImpl + : public EigenBase > +{ + typedef typename PermutationType::PlainPermutationType PlainPermutationType; + typedef internal::traits PermTraits; + protected: + InverseImpl() {} + public: + typedef Inverse InverseType; + using EigenBase >::derived; + + #ifndef EIGEN_PARSED_BY_DOXYGEN + typedef typename PermutationType::DenseMatrixType DenseMatrixType; + enum { + RowsAtCompileTime = PermTraits::RowsAtCompileTime, + ColsAtCompileTime = PermTraits::ColsAtCompileTime, + MaxRowsAtCompileTime = PermTraits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = PermTraits::MaxColsAtCompileTime + }; + #endif + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + void evalTo(MatrixBase& other) const + { + other.setZero(); + for (Index i=0; i friend + const Product + operator*(const MatrixBase& matrix, const InverseType& trPerm) + { + return Product(matrix.derived(), trPerm.derived()); + } + + /** \returns the matrix with the inverse permutation applied to the rows. + */ + template + const Product + operator*(const MatrixBase& matrix) const + { + return Product(derived(), matrix.derived()); + } +}; + +template +const PermutationWrapper MatrixBase::asPermutation() const +{ + return derived(); +} + +namespace internal { + +template<> struct AssignmentKind { typedef EigenBase2EigenBase Kind; }; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_PERMUTATIONMATRIX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h new file mode 100644 index 00000000000..0f3632cfd14 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/PlainObjectBase.h @@ -0,0 +1,1037 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2009 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_DENSESTORAGEBASE_H +#define EIGEN_DENSESTORAGEBASE_H + +#if defined(EIGEN_INITIALIZE_MATRICES_BY_ZERO) +# define EIGEN_INITIALIZE_COEFFS +# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED for(int i=0;i::quiet_NaN(); +#else +# undef EIGEN_INITIALIZE_COEFFS +# define EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED +#endif + +namespace Eigen { + +namespace internal { + +template struct check_rows_cols_for_overflow { + template + EIGEN_DEVICE_FUNC + static EIGEN_ALWAYS_INLINE void run(Index, Index) + { + } +}; + +template<> struct check_rows_cols_for_overflow { + template + EIGEN_DEVICE_FUNC + static EIGEN_ALWAYS_INLINE void run(Index rows, Index cols) + { + // http://hg.mozilla.org/mozilla-central/file/6c8a909977d3/xpcom/ds/CheckedInt.h#l242 + // we assume Index is signed + Index max_index = (std::size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed + bool error = (rows == 0 || cols == 0) ? false + : (rows > max_index / cols); + if (error) + throw_std_bad_alloc(); + } +}; + +template +struct conservative_resize_like_impl; + +template struct matrix_swap_impl; + +} // end namespace internal + +#ifdef EIGEN_PARSED_BY_DOXYGEN +namespace doxygen { + +// This is a workaround to doxygen not being able to understand the inheritance logic +// when it is hidden by the dense_xpr_base helper struct. +// Moreover, doxygen fails to include members that are not documented in the declaration body of +// MatrixBase if we inherits MatrixBase >, +// this is why we simply inherits MatrixBase, though this does not make sense. + +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template struct dense_xpr_base_dispatcher; +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template +struct dense_xpr_base_dispatcher > + : public MatrixBase {}; +/** This class is just a workaround for Doxygen and it does not not actually exist. */ +template +struct dense_xpr_base_dispatcher > + : public ArrayBase {}; + +} // namespace doxygen + +/** \class PlainObjectBase + * \ingroup Core_Module + * \brief %Dense storage base class for matrices and arrays. + * + * This class can be extended with the help of the plugin mechanism described on the page + * \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_PLAINOBJECTBASE_PLUGIN. + * + * \tparam Derived is the derived type, e.g., a Matrix or Array + * + * \sa \ref TopicClassHierarchy + */ +template +class PlainObjectBase : public doxygen::dense_xpr_base_dispatcher +#else +template +class PlainObjectBase : public internal::dense_xpr_base::type +#endif +{ + public: + enum { Options = internal::traits::Options }; + typedef typename internal::dense_xpr_base::type Base; + + typedef typename internal::traits::StorageKind StorageKind; + typedef typename internal::traits::Scalar Scalar; + + typedef typename internal::packet_traits::type PacketScalar; + typedef typename NumTraits::Real RealScalar; + typedef Derived DenseType; + + using Base::RowsAtCompileTime; + using Base::ColsAtCompileTime; + using Base::SizeAtCompileTime; + using Base::MaxRowsAtCompileTime; + using Base::MaxColsAtCompileTime; + using Base::MaxSizeAtCompileTime; + using Base::IsVectorAtCompileTime; + using Base::Flags; + + template friend class Eigen::Map; + friend class Eigen::Map; + typedef Eigen::Map MapType; + friend class Eigen::Map; + typedef const Eigen::Map ConstMapType; +#if EIGEN_MAX_ALIGN_BYTES>0 + // for EIGEN_MAX_ALIGN_BYTES==0, AlignedMax==Unaligned, and many compilers generate warnings for friend-ing a class twice. + friend class Eigen::Map; + friend class Eigen::Map; +#endif + typedef Eigen::Map AlignedMapType; + typedef const Eigen::Map ConstAlignedMapType; + template struct StridedMapType { typedef Eigen::Map type; }; + template struct StridedConstMapType { typedef Eigen::Map type; }; + template struct StridedAlignedMapType { typedef Eigen::Map type; }; + template struct StridedConstAlignedMapType { typedef Eigen::Map type; }; + + protected: + DenseStorage m_storage; + + public: + enum { NeedsToAlign = (SizeAtCompileTime != Dynamic) && (internal::traits::Alignment>0) }; + EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) + + EIGEN_DEVICE_FUNC + Base& base() { return *static_cast(this); } + EIGEN_DEVICE_FUNC + const Base& base() const { return *static_cast(this); } + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index rows() const { return m_storage.rows(); } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Index cols() const { return m_storage.cols(); } + + /** This is an overloaded version of DenseCoeffsBase::coeff(Index,Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeff(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeff(Index rowId, Index colId) const + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeff(Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeff(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeff(Index index) const + { + return m_storage.data()[index]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeffRef(Index,Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeffRef(Index,Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index rowId, Index colId) + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is an overloaded version of DenseCoeffsBase::coeffRef(Index) const + * provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts. + * + * See DenseCoeffsBase::coeffRef(Index) const for details. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Scalar& coeffRef(Index index) + { + return m_storage.data()[index]; + } + + /** This is the const version of coeffRef(Index,Index) which is thus synonym of coeff(Index,Index). + * It is provided for convenience. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeffRef(Index rowId, Index colId) const + { + if(Flags & RowMajorBit) + return m_storage.data()[colId + rowId * m_storage.cols()]; + else // column-major + return m_storage.data()[rowId + colId * m_storage.rows()]; + } + + /** This is the const version of coeffRef(Index) which is thus synonym of coeff(Index). + * It is provided for convenience. */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE const Scalar& coeffRef(Index index) const + { + return m_storage.data()[index]; + } + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const + { + return internal::ploadt + (m_storage.data() + (Flags & RowMajorBit + ? colId + rowId * m_storage.cols() + : rowId + colId * m_storage.rows())); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE PacketScalar packet(Index index) const + { + return internal::ploadt(m_storage.data() + index); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE void writePacket(Index rowId, Index colId, const PacketScalar& val) + { + internal::pstoret + (m_storage.data() + (Flags & RowMajorBit + ? colId + rowId * m_storage.cols() + : rowId + colId * m_storage.rows()), val); + } + + /** \internal */ + template + EIGEN_STRONG_INLINE void writePacket(Index index, const PacketScalar& val) + { + internal::pstoret(m_storage.data() + index, val); + } + + /** \returns a const pointer to the data array of this matrix */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar *data() const + { return m_storage.data(); } + + /** \returns a pointer to the data array of this matrix */ + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar *data() + { return m_storage.data(); } + + /** Resizes \c *this to a \a rows x \a cols matrix. + * + * This method is intended for dynamic-size matrices, although it is legal to call it on any + * matrix as long as fixed dimensions are left unchanged. If you only want to change the number + * of rows and/or of columns, you can use resize(NoChange_t, Index), resize(Index, NoChange_t). + * + * If the current number of coefficients of \c *this exactly matches the + * product \a rows * \a cols, then no memory allocation is performed and + * the current values are left unchanged. In all other cases, including + * shrinking, the data is reallocated and all previous values are lost. + * + * Example: \include Matrix_resize_int_int.cpp + * Output: \verbinclude Matrix_resize_int_int.out + * + * \sa resize(Index) for vectors, resize(NoChange_t, Index), resize(Index, NoChange_t) + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void resize(Index rows, Index cols) + { + eigen_assert( EIGEN_IMPLIES(RowsAtCompileTime!=Dynamic,rows==RowsAtCompileTime) + && EIGEN_IMPLIES(ColsAtCompileTime!=Dynamic,cols==ColsAtCompileTime) + && EIGEN_IMPLIES(RowsAtCompileTime==Dynamic && MaxRowsAtCompileTime!=Dynamic,rows<=MaxRowsAtCompileTime) + && EIGEN_IMPLIES(ColsAtCompileTime==Dynamic && MaxColsAtCompileTime!=Dynamic,cols<=MaxColsAtCompileTime) + && rows>=0 && cols>=0 && "Invalid sizes when resizing a matrix or array."); + internal::check_rows_cols_for_overflow::run(rows, cols); + #ifdef EIGEN_INITIALIZE_COEFFS + Index size = rows*cols; + bool size_changed = size != this->size(); + m_storage.resize(size, rows, cols); + if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + #else + m_storage.resize(rows*cols, rows, cols); + #endif + } + + /** Resizes \c *this to a vector of length \a size + * + * \only_for_vectors. This method does not work for + * partially dynamic matrices when the static dimension is anything other + * than 1. For example it will not work with Matrix. + * + * Example: \include Matrix_resize_int.cpp + * Output: \verbinclude Matrix_resize_int.out + * + * \sa resize(Index,Index), resize(NoChange_t, Index), resize(Index, NoChange_t) + */ + EIGEN_DEVICE_FUNC + inline void resize(Index size) + { + EIGEN_STATIC_ASSERT_VECTOR_ONLY(PlainObjectBase) + eigen_assert(((SizeAtCompileTime == Dynamic && (MaxSizeAtCompileTime==Dynamic || size<=MaxSizeAtCompileTime)) || SizeAtCompileTime == size) && size>=0); + #ifdef EIGEN_INITIALIZE_COEFFS + bool size_changed = size != this->size(); + #endif + if(RowsAtCompileTime == 1) + m_storage.resize(size, 1, size); + else + m_storage.resize(size, size, 1); + #ifdef EIGEN_INITIALIZE_COEFFS + if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + #endif + } + + /** Resizes the matrix, changing only the number of columns. For the parameter of type NoChange_t, just pass the special value \c NoChange + * as in the example below. + * + * Example: \include Matrix_resize_NoChange_int.cpp + * Output: \verbinclude Matrix_resize_NoChange_int.out + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + inline void resize(NoChange_t, Index cols) + { + resize(rows(), cols); + } + + /** Resizes the matrix, changing only the number of rows. For the parameter of type NoChange_t, just pass the special value \c NoChange + * as in the example below. + * + * Example: \include Matrix_resize_int_NoChange.cpp + * Output: \verbinclude Matrix_resize_int_NoChange.out + * + * \sa resize(Index,Index) + */ + EIGEN_DEVICE_FUNC + inline void resize(Index rows, NoChange_t) + { + resize(rows, cols()); + } + + /** Resizes \c *this to have the same dimensions as \a other. + * Takes care of doing all the checking that's needed. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void resizeLike(const EigenBase& _other) + { + const OtherDerived& other = _other.derived(); + internal::check_rows_cols_for_overflow::run(other.rows(), other.cols()); + const Index othersize = other.rows()*other.cols(); + if(RowsAtCompileTime == 1) + { + eigen_assert(other.rows() == 1 || other.cols() == 1); + resize(1, othersize); + } + else if(ColsAtCompileTime == 1) + { + eigen_assert(other.rows() == 1 || other.cols() == 1); + resize(othersize, 1); + } + else resize(other.rows(), other.cols()); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * The method is intended for matrices of dynamic size. If you only want to change the number + * of rows and/or of columns, you can use conservativeResize(NoChange_t, Index) or + * conservativeResize(Index, NoChange_t). + * + * Matrices are resized relative to the top-left element. In case values need to be + * appended to the matrix they will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index rows, Index cols) + { + internal::conservative_resize_like_impl::run(*this, rows, cols); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * As opposed to conservativeResize(Index rows, Index cols), this version leaves + * the number of columns unchanged. + * + * In case the matrix is growing, new rows will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index rows, NoChange_t) + { + // Note: see the comment in conservativeResize(Index,Index) + conservativeResize(rows, cols()); + } + + /** Resizes the matrix to \a rows x \a cols while leaving old values untouched. + * + * As opposed to conservativeResize(Index rows, Index cols), this version leaves + * the number of rows unchanged. + * + * In case the matrix is growing, new columns will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index cols) + { + // Note: see the comment in conservativeResize(Index,Index) + conservativeResize(rows(), cols); + } + + /** Resizes the vector to \a size while retaining old values. + * + * \only_for_vectors. This method does not work for + * partially dynamic matrices when the static dimension is anything other + * than 1. For example it will not work with Matrix. + * + * When values are appended, they will be uninitialized. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResize(Index size) + { + internal::conservative_resize_like_impl::run(*this, size); + } + + /** Resizes the matrix to \a rows x \a cols of \c other, while leaving old values untouched. + * + * The method is intended for matrices of dynamic size. If you only want to change the number + * of rows and/or of columns, you can use conservativeResize(NoChange_t, Index) or + * conservativeResize(Index, NoChange_t). + * + * Matrices are resized relative to the top-left element. In case values need to be + * appended to the matrix they will copied from \c other. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void conservativeResizeLike(const DenseBase& other) + { + internal::conservative_resize_like_impl::run(*this, other); + } + + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const PlainObjectBase& other) + { + return _set(other); + } + + /** \sa MatrixBase::lazyAssign() */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& lazyAssign(const DenseBase& other) + { + _resize_to_match(other); + return Base::lazyAssign(other.derived()); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const ReturnByValue& func) + { + resize(func.rows(), func.cols()); + return Base::operator=(func); + } + + // Prevent user from trying to instantiate PlainObjectBase objects + // by making all its constructor protected. See bug 1074. + protected: + + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase() : m_storage() + { +// _check_template_params(); +// EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + +#ifndef EIGEN_PARSED_BY_DOXYGEN + // FIXME is it still needed ? + /** \internal */ + EIGEN_DEVICE_FUNC + explicit PlainObjectBase(internal::constructor_without_unaligned_array_assert) + : m_storage(internal::constructor_without_unaligned_array_assert()) + { +// _check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } +#endif + +#if EIGEN_HAS_RVALUE_REFERENCES + EIGEN_DEVICE_FUNC + PlainObjectBase(PlainObjectBase&& other) EIGEN_NOEXCEPT + : m_storage( std::move(other.m_storage) ) + { + } + + EIGEN_DEVICE_FUNC + PlainObjectBase& operator=(PlainObjectBase&& other) EIGEN_NOEXCEPT + { + using std::swap; + swap(m_storage, other.m_storage); + return *this; + } +#endif + + /** Copy constructor */ + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const PlainObjectBase& other) + : Base(), m_storage(other.m_storage) { } + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(Index size, Index rows, Index cols) + : m_storage(size, rows, cols) + { +// _check_template_params(); +// EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED + } + + /** \sa PlainObjectBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase &other) + : m_storage() + { + _check_template_params(); + resizeLike(other); + _set_noalias(other); + } + + /** \sa PlainObjectBase::operator=(const EigenBase&) */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const EigenBase &other) + : m_storage() + { + _check_template_params(); + resizeLike(other); + *this = other.derived(); + } + /** \brief Copy constructor with in-place evaluation */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE PlainObjectBase(const ReturnByValue& other) + { + _check_template_params(); + // FIXME this does not automatically transpose vectors if necessary + resize(other.rows(), other.cols()); + other.evalTo(this->derived()); + } + + public: + + /** \brief Copies the generic expression \a other into *this. + * \copydetails DenseBase::operator=(const EigenBase &other) + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& operator=(const EigenBase &other) + { + _resize_to_match(other); + Base::operator=(other.derived()); + return this->derived(); + } + + /** \name Map + * These are convenience functions returning Map objects. The Map() static functions return unaligned Map objects, + * while the AlignedMap() functions return aligned Map objects and thus should be called only with 16-byte-aligned + * \a data pointers. + * + * Here is an example using strides: + * \include Matrix_Map_stride.cpp + * Output: \verbinclude Matrix_Map_stride.out + * + * \see class Map + */ + //@{ + static inline ConstMapType Map(const Scalar* data) + { return ConstMapType(data); } + static inline MapType Map(Scalar* data) + { return MapType(data); } + static inline ConstMapType Map(const Scalar* data, Index size) + { return ConstMapType(data, size); } + static inline MapType Map(Scalar* data, Index size) + { return MapType(data, size); } + static inline ConstMapType Map(const Scalar* data, Index rows, Index cols) + { return ConstMapType(data, rows, cols); } + static inline MapType Map(Scalar* data, Index rows, Index cols) + { return MapType(data, rows, cols); } + + static inline ConstAlignedMapType MapAligned(const Scalar* data) + { return ConstAlignedMapType(data); } + static inline AlignedMapType MapAligned(Scalar* data) + { return AlignedMapType(data); } + static inline ConstAlignedMapType MapAligned(const Scalar* data, Index size) + { return ConstAlignedMapType(data, size); } + static inline AlignedMapType MapAligned(Scalar* data, Index size) + { return AlignedMapType(data, size); } + static inline ConstAlignedMapType MapAligned(const Scalar* data, Index rows, Index cols) + { return ConstAlignedMapType(data, rows, cols); } + static inline AlignedMapType MapAligned(Scalar* data, Index rows, Index cols) + { return AlignedMapType(data, rows, cols); } + + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, const Stride& stride) + { return typename StridedConstMapType >::type(data, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, const Stride& stride) + { return typename StridedMapType >::type(data, stride); } + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, Index size, const Stride& stride) + { return typename StridedConstMapType >::type(data, size, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, Index size, const Stride& stride) + { return typename StridedMapType >::type(data, size, stride); } + template + static inline typename StridedConstMapType >::type Map(const Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedConstMapType >::type(data, rows, cols, stride); } + template + static inline typename StridedMapType >::type Map(Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedMapType >::type(data, rows, cols, stride); } + + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, stride); } + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, Index size, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, size, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, Index size, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, size, stride); } + template + static inline typename StridedConstAlignedMapType >::type MapAligned(const Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedConstAlignedMapType >::type(data, rows, cols, stride); } + template + static inline typename StridedAlignedMapType >::type MapAligned(Scalar* data, Index rows, Index cols, const Stride& stride) + { return typename StridedAlignedMapType >::type(data, rows, cols, stride); } + //@} + + using Base::setConstant; + EIGEN_DEVICE_FUNC Derived& setConstant(Index size, const Scalar& val); + EIGEN_DEVICE_FUNC Derived& setConstant(Index rows, Index cols, const Scalar& val); + + using Base::setZero; + EIGEN_DEVICE_FUNC Derived& setZero(Index size); + EIGEN_DEVICE_FUNC Derived& setZero(Index rows, Index cols); + + using Base::setOnes; + EIGEN_DEVICE_FUNC Derived& setOnes(Index size); + EIGEN_DEVICE_FUNC Derived& setOnes(Index rows, Index cols); + + using Base::setRandom; + Derived& setRandom(Index size); + Derived& setRandom(Index rows, Index cols); + + #ifdef EIGEN_PLAINOBJECTBASE_PLUGIN + #include EIGEN_PLAINOBJECTBASE_PLUGIN + #endif + + protected: + /** \internal Resizes *this in preparation for assigning \a other to it. + * Takes care of doing all the checking that's needed. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _resize_to_match(const EigenBase& other) + { + #ifdef EIGEN_NO_AUTOMATIC_RESIZING + eigen_assert((this->size()==0 || (IsVectorAtCompileTime ? (this->size() == other.size()) + : (rows() == other.rows() && cols() == other.cols()))) + && "Size mismatch. Automatic resizing is disabled because EIGEN_NO_AUTOMATIC_RESIZING is defined"); + EIGEN_ONLY_USED_FOR_DEBUG(other); + #else + resizeLike(other); + #endif + } + + /** + * \brief Copies the value of the expression \a other into \c *this with automatic resizing. + * + * *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized), + * it will be initialized. + * + * Note that copying a row-vector into a vector (and conversely) is allowed. + * The resizing, if any, is then done in the appropriate way so that row-vectors + * remain row-vectors and vectors remain vectors. + * + * \sa operator=(const MatrixBase&), _set_noalias() + * + * \internal + */ + // aliasing is dealt once in internall::call_assignment + // so at this stage we have to assume aliasing... and resising has to be done later. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& _set(const DenseBase& other) + { + internal::call_assignment(this->derived(), other.derived()); + return this->derived(); + } + + /** \internal Like _set() but additionally makes the assumption that no aliasing effect can happen (which + * is the case when creating a new matrix) so one can enforce lazy evaluation. + * + * \sa operator=(const MatrixBase&), _set() + */ + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE Derived& _set_noalias(const DenseBase& other) + { + // I don't think we need this resize call since the lazyAssign will anyways resize + // and lazyAssign will be called by the assign selector. + //_resize_to_match(other); + // the 'false' below means to enforce lazy evaluation. We don't use lazyAssign() because + // it wouldn't allow to copy a row-vector into a column-vector. + internal::call_assignment_no_alias(this->derived(), other.derived(), internal::assign_op()); + return this->derived(); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if::type* = 0) + { + const bool t0_is_integer_alike = internal::is_valid_index_type::value; + const bool t1_is_integer_alike = internal::is_valid_index_type::value; + EIGEN_STATIC_ASSERT(t0_is_integer_alike && + t1_is_integer_alike, + FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) + resize(rows,cols); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(const T0& val0, const T1& val1, typename internal::enable_if::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2) + m_storage.data()[0] = Scalar(val0); + m_storage.data()[1] = Scalar(val1); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init2(const Index& val0, const Index& val1, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime==2,T1>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2) + m_storage.data()[0] = Scalar(val0); + m_storage.data()[1] = Scalar(val1); + } + + // The argument is convertible to the Index type and we either have a non 1x1 Matrix, or a dynamic-sized Array, + // then the argument is meant to be the size of the object. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(Index size, typename internal::enable_if< (Base::SizeAtCompileTime!=1 || !internal::is_convertible::value) + && ((!internal::is_same::XprKind,ArrayXpr>::value || Base::SizeAtCompileTime==Dynamic)),T>::type* = 0) + { + // NOTE MSVC 2008 complains if we directly put bool(NumTraits::IsInteger) as the EIGEN_STATIC_ASSERT argument. + const bool is_integer_alike = internal::is_valid_index_type::value; + EIGEN_UNUSED_VARIABLE(is_integer_alike); + EIGEN_STATIC_ASSERT(is_integer_alike, + FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) + resize(size); + } + + // We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type can be implicitely converted) + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar& val0, typename internal::enable_if::value,T>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1) + m_storage.data()[0] = val0; + } + + // We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type match the index type) + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Index& val0, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime==1 + && internal::is_convertible::value,T*>::type* = 0) + { + EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1) + m_storage.data()[0] = Scalar(val0); + } + + // Initialize a fixed size matrix from a pointer to raw data + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar* data){ + this->_set_noalias(ConstMapType(data)); + } + + // Initialize an arbitrary matrix from a dense expression + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const DenseBase& other){ + this->_set_noalias(other); + } + + // Initialize an arbitrary matrix from an object convertible to the Derived type. + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Derived& other){ + this->_set_noalias(other); + } + + // Initialize an arbitrary matrix from a generic Eigen expression + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const EigenBase& other){ + this->derived() = other; + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const ReturnByValue& other) + { + resize(other.rows(), other.cols()); + other.evalTo(this->derived()); + } + + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const RotationBase& r) + { + this->derived() = r; + } + + // For fixed-size Array + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Scalar& val0, + typename internal::enable_if< Base::SizeAtCompileTime!=Dynamic + && Base::SizeAtCompileTime!=1 + && internal::is_convertible::value + && internal::is_same::XprKind,ArrayXpr>::value,T>::type* = 0) + { + Base::setConstant(val0); + } + + // For fixed-size Array + template + EIGEN_DEVICE_FUNC + EIGEN_STRONG_INLINE void _init1(const Index& val0, + typename internal::enable_if< (!internal::is_same::value) + && (internal::is_same::value) + && Base::SizeAtCompileTime!=Dynamic + && Base::SizeAtCompileTime!=1 + && internal::is_convertible::value + && internal::is_same::XprKind,ArrayXpr>::value,T*>::type* = 0) + { + Base::setConstant(val0); + } + + template + friend struct internal::matrix_swap_impl; + + public: + +#ifndef EIGEN_PARSED_BY_DOXYGEN + /** \internal + * \brief Override DenseBase::swap() since for dynamic-sized matrices + * of same type it is enough to swap the data pointers. + */ + template + EIGEN_DEVICE_FUNC + void swap(DenseBase & other) + { + enum { SwapPointers = internal::is_same::value && Base::SizeAtCompileTime==Dynamic }; + internal::matrix_swap_impl::run(this->derived(), other.derived()); + } + + /** \internal + * \brief const version forwarded to DenseBase::swap + */ + template + EIGEN_DEVICE_FUNC + void swap(DenseBase const & other) + { Base::swap(other.derived()); } + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE void _check_template_params() + { + EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, (Options&RowMajor)==RowMajor) + && EIGEN_IMPLIES(MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1, (Options&RowMajor)==0) + && ((RowsAtCompileTime == Dynamic) || (RowsAtCompileTime >= 0)) + && ((ColsAtCompileTime == Dynamic) || (ColsAtCompileTime >= 0)) + && ((MaxRowsAtCompileTime == Dynamic) || (MaxRowsAtCompileTime >= 0)) + && ((MaxColsAtCompileTime == Dynamic) || (MaxColsAtCompileTime >= 0)) + && (MaxRowsAtCompileTime == RowsAtCompileTime || RowsAtCompileTime==Dynamic) + && (MaxColsAtCompileTime == ColsAtCompileTime || ColsAtCompileTime==Dynamic) + && (Options & (DontAlign|RowMajor)) == Options), + INVALID_MATRIX_TEMPLATE_PARAMETERS) + } + + enum { IsPlainObjectBase = 1 }; +#endif +}; + +namespace internal { + +template +struct conservative_resize_like_impl +{ + static void run(DenseBase& _this, Index rows, Index cols) + { + if (_this.rows() == rows && _this.cols() == cols) return; + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(Derived) + + if ( ( Derived::IsRowMajor && _this.cols() == cols) || // row-major and we change only the number of rows + (!Derived::IsRowMajor && _this.rows() == rows) ) // column-major and we change only the number of columns + { + internal::check_rows_cols_for_overflow::run(rows, cols); + _this.derived().m_storage.conservativeResize(rows*cols,rows,cols); + } + else + { + // The storage order does not allow us to use reallocation. + typename Derived::PlainObject tmp(rows,cols); + const Index common_rows = numext::mini(rows, _this.rows()); + const Index common_cols = numext::mini(cols, _this.cols()); + tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols); + _this.derived().swap(tmp); + } + } + + static void run(DenseBase& _this, const DenseBase& other) + { + if (_this.rows() == other.rows() && _this.cols() == other.cols()) return; + + // Note: Here is space for improvement. Basically, for conservativeResize(Index,Index), + // neither RowsAtCompileTime or ColsAtCompileTime must be Dynamic. If only one of the + // dimensions is dynamic, one could use either conservativeResize(Index rows, NoChange_t) or + // conservativeResize(NoChange_t, Index cols). For these methods new static asserts like + // EIGEN_STATIC_ASSERT_DYNAMIC_ROWS and EIGEN_STATIC_ASSERT_DYNAMIC_COLS would be good. + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(Derived) + EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(OtherDerived) + + if ( ( Derived::IsRowMajor && _this.cols() == other.cols()) || // row-major and we change only the number of rows + (!Derived::IsRowMajor && _this.rows() == other.rows()) ) // column-major and we change only the number of columns + { + const Index new_rows = other.rows() - _this.rows(); + const Index new_cols = other.cols() - _this.cols(); + _this.derived().m_storage.conservativeResize(other.size(),other.rows(),other.cols()); + if (new_rows>0) + _this.bottomRightCorner(new_rows, other.cols()) = other.bottomRows(new_rows); + else if (new_cols>0) + _this.bottomRightCorner(other.rows(), new_cols) = other.rightCols(new_cols); + } + else + { + // The storage order does not allow us to use reallocation. + typename Derived::PlainObject tmp(other); + const Index common_rows = numext::mini(tmp.rows(), _this.rows()); + const Index common_cols = numext::mini(tmp.cols(), _this.cols()); + tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols); + _this.derived().swap(tmp); + } + } +}; + +// Here, the specialization for vectors inherits from the general matrix case +// to allow calling .conservativeResize(rows,cols) on vectors. +template +struct conservative_resize_like_impl + : conservative_resize_like_impl +{ + using conservative_resize_like_impl::run; + + static void run(DenseBase& _this, Index size) + { + const Index new_rows = Derived::RowsAtCompileTime==1 ? 1 : size; + const Index new_cols = Derived::RowsAtCompileTime==1 ? size : 1; + _this.derived().m_storage.conservativeResize(size,new_rows,new_cols); + } + + static void run(DenseBase& _this, const DenseBase& other) + { + if (_this.rows() == other.rows() && _this.cols() == other.cols()) return; + + const Index num_new_elements = other.size() - _this.size(); + + const Index new_rows = Derived::RowsAtCompileTime==1 ? 1 : other.rows(); + const Index new_cols = Derived::RowsAtCompileTime==1 ? other.cols() : 1; + _this.derived().m_storage.conservativeResize(other.size(),new_rows,new_cols); + + if (num_new_elements > 0) + _this.tail(num_new_elements) = other.tail(num_new_elements); + } +}; + +template +struct matrix_swap_impl +{ + EIGEN_DEVICE_FUNC + static inline void run(MatrixTypeA& a, MatrixTypeB& b) + { + a.base().swap(b); + } +}; + +template +struct matrix_swap_impl +{ + EIGEN_DEVICE_FUNC + static inline void run(MatrixTypeA& a, MatrixTypeB& b) + { + static_cast(a).m_storage.swap(static_cast(b).m_storage); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_DENSESTORAGEBASE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Product.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Product.h new file mode 100644 index 00000000000..676c4802779 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Product.h @@ -0,0 +1,186 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2011 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_PRODUCT_H +#define EIGEN_PRODUCT_H + +namespace Eigen { + +template class ProductImpl; + +namespace internal { + +template +struct traits > +{ + typedef typename remove_all::type LhsCleaned; + typedef typename remove_all::type RhsCleaned; + typedef traits LhsTraits; + typedef traits RhsTraits; + + typedef MatrixXpr XprKind; + + typedef typename ScalarBinaryOpTraits::Scalar, typename traits::Scalar>::ReturnType Scalar; + typedef typename product_promote_storage_type::ret>::ret StorageKind; + typedef typename promote_index_type::type StorageIndex; + + enum { + RowsAtCompileTime = LhsTraits::RowsAtCompileTime, + ColsAtCompileTime = RhsTraits::ColsAtCompileTime, + MaxRowsAtCompileTime = LhsTraits::MaxRowsAtCompileTime, + MaxColsAtCompileTime = RhsTraits::MaxColsAtCompileTime, + + // FIXME: only needed by GeneralMatrixMatrixTriangular + InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsTraits::ColsAtCompileTime, RhsTraits::RowsAtCompileTime), + + // The storage order is somewhat arbitrary here. The correct one will be determined through the evaluator. + Flags = (MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1) ? RowMajorBit + : (MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1) ? 0 + : ( ((LhsTraits::Flags&NoPreferredStorageOrderBit) && (RhsTraits::Flags&RowMajorBit)) + || ((RhsTraits::Flags&NoPreferredStorageOrderBit) && (LhsTraits::Flags&RowMajorBit)) ) ? RowMajorBit + : NoPreferredStorageOrderBit + }; +}; + +} // end namespace internal + +/** \class Product + * \ingroup Core_Module + * + * \brief Expression of the product of two arbitrary matrices or vectors + * + * \tparam _Lhs the type of the left-hand side expression + * \tparam _Rhs the type of the right-hand side expression + * + * This class represents an expression of the product of two arbitrary matrices. + * + * The other template parameters are: + * \tparam Option can be DefaultProduct, AliasFreeProduct, or LazyProduct + * + */ +template +class Product : public ProductImpl<_Lhs,_Rhs,Option, + typename internal::product_promote_storage_type::StorageKind, + typename internal::traits<_Rhs>::StorageKind, + internal::product_type<_Lhs,_Rhs>::ret>::ret> +{ + public: + + typedef _Lhs Lhs; + typedef _Rhs Rhs; + + typedef typename ProductImpl< + Lhs, Rhs, Option, + typename internal::product_promote_storage_type::StorageKind, + typename internal::traits::StorageKind, + internal::product_type::ret>::ret>::Base Base; + EIGEN_GENERIC_PUBLIC_INTERFACE(Product) + + typedef typename internal::ref_selector::type LhsNested; + typedef typename internal::ref_selector::type RhsNested; + typedef typename internal::remove_all::type LhsNestedCleaned; + typedef typename internal::remove_all::type RhsNestedCleaned; + + EIGEN_DEVICE_FUNC Product(const Lhs& lhs, const Rhs& rhs) : m_lhs(lhs), m_rhs(rhs) + { + eigen_assert(lhs.cols() == rhs.rows() + && "invalid matrix product" + && "if you wanted a coeff-wise or a dot product use the respective explicit functions"); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index rows() const { return m_lhs.rows(); } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index cols() const { return m_rhs.cols(); } + + EIGEN_DEVICE_FUNC const LhsNestedCleaned& lhs() const { return m_lhs; } + EIGEN_DEVICE_FUNC const RhsNestedCleaned& rhs() const { return m_rhs; } + + protected: + + LhsNested m_lhs; + RhsNested m_rhs; +}; + +namespace internal { + +template::ret> +class dense_product_base + : public internal::dense_xpr_base >::type +{}; + +/** Convertion to scalar for inner-products */ +template +class dense_product_base + : public internal::dense_xpr_base >::type +{ + typedef Product ProductXpr; + typedef typename internal::dense_xpr_base::type Base; +public: + using Base::derived; + typedef typename Base::Scalar Scalar; + + EIGEN_STRONG_INLINE operator const Scalar() const + { + return internal::evaluator(derived()).coeff(0,0); + } +}; + +} // namespace internal + +// Generic API dispatcher +template +class ProductImpl : public internal::generic_xpr_base, MatrixXpr, StorageKind>::type +{ + public: + typedef typename internal::generic_xpr_base, MatrixXpr, StorageKind>::type Base; +}; + +template +class ProductImpl + : public internal::dense_product_base +{ + typedef Product Derived; + + public: + + typedef typename internal::dense_product_base Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Derived) + protected: + enum { + IsOneByOne = (RowsAtCompileTime == 1 || RowsAtCompileTime == Dynamic) && + (ColsAtCompileTime == 1 || ColsAtCompileTime == Dynamic), + EnableCoeff = IsOneByOne || Option==LazyProduct + }; + + public: + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index row, Index col) const + { + EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS); + eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) ); + + return internal::evaluator(derived()).coeff(row,col); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index i) const + { + EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS); + eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) ); + + return internal::evaluator(derived()).coeff(i); + } + + +}; + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h new file mode 100644 index 00000000000..bce1310c96f --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ProductEvaluators.h @@ -0,0 +1,1138 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2008-2010 Gael Guennebaud +// Copyright (C) 2011 Jitse Niesen +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + + +#ifndef EIGEN_PRODUCTEVALUATORS_H +#define EIGEN_PRODUCTEVALUATORS_H + +namespace Eigen { + +namespace internal { + +/** \internal + * Evaluator of a product expression. + * Since products require special treatments to handle all possible cases, + * we simply deffer the evaluation logic to a product_evaluator class + * which offers more partial specialization possibilities. + * + * \sa class product_evaluator + */ +template +struct evaluator > + : public product_evaluator > +{ + typedef Product XprType; + typedef product_evaluator Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) : Base(xpr) {} +}; + +// Catch "scalar * ( A * B )" and transform it to "(A*scalar) * B" +// TODO we should apply that rule only if that's really helpful +template +struct evaluator_assume_aliasing, + const CwiseNullaryOp, Plain1>, + const Product > > +{ + static const bool value = true; +}; +template +struct evaluator, + const CwiseNullaryOp, Plain1>, + const Product > > + : public evaluator > +{ + typedef CwiseBinaryOp, + const CwiseNullaryOp, Plain1>, + const Product > XprType; + typedef evaluator > Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) + : Base(xpr.lhs().functor().m_other * xpr.rhs().lhs() * xpr.rhs().rhs()) + {} +}; + + +template +struct evaluator, DiagIndex> > + : public evaluator, DiagIndex> > +{ + typedef Diagonal, DiagIndex> XprType; + typedef evaluator, DiagIndex> > Base; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) + : Base(Diagonal, DiagIndex>( + Product(xpr.nestedExpression().lhs(), xpr.nestedExpression().rhs()), + xpr.index() )) + {} +}; + + +// Helper class to perform a matrix product with the destination at hand. +// Depending on the sizes of the factors, there are different evaluation strategies +// as controlled by internal::product_type. +template< typename Lhs, typename Rhs, + typename LhsShape = typename evaluator_traits::Shape, + typename RhsShape = typename evaluator_traits::Shape, + int ProductType = internal::product_type::value> +struct generic_product_impl; + +template +struct evaluator_assume_aliasing > { + static const bool value = true; +}; + +// This is the default evaluator implementation for products: +// It creates a temporary and call generic_product_impl +template +struct product_evaluator, ProductTag, LhsShape, RhsShape> + : public evaluator::PlainObject> +{ + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + typedef evaluator Base; + enum { + Flags = Base::Flags | EvalBeforeNestingBit + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit product_evaluator(const XprType& xpr) + : m_result(xpr.rows(), xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + +// FIXME shall we handle nested_eval here?, +// if so, then we must take care at removing the call to nested_eval in the specializations (e.g., in permutation_matrix_product, transposition_matrix_product, etc.) +// typedef typename internal::nested_eval::type LhsNested; +// typedef typename internal::nested_eval::type RhsNested; +// typedef typename internal::remove_all::type LhsNestedCleaned; +// typedef typename internal::remove_all::type RhsNestedCleaned; +// +// const LhsNested lhs(xpr.lhs()); +// const RhsNested rhs(xpr.rhs()); +// +// generic_product_impl::evalTo(m_result, lhs, rhs); + + generic_product_impl::evalTo(m_result, xpr.lhs(), xpr.rhs()); + } + +protected: + PlainObject m_result; +}; + +// The following three shortcuts are enabled only if the scalar types match excatly. +// TODO: we could enable them for different scalar types when the product is not vectorized. + +// Dense = Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) + { + Index dstRows = src.rows(); + Index dstCols = src.cols(); + if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) + dst.resize(dstRows, dstCols); + // FIXME shall we handle nested_eval here? + generic_product_impl::evalTo(dst, src.lhs(), src.rhs()); + } +}; + +// Dense += Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::add_assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &) + { + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + // FIXME shall we handle nested_eval here? + generic_product_impl::addTo(dst, src.lhs(), src.rhs()); + } +}; + +// Dense -= Product +template< typename DstXprType, typename Lhs, typename Rhs, int Options, typename Scalar> +struct Assignment, internal::sub_assign_op, Dense2Dense, + typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type> +{ + typedef Product SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &) + { + eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); + // FIXME shall we handle nested_eval here? + generic_product_impl::subTo(dst, src.lhs(), src.rhs()); + } +}; + + +// Dense ?= scalar * Product +// TODO we should apply that rule if that's really helpful +// for instance, this is not good for inner products +template< typename DstXprType, typename Lhs, typename Rhs, typename AssignFunc, typename Scalar, typename ScalarBis, typename Plain> +struct Assignment, const CwiseNullaryOp,Plain>, + const Product >, AssignFunc, Dense2Dense> +{ + typedef CwiseBinaryOp, + const CwiseNullaryOp,Plain>, + const Product > SrcXprType; + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const AssignFunc& func) + { + call_assignment_no_alias(dst, (src.lhs().functor().m_other * src.rhs().lhs())*src.rhs().rhs(), func); + } +}; + +//---------------------------------------- +// Catch "Dense ?= xpr + Product<>" expression to save one temporary +// FIXME we could probably enable these rules for any product, i.e., not only Dense and DefaultProduct + +template +struct evaluator_assume_aliasing::Scalar>, const OtherXpr, + const Product >, DenseShape > { + static const bool value = true; +}; + +template +struct evaluator_assume_aliasing::Scalar>, const OtherXpr, + const Product >, DenseShape > { + static const bool value = true; +}; + +template +struct assignment_from_xpr_op_product +{ + template + static EIGEN_STRONG_INLINE + void run(DstXprType &dst, const SrcXprType &src, const InitialFunc& /*func*/) + { + call_assignment_no_alias(dst, src.lhs(), Func1()); + call_assignment_no_alias(dst, src.rhs(), Func2()); + } +}; + +#define EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(ASSIGN_OP,BINOP,ASSIGN_OP2) \ + template< typename DstXprType, typename OtherXpr, typename Lhs, typename Rhs, typename DstScalar, typename SrcScalar, typename OtherScalar,typename ProdScalar> \ + struct Assignment, const OtherXpr, \ + const Product >, internal::ASSIGN_OP, Dense2Dense> \ + : assignment_from_xpr_op_product, internal::ASSIGN_OP, internal::ASSIGN_OP2 > \ + {} + +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(assign_op, scalar_sum_op,add_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(add_assign_op,scalar_sum_op,add_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(sub_assign_op,scalar_sum_op,sub_assign_op); + +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(assign_op, scalar_difference_op,sub_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(add_assign_op,scalar_difference_op,sub_assign_op); +EIGEN_CATCH_ASSIGN_XPR_OP_PRODUCT(sub_assign_op,scalar_difference_op,add_assign_op); + +//---------------------------------------- + +template +struct generic_product_impl +{ + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + dst.coeffRef(0,0) = (lhs.transpose().cwiseProduct(rhs)).sum(); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + dst.coeffRef(0,0) += (lhs.transpose().cwiseProduct(rhs)).sum(); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { dst.coeffRef(0,0) -= (lhs.transpose().cwiseProduct(rhs)).sum(); } +}; + + +/*********************************************************************** +* Implementation of outer dense * dense vector product +***********************************************************************/ + +// Column major result +template +void outer_product_selector_run(Dst& dst, const Lhs &lhs, const Rhs &rhs, const Func& func, const false_type&) +{ + evaluator rhsEval(rhs); + typename nested_eval::type actual_lhs(lhs); + // FIXME if cols is large enough, then it might be useful to make sure that lhs is sequentially stored + // FIXME not very good if rhs is real and lhs complex while alpha is real too + const Index cols = dst.cols(); + for (Index j=0; j +void outer_product_selector_run(Dst& dst, const Lhs &lhs, const Rhs &rhs, const Func& func, const true_type&) +{ + evaluator lhsEval(lhs); + typename nested_eval::type actual_rhs(rhs); + // FIXME if rows is large enough, then it might be useful to make sure that rhs is sequentially stored + // FIXME not very good if lhs is real and rhs complex while alpha is real too + const Index rows = dst.rows(); + for (Index i=0; i +struct generic_product_impl +{ + template struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {}; + typedef typename Product::Scalar Scalar; + + // TODO it would be nice to be able to exploit our *_assign_op functors for that purpose + struct set { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() = src; } }; + struct add { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() += src; } }; + struct sub { template void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() -= src; } }; + struct adds { + Scalar m_scale; + explicit adds(const Scalar& s) : m_scale(s) {} + template void operator()(const Dst& dst, const Src& src) const { + dst.const_cast_derived() += m_scale * src; + } + }; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, set(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, add(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + internal::outer_product_selector_run(dst, lhs, rhs, sub(), is_row_major()); + } + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + internal::outer_product_selector_run(dst, lhs, rhs, adds(alpha), is_row_major()); + } + +}; + + +// This base class provides default implementations for evalTo, addTo, subTo, in terms of scaleAndAddTo +template +struct generic_product_impl_base +{ + typedef typename Product::Scalar Scalar; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { dst.setZero(); scaleAndAddTo(dst, lhs, rhs, Scalar(1)); } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { scaleAndAddTo(dst,lhs, rhs, Scalar(1)); } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); } + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { Derived::scaleAndAddTo(dst,lhs,rhs,alpha); } + +}; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename nested_eval::type LhsNested; + typedef typename nested_eval::type RhsNested; + typedef typename Product::Scalar Scalar; + enum { Side = Lhs::IsVectorAtCompileTime ? OnTheLeft : OnTheRight }; + typedef typename internal::remove_all::type>::type MatrixType; + + template + static EIGEN_STRONG_INLINE void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + LhsNested actual_lhs(lhs); + RhsNested actual_rhs(rhs); + internal::gemv_dense_selector::HasUsableDirectAccess) + >::run(actual_lhs, actual_rhs, dst, alpha); + } +}; + +template +struct generic_product_impl +{ + typedef typename Product::Scalar Scalar; + + template + static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // Same as: dst.noalias() = lhs.lazyProduct(rhs); + // but easier on the compiler side + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::assign_op()); + } + + template + static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // dst.noalias() += lhs.lazyProduct(rhs); + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::add_assign_op()); + } + + template + static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) + { + // dst.noalias() -= lhs.lazyProduct(rhs); + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::sub_assign_op()); + } + + // Catch "dst {,+,-}= (s*A)*B" and evaluate it lazily by moving out the scalar factor: + // dst {,+,-}= s * (A.lazyProduct(B)) + // This is a huge benefit for heap-allocated matrix types as it save one costly allocation. + // For them, this strategy is also faster than simply by-passing the heap allocation through + // stack allocation. + // For fixed sizes matrices, this is less obvious, it is sometimes x2 faster, but sometimes x3 slower, + // and the behavior depends also a lot on the compiler... so let's be conservative and enable them for dynamic-size only, + // that is when coming from generic_product_impl<...,GemmProduct> in file GeneralMatrixMatrix.h + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const CwiseBinaryOp, + const CwiseNullaryOp, Plain1>, Xpr2>& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lhs().functor().m_other * lhs.rhs().lazyProduct(rhs), func); + } + + // Here, we we always have LhsT==Lhs, but we need to make it a template type to make the above + // overload more specialized. + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const LhsT& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), func); + } + + +// template +// static inline void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) +// { dst.noalias() += alpha * lhs.lazyProduct(rhs); } +}; + +// This specialization enforces the use of a coefficient-based evaluation strategy +template +struct generic_product_impl + : generic_product_impl {}; + +// Case 2: Evaluate coeff by coeff +// +// This is mostly taken from CoeffBasedProduct.h +// The main difference is that we add an extra argument to the etor_product_*_impl::run() function +// for the inner dimension of the product, because evaluator object do not know their size. + +template +struct etor_product_coeff_impl; + +template +struct etor_product_packet_impl; + +template +struct product_evaluator, ProductTag, DenseShape, DenseShape> + : evaluator_base > +{ + typedef Product XprType; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + explicit product_evaluator(const XprType& xpr) + : m_lhs(xpr.lhs()), + m_rhs(xpr.rhs()), + m_lhsImpl(m_lhs), // FIXME the creation of the evaluator objects should result in a no-op, but check that! + m_rhsImpl(m_rhs), // Moreover, they are only useful for the packet path, so we could completely disable them when not needed, + // or perhaps declare them on the fly on the packet method... We have experiment to check what's best. + m_innerDim(xpr.lhs().cols()) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::MulCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::AddCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); +#if 0 + std::cerr << "LhsOuterStrideBytes= " << LhsOuterStrideBytes << "\n"; + std::cerr << "RhsOuterStrideBytes= " << RhsOuterStrideBytes << "\n"; + std::cerr << "LhsAlignment= " << LhsAlignment << "\n"; + std::cerr << "RhsAlignment= " << RhsAlignment << "\n"; + std::cerr << "CanVectorizeLhs= " << CanVectorizeLhs << "\n"; + std::cerr << "CanVectorizeRhs= " << CanVectorizeRhs << "\n"; + std::cerr << "CanVectorizeInner= " << CanVectorizeInner << "\n"; + std::cerr << "EvalToRowMajor= " << EvalToRowMajor << "\n"; + std::cerr << "Alignment= " << Alignment << "\n"; + std::cerr << "Flags= " << Flags << "\n"; +#endif + } + + // Everything below here is taken from CoeffBasedProduct.h + + typedef typename internal::nested_eval::type LhsNested; + typedef typename internal::nested_eval::type RhsNested; + + typedef typename internal::remove_all::type LhsNestedCleaned; + typedef typename internal::remove_all::type RhsNestedCleaned; + + typedef evaluator LhsEtorType; + typedef evaluator RhsEtorType; + + enum { + RowsAtCompileTime = LhsNestedCleaned::RowsAtCompileTime, + ColsAtCompileTime = RhsNestedCleaned::ColsAtCompileTime, + InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsNestedCleaned::ColsAtCompileTime, RhsNestedCleaned::RowsAtCompileTime), + MaxRowsAtCompileTime = LhsNestedCleaned::MaxRowsAtCompileTime, + MaxColsAtCompileTime = RhsNestedCleaned::MaxColsAtCompileTime + }; + + typedef typename find_best_packet::type LhsVecPacketType; + typedef typename find_best_packet::type RhsVecPacketType; + + enum { + + LhsCoeffReadCost = LhsEtorType::CoeffReadCost, + RhsCoeffReadCost = RhsEtorType::CoeffReadCost, + CoeffReadCost = InnerSize==0 ? NumTraits::ReadCost + : InnerSize == Dynamic ? HugeCost + : InnerSize * (NumTraits::MulCost + LhsCoeffReadCost + RhsCoeffReadCost) + + (InnerSize - 1) * NumTraits::AddCost, + + Unroll = CoeffReadCost <= EIGEN_UNROLLING_LIMIT, + + LhsFlags = LhsEtorType::Flags, + RhsFlags = RhsEtorType::Flags, + + LhsRowMajor = LhsFlags & RowMajorBit, + RhsRowMajor = RhsFlags & RowMajorBit, + + LhsVecPacketSize = unpacket_traits::size, + RhsVecPacketSize = unpacket_traits::size, + + // Here, we don't care about alignment larger than the usable packet size. + LhsAlignment = EIGEN_PLAIN_ENUM_MIN(LhsEtorType::Alignment,LhsVecPacketSize*int(sizeof(typename LhsNestedCleaned::Scalar))), + RhsAlignment = EIGEN_PLAIN_ENUM_MIN(RhsEtorType::Alignment,RhsVecPacketSize*int(sizeof(typename RhsNestedCleaned::Scalar))), + + SameType = is_same::value, + + CanVectorizeRhs = bool(RhsRowMajor) && (RhsFlags & PacketAccessBit) && (ColsAtCompileTime!=1), + CanVectorizeLhs = (!LhsRowMajor) && (LhsFlags & PacketAccessBit) && (RowsAtCompileTime!=1), + + EvalToRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1 + : (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0 + : (bool(RhsRowMajor) && !CanVectorizeLhs), + + Flags = ((unsigned int)(LhsFlags | RhsFlags) & HereditaryBits & ~RowMajorBit) + | (EvalToRowMajor ? RowMajorBit : 0) + // TODO enable vectorization for mixed types + | (SameType && (CanVectorizeLhs || CanVectorizeRhs) ? PacketAccessBit : 0) + | (XprType::IsVectorAtCompileTime ? LinearAccessBit : 0), + + LhsOuterStrideBytes = int(LhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename LhsNestedCleaned::Scalar)), + RhsOuterStrideBytes = int(RhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename RhsNestedCleaned::Scalar)), + + Alignment = bool(CanVectorizeLhs) ? (LhsOuterStrideBytes<=0 || (int(LhsOuterStrideBytes) % EIGEN_PLAIN_ENUM_MAX(1,LhsAlignment))!=0 ? 0 : LhsAlignment) + : bool(CanVectorizeRhs) ? (RhsOuterStrideBytes<=0 || (int(RhsOuterStrideBytes) % EIGEN_PLAIN_ENUM_MAX(1,RhsAlignment))!=0 ? 0 : RhsAlignment) + : 0, + + /* CanVectorizeInner deserves special explanation. It does not affect the product flags. It is not used outside + * of Product. If the Product itself is not a packet-access expression, there is still a chance that the inner + * loop of the product might be vectorized. This is the meaning of CanVectorizeInner. Since it doesn't affect + * the Flags, it is safe to make this value depend on ActualPacketAccessBit, that doesn't affect the ABI. + */ + CanVectorizeInner = SameType + && LhsRowMajor + && (!RhsRowMajor) + && (LhsFlags & RhsFlags & ActualPacketAccessBit) + && (InnerSize % packet_traits::size == 0) + }; + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index row, Index col) const + { + return (m_lhs.row(row).transpose().cwiseProduct( m_rhs.col(col) )).sum(); + } + + /* Allow index-based non-packet access. It is impossible though to allow index-based packed access, + * which is why we don't set the LinearAccessBit. + * TODO: this seems possible when the result is a vector + */ + EIGEN_DEVICE_FUNC const CoeffReturnType coeff(Index index) const + { + const Index row = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? 0 : index; + const Index col = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? index : 0; + return (m_lhs.row(row).transpose().cwiseProduct( m_rhs.col(col) )).sum(); + } + + template + const PacketType packet(Index row, Index col) const + { + PacketType res; + typedef etor_product_packet_impl PacketImpl; + PacketImpl::run(row, col, m_lhsImpl, m_rhsImpl, m_innerDim, res); + return res; + } + + template + const PacketType packet(Index index) const + { + const Index row = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? 0 : index; + const Index col = (RowsAtCompileTime == 1 || MaxRowsAtCompileTime==1) ? index : 0; + return packet(row,col); + } + +protected: + typename internal::add_const_on_value_type::type m_lhs; + typename internal::add_const_on_value_type::type m_rhs; + + LhsEtorType m_lhsImpl; + RhsEtorType m_rhsImpl; + + // TODO: Get rid of m_innerDim if known at compile time + Index m_innerDim; +}; + +template +struct product_evaluator, LazyCoeffBasedProductMode, DenseShape, DenseShape> + : product_evaluator, CoeffBasedProductMode, DenseShape, DenseShape> +{ + typedef Product XprType; + typedef Product BaseProduct; + typedef product_evaluator Base; + enum { + Flags = Base::Flags | EvalBeforeNestingBit + }; + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(BaseProduct(xpr.lhs(),xpr.rhs())) + {} +}; + +/**************************************** +*** Coeff based product, Packet path *** +****************************************/ + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet &res) + { + etor_product_packet_impl::run(row, col, lhs, rhs, innerDim, res); + res = pmadd(pset1(lhs.coeff(row, Index(UnrollingIndex-1))), rhs.template packet(Index(UnrollingIndex-1), col), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet &res) + { + etor_product_packet_impl::run(row, col, lhs, rhs, innerDim, res); + res = pmadd(lhs.template packet(row, Index(UnrollingIndex-1)), pset1(rhs.coeff(Index(UnrollingIndex-1), col)), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index /*innerDim*/, Packet &res) + { + res = pmul(pset1(lhs.coeff(row, Index(0))),rhs.template packet(Index(0), col)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index /*innerDim*/, Packet &res) + { + res = pmul(lhs.template packet(row, Index(0)), pset1(rhs.coeff(Index(0), col))); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Index /*innerDim*/, Packet &res) + { + res = pset1(typename unpacket_traits::type(0)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index /*row*/, Index /*col*/, const Lhs& /*lhs*/, const Rhs& /*rhs*/, Index /*innerDim*/, Packet &res) + { + res = pset1(typename unpacket_traits::type(0)); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet& res) + { + res = pset1(typename unpacket_traits::type(0)); + for(Index i = 0; i < innerDim; ++i) + res = pmadd(pset1(lhs.coeff(row, i)), rhs.template packet(i, col), res); + } +}; + +template +struct etor_product_packet_impl +{ + static EIGEN_STRONG_INLINE void run(Index row, Index col, const Lhs& lhs, const Rhs& rhs, Index innerDim, Packet& res) + { + res = pset1(typename unpacket_traits::type(0)); + for(Index i = 0; i < innerDim; ++i) + res = pmadd(lhs.template packet(row, i), pset1(rhs.coeff(i, col)), res); + } +}; + + +/*************************************************************************** +* Triangular products +***************************************************************************/ +template +struct triangular_product_impl; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + triangular_product_impl + ::run(dst, lhs.nestedExpression(), rhs, alpha); + } +}; + +template +struct generic_product_impl +: generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + triangular_product_impl::run(dst, lhs, rhs.nestedExpression(), alpha); + } +}; + + +/*************************************************************************** +* SelfAdjoint products +***************************************************************************/ +template +struct selfadjoint_product_impl; + +template +struct generic_product_impl + : generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + selfadjoint_product_impl::run(dst, lhs.nestedExpression(), rhs, alpha); + } +}; + +template +struct generic_product_impl +: generic_product_impl_base > +{ + typedef typename Product::Scalar Scalar; + + template + static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) + { + selfadjoint_product_impl::run(dst, lhs, rhs.nestedExpression(), alpha); + } +}; + + +/*************************************************************************** +* Diagonal products +***************************************************************************/ + +template +struct diagonal_product_evaluator_base + : evaluator_base +{ + typedef typename ScalarBinaryOpTraits::ReturnType Scalar; +public: + enum { + CoeffReadCost = NumTraits::MulCost + evaluator::CoeffReadCost + evaluator::CoeffReadCost, + + MatrixFlags = evaluator::Flags, + DiagFlags = evaluator::Flags, + _StorageOrder = MatrixFlags & RowMajorBit ? RowMajor : ColMajor, + _ScalarAccessOnDiag = !((int(_StorageOrder) == ColMajor && int(ProductOrder) == OnTheLeft) + ||(int(_StorageOrder) == RowMajor && int(ProductOrder) == OnTheRight)), + _SameTypes = is_same::value, + // FIXME currently we need same types, but in the future the next rule should be the one + //_Vectorizable = bool(int(MatrixFlags)&PacketAccessBit) && ((!_PacketOnDiag) || (_SameTypes && bool(int(DiagFlags)&PacketAccessBit))), + _Vectorizable = bool(int(MatrixFlags)&PacketAccessBit) && _SameTypes && (_ScalarAccessOnDiag || (bool(int(DiagFlags)&PacketAccessBit))), + _LinearAccessMask = (MatrixType::RowsAtCompileTime==1 || MatrixType::ColsAtCompileTime==1) ? LinearAccessBit : 0, + Flags = ((HereditaryBits|_LinearAccessMask) & (unsigned int)(MatrixFlags)) | (_Vectorizable ? PacketAccessBit : 0), + Alignment = evaluator::Alignment, + + AsScalarProduct = (DiagonalType::SizeAtCompileTime==1) + || (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::RowsAtCompileTime==1 && ProductOrder==OnTheLeft) + || (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::ColsAtCompileTime==1 && ProductOrder==OnTheRight) + }; + + diagonal_product_evaluator_base(const MatrixType &mat, const DiagonalType &diag) + : m_diagImpl(diag), m_matImpl(mat) + { + EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits::MulCost); + EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost); + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index idx) const + { + if(AsScalarProduct) + return m_diagImpl.coeff(0) * m_matImpl.coeff(idx); + else + return m_diagImpl.coeff(idx) * m_matImpl.coeff(idx); + } + +protected: + template + EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::true_type) const + { + return internal::pmul(m_matImpl.template packet(row, col), + internal::pset1(m_diagImpl.coeff(id))); + } + + template + EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::false_type) const + { + enum { + InnerSize = (MatrixType::Flags & RowMajorBit) ? MatrixType::ColsAtCompileTime : MatrixType::RowsAtCompileTime, + DiagonalPacketLoadMode = EIGEN_PLAIN_ENUM_MIN(LoadMode,((InnerSize%16) == 0) ? int(Aligned16) : int(evaluator::Alignment)) // FIXME hardcoded 16!! + }; + return internal::pmul(m_matImpl.template packet(row, col), + m_diagImpl.template packet(id)); + } + + evaluator m_diagImpl; + evaluator m_matImpl; +}; + +// diagonal * dense +template +struct product_evaluator, ProductTag, DiagonalShape, DenseShape> + : diagonal_product_evaluator_base, OnTheLeft> +{ + typedef diagonal_product_evaluator_base, OnTheLeft> Base; + using Base::m_diagImpl; + using Base::m_matImpl; + using Base::coeff; + typedef typename Base::Scalar Scalar; + + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + + enum { + StorageOrder = int(Rhs::Flags) & RowMajorBit ? RowMajor : ColMajor + }; + + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(xpr.rhs(), xpr.lhs().diagonal()) + { + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const + { + return m_diagImpl.coeff(row) * m_matImpl.coeff(row, col); + } + +#ifndef __CUDACC__ + template + EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const + { + // FIXME: NVCC used to complain about the template keyword, but we have to check whether this is still the case. + // See also similar calls below. + return this->template packet_impl(row,col, row, + typename internal::conditional::type()); + } + + template + EIGEN_STRONG_INLINE PacketType packet(Index idx) const + { + return packet(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx); + } +#endif +}; + +// dense * diagonal +template +struct product_evaluator, ProductTag, DenseShape, DiagonalShape> + : diagonal_product_evaluator_base, OnTheRight> +{ + typedef diagonal_product_evaluator_base, OnTheRight> Base; + using Base::m_diagImpl; + using Base::m_matImpl; + using Base::coeff; + typedef typename Base::Scalar Scalar; + + typedef Product XprType; + typedef typename XprType::PlainObject PlainObject; + + enum { StorageOrder = int(Lhs::Flags) & RowMajorBit ? RowMajor : ColMajor }; + + EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr) + : Base(xpr.lhs(), xpr.rhs().diagonal()) + { + } + + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const + { + return m_matImpl.coeff(row, col) * m_diagImpl.coeff(col); + } + +#ifndef __CUDACC__ + template + EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const + { + return this->template packet_impl(row,col, col, + typename internal::conditional::type()); + } + + template + EIGEN_STRONG_INLINE PacketType packet(Index idx) const + { + return packet(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx); + } +#endif +}; + +/*************************************************************************** +* Products with permutation matrices +***************************************************************************/ + +/** \internal + * \class permutation_matrix_product + * Internal helper class implementing the product between a permutation matrix and a matrix. + * This class is specialized for DenseShape below and for SparseShape in SparseCore/SparsePermutation.h + */ +template +struct permutation_matrix_product; + +template +struct permutation_matrix_product +{ + typedef typename nested_eval::type MatrixType; + typedef typename remove_all::type MatrixTypeCleaned; + + template + static inline void run(Dest& dst, const PermutationType& perm, const ExpressionType& xpr) + { + MatrixType mat(xpr); + const Index n = Side==OnTheLeft ? mat.rows() : mat.cols(); + // FIXME we need an is_same for expression that is not sensitive to constness. For instance + // is_same_xpr, Block >::value should be true. + //if(is_same::value && extract_data(dst) == extract_data(mat)) + if(is_same_dense(dst, mat)) + { + // apply the permutation inplace + Matrix mask(perm.size()); + mask.fill(false); + Index r = 0; + while(r < perm.size()) + { + // search for the next seed + while(r=perm.size()) + break; + // we got one, let's follow it until we are back to the seed + Index k0 = r++; + Index kPrev = k0; + mask.coeffRef(k0) = true; + for(Index k=perm.indices().coeff(k0); k!=k0; k=perm.indices().coeff(k)) + { + Block(dst, k) + .swap(Block + (dst,((Side==OnTheLeft) ^ Transposed) ? k0 : kPrev)); + + mask.coeffRef(k) = true; + kPrev = k; + } + } + } + else + { + for(Index i = 0; i < n; ++i) + { + Block + (dst, ((Side==OnTheLeft) ^ Transposed) ? perm.indices().coeff(i) : i) + + = + + Block + (mat, ((Side==OnTheRight) ^ Transposed) ? perm.indices().coeff(i) : i); + } + } + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, lhs, rhs); + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, rhs, lhs); + } +}; + +template +struct generic_product_impl, Rhs, PermutationShape, MatrixShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Inverse& lhs, const Rhs& rhs) + { + permutation_matrix_product::run(dst, lhs.nestedExpression(), rhs); + } +}; + +template +struct generic_product_impl, MatrixShape, PermutationShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Inverse& rhs) + { + permutation_matrix_product::run(dst, rhs.nestedExpression(), lhs); + } +}; + + +/*************************************************************************** +* Products with transpositions matrices +***************************************************************************/ + +// FIXME could we unify Transpositions and Permutation into a single "shape"?? + +/** \internal + * \class transposition_matrix_product + * Internal helper class implementing the product between a permutation matrix and a matrix. + */ +template +struct transposition_matrix_product +{ + typedef typename nested_eval::type MatrixType; + typedef typename remove_all::type MatrixTypeCleaned; + + template + static inline void run(Dest& dst, const TranspositionType& tr, const ExpressionType& xpr) + { + MatrixType mat(xpr); + typedef typename TranspositionType::StorageIndex StorageIndex; + const Index size = tr.size(); + StorageIndex j = 0; + + if(!is_same_dense(dst,mat)) + dst = mat; + + for(Index k=(Transposed?size-1:0) ; Transposed?k>=0:k +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, lhs, rhs); + } +}; + +template +struct generic_product_impl +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, rhs, lhs); + } +}; + + +template +struct generic_product_impl, Rhs, TranspositionsShape, MatrixShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Transpose& lhs, const Rhs& rhs) + { + transposition_matrix_product::run(dst, lhs.nestedExpression(), rhs); + } +}; + +template +struct generic_product_impl, MatrixShape, TranspositionsShape, ProductTag> +{ + template + static void evalTo(Dest& dst, const Lhs& lhs, const Transpose& rhs) + { + transposition_matrix_product::run(dst, rhs.nestedExpression(), lhs); + } +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_PRODUCT_EVALUATORS_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Random.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Random.h new file mode 100644 index 00000000000..6faf789c761 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Random.h @@ -0,0 +1,182 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_RANDOM_H +#define EIGEN_RANDOM_H + +namespace Eigen { + +namespace internal { + +template struct scalar_random_op { + EIGEN_EMPTY_STRUCT_CTOR(scalar_random_op) + inline const Scalar operator() () const { return random(); } +}; + +template +struct functor_traits > +{ enum { Cost = 5 * NumTraits::MulCost, PacketAccess = false, IsRepeatable = false }; }; + +} // end namespace internal + +/** \returns a random matrix expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * The parameters \a rows and \a cols are the number of rows and of columns of + * the returned matrix. Must be compatible with this MatrixBase type. + * + * \not_reentrant + * + * This variant is meant to be used for dynamic-size matrix types. For fixed-size types, + * it is redundant to pass \a rows and \a cols as arguments, so Random() should be used + * instead. + * + * + * Example: \include MatrixBase_random_int_int.cpp + * Output: \verbinclude MatrixBase_random_int_int.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary matrix whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * See DenseBase::NullaryExpr(Index, const CustomNullaryOp&) for an example using C++11 random generators. + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index), DenseBase::Random() + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random(Index rows, Index cols) +{ + return NullaryExpr(rows, cols, internal::scalar_random_op()); +} + +/** \returns a random vector expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * The parameter \a size is the size of the returned vector. + * Must be compatible with this MatrixBase type. + * + * \only_for_vectors + * \not_reentrant + * + * This variant is meant to be used for dynamic-size vector types. For fixed-size types, + * it is redundant to pass \a size as argument, so Random() should be used + * instead. + * + * Example: \include MatrixBase_random_int.cpp + * Output: \verbinclude MatrixBase_random_int.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary vector whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random() + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random(Index size) +{ + return NullaryExpr(size, internal::scalar_random_op()); +} + +/** \returns a fixed-size random matrix or vector expression + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * This variant is only for fixed-size MatrixBase types. For dynamic-size types, you + * need to use the variants taking size arguments. + * + * Example: \include MatrixBase_random.cpp + * Output: \verbinclude MatrixBase_random.out + * + * This expression has the "evaluate before nesting" flag so that it will be evaluated into + * a temporary matrix whenever it is nested in a larger expression. This prevents unexpected + * behavior with expressions involving random matrices. + * + * \not_reentrant + * + * \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random(Index) + */ +template +inline const typename DenseBase::RandomReturnType +DenseBase::Random() +{ + return NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_random_op()); +} + +/** Sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \not_reentrant + * + * Example: \include MatrixBase_setRandom.cpp + * Output: \verbinclude MatrixBase_setRandom.out + * + * \sa class CwiseNullaryOp, setRandom(Index), setRandom(Index,Index) + */ +template +inline Derived& DenseBase::setRandom() +{ + return *this = Random(rows(), cols()); +} + +/** Resizes to the given \a newSize, and sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \only_for_vectors + * \not_reentrant + * + * Example: \include Matrix_setRandom_int.cpp + * Output: \verbinclude Matrix_setRandom_int.out + * + * \sa DenseBase::setRandom(), setRandom(Index,Index), class CwiseNullaryOp, DenseBase::Random() + */ +template +EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setRandom(Index newSize) +{ + resize(newSize); + return setRandom(); +} + +/** Resizes to the given size, and sets all coefficients in this expression to random values. + * + * Numbers are uniformly spread through their whole definition range for integer types, + * and in the [-1:1] range for floating point scalar types. + * + * \not_reentrant + * + * \param rows the new number of rows + * \param cols the new number of columns + * + * Example: \include Matrix_setRandom_int_int.cpp + * Output: \verbinclude Matrix_setRandom_int_int.out + * + * \sa DenseBase::setRandom(), setRandom(Index), class CwiseNullaryOp, DenseBase::Random() + */ +template +EIGEN_STRONG_INLINE Derived& +PlainObjectBase::setRandom(Index rows, Index cols) +{ + resize(rows, cols); + return setRandom(); +} + +} // end namespace Eigen + +#endif // EIGEN_RANDOM_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h new file mode 100644 index 00000000000..760e9f86154 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Redux.h @@ -0,0 +1,505 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008 Gael Guennebaud +// Copyright (C) 2006-2008 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REDUX_H +#define EIGEN_REDUX_H + +namespace Eigen { + +namespace internal { + +// TODO +// * implement other kind of vectorization +// * factorize code + +/*************************************************************************** +* Part 1 : the logic deciding a strategy for vectorization and unrolling +***************************************************************************/ + +template +struct redux_traits +{ +public: + typedef typename find_best_packet::type PacketType; + enum { + PacketSize = unpacket_traits::size, + InnerMaxSize = int(Derived::IsRowMajor) + ? Derived::MaxColsAtCompileTime + : Derived::MaxRowsAtCompileTime + }; + + enum { + MightVectorize = (int(Derived::Flags)&ActualPacketAccessBit) + && (functor_traits::PacketAccess), + MayLinearVectorize = bool(MightVectorize) && (int(Derived::Flags)&LinearAccessBit), + MaySliceVectorize = bool(MightVectorize) && int(InnerMaxSize)>=3*PacketSize + }; + +public: + enum { + Traversal = int(MayLinearVectorize) ? int(LinearVectorizedTraversal) + : int(MaySliceVectorize) ? int(SliceVectorizedTraversal) + : int(DefaultTraversal) + }; + +public: + enum { + Cost = Derived::SizeAtCompileTime == Dynamic ? HugeCost + : Derived::SizeAtCompileTime * Derived::CoeffReadCost + (Derived::SizeAtCompileTime-1) * functor_traits::Cost, + UnrollingLimit = EIGEN_UNROLLING_LIMIT * (int(Traversal) == int(DefaultTraversal) ? 1 : int(PacketSize)) + }; + +public: + enum { + Unrolling = Cost <= UnrollingLimit ? CompleteUnrolling : NoUnrolling + }; + +#ifdef EIGEN_DEBUG_ASSIGN + static void debug() + { + std::cerr << "Xpr: " << typeid(typename Derived::XprType).name() << std::endl; + std::cerr.setf(std::ios::hex, std::ios::basefield); + EIGEN_DEBUG_VAR(Derived::Flags) + std::cerr.unsetf(std::ios::hex); + EIGEN_DEBUG_VAR(InnerMaxSize) + EIGEN_DEBUG_VAR(PacketSize) + EIGEN_DEBUG_VAR(MightVectorize) + EIGEN_DEBUG_VAR(MayLinearVectorize) + EIGEN_DEBUG_VAR(MaySliceVectorize) + EIGEN_DEBUG_VAR(Traversal) + EIGEN_DEBUG_VAR(UnrollingLimit) + EIGEN_DEBUG_VAR(Unrolling) + std::cerr << std::endl; + } +#endif +}; + +/*************************************************************************** +* Part 2 : unrollers +***************************************************************************/ + +/*** no vectorization ***/ + +template +struct redux_novec_unroller +{ + enum { + HalfLength = Length/2 + }; + + typedef typename Derived::Scalar Scalar; + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + return func(redux_novec_unroller::run(mat,func), + redux_novec_unroller::run(mat,func)); + } +}; + +template +struct redux_novec_unroller +{ + enum { + outer = Start / Derived::InnerSizeAtCompileTime, + inner = Start % Derived::InnerSizeAtCompileTime + }; + + typedef typename Derived::Scalar Scalar; + + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func&) + { + return mat.coeffByOuterInner(outer, inner); + } +}; + +// This is actually dead code and will never be called. It is required +// to prevent false warnings regarding failed inlining though +// for 0 length run() will never be called at all. +template +struct redux_novec_unroller +{ + typedef typename Derived::Scalar Scalar; + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived&, const Func&) { return Scalar(); } +}; + +/*** vectorization ***/ + +template +struct redux_vec_unroller +{ + enum { + PacketSize = redux_traits::PacketSize, + HalfLength = Length/2 + }; + + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func& func) + { + return func.packetOp( + redux_vec_unroller::run(mat,func), + redux_vec_unroller::run(mat,func) ); + } +}; + +template +struct redux_vec_unroller +{ + enum { + index = Start * redux_traits::PacketSize, + outer = index / int(Derived::InnerSizeAtCompileTime), + inner = index % int(Derived::InnerSizeAtCompileTime), + alignment = Derived::Alignment + }; + + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func&) + { + return mat.template packetByOuterInner(outer, inner); + } +}; + +/*************************************************************************** +* Part 3 : implementation of all cases +***************************************************************************/ + +template::Traversal, + int Unrolling = redux_traits::Unrolling +> +struct redux_impl; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + EIGEN_DEVICE_FUNC + static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + Scalar res; + res = mat.coeffByOuterInner(0, 0); + for(Index i = 1; i < mat.innerSize(); ++i) + res = func(res, mat.coeffByOuterInner(0, i)); + for(Index i = 1; i < mat.outerSize(); ++i) + for(Index j = 0; j < mat.innerSize(); ++j) + res = func(res, mat.coeffByOuterInner(i, j)); + return res; + } +}; + +template +struct redux_impl + : public redux_novec_unroller +{}; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketScalar; + + static Scalar run(const Derived &mat, const Func& func) + { + const Index size = mat.size(); + + const Index packetSize = redux_traits::PacketSize; + const int packetAlignment = unpacket_traits::alignment; + enum { + alignment0 = (bool(Derived::Flags & DirectAccessBit) && bool(packet_traits::AlignedOnScalar)) ? int(packetAlignment) : int(Unaligned), + alignment = EIGEN_PLAIN_ENUM_MAX(alignment0, Derived::Alignment) + }; + const Index alignedStart = internal::first_default_aligned(mat.nestedExpression()); + const Index alignedSize2 = ((size-alignedStart)/(2*packetSize))*(2*packetSize); + const Index alignedSize = ((size-alignedStart)/(packetSize))*(packetSize); + const Index alignedEnd2 = alignedStart + alignedSize2; + const Index alignedEnd = alignedStart + alignedSize; + Scalar res; + if(alignedSize) + { + PacketScalar packet_res0 = mat.template packet(alignedStart); + if(alignedSize>packetSize) // we have at least two packets to partly unroll the loop + { + PacketScalar packet_res1 = mat.template packet(alignedStart+packetSize); + for(Index index = alignedStart + 2*packetSize; index < alignedEnd2; index += 2*packetSize) + { + packet_res0 = func.packetOp(packet_res0, mat.template packet(index)); + packet_res1 = func.packetOp(packet_res1, mat.template packet(index+packetSize)); + } + + packet_res0 = func.packetOp(packet_res0,packet_res1); + if(alignedEnd>alignedEnd2) + packet_res0 = func.packetOp(packet_res0, mat.template packet(alignedEnd2)); + } + res = func.predux(packet_res0); + + for(Index index = 0; index < alignedStart; ++index) + res = func(res,mat.coeff(index)); + + for(Index index = alignedEnd; index < size; ++index) + res = func(res,mat.coeff(index)); + } + else // too small to vectorize anything. + // since this is dynamic-size hence inefficient anyway for such small sizes, don't try to optimize. + { + res = mat.coeff(0); + for(Index index = 1; index < size; ++index) + res = func(res,mat.coeff(index)); + } + + return res; + } +}; + +// NOTE: for SliceVectorizedTraversal we simply bypass unrolling +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + typedef typename redux_traits::PacketType PacketType; + + EIGEN_DEVICE_FUNC static Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + const Index innerSize = mat.innerSize(); + const Index outerSize = mat.outerSize(); + enum { + packetSize = redux_traits::PacketSize + }; + const Index packetedInnerSize = ((innerSize)/packetSize)*packetSize; + Scalar res; + if(packetedInnerSize) + { + PacketType packet_res = mat.template packet(0,0); + for(Index j=0; j(j,i)); + + res = func.predux(packet_res); + for(Index j=0; j::run(mat, func); + } + + return res; + } +}; + +template +struct redux_impl +{ + typedef typename Derived::Scalar Scalar; + + typedef typename redux_traits::PacketType PacketScalar; + enum { + PacketSize = redux_traits::PacketSize, + Size = Derived::SizeAtCompileTime, + VectorizedSize = (Size / PacketSize) * PacketSize + }; + EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func) + { + eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix"); + if (VectorizedSize > 0) { + Scalar res = func.predux(redux_vec_unroller::run(mat,func)); + if (VectorizedSize != Size) + res = func(res,redux_novec_unroller::run(mat,func)); + return res; + } + else { + return redux_novec_unroller::run(mat,func); + } + } +}; + +// evaluator adaptor +template +class redux_evaluator +{ +public: + typedef _XprType XprType; + EIGEN_DEVICE_FUNC explicit redux_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {} + + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + typedef typename XprType::PacketScalar PacketScalar; + typedef typename XprType::PacketReturnType PacketReturnType; + + enum { + MaxRowsAtCompileTime = XprType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = XprType::MaxColsAtCompileTime, + // TODO we should not remove DirectAccessBit and rather find an elegant way to query the alignment offset at runtime from the evaluator + Flags = evaluator::Flags & ~DirectAccessBit, + IsRowMajor = XprType::IsRowMajor, + SizeAtCompileTime = XprType::SizeAtCompileTime, + InnerSizeAtCompileTime = XprType::InnerSizeAtCompileTime, + CoeffReadCost = evaluator::CoeffReadCost, + Alignment = evaluator::Alignment + }; + + EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } + EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } + EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); } + EIGEN_DEVICE_FUNC Index innerSize() const { return m_xpr.innerSize(); } + EIGEN_DEVICE_FUNC Index outerSize() const { return m_xpr.outerSize(); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeff(Index row, Index col) const + { return m_evaluator.coeff(row, col); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeff(Index index) const + { return m_evaluator.coeff(index); } + + template + PacketType packet(Index row, Index col) const + { return m_evaluator.template packet(row, col); } + + template + PacketType packet(Index index) const + { return m_evaluator.template packet(index); } + + EIGEN_DEVICE_FUNC + CoeffReturnType coeffByOuterInner(Index outer, Index inner) const + { return m_evaluator.coeff(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); } + + template + PacketType packetByOuterInner(Index outer, Index inner) const + { return m_evaluator.template packet(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); } + + const XprType & nestedExpression() const { return m_xpr; } + +protected: + internal::evaluator m_evaluator; + const XprType &m_xpr; +}; + +} // end namespace internal + +/*************************************************************************** +* Part 4 : public API +***************************************************************************/ + + +/** \returns the result of a full redux operation on the whole matrix or vector using \a func + * + * The template parameter \a BinaryOp is the type of the functor \a func which must be + * an associative operator. Both current C++98 and C++11 functor styles are handled. + * + * \sa DenseBase::sum(), DenseBase::minCoeff(), DenseBase::maxCoeff(), MatrixBase::colwise(), MatrixBase::rowwise() + */ +template +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::redux(const Func& func) const +{ + eigen_assert(this->rows()>0 && this->cols()>0 && "you are using an empty matrix"); + + typedef typename internal::redux_evaluator ThisEvaluator; + ThisEvaluator thisEval(derived()); + + return internal::redux_impl::run(thisEval, func); +} + +/** \returns the minimum of all coefficients of \c *this. + * \warning the result is undefined if \c *this contains NaN. + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::minCoeff() const +{ + return derived().redux(Eigen::internal::scalar_min_op()); +} + +/** \returns the maximum of all coefficients of \c *this. + * \warning the result is undefined if \c *this contains NaN. + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::maxCoeff() const +{ + return derived().redux(Eigen::internal::scalar_max_op()); +} + +/** \returns the sum of all coefficients of \c *this + * + * If \c *this is empty, then the value 0 is returned. + * + * \sa trace(), prod(), mean() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::sum() const +{ + if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0)) + return Scalar(0); + return derived().redux(Eigen::internal::scalar_sum_op()); +} + +/** \returns the mean of all coefficients of *this +* +* \sa trace(), prod(), sum() +*/ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::mean() const +{ +#ifdef __INTEL_COMPILER + #pragma warning push + #pragma warning ( disable : 2259 ) +#endif + return Scalar(derived().redux(Eigen::internal::scalar_sum_op())) / Scalar(this->size()); +#ifdef __INTEL_COMPILER + #pragma warning pop +#endif +} + +/** \returns the product of all coefficients of *this + * + * Example: \include MatrixBase_prod.cpp + * Output: \verbinclude MatrixBase_prod.out + * + * \sa sum(), mean(), trace() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +DenseBase::prod() const +{ + if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0)) + return Scalar(1); + return derived().redux(Eigen::internal::scalar_product_op()); +} + +/** \returns the trace of \c *this, i.e. the sum of the coefficients on the main diagonal. + * + * \c *this can be any matrix, not necessarily square. + * + * \sa diagonal(), sum() + */ +template +EIGEN_STRONG_INLINE typename internal::traits::Scalar +MatrixBase::trace() const +{ + return derived().diagonal().sum(); +} + +} // end namespace Eigen + +#endif // EIGEN_REDUX_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h new file mode 100644 index 00000000000..17a1496b842 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Ref.h @@ -0,0 +1,284 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2012 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REF_H +#define EIGEN_REF_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : public traits > +{ + typedef _PlainObjectType PlainObjectType; + typedef _StrideType StrideType; + enum { + Options = _Options, + Flags = traits >::Flags | NestByRefBit, + Alignment = traits >::Alignment + }; + + template struct match { + enum { + IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime, + HasDirectAccess = internal::has_direct_access::ret, + StorageOrderMatch = IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), + InnerStrideMatch = int(StrideType::InnerStrideAtCompileTime)==int(Dynamic) + || int(StrideType::InnerStrideAtCompileTime)==int(Derived::InnerStrideAtCompileTime) + || (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1), + OuterStrideMatch = IsVectorAtCompileTime + || int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime), + // NOTE, this indirection of evaluator::Alignment is needed + // to workaround a very strange bug in MSVC related to the instantiation + // of has_*ary_operator in evaluator. + // This line is surprisingly very sensitive. For instance, simply adding parenthesis + // as "DerivedAlignment = (int(evaluator::Alignment))," will make MSVC fail... + DerivedAlignment = int(evaluator::Alignment), + AlignmentMatch = (int(traits::Alignment)==int(Unaligned)) || (DerivedAlignment >= int(Alignment)), // FIXME the first condition is not very clear, it should be replaced by the required alignment + ScalarTypeMatch = internal::is_same::value, + MatchAtCompileTime = HasDirectAccess && StorageOrderMatch && InnerStrideMatch && OuterStrideMatch && AlignmentMatch && ScalarTypeMatch + }; + typedef typename internal::conditional::type type; + }; + +}; + +template +struct traits > : public traits {}; + +} + +template class RefBase + : public MapBase +{ + typedef typename internal::traits::PlainObjectType PlainObjectType; + typedef typename internal::traits::StrideType StrideType; + +public: + + typedef MapBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(RefBase) + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1; + } + + EIGEN_DEVICE_FUNC inline Index outerStride() const + { + return StrideType::OuterStrideAtCompileTime != 0 ? m_stride.outer() + : IsVectorAtCompileTime ? this->size() + : int(Flags)&RowMajorBit ? this->cols() + : this->rows(); + } + + EIGEN_DEVICE_FUNC RefBase() + : Base(0,RowsAtCompileTime==Dynamic?0:RowsAtCompileTime,ColsAtCompileTime==Dynamic?0:ColsAtCompileTime), + // Stride<> does not allow default ctor for Dynamic strides, so let' initialize it with dummy values: + m_stride(StrideType::OuterStrideAtCompileTime==Dynamic?0:StrideType::OuterStrideAtCompileTime, + StrideType::InnerStrideAtCompileTime==Dynamic?0:StrideType::InnerStrideAtCompileTime) + {} + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(RefBase) + +protected: + + typedef Stride StrideBase; + + template + EIGEN_DEVICE_FUNC void construct(Expression& expr) + { + EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(PlainObjectType,Expression); + + if(PlainObjectType::RowsAtCompileTime==1) + { + eigen_assert(expr.rows()==1 || expr.cols()==1); + ::new (static_cast(this)) Base(expr.data(), 1, expr.size()); + } + else if(PlainObjectType::ColsAtCompileTime==1) + { + eigen_assert(expr.rows()==1 || expr.cols()==1); + ::new (static_cast(this)) Base(expr.data(), expr.size(), 1); + } + else + ::new (static_cast(this)) Base(expr.data(), expr.rows(), expr.cols()); + + if(Expression::IsVectorAtCompileTime && (!PlainObjectType::IsVectorAtCompileTime) && ((Expression::Flags&RowMajorBit)!=(PlainObjectType::Flags&RowMajorBit))) + ::new (&m_stride) StrideBase(expr.innerStride(), StrideType::InnerStrideAtCompileTime==0?0:1); + else + ::new (&m_stride) StrideBase(StrideType::OuterStrideAtCompileTime==0?0:expr.outerStride(), + StrideType::InnerStrideAtCompileTime==0?0:expr.innerStride()); + } + + StrideBase m_stride; +}; + +/** \class Ref + * \ingroup Core_Module + * + * \brief A matrix or vector expression mapping an existing expression + * + * \tparam PlainObjectType the equivalent matrix type of the mapped data + * \tparam Options specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned. + * The default is \c #Unaligned. + * \tparam StrideType optionally specifies strides. By default, Ref implies a contiguous storage along the inner dimension (inner stride==1), + * but accepts a variable outer stride (leading dimension). + * This can be overridden by specifying strides. + * The type passed here must be a specialization of the Stride template, see examples below. + * + * This class provides a way to write non-template functions taking Eigen objects as parameters while limiting the number of copies. + * A Ref<> object can represent either a const expression or a l-value: + * \code + * // in-out argument: + * void foo1(Ref x); + * + * // read-only const argument: + * void foo2(const Ref& x); + * \endcode + * + * In the in-out case, the input argument must satisfy the constraints of the actual Ref<> type, otherwise a compilation issue will be triggered. + * By default, a Ref can reference any dense vector expression of float having a contiguous memory layout. + * Likewise, a Ref can reference any column-major dense matrix expression of float whose column's elements are contiguously stored with + * the possibility to have a constant space in-between each column, i.e. the inner stride must be equal to 1, but the outer stride (or leading dimension) + * can be greater than the number of rows. + * + * In the const case, if the input expression does not match the above requirement, then it is evaluated into a temporary before being passed to the function. + * Here are some examples: + * \code + * MatrixXf A; + * VectorXf a; + * foo1(a.head()); // OK + * foo1(A.col()); // OK + * foo1(A.row()); // Compilation error because here innerstride!=1 + * foo2(A.row()); // Compilation error because A.row() is a 1xN object while foo2 is expecting a Nx1 object + * foo2(A.row().transpose()); // The row is copied into a contiguous temporary + * foo2(2*a); // The expression is evaluated into a temporary + * foo2(A.col().segment(2,4)); // No temporary + * \endcode + * + * The range of inputs that can be referenced without temporary can be enlarged using the last two template parameters. + * Here is an example accepting an innerstride!=1: + * \code + * // in-out argument: + * void foo3(Ref > x); + * foo3(A.row()); // OK + * \endcode + * The downside here is that the function foo3 might be significantly slower than foo1 because it won't be able to exploit vectorization, and will involve more + * expensive address computations even if the input is contiguously stored in memory. To overcome this issue, one might propose to overload internally calling a + * template function, e.g.: + * \code + * // in the .h: + * void foo(const Ref& A); + * void foo(const Ref >& A); + * + * // in the .cpp: + * template void foo_impl(const TypeOfA& A) { + * ... // crazy code goes here + * } + * void foo(const Ref& A) { foo_impl(A); } + * void foo(const Ref >& A) { foo_impl(A); } + * \endcode + * + * + * \sa PlainObjectBase::Map(), \ref TopicStorageOrders + */ +template class Ref + : public RefBase > +{ + private: + typedef internal::traits Traits; + template + EIGEN_DEVICE_FUNC inline Ref(const PlainObjectBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0); + public: + + typedef RefBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Ref) + + + #ifndef EIGEN_PARSED_BY_DOXYGEN + template + EIGEN_DEVICE_FUNC inline Ref(PlainObjectBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0) + { + EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); + Base::construct(expr.derived()); + } + template + EIGEN_DEVICE_FUNC inline Ref(const DenseBase& expr, + typename internal::enable_if::MatchAtCompileTime),Derived>::type* = 0) + #else + /** Implicit constructor from any dense expression */ + template + inline Ref(DenseBase& expr) + #endif + { + EIGEN_STATIC_ASSERT(bool(internal::is_lvalue::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + EIGEN_STATIC_ASSERT(bool(Traits::template match::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH); + EIGEN_STATIC_ASSERT(!Derived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY); + Base::construct(expr.const_cast_derived()); + } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Ref) + +}; + +// this is the const ref version +template class Ref + : public RefBase > +{ + typedef internal::traits Traits; + public: + + typedef RefBase Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Ref) + + template + EIGEN_DEVICE_FUNC inline Ref(const DenseBase& expr, + typename internal::enable_if::ScalarTypeMatch),Derived>::type* = 0) + { +// std::cout << match_helper::HasDirectAccess << "," << match_helper::OuterStrideMatch << "," << match_helper::InnerStrideMatch << "\n"; +// std::cout << int(StrideType::OuterStrideAtCompileTime) << " - " << int(Derived::OuterStrideAtCompileTime) << "\n"; +// std::cout << int(StrideType::InnerStrideAtCompileTime) << " - " << int(Derived::InnerStrideAtCompileTime) << "\n"; + construct(expr.derived(), typename Traits::template match::type()); + } + + EIGEN_DEVICE_FUNC inline Ref(const Ref& other) : Base(other) { + // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy + } + + template + EIGEN_DEVICE_FUNC inline Ref(const RefBase& other) { + construct(other.derived(), typename Traits::template match::type()); + } + + protected: + + template + EIGEN_DEVICE_FUNC void construct(const Expression& expr,internal::true_type) + { + Base::construct(expr); + } + + template + EIGEN_DEVICE_FUNC void construct(const Expression& expr, internal::false_type) + { + internal::call_assignment_no_alias(m_object,expr,internal::assign_op()); + Base::construct(m_object); + } + + protected: + TPlainObjectType m_object; +}; + +} // end namespace Eigen + +#endif // EIGEN_REF_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h new file mode 100644 index 00000000000..9960ef884ef --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Replicate.h @@ -0,0 +1,142 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REPLICATE_H +#define EIGEN_REPLICATE_H + +namespace Eigen { + +namespace internal { +template +struct traits > + : traits +{ + typedef typename MatrixType::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + enum { + RowsAtCompileTime = RowFactor==Dynamic || int(MatrixType::RowsAtCompileTime)==Dynamic + ? Dynamic + : RowFactor * MatrixType::RowsAtCompileTime, + ColsAtCompileTime = ColFactor==Dynamic || int(MatrixType::ColsAtCompileTime)==Dynamic + ? Dynamic + : ColFactor * MatrixType::ColsAtCompileTime, + //FIXME we don't propagate the max sizes !!! + MaxRowsAtCompileTime = RowsAtCompileTime, + MaxColsAtCompileTime = ColsAtCompileTime, + IsRowMajor = MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1 ? 1 + : MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1 ? 0 + : (MatrixType::Flags & RowMajorBit) ? 1 : 0, + + // FIXME enable DirectAccess with negative strides? + Flags = IsRowMajor ? RowMajorBit : 0 + }; +}; +} + +/** + * \class Replicate + * \ingroup Core_Module + * + * \brief Expression of the multiple replication of a matrix or vector + * + * \tparam MatrixType the type of the object we are replicating + * \tparam RowFactor number of repetitions at compile time along the vertical direction, can be Dynamic. + * \tparam ColFactor number of repetitions at compile time along the horizontal direction, can be Dynamic. + * + * This class represents an expression of the multiple replication of a matrix or vector. + * It is the return type of DenseBase::replicate() and most of the time + * this is the only way it is used. + * + * \sa DenseBase::replicate() + */ +template class Replicate + : public internal::dense_xpr_base< Replicate >::type +{ + typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; + typedef typename internal::traits::_MatrixTypeNested _MatrixTypeNested; + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Replicate) + typedef typename internal::remove_all::type NestedExpression; + + template + EIGEN_DEVICE_FUNC + inline explicit Replicate(const OriginalMatrixType& matrix) + : m_matrix(matrix), m_rowFactor(RowFactor), m_colFactor(ColFactor) + { + EIGEN_STATIC_ASSERT((internal::is_same::type,OriginalMatrixType>::value), + THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE) + eigen_assert(RowFactor!=Dynamic && ColFactor!=Dynamic); + } + + template + EIGEN_DEVICE_FUNC + inline Replicate(const OriginalMatrixType& matrix, Index rowFactor, Index colFactor) + : m_matrix(matrix), m_rowFactor(rowFactor), m_colFactor(colFactor) + { + EIGEN_STATIC_ASSERT((internal::is_same::type,OriginalMatrixType>::value), + THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE) + } + + EIGEN_DEVICE_FUNC + inline Index rows() const { return m_matrix.rows() * m_rowFactor.value(); } + EIGEN_DEVICE_FUNC + inline Index cols() const { return m_matrix.cols() * m_colFactor.value(); } + + EIGEN_DEVICE_FUNC + const _MatrixTypeNested& nestedExpression() const + { + return m_matrix; + } + + protected: + MatrixTypeNested m_matrix; + const internal::variable_if_dynamic m_rowFactor; + const internal::variable_if_dynamic m_colFactor; +}; + +/** + * \return an expression of the replication of \c *this + * + * Example: \include MatrixBase_replicate.cpp + * Output: \verbinclude MatrixBase_replicate.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(Index,Index), class Replicate + */ +template +template +const Replicate +DenseBase::replicate() const +{ + return Replicate(derived()); +} + +/** + * \return an expression of the replication of each column (or row) of \c *this + * + * Example: \include DirectionWise_replicate_int.cpp + * Output: \verbinclude DirectionWise_replicate_int.out + * + * \sa VectorwiseOp::replicate(), DenseBase::replicate(), class Replicate + */ +template +const typename VectorwiseOp::ReplicateReturnType +VectorwiseOp::replicate(Index factor) const +{ + return typename VectorwiseOp::ReplicateReturnType + (_expression(),Direction==Vertical?factor:1,Direction==Horizontal?factor:1); +} + +} // end namespace Eigen + +#endif // EIGEN_REPLICATE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h new file mode 100644 index 00000000000..c44b7673bb3 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/ReturnByValue.h @@ -0,0 +1,117 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2009-2010 Gael Guennebaud +// Copyright (C) 2009-2010 Benoit Jacob +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_RETURNBYVALUE_H +#define EIGEN_RETURNBYVALUE_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : public traits::ReturnType> +{ + enum { + // We're disabling the DirectAccess because e.g. the constructor of + // the Block-with-DirectAccess expression requires to have a coeffRef method. + // Also, we don't want to have to implement the stride stuff. + Flags = (traits::ReturnType>::Flags + | EvalBeforeNestingBit) & ~DirectAccessBit + }; +}; + +/* The ReturnByValue object doesn't even have a coeff() method. + * So the only way that nesting it in an expression can work, is by evaluating it into a plain matrix. + * So internal::nested always gives the plain return matrix type. + * + * FIXME: I don't understand why we need this specialization: isn't this taken care of by the EvalBeforeNestingBit ?? + * Answer: EvalBeforeNestingBit should be deprecated since we have the evaluators + */ +template +struct nested_eval, n, PlainObject> +{ + typedef typename traits::ReturnType type; +}; + +} // end namespace internal + +/** \class ReturnByValue + * \ingroup Core_Module + * + */ +template class ReturnByValue + : public internal::dense_xpr_base< ReturnByValue >::type, internal::no_assignment_operator +{ + public: + typedef typename internal::traits::ReturnType ReturnType; + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(ReturnByValue) + + template + EIGEN_DEVICE_FUNC + inline void evalTo(Dest& dst) const + { static_cast(this)->evalTo(dst); } + EIGEN_DEVICE_FUNC inline Index rows() const { return static_cast(this)->rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return static_cast(this)->cols(); } + +#ifndef EIGEN_PARSED_BY_DOXYGEN +#define Unusable YOU_ARE_TRYING_TO_ACCESS_A_SINGLE_COEFFICIENT_IN_A_SPECIAL_EXPRESSION_WHERE_THAT_IS_NOT_ALLOWED_BECAUSE_THAT_WOULD_BE_INEFFICIENT + class Unusable{ + Unusable(const Unusable&) {} + Unusable& operator=(const Unusable&) {return *this;} + }; + const Unusable& coeff(Index) const { return *reinterpret_cast(this); } + const Unusable& coeff(Index,Index) const { return *reinterpret_cast(this); } + Unusable& coeffRef(Index) { return *reinterpret_cast(this); } + Unusable& coeffRef(Index,Index) { return *reinterpret_cast(this); } +#undef Unusable +#endif +}; + +template +template +Derived& DenseBase::operator=(const ReturnByValue& other) +{ + other.evalTo(derived()); + return derived(); +} + +namespace internal { + +// Expression is evaluated in a temporary; default implementation of Assignment is bypassed so that +// when a ReturnByValue expression is assigned, the evaluator is not constructed. +// TODO: Finalize port to new regime; ReturnByValue should not exist in the expression world + +template +struct evaluator > + : public evaluator::ReturnType> +{ + typedef ReturnByValue XprType; + typedef typename internal::traits::ReturnType PlainObject; + typedef evaluator Base; + + EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr) + : m_result(xpr.rows(), xpr.cols()) + { + ::new (static_cast(this)) Base(m_result); + xpr.evalTo(m_result); + } + +protected: + PlainObject m_result; +}; + +} // end namespace internal + +} // end namespace Eigen + +#endif // EIGEN_RETURNBYVALUE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h new file mode 100644 index 00000000000..0640cda2a15 --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Reverse.h @@ -0,0 +1,211 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2006-2008 Benoit Jacob +// Copyright (C) 2009 Ricard Marxer +// Copyright (C) 2009-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_REVERSE_H +#define EIGEN_REVERSE_H + +namespace Eigen { + +namespace internal { + +template +struct traits > + : traits +{ + typedef typename MatrixType::Scalar Scalar; + typedef typename traits::StorageKind StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ref_selector::type MatrixTypeNested; + typedef typename remove_reference::type _MatrixTypeNested; + enum { + RowsAtCompileTime = MatrixType::RowsAtCompileTime, + ColsAtCompileTime = MatrixType::ColsAtCompileTime, + MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime, + Flags = _MatrixTypeNested::Flags & (RowMajorBit | LvalueBit) + }; +}; + +template struct reverse_packet_cond +{ + static inline PacketType run(const PacketType& x) { return preverse(x); } +}; + +template struct reverse_packet_cond +{ + static inline PacketType run(const PacketType& x) { return x; } +}; + +} // end namespace internal + +/** \class Reverse + * \ingroup Core_Module + * + * \brief Expression of the reverse of a vector or matrix + * + * \tparam MatrixType the type of the object of which we are taking the reverse + * \tparam Direction defines the direction of the reverse operation, can be Vertical, Horizontal, or BothDirections + * + * This class represents an expression of the reverse of a vector. + * It is the return type of MatrixBase::reverse() and VectorwiseOp::reverse() + * and most of the time this is the only way it is used. + * + * \sa MatrixBase::reverse(), VectorwiseOp::reverse() + */ +template class Reverse + : public internal::dense_xpr_base< Reverse >::type +{ + public: + + typedef typename internal::dense_xpr_base::type Base; + EIGEN_DENSE_PUBLIC_INTERFACE(Reverse) + typedef typename internal::remove_all::type NestedExpression; + using Base::IsRowMajor; + + protected: + enum { + PacketSize = internal::packet_traits::size, + IsColMajor = !IsRowMajor, + ReverseRow = (Direction == Vertical) || (Direction == BothDirections), + ReverseCol = (Direction == Horizontal) || (Direction == BothDirections), + OffsetRow = ReverseRow && IsColMajor ? PacketSize : 1, + OffsetCol = ReverseCol && IsRowMajor ? PacketSize : 1, + ReversePacket = (Direction == BothDirections) + || ((Direction == Vertical) && IsColMajor) + || ((Direction == Horizontal) && IsRowMajor) + }; + typedef internal::reverse_packet_cond reverse_packet; + public: + + EIGEN_DEVICE_FUNC explicit inline Reverse(const MatrixType& matrix) : m_matrix(matrix) { } + + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Reverse) + + EIGEN_DEVICE_FUNC inline Index rows() const { return m_matrix.rows(); } + EIGEN_DEVICE_FUNC inline Index cols() const { return m_matrix.cols(); } + + EIGEN_DEVICE_FUNC inline Index innerStride() const + { + return -m_matrix.innerStride(); + } + + EIGEN_DEVICE_FUNC const typename internal::remove_all::type& + nestedExpression() const + { + return m_matrix; + } + + protected: + typename MatrixType::Nested m_matrix; +}; + +/** \returns an expression of the reverse of *this. + * + * Example: \include MatrixBase_reverse.cpp + * Output: \verbinclude MatrixBase_reverse.out + * + */ +template +inline typename DenseBase::ReverseReturnType +DenseBase::reverse() +{ + return ReverseReturnType(derived()); +} + + +//reverse const overload moved DenseBase.h due to a CUDA compiler bug + +/** This is the "in place" version of reverse: it reverses \c *this. + * + * In most cases it is probably better to simply use the reversed expression + * of a matrix. However, when reversing the matrix data itself is really needed, + * then this "in-place" version is probably the right choice because it provides + * the following additional benefits: + * - less error prone: doing the same operation with .reverse() requires special care: + * \code m = m.reverse().eval(); \endcode + * - this API enables reverse operations without the need for a temporary + * - it allows future optimizations (cache friendliness, etc.) + * + * \sa VectorwiseOp::reverseInPlace(), reverse() */ +template +inline void DenseBase::reverseInPlace() +{ + if(cols()>rows()) + { + Index half = cols()/2; + leftCols(half).swap(rightCols(half).reverse()); + if((cols()%2)==1) + { + Index half2 = rows()/2; + col(half).head(half2).swap(col(half).tail(half2).reverse()); + } + } + else + { + Index half = rows()/2; + topRows(half).swap(bottomRows(half).reverse()); + if((rows()%2)==1) + { + Index half2 = cols()/2; + row(half).head(half2).swap(row(half).tail(half2).reverse()); + } + } +} + +namespace internal { + +template +struct vectorwise_reverse_inplace_impl; + +template<> +struct vectorwise_reverse_inplace_impl +{ + template + static void run(ExpressionType &xpr) + { + Index half = xpr.rows()/2; + xpr.topRows(half).swap(xpr.bottomRows(half).colwise().reverse()); + } +}; + +template<> +struct vectorwise_reverse_inplace_impl +{ + template + static void run(ExpressionType &xpr) + { + Index half = xpr.cols()/2; + xpr.leftCols(half).swap(xpr.rightCols(half).rowwise().reverse()); + } +}; + +} // end namespace internal + +/** This is the "in place" version of VectorwiseOp::reverse: it reverses each column or row of \c *this. + * + * In most cases it is probably better to simply use the reversed expression + * of a matrix. However, when reversing the matrix data itself is really needed, + * then this "in-place" version is probably the right choice because it provides + * the following additional benefits: + * - less error prone: doing the same operation with .reverse() requires special care: + * \code m = m.reverse().eval(); \endcode + * - this API enables reverse operations without the need for a temporary + * + * \sa DenseBase::reverseInPlace(), reverse() */ +template +void VectorwiseOp::reverseInPlace() +{ + internal::vectorwise_reverse_inplace_impl::run(_expression().const_cast_derived()); +} + +} // end namespace Eigen + +#endif // EIGEN_REVERSE_H diff --git a/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Select.h b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Select.h new file mode 100644 index 00000000000..79eec1b5b0e --- /dev/null +++ b/third_party/cpgfunctionEP-0.2/third_party/eigen-3.3.9/Eigen/src/Core/Select.h @@ -0,0 +1,162 @@ +// This file is part of Eigen, a lightweight C++ template library +// for linear algebra. +// +// Copyright (C) 2008-2010 Gael Guennebaud +// +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. + +#ifndef EIGEN_SELECT_H +#define EIGEN_SELECT_H + +namespace Eigen { + +/** \class Select + * \ingroup Core_Module + * + * \brief Expression of a coefficient wise version of the C++ ternary operator ?: + * + * \param ConditionMatrixType the type of the \em condition expression which must be a boolean matrix + * \param ThenMatrixType the type of the \em then expression + * \param ElseMatrixType the type of the \em else expression + * + * This class represents an expression of a coefficient wise version of the C++ ternary operator ?:. + * It is the return type of DenseBase::select() and most of the time this is the only way it is used. + * + * \sa DenseBase::select(const DenseBase&, const DenseBase&) const + */ + +namespace internal { +template +struct traits > + : traits +{ + typedef typename traits::Scalar Scalar; + typedef Dense StorageKind; + typedef typename traits::XprKind XprKind; + typedef typename ConditionMatrixType::Nested ConditionMatrixNested; + typedef typename ThenMatrixType::Nested ThenMatrixNested; + typedef typename ElseMatrixType::Nested ElseMatrixNested; + enum { + RowsAtCompileTime = ConditionMatrixType::RowsAtCompileTime, + ColsAtCompileTime = ConditionMatrixType::ColsAtCompileTime, + MaxRowsAtCompileTime = ConditionMatrixType::MaxRowsAtCompileTime, + MaxColsAtCompileTime = ConditionMatrixType::MaxColsAtCompileTime, + Flags = (unsigned int)ThenMatrixType::Flags & ElseMatrixType::Flags & RowMajorBit + }; +}; +} + +template +class Select : public internal::dense_xpr_base< Select >::type, + internal::no_assignment_operator +{ + public: + + typedef typename internal::dense_xpr_base::type Base; - EIGEN_DENSE_PUBLIC_INTERFACE(Select) - - inline EIGEN_DEVICE_FUNC - Select(const ConditionMatrixType& a_conditionMatrix, - const ThenMatrixType& a_thenMatrix, - const ElseMatrixType& a_elseMatrix) - : m_condition(a_conditionMatrix), m_then(a_thenMatrix), m_else(a_elseMatrix) - { - eigen_assert(m_condition.rows() == m_then.rows() && m_condition.rows() == m_else.rows()); - eigen_assert(m_condition.cols() == m_then.cols() && m_condition.cols() == m_else.cols()); - } - - inline EIGEN_DEVICE_FUNC Index rows() const { return m_condition.rows(); } - inline EIGEN_DEVICE_FUNC Index cols() const { return m_condition.cols(); } - - inline EIGEN_DEVICE_FUNC - const Scalar coeff(Index i, Index j) const - { - if (m_condition.coeff(i,j)) - return m_then.coeff(i,j); - else - return m_else.coeff(i,j); - } - - inline EIGEN_DEVICE_FUNC - const Scalar coeff(Index i) const - { - if (m_condition.coeff(i)) - return m_then.coeff(i); - else - return m_else.coeff(i); - } - - inline EIGEN_DEVICE_FUNC const ConditionMatrixType& conditionMatrix() const - { - return m_condition; - } - - inline EIGEN_DEVICE_FUNC const ThenMatrixType& thenMatrix() const - { - return m_then; - } - - inline EIGEN_DEVICE_FUNC const ElseMatrixType& elseMatrix() const - { - return m_else; - } - - protected: - typename ConditionMatrixType::Nested m_condition; - typename ThenMatrixType::Nested m_then; - typename ElseMatrixType::Nested m_else; -}; - - -/** \returns a matrix where each coefficient (i,j) is equal to \a thenMatrix(i,j) - * if \c *this(i,j), and \a elseMatrix(i,j) otherwise. - * - * Example: \include MatrixBase_select.cpp - * Output: \verbinclude MatrixBase_select.out - * - * \sa class Select - */ -template -template -inline const Select -DenseBase::select(const DenseBase& thenMatrix, - const DenseBase& elseMatrix) const -{ - return Select(derived(), thenMatrix.derived(), elseMatrix.derived()); -} - -/** Version of DenseBase::select(const DenseBase&, const DenseBase&) with - * the \em else expression being a scalar value. - * - * \sa DenseBase::select(const DenseBase&, const DenseBase&) const, class Select - */ -template -template -inline const Select -DenseBase::select(const DenseBase& thenMatrix, - const typename ThenDerived::Scalar& elseScalar) const -{ - return Select( - derived(), thenMatrix.derived(), ThenDerived::Constant(rows(),cols(),elseScalar)); -} - -/** Version of DenseBase::select(const DenseBase&, const DenseBase&) with - * the \em then expression being a scalar value. - * - * \sa DenseBase::select(const DenseBase&, const DenseBase&) const, class Select - */ -template -template -inline const Select -DenseBase::select(const typename ElseDerived::Scalar& thenScalar, - const DenseBase& elseMatrix) const -{ - return Select( - derived(), ElseDerived::Constant(rows(),cols(),thenScalar), elseMatrix.derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SELECT_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h deleted file mode 100644 index b2e51f37ac4..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfAdjointView.h +++ /dev/null @@ -1,352 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINTMATRIX_H -#define EIGEN_SELFADJOINTMATRIX_H - -namespace Eigen { - -/** \class SelfAdjointView - * \ingroup Core_Module - * - * - * \brief Expression of a selfadjoint matrix from a triangular part of a dense matrix - * - * \param MatrixType the type of the dense matrix storing the coefficients - * \param TriangularPart can be either \c #Lower or \c #Upper - * - * This class is an expression of a sefladjoint matrix from a triangular part of a matrix - * with given dense storage of the coefficients. It is the return type of MatrixBase::selfadjointView() - * and most of the time this is the only way that it is used. - * - * \sa class TriangularBase, MatrixBase::selfadjointView() - */ - -namespace internal { -template -struct traits > : traits -{ - typedef typename ref_selector::non_const_type MatrixTypeNested; - typedef typename remove_all::type MatrixTypeNestedCleaned; - typedef MatrixType ExpressionType; - typedef typename MatrixType::PlainObject FullMatrixType; - enum { - Mode = UpLo | SelfAdjoint, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags = MatrixTypeNestedCleaned::Flags & (HereditaryBits|FlagsLvalueBit) - & (~(PacketAccessBit | DirectAccessBit | LinearAccessBit)) // FIXME these flags should be preserved - }; -}; -} - - -template class SelfAdjointView - : public TriangularBase > -{ - public: - - typedef _MatrixType MatrixType; - typedef TriangularBase Base; - typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; - typedef typename internal::traits::MatrixTypeNestedCleaned MatrixTypeNestedCleaned; - typedef MatrixTypeNestedCleaned NestedExpression; - - /** \brief The type of coefficients in this matrix */ - typedef typename internal::traits::Scalar Scalar; - typedef typename MatrixType::StorageIndex StorageIndex; - typedef typename internal::remove_all::type MatrixConjugateReturnType; - - enum { - Mode = internal::traits::Mode, - Flags = internal::traits::Flags, - TransposeMode = ((Mode & Upper) ? Lower : 0) | ((Mode & Lower) ? Upper : 0) - }; - typedef typename MatrixType::PlainObject PlainObject; - - EIGEN_DEVICE_FUNC - explicit inline SelfAdjointView(MatrixType& matrix) : m_matrix(matrix) - { - EIGEN_STATIC_ASSERT(UpLo==Lower || UpLo==Upper,SELFADJOINTVIEW_ACCEPTS_UPPER_AND_LOWER_MODE_ONLY); - } - - EIGEN_DEVICE_FUNC - inline Index rows() const { return m_matrix.rows(); } - EIGEN_DEVICE_FUNC - inline Index cols() const { return m_matrix.cols(); } - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return m_matrix.outerStride(); } - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return m_matrix.innerStride(); } - - /** \sa MatrixBase::coeff() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const - { - Base::check_coordinates_internal(row, col); - return m_matrix.coeff(row, col); - } - - /** \sa MatrixBase::coeffRef() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) - { - EIGEN_STATIC_ASSERT_LVALUE(SelfAdjointView); - Base::check_coordinates_internal(row, col); - return m_matrix.coeffRef(row, col); - } - - /** \internal */ - EIGEN_DEVICE_FUNC - const MatrixTypeNestedCleaned& _expression() const { return m_matrix; } - - EIGEN_DEVICE_FUNC - const MatrixTypeNestedCleaned& nestedExpression() const { return m_matrix; } - EIGEN_DEVICE_FUNC - MatrixTypeNestedCleaned& nestedExpression() { return m_matrix; } - - /** Efficient triangular matrix times vector/matrix product */ - template - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& rhs) const - { - return Product(*this, rhs.derived()); - } - - /** Efficient vector/matrix times triangular matrix product */ - template friend - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& lhs, const SelfAdjointView& rhs) - { - return Product(lhs.derived(),rhs); - } - - friend EIGEN_DEVICE_FUNC - const SelfAdjointView - operator*(const Scalar& s, const SelfAdjointView& mat) - { - return (s*mat.nestedExpression()).template selfadjointView(); - } - - /** Perform a symmetric rank 2 update of the selfadjoint matrix \c *this: - * \f$ this = this + \alpha u v^* + conj(\alpha) v u^* \f$ - * \returns a reference to \c *this - * - * The vectors \a u and \c v \b must be column vectors, however they can be - * a adjoint expression without any overhead. Only the meaningful triangular - * part of the matrix is updated, the rest is left unchanged. - * - * \sa rankUpdate(const MatrixBase&, Scalar) - */ - template - EIGEN_DEVICE_FUNC - SelfAdjointView& rankUpdate(const MatrixBase& u, const MatrixBase& v, const Scalar& alpha = Scalar(1)); - - /** Perform a symmetric rank K update of the selfadjoint matrix \c *this: - * \f$ this = this + \alpha ( u u^* ) \f$ where \a u is a vector or matrix. - * - * \returns a reference to \c *this - * - * Note that to perform \f$ this = this + \alpha ( u^* u ) \f$ you can simply - * call this function with u.adjoint(). - * - * \sa rankUpdate(const MatrixBase&, const MatrixBase&, Scalar) - */ - template - EIGEN_DEVICE_FUNC - SelfAdjointView& rankUpdate(const MatrixBase& u, const Scalar& alpha = Scalar(1)); - - /** \returns an expression of a triangular view extracted from the current selfadjoint view of a given triangular part - * - * The parameter \a TriMode can have the following values: \c #Upper, \c #StrictlyUpper, \c #UnitUpper, - * \c #Lower, \c #StrictlyLower, \c #UnitLower. - * - * If \c TriMode references the same triangular part than \c *this, then this method simply return a \c TriangularView of the nested expression, - * otherwise, the nested expression is first transposed, thus returning a \c TriangularView> object. - * - * \sa MatrixBase::triangularView(), class TriangularView - */ - template - EIGEN_DEVICE_FUNC - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), - TriangularView, - TriangularView >::type - triangularView() const - { - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), MatrixType&, typename MatrixType::ConstTransposeReturnType>::type tmp1(m_matrix); - typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), MatrixType&, typename MatrixType::AdjointReturnType>::type tmp2(tmp1); - return typename internal::conditional<(TriMode&(Upper|Lower))==(UpLo&(Upper|Lower)), - TriangularView, - TriangularView >::type(tmp2); - } - - typedef SelfAdjointView ConjugateReturnType; - /** \sa MatrixBase::conjugate() const */ - EIGEN_DEVICE_FUNC - inline const ConjugateReturnType conjugate() const - { return ConjugateReturnType(m_matrix.conjugate()); } - - typedef SelfAdjointView AdjointReturnType; - /** \sa MatrixBase::adjoint() const */ - EIGEN_DEVICE_FUNC - inline const AdjointReturnType adjoint() const - { return AdjointReturnType(m_matrix.adjoint()); } - - typedef SelfAdjointView TransposeReturnType; - /** \sa MatrixBase::transpose() */ - EIGEN_DEVICE_FUNC - inline TransposeReturnType transpose() - { - EIGEN_STATIC_ASSERT_LVALUE(MatrixType) - typename MatrixType::TransposeReturnType tmp(m_matrix); - return TransposeReturnType(tmp); - } - - typedef SelfAdjointView ConstTransposeReturnType; - /** \sa MatrixBase::transpose() const */ - EIGEN_DEVICE_FUNC - inline const ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(m_matrix.transpose()); - } - - /** \returns a const expression of the main diagonal of the matrix \c *this - * - * This method simply returns the diagonal of the nested expression, thus by-passing the SelfAdjointView decorator. - * - * \sa MatrixBase::diagonal(), class Diagonal */ - EIGEN_DEVICE_FUNC - typename MatrixType::ConstDiagonalReturnType diagonal() const - { - return typename MatrixType::ConstDiagonalReturnType(m_matrix); - } - -/////////// Cholesky module /////////// - - const LLT llt() const; - const LDLT ldlt() const; - -/////////// Eigenvalue module /////////// - - /** Real part of #Scalar */ - typedef typename NumTraits::Real RealScalar; - /** Return type of eigenvalues() */ - typedef Matrix::ColsAtCompileTime, 1> EigenvaluesReturnType; - - EIGEN_DEVICE_FUNC - EigenvaluesReturnType eigenvalues() const; - EIGEN_DEVICE_FUNC - RealScalar operatorNorm() const; - - protected: - MatrixTypeNested m_matrix; -}; - - -// template -// internal::selfadjoint_matrix_product_returntype > -// operator*(const MatrixBase& lhs, const SelfAdjointView& rhs) -// { -// return internal::matrix_selfadjoint_product_returntype >(lhs.derived(),rhs); -// } - -// selfadjoint to dense matrix - -namespace internal { - -// TODO currently a selfadjoint expression has the form SelfAdjointView<.,.> -// in the future selfadjoint-ness should be defined by the expression traits -// such that Transpose > is valid. (currently TriangularBase::transpose() is overloaded to make it work) -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef SelfAdjointShape Shape; -}; - -template -class triangular_dense_assignment_kernel - : public generic_dense_assignment_kernel -{ -protected: - typedef generic_dense_assignment_kernel Base; - typedef typename Base::DstXprType DstXprType; - typedef typename Base::SrcXprType SrcXprType; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; -public: - - typedef typename Base::DstEvaluatorType DstEvaluatorType; - typedef typename Base::SrcEvaluatorType SrcEvaluatorType; - typedef typename Base::Scalar Scalar; - typedef typename Base::AssignmentTraits AssignmentTraits; - - - EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - - EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - Scalar tmp = m_src.coeff(row,col); - m_functor.assignCoeff(m_dst.coeffRef(row,col), tmp); - m_functor.assignCoeff(m_dst.coeffRef(col,row), numext::conj(tmp)); - } - - EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id) - { - Base::assignCoeff(id,id); - } - - EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index, Index) - { eigen_internal_assert(false && "should never be called"); } -}; - -} // end namespace internal - -/*************************************************************************** -* Implementation of MatrixBase methods -***************************************************************************/ - -/** This is the const version of MatrixBase::selfadjointView() */ -template -template -typename MatrixBase::template ConstSelfAdjointViewReturnType::Type -MatrixBase::selfadjointView() const -{ - return typename ConstSelfAdjointViewReturnType::Type(derived()); -} - -/** \returns an expression of a symmetric/self-adjoint view extracted from the upper or lower triangular part of the current matrix - * - * The parameter \a UpLo can be either \c #Upper or \c #Lower - * - * Example: \include MatrixBase_selfadjointView.cpp - * Output: \verbinclude MatrixBase_selfadjointView.out - * - * \sa class SelfAdjointView - */ -template -template -typename MatrixBase::template SelfAdjointViewReturnType::Type -MatrixBase::selfadjointView() -{ - return typename SelfAdjointViewReturnType::Type(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINTMATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h deleted file mode 100644 index 7c89c2e23c6..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SelfCwiseBinaryOp.h +++ /dev/null @@ -1,47 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFCWISEBINARYOP_H -#define EIGEN_SELFCWISEBINARYOP_H - -namespace Eigen { - -// TODO generalize the scalar type of 'other' - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::operator*=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::mul_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& ArrayBase::operator+=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::add_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& ArrayBase::operator-=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::sub_assign_op()); - return derived(); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase::operator/=(const Scalar& other) -{ - internal::call_assignment(this->derived(), PlainObject::Constant(rows(),cols(),other), internal::div_assign_op()); - return derived(); -} - -} // end namespace Eigen - -#endif // EIGEN_SELFCWISEBINARYOP_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h deleted file mode 100644 index a8daea51135..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Solve.h +++ /dev/null @@ -1,188 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVE_H -#define EIGEN_SOLVE_H - -namespace Eigen { - -template class SolveImpl; - -/** \class Solve - * \ingroup Core_Module - * - * \brief Pseudo expression representing a solving operation - * - * \tparam Decomposition the type of the matrix or decomposion object - * \tparam Rhstype the type of the right-hand side - * - * This class represents an expression of A.solve(B) - * and most of the time this is the only way it is used. - * - */ -namespace internal { - -// this solve_traits class permits to determine the evaluation type with respect to storage kind (Dense vs Sparse) -template struct solve_traits; - -template -struct solve_traits -{ - typedef typename make_proper_matrix_type::type PlainObject; -}; - -template -struct traits > - : traits::StorageKind>::PlainObject> -{ - typedef typename solve_traits::StorageKind>::PlainObject PlainObject; - typedef typename promote_index_type::type StorageIndex; - typedef traits BaseTraits; - enum { - Flags = BaseTraits::Flags & RowMajorBit, - CoeffReadCost = HugeCost - }; -}; - -} - - -template -class Solve : public SolveImpl::StorageKind> -{ -public: - typedef typename internal::traits::PlainObject PlainObject; - typedef typename internal::traits::StorageIndex StorageIndex; - - Solve(const Decomposition &dec, const RhsType &rhs) - : m_dec(dec), m_rhs(rhs) - {} - - EIGEN_DEVICE_FUNC Index rows() const { return m_dec.cols(); } - EIGEN_DEVICE_FUNC Index cols() const { return m_rhs.cols(); } - - EIGEN_DEVICE_FUNC const Decomposition& dec() const { return m_dec; } - EIGEN_DEVICE_FUNC const RhsType& rhs() const { return m_rhs; } - -protected: - const Decomposition &m_dec; - const RhsType &m_rhs; -}; - - -// Specialization of the Solve expression for dense results -template -class SolveImpl - : public MatrixBase > -{ - typedef Solve Derived; - -public: - - typedef MatrixBase > Base; - EIGEN_DENSE_PUBLIC_INTERFACE(Derived) - -private: - - Scalar coeff(Index row, Index col) const; - Scalar coeff(Index i) const; -}; - -// Generic API dispatcher -template -class SolveImpl : public internal::generic_xpr_base, MatrixXpr, StorageKind>::type -{ - public: - typedef typename internal::generic_xpr_base, MatrixXpr, StorageKind>::type Base; -}; - -namespace internal { - -// Evaluator of Solve -> eval into a temporary -template -struct evaluator > - : public evaluator::PlainObject> -{ - typedef Solve SolveType; - typedef typename SolveType::PlainObject PlainObject; - typedef evaluator Base; - - enum { Flags = Base::Flags | EvalBeforeNestingBit }; - - EIGEN_DEVICE_FUNC explicit evaluator(const SolveType& solve) - : m_result(solve.rows(), solve.cols()) - { - ::new (static_cast(this)) Base(m_result); - solve.dec()._solve_impl(solve.rhs(), m_result); - } - -protected: - PlainObject m_result; -}; - -// Specialization for "dst = dec.solve(rhs)" -// NOTE we need to specialize it for Dense2Dense to avoid ambiguous specialization error and a Sparse2Sparse specialization must exist somewhere -template -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Solve SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec()._solve_impl(src.rhs(), dst); - } -}; - -// Specialization for "dst = dec.transpose().solve(rhs)" -template -struct Assignment,RhsType>, internal::assign_op, Dense2Dense> -{ - typedef Solve,RhsType> SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec().nestedExpression().template _solve_impl_transposed(src.rhs(), dst); - } -}; - -// Specialization for "dst = dec.adjoint().solve(rhs)" -template -struct Assignment, const Transpose >,RhsType>, - internal::assign_op, Dense2Dense> -{ - typedef Solve, const Transpose >,RhsType> SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - src.dec().nestedExpression().nestedExpression().template _solve_impl_transposed(src.rhs(), dst); - } -}; - -} // end namepsace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h deleted file mode 100644 index fd0acb1a58e..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolveTriangular.h +++ /dev/null @@ -1,235 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVETRIANGULAR_H -#define EIGEN_SOLVETRIANGULAR_H - -namespace Eigen { - -namespace internal { - -// Forward declarations: -// The following two routines are implemented in the products/TriangularSolver*.h files -template -struct triangular_solve_vector; - -template -struct triangular_solve_matrix; - -// small helper struct extracting some traits on the underlying solver operation -template -class trsolve_traits -{ - private: - enum { - RhsIsVectorAtCompileTime = (Side==OnTheLeft ? Rhs::ColsAtCompileTime : Rhs::RowsAtCompileTime)==1 - }; - public: - enum { - Unrolling = (RhsIsVectorAtCompileTime && Rhs::SizeAtCompileTime != Dynamic && Rhs::SizeAtCompileTime <= 8) - ? CompleteUnrolling : NoUnrolling, - RhsVectors = RhsIsVectorAtCompileTime ? 1 : Dynamic - }; -}; - -template::Unrolling, - int RhsVectors = trsolve_traits::RhsVectors - > -struct triangular_solver_selector; - -template -struct triangular_solver_selector -{ - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef blas_traits LhsProductTraits; - typedef typename LhsProductTraits::ExtractType ActualLhsType; - typedef Map, Aligned> MappedRhs; - static void run(const Lhs& lhs, Rhs& rhs) - { - ActualLhsType actualLhs = LhsProductTraits::extract(lhs); - - // FIXME find a way to allow an inner stride if packet_traits::size==1 - - bool useRhsDirectly = Rhs::InnerStrideAtCompileTime==1 || rhs.innerStride()==1; - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhs,rhs.size(), - (useRhsDirectly ? rhs.data() : 0)); - - if(!useRhsDirectly) - MappedRhs(actualRhs,rhs.size()) = rhs; - - triangular_solve_vector - ::run(actualLhs.cols(), actualLhs.data(), actualLhs.outerStride(), actualRhs); - - if(!useRhsDirectly) - rhs = MappedRhs(actualRhs, rhs.size()); - } -}; - -// the rhs is a matrix -template -struct triangular_solver_selector -{ - typedef typename Rhs::Scalar Scalar; - typedef blas_traits LhsProductTraits; - typedef typename LhsProductTraits::DirectLinearAccessType ActualLhsType; - - static void run(const Lhs& lhs, Rhs& rhs) - { - typename internal::add_const_on_value_type::type actualLhs = LhsProductTraits::extract(lhs); - - const Index size = lhs.rows(); - const Index othersize = Side==OnTheLeft? rhs.cols() : rhs.rows(); - - typedef internal::gemm_blocking_space<(Rhs::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Rhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxRowsAtCompileTime,4> BlockingType; - - BlockingType blocking(rhs.rows(), rhs.cols(), size, 1, false); - - triangular_solve_matrix - ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.innerStride(), rhs.outerStride(), blocking); - } -}; - -/*************************************************************************** -* meta-unrolling implementation -***************************************************************************/ - -template -struct triangular_solver_unroller; - -template -struct triangular_solver_unroller { - enum { - IsLower = ((Mode&Lower)==Lower), - DiagIndex = IsLower ? LoopIndex : Size - LoopIndex - 1, - StartIndex = IsLower ? 0 : DiagIndex+1 - }; - static void run(const Lhs& lhs, Rhs& rhs) - { - if (LoopIndex>0) - rhs.coeffRef(DiagIndex) -= lhs.row(DiagIndex).template segment(StartIndex).transpose() - .cwiseProduct(rhs.template segment(StartIndex)).sum(); - - if(!(Mode & UnitDiag)) - rhs.coeffRef(DiagIndex) /= lhs.coeff(DiagIndex,DiagIndex); - - triangular_solver_unroller::run(lhs,rhs); - } -}; - -template -struct triangular_solver_unroller { - static void run(const Lhs&, Rhs&) {} -}; - -template -struct triangular_solver_selector { - static void run(const Lhs& lhs, Rhs& rhs) - { triangular_solver_unroller::run(lhs,rhs); } -}; - -template -struct triangular_solver_selector { - static void run(const Lhs& lhs, Rhs& rhs) - { - Transpose trLhs(lhs); - Transpose trRhs(rhs); - - triangular_solver_unroller,Transpose, - ((Mode&Upper)==Upper ? Lower : Upper) | (Mode&UnitDiag), - 0,Rhs::SizeAtCompileTime>::run(trLhs,trRhs); - } -}; - -} // end namespace internal - -/*************************************************************************** -* TriangularView methods -***************************************************************************/ - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void TriangularViewImpl::solveInPlace(const MatrixBase& _other) const -{ - OtherDerived& other = _other.const_cast_derived(); - eigen_assert( derived().cols() == derived().rows() && ((Side==OnTheLeft && derived().cols() == other.rows()) || (Side==OnTheRight && derived().cols() == other.cols())) ); - eigen_assert((!(Mode & ZeroDiag)) && bool(Mode & (Upper|Lower))); - // If solving for a 0x0 matrix, nothing to do, simply return. - if (derived().cols() == 0) - return; - - enum { copy = (internal::traits::Flags & RowMajorBit) && OtherDerived::IsVectorAtCompileTime && OtherDerived::SizeAtCompileTime!=1}; - typedef typename internal::conditional::type, OtherDerived&>::type OtherCopy; - OtherCopy otherCopy(other); - - internal::triangular_solver_selector::type, - Side, Mode>::run(derived().nestedExpression(), otherCopy); - - if (copy) - other = otherCopy; -} - -template -template -const internal::triangular_solve_retval,Other> -TriangularViewImpl::solve(const MatrixBase& other) const -{ - return internal::triangular_solve_retval(derived(), other.derived()); -} -#endif - -namespace internal { - - -template -struct traits > -{ - typedef typename internal::plain_matrix_type_column_major::type ReturnType; -}; - -template struct triangular_solve_retval - : public ReturnByValue > -{ - typedef typename remove_all::type RhsNestedCleaned; - typedef ReturnByValue Base; - - triangular_solve_retval(const TriangularType& tri, const Rhs& rhs) - : m_triangularMatrix(tri), m_rhs(rhs) - {} - - inline Index rows() const { return m_rhs.rows(); } - inline Index cols() const { return m_rhs.cols(); } - - template inline void evalTo(Dest& dst) const - { - if(!is_same_dense(dst,m_rhs)) - dst = m_rhs; - m_triangularMatrix.template solveInPlace(dst); - } - - protected: - const TriangularType& m_triangularMatrix; - typename Rhs::Nested m_rhs; -}; - -} // namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVETRIANGULAR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h deleted file mode 100644 index 8a4adc22973..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/SolverBase.h +++ /dev/null @@ -1,130 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SOLVERBASE_H -#define EIGEN_SOLVERBASE_H - -namespace Eigen { - -namespace internal { - - - -} // end namespace internal - -/** \class SolverBase - * \brief A base class for matrix decomposition and solvers - * - * \tparam Derived the actual type of the decomposition/solver. - * - * Any matrix decomposition inheriting this base class provide the following API: - * - * \code - * MatrixType A, b, x; - * DecompositionType dec(A); - * x = dec.solve(b); // solve A * x = b - * x = dec.transpose().solve(b); // solve A^T * x = b - * x = dec.adjoint().solve(b); // solve A' * x = b - * \endcode - * - * \warning Currently, any other usage of transpose() and adjoint() are not supported and will produce compilation errors. - * - * \sa class PartialPivLU, class FullPivLU - */ -template -class SolverBase : public EigenBase -{ - public: - - typedef EigenBase Base; - typedef typename internal::traits::Scalar Scalar; - typedef Scalar CoeffReturnType; - - enum { - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, - internal::traits::ColsAtCompileTime>::ret), - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, - MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, - internal::traits::MaxColsAtCompileTime>::ret), - IsVectorAtCompileTime = internal::traits::MaxRowsAtCompileTime == 1 - || internal::traits::MaxColsAtCompileTime == 1 - }; - - /** Default constructor */ - SolverBase() - {} - - ~SolverBase() - {} - - using Base::derived; - - /** \returns an expression of the solution x of \f$ A x = b \f$ using the current decomposition of A. - */ - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(derived().rows()==b.rows() && "solve(): invalid number of rows of the right hand side matrix b"); - return Solve(derived(), b.derived()); - } - - /** \internal the return type of transpose() */ - typedef typename internal::add_const >::type ConstTransposeReturnType; - /** \returns an expression of the transposed of the factored matrix. - * - * A typical usage is to solve for the transposed problem A^T x = b: - * \code x = dec.transpose().solve(b); \endcode - * - * \sa adjoint(), solve() - */ - inline ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(derived()); - } - - /** \internal the return type of adjoint() */ - typedef typename internal::conditional::IsComplex, - CwiseUnaryOp, ConstTransposeReturnType>, - ConstTransposeReturnType - >::type AdjointReturnType; - /** \returns an expression of the adjoint of the factored matrix - * - * A typical usage is to solve for the adjoint problem A' x = b: - * \code x = dec.adjoint().solve(b); \endcode - * - * For real scalar types, this function is equivalent to transpose(). - * - * \sa transpose(), solve() - */ - inline AdjointReturnType adjoint() const - { - return AdjointReturnType(derived().transpose()); - } - - protected: -}; - -namespace internal { - -template -struct generic_xpr_base -{ - typedef SolverBase type; - -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SOLVERBASE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h deleted file mode 100644 index 88c8d989024..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/StableNorm.h +++ /dev/null @@ -1,221 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STABLENORM_H -#define EIGEN_STABLENORM_H - -namespace Eigen { - -namespace internal { - -template -inline void stable_norm_kernel(const ExpressionType& bl, Scalar& ssq, Scalar& scale, Scalar& invScale) -{ - Scalar maxCoeff = bl.cwiseAbs().maxCoeff(); - - if(maxCoeff>scale) - { - ssq = ssq * numext::abs2(scale/maxCoeff); - Scalar tmp = Scalar(1)/maxCoeff; - if(tmp > NumTraits::highest()) - { - invScale = NumTraits::highest(); - scale = Scalar(1)/invScale; - } - else if(maxCoeff>NumTraits::highest()) // we got a INF - { - invScale = Scalar(1); - scale = maxCoeff; - } - else - { - scale = maxCoeff; - invScale = tmp; - } - } - else if(maxCoeff!=maxCoeff) // we got a NaN - { - scale = maxCoeff; - } - - // TODO if the maxCoeff is much much smaller than the current scale, - // then we can neglect this sub vector - if(scale>Scalar(0)) // if scale==0, then bl is 0 - ssq += (bl*invScale).squaredNorm(); -} - -template -inline typename NumTraits::Scalar>::Real -blueNorm_impl(const EigenBase& _vec) -{ - typedef typename Derived::RealScalar RealScalar; - using std::pow; - using std::sqrt; - using std::abs; - const Derived& vec(_vec.derived()); - static bool initialized = false; - static RealScalar b1, b2, s1m, s2m, rbig, relerr; - if(!initialized) - { - int ibeta, it, iemin, iemax, iexp; - RealScalar eps; - // This program calculates the machine-dependent constants - // bl, b2, slm, s2m, relerr overfl - // from the "basic" machine-dependent numbers - // nbig, ibeta, it, iemin, iemax, rbig. - // The following define the basic machine-dependent constants. - // For portability, the PORT subprograms "ilmaeh" and "rlmach" - // are used. For any specific computer, each of the assignment - // statements can be replaced - ibeta = std::numeric_limits::radix; // base for floating-point numbers - it = std::numeric_limits::digits; // number of base-beta digits in mantissa - iemin = std::numeric_limits::min_exponent; // minimum exponent - iemax = std::numeric_limits::max_exponent; // maximum exponent - rbig = (std::numeric_limits::max)(); // largest floating-point number - - iexp = -((1-iemin)/2); - b1 = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // lower boundary of midrange - iexp = (iemax + 1 - it)/2; - b2 = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // upper boundary of midrange - - iexp = (2-iemin)/2; - s1m = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // scaling factor for lower range - iexp = - ((iemax+it)/2); - s2m = RealScalar(pow(RealScalar(ibeta),RealScalar(iexp))); // scaling factor for upper range - - eps = RealScalar(pow(double(ibeta), 1-it)); - relerr = sqrt(eps); // tolerance for neglecting asml - initialized = true; - } - Index n = vec.size(); - RealScalar ab2 = b2 / RealScalar(n); - RealScalar asml = RealScalar(0); - RealScalar amed = RealScalar(0); - RealScalar abig = RealScalar(0); - for(typename Derived::InnerIterator it(vec, 0); it; ++it) - { - RealScalar ax = abs(it.value()); - if(ax > ab2) abig += numext::abs2(ax*s2m); - else if(ax < b1) asml += numext::abs2(ax*s1m); - else amed += numext::abs2(ax); - } - if(amed!=amed) - return amed; // we got a NaN - if(abig > RealScalar(0)) - { - abig = sqrt(abig); - if(abig > rbig) // overflow, or *this contains INF values - return abig; // return INF - if(amed > RealScalar(0)) - { - abig = abig/s2m; - amed = sqrt(amed); - } - else - return abig/s2m; - } - else if(asml > RealScalar(0)) - { - if (amed > RealScalar(0)) - { - abig = sqrt(amed); - amed = sqrt(asml) / s1m; - } - else - return sqrt(asml)/s1m; - } - else - return sqrt(amed); - asml = numext::mini(abig, amed); - abig = numext::maxi(abig, amed); - if(asml <= abig*relerr) - return abig; - else - return abig * sqrt(RealScalar(1) + numext::abs2(asml/abig)); -} - -} // end namespace internal - -/** \returns the \em l2 norm of \c *this avoiding underflow and overflow. - * This version use a blockwise two passes algorithm: - * 1 - find the absolute largest coefficient \c s - * 2 - compute \f$ s \Vert \frac{*this}{s} \Vert \f$ in a standard way - * - * For architecture/scalar types supporting vectorization, this version - * is faster than blueNorm(). Otherwise the blueNorm() is much faster. - * - * \sa norm(), blueNorm(), hypotNorm() - */ -template -inline typename NumTraits::Scalar>::Real -MatrixBase::stableNorm() const -{ - using std::sqrt; - using std::abs; - const Index blockSize = 4096; - RealScalar scale(0); - RealScalar invScale(1); - RealScalar ssq(0); // sum of square - - typedef typename internal::nested_eval::type DerivedCopy; - typedef typename internal::remove_all::type DerivedCopyClean; - const DerivedCopy copy(derived()); - - enum { - CanAlign = ( (int(DerivedCopyClean::Flags)&DirectAccessBit) - || (int(internal::evaluator::Alignment)>0) // FIXME Alignment)>0 might not be enough - ) && (blockSize*sizeof(Scalar)*20) // if we cannot allocate on the stack, then let's not bother about this optimization - }; - typedef typename internal::conditional, internal::evaluator::Alignment>, - typename DerivedCopyClean::ConstSegmentReturnType>::type SegmentWrapper; - Index n = size(); - - if(n==1) - return abs(this->coeff(0)); - - Index bi = internal::first_default_aligned(copy); - if (bi>0) - internal::stable_norm_kernel(copy.head(bi), ssq, scale, invScale); - for (; bi -inline typename NumTraits::Scalar>::Real -MatrixBase::blueNorm() const -{ - return internal::blueNorm_impl(*this); -} - -/** \returns the \em l2 norm of \c *this avoiding undeflow and overflow. - * This version use a concatenation of hypot() calls, and it is very slow. - * - * \sa norm(), stableNorm() - */ -template -inline typename NumTraits::Scalar>::Real -MatrixBase::hypotNorm() const -{ - return this->cwiseAbs().redux(internal::scalar_hypot_op()); -} - -} // end namespace Eigen - -#endif // EIGEN_STABLENORM_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h deleted file mode 100644 index 513742f34b9..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Stride.h +++ /dev/null @@ -1,111 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STRIDE_H -#define EIGEN_STRIDE_H - -namespace Eigen { - -/** \class Stride - * \ingroup Core_Module - * - * \brief Holds strides information for Map - * - * This class holds the strides information for mapping arrays with strides with class Map. - * - * It holds two values: the inner stride and the outer stride. - * - * The inner stride is the pointer increment between two consecutive entries within a given row of a - * row-major matrix or within a given column of a column-major matrix. - * - * The outer stride is the pointer increment between two consecutive rows of a row-major matrix or - * between two consecutive columns of a column-major matrix. - * - * These two values can be passed either at compile-time as template parameters, or at runtime as - * arguments to the constructor. - * - * Indeed, this class takes two template parameters: - * \tparam _OuterStrideAtCompileTime the outer stride, or Dynamic if you want to specify it at runtime. - * \tparam _InnerStrideAtCompileTime the inner stride, or Dynamic if you want to specify it at runtime. - * - * Here is an example: - * \include Map_general_stride.cpp - * Output: \verbinclude Map_general_stride.out - * - * \sa class InnerStride, class OuterStride, \ref TopicStorageOrders - */ -template -class Stride -{ - public: - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - enum { - InnerStrideAtCompileTime = _InnerStrideAtCompileTime, - OuterStrideAtCompileTime = _OuterStrideAtCompileTime - }; - - /** Default constructor, for use when strides are fixed at compile time */ - EIGEN_DEVICE_FUNC - Stride() - : m_outer(OuterStrideAtCompileTime), m_inner(InnerStrideAtCompileTime) - { - eigen_assert(InnerStrideAtCompileTime != Dynamic && OuterStrideAtCompileTime != Dynamic); - } - - /** Constructor allowing to pass the strides at runtime */ - EIGEN_DEVICE_FUNC - Stride(Index outerStride, Index innerStride) - : m_outer(outerStride), m_inner(innerStride) - { - eigen_assert(innerStride>=0 && outerStride>=0); - } - - /** Copy constructor */ - EIGEN_DEVICE_FUNC - Stride(const Stride& other) - : m_outer(other.outer()), m_inner(other.inner()) - {} - - /** \returns the outer stride */ - EIGEN_DEVICE_FUNC - inline Index outer() const { return m_outer.value(); } - /** \returns the inner stride */ - EIGEN_DEVICE_FUNC - inline Index inner() const { return m_inner.value(); } - - protected: - internal::variable_if_dynamic m_outer; - internal::variable_if_dynamic m_inner; -}; - -/** \brief Convenience specialization of Stride to specify only an inner stride - * See class Map for some examples */ -template -class InnerStride : public Stride<0, Value> -{ - typedef Stride<0, Value> Base; - public: - EIGEN_DEVICE_FUNC InnerStride() : Base() {} - EIGEN_DEVICE_FUNC InnerStride(Index v) : Base(0, v) {} // FIXME making this explicit could break valid code -}; - -/** \brief Convenience specialization of Stride to specify only an outer stride - * See class Map for some examples */ -template -class OuterStride : public Stride -{ - typedef Stride Base; - public: - EIGEN_DEVICE_FUNC OuterStride() : Base() {} - EIGEN_DEVICE_FUNC OuterStride(Index v) : Base(v,0) {} // FIXME making this explicit could break valid code -}; - -} // end namespace Eigen - -#endif // EIGEN_STRIDE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h deleted file mode 100644 index d702009185e..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Swap.h +++ /dev/null @@ -1,67 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SWAP_H -#define EIGEN_SWAP_H - -namespace Eigen { - -namespace internal { - -// Overload default assignPacket behavior for swapping them -template -class generic_dense_assignment_kernel, Specialized> - : public generic_dense_assignment_kernel, BuiltIn> -{ -protected: - typedef generic_dense_assignment_kernel, BuiltIn> Base; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; - -public: - typedef typename Base::Scalar Scalar; - typedef typename Base::DstXprType DstXprType; - typedef swap_assign_op Functor; - - EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorTypeT &dst, const SrcEvaluatorTypeT &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - - template - void assignPacket(Index row, Index col) - { - PacketType tmp = m_src.template packet(row,col); - const_cast(m_src).template writePacket(row,col, m_dst.template packet(row,col)); - m_dst.template writePacket(row,col,tmp); - } - - template - void assignPacket(Index index) - { - PacketType tmp = m_src.template packet(index); - const_cast(m_src).template writePacket(index, m_dst.template packet(index)); - m_dst.template writePacket(index,tmp); - } - - // TODO find a simple way not to have to copy/paste this function from generic_dense_assignment_kernel, by simple I mean no CRTP (Gael) - template - void assignPacketByOuterInner(Index outer, Index inner) - { - Index row = Base::rowIndexByOuterInner(outer, inner); - Index col = Base::colIndexByOuterInner(outer, inner); - assignPacket(row, col); - } -}; - -} // namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SWAP_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h deleted file mode 100644 index 960dc4510c7..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpose.h +++ /dev/null @@ -1,405 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2008 Benoit Jacob -// Copyright (C) 2009-2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSPOSE_H -#define EIGEN_TRANSPOSE_H - -namespace Eigen { - -namespace internal { -template -struct traits > : public traits -{ - typedef typename ref_selector::type MatrixTypeNested; - typedef typename remove_reference::type MatrixTypeNestedPlain; - enum { - RowsAtCompileTime = MatrixType::ColsAtCompileTime, - ColsAtCompileTime = MatrixType::RowsAtCompileTime, - MaxRowsAtCompileTime = MatrixType::MaxColsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags0 = traits::Flags & ~(LvalueBit | NestByRefBit), - Flags1 = Flags0 | FlagsLvalueBit, - Flags = Flags1 ^ RowMajorBit, - InnerStrideAtCompileTime = inner_stride_at_compile_time::ret, - OuterStrideAtCompileTime = outer_stride_at_compile_time::ret - }; -}; -} - -template class TransposeImpl; - -/** \class Transpose - * \ingroup Core_Module - * - * \brief Expression of the transpose of a matrix - * - * \tparam MatrixType the type of the object of which we are taking the transpose - * - * This class represents an expression of the transpose of a matrix. - * It is the return type of MatrixBase::transpose() and MatrixBase::adjoint() - * and most of the time this is the only way it is used. - * - * \sa MatrixBase::transpose(), MatrixBase::adjoint() - */ -template class Transpose - : public TransposeImpl::StorageKind> -{ - public: - - typedef typename internal::ref_selector::non_const_type MatrixTypeNested; - - typedef typename TransposeImpl::StorageKind>::Base Base; - EIGEN_GENERIC_PUBLIC_INTERFACE(Transpose) - typedef typename internal::remove_all::type NestedExpression; - - EIGEN_DEVICE_FUNC - explicit inline Transpose(MatrixType& matrix) : m_matrix(matrix) {} - - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Transpose) - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_matrix.cols(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_matrix.rows(); } - - /** \returns the nested expression */ - EIGEN_DEVICE_FUNC - const typename internal::remove_all::type& - nestedExpression() const { return m_matrix; } - - /** \returns the nested expression */ - EIGEN_DEVICE_FUNC - typename internal::remove_reference::type& - nestedExpression() { return m_matrix; } - - /** \internal */ - void resize(Index nrows, Index ncols) { - m_matrix.resize(ncols,nrows); - } - - protected: - typename internal::ref_selector::non_const_type m_matrix; -}; - -namespace internal { - -template::ret> -struct TransposeImpl_base -{ - typedef typename dense_xpr_base >::type type; -}; - -template -struct TransposeImpl_base -{ - typedef typename dense_xpr_base >::type type; -}; - -} // end namespace internal - -// Generic API dispatcher -template -class TransposeImpl - : public internal::generic_xpr_base >::type -{ -public: - typedef typename internal::generic_xpr_base >::type Base; -}; - -template class TransposeImpl - : public internal::TransposeImpl_base::type -{ - public: - - typedef typename internal::TransposeImpl_base::type Base; - using Base::coeffRef; - EIGEN_DENSE_PUBLIC_INTERFACE(Transpose) - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TransposeImpl) - - EIGEN_DEVICE_FUNC inline Index innerStride() const { return derived().nestedExpression().innerStride(); } - EIGEN_DEVICE_FUNC inline Index outerStride() const { return derived().nestedExpression().outerStride(); } - - typedef typename internal::conditional< - internal::is_lvalue::value, - Scalar, - const Scalar - >::type ScalarWithConstIfNotLvalue; - - EIGEN_DEVICE_FUNC inline ScalarWithConstIfNotLvalue* data() { return derived().nestedExpression().data(); } - EIGEN_DEVICE_FUNC inline const Scalar* data() const { return derived().nestedExpression().data(); } - - // FIXME: shall we keep the const version of coeffRef? - EIGEN_DEVICE_FUNC - inline const Scalar& coeffRef(Index rowId, Index colId) const - { - return derived().nestedExpression().coeffRef(colId, rowId); - } - - EIGEN_DEVICE_FUNC - inline const Scalar& coeffRef(Index index) const - { - return derived().nestedExpression().coeffRef(index); - } - protected: - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TransposeImpl) -}; - -/** \returns an expression of the transpose of *this. - * - * Example: \include MatrixBase_transpose.cpp - * Output: \verbinclude MatrixBase_transpose.out - * - * \warning If you want to replace a matrix by its own transpose, do \b NOT do this: - * \code - * m = m.transpose(); // bug!!! caused by aliasing effect - * \endcode - * Instead, use the transposeInPlace() method: - * \code - * m.transposeInPlace(); - * \endcode - * which gives Eigen good opportunities for optimization, or alternatively you can also do: - * \code - * m = m.transpose().eval(); - * \endcode - * - * \sa transposeInPlace(), adjoint() */ -template -inline Transpose -DenseBase::transpose() -{ - return TransposeReturnType(derived()); -} - -/** This is the const version of transpose(). - * - * Make sure you read the warning for transpose() ! - * - * \sa transposeInPlace(), adjoint() */ -template -inline typename DenseBase::ConstTransposeReturnType -DenseBase::transpose() const -{ - return ConstTransposeReturnType(derived()); -} - -/** \returns an expression of the adjoint (i.e. conjugate transpose) of *this. - * - * Example: \include MatrixBase_adjoint.cpp - * Output: \verbinclude MatrixBase_adjoint.out - * - * \warning If you want to replace a matrix by its own adjoint, do \b NOT do this: - * \code - * m = m.adjoint(); // bug!!! caused by aliasing effect - * \endcode - * Instead, use the adjointInPlace() method: - * \code - * m.adjointInPlace(); - * \endcode - * which gives Eigen good opportunities for optimization, or alternatively you can also do: - * \code - * m = m.adjoint().eval(); - * \endcode - * - * \sa adjointInPlace(), transpose(), conjugate(), class Transpose, class internal::scalar_conjugate_op */ -template -inline const typename MatrixBase::AdjointReturnType -MatrixBase::adjoint() const -{ - return AdjointReturnType(this->transpose()); -} - -/*************************************************************************** -* "in place" transpose implementation -***************************************************************************/ - -namespace internal { - -template::size)) - && (internal::evaluator::Flags&PacketAccessBit) > -struct inplace_transpose_selector; - -template -struct inplace_transpose_selector { // square matrix - static void run(MatrixType& m) { - m.matrix().template triangularView().swap(m.matrix().transpose()); - } -}; - -// TODO: vectorized path is currently limited to LargestPacketSize x LargestPacketSize cases only. -template -struct inplace_transpose_selector { // PacketSize x PacketSize - static void run(MatrixType& m) { - typedef typename MatrixType::Scalar Scalar; - typedef typename internal::packet_traits::type Packet; - const Index PacketSize = internal::packet_traits::size; - const Index Alignment = internal::evaluator::Alignment; - PacketBlock A; - for (Index i=0; i(i,0); - internal::ptranspose(A); - for (Index i=0; i(m.rowIndexByOuterInner(i,0), m.colIndexByOuterInner(i,0), A.packet[i]); - } -}; - -template -struct inplace_transpose_selector { // non square matrix - static void run(MatrixType& m) { - if (m.rows()==m.cols()) - m.matrix().template triangularView().swap(m.matrix().transpose()); - else - m = m.transpose().eval(); - } -}; - -} // end namespace internal - -/** This is the "in place" version of transpose(): it replaces \c *this by its own transpose. - * Thus, doing - * \code - * m.transposeInPlace(); - * \endcode - * has the same effect on m as doing - * \code - * m = m.transpose().eval(); - * \endcode - * and is faster and also safer because in the latter line of code, forgetting the eval() results - * in a bug caused by \ref TopicAliasing "aliasing". - * - * Notice however that this method is only useful if you want to replace a matrix by its own transpose. - * If you just need the transpose of a matrix, use transpose(). - * - * \note if the matrix is not square, then \c *this must be a resizable matrix. - * This excludes (non-square) fixed-size matrices, block-expressions and maps. - * - * \sa transpose(), adjoint(), adjointInPlace() */ -template -inline void DenseBase::transposeInPlace() -{ - eigen_assert((rows() == cols() || (RowsAtCompileTime == Dynamic && ColsAtCompileTime == Dynamic)) - && "transposeInPlace() called on a non-square non-resizable matrix"); - internal::inplace_transpose_selector::run(derived()); -} - -/*************************************************************************** -* "in place" adjoint implementation -***************************************************************************/ - -/** This is the "in place" version of adjoint(): it replaces \c *this by its own transpose. - * Thus, doing - * \code - * m.adjointInPlace(); - * \endcode - * has the same effect on m as doing - * \code - * m = m.adjoint().eval(); - * \endcode - * and is faster and also safer because in the latter line of code, forgetting the eval() results - * in a bug caused by aliasing. - * - * Notice however that this method is only useful if you want to replace a matrix by its own adjoint. - * If you just need the adjoint of a matrix, use adjoint(). - * - * \note if the matrix is not square, then \c *this must be a resizable matrix. - * This excludes (non-square) fixed-size matrices, block-expressions and maps. - * - * \sa transpose(), adjoint(), transposeInPlace() */ -template -inline void MatrixBase::adjointInPlace() -{ - derived() = adjoint().eval(); -} - -#ifndef EIGEN_NO_DEBUG - -// The following is to detect aliasing problems in most common cases. - -namespace internal { - -template -struct check_transpose_aliasing_compile_time_selector -{ - enum { ret = bool(blas_traits::IsTransposed) != DestIsTransposed }; -}; - -template -struct check_transpose_aliasing_compile_time_selector > -{ - enum { ret = bool(blas_traits::IsTransposed) != DestIsTransposed - || bool(blas_traits::IsTransposed) != DestIsTransposed - }; -}; - -template -struct check_transpose_aliasing_run_time_selector -{ - static bool run(const Scalar* dest, const OtherDerived& src) - { - return (bool(blas_traits::IsTransposed) != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src)); - } -}; - -template -struct check_transpose_aliasing_run_time_selector > -{ - static bool run(const Scalar* dest, const CwiseBinaryOp& src) - { - return ((blas_traits::IsTransposed != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src.lhs()))) - || ((blas_traits::IsTransposed != DestIsTransposed) && (dest!=0 && dest==(const Scalar*)extract_data(src.rhs()))); - } -}; - -// the following selector, checkTransposeAliasing_impl, based on MightHaveTransposeAliasing, -// is because when the condition controlling the assert is known at compile time, ICC emits a warning. -// This is actually a good warning: in expressions that don't have any transposing, the condition is -// known at compile time to be false, and using that, we can avoid generating the code of the assert again -// and again for all these expressions that don't need it. - -template::IsTransposed,OtherDerived>::ret - > -struct checkTransposeAliasing_impl -{ - static void run(const Derived& dst, const OtherDerived& other) - { - eigen_assert((!check_transpose_aliasing_run_time_selector - ::IsTransposed,OtherDerived> - ::run(extract_data(dst), other)) - && "aliasing detected during transposition, use transposeInPlace() " - "or evaluate the rhs into a temporary using .eval()"); - - } -}; - -template -struct checkTransposeAliasing_impl -{ - static void run(const Derived&, const OtherDerived&) - { - } -}; - -template -void check_for_aliasing(const Dst &dst, const Src &src) -{ - internal::checkTransposeAliasing_impl::run(dst, src); -} - -} // end namespace internal - -#endif // EIGEN_NO_DEBUG - -} // end namespace Eigen - -#endif // EIGEN_TRANSPOSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h deleted file mode 100644 index 7718625e80b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Transpositions.h +++ /dev/null @@ -1,368 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010-2011 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRANSPOSITIONS_H -#define EIGEN_TRANSPOSITIONS_H - -namespace Eigen { - -template -class TranspositionsBase -{ - typedef internal::traits Traits; - - public: - - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - - Derived& derived() { return *static_cast(this); } - const Derived& derived() const { return *static_cast(this); } - - /** Copies the \a other transpositions into \c *this */ - template - Derived& operator=(const TranspositionsBase& other) - { - indices() = other.indices(); - return derived(); - } - - /** \returns the number of transpositions */ - Index size() const { return indices().size(); } - /** \returns the number of rows of the equivalent permutation matrix */ - Index rows() const { return indices().size(); } - /** \returns the number of columns of the equivalent permutation matrix */ - Index cols() const { return indices().size(); } - - /** Direct access to the underlying index vector */ - inline const StorageIndex& coeff(Index i) const { return indices().coeff(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& coeffRef(Index i) { return indices().coeffRef(i); } - /** Direct access to the underlying index vector */ - inline const StorageIndex& operator()(Index i) const { return indices()(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& operator()(Index i) { return indices()(i); } - /** Direct access to the underlying index vector */ - inline const StorageIndex& operator[](Index i) const { return indices()(i); } - /** Direct access to the underlying index vector */ - inline StorageIndex& operator[](Index i) { return indices()(i); } - - /** const version of indices(). */ - const IndicesType& indices() const { return derived().indices(); } - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return derived().indices(); } - - /** Resizes to given size. */ - inline void resize(Index newSize) - { - indices().resize(newSize); - } - - /** Sets \c *this to represents an identity transformation */ - void setIdentity() - { - for(StorageIndex i = 0; i < indices().size(); ++i) - coeffRef(i) = i; - } - - // FIXME: do we want such methods ? - // might be usefull when the target matrix expression is complex, e.g.: - // object.matrix().block(..,..,..,..) = trans * object.matrix().block(..,..,..,..); - /* - template - void applyForwardToRows(MatrixType& mat) const - { - for(Index k=0 ; k - void applyBackwardToRows(MatrixType& mat) const - { - for(Index k=size()-1 ; k>=0 ; --k) - if(m_indices(k)!=k) - mat.row(k).swap(mat.row(m_indices(k))); - } - */ - - /** \returns the inverse transformation */ - inline Transpose inverse() const - { return Transpose(derived()); } - - /** \returns the tranpose transformation */ - inline Transpose transpose() const - { return Transpose(derived()); } - - protected: -}; - -namespace internal { -template -struct traits > - : traits > -{ - typedef Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType; - typedef TranspositionsStorage StorageKind; -}; -} - -/** \class Transpositions - * \ingroup Core_Module - * - * \brief Represents a sequence of transpositions (row/column interchange) - * - * \tparam SizeAtCompileTime the number of transpositions, or Dynamic - * \tparam MaxSizeAtCompileTime the maximum number of transpositions, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it. - * - * This class represents a permutation transformation as a sequence of \em n transpositions - * \f$[T_{n-1} \ldots T_{i} \ldots T_{0}]\f$. It is internally stored as a vector of integers \c indices. - * Each transposition \f$ T_{i} \f$ applied on the left of a matrix (\f$ T_{i} M\f$) interchanges - * the rows \c i and \c indices[i] of the matrix \c M. - * A transposition applied on the right (e.g., \f$ M T_{i}\f$) yields a column interchange. - * - * Compared to the class PermutationMatrix, such a sequence of transpositions is what is - * computed during a decomposition with pivoting, and it is faster when applying the permutation in-place. - * - * To apply a sequence of transpositions to a matrix, simply use the operator * as in the following example: - * \code - * Transpositions tr; - * MatrixXf mat; - * mat = tr * mat; - * \endcode - * In this example, we detect that the matrix appears on both side, and so the transpositions - * are applied in-place without any temporary or extra copy. - * - * \sa class PermutationMatrix - */ - -template -class Transpositions : public TranspositionsBase > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - inline Transpositions() {} - - /** Copy constructor. */ - template - inline Transpositions(const TranspositionsBase& other) - : m_indices(other.indices()) {} - - /** Generic constructor from expression of the transposition indices. */ - template - explicit inline Transpositions(const MatrixBase& indices) : m_indices(indices) - {} - - /** Copies the \a other transpositions into \c *this */ - template - Transpositions& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - /** Constructs an uninitialized permutation matrix of given size. - */ - inline Transpositions(Index size) : m_indices(size) - {} - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - IndicesType m_indices; -}; - - -namespace internal { -template -struct traits,_PacketAccess> > - : traits > -{ - typedef Map, _PacketAccess> IndicesType; - typedef _StorageIndex StorageIndex; - typedef TranspositionsStorage StorageKind; -}; -} - -template -class Map,PacketAccess> - : public TranspositionsBase,PacketAccess> > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - explicit inline Map(const StorageIndex* indicesPtr) - : m_indices(indicesPtr) - {} - - inline Map(const StorageIndex* indicesPtr, Index size) - : m_indices(indicesPtr,size) - {} - - /** Copies the \a other transpositions into \c *this */ - template - Map& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** This is a special case of the templated operator=. Its purpose is to - * prevent a default operator= from hiding the templated operator=. - */ - Map& operator=(const Map& other) - { - m_indices = other.m_indices; - return *this; - } - #endif - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - IndicesType m_indices; -}; - -namespace internal { -template -struct traits > - : traits > -{ - typedef TranspositionsStorage StorageKind; -}; -} - -template -class TranspositionsWrapper - : public TranspositionsBase > -{ - typedef internal::traits Traits; - public: - - typedef TranspositionsBase Base; - typedef typename Traits::IndicesType IndicesType; - typedef typename IndicesType::Scalar StorageIndex; - - explicit inline TranspositionsWrapper(IndicesType& indices) - : m_indices(indices) - {} - - /** Copies the \a other transpositions into \c *this */ - template - TranspositionsWrapper& operator=(const TranspositionsBase& other) - { - return Base::operator=(other); - } - - /** const version of indices(). */ - const IndicesType& indices() const { return m_indices; } - - /** \returns a reference to the stored array representing the transpositions. */ - IndicesType& indices() { return m_indices; } - - protected: - - typename IndicesType::Nested m_indices; -}; - - - -/** \returns the \a matrix with the \a transpositions applied to the columns. - */ -template -EIGEN_DEVICE_FUNC -const Product -operator*(const MatrixBase &matrix, - const TranspositionsBase& transpositions) -{ - return Product - (matrix.derived(), transpositions.derived()); -} - -/** \returns the \a matrix with the \a transpositions applied to the rows. - */ -template -EIGEN_DEVICE_FUNC -const Product -operator*(const TranspositionsBase &transpositions, - const MatrixBase& matrix) -{ - return Product - (transpositions.derived(), matrix.derived()); -} - -// Template partial specialization for transposed/inverse transpositions - -namespace internal { - -template -struct traits > > - : traits -{}; - -} // end namespace internal - -template -class Transpose > -{ - typedef TranspositionsDerived TranspositionType; - typedef typename TranspositionType::IndicesType IndicesType; - public: - - explicit Transpose(const TranspositionType& t) : m_transpositions(t) {} - - Index size() const { return m_transpositions.size(); } - Index rows() const { return m_transpositions.size(); } - Index cols() const { return m_transpositions.size(); } - - /** \returns the \a matrix with the inverse transpositions applied to the columns. - */ - template friend - const Product - operator*(const MatrixBase& matrix, const Transpose& trt) - { - return Product(matrix.derived(), trt); - } - - /** \returns the \a matrix with the inverse transpositions applied to the rows. - */ - template - const Product - operator*(const MatrixBase& matrix) const - { - return Product(*this, matrix.derived()); - } - - const TranspositionType& nestedExpression() const { return m_transpositions; } - - protected: - const TranspositionType& m_transpositions; -}; - -} // end namespace Eigen - -#endif // EIGEN_TRANSPOSITIONS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h deleted file mode 100644 index 9abb7e31a8e..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/TriangularMatrix.h +++ /dev/null @@ -1,985 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Benoit Jacob -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULARMATRIX_H -#define EIGEN_TRIANGULARMATRIX_H - -namespace Eigen { - -namespace internal { - -template struct triangular_solve_retval; - -} - -/** \class TriangularBase - * \ingroup Core_Module - * - * \brief Base class for triangular part in a matrix - */ -template class TriangularBase : public EigenBase -{ - public: - - enum { - Mode = internal::traits::Mode, - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime, - - SizeAtCompileTime = (internal::size_at_compile_time::RowsAtCompileTime, - internal::traits::ColsAtCompileTime>::ret), - /**< This is equal to the number of coefficients, i.e. the number of - * rows times the number of columns, or to \a Dynamic if this is not - * known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */ - - MaxSizeAtCompileTime = (internal::size_at_compile_time::MaxRowsAtCompileTime, - internal::traits::MaxColsAtCompileTime>::ret) - - }; - typedef typename internal::traits::Scalar Scalar; - typedef typename internal::traits::StorageKind StorageKind; - typedef typename internal::traits::StorageIndex StorageIndex; - typedef typename internal::traits::FullMatrixType DenseMatrixType; - typedef DenseMatrixType DenseType; - typedef Derived const& Nested; - - EIGEN_DEVICE_FUNC - inline TriangularBase() { eigen_assert(!((Mode&UnitDiag) && (Mode&ZeroDiag))); } - - EIGEN_DEVICE_FUNC - inline Index rows() const { return derived().rows(); } - EIGEN_DEVICE_FUNC - inline Index cols() const { return derived().cols(); } - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return derived().outerStride(); } - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return derived().innerStride(); } - - // dummy resize function - void resize(Index rows, Index cols) - { - EIGEN_UNUSED_VARIABLE(rows); - EIGEN_UNUSED_VARIABLE(cols); - eigen_assert(rows==this->rows() && cols==this->cols()); - } - - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const { return derived().coeff(row,col); } - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) { return derived().coeffRef(row,col); } - - /** \see MatrixBase::copyCoeff(row,col) - */ - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE void copyCoeff(Index row, Index col, Other& other) - { - derived().coeffRef(row, col) = other.coeff(row, col); - } - - EIGEN_DEVICE_FUNC - inline Scalar operator()(Index row, Index col) const - { - check_coordinates(row, col); - return coeff(row,col); - } - EIGEN_DEVICE_FUNC - inline Scalar& operator()(Index row, Index col) - { - check_coordinates(row, col); - return coeffRef(row,col); - } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_DEVICE_FUNC - inline const Derived& derived() const { return *static_cast(this); } - EIGEN_DEVICE_FUNC - inline Derived& derived() { return *static_cast(this); } - #endif // not EIGEN_PARSED_BY_DOXYGEN - - template - EIGEN_DEVICE_FUNC - void evalTo(MatrixBase &other) const; - template - EIGEN_DEVICE_FUNC - void evalToLazy(MatrixBase &other) const; - - EIGEN_DEVICE_FUNC - DenseMatrixType toDenseMatrix() const - { - DenseMatrixType res(rows(), cols()); - evalToLazy(res); - return res; - } - - protected: - - void check_coordinates(Index row, Index col) const - { - EIGEN_ONLY_USED_FOR_DEBUG(row); - EIGEN_ONLY_USED_FOR_DEBUG(col); - eigen_assert(col>=0 && col=0 && row=row) - || (mode==Lower && col<=row) - || ((mode==StrictlyUpper || mode==UnitUpper) && col>row) - || ((mode==StrictlyLower || mode==UnitLower) && col -struct traits > : traits -{ - typedef typename ref_selector::non_const_type MatrixTypeNested; - typedef typename remove_reference::type MatrixTypeNestedNonRef; - typedef typename remove_all::type MatrixTypeNestedCleaned; - typedef typename MatrixType::PlainObject FullMatrixType; - typedef MatrixType ExpressionType; - enum { - Mode = _Mode, - FlagsLvalueBit = is_lvalue::value ? LvalueBit : 0, - Flags = (MatrixTypeNestedCleaned::Flags & (HereditaryBits | FlagsLvalueBit) & (~(PacketAccessBit | DirectAccessBit | LinearAccessBit))) - }; -}; -} - -template class TriangularViewImpl; - -template class TriangularView - : public TriangularViewImpl<_MatrixType, _Mode, typename internal::traits<_MatrixType>::StorageKind > -{ - public: - - typedef TriangularViewImpl<_MatrixType, _Mode, typename internal::traits<_MatrixType>::StorageKind > Base; - typedef typename internal::traits::Scalar Scalar; - typedef _MatrixType MatrixType; - - protected: - typedef typename internal::traits::MatrixTypeNested MatrixTypeNested; - typedef typename internal::traits::MatrixTypeNestedNonRef MatrixTypeNestedNonRef; - - typedef typename internal::remove_all::type MatrixConjugateReturnType; - - public: - - typedef typename internal::traits::StorageKind StorageKind; - typedef typename internal::traits::MatrixTypeNestedCleaned NestedExpression; - - enum { - Mode = _Mode, - Flags = internal::traits::Flags, - TransposeMode = (Mode & Upper ? Lower : 0) - | (Mode & Lower ? Upper : 0) - | (Mode & (UnitDiag)) - | (Mode & (ZeroDiag)), - IsVectorAtCompileTime = false - }; - - EIGEN_DEVICE_FUNC - explicit inline TriangularView(MatrixType& matrix) : m_matrix(matrix) - {} - - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TriangularView) - - /** \copydoc EigenBase::rows() */ - EIGEN_DEVICE_FUNC - inline Index rows() const { return m_matrix.rows(); } - /** \copydoc EigenBase::cols() */ - EIGEN_DEVICE_FUNC - inline Index cols() const { return m_matrix.cols(); } - - /** \returns a const reference to the nested expression */ - EIGEN_DEVICE_FUNC - const NestedExpression& nestedExpression() const { return m_matrix; } - - /** \returns a reference to the nested expression */ - EIGEN_DEVICE_FUNC - NestedExpression& nestedExpression() { return m_matrix; } - - typedef TriangularView ConjugateReturnType; - /** \sa MatrixBase::conjugate() const */ - EIGEN_DEVICE_FUNC - inline const ConjugateReturnType conjugate() const - { return ConjugateReturnType(m_matrix.conjugate()); } - - typedef TriangularView AdjointReturnType; - /** \sa MatrixBase::adjoint() const */ - EIGEN_DEVICE_FUNC - inline const AdjointReturnType adjoint() const - { return AdjointReturnType(m_matrix.adjoint()); } - - typedef TriangularView TransposeReturnType; - /** \sa MatrixBase::transpose() */ - EIGEN_DEVICE_FUNC - inline TransposeReturnType transpose() - { - EIGEN_STATIC_ASSERT_LVALUE(MatrixType) - typename MatrixType::TransposeReturnType tmp(m_matrix); - return TransposeReturnType(tmp); - } - - typedef TriangularView ConstTransposeReturnType; - /** \sa MatrixBase::transpose() const */ - EIGEN_DEVICE_FUNC - inline const ConstTransposeReturnType transpose() const - { - return ConstTransposeReturnType(m_matrix.transpose()); - } - - template - EIGEN_DEVICE_FUNC - inline const Solve - solve(const MatrixBase& other) const - { return Solve(*this, other.derived()); } - - // workaround MSVC ICE - #if EIGEN_COMP_MSVC - template - EIGEN_DEVICE_FUNC - inline const internal::triangular_solve_retval - solve(const MatrixBase& other) const - { return Base::template solve(other); } - #else - using Base::solve; - #endif - - /** \returns a selfadjoint view of the referenced triangular part which must be either \c #Upper or \c #Lower. - * - * This is a shortcut for \code this->nestedExpression().selfadjointView<(*this)::Mode>() \endcode - * \sa MatrixBase::selfadjointView() */ - EIGEN_DEVICE_FUNC - SelfAdjointView selfadjointView() - { - EIGEN_STATIC_ASSERT((Mode&(UnitDiag|ZeroDiag))==0,PROGRAMMING_ERROR); - return SelfAdjointView(m_matrix); - } - - /** This is the const version of selfadjointView() */ - EIGEN_DEVICE_FUNC - const SelfAdjointView selfadjointView() const - { - EIGEN_STATIC_ASSERT((Mode&(UnitDiag|ZeroDiag))==0,PROGRAMMING_ERROR); - return SelfAdjointView(m_matrix); - } - - - /** \returns the determinant of the triangular matrix - * \sa MatrixBase::determinant() */ - EIGEN_DEVICE_FUNC - Scalar determinant() const - { - if (Mode & UnitDiag) - return 1; - else if (Mode & ZeroDiag) - return 0; - else - return m_matrix.diagonal().prod(); - } - - protected: - - MatrixTypeNested m_matrix; -}; - -/** \ingroup Core_Module - * - * \brief Base class for a triangular part in a \b dense matrix - * - * This class is an abstract base class of class TriangularView, and objects of type TriangularViewImpl cannot be instantiated. - * It extends class TriangularView with additional methods which available for dense expressions only. - * - * \sa class TriangularView, MatrixBase::triangularView() - */ -template class TriangularViewImpl<_MatrixType,_Mode,Dense> - : public TriangularBase > -{ - public: - - typedef TriangularView<_MatrixType, _Mode> TriangularViewType; - typedef TriangularBase Base; - typedef typename internal::traits::Scalar Scalar; - - typedef _MatrixType MatrixType; - typedef typename MatrixType::PlainObject DenseMatrixType; - typedef DenseMatrixType PlainObject; - - public: - using Base::evalToLazy; - using Base::derived; - - typedef typename internal::traits::StorageKind StorageKind; - - enum { - Mode = _Mode, - Flags = internal::traits::Flags - }; - - /** \returns the outer-stride of the underlying dense matrix - * \sa DenseCoeffsBase::outerStride() */ - EIGEN_DEVICE_FUNC - inline Index outerStride() const { return derived().nestedExpression().outerStride(); } - /** \returns the inner-stride of the underlying dense matrix - * \sa DenseCoeffsBase::innerStride() */ - EIGEN_DEVICE_FUNC - inline Index innerStride() const { return derived().nestedExpression().innerStride(); } - - /** \sa MatrixBase::operator+=() */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator+=(const DenseBase& other) { - internal::call_assignment_no_alias(derived(), other.derived(), internal::add_assign_op()); - return derived(); - } - /** \sa MatrixBase::operator-=() */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator-=(const DenseBase& other) { - internal::call_assignment_no_alias(derived(), other.derived(), internal::sub_assign_op()); - return derived(); - } - - /** \sa MatrixBase::operator*=() */ - EIGEN_DEVICE_FUNC - TriangularViewType& operator*=(const typename internal::traits::Scalar& other) { return *this = derived().nestedExpression() * other; } - /** \sa DenseBase::operator/=() */ - EIGEN_DEVICE_FUNC - TriangularViewType& operator/=(const typename internal::traits::Scalar& other) { return *this = derived().nestedExpression() / other; } - - /** \sa MatrixBase::fill() */ - EIGEN_DEVICE_FUNC - void fill(const Scalar& value) { setConstant(value); } - /** \sa MatrixBase::setConstant() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setConstant(const Scalar& value) - { return *this = MatrixType::Constant(derived().rows(), derived().cols(), value); } - /** \sa MatrixBase::setZero() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setZero() { return setConstant(Scalar(0)); } - /** \sa MatrixBase::setOnes() */ - EIGEN_DEVICE_FUNC - TriangularViewType& setOnes() { return setConstant(Scalar(1)); } - - /** \sa MatrixBase::coeff() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar coeff(Index row, Index col) const - { - Base::check_coordinates_internal(row, col); - return derived().nestedExpression().coeff(row, col); - } - - /** \sa MatrixBase::coeffRef() - * \warning the coordinates must fit into the referenced triangular part - */ - EIGEN_DEVICE_FUNC - inline Scalar& coeffRef(Index row, Index col) - { - EIGEN_STATIC_ASSERT_LVALUE(TriangularViewType); - Base::check_coordinates_internal(row, col); - return derived().nestedExpression().coeffRef(row, col); - } - - /** Assigns a triangular matrix to a triangular part of a dense matrix */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const TriangularBase& other); - - /** Shortcut for\code *this = other.other.triangularView<(*this)::Mode>() \endcode */ - template - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const MatrixBase& other); - -#ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_DEVICE_FUNC - TriangularViewType& operator=(const TriangularViewImpl& other) - { return *this = other.derived().nestedExpression(); } - - /** \deprecated */ - template - EIGEN_DEVICE_FUNC - void lazyAssign(const TriangularBase& other); - - /** \deprecated */ - template - EIGEN_DEVICE_FUNC - void lazyAssign(const MatrixBase& other); -#endif - - /** Efficient triangular matrix times vector/matrix product */ - template - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& rhs) const - { - return Product(derived(), rhs.derived()); - } - - /** Efficient vector/matrix times triangular matrix product */ - template friend - EIGEN_DEVICE_FUNC - const Product - operator*(const MatrixBase& lhs, const TriangularViewImpl& rhs) - { - return Product(lhs.derived(),rhs.derived()); - } - - /** \returns the product of the inverse of \c *this with \a other, \a *this being triangular. - * - * This function computes the inverse-matrix matrix product inverse(\c *this) * \a other if - * \a Side==OnTheLeft (the default), or the right-inverse-multiply \a other * inverse(\c *this) if - * \a Side==OnTheRight. - * - * Note that the template parameter \c Side can be ommitted, in which case \c Side==OnTheLeft - * - * The matrix \c *this must be triangular and invertible (i.e., all the coefficients of the - * diagonal must be non zero). It works as a forward (resp. backward) substitution if \c *this - * is an upper (resp. lower) triangular matrix. - * - * Example: \include Triangular_solve.cpp - * Output: \verbinclude Triangular_solve.out - * - * This function returns an expression of the inverse-multiply and can works in-place if it is assigned - * to the same matrix or vector \a other. - * - * For users coming from BLAS, this function (and more specifically solveInPlace()) offer - * all the operations supported by the \c *TRSV and \c *TRSM BLAS routines. - * - * \sa TriangularView::solveInPlace() - */ - template - EIGEN_DEVICE_FUNC - inline const internal::triangular_solve_retval - solve(const MatrixBase& other) const; - - /** "in-place" version of TriangularView::solve() where the result is written in \a other - * - * \warning The parameter is only marked 'const' to make the C++ compiler accept a temporary expression here. - * This function will const_cast it, so constness isn't honored here. - * - * Note that the template parameter \c Side can be ommitted, in which case \c Side==OnTheLeft - * - * See TriangularView:solve() for the details. - */ - template - EIGEN_DEVICE_FUNC - void solveInPlace(const MatrixBase& other) const; - - template - EIGEN_DEVICE_FUNC - void solveInPlace(const MatrixBase& other) const - { return solveInPlace(other); } - - /** Swaps the coefficients of the common triangular parts of two matrices */ - template - EIGEN_DEVICE_FUNC -#ifdef EIGEN_PARSED_BY_DOXYGEN - void swap(TriangularBase &other) -#else - void swap(TriangularBase const & other) -#endif - { - EIGEN_STATIC_ASSERT_LVALUE(OtherDerived); - call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); - } - - /** \deprecated - * Shortcut for \code (*this).swap(other.triangularView<(*this)::Mode>()) \endcode */ - template - EIGEN_DEVICE_FUNC - void swap(MatrixBase const & other) - { - EIGEN_STATIC_ASSERT_LVALUE(OtherDerived); - call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op()); - } - - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE void _solve_impl(const RhsType &rhs, DstType &dst) const { - if(!internal::is_same_dense(dst,rhs)) - dst = rhs; - this->solveInPlace(dst); - } - - template - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE TriangularViewType& _assignProduct(const ProductType& prod, const Scalar& alpha, bool beta); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(TriangularViewImpl) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TriangularViewImpl) - -}; - -/*************************************************************************** -* Implementation of triangular evaluation/assignment -***************************************************************************/ - -#ifndef EIGEN_PARSED_BY_DOXYGEN -// FIXME should we keep that possibility -template -template -inline TriangularView& -TriangularViewImpl::operator=(const MatrixBase& other) -{ - internal::call_assignment_no_alias(derived(), other.derived(), internal::assign_op()); - return derived(); -} - -// FIXME should we keep that possibility -template -template -void TriangularViewImpl::lazyAssign(const MatrixBase& other) -{ - internal::call_assignment_no_alias(derived(), other.template triangularView()); -} - - - -template -template -inline TriangularView& -TriangularViewImpl::operator=(const TriangularBase& other) -{ - eigen_assert(Mode == int(OtherDerived::Mode)); - internal::call_assignment(derived(), other.derived()); - return derived(); -} - -template -template -void TriangularViewImpl::lazyAssign(const TriangularBase& other) -{ - eigen_assert(Mode == int(OtherDerived::Mode)); - internal::call_assignment_no_alias(derived(), other.derived()); -} -#endif - -/*************************************************************************** -* Implementation of TriangularBase methods -***************************************************************************/ - -/** Assigns a triangular or selfadjoint matrix to a dense matrix. - * If the matrix is triangular, the opposite part is set to zero. */ -template -template -void TriangularBase::evalTo(MatrixBase &other) const -{ - evalToLazy(other.derived()); -} - -/*************************************************************************** -* Implementation of TriangularView methods -***************************************************************************/ - -/*************************************************************************** -* Implementation of MatrixBase methods -***************************************************************************/ - -/** - * \returns an expression of a triangular view extracted from the current matrix - * - * The parameter \a Mode can have the following values: \c #Upper, \c #StrictlyUpper, \c #UnitUpper, - * \c #Lower, \c #StrictlyLower, \c #UnitLower. - * - * Example: \include MatrixBase_triangularView.cpp - * Output: \verbinclude MatrixBase_triangularView.out - * - * \sa class TriangularView - */ -template -template -typename MatrixBase::template TriangularViewReturnType::Type -MatrixBase::triangularView() -{ - return typename TriangularViewReturnType::Type(derived()); -} - -/** This is the const version of MatrixBase::triangularView() */ -template -template -typename MatrixBase::template ConstTriangularViewReturnType::Type -MatrixBase::triangularView() const -{ - return typename ConstTriangularViewReturnType::Type(derived()); -} - -/** \returns true if *this is approximately equal to an upper triangular matrix, - * within the precision given by \a prec. - * - * \sa isLowerTriangular() - */ -template -bool MatrixBase::isUpperTriangular(const RealScalar& prec) const -{ - RealScalar maxAbsOnUpperPart = static_cast(-1); - for(Index j = 0; j < cols(); ++j) - { - Index maxi = numext::mini(j, rows()-1); - for(Index i = 0; i <= maxi; ++i) - { - RealScalar absValue = numext::abs(coeff(i,j)); - if(absValue > maxAbsOnUpperPart) maxAbsOnUpperPart = absValue; - } - } - RealScalar threshold = maxAbsOnUpperPart * prec; - for(Index j = 0; j < cols(); ++j) - for(Index i = j+1; i < rows(); ++i) - if(numext::abs(coeff(i, j)) > threshold) return false; - return true; -} - -/** \returns true if *this is approximately equal to a lower triangular matrix, - * within the precision given by \a prec. - * - * \sa isUpperTriangular() - */ -template -bool MatrixBase::isLowerTriangular(const RealScalar& prec) const -{ - RealScalar maxAbsOnLowerPart = static_cast(-1); - for(Index j = 0; j < cols(); ++j) - for(Index i = j; i < rows(); ++i) - { - RealScalar absValue = numext::abs(coeff(i,j)); - if(absValue > maxAbsOnLowerPart) maxAbsOnLowerPart = absValue; - } - RealScalar threshold = maxAbsOnLowerPart * prec; - for(Index j = 1; j < cols(); ++j) - { - Index maxi = numext::mini(j, rows()-1); - for(Index i = 0; i < maxi; ++i) - if(numext::abs(coeff(i, j)) > threshold) return false; - } - return true; -} - - -/*************************************************************************** -**************************************************************************** -* Evaluators and Assignment of triangular expressions -*************************************************************************** -***************************************************************************/ - -namespace internal { - - -// TODO currently a triangular expression has the form TriangularView<.,.> -// in the future triangular-ness should be defined by the expression traits -// such that Transpose > is valid. (currently TriangularBase::transpose() is overloaded to make it work) -template -struct evaluator_traits > -{ - typedef typename storage_kind_to_evaluator_kind::Kind Kind; - typedef typename glue_shapes::Shape, TriangularShape>::type Shape; -}; - -template -struct unary_evaluator, IndexBased> - : evaluator::type> -{ - typedef TriangularView XprType; - typedef evaluator::type> Base; - unary_evaluator(const XprType &xpr) : Base(xpr.nestedExpression()) {} -}; - -// Additional assignment kinds: -struct Triangular2Triangular {}; -struct Triangular2Dense {}; -struct Dense2Triangular {}; - - -template struct triangular_assignment_loop; - - -/** \internal Specialization of the dense assignment kernel for triangular matrices. - * The main difference is that the triangular, diagonal, and opposite parts are processed through three different functions. - * \tparam UpLo must be either Lower or Upper - * \tparam Mode must be either 0, UnitDiag, ZeroDiag, or SelfAdjoint - */ -template -class triangular_dense_assignment_kernel : public generic_dense_assignment_kernel -{ -protected: - typedef generic_dense_assignment_kernel Base; - typedef typename Base::DstXprType DstXprType; - typedef typename Base::SrcXprType SrcXprType; - using Base::m_dst; - using Base::m_src; - using Base::m_functor; -public: - - typedef typename Base::DstEvaluatorType DstEvaluatorType; - typedef typename Base::SrcEvaluatorType SrcEvaluatorType; - typedef typename Base::Scalar Scalar; - typedef typename Base::AssignmentTraits AssignmentTraits; - - - EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr) - : Base(dst, src, func, dstExpr) - {} - -#ifdef EIGEN_INTERNAL_DEBUGGING - EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - Base::assignCoeff(row,col); - } -#else - using Base::assignCoeff; -#endif - - EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id) - { - if(Mode==UnitDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(1)); - else if(Mode==ZeroDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(0)); - else if(Mode==0) Base::assignCoeff(id,id); - } - - EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index row, Index col) - { - eigen_internal_assert(row!=col); - if(SetOpposite) - m_functor.assignCoeff(m_dst.coeffRef(row,col), Scalar(0)); - } -}; - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -void call_triangular_assignment_loop(DstXprType& dst, const SrcXprType& src, const Functor &func) -{ - typedef evaluator DstEvaluatorType; - typedef evaluator SrcEvaluatorType; - - SrcEvaluatorType srcEvaluator(src); - - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - DstEvaluatorType dstEvaluator(dst); - - typedef triangular_dense_assignment_kernel< Mode&(Lower|Upper),Mode&(UnitDiag|ZeroDiag|SelfAdjoint),SetOpposite, - DstEvaluatorType,SrcEvaluatorType,Functor> Kernel; - Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived()); - - enum { - unroll = DstXprType::SizeAtCompileTime != Dynamic - && SrcEvaluatorType::CoeffReadCost < HugeCost - && DstXprType::SizeAtCompileTime * (DstEvaluatorType::CoeffReadCost+SrcEvaluatorType::CoeffReadCost) / 2 <= EIGEN_UNROLLING_LIMIT - }; - - triangular_assignment_loop::run(kernel); -} - -template -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE -void call_triangular_assignment_loop(DstXprType& dst, const SrcXprType& src) -{ - call_triangular_assignment_loop(dst, src, internal::assign_op()); -} - -template<> struct AssignmentKind { typedef Triangular2Triangular Kind; }; -template<> struct AssignmentKind { typedef Triangular2Dense Kind; }; -template<> struct AssignmentKind { typedef Dense2Triangular Kind; }; - - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - eigen_assert(int(DstXprType::Mode) == int(SrcXprType::Mode)); - - call_triangular_assignment_loop(dst, src, func); - } -}; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - call_triangular_assignment_loop(dst, src, func); - } -}; - -template< typename DstXprType, typename SrcXprType, typename Functor> -struct Assignment -{ - EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func) - { - call_triangular_assignment_loop(dst, src, func); - } -}; - - -template -struct triangular_assignment_loop -{ - // FIXME: this is not very clean, perhaps this information should be provided by the kernel? - typedef typename Kernel::DstEvaluatorType DstEvaluatorType; - typedef typename DstEvaluatorType::XprType DstXprType; - - enum { - col = (UnrollCount-1) / DstXprType::RowsAtCompileTime, - row = (UnrollCount-1) % DstXprType::RowsAtCompileTime - }; - - typedef typename Kernel::Scalar Scalar; - - EIGEN_DEVICE_FUNC - static inline void run(Kernel &kernel) - { - triangular_assignment_loop::run(kernel); - - if(row==col) - kernel.assignDiagonalCoeff(row); - else if( ((Mode&Lower) && row>col) || ((Mode&Upper) && row -struct triangular_assignment_loop -{ - EIGEN_DEVICE_FUNC - static inline void run(Kernel &) {} -}; - - - -// TODO: experiment with a recursive assignment procedure splitting the current -// triangular part into one rectangular and two triangular parts. - - -template -struct triangular_assignment_loop -{ - typedef typename Kernel::Scalar Scalar; - EIGEN_DEVICE_FUNC - static inline void run(Kernel &kernel) - { - for(Index j = 0; j < kernel.cols(); ++j) - { - Index maxi = numext::mini(j, kernel.rows()); - Index i = 0; - if (((Mode&Lower) && SetOpposite) || (Mode&Upper)) - { - for(; i < maxi; ++i) - if(Mode&Upper) kernel.assignCoeff(i, j); - else kernel.assignOppositeCoeff(i, j); - } - else - i = maxi; - - if(i -template -void TriangularBase::evalToLazy(MatrixBase &other) const -{ - other.derived().resize(this->rows(), this->cols()); - internal::call_triangular_assignment_loop(other.derived(), derived().nestedExpression()); -} - -namespace internal { - -// Triangular = Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - dst._assignProduct(src, 1, 0); - } -}; - -// Triangular += Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::add_assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op &) - { - dst._assignProduct(src, 1, 1); - } -}; - -// Triangular -= Product -template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar> -struct Assignment, internal::sub_assign_op::Scalar>, Dense2Triangular> -{ - typedef Product SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op &) - { - dst._assignProduct(src, -1, 1); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULARMATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h deleted file mode 100644 index d72fbf7e99d..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorBlock.h +++ /dev/null @@ -1,96 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_VECTORBLOCK_H -#define EIGEN_VECTORBLOCK_H - -namespace Eigen { - -namespace internal { -template -struct traits > - : public traits::Flags & RowMajorBit ? 1 : Size, - traits::Flags & RowMajorBit ? Size : 1> > -{ -}; -} - -/** \class VectorBlock - * \ingroup Core_Module - * - * \brief Expression of a fixed-size or dynamic-size sub-vector - * - * \tparam VectorType the type of the object in which we are taking a sub-vector - * \tparam Size size of the sub-vector we are taking at compile time (optional) - * - * This class represents an expression of either a fixed-size or dynamic-size sub-vector. - * It is the return type of DenseBase::segment(Index,Index) and DenseBase::segment(Index) and - * most of the time this is the only way it is used. - * - * However, if you want to directly maniputate sub-vector expressions, - * for instance if you want to write a function returning such an expression, you - * will need to use this class. - * - * Here is an example illustrating the dynamic case: - * \include class_VectorBlock.cpp - * Output: \verbinclude class_VectorBlock.out - * - * \note Even though this expression has dynamic size, in the case where \a VectorType - * has fixed size, this expression inherits a fixed maximal size which means that evaluating - * it does not cause a dynamic memory allocation. - * - * Here is an example illustrating the fixed-size case: - * \include class_FixedVectorBlock.cpp - * Output: \verbinclude class_FixedVectorBlock.out - * - * \sa class Block, DenseBase::segment(Index,Index,Index,Index), DenseBase::segment(Index,Index) - */ -template class VectorBlock - : public Block::Flags & RowMajorBit ? 1 : Size, - internal::traits::Flags & RowMajorBit ? Size : 1> -{ - typedef Block::Flags & RowMajorBit ? 1 : Size, - internal::traits::Flags & RowMajorBit ? Size : 1> Base; - enum { - IsColVector = !(internal::traits::Flags & RowMajorBit) - }; - public: - EIGEN_DENSE_PUBLIC_INTERFACE(VectorBlock) - - using Base::operator=; - - /** Dynamic-size constructor - */ - EIGEN_DEVICE_FUNC - inline VectorBlock(VectorType& vector, Index start, Index size) - : Base(vector, - IsColVector ? start : 0, IsColVector ? 0 : start, - IsColVector ? size : 1, IsColVector ? 1 : size) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorBlock); - } - - /** Fixed-size constructor - */ - EIGEN_DEVICE_FUNC - inline VectorBlock(VectorType& vector, Index start) - : Base(vector, IsColVector ? start : 0, IsColVector ? 0 : start) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorBlock); - } -}; - - -} // end namespace Eigen - -#endif // EIGEN_VECTORBLOCK_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h deleted file mode 100644 index 4fe267e9f11..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/VectorwiseOp.h +++ /dev/null @@ -1,695 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARTIAL_REDUX_H -#define EIGEN_PARTIAL_REDUX_H - -namespace Eigen { - -/** \class PartialReduxExpr - * \ingroup Core_Module - * - * \brief Generic expression of a partially reduxed matrix - * - * \tparam MatrixType the type of the matrix we are applying the redux operation - * \tparam MemberOp type of the member functor - * \tparam Direction indicates the direction of the redux (#Vertical or #Horizontal) - * - * This class represents an expression of a partial redux operator of a matrix. - * It is the return type of some VectorwiseOp functions, - * and most of the time this is the only way it is used. - * - * \sa class VectorwiseOp - */ - -template< typename MatrixType, typename MemberOp, int Direction> -class PartialReduxExpr; - -namespace internal { -template -struct traits > - : traits -{ - typedef typename MemberOp::result_type Scalar; - typedef typename traits::StorageKind StorageKind; - typedef typename traits::XprKind XprKind; - typedef typename MatrixType::Scalar InputScalar; - enum { - RowsAtCompileTime = Direction==Vertical ? 1 : MatrixType::RowsAtCompileTime, - ColsAtCompileTime = Direction==Horizontal ? 1 : MatrixType::ColsAtCompileTime, - MaxRowsAtCompileTime = Direction==Vertical ? 1 : MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = Direction==Horizontal ? 1 : MatrixType::MaxColsAtCompileTime, - Flags = RowsAtCompileTime == 1 ? RowMajorBit : 0, - TraversalSize = Direction==Vertical ? MatrixType::RowsAtCompileTime : MatrixType::ColsAtCompileTime - }; -}; -} - -template< typename MatrixType, typename MemberOp, int Direction> -class PartialReduxExpr : public internal::dense_xpr_base< PartialReduxExpr >::type, - internal::no_assignment_operator -{ - public: - - typedef typename internal::dense_xpr_base::type Base; - EIGEN_DENSE_PUBLIC_INTERFACE(PartialReduxExpr) - - EIGEN_DEVICE_FUNC - explicit PartialReduxExpr(const MatrixType& mat, const MemberOp& func = MemberOp()) - : m_matrix(mat), m_functor(func) {} - - EIGEN_DEVICE_FUNC - Index rows() const { return (Direction==Vertical ? 1 : m_matrix.rows()); } - EIGEN_DEVICE_FUNC - Index cols() const { return (Direction==Horizontal ? 1 : m_matrix.cols()); } - - EIGEN_DEVICE_FUNC - typename MatrixType::Nested nestedExpression() const { return m_matrix; } - - EIGEN_DEVICE_FUNC - const MemberOp& functor() const { return m_functor; } - - protected: - typename MatrixType::Nested m_matrix; - const MemberOp m_functor; -}; - -#define EIGEN_MEMBER_FUNCTOR(MEMBER,COST) \ - template \ - struct member_##MEMBER { \ - EIGEN_EMPTY_STRUCT_CTOR(member_##MEMBER) \ - typedef ResultType result_type; \ - template struct Cost \ - { enum { value = COST }; }; \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE \ - ResultType operator()(const XprType& mat) const \ - { return mat.MEMBER(); } \ - } - -namespace internal { - -EIGEN_MEMBER_FUNCTOR(squaredNorm, Size * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(norm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(stableNorm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(blueNorm, (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(hypotNorm, (Size-1) * functor_traits >::Cost ); -EIGEN_MEMBER_FUNCTOR(sum, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(mean, (Size-1)*NumTraits::AddCost + NumTraits::MulCost); -EIGEN_MEMBER_FUNCTOR(minCoeff, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(maxCoeff, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(all, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(any, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(count, (Size-1)*NumTraits::AddCost); -EIGEN_MEMBER_FUNCTOR(prod, (Size-1)*NumTraits::MulCost); - -template -struct member_lpnorm { - typedef ResultType result_type; - template struct Cost - { enum { value = (Size+5) * NumTraits::MulCost + (Size-1)*NumTraits::AddCost }; }; - EIGEN_DEVICE_FUNC member_lpnorm() {} - template - EIGEN_DEVICE_FUNC inline ResultType operator()(const XprType& mat) const - { return mat.template lpNorm

(); } -}; - -template -struct member_redux { - typedef typename result_of< - BinaryOp(const Scalar&,const Scalar&) - >::type result_type; - template struct Cost - { enum { value = (Size-1) * functor_traits::Cost }; }; - EIGEN_DEVICE_FUNC explicit member_redux(const BinaryOp func) : m_functor(func) {} - template - EIGEN_DEVICE_FUNC inline result_type operator()(const DenseBase& mat) const - { return mat.redux(m_functor); } - const BinaryOp m_functor; -}; -} - -/** \class VectorwiseOp - * \ingroup Core_Module - * - * \brief Pseudo expression providing partial reduction operations - * - * \tparam ExpressionType the type of the object on which to do partial reductions - * \tparam Direction indicates the direction of the redux (#Vertical or #Horizontal) - * - * This class represents a pseudo expression with partial reduction features. - * It is the return type of DenseBase::colwise() and DenseBase::rowwise() - * and most of the time this is the only way it is used. - * - * Example: \include MatrixBase_colwise.cpp - * Output: \verbinclude MatrixBase_colwise.out - * - * \sa DenseBase::colwise(), DenseBase::rowwise(), class PartialReduxExpr - */ -template class VectorwiseOp -{ - public: - - typedef typename ExpressionType::Scalar Scalar; - typedef typename ExpressionType::RealScalar RealScalar; - typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3 - typedef typename internal::ref_selector::non_const_type ExpressionTypeNested; - typedef typename internal::remove_all::type ExpressionTypeNestedCleaned; - - template class Functor, - typename Scalar_=Scalar> struct ReturnType - { - typedef PartialReduxExpr, - Direction - > Type; - }; - - template struct ReduxReturnType - { - typedef PartialReduxExpr, - Direction - > Type; - }; - - enum { - isVertical = (Direction==Vertical) ? 1 : 0, - isHorizontal = (Direction==Horizontal) ? 1 : 0 - }; - - protected: - - typedef typename internal::conditional::type SubVector; - /** \internal - * \returns the i-th subvector according to the \c Direction */ - EIGEN_DEVICE_FUNC - SubVector subVector(Index i) - { - return SubVector(m_matrix.derived(),i); - } - - /** \internal - * \returns the number of subvectors in the direction \c Direction */ - EIGEN_DEVICE_FUNC - Index subVectors() const - { return isVertical?m_matrix.cols():m_matrix.rows(); } - - template struct ExtendedType { - typedef Replicate Type; - }; - - /** \internal - * Replicates a vector to match the size of \c *this */ - template - EIGEN_DEVICE_FUNC - typename ExtendedType::Type - extendedTo(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isVertical, OtherDerived::MaxColsAtCompileTime==1), - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED) - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isHorizontal, OtherDerived::MaxRowsAtCompileTime==1), - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED) - return typename ExtendedType::Type - (other.derived(), - isVertical ? 1 : m_matrix.rows(), - isHorizontal ? 1 : m_matrix.cols()); - } - - template struct OppositeExtendedType { - typedef Replicate Type; - }; - - /** \internal - * Replicates a vector in the opposite direction to match the size of \c *this */ - template - EIGEN_DEVICE_FUNC - typename OppositeExtendedType::Type - extendedToOpposite(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isHorizontal, OtherDerived::MaxColsAtCompileTime==1), - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED) - EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(isVertical, OtherDerived::MaxRowsAtCompileTime==1), - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED) - return typename OppositeExtendedType::Type - (other.derived(), - isHorizontal ? 1 : m_matrix.rows(), - isVertical ? 1 : m_matrix.cols()); - } - - public: - EIGEN_DEVICE_FUNC - explicit inline VectorwiseOp(ExpressionType& matrix) : m_matrix(matrix) {} - - /** \internal */ - EIGEN_DEVICE_FUNC - inline const ExpressionType& _expression() const { return m_matrix; } - - /** \returns a row or column vector expression of \c *this reduxed by \a func - * - * The template parameter \a BinaryOp is the type of the functor - * of the custom redux operator. Note that func must be an associative operator. - * - * \sa class VectorwiseOp, DenseBase::colwise(), DenseBase::rowwise() - */ - template - EIGEN_DEVICE_FUNC - const typename ReduxReturnType::Type - redux(const BinaryOp& func = BinaryOp()) const - { return typename ReduxReturnType::Type(_expression(), internal::member_redux(func)); } - - typedef typename ReturnType::Type MinCoeffReturnType; - typedef typename ReturnType::Type MaxCoeffReturnType; - typedef typename ReturnType::Type SquaredNormReturnType; - typedef typename ReturnType::Type NormReturnType; - typedef typename ReturnType::Type BlueNormReturnType; - typedef typename ReturnType::Type StableNormReturnType; - typedef typename ReturnType::Type HypotNormReturnType; - typedef typename ReturnType::Type SumReturnType; - typedef typename ReturnType::Type MeanReturnType; - typedef typename ReturnType::Type AllReturnType; - typedef typename ReturnType::Type AnyReturnType; - typedef PartialReduxExpr, Direction> CountReturnType; - typedef typename ReturnType::Type ProdReturnType; - typedef Reverse ConstReverseReturnType; - typedef Reverse ReverseReturnType; - - template struct LpNormReturnType { - typedef PartialReduxExpr,Direction> Type; - }; - - /** \returns a row (or column) vector expression of the smallest coefficient - * of each column (or row) of the referenced expression. - * - * \warning the result is undefined if \c *this contains NaN. - * - * Example: \include PartialRedux_minCoeff.cpp - * Output: \verbinclude PartialRedux_minCoeff.out - * - * \sa DenseBase::minCoeff() */ - EIGEN_DEVICE_FUNC - const MinCoeffReturnType minCoeff() const - { return MinCoeffReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the largest coefficient - * of each column (or row) of the referenced expression. - * - * \warning the result is undefined if \c *this contains NaN. - * - * Example: \include PartialRedux_maxCoeff.cpp - * Output: \verbinclude PartialRedux_maxCoeff.out - * - * \sa DenseBase::maxCoeff() */ - EIGEN_DEVICE_FUNC - const MaxCoeffReturnType maxCoeff() const - { return MaxCoeffReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the squared norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_squaredNorm.cpp - * Output: \verbinclude PartialRedux_squaredNorm.out - * - * \sa DenseBase::squaredNorm() */ - EIGEN_DEVICE_FUNC - const SquaredNormReturnType squaredNorm() const - { return SquaredNormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_norm.cpp - * Output: \verbinclude PartialRedux_norm.out - * - * \sa DenseBase::norm() */ - EIGEN_DEVICE_FUNC - const NormReturnType norm() const - { return NormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * Example: \include PartialRedux_norm.cpp - * Output: \verbinclude PartialRedux_norm.out - * - * \sa DenseBase::norm() */ - template - EIGEN_DEVICE_FUNC - const typename LpNormReturnType

::Type lpNorm() const - { return typename LpNormReturnType

::Type(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, using - * Blue's algorithm. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::blueNorm() */ - EIGEN_DEVICE_FUNC - const BlueNormReturnType blueNorm() const - { return BlueNormReturnType(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, avoiding - * underflow and overflow. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::stableNorm() */ - EIGEN_DEVICE_FUNC - const StableNormReturnType stableNorm() const - { return StableNormReturnType(_expression()); } - - - /** \returns a row (or column) vector expression of the norm - * of each column (or row) of the referenced expression, avoiding - * underflow and overflow using a concatenation of hypot() calls. - * This is a vector with real entries, even if the original matrix has complex entries. - * - * \sa DenseBase::hypotNorm() */ - EIGEN_DEVICE_FUNC - const HypotNormReturnType hypotNorm() const - { return HypotNormReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the sum - * of each column (or row) of the referenced expression. - * - * Example: \include PartialRedux_sum.cpp - * Output: \verbinclude PartialRedux_sum.out - * - * \sa DenseBase::sum() */ - EIGEN_DEVICE_FUNC - const SumReturnType sum() const - { return SumReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the mean - * of each column (or row) of the referenced expression. - * - * \sa DenseBase::mean() */ - EIGEN_DEVICE_FUNC - const MeanReturnType mean() const - { return MeanReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * whether \b all coefficients of each respective column (or row) are \c true. - * This expression can be assigned to a vector with entries of type \c bool. - * - * \sa DenseBase::all() */ - EIGEN_DEVICE_FUNC - const AllReturnType all() const - { return AllReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * whether \b at \b least one coefficient of each respective column (or row) is \c true. - * This expression can be assigned to a vector with entries of type \c bool. - * - * \sa DenseBase::any() */ - EIGEN_DEVICE_FUNC - const AnyReturnType any() const - { return AnyReturnType(_expression()); } - - /** \returns a row (or column) vector expression representing - * the number of \c true coefficients of each respective column (or row). - * This expression can be assigned to a vector whose entries have the same type as is used to - * index entries of the original matrix; for dense matrices, this is \c std::ptrdiff_t . - * - * Example: \include PartialRedux_count.cpp - * Output: \verbinclude PartialRedux_count.out - * - * \sa DenseBase::count() */ - EIGEN_DEVICE_FUNC - const CountReturnType count() const - { return CountReturnType(_expression()); } - - /** \returns a row (or column) vector expression of the product - * of each column (or row) of the referenced expression. - * - * Example: \include PartialRedux_prod.cpp - * Output: \verbinclude PartialRedux_prod.out - * - * \sa DenseBase::prod() */ - EIGEN_DEVICE_FUNC - const ProdReturnType prod() const - { return ProdReturnType(_expression()); } - - - /** \returns a matrix expression - * where each column (or row) are reversed. - * - * Example: \include Vectorwise_reverse.cpp - * Output: \verbinclude Vectorwise_reverse.out - * - * \sa DenseBase::reverse() */ - EIGEN_DEVICE_FUNC - const ConstReverseReturnType reverse() const - { return ConstReverseReturnType( _expression() ); } - - /** \returns a writable matrix expression - * where each column (or row) are reversed. - * - * \sa reverse() const */ - EIGEN_DEVICE_FUNC - ReverseReturnType reverse() - { return ReverseReturnType( _expression() ); } - - typedef Replicate ReplicateReturnType; - EIGEN_DEVICE_FUNC - const ReplicateReturnType replicate(Index factor) const; - - /** - * \return an expression of the replication of each column (or row) of \c *this - * - * Example: \include DirectionWise_replicate.cpp - * Output: \verbinclude DirectionWise_replicate.out - * - * \sa VectorwiseOp::replicate(Index), DenseBase::replicate(), class Replicate - */ - // NOTE implemented here because of sunstudio's compilation errors - // isVertical*Factor+isHorizontal instead of (isVertical?Factor:1) to handle CUDA bug with ternary operator - template const Replicate - EIGEN_DEVICE_FUNC - replicate(Index factor = Factor) const - { - return Replicate - (_expression(),isVertical?factor:1,isHorizontal?factor:1); - } - -/////////// Artithmetic operators /////////// - - /** Copies the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - //eigen_assert((m_matrix.isNull()) == (other.isNull())); FIXME - return const_cast(m_matrix = extendedTo(other.derived())); - } - - /** Adds the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator+=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return const_cast(m_matrix += extendedTo(other.derived())); - } - - /** Substracts the vector \a other to each subvector of \c *this */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator-=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return const_cast(m_matrix -= extendedTo(other.derived())); - } - - /** Multiples each subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator*=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - m_matrix *= extendedTo(other.derived()); - return const_cast(m_matrix); - } - - /** Divides each subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - ExpressionType& operator/=(const DenseBase& other) - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - m_matrix /= extendedTo(other.derived()); - return const_cast(m_matrix); - } - - /** Returns the expression of the sum of the vector \a other to each subvector of \c *this */ - template EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator+(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix + extendedTo(other.derived()); - } - - /** Returns the expression of the difference between each subvector of \c *this and the vector \a other */ - template - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator-(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix - extendedTo(other.derived()); - } - - /** Returns the expression where each subvector is the product of the vector \a other - * by the corresponding subvector of \c *this */ - template EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - EIGEN_DEVICE_FUNC - operator*(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix * extendedTo(other.derived()); - } - - /** Returns the expression where each subvector is the quotient of the corresponding - * subvector of \c *this by the vector \a other */ - template - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename ExtendedType::Type> - operator/(const DenseBase& other) const - { - EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived) - EIGEN_STATIC_ASSERT_ARRAYXPR(ExpressionType) - EIGEN_STATIC_ASSERT_SAME_XPR_KIND(ExpressionType, OtherDerived) - return m_matrix / extendedTo(other.derived()); - } - - /** \returns an expression where each column (or row) of the referenced matrix are normalized. - * The referenced matrix is \b not modified. - * \sa MatrixBase::normalized(), normalize() - */ - EIGEN_DEVICE_FUNC - CwiseBinaryOp, - const ExpressionTypeNestedCleaned, - const typename OppositeExtendedType::Type>::Type> - normalized() const { return m_matrix.cwiseQuotient(extendedToOpposite(this->norm())); } - - - /** Normalize in-place each row or columns of the referenced matrix. - * \sa MatrixBase::normalize(), normalized() - */ - EIGEN_DEVICE_FUNC void normalize() { - m_matrix = this->normalized(); - } - - EIGEN_DEVICE_FUNC inline void reverseInPlace(); - -/////////// Geometry module /////////// - - typedef Homogeneous HomogeneousReturnType; - EIGEN_DEVICE_FUNC - HomogeneousReturnType homogeneous() const; - - typedef typename ExpressionType::PlainObject CrossReturnType; - template - EIGEN_DEVICE_FUNC - const CrossReturnType cross(const MatrixBase& other) const; - - enum { - HNormalized_Size = Direction==Vertical ? internal::traits::RowsAtCompileTime - : internal::traits::ColsAtCompileTime, - HNormalized_SizeMinusOne = HNormalized_Size==Dynamic ? Dynamic : HNormalized_Size-1 - }; - typedef Block::RowsAtCompileTime), - Direction==Horizontal ? int(HNormalized_SizeMinusOne) - : int(internal::traits::ColsAtCompileTime)> - HNormalized_Block; - typedef Block::RowsAtCompileTime), - Direction==Horizontal ? 1 : int(internal::traits::ColsAtCompileTime)> - HNormalized_Factors; - typedef CwiseBinaryOp::Scalar>, - const HNormalized_Block, - const Replicate > - HNormalizedReturnType; - - EIGEN_DEVICE_FUNC - const HNormalizedReturnType hnormalized() const; - - protected: - ExpressionTypeNested m_matrix; -}; - -//const colwise moved to DenseBase.h due to CUDA compiler bug - - -/** \returns a writable VectorwiseOp wrapper of *this providing additional partial reduction operations - * - * \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting - */ -template -inline typename DenseBase::ColwiseReturnType -DenseBase::colwise() -{ - return ColwiseReturnType(derived()); -} - -//const rowwise moved to DenseBase.h due to CUDA compiler bug - - -/** \returns a writable VectorwiseOp wrapper of *this providing additional partial reduction operations - * - * \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting - */ -template -inline typename DenseBase::RowwiseReturnType -DenseBase::rowwise() -{ - return RowwiseReturnType(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_PARTIAL_REDUX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h deleted file mode 100644 index 54c1883d98f..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/Visitor.h +++ /dev/null @@ -1,273 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_VISITOR_H -#define EIGEN_VISITOR_H - -namespace Eigen { - -namespace internal { - -template -struct visitor_impl -{ - enum { - col = (UnrollCount-1) / Derived::RowsAtCompileTime, - row = (UnrollCount-1) % Derived::RowsAtCompileTime - }; - - EIGEN_DEVICE_FUNC - static inline void run(const Derived &mat, Visitor& visitor) - { - visitor_impl::run(mat, visitor); - visitor(mat.coeff(row, col), row, col); - } -}; - -template -struct visitor_impl -{ - EIGEN_DEVICE_FUNC - static inline void run(const Derived &mat, Visitor& visitor) - { - return visitor.init(mat.coeff(0, 0), 0, 0); - } -}; - -template -struct visitor_impl -{ - EIGEN_DEVICE_FUNC - static inline void run(const Derived& mat, Visitor& visitor) - { - visitor.init(mat.coeff(0,0), 0, 0); - for(Index i = 1; i < mat.rows(); ++i) - visitor(mat.coeff(i, 0), i, 0); - for(Index j = 1; j < mat.cols(); ++j) - for(Index i = 0; i < mat.rows(); ++i) - visitor(mat.coeff(i, j), i, j); - } -}; - -// evaluator adaptor -template -class visitor_evaluator -{ -public: - EIGEN_DEVICE_FUNC - explicit visitor_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {} - - typedef typename XprType::Scalar Scalar; - typedef typename XprType::CoeffReturnType CoeffReturnType; - - enum { - RowsAtCompileTime = XprType::RowsAtCompileTime, - CoeffReadCost = internal::evaluator::CoeffReadCost - }; - - EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); } - EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); } - EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); } - - EIGEN_DEVICE_FUNC CoeffReturnType coeff(Index row, Index col) const - { return m_evaluator.coeff(row, col); } - -protected: - internal::evaluator m_evaluator; - const XprType &m_xpr; -}; -} // end namespace internal - -/** Applies the visitor \a visitor to the whole coefficients of the matrix or vector. - * - * The template parameter \a Visitor is the type of the visitor and provides the following interface: - * \code - * struct MyVisitor { - * // called for the first coefficient - * void init(const Scalar& value, Index i, Index j); - * // called for all other coefficients - * void operator() (const Scalar& value, Index i, Index j); - * }; - * \endcode - * - * \note compared to one or two \em for \em loops, visitors offer automatic - * unrolling for small fixed size matrix. - * - * \sa minCoeff(Index*,Index*), maxCoeff(Index*,Index*), DenseBase::redux() - */ -template -template -EIGEN_DEVICE_FUNC -void DenseBase::visit(Visitor& visitor) const -{ - typedef typename internal::visitor_evaluator ThisEvaluator; - ThisEvaluator thisEval(derived()); - - enum { - unroll = SizeAtCompileTime != Dynamic - && SizeAtCompileTime * ThisEvaluator::CoeffReadCost + (SizeAtCompileTime-1) * internal::functor_traits::Cost <= EIGEN_UNROLLING_LIMIT - }; - return internal::visitor_impl::run(thisEval, visitor); -} - -namespace internal { - -/** \internal - * \brief Base class to implement min and max visitors - */ -template -struct coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - Index row, col; - Scalar res; - EIGEN_DEVICE_FUNC - inline void init(const Scalar& value, Index i, Index j) - { - res = value; - row = i; - col = j; - } -}; - -/** \internal - * \brief Visitor computing the min coefficient with its value and coordinates - * - * \sa DenseBase::minCoeff(Index*, Index*) - */ -template -struct min_coeff_visitor : coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - EIGEN_DEVICE_FUNC - void operator() (const Scalar& value, Index i, Index j) - { - if(value < this->res) - { - this->res = value; - this->row = i; - this->col = j; - } - } -}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost - }; -}; - -/** \internal - * \brief Visitor computing the max coefficient with its value and coordinates - * - * \sa DenseBase::maxCoeff(Index*, Index*) - */ -template -struct max_coeff_visitor : coeff_visitor -{ - typedef typename Derived::Scalar Scalar; - EIGEN_DEVICE_FUNC - void operator() (const Scalar& value, Index i, Index j) - { - if(value > this->res) - { - this->res = value; - this->row = i; - this->col = j; - } - } -}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost - }; -}; - -} // end namespace internal - -/** \fn DenseBase::minCoeff(IndexType* rowId, IndexType* colId) const - * \returns the minimum of all coefficients of *this and puts in *row and *col its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(Index*), DenseBase::maxCoeff(Index*,Index*), DenseBase::visit(), DenseBase::minCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::minCoeff(IndexType* rowId, IndexType* colId) const -{ - internal::min_coeff_visitor minVisitor; - this->visit(minVisitor); - *rowId = minVisitor.row; - if (colId) *colId = minVisitor.col; - return minVisitor.res; -} - -/** \returns the minimum of all coefficients of *this and puts in *index its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::visit(), DenseBase::minCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::minCoeff(IndexType* index) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - internal::min_coeff_visitor minVisitor; - this->visit(minVisitor); - *index = IndexType((RowsAtCompileTime==1) ? minVisitor.col : minVisitor.row); - return minVisitor.res; -} - -/** \fn DenseBase::maxCoeff(IndexType* rowId, IndexType* colId) const - * \returns the maximum of all coefficients of *this and puts in *row and *col its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visit(), DenseBase::maxCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::maxCoeff(IndexType* rowPtr, IndexType* colPtr) const -{ - internal::max_coeff_visitor maxVisitor; - this->visit(maxVisitor); - *rowPtr = maxVisitor.row; - if (colPtr) *colPtr = maxVisitor.col; - return maxVisitor.res; -} - -/** \returns the maximum of all coefficients of *this and puts in *index its location. - * \warning the result is undefined if \c *this contains NaN. - * - * \sa DenseBase::maxCoeff(IndexType*,IndexType*), DenseBase::minCoeff(IndexType*,IndexType*), DenseBase::visitor(), DenseBase::maxCoeff() - */ -template -template -EIGEN_DEVICE_FUNC -typename internal::traits::Scalar -DenseBase::maxCoeff(IndexType* index) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - internal::max_coeff_visitor maxVisitor; - this->visit(maxVisitor); - *index = (RowsAtCompileTime==1) ? maxVisitor.col : maxVisitor.row; - return maxVisitor.res; -} - -} // end namespace Eigen - -#endif // EIGEN_VISITOR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h deleted file mode 100644 index 9c25365090b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Complex.h +++ /dev/null @@ -1,103 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2014 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_CUDA_H -#define EIGEN_COMPLEX_CUDA_H - -// clang-format off - -namespace Eigen { - -namespace internal { - -#if defined(__CUDACC__) && defined(EIGEN_USE_GPU) - -// Many std::complex methods such as operator+, operator-, operator* and -// operator/ are not constexpr. Due to this, clang does not treat them as device -// functions and thus Eigen functors making use of these operators fail to -// compile. Here, we manually specialize these functors for complex types when -// building for CUDA to avoid non-constexpr methods. - -// Sum -template struct scalar_sum_op, const std::complex > : binary_op_base, const std::complex > { - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_sum_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - return std::complex(numext::real(a) + numext::real(b), - numext::imag(a) + numext::imag(b)); - } -}; - -template struct scalar_sum_op, std::complex > : scalar_sum_op, const std::complex > {}; - - -// Difference -template struct scalar_difference_op, const std::complex > : binary_op_base, const std::complex > { - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_difference_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - return std::complex(numext::real(a) - numext::real(b), - numext::imag(a) - numext::imag(b)); - } -}; - -template struct scalar_difference_op, std::complex > : scalar_difference_op, const std::complex > {}; - - -// Product -template struct scalar_product_op, const std::complex > : binary_op_base, const std::complex > { - enum { - Vectorizable = packet_traits>::HasMul - }; - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_product_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - const T a_real = numext::real(a); - const T a_imag = numext::imag(a); - const T b_real = numext::real(b); - const T b_imag = numext::imag(b); - return std::complex(a_real * b_real - a_imag * b_imag, - a_real * b_imag + a_imag * b_real); - } -}; - -template struct scalar_product_op, std::complex > : scalar_product_op, const std::complex > {}; - - -// Quotient -template struct scalar_quotient_op, const std::complex > : binary_op_base, const std::complex > { - enum { - Vectorizable = packet_traits>::HasDiv - }; - typedef typename std::complex result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex operator() (const std::complex& a, const std::complex& b) const { - const T a_real = numext::real(a); - const T a_imag = numext::imag(a); - const T b_real = numext::real(b); - const T b_imag = numext::imag(b); - const T norm = T(1) / (b_real * b_real + b_imag * b_imag); - return std::complex((a_real * b_real + a_imag * b_imag) * norm, - (a_imag * b_real - a_real * b_imag) * norm); - } -}; - -template struct scalar_quotient_op, std::complex > : scalar_quotient_op, const std::complex > {}; - -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_CUDA_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h deleted file mode 100644 index 59717b4fe6c..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/Half.h +++ /dev/null @@ -1,675 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -// -// The conversion routines are Copyright (c) Fabian Giesen, 2016. -// The original license follows: -// -// Copyright (c) Fabian Giesen, 2016 -// All rights reserved. -// Redistribution and use in source and binary forms, with or without -// modification, are permitted. -// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS -// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT -// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR -// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT -// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, -// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT -// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, -// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY -// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT -// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE -// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - -// Standard 16-bit float type, mostly useful for GPUs. Defines a new -// type Eigen::half (inheriting from CUDA's __half struct) with -// operator overloads such that it behaves basically as an arithmetic -// type. It will be quite slow on CPUs (so it is recommended to stay -// in float32_bits for CPUs, except for simple parameter conversions, I/O -// to disk and the likes), but fast on GPUs. - - -#ifndef EIGEN_HALF_CUDA_H -#define EIGEN_HALF_CUDA_H - -#if __cplusplus > 199711L -#define EIGEN_EXPLICIT_CAST(tgt_type) explicit operator tgt_type() -#else -#define EIGEN_EXPLICIT_CAST(tgt_type) operator tgt_type() -#endif - -#include - -namespace Eigen { - -struct half; - -namespace half_impl { - -#if !defined(EIGEN_HAS_CUDA_FP16) -// Make our own __half_raw definition that is similar to CUDA's. -struct __half_raw { - EIGEN_DEVICE_FUNC __half_raw() : x(0) {} - explicit EIGEN_DEVICE_FUNC __half_raw(unsigned short raw) : x(raw) {} - unsigned short x; -}; -#elif defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER < 90000 -// In CUDA < 9.0, __half is the equivalent of CUDA 9's __half_raw -typedef __half __half_raw; -#endif - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw raw_uint16_to_half(unsigned short x); -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw float_to_half_rtne(float ff); -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC float half_to_float(__half_raw h); - -struct half_base : public __half_raw { - EIGEN_DEVICE_FUNC half_base() {} - EIGEN_DEVICE_FUNC half_base(const half_base& h) : __half_raw(h) {} - EIGEN_DEVICE_FUNC half_base(const __half_raw& h) : __half_raw(h) {} -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER >= 90000 - EIGEN_DEVICE_FUNC half_base(const __half& h) : __half_raw(*(__half_raw*)&h) {} -#endif -}; - -} // namespace half_impl - -// Class definition. -struct half : public half_impl::half_base { - #if !defined(EIGEN_HAS_CUDA_FP16) || (defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER < 90000) - typedef half_impl::__half_raw __half_raw; - #endif - - EIGEN_DEVICE_FUNC half() {} - - EIGEN_DEVICE_FUNC half(const __half_raw& h) : half_impl::half_base(h) {} - EIGEN_DEVICE_FUNC half(const half& h) : half_impl::half_base(h) {} -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDACC_VER) && EIGEN_CUDACC_VER >= 90000 - EIGEN_DEVICE_FUNC half(const __half& h) : half_impl::half_base(h) {} -#endif - - explicit EIGEN_DEVICE_FUNC half(bool b) - : half_impl::half_base(half_impl::raw_uint16_to_half(b ? 0x3c00 : 0)) {} - template - explicit EIGEN_DEVICE_FUNC half(const T& val) - : half_impl::half_base(half_impl::float_to_half_rtne(static_cast(val))) {} - explicit EIGEN_DEVICE_FUNC half(float f) - : half_impl::half_base(half_impl::float_to_half_rtne(f)) {} - - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(bool) const { - // +0.0 and -0.0 become false, everything else becomes true. - return (x & 0x7fff) != 0; - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(signed char) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned char) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(short) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned short) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(int) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned int) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(long long) const { - return static_cast(half_impl::half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(unsigned long long) const { - return static_cast(half_to_float(*this)); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(float) const { - return half_impl::half_to_float(*this); - } - EIGEN_DEVICE_FUNC EIGEN_EXPLICIT_CAST(double) const { - return static_cast(half_impl::half_to_float(*this)); - } - - EIGEN_DEVICE_FUNC half& operator=(const half& other) { - x = other.x; - return *this; - } -}; - -} // end namespace Eigen - -namespace std { -template<> -struct numeric_limits { - static const bool is_specialized = true; - static const bool is_signed = true; - static const bool is_integer = false; - static const bool is_exact = false; - static const bool has_infinity = true; - static const bool has_quiet_NaN = true; - static const bool has_signaling_NaN = true; - static const float_denorm_style has_denorm = denorm_present; - static const bool has_denorm_loss = false; - static const std::float_round_style round_style = std::round_to_nearest; - static const bool is_iec559 = false; - static const bool is_bounded = false; - static const bool is_modulo = false; - static const int digits = 11; - static const int digits10 = 3; // according to http://half.sourceforge.net/structstd_1_1numeric__limits_3_01half__float_1_1half_01_4.html - static const int max_digits10 = 5; // according to http://half.sourceforge.net/structstd_1_1numeric__limits_3_01half__float_1_1half_01_4.html - static const int radix = 2; - static const int min_exponent = -13; - static const int min_exponent10 = -4; - static const int max_exponent = 16; - static const int max_exponent10 = 4; - static const bool traps = true; - static const bool tinyness_before = false; - - static Eigen::half (min)() { return Eigen::half_impl::raw_uint16_to_half(0x400); } - static Eigen::half lowest() { return Eigen::half_impl::raw_uint16_to_half(0xfbff); } - static Eigen::half (max)() { return Eigen::half_impl::raw_uint16_to_half(0x7bff); } - static Eigen::half epsilon() { return Eigen::half_impl::raw_uint16_to_half(0x0800); } - static Eigen::half round_error() { return Eigen::half(0.5); } - static Eigen::half infinity() { return Eigen::half_impl::raw_uint16_to_half(0x7c00); } - static Eigen::half quiet_NaN() { return Eigen::half_impl::raw_uint16_to_half(0x7e00); } - static Eigen::half signaling_NaN() { return Eigen::half_impl::raw_uint16_to_half(0x7e00); } - static Eigen::half denorm_min() { return Eigen::half_impl::raw_uint16_to_half(0x1); } -}; - -// If std::numeric_limits is specialized, should also specialize -// std::numeric_limits, std::numeric_limits, and -// std::numeric_limits -// https://stackoverflow.com/a/16519653/ -template<> -struct numeric_limits : numeric_limits {}; -template<> -struct numeric_limits : numeric_limits {}; -template<> -struct numeric_limits : numeric_limits {}; -} // end namespace std - -namespace Eigen { - -namespace half_impl { - -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - -// Intrinsics for native fp16 support. Note that on current hardware, -// these are no faster than float32_bits arithmetic (you need to use the half2 -// versions to get the ALU speed increased), but you do save the -// conversion steps back and forth. - -EIGEN_STRONG_INLINE __device__ half operator + (const half& a, const half& b) { - return __hadd(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator * (const half& a, const half& b) { - return __hmul(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator - (const half& a, const half& b) { - return __hsub(a, b); -} -EIGEN_STRONG_INLINE __device__ half operator / (const half& a, const half& b) { - float num = __half2float(a); - float denom = __half2float(b); - return __float2half(num / denom); -} -EIGEN_STRONG_INLINE __device__ half operator - (const half& a) { - return __hneg(a); -} -EIGEN_STRONG_INLINE __device__ half& operator += (half& a, const half& b) { - a = a + b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator *= (half& a, const half& b) { - a = a * b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator -= (half& a, const half& b) { - a = a - b; - return a; -} -EIGEN_STRONG_INLINE __device__ half& operator /= (half& a, const half& b) { - a = a / b; - return a; -} -EIGEN_STRONG_INLINE __device__ bool operator == (const half& a, const half& b) { - return __heq(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator != (const half& a, const half& b) { - return __hne(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator < (const half& a, const half& b) { - return __hlt(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator <= (const half& a, const half& b) { - return __hle(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator > (const half& a, const half& b) { - return __hgt(a, b); -} -EIGEN_STRONG_INLINE __device__ bool operator >= (const half& a, const half& b) { - return __hge(a, b); -} - -#else // Emulate support for half floats - -// Definitions for CPUs and older CUDA, mostly working through conversion -// to/from float32_bits. - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator + (const half& a, const half& b) { - return half(float(a) + float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator * (const half& a, const half& b) { - return half(float(a) * float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator - (const half& a, const half& b) { - return half(float(a) - float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator / (const half& a, const half& b) { - return half(float(a) / float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator - (const half& a) { - half result; - result.x = a.x ^ 0x8000; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator += (half& a, const half& b) { - a = half(float(a) + float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator *= (half& a, const half& b) { - a = half(float(a) * float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator -= (half& a, const half& b) { - a = half(float(a) - float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half& operator /= (half& a, const half& b) { - a = half(float(a) / float(b)); - return a; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator == (const half& a, const half& b) { - return numext::equal_strict(float(a),float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator != (const half& a, const half& b) { - return numext::not_equal_strict(float(a), float(b)); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator < (const half& a, const half& b) { - return float(a) < float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator <= (const half& a, const half& b) { - return float(a) <= float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator > (const half& a, const half& b) { - return float(a) > float(b); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool operator >= (const half& a, const half& b) { - return float(a) >= float(b); -} - -#endif // Emulate support for half floats - -// Division by an index. Do it in full float precision to avoid accuracy -// issues in converting the denominator to half. -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half operator / (const half& a, Index b) { - return half(static_cast(a) / static_cast(b)); -} - -// Conversion routines, including fallbacks for the host or older CUDA. -// Note that newer Intel CPUs (Haswell or newer) have vectorized versions of -// these in hardware. If we need more performance on older/other CPUs, they are -// also possible to vectorize directly. - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw raw_uint16_to_half(unsigned short x) { - __half_raw h; - h.x = x; - return h; -} - -union float32_bits { - unsigned int u; - float f; -}; - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC __half_raw float_to_half_rtne(float ff) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 - __half tmp_ff = __float2half(ff); - return *(__half_raw*)&tmp_ff; - -#elif defined(EIGEN_HAS_FP16_C) - __half_raw h; - h.x = _cvtss_sh(ff, 0); - return h; - -#else - float32_bits f; f.f = ff; - - const float32_bits f32infty = { 255 << 23 }; - const float32_bits f16max = { (127 + 16) << 23 }; - const float32_bits denorm_magic = { ((127 - 15) + (23 - 10) + 1) << 23 }; - unsigned int sign_mask = 0x80000000u; - __half_raw o; - o.x = static_cast(0x0u); - - unsigned int sign = f.u & sign_mask; - f.u ^= sign; - - // NOTE all the integer compares in this function can be safely - // compiled into signed compares since all operands are below - // 0x80000000. Important if you want fast straight SSE2 code - // (since there's no unsigned PCMPGTD). - - if (f.u >= f16max.u) { // result is Inf or NaN (all exponent bits set) - o.x = (f.u > f32infty.u) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf - } else { // (De)normalized number or zero - if (f.u < (113 << 23)) { // resulting FP16 is subnormal or zero - // use a magic value to align our 10 mantissa bits at the bottom of - // the float. as long as FP addition is round-to-nearest-even this - // just works. - f.f += denorm_magic.f; - - // and one integer subtract of the bias later, we have our final float! - o.x = static_cast(f.u - denorm_magic.u); - } else { - unsigned int mant_odd = (f.u >> 13) & 1; // resulting mantissa is odd - - // update exponent, rounding bias part 1 - f.u += ((unsigned int)(15 - 127) << 23) + 0xfff; - // rounding bias part 2 - f.u += mant_odd; - // take the bits! - o.x = static_cast(f.u >> 13); - } - } - - o.x |= static_cast(sign >> 16); - return o; -#endif -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC float half_to_float(__half_raw h) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 - return __half2float(h); - -#elif defined(EIGEN_HAS_FP16_C) - return _cvtsh_ss(h.x); - -#else - const float32_bits magic = { 113 << 23 }; - const unsigned int shifted_exp = 0x7c00 << 13; // exponent mask after shift - float32_bits o; - - o.u = (h.x & 0x7fff) << 13; // exponent/mantissa bits - unsigned int exp = shifted_exp & o.u; // just the exponent - o.u += (127 - 15) << 23; // exponent adjust - - // handle exponent special cases - if (exp == shifted_exp) { // Inf/NaN? - o.u += (128 - 16) << 23; // extra exp adjust - } else if (exp == 0) { // Zero/Denormal? - o.u += 1 << 23; // extra exp adjust - o.f -= magic.f; // renormalize - } - - o.u |= (h.x & 0x8000) << 16; // sign bit - return o.f; -#endif -} - -// --- standard functions --- - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isinf)(const half& a) { - return (a.x & 0x7fff) == 0x7c00; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isnan)(const half& a) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hisnan(a); -#else - return (a.x & 0x7fff) > 0x7c00; -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC bool (isfinite)(const half& a) { - return !(isinf EIGEN_NOT_A_MACRO (a)) && !(isnan EIGEN_NOT_A_MACRO (a)); -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half abs(const half& a) { - half result; - result.x = a.x & 0x7FFF; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half exp(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - return half(hexp(a)); -#else - return half(::expf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log(const half& a) { -#if defined(EIGEN_HAS_CUDA_FP16) && EIGEN_CUDACC_VER >= 80000 && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return half(::hlog(a)); -#else - return half(::logf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log1p(const half& a) { - return half(numext::log1p(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half log10(const half& a) { - return half(::log10f(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half sqrt(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - return half(hsqrt(a)); -#else - return half(::sqrtf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half pow(const half& a, const half& b) { - return half(::powf(float(a), float(b))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half sin(const half& a) { - return half(::sinf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half cos(const half& a) { - return half(::cosf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half tan(const half& a) { - return half(::tanf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half tanh(const half& a) { - return half(::tanhf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half floor(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 300 - return half(hfloor(a)); -#else - return half(::floorf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half ceil(const half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 300 - return half(hceil(a)); -#else - return half(::ceilf(float(a))); -#endif -} - -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half (min)(const half& a, const half& b) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hlt(b, a) ? b : a; -#else - const float f1 = static_cast(a); - const float f2 = static_cast(b); - return f2 < f1 ? b : a; -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC half (max)(const half& a, const half& b) { -#if defined(EIGEN_HAS_CUDA_FP16) && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return __hlt(a, b) ? b : a; -#else - const float f1 = static_cast(a); - const float f2 = static_cast(b); - return f1 < f2 ? b : a; -#endif -} - -EIGEN_ALWAYS_INLINE std::ostream& operator << (std::ostream& os, const half& v) { - os << static_cast(v); - return os; -} - -} // end namespace half_impl - -// import Eigen::half_impl::half into Eigen namespace -// using half_impl::half; - -namespace internal { - -template<> -struct random_default_impl -{ - static inline half run(const half& x, const half& y) - { - return x + (y-x) * half(float(std::rand()) / float(RAND_MAX)); - } - static inline half run() - { - return run(half(-1.f), half(1.f)); - } -}; - -template<> struct is_arithmetic { enum { value = true }; }; - -} // end namespace internal - -template<> struct NumTraits - : GenericNumTraits -{ - enum { - IsSigned = true, - IsInteger = false, - IsComplex = false, - RequireInitialization = false - }; - - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half epsilon() { - return half_impl::raw_uint16_to_half(0x0800); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half dummy_precision() { return Eigen::half(1e-2f); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half highest() { - return half_impl::raw_uint16_to_half(0x7bff); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half lowest() { - return half_impl::raw_uint16_to_half(0xfbff); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half infinity() { - return half_impl::raw_uint16_to_half(0x7c00); - } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Eigen::half quiet_NaN() { - return half_impl::raw_uint16_to_half(0x7c01); - } -}; - -} // end namespace Eigen - -// C-like standard mathematical functions and trancendentals. -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half fabsh(const Eigen::half& a) { - Eigen::half result; - result.x = a.x & 0x7FFF; - return result; -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half exph(const Eigen::half& a) { - return Eigen::half(::expf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half logh(const Eigen::half& a) { -#if EIGEN_CUDACC_VER >= 80000 && defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 530 - return Eigen::half(::hlog(a)); -#else - return Eigen::half(::logf(float(a))); -#endif -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half sqrth(const Eigen::half& a) { - return Eigen::half(::sqrtf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half powh(const Eigen::half& a, const Eigen::half& b) { - return Eigen::half(::powf(float(a), float(b))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half floorh(const Eigen::half& a) { - return Eigen::half(::floorf(float(a))); -} -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half ceilh(const Eigen::half& a) { - return Eigen::half(::ceilf(float(a))); -} - -namespace std { - -#if __cplusplus > 199711L -template <> -struct hash { - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::size_t operator()(const Eigen::half& a) const { - return static_cast(a.x); - } -}; -#endif - -} // end namespace std - - -// Add the missing shfl_xor intrinsic -#if defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 300 -__device__ EIGEN_STRONG_INLINE Eigen::half __shfl_xor(Eigen::half var, int laneMask, int width=warpSize) { - #if EIGEN_CUDACC_VER < 90000 - return static_cast(__shfl_xor(static_cast(var), laneMask, width)); - #else - return static_cast(__shfl_xor_sync(0xFFFFFFFF, static_cast(var), laneMask, width)); - #endif -} -#endif - -// ldg() has an overload for __half_raw, but we also need one for Eigen::half. -#if defined(EIGEN_CUDA_ARCH) && EIGEN_CUDA_ARCH >= 350 -EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC Eigen::half __ldg(const Eigen::half* ptr) { - return Eigen::half_impl::raw_uint16_to_half( - __ldg(reinterpret_cast(ptr))); -} -#endif - - -#if defined(EIGEN_CUDA_ARCH) -namespace Eigen { -namespace numext { - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isnan)(const Eigen::half& h) { - return (half_impl::isnan)(h); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isinf)(const Eigen::half& h) { - return (half_impl::isinf)(h); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -bool (isfinite)(const Eigen::half& h) { - return (half_impl::isfinite)(h); -} - -} // namespace Eigen -} // namespace numext -#endif - -#endif // EIGEN_HALF_CUDA_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h deleted file mode 100644 index f749c573ff6..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +++ /dev/null @@ -1,1124 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_HALF_CUDA_H -#define EIGEN_PACKET_MATH_HALF_CUDA_H - - -namespace Eigen { -namespace internal { - -// Most of the following operations require arch >= 3.0 -#if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDACC__) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - -template<> struct is_arithmetic { enum { value = true }; }; - -template<> struct packet_traits : default_packet_traits -{ - typedef half2 type; - typedef half2 half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 0, - HasAdd = 1, - HasMul = 1, - HasDiv = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasExp = 1, - HasLog = 1, - HasLog1p = 1 - }; -}; - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=2, alignment=Aligned16}; typedef half2 half; }; - -template<> __device__ EIGEN_STRONG_INLINE half2 pset1(const Eigen::half& from) { - return __half2half2(from); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pload(const Eigen::half* from) { - return *reinterpret_cast(from); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 ploadu(const Eigen::half* from) { - return __halves2half2(from[0], from[1]); -} - -template<> EIGEN_STRONG_INLINE half2 ploaddup(const Eigen::half* from) { - return __halves2half2(from[0], from[0]); -} - -template<> __device__ EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const half2& from) { - *reinterpret_cast(to) = from; -} - -template<> __device__ EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const half2& from) { - to[0] = __low2half(from); - to[1] = __high2half(from); -} - -template<> - __device__ EIGEN_ALWAYS_INLINE half2 ploadt_ro(const Eigen::half* from) { -#if __CUDA_ARCH__ >= 350 - return __ldg((const half2*)from); -#else - return __halves2half2(*(from+0), *(from+1)); -#endif -} - -template<> -__device__ EIGEN_ALWAYS_INLINE half2 ploadt_ro(const Eigen::half* from) { -#if __CUDA_ARCH__ >= 350 - return __halves2half2(__ldg(from+0), __ldg(from+1)); -#else - return __halves2half2(*(from+0), *(from+1)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pgather(const Eigen::half* from, Index stride) { - return __halves2half2(from[0*stride], from[1*stride]); -} - -template<> __device__ EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const half2& from, Index stride) { - to[stride*0] = __low2half(from); - to[stride*1] = __high2half(from); -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half pfirst(const half2& a) { - return __low2half(a); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pabs(const half2& a) { - half2 result; - unsigned temp = *(reinterpret_cast(&(a))); - *(reinterpret_cast(&(result))) = temp & 0x7FFF7FFF; - return result; -} - - -__device__ EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __half a1 = __low2half(kernel.packet[0]); - __half a2 = __high2half(kernel.packet[0]); - __half b1 = __low2half(kernel.packet[1]); - __half b2 = __high2half(kernel.packet[1]); - kernel.packet[0] = __halves2half2(a1, b1); - kernel.packet[1] = __halves2half2(a2, b2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 plset(const Eigen::half& a) { -#if __CUDA_ARCH__ >= 530 - return __halves2half2(a, __hadd(a, __float2half(1.0f))); -#else - float f = __half2float(a) + 1.0f; - return __halves2half2(a, __float2half(f)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 padd(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hadd2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 + b1; - float r2 = a2 + b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 psub(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hsub2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 - b1; - float r2 = a2 - b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pnegate(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hneg2(a); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return __floats2half2_rn(-a1, -a2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pconj(const half2& a) { return a; } - -template<> __device__ EIGEN_STRONG_INLINE half2 pmul(const half2& a, const half2& b) { -#if __CUDA_ARCH__ >= 530 - return __hmul2(a, b); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 * b1; - float r2 = a2 * b2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmadd(const half2& a, const half2& b, const half2& c) { -#if __CUDA_ARCH__ >= 530 - return __hfma2(a, b, c); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float c1 = __low2float(c); - float c2 = __high2float(c); - float r1 = a1 * b1 + c1; - float r2 = a2 * b2 + c2; - return __floats2half2_rn(r1, r2); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pdiv(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - float r1 = a1 / b1; - float r2 = a2 / b2; - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmin(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - __half r1 = a1 < b1 ? __low2half(a) : __low2half(b); - __half r2 = a2 < b2 ? __high2half(a) : __high2half(b); - return __halves2half2(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pmax(const half2& a, const half2& b) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float b1 = __low2float(b); - float b2 = __high2float(b); - __half r1 = a1 > b1 ? __low2half(a) : __low2half(b); - __half r2 = a2 > b2 ? __high2half(a) : __high2half(b); - return __halves2half2(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hadd(__low2half(a), __high2half(a)); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 + a2)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_max(const half2& a) { -#if __CUDA_ARCH__ >= 530 - __half first = __low2half(a); - __half second = __high2half(a); - return __hgt(first, second) ? first : second; -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return a1 > a2 ? __low2half(a) : __high2half(a); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_min(const half2& a) { -#if __CUDA_ARCH__ >= 530 - __half first = __low2half(a); - __half second = __high2half(a); - return __hlt(first, second) ? first : second; -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return a1 < a2 ? __low2half(a) : __high2half(a); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_mul(const half2& a) { -#if __CUDA_ARCH__ >= 530 - return __hmul(__low2half(a), __high2half(a)); -#else - float a1 = __low2float(a); - float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 * a2)); -#endif -} - -template<> __device__ EIGEN_STRONG_INLINE half2 plog1p(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = log1pf(a1); - float r2 = log1pf(a2); - return __floats2half2_rn(r1, r2); -} - -#if EIGEN_CUDACC_VER >= 80000 && defined EIGEN_CUDA_ARCH && EIGEN_CUDA_ARCH >= 530 - -template<> __device__ EIGEN_STRONG_INLINE -half2 plog(const half2& a) { - return h2log(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 pexp(const half2& a) { - return h2exp(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 psqrt(const half2& a) { - return h2sqrt(a); -} - -template<> __device__ EIGEN_STRONG_INLINE -half2 prsqrt(const half2& a) { - return h2rsqrt(a); -} - -#else - -template<> __device__ EIGEN_STRONG_INLINE half2 plog(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = logf(a1); - float r2 = logf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 pexp(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = expf(a1); - float r2 = expf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 psqrt(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = sqrtf(a1); - float r2 = sqrtf(a2); - return __floats2half2_rn(r1, r2); -} - -template<> __device__ EIGEN_STRONG_INLINE half2 prsqrt(const half2& a) { - float a1 = __low2float(a); - float a2 = __high2float(a); - float r1 = rsqrtf(a1); - float r2 = rsqrtf(a2); - return __floats2half2_rn(r1, r2); -} - -#endif - -#elif defined EIGEN_VECTORIZE_AVX512 - -typedef struct { - __m256i x; -} Packet16h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet16h type; - // There is no half-size packet for Packet16h. - typedef Packet16h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 16, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=16, alignment=Aligned32}; typedef Packet16h half; }; - -template<> EIGEN_STRONG_INLINE Packet16h pset1(const Eigen::half& from) { - Packet16h result; - result.x = _mm256_set1_epi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet16h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm256_extract_epi16(from.x, 0))); -} - -template<> EIGEN_STRONG_INLINE Packet16h pload(const Eigen::half* from) { - Packet16h result; - result.x = _mm256_load_si256(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet16h ploadu(const Eigen::half* from) { - Packet16h result; - result.x = _mm256_loadu_si256(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet16h& from) { - _mm256_store_si256((__m256i*)to, from.x); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet16h& from) { - _mm256_storeu_si256((__m256i*)to, from.x); -} - -template<> EIGEN_STRONG_INLINE Packet16h -ploadquad(const Eigen::half* from) { - Packet16h result; - unsigned short a = from[0].x; - unsigned short b = from[1].x; - unsigned short c = from[2].x; - unsigned short d = from[3].x; - result.x = _mm256_set_epi16(d, d, d, d, c, c, c, c, b, b, b, b, a, a, a, a); - return result; -} - -EIGEN_STRONG_INLINE Packet16f half2float(const Packet16h& a) { -#ifdef EIGEN_HAS_FP16_C - return _mm512_cvtph_ps(a.x); -#else - EIGEN_ALIGN64 half aux[16]; - pstore(aux, a); - float f0(aux[0]); - float f1(aux[1]); - float f2(aux[2]); - float f3(aux[3]); - float f4(aux[4]); - float f5(aux[5]); - float f6(aux[6]); - float f7(aux[7]); - float f8(aux[8]); - float f9(aux[9]); - float fa(aux[10]); - float fb(aux[11]); - float fc(aux[12]); - float fd(aux[13]); - float fe(aux[14]); - float ff(aux[15]); - - return _mm512_set_ps( - ff, fe, fd, fc, fb, fa, f9, f8, f7, f6, f5, f4, f3, f2, f1, f0); -#endif -} - -EIGEN_STRONG_INLINE Packet16h float2half(const Packet16f& a) { -#ifdef EIGEN_HAS_FP16_C - Packet16h result; - result.x = _mm512_cvtps_ph(a, _MM_FROUND_TO_NEAREST_INT|_MM_FROUND_NO_EXC); - return result; -#else - EIGEN_ALIGN64 float aux[16]; - pstore(aux, a); - half h0(aux[0]); - half h1(aux[1]); - half h2(aux[2]); - half h3(aux[3]); - half h4(aux[4]); - half h5(aux[5]); - half h6(aux[6]); - half h7(aux[7]); - half h8(aux[8]); - half h9(aux[9]); - half ha(aux[10]); - half hb(aux[11]); - half hc(aux[12]); - half hd(aux[13]); - half he(aux[14]); - half hf(aux[15]); - - Packet16h result; - result.x = _mm256_set_epi16( - hf.x, he.x, hd.x, hc.x, hb.x, ha.x, h9.x, h8.x, - h7.x, h6.x, h5.x, h4.x, h3.x, h2.x, h1.x, h0.x); - return result; -#endif -} - -template<> EIGEN_STRONG_INLINE Packet16h padd(const Packet16h& a, const Packet16h& b) { - Packet16f af = half2float(a); - Packet16f bf = half2float(b); - Packet16f rf = padd(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet16h pmul(const Packet16h& a, const Packet16h& b) { - Packet16f af = half2float(a); - Packet16f bf = half2float(b); - Packet16f rf = pmul(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE half predux(const Packet16h& from) { - Packet16f from_float = half2float(from); - return half(predux(from_float)); -} - -template<> EIGEN_STRONG_INLINE Packet16h pgather(const Eigen::half* from, Index stride) -{ - Packet16h result; - result.x = _mm256_set_epi16( - from[15*stride].x, from[14*stride].x, from[13*stride].x, from[12*stride].x, - from[11*stride].x, from[10*stride].x, from[9*stride].x, from[8*stride].x, - from[7*stride].x, from[6*stride].x, from[5*stride].x, from[4*stride].x, - from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(half* to, const Packet16h& from, Index stride) -{ - EIGEN_ALIGN64 half aux[16]; - pstore(aux, from); - to[stride*0].x = aux[0].x; - to[stride*1].x = aux[1].x; - to[stride*2].x = aux[2].x; - to[stride*3].x = aux[3].x; - to[stride*4].x = aux[4].x; - to[stride*5].x = aux[5].x; - to[stride*6].x = aux[6].x; - to[stride*7].x = aux[7].x; - to[stride*8].x = aux[8].x; - to[stride*9].x = aux[9].x; - to[stride*10].x = aux[10].x; - to[stride*11].x = aux[11].x; - to[stride*12].x = aux[12].x; - to[stride*13].x = aux[13].x; - to[stride*14].x = aux[14].x; - to[stride*15].x = aux[15].x; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m256i a = kernel.packet[0].x; - __m256i b = kernel.packet[1].x; - __m256i c = kernel.packet[2].x; - __m256i d = kernel.packet[3].x; - __m256i e = kernel.packet[4].x; - __m256i f = kernel.packet[5].x; - __m256i g = kernel.packet[6].x; - __m256i h = kernel.packet[7].x; - __m256i i = kernel.packet[8].x; - __m256i j = kernel.packet[9].x; - __m256i k = kernel.packet[10].x; - __m256i l = kernel.packet[11].x; - __m256i m = kernel.packet[12].x; - __m256i n = kernel.packet[13].x; - __m256i o = kernel.packet[14].x; - __m256i p = kernel.packet[15].x; - - __m256i ab_07 = _mm256_unpacklo_epi16(a, b); - __m256i cd_07 = _mm256_unpacklo_epi16(c, d); - __m256i ef_07 = _mm256_unpacklo_epi16(e, f); - __m256i gh_07 = _mm256_unpacklo_epi16(g, h); - __m256i ij_07 = _mm256_unpacklo_epi16(i, j); - __m256i kl_07 = _mm256_unpacklo_epi16(k, l); - __m256i mn_07 = _mm256_unpacklo_epi16(m, n); - __m256i op_07 = _mm256_unpacklo_epi16(o, p); - - __m256i ab_8f = _mm256_unpackhi_epi16(a, b); - __m256i cd_8f = _mm256_unpackhi_epi16(c, d); - __m256i ef_8f = _mm256_unpackhi_epi16(e, f); - __m256i gh_8f = _mm256_unpackhi_epi16(g, h); - __m256i ij_8f = _mm256_unpackhi_epi16(i, j); - __m256i kl_8f = _mm256_unpackhi_epi16(k, l); - __m256i mn_8f = _mm256_unpackhi_epi16(m, n); - __m256i op_8f = _mm256_unpackhi_epi16(o, p); - - __m256i abcd_03 = _mm256_unpacklo_epi32(ab_07, cd_07); - __m256i abcd_47 = _mm256_unpackhi_epi32(ab_07, cd_07); - __m256i efgh_03 = _mm256_unpacklo_epi32(ef_07, gh_07); - __m256i efgh_47 = _mm256_unpackhi_epi32(ef_07, gh_07); - __m256i ijkl_03 = _mm256_unpacklo_epi32(ij_07, kl_07); - __m256i ijkl_47 = _mm256_unpackhi_epi32(ij_07, kl_07); - __m256i mnop_03 = _mm256_unpacklo_epi32(mn_07, op_07); - __m256i mnop_47 = _mm256_unpackhi_epi32(mn_07, op_07); - - __m256i abcd_8b = _mm256_unpacklo_epi32(ab_8f, cd_8f); - __m256i abcd_cf = _mm256_unpackhi_epi32(ab_8f, cd_8f); - __m256i efgh_8b = _mm256_unpacklo_epi32(ef_8f, gh_8f); - __m256i efgh_cf = _mm256_unpackhi_epi32(ef_8f, gh_8f); - __m256i ijkl_8b = _mm256_unpacklo_epi32(ij_8f, kl_8f); - __m256i ijkl_cf = _mm256_unpackhi_epi32(ij_8f, kl_8f); - __m256i mnop_8b = _mm256_unpacklo_epi32(mn_8f, op_8f); - __m256i mnop_cf = _mm256_unpackhi_epi32(mn_8f, op_8f); - - __m256i abcdefgh_01 = _mm256_unpacklo_epi64(abcd_03, efgh_03); - __m256i abcdefgh_23 = _mm256_unpackhi_epi64(abcd_03, efgh_03); - __m256i ijklmnop_01 = _mm256_unpacklo_epi64(ijkl_03, mnop_03); - __m256i ijklmnop_23 = _mm256_unpackhi_epi64(ijkl_03, mnop_03); - __m256i abcdefgh_45 = _mm256_unpacklo_epi64(abcd_47, efgh_47); - __m256i abcdefgh_67 = _mm256_unpackhi_epi64(abcd_47, efgh_47); - __m256i ijklmnop_45 = _mm256_unpacklo_epi64(ijkl_47, mnop_47); - __m256i ijklmnop_67 = _mm256_unpackhi_epi64(ijkl_47, mnop_47); - __m256i abcdefgh_89 = _mm256_unpacklo_epi64(abcd_8b, efgh_8b); - __m256i abcdefgh_ab = _mm256_unpackhi_epi64(abcd_8b, efgh_8b); - __m256i ijklmnop_89 = _mm256_unpacklo_epi64(ijkl_8b, mnop_8b); - __m256i ijklmnop_ab = _mm256_unpackhi_epi64(ijkl_8b, mnop_8b); - __m256i abcdefgh_cd = _mm256_unpacklo_epi64(abcd_cf, efgh_cf); - __m256i abcdefgh_ef = _mm256_unpackhi_epi64(abcd_cf, efgh_cf); - __m256i ijklmnop_cd = _mm256_unpacklo_epi64(ijkl_cf, mnop_cf); - __m256i ijklmnop_ef = _mm256_unpackhi_epi64(ijkl_cf, mnop_cf); - - // NOTE: no unpacklo/hi instr in this case, so using permute instr. - __m256i a_p_0 = _mm256_permute2x128_si256(abcdefgh_01, ijklmnop_01, 0x20); - __m256i a_p_1 = _mm256_permute2x128_si256(abcdefgh_01, ijklmnop_01, 0x31); - __m256i a_p_2 = _mm256_permute2x128_si256(abcdefgh_23, ijklmnop_23, 0x20); - __m256i a_p_3 = _mm256_permute2x128_si256(abcdefgh_23, ijklmnop_23, 0x31); - __m256i a_p_4 = _mm256_permute2x128_si256(abcdefgh_45, ijklmnop_45, 0x20); - __m256i a_p_5 = _mm256_permute2x128_si256(abcdefgh_45, ijklmnop_45, 0x31); - __m256i a_p_6 = _mm256_permute2x128_si256(abcdefgh_67, ijklmnop_67, 0x20); - __m256i a_p_7 = _mm256_permute2x128_si256(abcdefgh_67, ijklmnop_67, 0x31); - __m256i a_p_8 = _mm256_permute2x128_si256(abcdefgh_89, ijklmnop_89, 0x20); - __m256i a_p_9 = _mm256_permute2x128_si256(abcdefgh_89, ijklmnop_89, 0x31); - __m256i a_p_a = _mm256_permute2x128_si256(abcdefgh_ab, ijklmnop_ab, 0x20); - __m256i a_p_b = _mm256_permute2x128_si256(abcdefgh_ab, ijklmnop_ab, 0x31); - __m256i a_p_c = _mm256_permute2x128_si256(abcdefgh_cd, ijklmnop_cd, 0x20); - __m256i a_p_d = _mm256_permute2x128_si256(abcdefgh_cd, ijklmnop_cd, 0x31); - __m256i a_p_e = _mm256_permute2x128_si256(abcdefgh_ef, ijklmnop_ef, 0x20); - __m256i a_p_f = _mm256_permute2x128_si256(abcdefgh_ef, ijklmnop_ef, 0x31); - - kernel.packet[0].x = a_p_0; - kernel.packet[1].x = a_p_1; - kernel.packet[2].x = a_p_2; - kernel.packet[3].x = a_p_3; - kernel.packet[4].x = a_p_4; - kernel.packet[5].x = a_p_5; - kernel.packet[6].x = a_p_6; - kernel.packet[7].x = a_p_7; - kernel.packet[8].x = a_p_8; - kernel.packet[9].x = a_p_9; - kernel.packet[10].x = a_p_a; - kernel.packet[11].x = a_p_b; - kernel.packet[12].x = a_p_c; - kernel.packet[13].x = a_p_d; - kernel.packet[14].x = a_p_e; - kernel.packet[15].x = a_p_f; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN64 half in[8][16]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - pstore(in[4], kernel.packet[4]); - pstore(in[5], kernel.packet[5]); - pstore(in[6], kernel.packet[6]); - pstore(in[7], kernel.packet[7]); - - EIGEN_ALIGN64 half out[8][16]; - - for (int i = 0; i < 8; ++i) { - for (int j = 0; j < 8; ++j) { - out[i][j] = in[j][2*i]; - } - for (int j = 0; j < 8; ++j) { - out[i][j+8] = in[j][2*i+1]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); - kernel.packet[4] = pload(out[4]); - kernel.packet[5] = pload(out[5]); - kernel.packet[6] = pload(out[6]); - kernel.packet[7] = pload(out[7]); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN64 half in[4][16]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - - EIGEN_ALIGN64 half out[4][16]; - - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 4; ++j) { - out[i][j] = in[j][4*i]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+4] = in[j][4*i+1]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+8] = in[j][4*i+2]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+12] = in[j][4*i+3]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); -} - - -#elif defined EIGEN_VECTORIZE_AVX - -typedef struct { - __m128i x; -} Packet8h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet8h type; - // There is no half-size packet for Packet8h. - typedef Packet8h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 8, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=8, alignment=Aligned16}; typedef Packet8h half; }; - -template<> EIGEN_STRONG_INLINE Packet8h pset1(const Eigen::half& from) { - Packet8h result; - result.x = _mm_set1_epi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet8h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm_extract_epi16(from.x, 0))); -} - -template<> EIGEN_STRONG_INLINE Packet8h pload(const Eigen::half* from) { - Packet8h result; - result.x = _mm_load_si128(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet8h ploadu(const Eigen::half* from) { - Packet8h result; - result.x = _mm_loadu_si128(reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet8h& from) { - _mm_store_si128(reinterpret_cast<__m128i*>(to), from.x); -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet8h& from) { - _mm_storeu_si128(reinterpret_cast<__m128i*>(to), from.x); -} - -template<> EIGEN_STRONG_INLINE Packet8h -ploadquad(const Eigen::half* from) { - Packet8h result; - unsigned short a = from[0].x; - unsigned short b = from[1].x; - result.x = _mm_set_epi16(b, b, b, b, a, a, a, a); - return result; -} - -EIGEN_STRONG_INLINE Packet8f half2float(const Packet8h& a) { -#ifdef EIGEN_HAS_FP16_C - return _mm256_cvtph_ps(a.x); -#else - EIGEN_ALIGN32 Eigen::half aux[8]; - pstore(aux, a); - float f0(aux[0]); - float f1(aux[1]); - float f2(aux[2]); - float f3(aux[3]); - float f4(aux[4]); - float f5(aux[5]); - float f6(aux[6]); - float f7(aux[7]); - - return _mm256_set_ps(f7, f6, f5, f4, f3, f2, f1, f0); -#endif -} - -EIGEN_STRONG_INLINE Packet8h float2half(const Packet8f& a) { -#ifdef EIGEN_HAS_FP16_C - Packet8h result; - result.x = _mm256_cvtps_ph(a, _MM_FROUND_TO_NEAREST_INT|_MM_FROUND_NO_EXC); - return result; -#else - EIGEN_ALIGN32 float aux[8]; - pstore(aux, a); - Eigen::half h0(aux[0]); - Eigen::half h1(aux[1]); - Eigen::half h2(aux[2]); - Eigen::half h3(aux[3]); - Eigen::half h4(aux[4]); - Eigen::half h5(aux[5]); - Eigen::half h6(aux[6]); - Eigen::half h7(aux[7]); - - Packet8h result; - result.x = _mm_set_epi16(h7.x, h6.x, h5.x, h4.x, h3.x, h2.x, h1.x, h0.x); - return result; -#endif -} - -template<> EIGEN_STRONG_INLINE Packet8h pconj(const Packet8h& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet8h padd(const Packet8h& a, const Packet8h& b) { - Packet8f af = half2float(a); - Packet8f bf = half2float(b); - Packet8f rf = padd(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet8h pmul(const Packet8h& a, const Packet8h& b) { - Packet8f af = half2float(a); - Packet8f bf = half2float(b); - Packet8f rf = pmul(af, bf); - return float2half(rf); -} - -template<> EIGEN_STRONG_INLINE Packet8h pgather(const Eigen::half* from, Index stride) -{ - Packet8h result; - result.x = _mm_set_epi16(from[7*stride].x, from[6*stride].x, from[5*stride].x, from[4*stride].x, from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const Packet8h& from, Index stride) -{ - EIGEN_ALIGN32 Eigen::half aux[8]; - pstore(aux, from); - to[stride*0].x = aux[0].x; - to[stride*1].x = aux[1].x; - to[stride*2].x = aux[2].x; - to[stride*3].x = aux[3].x; - to[stride*4].x = aux[4].x; - to[stride*5].x = aux[5].x; - to[stride*6].x = aux[6].x; - to[stride*7].x = aux[7].x; -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_max(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_max(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_min(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_min(af); - return Eigen::half(reduced); -} - -template<> EIGEN_STRONG_INLINE Eigen::half predux_mul(const Packet8h& a) { - Packet8f af = half2float(a); - float reduced = predux_mul(af); - return Eigen::half(reduced); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m128i a = kernel.packet[0].x; - __m128i b = kernel.packet[1].x; - __m128i c = kernel.packet[2].x; - __m128i d = kernel.packet[3].x; - __m128i e = kernel.packet[4].x; - __m128i f = kernel.packet[5].x; - __m128i g = kernel.packet[6].x; - __m128i h = kernel.packet[7].x; - - __m128i a03b03 = _mm_unpacklo_epi16(a, b); - __m128i c03d03 = _mm_unpacklo_epi16(c, d); - __m128i e03f03 = _mm_unpacklo_epi16(e, f); - __m128i g03h03 = _mm_unpacklo_epi16(g, h); - __m128i a47b47 = _mm_unpackhi_epi16(a, b); - __m128i c47d47 = _mm_unpackhi_epi16(c, d); - __m128i e47f47 = _mm_unpackhi_epi16(e, f); - __m128i g47h47 = _mm_unpackhi_epi16(g, h); - - __m128i a01b01c01d01 = _mm_unpacklo_epi32(a03b03, c03d03); - __m128i a23b23c23d23 = _mm_unpackhi_epi32(a03b03, c03d03); - __m128i e01f01g01h01 = _mm_unpacklo_epi32(e03f03, g03h03); - __m128i e23f23g23h23 = _mm_unpackhi_epi32(e03f03, g03h03); - __m128i a45b45c45d45 = _mm_unpacklo_epi32(a47b47, c47d47); - __m128i a67b67c67d67 = _mm_unpackhi_epi32(a47b47, c47d47); - __m128i e45f45g45h45 = _mm_unpacklo_epi32(e47f47, g47h47); - __m128i e67f67g67h67 = _mm_unpackhi_epi32(e47f47, g47h47); - - __m128i a0b0c0d0e0f0g0h0 = _mm_unpacklo_epi64(a01b01c01d01, e01f01g01h01); - __m128i a1b1c1d1e1f1g1h1 = _mm_unpackhi_epi64(a01b01c01d01, e01f01g01h01); - __m128i a2b2c2d2e2f2g2h2 = _mm_unpacklo_epi64(a23b23c23d23, e23f23g23h23); - __m128i a3b3c3d3e3f3g3h3 = _mm_unpackhi_epi64(a23b23c23d23, e23f23g23h23); - __m128i a4b4c4d4e4f4g4h4 = _mm_unpacklo_epi64(a45b45c45d45, e45f45g45h45); - __m128i a5b5c5d5e5f5g5h5 = _mm_unpackhi_epi64(a45b45c45d45, e45f45g45h45); - __m128i a6b6c6d6e6f6g6h6 = _mm_unpacklo_epi64(a67b67c67d67, e67f67g67h67); - __m128i a7b7c7d7e7f7g7h7 = _mm_unpackhi_epi64(a67b67c67d67, e67f67g67h67); - - kernel.packet[0].x = a0b0c0d0e0f0g0h0; - kernel.packet[1].x = a1b1c1d1e1f1g1h1; - kernel.packet[2].x = a2b2c2d2e2f2g2h2; - kernel.packet[3].x = a3b3c3d3e3f3g3h3; - kernel.packet[4].x = a4b4c4d4e4f4g4h4; - kernel.packet[5].x = a5b5c5d5e5f5g5h5; - kernel.packet[6].x = a6b6c6d6e6f6g6h6; - kernel.packet[7].x = a7b7c7d7e7f7g7h7; -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - EIGEN_ALIGN32 Eigen::half in[4][8]; - pstore(in[0], kernel.packet[0]); - pstore(in[1], kernel.packet[1]); - pstore(in[2], kernel.packet[2]); - pstore(in[3], kernel.packet[3]); - - EIGEN_ALIGN32 Eigen::half out[4][8]; - - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 4; ++j) { - out[i][j] = in[j][2*i]; - } - for (int j = 0; j < 4; ++j) { - out[i][j+4] = in[j][2*i+1]; - } - } - - kernel.packet[0] = pload(out[0]); - kernel.packet[1] = pload(out[1]); - kernel.packet[2] = pload(out[2]); - kernel.packet[3] = pload(out[3]); -} - - -// Disable the following code since it's broken on too many platforms / compilers. -//#elif defined(EIGEN_VECTORIZE_SSE) && (!EIGEN_ARCH_x86_64) && (!EIGEN_COMP_MSVC) -#elif 0 - -typedef struct { - __m64 x; -} Packet4h; - - -template<> struct is_arithmetic { enum { value = true }; }; - -template <> -struct packet_traits : default_packet_traits { - typedef Packet4h type; - // There is no half-size packet for Packet4h. - typedef Packet4h half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 4, - HasHalfPacket = 0, - HasAdd = 0, - HasSub = 0, - HasMul = 0, - HasNegate = 0, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasConj = 0, - HasSetLinear = 0, - HasDiv = 0, - HasSqrt = 0, - HasRsqrt = 0, - HasExp = 0, - HasLog = 0, - HasBlend = 0 - }; -}; - - -template<> struct unpacket_traits { typedef Eigen::half type; enum {size=4, alignment=Aligned16}; typedef Packet4h half; }; - -template<> EIGEN_STRONG_INLINE Packet4h pset1(const Eigen::half& from) { - Packet4h result; - result.x = _mm_set1_pi16(from.x); - return result; -} - -template<> EIGEN_STRONG_INLINE Eigen::half pfirst(const Packet4h& from) { - return half_impl::raw_uint16_to_half(static_cast(_mm_cvtsi64_si32(from.x))); -} - -template<> EIGEN_STRONG_INLINE Packet4h pconj(const Packet4h& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4h padd(const Packet4h& a, const Packet4h& b) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - __int64_t b64 = _mm_cvtm64_si64(b.x); - - Eigen::half h[4]; - - Eigen::half ha = half_impl::raw_uint16_to_half(static_cast(a64)); - Eigen::half hb = half_impl::raw_uint16_to_half(static_cast(b64)); - h[0] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 16)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 16)); - h[1] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 32)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 32)); - h[2] = ha + hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 48)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 48)); - h[3] = ha + hb; - Packet4h result; - result.x = _mm_set_pi16(h[3].x, h[2].x, h[1].x, h[0].x); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h pmul(const Packet4h& a, const Packet4h& b) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - __int64_t b64 = _mm_cvtm64_si64(b.x); - - Eigen::half h[4]; - - Eigen::half ha = half_impl::raw_uint16_to_half(static_cast(a64)); - Eigen::half hb = half_impl::raw_uint16_to_half(static_cast(b64)); - h[0] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 16)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 16)); - h[1] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 32)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 32)); - h[2] = ha * hb; - ha = half_impl::raw_uint16_to_half(static_cast(a64 >> 48)); - hb = half_impl::raw_uint16_to_half(static_cast(b64 >> 48)); - h[3] = ha * hb; - Packet4h result; - result.x = _mm_set_pi16(h[3].x, h[2].x, h[1].x, h[0].x); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h pload(const Eigen::half* from) { - Packet4h result; - result.x = _mm_cvtsi64_m64(*reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE Packet4h ploadu(const Eigen::half* from) { - Packet4h result; - result.x = _mm_cvtsi64_m64(*reinterpret_cast(from)); - return result; -} - -template<> EIGEN_STRONG_INLINE void pstore(Eigen::half* to, const Packet4h& from) { - __int64_t r = _mm_cvtm64_si64(from.x); - *(reinterpret_cast<__int64_t*>(to)) = r; -} - -template<> EIGEN_STRONG_INLINE void pstoreu(Eigen::half* to, const Packet4h& from) { - __int64_t r = _mm_cvtm64_si64(from.x); - *(reinterpret_cast<__int64_t*>(to)) = r; -} - -template<> EIGEN_STRONG_INLINE Packet4h -ploadquad(const Eigen::half* from) { - return pset1(*from); -} - -template<> EIGEN_STRONG_INLINE Packet4h pgather(const Eigen::half* from, Index stride) -{ - Packet4h result; - result.x = _mm_set_pi16(from[3*stride].x, from[2*stride].x, from[1*stride].x, from[0*stride].x); - return result; -} - -template<> EIGEN_STRONG_INLINE void pscatter(Eigen::half* to, const Packet4h& from, Index stride) -{ - __int64_t a = _mm_cvtm64_si64(from.x); - to[stride*0].x = static_cast(a); - to[stride*1].x = static_cast(a >> 16); - to[stride*2].x = static_cast(a >> 32); - to[stride*3].x = static_cast(a >> 48); -} - -EIGEN_STRONG_INLINE void -ptranspose(PacketBlock& kernel) { - __m64 T0 = _mm_unpacklo_pi16(kernel.packet[0].x, kernel.packet[1].x); - __m64 T1 = _mm_unpacklo_pi16(kernel.packet[2].x, kernel.packet[3].x); - __m64 T2 = _mm_unpackhi_pi16(kernel.packet[0].x, kernel.packet[1].x); - __m64 T3 = _mm_unpackhi_pi16(kernel.packet[2].x, kernel.packet[3].x); - - kernel.packet[0].x = _mm_unpacklo_pi32(T0, T1); - kernel.packet[1].x = _mm_unpackhi_pi32(T0, T1); - kernel.packet[2].x = _mm_unpacklo_pi32(T2, T3); - kernel.packet[3].x = _mm_unpackhi_pi32(T2, T3); -} - -#endif - -} -} - -#endif // EIGEN_PACKET_MATH_HALF_CUDA_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h deleted file mode 100644 index aa5fbce8eac..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/CUDA/TypeCasting.h +++ /dev/null @@ -1,212 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TYPE_CASTING_CUDA_H -#define EIGEN_TYPE_CASTING_CUDA_H - -namespace Eigen { - -namespace internal { - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef Eigen::half result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Eigen::half operator() (const float& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __float2half(a); - #else - return Eigen::half(a); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef Eigen::half result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Eigen::half operator() (const int& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __float2half(static_cast(a)); - #else - return Eigen::half(static_cast(a)); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - -template<> -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef float result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float operator() (const Eigen::half& a) const { - #if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - return __half2float(a); - #else - return static_cast(a); - #endif - } -}; - -template<> -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - - - -#if defined(EIGEN_HAS_CUDA_FP16) && defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 300 - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 2, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pcast(const half2& a, const half2& b) { - float2 r1 = __half22float2(a); - float2 r2 = __half22float2(b); - return make_float4(r1.x, r1.y, r2.x, r2.y); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 2 - }; -}; - -template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE half2 pcast(const float4& a) { - // Simply discard the second half of the input - return __floats2half2_rn(a.x, a.y); -} - -#elif defined EIGEN_VECTORIZE_AVX512 -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet16f pcast(const Packet16h& a) { - return half2float(a); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet16h pcast(const Packet16f& a) { - return float2half(a); -} - -#elif defined EIGEN_VECTORIZE_AVX - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet8f pcast(const Packet8h& a) { - return half2float(a); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet8h pcast(const Packet8f& a) { - return float2half(a); -} - -// Disable the following code since it's broken on too many platforms / compilers. -//#elif defined(EIGEN_VECTORIZE_SSE) && (!EIGEN_ARCH_x86_64) && (!EIGEN_COMP_MSVC) -#elif 0 - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet4h& a) { - __int64_t a64 = _mm_cvtm64_si64(a.x); - Eigen::half h = raw_uint16_to_half(static_cast(a64)); - float f1 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 16)); - float f2 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 32)); - float f3 = static_cast(h); - h = raw_uint16_to_half(static_cast(a64 >> 48)); - float f4 = static_cast(h); - return _mm_set_ps(f4, f3, f2, f1); -} - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template<> EIGEN_STRONG_INLINE Packet4h pcast(const Packet4f& a) { - EIGEN_ALIGN16 float aux[4]; - pstore(aux, a); - Eigen::half h0(aux[0]); - Eigen::half h1(aux[1]); - Eigen::half h2(aux[2]); - Eigen::half h3(aux[3]); - - Packet4h result; - result.x = _mm_set_pi16(h3.x, h2.x, h1.x, h0.x); - return result; -} - -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TYPE_CASTING_CUDA_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h deleted file mode 100644 index 4cfe34e0526..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/ConjHelper.h +++ /dev/null @@ -1,29 +0,0 @@ - -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2017 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ARCH_CONJ_HELPER_H -#define EIGEN_ARCH_CONJ_HELPER_H - -#define EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(PACKET_CPLX, PACKET_REAL) \ - template<> struct conj_helper { \ - EIGEN_STRONG_INLINE PACKET_CPLX pmadd(const PACKET_REAL& x, const PACKET_CPLX& y, const PACKET_CPLX& c) const \ - { return padd(c, pmul(x,y)); } \ - EIGEN_STRONG_INLINE PACKET_CPLX pmul(const PACKET_REAL& x, const PACKET_CPLX& y) const \ - { return PACKET_CPLX(Eigen::internal::pmul(x, y.v)); } \ - }; \ - \ - template<> struct conj_helper { \ - EIGEN_STRONG_INLINE PACKET_CPLX pmadd(const PACKET_CPLX& x, const PACKET_REAL& y, const PACKET_CPLX& c) const \ - { return padd(c, pmul(x,y)); } \ - EIGEN_STRONG_INLINE PACKET_CPLX pmul(const PACKET_CPLX& x, const PACKET_REAL& y) const \ - { return PACKET_CPLX(Eigen::internal::pmul(x.v, y)); } \ - }; - -#endif // EIGEN_ARCH_CONJ_HELPER_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h deleted file mode 100644 index 097373c84dc..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/Default/Settings.h +++ /dev/null @@ -1,49 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -/* All the parameters defined in this file can be specialized in the - * architecture specific files, and/or by the user. - * More to come... */ - -#ifndef EIGEN_DEFAULT_SETTINGS_H -#define EIGEN_DEFAULT_SETTINGS_H - -/** Defines the maximal loop size to enable meta unrolling of loops. - * Note that the value here is expressed in Eigen's own notion of "number of FLOPS", - * it does not correspond to the number of iterations or the number of instructions - */ -#ifndef EIGEN_UNROLLING_LIMIT -#define EIGEN_UNROLLING_LIMIT 100 -#endif - -/** Defines the threshold between a "small" and a "large" matrix. - * This threshold is mainly used to select the proper product implementation. - */ -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -/** Defines the maximal width of the blocks used in the triangular product and solver - * for vectors (level 2 blas xTRMV and xTRSV). The default is 8. - */ -#ifndef EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH -#define EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH 8 -#endif - - -/** Defines the default number of registers available for that architecture. - * Currently it must be 8 or 16. Other values will fail. - */ -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 8 -#endif - -#endif // EIGEN_DEFAULT_SETTINGS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h deleted file mode 100644 index d075043ce1b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/Complex.h +++ /dev/null @@ -1,471 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_COMPLEX_SSE_H -#define EIGEN_COMPLEX_SSE_H - -namespace Eigen { - -namespace internal { - -//---------- float ---------- -struct Packet2cf -{ - EIGEN_STRONG_INLINE Packet2cf() {} - EIGEN_STRONG_INLINE explicit Packet2cf(const __m128& a) : v(a) {} - __m128 v; -}; - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet2cf type; - typedef Packet2cf half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size = 2, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0, - HasBlend = 1 - }; -}; -#endif - -template<> struct unpacket_traits { typedef std::complex type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; - -template<> EIGEN_STRONG_INLINE Packet2cf padd(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf psub(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a) -{ - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); - return Packet2cf(_mm_xor_ps(a.v,mask)); -} -template<> EIGEN_STRONG_INLINE Packet2cf pconj(const Packet2cf& a) -{ - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_xor_ps(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) -{ - #ifdef EIGEN_VECTORIZE_SSE3 - return Packet2cf(_mm_addsub_ps(_mm_mul_ps(_mm_moveldup_ps(a.v), b.v), - _mm_mul_ps(_mm_movehdup_ps(a.v), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); -// return Packet2cf(_mm_addsub_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), -// _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), -// vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x00000000,0x80000000,0x00000000)); - return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), - _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); - #endif -} - -template<> EIGEN_STRONG_INLINE Packet2cf pand (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_and_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf por (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_or_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pxor (const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_xor_ps(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet2cf pandnot(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_andnot_ps(a.v,b.v)); } - -template<> EIGEN_STRONG_INLINE Packet2cf pload (const std::complex* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet2cf(pload(&numext::real_ref(*from))); } -template<> EIGEN_STRONG_INLINE Packet2cf ploadu(const std::complex* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet2cf(ploadu(&numext::real_ref(*from))); } - -template<> EIGEN_STRONG_INLINE Packet2cf pset1(const std::complex& from) -{ - Packet2cf res; -#if EIGEN_GNUC_AT_MOST(4,2) - // Workaround annoying "may be used uninitialized in this function" warning with gcc 4.2 - res.v = _mm_loadl_pi(_mm_set1_ps(0.0f), reinterpret_cast(&from)); -#elif EIGEN_GNUC_AT_LEAST(4,6) - // Suppress annoying "may be used uninitialized in this function" warning with gcc >= 4.6 - #pragma GCC diagnostic push - #pragma GCC diagnostic ignored "-Wuninitialized" - res.v = _mm_loadl_pi(res.v, (const __m64*)&from); - #pragma GCC diagnostic pop -#else - res.v = _mm_loadl_pi(res.v, (const __m64*)&from); -#endif - return Packet2cf(_mm_movelh_ps(res.v,res.v)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf ploaddup(const std::complex* from) { return pset1(*from); } - -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_ALIGNED_STORE pstore(&numext::real_ref(*to), Packet4f(from.v)); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet2cf& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(&numext::real_ref(*to), Packet4f(from.v)); } - - -template<> EIGEN_DEVICE_FUNC inline Packet2cf pgather, Packet2cf>(const std::complex* from, Index stride) -{ - return Packet2cf(_mm_set_ps(std::imag(from[1*stride]), std::real(from[1*stride]), - std::imag(from[0*stride]), std::real(from[0*stride]))); -} - -template<> EIGEN_DEVICE_FUNC inline void pscatter, Packet2cf>(std::complex* to, const Packet2cf& from, Index stride) -{ - to[stride*0] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 0)), - _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 1))); - to[stride*1] = std::complex(_mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 2)), - _mm_cvtss_f32(_mm_shuffle_ps(from.v, from.v, 3))); -} - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet2cf& a) -{ - #if EIGEN_GNUC_AT_MOST(4,3) - // Workaround gcc 4.2 ICE - this is not performance wise ideal, but who cares... - // This workaround also fix invalid code generation with gcc 4.3 - EIGEN_ALIGN16 std::complex res[2]; - _mm_store_ps((float*)res, a.v); - return res[0]; - #else - std::complex res; - _mm_storel_pi((__m64*)&res, a.v); - return res; - #endif -} - -template<> EIGEN_STRONG_INLINE Packet2cf preverse(const Packet2cf& a) { return Packet2cf(_mm_castpd_ps(preverse(Packet2d(_mm_castps_pd(a.v))))); } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet2cf& a) -{ - return pfirst(Packet2cf(_mm_add_ps(a.v, _mm_movehl_ps(a.v,a.v)))); -} - -template<> EIGEN_STRONG_INLINE Packet2cf preduxp(const Packet2cf* vecs) -{ - return Packet2cf(_mm_add_ps(_mm_movelh_ps(vecs[0].v,vecs[1].v), _mm_movehl_ps(vecs[1].v,vecs[0].v))); -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet2cf& a) -{ - return pfirst(pmul(a, Packet2cf(_mm_movehl_ps(a.v,a.v)))); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2cf& first, const Packet2cf& second) - { - if (Offset==1) - { - first.v = _mm_movehl_ps(first.v, first.v); - first.v = _mm_movelh_ps(first.v, second.v); - } - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(a, pconj(b)); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_add_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), - _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(pconj(a), b); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_add_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), - _mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)), mask))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet2cf pmadd(const Packet2cf& x, const Packet2cf& y, const Packet2cf& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet2cf pmul(const Packet2cf& a, const Packet2cf& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return pconj(internal::pmul(a, b)); - #else - const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x00000000,0x80000000,0x00000000,0x80000000)); - return Packet2cf(_mm_sub_ps(_mm_xor_ps(_mm_mul_ps(vec4f_swizzle1(a.v, 0, 0, 2, 2), b.v), mask), - _mm_mul_ps(vec4f_swizzle1(a.v, 1, 1, 3, 3), - vec4f_swizzle1(b.v, 1, 0, 3, 2)))); - #endif - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet2cf,Packet4f) - -template<> EIGEN_STRONG_INLINE Packet2cf pdiv(const Packet2cf& a, const Packet2cf& b) -{ - // TODO optimize it for SSE3 and 4 - Packet2cf res = conj_helper().pmul(a,b); - __m128 s = _mm_mul_ps(b.v,b.v); - return Packet2cf(_mm_div_ps(res.v,_mm_add_ps(s,_mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(s), 0xb1))))); -} - -EIGEN_STRONG_INLINE Packet2cf pcplxflip/* */(const Packet2cf& x) -{ - return Packet2cf(vec4f_swizzle1(x.v, 1, 0, 3, 2)); -} - - -//---------- double ---------- -struct Packet1cd -{ - EIGEN_STRONG_INLINE Packet1cd() {} - EIGEN_STRONG_INLINE explicit Packet1cd(const __m128d& a) : v(a) {} - __m128d v; -}; - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits > : default_packet_traits -{ - typedef Packet1cd type; - typedef Packet1cd half; - enum { - Vectorizable = 1, - AlignedOnScalar = 0, - size = 1, - HasHalfPacket = 0, - - HasAdd = 1, - HasSub = 1, - HasMul = 1, - HasDiv = 1, - HasNegate = 1, - HasAbs = 0, - HasAbs2 = 0, - HasMin = 0, - HasMax = 0, - HasSetLinear = 0 - }; -}; -#endif - -template<> struct unpacket_traits { typedef std::complex type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; - -template<> EIGEN_STRONG_INLINE Packet1cd padd(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd psub(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pnegate(const Packet1cd& a) { return Packet1cd(pnegate(Packet2d(a.v))); } -template<> EIGEN_STRONG_INLINE Packet1cd pconj(const Packet1cd& a) -{ - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_xor_pd(a.v,mask)); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) -{ - #ifdef EIGEN_VECTORIZE_SSE3 - return Packet1cd(_mm_addsub_pd(_mm_mul_pd(_mm_movedup_pd(a.v), b.v), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); - return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), - _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)), mask))); - #endif -} - -template<> EIGEN_STRONG_INLINE Packet1cd pand (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_and_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd por (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_or_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pxor (const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_xor_pd(a.v,b.v)); } -template<> EIGEN_STRONG_INLINE Packet1cd pandnot(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_andnot_pd(a.v,b.v)); } - -// FIXME force unaligned load, this is a temporary fix -template<> EIGEN_STRONG_INLINE Packet1cd pload (const std::complex* from) -{ EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd ploadu(const std::complex* from) -{ EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu((const double*)from)); } -template<> EIGEN_STRONG_INLINE Packet1cd pset1(const std::complex& from) -{ /* here we really have to use unaligned loads :( */ return ploadu(&from); } - -template<> EIGEN_STRONG_INLINE Packet1cd ploaddup(const std::complex* from) { return pset1(*from); } - -// FIXME force unaligned store, this is a temporary fix -template<> EIGEN_STRONG_INLINE void pstore >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_ALIGNED_STORE pstore((double*)to, Packet2d(from.v)); } -template<> EIGEN_STRONG_INLINE void pstoreu >(std::complex * to, const Packet1cd& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu((double*)to, Packet2d(from.v)); } - -template<> EIGEN_STRONG_INLINE void prefetch >(const std::complex * addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } - -template<> EIGEN_STRONG_INLINE std::complex pfirst(const Packet1cd& a) -{ - EIGEN_ALIGN16 double res[2]; - _mm_store_pd(res, a.v); - return std::complex(res[0],res[1]); -} - -template<> EIGEN_STRONG_INLINE Packet1cd preverse(const Packet1cd& a) { return a; } - -template<> EIGEN_STRONG_INLINE std::complex predux(const Packet1cd& a) -{ - return pfirst(a); -} - -template<> EIGEN_STRONG_INLINE Packet1cd preduxp(const Packet1cd* vecs) -{ - return vecs[0]; -} - -template<> EIGEN_STRONG_INLINE std::complex predux_mul(const Packet1cd& a) -{ - return pfirst(a); -} - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet1cd& /*first*/, const Packet1cd& /*second*/) - { - // FIXME is it sure we never have to align a Packet1cd? - // Even though a std::complex has 16 bytes, it is not necessarily aligned on a 16 bytes boundary... - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(a, pconj(b)); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_add_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return internal::pmul(pconj(a), b); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_add_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), - _mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)), mask))); - #endif - } -}; - -template<> struct conj_helper -{ - EIGEN_STRONG_INLINE Packet1cd pmadd(const Packet1cd& x, const Packet1cd& y, const Packet1cd& c) const - { return padd(pmul(x,y),c); } - - EIGEN_STRONG_INLINE Packet1cd pmul(const Packet1cd& a, const Packet1cd& b) const - { - #ifdef EIGEN_VECTORIZE_SSE3 - return pconj(internal::pmul(a, b)); - #else - const __m128d mask = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - return Packet1cd(_mm_sub_pd(_mm_xor_pd(_mm_mul_pd(vec2d_swizzle1(a.v, 0, 0), b.v), mask), - _mm_mul_pd(vec2d_swizzle1(a.v, 1, 1), - vec2d_swizzle1(b.v, 1, 0)))); - #endif - } -}; - -EIGEN_MAKE_CONJ_HELPER_CPLX_REAL(Packet1cd,Packet2d) - -template<> EIGEN_STRONG_INLINE Packet1cd pdiv(const Packet1cd& a, const Packet1cd& b) -{ - // TODO optimize it for SSE3 and 4 - Packet1cd res = conj_helper().pmul(a,b); - __m128d s = _mm_mul_pd(b.v,b.v); - return Packet1cd(_mm_div_pd(res.v, _mm_add_pd(s,_mm_shuffle_pd(s, s, 0x1)))); -} - -EIGEN_STRONG_INLINE Packet1cd pcplxflip/* */(const Packet1cd& x) -{ - return Packet1cd(preverse(Packet2d(x.v))); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128d w1 = _mm_castps_pd(kernel.packet[0].v); - __m128d w2 = _mm_castps_pd(kernel.packet[1].v); - - __m128 tmp = _mm_castpd_ps(_mm_unpackhi_pd(w1, w2)); - kernel.packet[0].v = _mm_castpd_ps(_mm_unpacklo_pd(w1, w2)); - kernel.packet[1].v = tmp; -} - -template<> EIGEN_STRONG_INLINE Packet2cf pblend(const Selector<2>& ifPacket, const Packet2cf& thenPacket, const Packet2cf& elsePacket) { - __m128d result = pblend(ifPacket, _mm_castps_pd(thenPacket.v), _mm_castps_pd(elsePacket.v)); - return Packet2cf(_mm_castpd_ps(result)); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pinsertfirst(const Packet2cf& a, std::complex b) -{ - return Packet2cf(_mm_loadl_pi(a.v, reinterpret_cast(&b))); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pinsertfirst(const Packet1cd&, std::complex b) -{ - return pset1(b); -} - -template<> EIGEN_STRONG_INLINE Packet2cf pinsertlast(const Packet2cf& a, std::complex b) -{ - return Packet2cf(_mm_loadh_pi(a.v, reinterpret_cast(&b))); -} - -template<> EIGEN_STRONG_INLINE Packet1cd pinsertlast(const Packet1cd&, std::complex b) -{ - return pset1(b); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_COMPLEX_SSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h deleted file mode 100644 index 7b5f948e119..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/MathFunctions.h +++ /dev/null @@ -1,562 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007 Julien Pommier -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -/* The sin, cos, exp, and log functions of this file come from - * Julien Pommier's sse math library: http://gruntthepeon.free.fr/ssemath/ - */ - -#ifndef EIGEN_MATH_FUNCTIONS_SSE_H -#define EIGEN_MATH_FUNCTIONS_SSE_H - -namespace Eigen { - -namespace internal { - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f plog(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); - - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(inv_mant_mask, ~0x7f800000); - - /* the smallest non denormalized float number */ - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(min_norm_pos, 0x00800000); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(minus_inf, 0xff800000);//-1.f/0.f); - - /* natural logarithm computed for 4 simultaneous float - return NaN for x <= 0 - */ - _EIGEN_DECLARE_CONST_Packet4f(cephes_SQRTHF, 0.707106781186547524f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p0, 7.0376836292E-2f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p1, - 1.1514610310E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p2, 1.1676998740E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p3, - 1.2420140846E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p4, + 1.4249322787E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p5, - 1.6668057665E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p6, + 2.0000714765E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p7, - 2.4999993993E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_p8, + 3.3333331174E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q1, -2.12194440e-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_log_q2, 0.693359375f); - - - Packet4i emm0; - - Packet4f invalid_mask = _mm_cmpnge_ps(x, _mm_setzero_ps()); // not greater equal is true if x is NaN - Packet4f iszero_mask = _mm_cmpeq_ps(x, _mm_setzero_ps()); - - x = pmax(x, p4f_min_norm_pos); /* cut off denormalized stuff */ - emm0 = _mm_srli_epi32(_mm_castps_si128(x), 23); - - /* keep only the fractional part */ - x = _mm_and_ps(x, p4f_inv_mant_mask); - x = _mm_or_ps(x, p4f_half); - - emm0 = _mm_sub_epi32(emm0, p4i_0x7f); - Packet4f e = padd(Packet4f(_mm_cvtepi32_ps(emm0)), p4f_1); - - /* part2: - if( x < SQRTHF ) { - e -= 1; - x = x + x - 1.0; - } else { x = x - 1.0; } - */ - Packet4f mask = _mm_cmplt_ps(x, p4f_cephes_SQRTHF); - Packet4f tmp = pand(x, mask); - x = psub(x, p4f_1); - e = psub(e, pand(p4f_1, mask)); - x = padd(x, tmp); - - Packet4f x2 = pmul(x,x); - Packet4f x3 = pmul(x2,x); - - Packet4f y, y1, y2; - y = pmadd(p4f_cephes_log_p0, x, p4f_cephes_log_p1); - y1 = pmadd(p4f_cephes_log_p3, x, p4f_cephes_log_p4); - y2 = pmadd(p4f_cephes_log_p6, x, p4f_cephes_log_p7); - y = pmadd(y , x, p4f_cephes_log_p2); - y1 = pmadd(y1, x, p4f_cephes_log_p5); - y2 = pmadd(y2, x, p4f_cephes_log_p8); - y = pmadd(y, x3, y1); - y = pmadd(y, x3, y2); - y = pmul(y, x3); - - y1 = pmul(e, p4f_cephes_log_q1); - tmp = pmul(x2, p4f_half); - y = padd(y, y1); - x = psub(x, tmp); - y2 = pmul(e, p4f_cephes_log_q2); - x = padd(x, y); - x = padd(x, y2); - // negative arg will be NAN, 0 will be -INF - return _mm_or_ps(_mm_andnot_ps(iszero_mask, _mm_or_ps(x, invalid_mask)), - _mm_and_ps(iszero_mask, p4f_minus_inf)); -} - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pexp(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - _EIGEN_DECLARE_CONST_Packet4i(0x7f, 0x7f); - - - _EIGEN_DECLARE_CONST_Packet4f(exp_hi, 88.3762626647950f); - _EIGEN_DECLARE_CONST_Packet4f(exp_lo, -88.3762626647949f); - - _EIGEN_DECLARE_CONST_Packet4f(cephes_LOG2EF, 1.44269504088896341f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C1, 0.693359375f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C2, -2.12194440e-4f); - - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p0, 1.9875691500E-4f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p1, 1.3981999507E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p2, 8.3334519073E-3f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p3, 4.1665795894E-2f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459E-1f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201E-1f); - - Packet4f tmp, fx; - Packet4i emm0; - - // clamp x - x = pmax(pmin(x, p4f_exp_hi), p4f_exp_lo); - - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(x, p4f_cephes_LOG2EF, p4f_half); - -#ifdef EIGEN_VECTORIZE_SSE4_1 - fx = _mm_floor_ps(fx); -#else - emm0 = _mm_cvttps_epi32(fx); - tmp = _mm_cvtepi32_ps(emm0); - /* if greater, substract 1 */ - Packet4f mask = _mm_cmpgt_ps(tmp, fx); - mask = _mm_and_ps(mask, p4f_1); - fx = psub(tmp, mask); -#endif - - tmp = pmul(fx, p4f_cephes_exp_C1); - Packet4f z = pmul(fx, p4f_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - z = pmul(x,x); - - Packet4f y = p4f_cephes_exp_p0; - y = pmadd(y, x, p4f_cephes_exp_p1); - y = pmadd(y, x, p4f_cephes_exp_p2); - y = pmadd(y, x, p4f_cephes_exp_p3); - y = pmadd(y, x, p4f_cephes_exp_p4); - y = pmadd(y, x, p4f_cephes_exp_p5); - y = pmadd(y, z, x); - y = padd(y, p4f_1); - - // build 2^n - emm0 = _mm_cvttps_epi32(fx); - emm0 = _mm_add_epi32(emm0, p4i_0x7f); - emm0 = _mm_slli_epi32(emm0, 23); - return pmax(pmul(y, Packet4f(_mm_castsi128_ps(emm0))), _x); -} -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d pexp(const Packet2d& _x) -{ - Packet2d x = _x; - - _EIGEN_DECLARE_CONST_Packet2d(1 , 1.0); - _EIGEN_DECLARE_CONST_Packet2d(2 , 2.0); - _EIGEN_DECLARE_CONST_Packet2d(half, 0.5); - - _EIGEN_DECLARE_CONST_Packet2d(exp_hi, 709.437); - _EIGEN_DECLARE_CONST_Packet2d(exp_lo, -709.436139303); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_LOG2EF, 1.4426950408889634073599); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p0, 1.26177193074810590878e-4); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p1, 3.02994407707441961300e-2); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_p2, 9.99999999999999999910e-1); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q0, 3.00198505138664455042e-6); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q1, 2.52448340349684104192e-3); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q2, 2.27265548208155028766e-1); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_q3, 2.00000000000000000009e0); - - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C1, 0.693145751953125); - _EIGEN_DECLARE_CONST_Packet2d(cephes_exp_C2, 1.42860682030941723212e-6); - static const __m128i p4i_1023_0 = _mm_setr_epi32(1023, 1023, 0, 0); - - Packet2d tmp, fx; - Packet4i emm0; - - // clamp x - x = pmax(pmin(x, p2d_exp_hi), p2d_exp_lo); - /* express exp(x) as exp(g + n*log(2)) */ - fx = pmadd(p2d_cephes_LOG2EF, x, p2d_half); - -#ifdef EIGEN_VECTORIZE_SSE4_1 - fx = _mm_floor_pd(fx); -#else - emm0 = _mm_cvttpd_epi32(fx); - tmp = _mm_cvtepi32_pd(emm0); - /* if greater, substract 1 */ - Packet2d mask = _mm_cmpgt_pd(tmp, fx); - mask = _mm_and_pd(mask, p2d_1); - fx = psub(tmp, mask); -#endif - - tmp = pmul(fx, p2d_cephes_exp_C1); - Packet2d z = pmul(fx, p2d_cephes_exp_C2); - x = psub(x, tmp); - x = psub(x, z); - - Packet2d x2 = pmul(x,x); - - Packet2d px = p2d_cephes_exp_p0; - px = pmadd(px, x2, p2d_cephes_exp_p1); - px = pmadd(px, x2, p2d_cephes_exp_p2); - px = pmul (px, x); - - Packet2d qx = p2d_cephes_exp_q0; - qx = pmadd(qx, x2, p2d_cephes_exp_q1); - qx = pmadd(qx, x2, p2d_cephes_exp_q2); - qx = pmadd(qx, x2, p2d_cephes_exp_q3); - - x = pdiv(px,psub(qx,px)); - x = pmadd(p2d_2,x,p2d_1); - - // build 2^n - emm0 = _mm_cvttpd_epi32(fx); - emm0 = _mm_add_epi32(emm0, p4i_1023_0); - emm0 = _mm_slli_epi32(emm0, 20); - emm0 = _mm_shuffle_epi32(emm0, _MM_SHUFFLE(1,2,0,3)); - return pmax(pmul(x, Packet2d(_mm_castsi128_pd(emm0))), _x); -} - -/* evaluation of 4 sines at onces, using SSE2 intrinsics. - - The code is the exact rewriting of the cephes sinf function. - Precision is excellent as long as x < 8192 (I did not bother to - take into account the special handling they have for greater values - -- it does not return garbage for arguments over 8192, though, but - the extra precision is missing). - - Note that it is such that sinf((float)M_PI) = 8.74e-8, which is the - surprising but correct result. -*/ - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psin(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - - _EIGEN_DECLARE_CONST_Packet4i(1, 1); - _EIGEN_DECLARE_CONST_Packet4i(not1, ~1); - _EIGEN_DECLARE_CONST_Packet4i(2, 2); - _EIGEN_DECLARE_CONST_Packet4i(4, 4); - - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(sign_mask, 0x80000000); - - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP1,-0.78515625f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP2, -2.4187564849853515625e-4f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP3, -3.77489497744594108e-8f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p0, -1.9515295891E-4f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p1, 8.3321608736E-3f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p2, -1.6666654611E-1f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p0, 2.443315711809948E-005f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p1, -1.388731625493765E-003f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p2, 4.166664568298827E-002f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI - - Packet4f xmm1, xmm2, xmm3, sign_bit, y; - - Packet4i emm0, emm2; - sign_bit = x; - /* take the absolute value */ - x = pabs(x); - - /* take the modulo */ - - /* extract the sign bit (upper one) */ - sign_bit = _mm_and_ps(sign_bit, p4f_sign_mask); - - /* scale by 4/Pi */ - y = pmul(x, p4f_cephes_FOPI); - - /* store the integer part of y in mm0 */ - emm2 = _mm_cvttps_epi32(y); - /* j=(j+1) & (~1) (see the cephes sources) */ - emm2 = _mm_add_epi32(emm2, p4i_1); - emm2 = _mm_and_si128(emm2, p4i_not1); - y = _mm_cvtepi32_ps(emm2); - /* get the swap sign flag */ - emm0 = _mm_and_si128(emm2, p4i_4); - emm0 = _mm_slli_epi32(emm0, 29); - /* get the polynom selection mask - there is one polynom for 0 <= x <= Pi/4 - and another one for Pi/4 EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f pcos(const Packet4f& _x) -{ - Packet4f x = _x; - _EIGEN_DECLARE_CONST_Packet4f(1 , 1.0f); - _EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); - - _EIGEN_DECLARE_CONST_Packet4i(1, 1); - _EIGEN_DECLARE_CONST_Packet4i(not1, ~1); - _EIGEN_DECLARE_CONST_Packet4i(2, 2); - _EIGEN_DECLARE_CONST_Packet4i(4, 4); - - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP1,-0.78515625f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP2, -2.4187564849853515625e-4f); - _EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP3, -3.77489497744594108e-8f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p0, -1.9515295891E-4f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p1, 8.3321608736E-3f); - _EIGEN_DECLARE_CONST_Packet4f(sincof_p2, -1.6666654611E-1f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p0, 2.443315711809948E-005f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p1, -1.388731625493765E-003f); - _EIGEN_DECLARE_CONST_Packet4f(coscof_p2, 4.166664568298827E-002f); - _EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4 / M_PI - - Packet4f xmm1, xmm2, xmm3, y; - Packet4i emm0, emm2; - - x = pabs(x); - - /* scale by 4/Pi */ - y = pmul(x, p4f_cephes_FOPI); - - /* get the integer part of y */ - emm2 = _mm_cvttps_epi32(y); - /* j=(j+1) & (~1) (see the cephes sources) */ - emm2 = _mm_add_epi32(emm2, p4i_1); - emm2 = _mm_and_si128(emm2, p4i_not1); - y = _mm_cvtepi32_ps(emm2); - - emm2 = _mm_sub_epi32(emm2, p4i_2); - - /* get the swap sign flag */ - emm0 = _mm_andnot_si128(emm2, p4i_4); - emm0 = _mm_slli_epi32(emm0, 29); - /* get the polynom selection mask */ - emm2 = _mm_and_si128(emm2, p4i_2); - emm2 = _mm_cmpeq_epi32(emm2, _mm_setzero_si128()); - - Packet4f sign_bit = _mm_castsi128_ps(emm0); - Packet4f poly_mask = _mm_castsi128_ps(emm2); - - /* The magic pass: "Extended precision modular arithmetic" - x = ((x - y * DP1) - y * DP2) - y * DP3; */ - xmm1 = pmul(y, p4f_minus_cephes_DP1); - xmm2 = pmul(y, p4f_minus_cephes_DP2); - xmm3 = pmul(y, p4f_minus_cephes_DP3); - x = padd(x, xmm1); - x = padd(x, xmm2); - x = padd(x, xmm3); - - /* Evaluate the first polynom (0 <= x <= Pi/4) */ - y = p4f_coscof_p0; - Packet4f z = pmul(x,x); - - y = pmadd(y,z,p4f_coscof_p1); - y = pmadd(y,z,p4f_coscof_p2); - y = pmul(y, z); - y = pmul(y, z); - Packet4f tmp = _mm_mul_ps(z, p4f_half); - y = psub(y, tmp); - y = padd(y, p4f_1); - - /* Evaluate the second polynom (Pi/4 <= x <= 0) */ - Packet4f y2 = p4f_sincof_p0; - y2 = pmadd(y2, z, p4f_sincof_p1); - y2 = pmadd(y2, z, p4f_sincof_p2); - y2 = pmul(y2, z); - y2 = pmadd(y2, x, x); - - /* select the correct result from the two polynoms */ - y2 = _mm_and_ps(poly_mask, y2); - y = _mm_andnot_ps(poly_mask, y); - y = _mm_or_ps(y,y2); - - /* update the sign */ - return _mm_xor_ps(y, sign_bit); -} - -#if EIGEN_FAST_MATH - -// Functions for sqrt. -// The EIGEN_FAST_MATH version uses the _mm_rsqrt_ps approximation and one step -// of Newton's method, at a cost of 1-2 bits of precision as opposed to the -// exact solution. It does not handle +inf, or denormalized numbers correctly. -// The main advantage of this approach is not just speed, but also the fact that -// it can be inlined and pipelined with other computations, further reducing its -// effective latency. This is similar to Quake3's fast inverse square root. -// For detail see here: http://www.beyond3d.com/content/articles/8/ -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& _x) -{ - Packet4f half = pmul(_x, pset1(.5f)); - Packet4f denormal_mask = _mm_and_ps( - _mm_cmpge_ps(_x, _mm_setzero_ps()), - _mm_cmplt_ps(_x, pset1((std::numeric_limits::min)()))); - - // Compute approximate reciprocal sqrt. - Packet4f x = _mm_rsqrt_ps(_x); - // Do a single step of Newton's iteration. - x = pmul(x, psub(pset1(1.5f), pmul(half, pmul(x,x)))); - // Flush results for denormals to zero. - return _mm_andnot_ps(denormal_mask, pmul(_x,x)); -} - -#else - -template<>EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f psqrt(const Packet4f& x) { return _mm_sqrt_ps(x); } - -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d psqrt(const Packet2d& x) { return _mm_sqrt_pd(x); } - -#if EIGEN_FAST_MATH - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& _x) { - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(inf, 0x7f800000); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(nan, 0x7fc00000); - _EIGEN_DECLARE_CONST_Packet4f(one_point_five, 1.5f); - _EIGEN_DECLARE_CONST_Packet4f(minus_half, -0.5f); - _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(flt_min, 0x00800000); - - Packet4f neg_half = pmul(_x, p4f_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - Packet4f le_zero_mask = _mm_cmple_ps(_x, p4f_flt_min); - Packet4f x = _mm_andnot_ps(le_zero_mask, _mm_rsqrt_ps(_x)); - - // Fill in NaNs and Infs for the negative/zero entries. - Packet4f neg_mask = _mm_cmplt_ps(_x, _mm_setzero_ps()); - Packet4f zero_mask = _mm_andnot_ps(neg_mask, le_zero_mask); - Packet4f infs_and_nans = _mm_or_ps(_mm_and_ps(neg_mask, p4f_nan), - _mm_and_ps(zero_mask, p4f_inf)); - - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p4f_one_point_five)); - - // Insert NaNs and Infs in all the right places. - return _mm_or_ps(x, infs_and_nans); -} - -#else - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet4f prsqrt(const Packet4f& x) { - // Unfortunately we can't use the much faster mm_rqsrt_ps since it only provides an approximation. - return _mm_div_ps(pset1(1.0f), _mm_sqrt_ps(x)); -} - -#endif - -template<> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED -Packet2d prsqrt(const Packet2d& x) { - // Unfortunately we can't use the much faster mm_rqsrt_pd since it only provides an approximation. - return _mm_div_pd(pset1(1.0), _mm_sqrt_pd(x)); -} - -// Hyperbolic Tangent function. -template <> -EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet4f -ptanh(const Packet4f& x) { - return internal::generic_fast_tanh_float(x); -} - -} // end namespace internal - -namespace numext { - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -float sqrt(const float &x) -{ - return internal::pfirst(internal::Packet4f(_mm_sqrt_ss(_mm_set_ss(x)))); -} - -template<> -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE -double sqrt(const double &x) -{ -#if EIGEN_COMP_GNUC_STRICT - // This works around a GCC bug generating poor code for _mm_sqrt_pd - // See https://bitbucket.org/eigen/eigen/commits/14f468dba4d350d7c19c9b93072e19f7b3df563b - return internal::pfirst(internal::Packet2d(__builtin_ia32_sqrtsd(_mm_set_sd(x)))); -#else - return internal::pfirst(internal::Packet2d(_mm_sqrt_pd(_mm_set_sd(x)))); -#endif -} - -} // end namespace numex - -} // end namespace Eigen - -#endif // EIGEN_MATH_FUNCTIONS_SSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h deleted file mode 100644 index 60e2517e4bd..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/PacketMath.h +++ /dev/null @@ -1,895 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PACKET_MATH_SSE_H -#define EIGEN_PACKET_MATH_SSE_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD -#define EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD 8 -#endif - -#ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS (2*sizeof(void*)) -#endif - -#ifdef __FMA__ -#ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD -#define EIGEN_HAS_SINGLE_INSTRUCTION_MADD 1 -#endif -#endif - -#if ((defined EIGEN_VECTORIZE_AVX) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_MINGW) && (__GXX_ABI_VERSION < 1004)) || EIGEN_OS_QNX -// With GCC's default ABI version, a __m128 or __m256 are the same types and therefore we cannot -// have overloads for both types without linking error. -// One solution is to increase ABI version using -fabi-version=4 (or greater). -// Otherwise, we workaround this inconvenience by wrapping 128bit types into the following helper -// structure: -template -struct eigen_packet_wrapper -{ - EIGEN_ALWAYS_INLINE operator T&() { return m_val; } - EIGEN_ALWAYS_INLINE operator const T&() const { return m_val; } - EIGEN_ALWAYS_INLINE eigen_packet_wrapper() {} - EIGEN_ALWAYS_INLINE eigen_packet_wrapper(const T &v) : m_val(v) {} - EIGEN_ALWAYS_INLINE eigen_packet_wrapper& operator=(const T &v) { - m_val = v; - return *this; - } - - T m_val; -}; -typedef eigen_packet_wrapper<__m128> Packet4f; -typedef eigen_packet_wrapper<__m128i> Packet4i; -typedef eigen_packet_wrapper<__m128d> Packet2d; -#else -typedef __m128 Packet4f; -typedef __m128i Packet4i; -typedef __m128d Packet2d; -#endif - -template<> struct is_arithmetic<__m128> { enum { value = true }; }; -template<> struct is_arithmetic<__m128i> { enum { value = true }; }; -template<> struct is_arithmetic<__m128d> { enum { value = true }; }; - -#define vec4f_swizzle1(v,p,q,r,s) \ - (_mm_castsi128_ps(_mm_shuffle_epi32( _mm_castps_si128(v), ((s)<<6|(r)<<4|(q)<<2|(p))))) - -#define vec4i_swizzle1(v,p,q,r,s) \ - (_mm_shuffle_epi32( v, ((s)<<6|(r)<<4|(q)<<2|(p)))) - -#define vec2d_swizzle1(v,p,q) \ - (_mm_castsi128_pd(_mm_shuffle_epi32( _mm_castpd_si128(v), ((q*2+1)<<6|(q*2)<<4|(p*2+1)<<2|(p*2))))) - -#define vec4f_swizzle2(a,b,p,q,r,s) \ - (_mm_shuffle_ps( (a), (b), ((s)<<6|(r)<<4|(q)<<2|(p)))) - -#define vec4i_swizzle2(a,b,p,q,r,s) \ - (_mm_castps_si128( (_mm_shuffle_ps( _mm_castsi128_ps(a), _mm_castsi128_ps(b), ((s)<<6|(r)<<4|(q)<<2|(p)))))) - -#define _EIGEN_DECLARE_CONST_Packet4f(NAME,X) \ - const Packet4f p4f_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet2d(NAME,X) \ - const Packet2d p2d_##NAME = pset1(X) - -#define _EIGEN_DECLARE_CONST_Packet4f_FROM_INT(NAME,X) \ - const Packet4f p4f_##NAME = _mm_castsi128_ps(pset1(X)) - -#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ - const Packet4i p4i_##NAME = pset1(X) - - -// Use the packet_traits defined in AVX/PacketMath.h instead if we're going -// to leverage AVX instructions. -#ifndef EIGEN_VECTORIZE_AVX -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4f type; - typedef Packet4f half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - HasHalfPacket = 0, - - HasDiv = 1, - HasSin = EIGEN_FAST_MATH, - HasCos = EIGEN_FAST_MATH, - HasLog = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasTanh = EIGEN_FAST_MATH, - HasBlend = 1 - -#ifdef EIGEN_VECTORIZE_SSE4_1 - , - HasRound = 1, - HasFloor = 1, - HasCeil = 1 -#endif - }; -}; -template<> struct packet_traits : default_packet_traits -{ - typedef Packet2d type; - typedef Packet2d half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=2, - HasHalfPacket = 0, - - HasDiv = 1, - HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, - HasBlend = 1 - -#ifdef EIGEN_VECTORIZE_SSE4_1 - , - HasRound = 1, - HasFloor = 1, - HasCeil = 1 -#endif - }; -}; -#endif -template<> struct packet_traits : default_packet_traits -{ - typedef Packet4i type; - typedef Packet4i half; - enum { - Vectorizable = 1, - AlignedOnScalar = 1, - size=4, - - HasBlend = 1 - }; -}; - -template<> struct unpacket_traits { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; -template<> struct unpacket_traits { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; -template<> struct unpacket_traits { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; - -#ifndef EIGEN_VECTORIZE_AVX -template<> struct scalar_div_cost { enum { value = 7 }; }; -template<> struct scalar_div_cost { enum { value = 8 }; }; -#endif - -#if EIGEN_COMP_MSVC==1500 -// Workaround MSVC 9 internal compiler error. -// TODO: It has been detected with win64 builds (amd64), so let's check whether it also happens in 32bits+SSE mode -// TODO: let's check whether there does not exist a better fix, like adding a pset0() function. (it crashed on pset1(0)). -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { return _mm_set_ps(from,from,from,from); } -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { return _mm_set_pd(from,from); } -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) { return _mm_set_epi32(from,from,from,from); } -#else -template<> EIGEN_STRONG_INLINE Packet4f pset1(const float& from) { return _mm_set_ps1(from); } -template<> EIGEN_STRONG_INLINE Packet2d pset1(const double& from) { return _mm_set1_pd(from); } -template<> EIGEN_STRONG_INLINE Packet4i pset1(const int& from) { return _mm_set1_epi32(from); } -#endif - -// GCC generates a shufps instruction for _mm_set1_ps/_mm_load1_ps instead of the more efficient pshufd instruction. -// However, using inrinsics for pset1 makes gcc to generate crappy code in some cases (see bug 203) -// Using inline assembly is also not an option because then gcc fails to reorder properly the instructions. -// Therefore, we introduced the pload1 functions to be used in product kernels for which bug 203 does not apply. -// Also note that with AVX, we want it to generate a vbroadcastss. -#if EIGEN_COMP_GNUC_STRICT && (!defined __AVX__) -template<> EIGEN_STRONG_INLINE Packet4f pload1(const float *from) { - return vec4f_swizzle1(_mm_load_ss(from),0,0,0,0); -} -#endif - -template<> EIGEN_STRONG_INLINE Packet4f plset(const float& a) { return _mm_add_ps(pset1(a), _mm_set_ps(3,2,1,0)); } -template<> EIGEN_STRONG_INLINE Packet2d plset(const double& a) { return _mm_add_pd(pset1(a),_mm_set_pd(1,0)); } -template<> EIGEN_STRONG_INLINE Packet4i plset(const int& a) { return _mm_add_epi32(pset1(a),_mm_set_epi32(3,2,1,0)); } - -template<> EIGEN_STRONG_INLINE Packet4f padd(const Packet4f& a, const Packet4f& b) { return _mm_add_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d padd(const Packet2d& a, const Packet2d& b) { return _mm_add_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i padd(const Packet4i& a, const Packet4i& b) { return _mm_add_epi32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f psub(const Packet4f& a, const Packet4f& b) { return _mm_sub_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d psub(const Packet2d& a, const Packet2d& b) { return _mm_sub_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i psub(const Packet4i& a, const Packet4i& b) { return _mm_sub_epi32(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a) -{ - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000)); - return _mm_xor_ps(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet2d pnegate(const Packet2d& a) -{ - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x80000000,0x0,0x80000000)); - return _mm_xor_pd(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet4i pnegate(const Packet4i& a) -{ - return psub(Packet4i(_mm_setr_epi32(0,0,0,0)), a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pconj(const Packet4f& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet2d pconj(const Packet2d& a) { return a; } -template<> EIGEN_STRONG_INLINE Packet4i pconj(const Packet4i& a) { return a; } - -template<> EIGEN_STRONG_INLINE Packet4f pmul(const Packet4f& a, const Packet4f& b) { return _mm_mul_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmul(const Packet2d& a, const Packet2d& b) { return _mm_mul_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmul(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_mullo_epi32(a,b); -#else - // this version is slightly faster than 4 scalar products - return vec4i_swizzle1( - vec4i_swizzle2( - _mm_mul_epu32(a,b), - _mm_mul_epu32(vec4i_swizzle1(a,1,0,3,2), - vec4i_swizzle1(b,1,0,3,2)), - 0,2,0,2), - 0,2,1,3); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pdiv(const Packet4f& a, const Packet4f& b) { return _mm_div_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pdiv(const Packet2d& a, const Packet2d& b) { return _mm_div_pd(a,b); } - -// for some weird raisons, it has to be overloaded for packet of integers -template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i& b, const Packet4i& c) { return padd(pmul(a,b), c); } -#ifdef __FMA__ -template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fmadd_ps(a,b,c); } -template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fmadd_pd(a,b,c); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f pmin(const Packet4f& a, const Packet4f& b) { return _mm_min_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmin(const Packet2d& a, const Packet2d& b) { return _mm_min_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmin(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_min_epi32(a,b); -#else - // after some bench, this version *is* faster than a scalar implementation - Packet4i mask = _mm_cmplt_epi32(a,b); - return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pmax(const Packet4f& a, const Packet4f& b) { return _mm_max_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pmax(const Packet2d& a, const Packet2d& b) { return _mm_max_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pmax(const Packet4i& a, const Packet4i& b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_max_epi32(a,b); -#else - // after some bench, this version *is* faster than a scalar implementation - Packet4i mask = _mm_cmpgt_epi32(a,b); - return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b)); -#endif -} - -#ifdef EIGEN_VECTORIZE_SSE4_1 -template<> EIGEN_STRONG_INLINE Packet4f pround(const Packet4f& a) { return _mm_round_ps(a, 0); } -template<> EIGEN_STRONG_INLINE Packet2d pround(const Packet2d& a) { return _mm_round_pd(a, 0); } - -template<> EIGEN_STRONG_INLINE Packet4f pceil(const Packet4f& a) { return _mm_ceil_ps(a); } -template<> EIGEN_STRONG_INLINE Packet2d pceil(const Packet2d& a) { return _mm_ceil_pd(a); } - -template<> EIGEN_STRONG_INLINE Packet4f pfloor(const Packet4f& a) { return _mm_floor_ps(a); } -template<> EIGEN_STRONG_INLINE Packet2d pfloor(const Packet2d& a) { return _mm_floor_pd(a); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f pand(const Packet4f& a, const Packet4f& b) { return _mm_and_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pand(const Packet2d& a, const Packet2d& b) { return _mm_and_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pand(const Packet4i& a, const Packet4i& b) { return _mm_and_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f por(const Packet4f& a, const Packet4f& b) { return _mm_or_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d por(const Packet2d& a, const Packet2d& b) { return _mm_or_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i por(const Packet4i& a, const Packet4i& b) { return _mm_or_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pxor(const Packet4f& a, const Packet4f& b) { return _mm_xor_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pxor(const Packet2d& a, const Packet2d& b) { return _mm_xor_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pxor(const Packet4i& a, const Packet4i& b) { return _mm_xor_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pandnot(const Packet4f& a, const Packet4f& b) { return _mm_andnot_ps(a,b); } -template<> EIGEN_STRONG_INLINE Packet2d pandnot(const Packet2d& a, const Packet2d& b) { return _mm_andnot_pd(a,b); } -template<> EIGEN_STRONG_INLINE Packet4i pandnot(const Packet4i& a, const Packet4i& b) { return _mm_andnot_si128(a,b); } - -template<> EIGEN_STRONG_INLINE Packet4f pload(const float* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_ps(from); } -template<> EIGEN_STRONG_INLINE Packet2d pload(const double* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_pd(from); } -template<> EIGEN_STRONG_INLINE Packet4i pload(const int* from) { EIGEN_DEBUG_ALIGNED_LOAD return _mm_load_si128(reinterpret_cast(from)); } - -#if EIGEN_COMP_MSVC - template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) { - EIGEN_DEBUG_UNALIGNED_LOAD - #if (EIGEN_COMP_MSVC==1600) - // NOTE Some version of MSVC10 generates bad code when using _mm_loadu_ps - // (i.e., it does not generate an unaligned load!! - __m128 res = _mm_loadl_pi(_mm_set1_ps(0.0f), (const __m64*)(from)); - res = _mm_loadh_pi(res, (const __m64*)(from+2)); - return res; - #else - return _mm_loadu_ps(from); - #endif - } -#else -// NOTE: with the code below, MSVC's compiler crashes! - -template<> EIGEN_STRONG_INLINE Packet4f ploadu(const float* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_ps(from); -} -#endif - -template<> EIGEN_STRONG_INLINE Packet2d ploadu(const double* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_pd(from); -} -template<> EIGEN_STRONG_INLINE Packet4i ploadu(const int* from) -{ - EIGEN_DEBUG_UNALIGNED_LOAD - return _mm_loadu_si128(reinterpret_cast(from)); -} - - -template<> EIGEN_STRONG_INLINE Packet4f ploaddup(const float* from) -{ - return vec4f_swizzle1(_mm_castpd_ps(_mm_load_sd(reinterpret_cast(from))), 0, 0, 1, 1); -} -template<> EIGEN_STRONG_INLINE Packet2d ploaddup(const double* from) -{ return pset1(from[0]); } -template<> EIGEN_STRONG_INLINE Packet4i ploaddup(const int* from) -{ - Packet4i tmp; - tmp = _mm_loadl_epi64(reinterpret_cast(from)); - return vec4i_swizzle1(tmp, 0, 0, 1, 1); -} - -template<> EIGEN_STRONG_INLINE void pstore(float* to, const Packet4f& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(double* to, const Packet2d& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstore(int* to, const Packet4i& from) { EIGEN_DEBUG_ALIGNED_STORE _mm_store_si128(reinterpret_cast<__m128i*>(to), from); } - -template<> EIGEN_STRONG_INLINE void pstoreu(double* to, const Packet2d& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_pd(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_ps(to, from); } -template<> EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm_storeu_si128(reinterpret_cast<__m128i*>(to), from); } - -template<> EIGEN_DEVICE_FUNC inline Packet4f pgather(const float* from, Index stride) -{ - return _mm_set_ps(from[3*stride], from[2*stride], from[1*stride], from[0*stride]); -} -template<> EIGEN_DEVICE_FUNC inline Packet2d pgather(const double* from, Index stride) -{ - return _mm_set_pd(from[1*stride], from[0*stride]); -} -template<> EIGEN_DEVICE_FUNC inline Packet4i pgather(const int* from, Index stride) -{ - return _mm_set_epi32(from[3*stride], from[2*stride], from[1*stride], from[0*stride]); - } - -template<> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet4f& from, Index stride) -{ - to[stride*0] = _mm_cvtss_f32(from); - to[stride*1] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 1)); - to[stride*2] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 2)); - to[stride*3] = _mm_cvtss_f32(_mm_shuffle_ps(from, from, 3)); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet2d& from, Index stride) -{ - to[stride*0] = _mm_cvtsd_f64(from); - to[stride*1] = _mm_cvtsd_f64(_mm_shuffle_pd(from, from, 1)); -} -template<> EIGEN_DEVICE_FUNC inline void pscatter(int* to, const Packet4i& from, Index stride) -{ - to[stride*0] = _mm_cvtsi128_si32(from); - to[stride*1] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 1)); - to[stride*2] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 2)); - to[stride*3] = _mm_cvtsi128_si32(_mm_shuffle_epi32(from, 3)); -} - -// some compilers might be tempted to perform multiple moves instead of using a vector path. -template<> EIGEN_STRONG_INLINE void pstore1(float* to, const float& a) -{ - Packet4f pa = _mm_set_ss(a); - pstore(to, Packet4f(vec4f_swizzle1(pa,0,0,0,0))); -} -// some compilers might be tempted to perform multiple moves instead of using a vector path. -template<> EIGEN_STRONG_INLINE void pstore1(double* to, const double& a) -{ - Packet2d pa = _mm_set_sd(a); - pstore(to, Packet2d(vec2d_swizzle1(pa,0,0))); -} - -#if EIGEN_COMP_PGI -typedef const void * SsePrefetchPtrType; -#else -typedef const char * SsePrefetchPtrType; -#endif - -#ifndef EIGEN_VECTORIZE_AVX -template<> EIGEN_STRONG_INLINE void prefetch(const float* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const double* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -template<> EIGEN_STRONG_INLINE void prefetch(const int* addr) { _mm_prefetch((SsePrefetchPtrType)(addr), _MM_HINT_T0); } -#endif - -#if EIGEN_COMP_MSVC_STRICT && EIGEN_OS_WIN64 -// The temporary variable fixes an internal compilation error in vs <= 2008 and a wrong-result bug in vs 2010 -// Direct of the struct members fixed bug #62. -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { return a.m128_f32[0]; } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { return a.m128d_f64[0]; } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int x = _mm_cvtsi128_si32(a); return x; } -#elif EIGEN_COMP_MSVC_STRICT -// The temporary variable fixes an internal compilation error in vs <= 2008 and a wrong-result bug in vs 2010 -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { float x = _mm_cvtss_f32(a); return x; } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { double x = _mm_cvtsd_f64(a); return x; } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { int x = _mm_cvtsi128_si32(a); return x; } -#else -template<> EIGEN_STRONG_INLINE float pfirst(const Packet4f& a) { return _mm_cvtss_f32(a); } -template<> EIGEN_STRONG_INLINE double pfirst(const Packet2d& a) { return _mm_cvtsd_f64(a); } -template<> EIGEN_STRONG_INLINE int pfirst(const Packet4i& a) { return _mm_cvtsi128_si32(a); } -#endif - -template<> EIGEN_STRONG_INLINE Packet4f preverse(const Packet4f& a) -{ return _mm_shuffle_ps(a,a,0x1B); } -template<> EIGEN_STRONG_INLINE Packet2d preverse(const Packet2d& a) -{ return _mm_shuffle_pd(a,a,0x1); } -template<> EIGEN_STRONG_INLINE Packet4i preverse(const Packet4i& a) -{ return _mm_shuffle_epi32(a,0x1B); } - -template<> EIGEN_STRONG_INLINE Packet4f pabs(const Packet4f& a) -{ - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF,0x7FFFFFFF)); - return _mm_and_ps(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet2d pabs(const Packet2d& a) -{ - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0xFFFFFFFF,0x7FFFFFFF,0xFFFFFFFF,0x7FFFFFFF)); - return _mm_and_pd(a,mask); -} -template<> EIGEN_STRONG_INLINE Packet4i pabs(const Packet4i& a) -{ - #ifdef EIGEN_VECTORIZE_SSSE3 - return _mm_abs_epi32(a); - #else - Packet4i aux = _mm_srai_epi32(a,31); - return _mm_sub_epi32(_mm_xor_si128(a,aux),aux); - #endif -} - -// with AVX, the default implementations based on pload1 are faster -#ifndef __AVX__ -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const float *a, - Packet4f& a0, Packet4f& a1, Packet4f& a2, Packet4f& a3) -{ - a3 = pload(a); - a0 = vec4f_swizzle1(a3, 0,0,0,0); - a1 = vec4f_swizzle1(a3, 1,1,1,1); - a2 = vec4f_swizzle1(a3, 2,2,2,2); - a3 = vec4f_swizzle1(a3, 3,3,3,3); -} -template<> EIGEN_STRONG_INLINE void -pbroadcast4(const double *a, - Packet2d& a0, Packet2d& a1, Packet2d& a2, Packet2d& a3) -{ -#ifdef EIGEN_VECTORIZE_SSE3 - a0 = _mm_loaddup_pd(a+0); - a1 = _mm_loaddup_pd(a+1); - a2 = _mm_loaddup_pd(a+2); - a3 = _mm_loaddup_pd(a+3); -#else - a1 = pload(a); - a0 = vec2d_swizzle1(a1, 0,0); - a1 = vec2d_swizzle1(a1, 1,1); - a3 = pload(a+2); - a2 = vec2d_swizzle1(a3, 0,0); - a3 = vec2d_swizzle1(a3, 1,1); -#endif -} -#endif - -EIGEN_STRONG_INLINE void punpackp(Packet4f* vecs) -{ - vecs[1] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0x55)); - vecs[2] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0xAA)); - vecs[3] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0xFF)); - vecs[0] = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(vecs[0]), 0x00)); -} - -#ifdef EIGEN_VECTORIZE_SSE3 -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - return _mm_hadd_ps(_mm_hadd_ps(vecs[0], vecs[1]),_mm_hadd_ps(vecs[2], vecs[3])); -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - return _mm_hadd_pd(vecs[0], vecs[1]); -} - -#else -template<> EIGEN_STRONG_INLINE Packet4f preduxp(const Packet4f* vecs) -{ - Packet4f tmp0, tmp1, tmp2; - tmp0 = _mm_unpacklo_ps(vecs[0], vecs[1]); - tmp1 = _mm_unpackhi_ps(vecs[0], vecs[1]); - tmp2 = _mm_unpackhi_ps(vecs[2], vecs[3]); - tmp0 = _mm_add_ps(tmp0, tmp1); - tmp1 = _mm_unpacklo_ps(vecs[2], vecs[3]); - tmp1 = _mm_add_ps(tmp1, tmp2); - tmp2 = _mm_movehl_ps(tmp1, tmp0); - tmp0 = _mm_movelh_ps(tmp0, tmp1); - return _mm_add_ps(tmp0, tmp2); -} - -template<> EIGEN_STRONG_INLINE Packet2d preduxp(const Packet2d* vecs) -{ - return _mm_add_pd(_mm_unpacklo_pd(vecs[0], vecs[1]), _mm_unpackhi_pd(vecs[0], vecs[1])); -} -#endif // SSE3 - -template<> EIGEN_STRONG_INLINE float predux(const Packet4f& a) -{ - // Disable SSE3 _mm_hadd_pd that is extremely slow on all existing Intel's architectures - // (from Nehalem to Haswell) -// #ifdef EIGEN_VECTORIZE_SSE3 -// Packet4f tmp = _mm_add_ps(a, vec4f_swizzle1(a,2,3,2,3)); -// return pfirst(_mm_hadd_ps(tmp, tmp)); -// #else - Packet4f tmp = _mm_add_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_add_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -// #endif -} - -template<> EIGEN_STRONG_INLINE double predux(const Packet2d& a) -{ - // Disable SSE3 _mm_hadd_pd that is extremely slow on all existing Intel's architectures - // (from Nehalem to Haswell) -// #ifdef EIGEN_VECTORIZE_SSE3 -// return pfirst(_mm_hadd_pd(a, a)); -// #else - return pfirst(_mm_add_sd(a, _mm_unpackhi_pd(a,a))); -// #endif -} - -#ifdef EIGEN_VECTORIZE_SSSE3 -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - return _mm_hadd_epi32(_mm_hadd_epi32(vecs[0], vecs[1]),_mm_hadd_epi32(vecs[2], vecs[3])); -} -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i tmp0 = _mm_hadd_epi32(a,a); - return pfirst(_mm_hadd_epi32(tmp0,tmp0)); -} -#else -template<> EIGEN_STRONG_INLINE int predux(const Packet4i& a) -{ - Packet4i tmp = _mm_add_epi32(a, _mm_unpackhi_epi64(a,a)); - return pfirst(tmp) + pfirst(_mm_shuffle_epi32(tmp, 1)); -} - -template<> EIGEN_STRONG_INLINE Packet4i preduxp(const Packet4i* vecs) -{ - Packet4i tmp0, tmp1, tmp2; - tmp0 = _mm_unpacklo_epi32(vecs[0], vecs[1]); - tmp1 = _mm_unpackhi_epi32(vecs[0], vecs[1]); - tmp2 = _mm_unpackhi_epi32(vecs[2], vecs[3]); - tmp0 = _mm_add_epi32(tmp0, tmp1); - tmp1 = _mm_unpacklo_epi32(vecs[2], vecs[3]); - tmp1 = _mm_add_epi32(tmp1, tmp2); - tmp2 = _mm_unpacklo_epi64(tmp0, tmp1); - tmp0 = _mm_unpackhi_epi64(tmp0, tmp1); - return _mm_add_epi32(tmp0, tmp2); -} -#endif -// Other reduction functions: - -// mul -template<> EIGEN_STRONG_INLINE float predux_mul(const Packet4f& a) -{ - Packet4f tmp = _mm_mul_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_mul_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_mul(const Packet2d& a) -{ - return pfirst(_mm_mul_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_mul(const Packet4i& a) -{ - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., reusing pmul is very slow !) - // TODO try to call _mm_mul_epu32 directly - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - return (aux[0] * aux[1]) * (aux[2] * aux[3]);; -} - -// min -template<> EIGEN_STRONG_INLINE float predux_min(const Packet4f& a) -{ - Packet4f tmp = _mm_min_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_min_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_min(const Packet2d& a) -{ - return pfirst(_mm_min_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_min(const Packet4i& a) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - Packet4i tmp = _mm_min_epi32(a, _mm_shuffle_epi32(a, _MM_SHUFFLE(0,0,3,2))); - return pfirst(_mm_min_epi32(tmp,_mm_shuffle_epi32(tmp, 1))); -#else - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., it does not like using std::min after the pstore !!) - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - int aux0 = aux[0] EIGEN_STRONG_INLINE float predux_max(const Packet4f& a) -{ - Packet4f tmp = _mm_max_ps(a, _mm_movehl_ps(a,a)); - return pfirst(_mm_max_ss(tmp, _mm_shuffle_ps(tmp,tmp, 1))); -} -template<> EIGEN_STRONG_INLINE double predux_max(const Packet2d& a) -{ - return pfirst(_mm_max_sd(a, _mm_unpackhi_pd(a,a))); -} -template<> EIGEN_STRONG_INLINE int predux_max(const Packet4i& a) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - Packet4i tmp = _mm_max_epi32(a, _mm_shuffle_epi32(a, _MM_SHUFFLE(0,0,3,2))); - return pfirst(_mm_max_epi32(tmp,_mm_shuffle_epi32(tmp, 1))); -#else - // after some experiments, it is seems this is the fastest way to implement it - // for GCC (eg., it does not like using std::min after the pstore !!) - EIGEN_ALIGN16 int aux[4]; - pstore(aux, a); - int aux0 = aux[0]>aux[1] ? aux[0] : aux[1]; - int aux2 = aux[2]>aux[3] ? aux[2] : aux[3]; - return aux0>aux2 ? aux0 : aux2; -#endif // EIGEN_VECTORIZE_SSE4_1 -} - -#if EIGEN_COMP_GNUC -// template <> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) -// { -// Packet4f res = b; -// asm("mulps %[a], %[b] \n\taddps %[c], %[b]" : [b] "+x" (res) : [a] "x" (a), [c] "x" (c)); -// return res; -// } -// EIGEN_STRONG_INLINE Packet4i _mm_alignr_epi8(const Packet4i& a, const Packet4i& b, const int i) -// { -// Packet4i res = a; -// asm("palignr %[i], %[a], %[b] " : [b] "+x" (res) : [a] "x" (a), [i] "i" (i)); -// return res; -// } -#endif - -#ifdef EIGEN_VECTORIZE_SSSE3 -// SSSE3 versions -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { - if (Offset!=0) - first = _mm_castsi128_ps(_mm_alignr_epi8(_mm_castps_si128(second), _mm_castps_si128(first), Offset*4)); - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { - if (Offset!=0) - first = _mm_alignr_epi8(second,first, Offset*4); - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset==1) - first = _mm_castsi128_pd(_mm_alignr_epi8(_mm_castpd_si128(second), _mm_castpd_si128(first), 8)); - } -}; -#else -// SSE2 versions -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4f& first, const Packet4f& second) - { - if (Offset==1) - { - first = _mm_move_ss(first,second); - first = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(first),0x39)); - } - else if (Offset==2) - { - first = _mm_movehl_ps(first,first); - first = _mm_movelh_ps(first,second); - } - else if (Offset==3) - { - first = _mm_move_ss(first,second); - first = _mm_shuffle_ps(first,second,0x93); - } - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet4i& first, const Packet4i& second) - { - if (Offset==1) - { - first = _mm_castps_si128(_mm_move_ss(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - first = _mm_shuffle_epi32(first,0x39); - } - else if (Offset==2) - { - first = _mm_castps_si128(_mm_movehl_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(first))); - first = _mm_castps_si128(_mm_movelh_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - } - else if (Offset==3) - { - first = _mm_castps_si128(_mm_move_ss(_mm_castsi128_ps(first),_mm_castsi128_ps(second))); - first = _mm_castps_si128(_mm_shuffle_ps(_mm_castsi128_ps(first),_mm_castsi128_ps(second),0x93)); - } - } -}; - -template -struct palign_impl -{ - static EIGEN_STRONG_INLINE void run(Packet2d& first, const Packet2d& second) - { - if (Offset==1) - { - first = _mm_castps_pd(_mm_movehl_ps(_mm_castpd_ps(first),_mm_castpd_ps(first))); - first = _mm_castps_pd(_mm_movelh_ps(_mm_castpd_ps(first),_mm_castpd_ps(second))); - } - } -}; -#endif - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - _MM_TRANSPOSE4_PS(kernel.packet[0], kernel.packet[1], kernel.packet[2], kernel.packet[3]); -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128d tmp = _mm_unpackhi_pd(kernel.packet[0], kernel.packet[1]); - kernel.packet[0] = _mm_unpacklo_pd(kernel.packet[0], kernel.packet[1]); - kernel.packet[1] = tmp; -} - -EIGEN_DEVICE_FUNC inline void -ptranspose(PacketBlock& kernel) { - __m128i T0 = _mm_unpacklo_epi32(kernel.packet[0], kernel.packet[1]); - __m128i T1 = _mm_unpacklo_epi32(kernel.packet[2], kernel.packet[3]); - __m128i T2 = _mm_unpackhi_epi32(kernel.packet[0], kernel.packet[1]); - __m128i T3 = _mm_unpackhi_epi32(kernel.packet[2], kernel.packet[3]); - - kernel.packet[0] = _mm_unpacklo_epi64(T0, T1); - kernel.packet[1] = _mm_unpackhi_epi64(T0, T1); - kernel.packet[2] = _mm_unpacklo_epi64(T2, T3); - kernel.packet[3] = _mm_unpackhi_epi64(T2, T3); -} - -template<> EIGEN_STRONG_INLINE Packet4i pblend(const Selector<4>& ifPacket, const Packet4i& thenPacket, const Packet4i& elsePacket) { - const __m128i zero = _mm_setzero_si128(); - const __m128i select = _mm_set_epi32(ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m128i false_mask = _mm_cmpeq_epi32(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_epi8(thenPacket, elsePacket, false_mask); -#else - return _mm_or_si128(_mm_andnot_si128(false_mask, thenPacket), _mm_and_si128(false_mask, elsePacket)); -#endif -} -template<> EIGEN_STRONG_INLINE Packet4f pblend(const Selector<4>& ifPacket, const Packet4f& thenPacket, const Packet4f& elsePacket) { - const __m128 zero = _mm_setzero_ps(); - const __m128 select = _mm_set_ps(ifPacket.select[3], ifPacket.select[2], ifPacket.select[1], ifPacket.select[0]); - __m128 false_mask = _mm_cmpeq_ps(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_ps(thenPacket, elsePacket, false_mask); -#else - return _mm_or_ps(_mm_andnot_ps(false_mask, thenPacket), _mm_and_ps(false_mask, elsePacket)); -#endif -} -template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, const Packet2d& thenPacket, const Packet2d& elsePacket) { - const __m128d zero = _mm_setzero_pd(); - const __m128d select = _mm_set_pd(ifPacket.select[1], ifPacket.select[0]); - __m128d false_mask = _mm_cmpeq_pd(select, zero); -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blendv_pd(thenPacket, elsePacket, false_mask); -#else - return _mm_or_pd(_mm_andnot_pd(false_mask, thenPacket), _mm_and_pd(false_mask, elsePacket)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pinsertfirst(const Packet4f& a, float b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_ps(a,pset1(b),1); -#else - return _mm_move_ss(a, _mm_load_ss(&b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet2d pinsertfirst(const Packet2d& a, double b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_pd(a,pset1(b),1); -#else - return _mm_move_sd(a, _mm_load_sd(&b)); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet4f pinsertlast(const Packet4f& a, float b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_ps(a,pset1(b),(1<<3)); -#else - const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x0,0x0,0x0,0xFFFFFFFF)); - return _mm_or_ps(_mm_andnot_ps(mask, a), _mm_and_ps(mask, pset1(b))); -#endif -} - -template<> EIGEN_STRONG_INLINE Packet2d pinsertlast(const Packet2d& a, double b) -{ -#ifdef EIGEN_VECTORIZE_SSE4_1 - return _mm_blend_pd(a,pset1(b),(1<<1)); -#else - const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x0,0xFFFFFFFF,0xFFFFFFFF)); - return _mm_or_pd(_mm_andnot_pd(mask, a), _mm_and_pd(mask, pset1(b))); -#endif -} - -// Scalar path for pmadd with FMA to ensure consistency with vectorized path. -#ifdef __FMA__ -template<> EIGEN_STRONG_INLINE float pmadd(const float& a, const float& b, const float& c) { - return ::fmaf(a,b,c); -} -template<> EIGEN_STRONG_INLINE double pmadd(const double& a, const double& b, const double& c) { - return ::fma(a,b,c); -} -#endif - -} // end namespace internal - -} // end namespace Eigen - -#if EIGEN_COMP_PGI -// PGI++ does not define the following intrinsics in C++ mode. -static inline __m128 _mm_castpd_ps (__m128d x) { return reinterpret_cast<__m128&>(x); } -static inline __m128i _mm_castpd_si128(__m128d x) { return reinterpret_cast<__m128i&>(x); } -static inline __m128d _mm_castps_pd (__m128 x) { return reinterpret_cast<__m128d&>(x); } -static inline __m128i _mm_castps_si128(__m128 x) { return reinterpret_cast<__m128i&>(x); } -static inline __m128 _mm_castsi128_ps(__m128i x) { return reinterpret_cast<__m128&>(x); } -static inline __m128d _mm_castsi128_pd(__m128i x) { return reinterpret_cast<__m128d&>(x); } -#endif - -#endif // EIGEN_PACKET_MATH_SSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h deleted file mode 100644 index c6ca8c716c0..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/arch/SSE/TypeCasting.h +++ /dev/null @@ -1,77 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2015 Benoit Steiner -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TYPE_CASTING_SSE_H -#define EIGEN_TYPE_CASTING_SSE_H - -namespace Eigen { - -namespace internal { - -#ifndef EIGEN_VECTORIZE_AVX -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 2, - TgtCoeffRatio = 1 - }; -}; - -template <> -struct type_casting_traits { - enum { - VectorizedCast = 1, - SrcCoeffRatio = 1, - TgtCoeffRatio = 2 - }; -}; -#endif - -template<> EIGEN_STRONG_INLINE Packet4i pcast(const Packet4f& a) { - return _mm_cvttps_epi32(a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet4i& a) { - return _mm_cvtepi32_ps(a); -} - -template<> EIGEN_STRONG_INLINE Packet4f pcast(const Packet2d& a, const Packet2d& b) { - return _mm_shuffle_ps(_mm_cvtpd_ps(a), _mm_cvtpd_ps(b), (1 << 2) | (1 << 6)); -} - -template<> EIGEN_STRONG_INLINE Packet2d pcast(const Packet4f& a) { - // Simply discard the second half of the input - return _mm_cvtps_pd(a); -} - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TYPE_CASTING_SSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h deleted file mode 100644 index 4153b877cff..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/AssignmentFunctors.h +++ /dev/null @@ -1,168 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_ASSIGNMENT_FUNCTORS_H -#define EIGEN_ASSIGNMENT_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -/** \internal - * \brief Template functor for scalar/packet assignment - * - */ -template struct assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a = b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,b); } -}; - -// Empty overload for void type (used by PermutationMatrix) -template struct assign_op {}; - -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost, - PacketAccess = is_same::value && packet_traits::Vectorizable && packet_traits::Vectorizable - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with addition - * - */ -template struct add_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(add_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a += b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::padd(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::AddCost, - PacketAccess = is_same::value && packet_traits::HasAdd - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with subtraction - * - */ -template struct sub_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(sub_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a -= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::psub(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::AddCost, - PacketAccess = is_same::value && packet_traits::HasSub - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with multiplication - * - */ -template -struct mul_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(mul_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a *= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::pmul(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::MulCost, - PacketAccess = is_same::value && packet_traits::HasMul - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with diviving - * - */ -template struct div_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(div_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(DstScalar& a, const SrcScalar& b) const { a /= b; } - - template - EIGEN_STRONG_INLINE void assignPacket(DstScalar* a, const Packet& b) const - { internal::pstoret(a,internal::pdiv(internal::ploadt(a),b)); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::ReadCost + NumTraits::MulCost, - PacketAccess = is_same::value && packet_traits::HasDiv - }; -}; - -/** \internal - * \brief Template functor for scalar/packet assignment with swapping - * - * It works as follow. For a non-vectorized evaluation loop, we have: - * for(i) func(A.coeffRef(i), B.coeff(i)); - * where B is a SwapWrapper expression. The trick is to make SwapWrapper::coeff behaves like a non-const coeffRef. - * Actually, SwapWrapper might not even be needed since even if B is a plain expression, since it has to be writable - * B.coeff already returns a const reference to the underlying scalar value. - * - * The case of a vectorized loop is more tricky: - * for(i,j) func.assignPacket(&A.coeffRef(i,j), B.packet(i,j)); - * Here, B must be a SwapWrapper whose packet function actually returns a proxy object holding a Scalar*, - * the actual alignment and Packet type. - * - */ -template struct swap_assign_op { - - EIGEN_EMPTY_STRUCT_CTOR(swap_assign_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Scalar& a, const Scalar& b) const - { -#ifdef __CUDACC__ - // FIXME is there some kind of cuda::swap? - Scalar t=b; const_cast(b)=a; a=t; -#else - using std::swap; - swap(a,const_cast(b)); -#endif - } -}; -template -struct functor_traits > { - enum { - Cost = 3 * NumTraits::ReadCost, - PacketAccess = packet_traits::Vectorizable - }; -}; - -} // namespace internal - -} // namespace Eigen - -#endif // EIGEN_ASSIGNMENT_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h deleted file mode 100644 index 3eae6b8cada..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/BinaryFunctors.h +++ /dev/null @@ -1,475 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BINARY_FUNCTORS_H -#define EIGEN_BINARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -//---------- associative binary functors ---------- - -template -struct binary_op_base -{ - typedef Arg1 first_argument_type; - typedef Arg2 second_argument_type; -}; - -/** \internal - * \brief Template functor to compute the sum of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::operator+, class VectorwiseOp, DenseBase::sum() - */ -template -struct scalar_sum_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_sum_op) -#else - scalar_sum_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a + b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::padd(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, // rough estimate! - PacketAccess = is_same::value && packet_traits::HasAdd && packet_traits::HasAdd - // TODO vectorize mixed sum - }; -}; - -/** \internal - * \brief Template specialization to deprecate the summation of boolean expressions. - * This is required to solve Bug 426. - * \sa DenseBase::count(), DenseBase::any(), ArrayBase::cast(), MatrixBase::cast() - */ -template<> struct scalar_sum_op : scalar_sum_op { - EIGEN_DEPRECATED - scalar_sum_op() {} -}; - - -/** \internal - * \brief Template functor to compute the product of two scalars - * - * \sa class CwiseBinaryOp, Cwise::operator*(), class VectorwiseOp, MatrixBase::redux() - */ -template -struct scalar_product_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_product_op) -#else - scalar_product_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a * b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmul(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_mul(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::MulCost + NumTraits::MulCost)/2, // rough estimate! - PacketAccess = is_same::value && packet_traits::HasMul && packet_traits::HasMul - // TODO vectorize mixed product - }; -}; - -/** \internal - * \brief Template functor to compute the conjugate product of two scalars - * - * This is a short cut for conj(x) * y which is needed for optimization purpose; in Eigen2 support mode, this becomes x * conj(y) - */ -template -struct scalar_conj_product_op : binary_op_base -{ - - enum { - Conj = NumTraits::IsComplex - }; - - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - - EIGEN_EMPTY_STRUCT_CTOR(scalar_conj_product_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const - { return conj_helper().pmul(a,b); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return conj_helper().pmul(a,b); } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::MulCost, - PacketAccess = internal::is_same::value && packet_traits::HasMul - }; -}; - -/** \internal - * \brief Template functor to compute the min of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::cwiseMin, class VectorwiseOp, MatrixBase::minCoeff() - */ -template -struct scalar_min_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_min_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return numext::mini(a, b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmin(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_min(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = internal::is_same::value && packet_traits::HasMin - }; -}; - -/** \internal - * \brief Template functor to compute the max of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::cwiseMax, class VectorwiseOp, MatrixBase::maxCoeff() - */ -template -struct scalar_max_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_max_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return numext::maxi(a, b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pmax(a,b); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type predux(const Packet& a) const - { return internal::predux_max(a); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = internal::is_same::value && packet_traits::HasMax - }; -}; - -/** \internal - * \brief Template functors for comparison of two scalars - * \todo Implement packet-comparisons - */ -template struct scalar_cmp_op; - -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = false - }; -}; - -template -struct result_of(LhsScalar,RhsScalar)> { - typedef bool type; -}; - - -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a==b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a<=b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a>b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a>=b;} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return !(a<=b || b<=a);} -}; -template -struct scalar_cmp_op : binary_op_base -{ - typedef bool result_type; - EIGEN_EMPTY_STRUCT_CTOR(scalar_cmp_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator()(const LhsScalar& a, const RhsScalar& b) const {return a!=b;} -}; - - -/** \internal - * \brief Template functor to compute the hypot of two \b positive \b and \b real scalars - * - * \sa MatrixBase::stableNorm(), class Redux - */ -template -struct scalar_hypot_op : binary_op_base -{ - EIGEN_EMPTY_STRUCT_CTOR(scalar_hypot_op) - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar &x, const Scalar &y) const - { - // This functor is used by hypotNorm only for which it is faster to first apply abs - // on all coefficients prior to reduction through hypot. - // This way we avoid calling abs on positive and real entries, and this also permits - // to seamlessly handle complexes. Otherwise we would have to handle both real and complexes - // through the same functor... - return internal::positive_real_hypot(x,y); - } -}; -template -struct functor_traits > { - enum - { - Cost = 3 * NumTraits::AddCost + - 2 * NumTraits::MulCost + - 2 * scalar_div_cost::value, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the pow of two scalars - */ -template -struct scalar_pow_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_pow_op) -#else - scalar_pow_op() { - typedef Scalar LhsScalar; - typedef Exponent RhsScalar; - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC - inline result_type operator() (const Scalar& a, const Exponent& b) const { return numext::pow(a, b); } -}; -template -struct functor_traits > { - enum { Cost = 5 * NumTraits::MulCost, PacketAccess = false }; -}; - - - -//---------- non associative binary functors ---------- - -/** \internal - * \brief Template functor to compute the difference of two scalars - * - * \sa class CwiseBinaryOp, MatrixBase::operator- - */ -template -struct scalar_difference_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_difference_op) -#else - scalar_difference_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a - b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::psub(a,b); } -}; -template -struct functor_traits > { - enum { - Cost = (NumTraits::AddCost+NumTraits::AddCost)/2, - PacketAccess = is_same::value && packet_traits::HasSub && packet_traits::HasSub - }; -}; - -/** \internal - * \brief Template functor to compute the quotient of two scalars - * - * \sa class CwiseBinaryOp, Cwise::operator/() - */ -template -struct scalar_quotient_op : binary_op_base -{ - typedef typename ScalarBinaryOpTraits::ReturnType result_type; -#ifndef EIGEN_SCALAR_BINARY_OP_PLUGIN - EIGEN_EMPTY_STRUCT_CTOR(scalar_quotient_op) -#else - scalar_quotient_op() { - EIGEN_SCALAR_BINARY_OP_PLUGIN - } -#endif - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const LhsScalar& a, const RhsScalar& b) const { return a / b; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a, const Packet& b) const - { return internal::pdiv(a,b); } -}; -template -struct functor_traits > { - typedef typename scalar_quotient_op::result_type result_type; - enum { - PacketAccess = is_same::value && packet_traits::HasDiv && packet_traits::HasDiv, - Cost = scalar_div_cost::value - }; -}; - - - -/** \internal - * \brief Template functor to compute the and of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator&& - */ -struct scalar_boolean_and_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_and_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a && b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the or of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator|| - */ -struct scalar_boolean_or_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_or_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a || b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the xor of two booleans - * - * \sa class CwiseBinaryOp, ArrayBase::operator^ - */ -struct scalar_boolean_xor_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_xor_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a, const bool& b) const { return a ^ b; } -}; -template<> struct functor_traits { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - - - -//---------- binary functors bound to a constant, thus appearing as a unary functor ---------- - -// The following two classes permits to turn any binary functor into a unary one with one argument bound to a constant value. -// They are analogues to std::binder1st/binder2nd but with the following differences: -// - they are compatible with packetOp -// - they are portable across C++ versions (the std::binder* are deprecated in C++11) -template struct bind1st_op : BinaryOp { - - typedef typename BinaryOp::first_argument_type first_argument_type; - typedef typename BinaryOp::second_argument_type second_argument_type; - typedef typename BinaryOp::result_type result_type; - - bind1st_op(const first_argument_type &val) : m_value(val) {} - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const second_argument_type& b) const { return BinaryOp::operator()(m_value,b); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& b) const - { return BinaryOp::packetOp(internal::pset1(m_value), b); } - - first_argument_type m_value; -}; -template struct functor_traits > : functor_traits {}; - - -template struct bind2nd_op : BinaryOp { - - typedef typename BinaryOp::first_argument_type first_argument_type; - typedef typename BinaryOp::second_argument_type second_argument_type; - typedef typename BinaryOp::result_type result_type; - - bind2nd_op(const second_argument_type &val) : m_value(val) {} - - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const first_argument_type& a) const { return BinaryOp::operator()(a,m_value); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return BinaryOp::packetOp(a,internal::pset1(m_value)); } - - second_argument_type m_value; -}; -template struct functor_traits > : functor_traits {}; - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BINARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h deleted file mode 100644 index b03be0269c9..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/NullaryFunctors.h +++ /dev/null @@ -1,188 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_NULLARY_FUNCTORS_H -#define EIGEN_NULLARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -template -struct scalar_constant_op { - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE scalar_constant_op(const scalar_constant_op& other) : m_other(other.m_other) { } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE scalar_constant_op(const Scalar& other) : m_other(other) { } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() () const { return m_other; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const PacketType packetOp() const { return internal::pset1(m_other); } - const Scalar m_other; -}; -template -struct functor_traits > -{ enum { Cost = 0 /* as the constant value should be loaded in register only once for the whole expression */, - PacketAccess = packet_traits::Vectorizable, IsRepeatable = true }; }; - -template struct scalar_identity_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_identity_op) - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType row, IndexType col) const { return row==col ? Scalar(1) : Scalar(0); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false, IsRepeatable = true }; }; - -template struct linspaced_op_impl; - -template -struct linspaced_op_impl -{ - linspaced_op_impl(const Scalar& low, const Scalar& high, Index num_steps) : - m_low(low), m_high(high), m_size1(num_steps==1 ? 1 : num_steps-1), m_step(num_steps==1 ? Scalar() : (high-low)/Scalar(num_steps-1)), - m_flip(numext::abs(high) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType i) const { - typedef typename NumTraits::Real RealScalar; - if(m_flip) - return (i==0)? m_low : (m_high - RealScalar(m_size1-i)*m_step); - else - return (i==m_size1)? m_high : (m_low + RealScalar(i)*m_step); - } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(IndexType i) const - { - // Principle: - // [low, ..., low] + ( [step, ..., step] * ( [i, ..., i] + [0, ..., size] ) ) - if(m_flip) - { - Packet pi = plset(Scalar(i-m_size1)); - Packet res = padd(pset1(m_high), pmul(pset1(m_step), pi)); - if(i==0) - res = pinsertfirst(res, m_low); - return res; - } - else - { - Packet pi = plset(Scalar(i)); - Packet res = padd(pset1(m_low), pmul(pset1(m_step), pi)); - if(i==m_size1-unpacket_traits::size+1) - res = pinsertlast(res, m_high); - return res; - } - } - - const Scalar m_low; - const Scalar m_high; - const Index m_size1; - const Scalar m_step; - const bool m_flip; -}; - -template -struct linspaced_op_impl -{ - linspaced_op_impl(const Scalar& low, const Scalar& high, Index num_steps) : - m_low(low), - m_multiplier((high-low)/convert_index(num_steps<=1 ? 1 : num_steps-1)), - m_divisor(convert_index((high>=low?num_steps:-num_steps)+(high-low))/((numext::abs(high-low)+1)==0?1:(numext::abs(high-low)+1))), - m_use_divisor(num_steps>1 && (numext::abs(high-low)+1) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE - const Scalar operator() (IndexType i) const - { - if(m_use_divisor) return m_low + convert_index(i)/m_divisor; - else return m_low + convert_index(i)*m_multiplier; - } - - const Scalar m_low; - const Scalar m_multiplier; - const Scalar m_divisor; - const bool m_use_divisor; -}; - -// ----- Linspace functor ---------------------------------------------------------------- - -// Forward declaration (we default to random access which does not really give -// us a speed gain when using packet access but it allows to use the functor in -// nested expressions). -template struct linspaced_op; -template struct functor_traits< linspaced_op > -{ - enum - { - Cost = 1, - PacketAccess = (!NumTraits::IsInteger) && packet_traits::HasSetLinear && packet_traits::HasBlend, - /*&& ((!NumTraits::IsInteger) || packet_traits::HasDiv),*/ // <- vectorization for integer is currently disabled - IsRepeatable = true - }; -}; -template struct linspaced_op -{ - linspaced_op(const Scalar& low, const Scalar& high, Index num_steps) - : impl((num_steps==1 ? high : low),high,num_steps) - {} - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (IndexType i) const { return impl(i); } - - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(IndexType i) const { return impl.packetOp(i); } - - // This proxy object handles the actual required temporaries and the different - // implementations (integer vs. floating point). - const linspaced_op_impl::IsInteger> impl; -}; - -// Linear access is automatically determined from the operator() prototypes available for the given functor. -// If it exposes an operator()(i,j), then we assume the i and j coefficients are required independently -// and linear access is not possible. In all other cases, linear access is enabled. -// Users should not have to deal with this structure. -template struct functor_has_linear_access { enum { ret = !has_binary_operator::value }; }; - -// For unreliable compilers, let's specialize the has_*ary_operator -// helpers so that at least built-in nullary functors work fine. -#if !( (EIGEN_COMP_MSVC>1600) || (EIGEN_GNUC_AT_LEAST(4,8)) || (EIGEN_COMP_ICC>=1600)) -template -struct has_nullary_operator,IndexType> { enum { value = 1}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 0}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 1}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 0}; }; -template -struct has_unary_operator,IndexType> { enum { value = 1}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; - -template -struct has_nullary_operator,IndexType> { enum { value = 1}; }; -template -struct has_unary_operator,IndexType> { enum { value = 0}; }; -template -struct has_binary_operator,IndexType> { enum { value = 0}; }; -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_NULLARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h deleted file mode 100644 index 9c1d75850be..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/StlFunctors.h +++ /dev/null @@ -1,136 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STL_FUNCTORS_H -#define EIGEN_STL_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -// default functor traits for STL functors: - -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = NumTraits::AddCost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 1, PacketAccess = false }; }; - -#if (__cplusplus < 201103L) && (EIGEN_COMP_MSVC <= 1900) -// std::binder* are deprecated since c++11 and will be removed in c++17 -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost, PacketAccess = false }; }; -#endif - -#if (__cplusplus < 201703L) && (EIGEN_COMP_MSVC < 1910) -// std::unary_negate is deprecated since c++17 and will be removed in c++20 -template -struct functor_traits > -{ enum { Cost = 1 + functor_traits::Cost, PacketAccess = false }; }; - -// std::binary_negate is deprecated since c++17 and will be removed in c++20 -template -struct functor_traits > -{ enum { Cost = 1 + functor_traits::Cost, PacketAccess = false }; }; -#endif - -#ifdef EIGEN_STDEXT_SUPPORT - -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > > -{ enum { Cost = 0, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost + functor_traits::Cost, PacketAccess = false }; }; - -template -struct functor_traits > -{ enum { Cost = functor_traits::Cost + functor_traits::Cost + functor_traits::Cost, PacketAccess = false }; }; - -#endif // EIGEN_STDEXT_SUPPORT - -// allow to add new functors and specializations of functor_traits from outside Eigen. -// this macro is really needed because functor_traits must be specialized after it is declared but before it is used... -#ifdef EIGEN_FUNCTORS_PLUGIN -#include EIGEN_FUNCTORS_PLUGIN -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_STL_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h deleted file mode 100644 index b254e96c6ad..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/TernaryFunctors.h +++ /dev/null @@ -1,25 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2016 Eugene Brevdo -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TERNARY_FUNCTORS_H -#define EIGEN_TERNARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -//---------- associative ternary functors ---------- - - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TERNARY_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h deleted file mode 100644 index b56e7afd2c5..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/functors/UnaryFunctors.h +++ /dev/null @@ -1,792 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_UNARY_FUNCTORS_H -#define EIGEN_UNARY_FUNCTORS_H - -namespace Eigen { - -namespace internal { - -/** \internal - * \brief Template functor to compute the opposite of a scalar - * - * \sa class CwiseUnaryOp, MatrixBase::operator- - */ -template struct scalar_opposite_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_opposite_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return -a; } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pnegate(a); } -}; -template -struct functor_traits > -{ enum { - Cost = NumTraits::AddCost, - PacketAccess = packet_traits::HasNegate }; -}; - -/** \internal - * \brief Template functor to compute the absolute value of a scalar - * - * \sa class CwiseUnaryOp, Cwise::abs - */ -template struct scalar_abs_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_abs_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pabs(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::AddCost, - PacketAccess = packet_traits::HasAbs - }; -}; - -/** \internal - * \brief Template functor to compute the score of a scalar, to chose a pivot - * - * \sa class CwiseUnaryOp - */ -template struct scalar_score_coeff_op : scalar_abs_op -{ - typedef void Score_is_abs; -}; -template -struct functor_traits > : functor_traits > {}; - -/* Avoid recomputing abs when we know the score and they are the same. Not a true Eigen functor. */ -template struct abs_knowing_score -{ - EIGEN_EMPTY_STRUCT_CTOR(abs_knowing_score) - typedef typename NumTraits::Real result_type; - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a, const Score&) const { return numext::abs(a); } -}; -template struct abs_knowing_score::Score_is_abs> -{ - EIGEN_EMPTY_STRUCT_CTOR(abs_knowing_score) - typedef typename NumTraits::Real result_type; - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scal&, const result_type& a) const { return a; } -}; - -/** \internal - * \brief Template functor to compute the squared absolute value of a scalar - * - * \sa class CwiseUnaryOp, Cwise::abs2 - */ -template struct scalar_abs2_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_abs2_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { return numext::abs2(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::pmul(a,a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasAbs2 }; }; - -/** \internal - * \brief Template functor to compute the conjugate of a complex value - * - * \sa class CwiseUnaryOp, MatrixBase::conjugate() - */ -template struct scalar_conjugate_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_conjugate_op) - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { using numext::conj; return conj(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const { return internal::pconj(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::IsComplex ? NumTraits::AddCost : 0, - PacketAccess = packet_traits::HasConj - }; -}; - -/** \internal - * \brief Template functor to compute the phase angle of a complex - * - * \sa class CwiseUnaryOp, Cwise::arg - */ -template struct scalar_arg_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_arg_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const result_type operator() (const Scalar& a) const { using numext::arg; return arg(a); } - template - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Packet packetOp(const Packet& a) const - { return internal::parg(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::IsComplex ? 5 * NumTraits::MulCost : NumTraits::AddCost, - PacketAccess = packet_traits::HasArg - }; -}; -/** \internal - * \brief Template functor to cast a scalar to another type - * - * \sa class CwiseUnaryOp, MatrixBase::cast() - */ -template -struct scalar_cast_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cast_op) - typedef NewType result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const NewType operator() (const Scalar& a) const { return cast(a); } -}; -template -struct functor_traits > -{ enum { Cost = is_same::value ? 0 : NumTraits::AddCost, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the real part of a complex - * - * \sa class CwiseUnaryOp, MatrixBase::real() - */ -template -struct scalar_real_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_real_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::real(a); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the imaginary part of a complex - * - * \sa class CwiseUnaryOp, MatrixBase::imag() - */ -template -struct scalar_imag_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return numext::imag(a); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the real part of a complex as a reference - * - * \sa class CwiseUnaryOp, MatrixBase::real() - */ -template -struct scalar_real_ref_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_real_ref_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::real_ref(*const_cast(&a)); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * \brief Template functor to extract the imaginary part of a complex as a reference - * - * \sa class CwiseUnaryOp, MatrixBase::imag() - */ -template -struct scalar_imag_ref_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_imag_ref_op) - typedef typename NumTraits::Real result_type; - EIGEN_DEVICE_FUNC - EIGEN_STRONG_INLINE result_type& operator() (const Scalar& a) const { return numext::imag_ref(*const_cast(&a)); } -}; -template -struct functor_traits > -{ enum { Cost = 0, PacketAccess = false }; }; - -/** \internal - * - * \brief Template functor to compute the exponential of a scalar - * - * \sa class CwiseUnaryOp, Cwise::exp() - */ -template struct scalar_exp_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_exp_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::exp(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pexp(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasExp, - // The following numbers are based on the AVX implementation. -#ifdef EIGEN_VECTORIZE_FMA - // Haswell can issue 2 add/mul/madd per cycle. - Cost = - (sizeof(Scalar) == 4 - // float: 8 pmadd, 4 pmul, 2 padd/psub, 6 other - ? (8 * NumTraits::AddCost + 6 * NumTraits::MulCost) - // double: 7 pmadd, 5 pmul, 3 padd/psub, 1 div, 13 other - : (14 * NumTraits::AddCost + - 6 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value)) -#else - Cost = - (sizeof(Scalar) == 4 - // float: 7 pmadd, 6 pmul, 4 padd/psub, 10 other - ? (21 * NumTraits::AddCost + 13 * NumTraits::MulCost) - // double: 7 pmadd, 5 pmul, 3 padd/psub, 1 div, 13 other - : (23 * NumTraits::AddCost + - 12 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value)) -#endif - }; -}; - -/** \internal - * - * \brief Template functor to compute the logarithm of a scalar - * - * \sa class CwiseUnaryOp, ArrayBase::log() - */ -template struct scalar_log_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::log(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasLog, - Cost = - (PacketAccess - // The following numbers are based on the AVX implementation. -#ifdef EIGEN_VECTORIZE_FMA - // 8 pmadd, 6 pmul, 8 padd/psub, 16 other, can issue 2 add/mul/madd per cycle. - ? (20 * NumTraits::AddCost + 7 * NumTraits::MulCost) -#else - // 8 pmadd, 6 pmul, 8 padd/psub, 20 other - ? (36 * NumTraits::AddCost + 14 * NumTraits::MulCost) -#endif - // Measured cost of std::log. - : sizeof(Scalar)==4 ? 40 : 85) - }; -}; - -/** \internal - * - * \brief Template functor to compute the logarithm of 1 plus a scalar value - * - * \sa class CwiseUnaryOp, ArrayBase::log1p() - */ -template struct scalar_log1p_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log1p_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::log1p(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog1p(a); } -}; -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasLog1p, - Cost = functor_traits >::Cost // TODO measure cost of log1p - }; -}; - -/** \internal - * - * \brief Template functor to compute the base-10 logarithm of a scalar - * - * \sa class CwiseUnaryOp, Cwise::log10() - */ -template struct scalar_log10_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_log10_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { EIGEN_USING_STD_MATH(log10) return log10(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::plog10(a); } -}; -template -struct functor_traits > -{ enum { Cost = 5 * NumTraits::MulCost, PacketAccess = packet_traits::HasLog10 }; }; - -/** \internal - * \brief Template functor to compute the square root of a scalar - * \sa class CwiseUnaryOp, Cwise::sqrt() - */ -template struct scalar_sqrt_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sqrt_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sqrt(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psqrt(a); } -}; -template -struct functor_traits > { - enum { -#if EIGEN_FAST_MATH - // The following numbers are based on the AVX implementation. - Cost = (sizeof(Scalar) == 8 ? 28 - // 4 pmul, 1 pmadd, 3 other - : (3 * NumTraits::AddCost + - 5 * NumTraits::MulCost)), -#else - // The following numbers are based on min VSQRT throughput on Haswell. - Cost = (sizeof(Scalar) == 8 ? 28 : 14), -#endif - PacketAccess = packet_traits::HasSqrt - }; -}; - -/** \internal - * \brief Template functor to compute the reciprocal square root of a scalar - * \sa class CwiseUnaryOp, Cwise::rsqrt() - */ -template struct scalar_rsqrt_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_rsqrt_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return Scalar(1)/numext::sqrt(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::prsqrt(a); } -}; - -template -struct functor_traits > -{ enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasRsqrt - }; -}; - -/** \internal - * \brief Template functor to compute the cosine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::cos() - */ -template struct scalar_cos_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cos_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return numext::cos(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pcos(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasCos - }; -}; - -/** \internal - * \brief Template functor to compute the sine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::sin() - */ -template struct scalar_sin_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sin_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sin(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psin(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasSin - }; -}; - - -/** \internal - * \brief Template functor to compute the tan of a scalar - * \sa class CwiseUnaryOp, ArrayBase::tan() - */ -template struct scalar_tan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_tan_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::tan(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::ptan(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasTan - }; -}; - -/** \internal - * \brief Template functor to compute the arc cosine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::acos() - */ -template struct scalar_acos_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_acos_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::acos(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pacos(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasACos - }; -}; - -/** \internal - * \brief Template functor to compute the arc sine of a scalar - * \sa class CwiseUnaryOp, ArrayBase::asin() - */ -template struct scalar_asin_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_asin_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::asin(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pasin(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasASin - }; -}; - - -/** \internal - * \brief Template functor to compute the atan of a scalar - * \sa class CwiseUnaryOp, ArrayBase::atan() - */ -template struct scalar_atan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_atan_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::atan(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::patan(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasATan - }; -}; - -/** \internal - * \brief Template functor to compute the tanh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::tanh() - */ -template -struct scalar_tanh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_tanh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator()(const Scalar& a) const { return numext::tanh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& x) const { return ptanh(x); } -}; - -template -struct functor_traits > { - enum { - PacketAccess = packet_traits::HasTanh, - Cost = ( (EIGEN_FAST_MATH && is_same::value) -// The following numbers are based on the AVX implementation, -#ifdef EIGEN_VECTORIZE_FMA - // Haswell can issue 2 add/mul/madd per cycle. - // 9 pmadd, 2 pmul, 1 div, 2 other - ? (2 * NumTraits::AddCost + - 6 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value) -#else - ? (11 * NumTraits::AddCost + - 11 * NumTraits::MulCost + - scalar_div_cost::HasDiv>::value) -#endif - // This number assumes a naive implementation of tanh - : (6 * NumTraits::AddCost + - 3 * NumTraits::MulCost + - 2 * scalar_div_cost::HasDiv>::value + - functor_traits >::Cost)) - }; -}; - -/** \internal - * \brief Template functor to compute the sinh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::sinh() - */ -template struct scalar_sinh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sinh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::sinh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psinh(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasSinh - }; -}; - -/** \internal - * \brief Template functor to compute the cosh of a scalar - * \sa class CwiseUnaryOp, ArrayBase::cosh() - */ -template struct scalar_cosh_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cosh_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const { return numext::cosh(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pcosh(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = 5 * NumTraits::MulCost, - PacketAccess = packet_traits::HasCosh - }; -}; - -/** \internal - * \brief Template functor to compute the inverse of a scalar - * \sa class CwiseUnaryOp, Cwise::inverse() - */ -template -struct scalar_inverse_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_inverse_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return Scalar(1)/a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pdiv(pset1(Scalar(1)),a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasDiv }; }; - -/** \internal - * \brief Template functor to compute the square of a scalar - * \sa class CwiseUnaryOp, Cwise::square() - */ -template -struct scalar_square_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_square_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return a*a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pmul(a,a); } -}; -template -struct functor_traits > -{ enum { Cost = NumTraits::MulCost, PacketAccess = packet_traits::HasMul }; }; - -/** \internal - * \brief Template functor to compute the cube of a scalar - * \sa class CwiseUnaryOp, Cwise::cube() - */ -template -struct scalar_cube_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_cube_op) - EIGEN_DEVICE_FUNC inline Scalar operator() (const Scalar& a) const { return a*a*a; } - template - EIGEN_DEVICE_FUNC inline const Packet packetOp(const Packet& a) const - { return internal::pmul(a,pmul(a,a)); } -}; -template -struct functor_traits > -{ enum { Cost = 2*NumTraits::MulCost, PacketAccess = packet_traits::HasMul }; }; - -/** \internal - * \brief Template functor to compute the rounded value of a scalar - * \sa class CwiseUnaryOp, ArrayBase::round() - */ -template struct scalar_round_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_round_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::round(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pround(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasRound - }; -}; - -/** \internal - * \brief Template functor to compute the floor of a scalar - * \sa class CwiseUnaryOp, ArrayBase::floor() - */ -template struct scalar_floor_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_floor_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::floor(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pfloor(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasFloor - }; -}; - -/** \internal - * \brief Template functor to compute the ceil of a scalar - * \sa class CwiseUnaryOp, ArrayBase::ceil() - */ -template struct scalar_ceil_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_ceil_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar operator() (const Scalar& a) const { return numext::ceil(a); } - template - EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::pceil(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = packet_traits::HasCeil - }; -}; - -/** \internal - * \brief Template functor to compute whether a scalar is NaN - * \sa class CwiseUnaryOp, ArrayBase::isnan() - */ -template struct scalar_isnan_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isnan_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isnan)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to check whether a scalar is +/-inf - * \sa class CwiseUnaryOp, ArrayBase::isinf() - */ -template struct scalar_isinf_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isinf_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isinf)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to check whether a scalar has a finite value - * \sa class CwiseUnaryOp, ArrayBase::isfinite() - */ -template struct scalar_isfinite_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_isfinite_op) - typedef bool result_type; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE result_type operator() (const Scalar& a) const { return (numext::isfinite)(a); } -}; -template -struct functor_traits > -{ - enum { - Cost = NumTraits::MulCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the logical not of a boolean - * - * \sa class CwiseUnaryOp, ArrayBase::operator! - */ -template struct scalar_boolean_not_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_boolean_not_op) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool operator() (const bool& a) const { return !a; } -}; -template -struct functor_traits > { - enum { - Cost = NumTraits::AddCost, - PacketAccess = false - }; -}; - -/** \internal - * \brief Template functor to compute the signum of a scalar - * \sa class CwiseUnaryOp, Cwise::sign() - */ -template::IsComplex!=0) > struct scalar_sign_op; -template -struct scalar_sign_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sign_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const - { - return Scalar( (a>Scalar(0)) - (a - //EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psign(a); } -}; -template -struct scalar_sign_op { - EIGEN_EMPTY_STRUCT_CTOR(scalar_sign_op) - EIGEN_DEVICE_FUNC inline const Scalar operator() (const Scalar& a) const - { - typedef typename NumTraits::Real real_type; - real_type aa = numext::abs(a); - if (aa==real_type(0)) - return Scalar(0); - aa = real_type(1)/aa; - return Scalar(a.real()*aa, a.imag()*aa ); - } - //TODO - //template - //EIGEN_DEVICE_FUNC inline Packet packetOp(const Packet& a) const { return internal::psign(a); } -}; -template -struct functor_traits > -{ enum { - Cost = - NumTraits::IsComplex - ? ( 8*NumTraits::MulCost ) // roughly - : ( 3*NumTraits::AddCost), - PacketAccess = packet_traits::HasSign - }; -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_FUNCTORS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h deleted file mode 100644 index 681451cc309..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ /dev/null @@ -1,2157 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_BLOCK_PANEL_H -#define EIGEN_GENERAL_BLOCK_PANEL_H - - -namespace Eigen { - -namespace internal { - -template -class gebp_traits; - - -/** \internal \returns b if a<=0, and returns a otherwise. */ -inline std::ptrdiff_t manage_caching_sizes_helper(std::ptrdiff_t a, std::ptrdiff_t b) -{ - return a<=0 ? b : a; -} - -#if EIGEN_ARCH_i386_OR_x86_64 -const std::ptrdiff_t defaultL1CacheSize = 32*1024; -const std::ptrdiff_t defaultL2CacheSize = 256*1024; -const std::ptrdiff_t defaultL3CacheSize = 2*1024*1024; -#else -const std::ptrdiff_t defaultL1CacheSize = 16*1024; -const std::ptrdiff_t defaultL2CacheSize = 512*1024; -const std::ptrdiff_t defaultL3CacheSize = 512*1024; -#endif - -/** \internal */ -struct CacheSizes { - CacheSizes(): m_l1(-1),m_l2(-1),m_l3(-1) { - int l1CacheSize, l2CacheSize, l3CacheSize; - queryCacheSizes(l1CacheSize, l2CacheSize, l3CacheSize); - m_l1 = manage_caching_sizes_helper(l1CacheSize, defaultL1CacheSize); - m_l2 = manage_caching_sizes_helper(l2CacheSize, defaultL2CacheSize); - m_l3 = manage_caching_sizes_helper(l3CacheSize, defaultL3CacheSize); - } - - std::ptrdiff_t m_l1; - std::ptrdiff_t m_l2; - std::ptrdiff_t m_l3; -}; - - -/** \internal */ -inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff_t* l2, std::ptrdiff_t* l3) -{ - static CacheSizes m_cacheSizes; - - if(action==SetAction) - { - // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1!=0 && l2!=0); - m_cacheSizes.m_l1 = *l1; - m_cacheSizes.m_l2 = *l2; - m_cacheSizes.m_l3 = *l3; - } - else if(action==GetAction) - { - eigen_internal_assert(l1!=0 && l2!=0); - *l1 = m_cacheSizes.m_l1; - *l2 = m_cacheSizes.m_l2; - *l3 = m_cacheSizes.m_l3; - } - else - { - eigen_internal_assert(false); - } -} - -/* Helper for computeProductBlockingSizes. - * - * Given a m x k times k x n matrix product of scalar types \c LhsScalar and \c RhsScalar, - * this function computes the blocking size parameters along the respective dimensions - * for matrix products and related algorithms. The blocking sizes depends on various - * parameters: - * - the L1 and L2 cache sizes, - * - the register level blocking sizes defined by gebp_traits, - * - the number of scalars that fit into a packet (when vectorization is enabled). - * - * \sa setCpuCacheSizes */ - -template -void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - typedef gebp_traits Traits; - - // Explanations: - // Let's recall that the product algorithms form mc x kc vertical panels A' on the lhs and - // kc x nc blocks B' on the rhs. B' has to fit into L2/L3 cache. Moreover, A' is processed - // per mr x kc horizontal small panels where mr is the blocking size along the m dimension - // at the register level. This small horizontal panel has to stay within L1 cache. - std::ptrdiff_t l1, l2, l3; - manage_caching_sizes(GetAction, &l1, &l2, &l3); - - if (num_threads > 1) { - typedef typename Traits::ResScalar ResScalar; - enum { - kdiv = KcFactor * (Traits::mr * sizeof(LhsScalar) + Traits::nr * sizeof(RhsScalar)), - ksub = Traits::mr * Traits::nr * sizeof(ResScalar), - kr = 8, - mr = Traits::mr, - nr = Traits::nr - }; - // Increasing k gives us more time to prefetch the content of the "C" - // registers. However once the latency is hidden there is no point in - // increasing the value of k, so we'll cap it at 320 (value determined - // experimentally). - // To avoid that k vanishes, we make k_cache at least as big as kr - const Index k_cache = numext::maxi(kr, (numext::mini)((l1-ksub)/kdiv, 320)); - if (k_cache < k) { - k = k_cache - (k_cache % kr); - eigen_internal_assert(k > 0); - } - - const Index n_cache = (l2-l1) / (nr * sizeof(RhsScalar) * k); - const Index n_per_thread = numext::div_ceil(n, num_threads); - if (n_cache <= n_per_thread) { - // Don't exceed the capacity of the l2 cache. - eigen_internal_assert(n_cache >= static_cast(nr)); - n = n_cache - (n_cache % nr); - eigen_internal_assert(n > 0); - } else { - n = (numext::mini)(n, (n_per_thread + nr - 1) - ((n_per_thread + nr - 1) % nr)); - } - - if (l3 > l2) { - // l3 is shared between all cores, so we'll give each thread its own chunk of l3. - const Index m_cache = (l3-l2) / (sizeof(LhsScalar) * k * num_threads); - const Index m_per_thread = numext::div_ceil(m, num_threads); - if(m_cache < m_per_thread && m_cache >= static_cast(mr)) { - m = m_cache - (m_cache % mr); - eigen_internal_assert(m > 0); - } else { - m = (numext::mini)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); - } - } - } - else { - // In unit tests we do not want to use extra large matrices, - // so we reduce the cache size to check the blocking strategy is not flawed -#ifdef EIGEN_DEBUG_SMALL_PRODUCT_BLOCKS - l1 = 9*1024; - l2 = 32*1024; - l3 = 512*1024; -#endif - - // Early return for small problems because the computation below are time consuming for small problems. - // Perhaps it would make more sense to consider k*n*m?? - // Note that for very tiny problem, this function should be bypassed anyway - // because we use the coefficient-based implementation for them. - if((numext::maxi)(k,(numext::maxi)(m,n))<48) - return; - - typedef typename Traits::ResScalar ResScalar; - enum { - k_peeling = 8, - k_div = KcFactor * (Traits::mr * sizeof(LhsScalar) + Traits::nr * sizeof(RhsScalar)), - k_sub = Traits::mr * Traits::nr * sizeof(ResScalar) - }; - - // ---- 1st level of blocking on L1, yields kc ---- - - // Blocking on the third dimension (i.e., k) is chosen so that an horizontal panel - // of size mr x kc of the lhs plus a vertical panel of kc x nr of the rhs both fits within L1 cache. - // We also include a register-level block of the result (mx x nr). - // (In an ideal world only the lhs panel would stay in L1) - // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi(((l1-k_sub)/k_div) & (~(k_peeling-1)),1); - const Index old_k = k; - if(k>max_kc) - { - // We are really blocking on the third dimension: - // -> reduce blocking size to make sure the last block is as large as possible - // while keeping the same number of sweeps over the result. - k = (k%max_kc)==0 ? max_kc - : max_kc - k_peeling * ((max_kc-1-(k%max_kc))/(k_peeling*(k/max_kc+1))); - - eigen_internal_assert(((old_k/k) == (old_k/max_kc)) && "the number of sweeps has to remain the same"); - } - - // ---- 2nd level of blocking on max(L2,L3), yields nc ---- - - // TODO find a reliable way to get the actual amount of cache per core to use for 2nd level blocking, that is: - // actual_l2 = max(l2, l3/nb_core_sharing_l3) - // The number below is quite conservative: it is better to underestimate the cache size rather than overestimating it) - // For instance, it corresponds to 6MB of L3 shared among 4 cores. - #ifdef EIGEN_DEBUG_SMALL_PRODUCT_BLOCKS - const Index actual_l2 = l3; - #else - const Index actual_l2 = 1572864; // == 1.5 MB - #endif - - // Here, nc is chosen such that a block of kc x nc of the rhs fit within half of L2. - // The second half is implicitly reserved to access the result and lhs coefficients. - // When k= Index(Traits::nr*sizeof(RhsScalar))*k) - { - // L1 blocking - max_nc = remaining_l1 / (k*sizeof(RhsScalar)); - } - else - { - // L2 blocking - max_nc = (3*actual_l2)/(2*2*max_kc*sizeof(RhsScalar)); - } - // WARNING Below, we assume that Traits::nr is a power of two. - Index nc = numext::mini(actual_l2/(2*k*sizeof(RhsScalar)), max_nc) & (~(Traits::nr-1)); - if(n>nc) - { - // We are really blocking over the columns: - // -> reduce blocking size to make sure the last block is as large as possible - // while keeping the same number of sweeps over the packed lhs. - // Here we allow one more sweep if this gives us a perfect match, thus the commented "-1" - n = (n%nc)==0 ? nc - : (nc - Traits::nr * ((nc/*-1*/-(n%nc))/(Traits::nr*(n/nc+1)))); - } - else if(old_k==k) - { - // So far, no blocking at all, i.e., kc==k, and nc==n. - // In this case, let's perform a blocking over the rows such that the packed lhs data is kept in cache L1/L2 - // TODO: part of this blocking strategy is now implemented within the kernel itself, so the L1-based heuristic here should be obsolete. - Index problem_size = k*n*sizeof(LhsScalar); - Index actual_lm = actual_l2; - Index max_mc = m; - if(problem_size<=1024) - { - // problem is small enough to keep in L1 - // Let's choose m such that lhs's block fit in 1/3 of L1 - actual_lm = l1; - } - else if(l3!=0 && problem_size<=32768) - { - // we have both L2 and L3, and problem is small enough to be kept in L2 - // Let's choose m such that lhs's block fit in 1/3 of L2 - actual_lm = l2; - max_mc = (numext::mini)(576,max_mc); - } - Index mc = (numext::mini)(actual_lm/(3*k*sizeof(LhsScalar)), max_mc); - if (mc > Traits::mr) mc -= mc % Traits::mr; - else if (mc==0) return; - m = (m%mc)==0 ? mc - : (mc - Traits::mr * ((mc/*-1*/-(m%mc))/(Traits::mr*(m/mc+1)))); - } - } -} - -template -inline bool useSpecificBlockingSizes(Index& k, Index& m, Index& n) -{ -#ifdef EIGEN_TEST_SPECIFIC_BLOCKING_SIZES - if (EIGEN_TEST_SPECIFIC_BLOCKING_SIZES) { - k = numext::mini(k, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_K); - m = numext::mini(m, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_M); - n = numext::mini(n, EIGEN_TEST_SPECIFIC_BLOCKING_SIZE_N); - return true; - } -#else - EIGEN_UNUSED_VARIABLE(k) - EIGEN_UNUSED_VARIABLE(m) - EIGEN_UNUSED_VARIABLE(n) -#endif - return false; -} - -/** \brief Computes the blocking parameters for a m x k times k x n matrix product - * - * \param[in,out] k Input: the third dimension of the product. Output: the blocking size along the same dimension. - * \param[in,out] m Input: the number of rows of the left hand side. Output: the blocking size along the same dimension. - * \param[in,out] n Input: the number of columns of the right hand side. Output: the blocking size along the same dimension. - * - * Given a m x k times k x n matrix product of scalar types \c LhsScalar and \c RhsScalar, - * this function computes the blocking size parameters along the respective dimensions - * for matrix products and related algorithms. - * - * The blocking size parameters may be evaluated: - * - either by a heuristic based on cache sizes; - * - or using fixed prescribed values (for testing purposes). - * - * \sa setCpuCacheSizes */ - -template -void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - if (!useSpecificBlockingSizes(k, m, n)) { - evaluateProductBlockingSizesHeuristic(k, m, n, num_threads); - } -} - -template -inline void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_threads = 1) -{ - computeProductBlockingSizes(k, m, n, num_threads); -} - -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_CJMADD - #define CJMADD(CJ,A,B,C,T) C = CJ.pmadd(A,B,C); -#else - - // FIXME (a bit overkill maybe ?) - - template struct gebp_madd_selector { - EIGEN_ALWAYS_INLINE static void run(const CJ& cj, A& a, B& b, C& c, T& /*t*/) - { - c = cj.pmadd(a,b,c); - } - }; - - template struct gebp_madd_selector { - EIGEN_ALWAYS_INLINE static void run(const CJ& cj, T& a, T& b, T& c, T& t) - { - t = b; t = cj.pmul(a,t); c = padd(c,t); - } - }; - - template - EIGEN_STRONG_INLINE void gebp_madd(const CJ& cj, A& a, B& b, C& c, T& t) - { - gebp_madd_selector::run(cj,a,b,c,t); - } - - #define CJMADD(CJ,A,B,C,T) gebp_madd(CJ,A,B,C,T); -// #define CJMADD(CJ,A,B,C,T) T = B; T = CJ.pmul(A,T); C = padd(C,T); -#endif - -/* Vectorization logic - * real*real: unpack rhs to constant packets, ... - * - * cd*cd : unpack rhs to (b_r,b_r), (b_i,b_i), mul to get (a_r b_r,a_i b_r) (a_r b_i,a_i b_i), - * storing each res packet into two packets (2x2), - * at the end combine them: swap the second and addsub them - * cf*cf : same but with 2x4 blocks - * cplx*real : unpack rhs to constant packets, ... - * real*cplx : load lhs as (a0,a0,a1,a1), and mul as usual - */ -template -class gebp_traits -{ -public: - typedef _LhsScalar LhsScalar; - typedef _RhsScalar RhsScalar; - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - - // register block size along the N direction must be 1 or 4 - nr = 4, - - // register block size along the M direction (currently, this one cannot be modified) - default_mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*LhsPacketSize, -#if defined(EIGEN_HAS_SINGLE_INSTRUCTION_MADD) && !defined(EIGEN_VECTORIZE_ALTIVEC) && !defined(EIGEN_VECTORIZE_VSX) - // we assume 16 registers - // See bug 992, if the scalar type is not vectorizable but that EIGEN_HAS_SINGLE_INSTRUCTION_MADD is defined, - // then using 3*LhsPacketSize triggers non-implemented paths in syrk. - mr = Vectorizable ? 3*LhsPacketSize : default_mr, -#else - mr = default_mr, -#endif - - LhsProgress = LhsPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// pbroadcast2(b, b0, b1); -// } - - template - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacketType& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - dest = ploadquad(b); - } - - template - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacketType& dest) const - { - dest = pload(a); - } - - template - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacketType& dest) const - { - dest = ploadu(a); - } - - template - EIGEN_STRONG_INLINE void madd(const LhsPacketType& a, const RhsPacketType& b, AccPacketType& c, AccPacketType& tmp) const - { - conj_helper cj; - // It would be a lot cleaner to call pmadd all the time. Unfortunately if we - // let gcc allocate the register in which to store the result of the pmul - // (in the case where there is no FMA) gcc fails to figure out how to avoid - // spilling register. -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c = cj.pmadd(a,b,c); -#else - tmp = b; tmp = cj.pmul(a,tmp); c = padd(c,tmp); -#endif - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = pmadd(c,alpha,r); - } - - template - EIGEN_STRONG_INLINE void acc(const ResPacketHalf& c, const ResPacketHalf& alpha, ResPacketHalf& r) const - { - r = pmadd(c,alpha,r); - } - -}; - -template -class gebp_traits, RealScalar, _ConjLhs, false> -{ -public: - typedef std::complex LhsScalar; - typedef RealScalar RhsScalar; - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = false, - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - nr = 4, -#if defined(EIGEN_HAS_SINGLE_INSTRUCTION_MADD) && !defined(EIGEN_VECTORIZE_ALTIVEC) && !defined(EIGEN_VECTORIZE_VSX) - // we assume 16 registers - mr = 3*LhsPacketSize, -#else - mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*LhsPacketSize, -#endif - - LhsProgress = LhsPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = pload(a); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploadu(a); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// pbroadcast2(b, b0, b1); -// } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp) const - { - madd_impl(a, b, c, tmp, typename conditional::type()); - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp, const true_type&) const - { -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c.v = pmadd(a.v,b,c.v); -#else - tmp = b; tmp = pmul(a.v,tmp); c.v = padd(c.v,tmp); -#endif - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsScalar& a, const RhsScalar& b, ResScalar& c, RhsScalar& /*tmp*/, const false_type&) const - { - c += a * b; - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = cj.pmadd(c,alpha,r); - } - -protected: - conj_helper cj; -}; - -template -struct DoublePacket -{ - Packet first; - Packet second; -}; - -template -DoublePacket padd(const DoublePacket &a, const DoublePacket &b) -{ - DoublePacket res; - res.first = padd(a.first, b.first); - res.second = padd(a.second,b.second); - return res; -} - -template -const DoublePacket& predux_downto4(const DoublePacket &a) -{ - return a; -} - -template struct unpacket_traits > { typedef DoublePacket half; }; -// template -// DoublePacket pmadd(const DoublePacket &a, const DoublePacket &b) -// { -// DoublePacket res; -// res.first = padd(a.first, b.first); -// res.second = padd(a.second,b.second); -// return res; -// } - -template -class gebp_traits, std::complex, _ConjLhs, _ConjRhs > -{ -public: - typedef std::complex Scalar; - typedef std::complex LhsScalar; - typedef std::complex RhsScalar; - typedef std::complex ResScalar; - - enum { - ConjLhs = _ConjLhs, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable - && packet_traits::Vectorizable, - RealPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - - // FIXME: should depend on NumberOfRegisters - nr = 4, - mr = ResPacketSize, - - LhsProgress = ResPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type RealPacket; - typedef typename packet_traits::type ScalarPacket; - typedef DoublePacket DoublePacketType; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - typedef typename conditional::type AccPacket; - - EIGEN_STRONG_INLINE void initAcc(Scalar& p) { p = Scalar(0); } - - EIGEN_STRONG_INLINE void initAcc(DoublePacketType& p) - { - p.first = pset1(RealScalar(0)); - p.second = pset1(RealScalar(0)); - } - - // Scalar path - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, ResPacket& dest) const - { - dest = pset1(*b); - } - - // Vectorized path - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, DoublePacketType& dest) const - { - dest.first = pset1(numext::real(*b)); - dest.second = pset1(numext::imag(*b)); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, ResPacket& dest) const - { - loadRhs(b,dest); - } - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, DoublePacketType& dest) const - { - eigen_internal_assert(unpacket_traits::size<=4); - loadRhs(b,dest); - } - - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - loadRhs(b+2, b2); - loadRhs(b+3, b3); - } - - // Vectorized path - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, DoublePacketType& b0, DoublePacketType& b1) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - } - - // Scalar path - EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsScalar& b0, RhsScalar& b1) - { - // FIXME not sure that's the best way to implement it! - loadRhs(b+0, b0); - loadRhs(b+1, b1); - } - - // nothing special here - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = pload((const typename unpacket_traits::type*)(a)); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploadu((const typename unpacket_traits::type*)(a)); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, DoublePacketType& c, RhsPacket& /*tmp*/) const - { - c.first = padd(pmul(a,b.first), c.first); - c.second = padd(pmul(a,b.second),c.second); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, ResPacket& c, RhsPacket& /*tmp*/) const - { - c = cj.pmadd(a,b,c); - } - - EIGEN_STRONG_INLINE void acc(const Scalar& c, const Scalar& alpha, Scalar& r) const { r += alpha * c; } - - EIGEN_STRONG_INLINE void acc(const DoublePacketType& c, const ResPacket& alpha, ResPacket& r) const - { - // assemble c - ResPacket tmp; - if((!ConjLhs)&&(!ConjRhs)) - { - tmp = pcplxflip(pconj(ResPacket(c.second))); - tmp = padd(ResPacket(c.first),tmp); - } - else if((!ConjLhs)&&(ConjRhs)) - { - tmp = pconj(pcplxflip(ResPacket(c.second))); - tmp = padd(ResPacket(c.first),tmp); - } - else if((ConjLhs)&&(!ConjRhs)) - { - tmp = pcplxflip(ResPacket(c.second)); - tmp = padd(pconj(ResPacket(c.first)),tmp); - } - else if((ConjLhs)&&(ConjRhs)) - { - tmp = pcplxflip(ResPacket(c.second)); - tmp = psub(pconj(ResPacket(c.first)),tmp); - } - - r = pmadd(tmp,alpha,r); - } - -protected: - conj_helper cj; -}; - -template -class gebp_traits, false, _ConjRhs > -{ -public: - typedef std::complex Scalar; - typedef RealScalar LhsScalar; - typedef Scalar RhsScalar; - typedef Scalar ResScalar; - - enum { - ConjLhs = false, - ConjRhs = _ConjRhs, - Vectorizable = packet_traits::Vectorizable - && packet_traits::Vectorizable, - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1, - - NumberOfRegisters = EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS, - // FIXME: should depend on NumberOfRegisters - nr = 4, - mr = (EIGEN_PLAIN_ENUM_MIN(16,NumberOfRegisters)/2/nr)*ResPacketSize, - - LhsProgress = ResPacketSize, - RhsProgress = 1 - }; - - typedef typename packet_traits::type _LhsPacket; - typedef typename packet_traits::type _RhsPacket; - typedef typename packet_traits::type _ResPacket; - - typedef typename conditional::type LhsPacket; - typedef typename conditional::type RhsPacket; - typedef typename conditional::type ResPacket; - - typedef ResPacket AccPacket; - - EIGEN_STRONG_INLINE void initAcc(AccPacket& p) - { - p = pset1(ResScalar(0)); - } - - EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, RhsPacket& dest) const - { - dest = pset1(*b); - } - - void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1, RhsPacket& b2, RhsPacket& b3) - { - pbroadcast4(b, b0, b1, b2, b3); - } - -// EIGEN_STRONG_INLINE void broadcastRhs(const RhsScalar* b, RhsPacket& b0, RhsPacket& b1) -// { -// // FIXME not sure that's the best way to implement it! -// b0 = pload1(b+0); -// b1 = pload1(b+1); -// } - - EIGEN_STRONG_INLINE void loadLhs(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploaddup(a); - } - - EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, RhsPacket& dest) const - { - eigen_internal_assert(unpacket_traits::size<=4); - loadRhs(b,dest); - } - - EIGEN_STRONG_INLINE void loadLhsUnaligned(const LhsScalar* a, LhsPacket& dest) const - { - dest = ploaddup(a); - } - - EIGEN_STRONG_INLINE void madd(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp) const - { - madd_impl(a, b, c, tmp, typename conditional::type()); - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsPacket& a, const RhsPacket& b, AccPacket& c, RhsPacket& tmp, const true_type&) const - { -#ifdef EIGEN_HAS_SINGLE_INSTRUCTION_MADD - EIGEN_UNUSED_VARIABLE(tmp); - c.v = pmadd(a,b.v,c.v); -#else - tmp = b; tmp.v = pmul(a,tmp.v); c = padd(c,tmp); -#endif - - } - - EIGEN_STRONG_INLINE void madd_impl(const LhsScalar& a, const RhsScalar& b, ResScalar& c, RhsScalar& /*tmp*/, const false_type&) const - { - c += a * b; - } - - EIGEN_STRONG_INLINE void acc(const AccPacket& c, const ResPacket& alpha, ResPacket& r) const - { - r = cj.pmadd(alpha,c,r); - } - -protected: - conj_helper cj; -}; - -/* optimized GEneral packed Block * packed Panel product kernel - * - * Mixing type logic: C += A * B - * | A | B | comments - * |real |cplx | no vectorization yet, would require to pack A with duplication - * |cplx |real | easy vectorization - */ -template -struct gebp_kernel -{ - typedef gebp_traits Traits; - typedef typename Traits::ResScalar ResScalar; - typedef typename Traits::LhsPacket LhsPacket; - typedef typename Traits::RhsPacket RhsPacket; - typedef typename Traits::ResPacket ResPacket; - typedef typename Traits::AccPacket AccPacket; - - typedef gebp_traits SwappedTraits; - typedef typename SwappedTraits::ResScalar SResScalar; - typedef typename SwappedTraits::LhsPacket SLhsPacket; - typedef typename SwappedTraits::RhsPacket SRhsPacket; - typedef typename SwappedTraits::ResPacket SResPacket; - typedef typename SwappedTraits::AccPacket SAccPacket; - - typedef typename DataMapper::LinearMapper LinearMapper; - - enum { - Vectorizable = Traits::Vectorizable, - LhsProgress = Traits::LhsProgress, - RhsProgress = Traits::RhsProgress, - ResPacketSize = Traits::ResPacketSize - }; - - EIGEN_DONT_INLINE - void operator()(const DataMapper& res, const LhsScalar* blockA, const RhsScalar* blockB, - Index rows, Index depth, Index cols, ResScalar alpha, - Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0); -}; - -template -EIGEN_DONT_INLINE -void gebp_kernel - ::operator()(const DataMapper& res, const LhsScalar* blockA, const RhsScalar* blockB, - Index rows, Index depth, Index cols, ResScalar alpha, - Index strideA, Index strideB, Index offsetA, Index offsetB) - { - Traits traits; - SwappedTraits straits; - - if(strideA==-1) strideA = depth; - if(strideB==-1) strideB = depth; - conj_helper cj; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - const Index peeled_mc3 = mr>=3*Traits::LhsProgress ? (rows/(3*LhsProgress))*(3*LhsProgress) : 0; - const Index peeled_mc2 = mr>=2*Traits::LhsProgress ? peeled_mc3+((rows-peeled_mc3)/(2*LhsProgress))*(2*LhsProgress) : 0; - const Index peeled_mc1 = mr>=1*Traits::LhsProgress ? (rows/(1*LhsProgress))*(1*LhsProgress) : 0; - enum { pk = 8 }; // NOTE Such a large peeling factor is important for large matrices (~ +5% when >1000 on Haswell) - const Index peeled_kc = depth & ~(pk-1); - const Index prefetch_res_offset = 32/sizeof(ResScalar); -// const Index depth2 = depth & ~1; - - //---------- Process 3 * LhsProgress rows at once ---------- - // This corresponds to 3*LhsProgress x nr register blocks. - // Usually, make sense only with FMA - if(mr>=3*Traits::LhsProgress) - { - // Here, the general idea is to loop on each largest micro horizontal panel of the lhs (3*Traits::LhsProgress x depth) - // and on each largest micro vertical panel of the rhs (depth * nr). - // Blocking sizes, i.e., 'depth' has been computed so that the micro horizontal panel of the lhs fit in L1. - // However, if depth is too small, we can extend the number of rows of these horizontal panels. - // This actual number of rows is computed as follow: - const Index l1 = defaultL1CacheSize; // in Bytes, TODO, l1 should be passed to this function. - // The max(1, ...) here is needed because we may be using blocking params larger than what our known l1 cache size - // suggests we should be using: either because our known l1 cache size is inaccurate (e.g. on Android, we can only guess), - // or because we are testing specific blocking sizes. - const Index actual_panel_rows = (3*LhsProgress) * std::max(1,( (l1 - sizeof(ResScalar)*mr*nr - depth*nr*sizeof(RhsScalar)) / (depth * sizeof(LhsScalar) * 3*LhsProgress) )); - for(Index i1=0; i1(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r0.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C8, alphav, R2); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r0.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r1.loadPacket(0 * Traits::ResPacketSize); - R1 = r1.loadPacket(1 * Traits::ResPacketSize); - R2 = r1.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C1, alphav, R0); - traits.acc(C5, alphav, R1); - traits.acc(C9, alphav, R2); - r1.storePacket(0 * Traits::ResPacketSize, R0); - r1.storePacket(1 * Traits::ResPacketSize, R1); - r1.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r2.loadPacket(1 * Traits::ResPacketSize); - R2 = r2.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C6, alphav, R1); - traits.acc(C10, alphav, R2); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r2.storePacket(1 * Traits::ResPacketSize, R1); - r2.storePacket(2 * Traits::ResPacketSize, R2); - - R0 = r3.loadPacket(0 * Traits::ResPacketSize); - R1 = r3.loadPacket(1 * Traits::ResPacketSize); - R2 = r3.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C3, alphav, R0); - traits.acc(C7, alphav, R1); - traits.acc(C11, alphav, R2); - r3.storePacket(0 * Traits::ResPacketSize, R0); - r3.storePacket(1 * Traits::ResPacketSize, R1); - r3.storePacket(2 * Traits::ResPacketSize, R2); - } - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r0.loadPacket(2 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C8, alphav, R2); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r0.storePacket(2 * Traits::ResPacketSize, R2); - } - } - } - } - - //---------- Process 2 * LhsProgress rows at once ---------- - if(mr>=2*Traits::LhsProgress) - { - const Index l1 = defaultL1CacheSize; // in Bytes, TODO, l1 should be passed to this function. - // The max(1, ...) here is needed because we may be using blocking params larger than what our known l1 cache size - // suggests we should be using: either because our known l1 cache size is inaccurate (e.g. on Android, we can only guess), - // or because we are testing specific blocking sizes. - Index actual_panel_rows = (2*LhsProgress) * std::max(1,( (l1 - sizeof(ResScalar)*mr*nr - depth*nr*sizeof(RhsScalar)) / (depth * sizeof(LhsScalar) * 2*LhsProgress) )); - - for(Index i1=peeled_mc3; i1=6 without FMA (bug 1637) - #if EIGEN_GNUC_AT_LEAST(6,0) && defined(EIGEN_VECTORIZE_SSE) - #define EIGEN_GEBP_2PX4_SPILLING_WORKAROUND __asm__ ("" : [a0] "+x,m" (A0),[a1] "+x,m" (A1)); - #else - #define EIGEN_GEBP_2PX4_SPILLING_WORKAROUND - #endif - #define EIGEN_GEBGP_ONESTEP(K) \ - do { \ - EIGEN_ASM_COMMENT("begin step of gebp micro kernel 2pX4"); \ - traits.loadLhs(&blA[(0+2*K)*LhsProgress], A0); \ - traits.loadLhs(&blA[(1+2*K)*LhsProgress], A1); \ - traits.broadcastRhs(&blB[(0+4*K)*RhsProgress], B_0, B1, B2, B3); \ - traits.madd(A0, B_0, C0, T0); \ - traits.madd(A1, B_0, C4, B_0); \ - traits.madd(A0, B1, C1, T0); \ - traits.madd(A1, B1, C5, B1); \ - traits.madd(A0, B2, C2, T0); \ - traits.madd(A1, B2, C6, B2); \ - traits.madd(A0, B3, C3, T0); \ - traits.madd(A1, B3, C7, B3); \ - EIGEN_GEBP_2PX4_SPILLING_WORKAROUND \ - EIGEN_ASM_COMMENT("end step of gebp micro kernel 2pX4"); \ - } while(false) - - internal::prefetch(blB+(48+0)); - EIGEN_GEBGP_ONESTEP(0); - EIGEN_GEBGP_ONESTEP(1); - EIGEN_GEBGP_ONESTEP(2); - EIGEN_GEBGP_ONESTEP(3); - internal::prefetch(blB+(48+16)); - EIGEN_GEBGP_ONESTEP(4); - EIGEN_GEBGP_ONESTEP(5); - EIGEN_GEBGP_ONESTEP(6); - EIGEN_GEBGP_ONESTEP(7); - - blB += pk*4*RhsProgress; - blA += pk*(2*Traits::LhsProgress); - - EIGEN_ASM_COMMENT("end gebp micro kernel 2pX4"); - } - // process remaining peeled loop - for(Index k=peeled_kc; k(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - R2 = r1.loadPacket(0 * Traits::ResPacketSize); - R3 = r1.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - traits.acc(C1, alphav, R2); - traits.acc(C5, alphav, R3); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - r1.storePacket(0 * Traits::ResPacketSize, R2); - r1.storePacket(1 * Traits::ResPacketSize, R3); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r2.loadPacket(1 * Traits::ResPacketSize); - R2 = r3.loadPacket(0 * Traits::ResPacketSize); - R3 = r3.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C6, alphav, R1); - traits.acc(C3, alphav, R2); - traits.acc(C7, alphav, R3); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r2.storePacket(1 * Traits::ResPacketSize, R1); - r3.storePacket(0 * Traits::ResPacketSize, R2); - r3.storePacket(1 * Traits::ResPacketSize, R3); - } - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r0.loadPacket(1 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C4, alphav, R1); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r0.storePacket(1 * Traits::ResPacketSize, R1); - } - } - } - } - //---------- Process 1 * LhsProgress rows at once ---------- - if(mr>=1*Traits::LhsProgress) - { - // loops on each largest micro horizontal panel of lhs (1*LhsProgress x depth) - for(Index i=peeled_mc2; i(alpha); - - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - R1 = r1.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - traits.acc(C1, alphav, R1); - r0.storePacket(0 * Traits::ResPacketSize, R0); - r1.storePacket(0 * Traits::ResPacketSize, R1); - - R0 = r2.loadPacket(0 * Traits::ResPacketSize); - R1 = r3.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C2, alphav, R0); - traits.acc(C3, alphav, R1); - r2.storePacket(0 * Traits::ResPacketSize, R0); - r3.storePacket(0 * Traits::ResPacketSize, R1); - } - - // Deal with remaining columns of the rhs - for(Index j2=packet_cols4; j2(alpha); - R0 = r0.loadPacket(0 * Traits::ResPacketSize); - traits.acc(C0, alphav, R0); - r0.storePacket(0 * Traits::ResPacketSize, R0); - } - } - } - //---------- Process remaining rows, 1 at once ---------- - if(peeled_mc1::half>::size; - if ((SwappedTraits::LhsProgress % 4) == 0 && - (SwappedTraits::LhsProgress <= 8) && - (SwappedTraits::LhsProgress!=8 || SResPacketHalfSize==nr)) - { - SAccPacket C0, C1, C2, C3; - straits.initAcc(C0); - straits.initAcc(C1); - straits.initAcc(C2); - straits.initAcc(C3); - - const Index spk = (std::max)(1,SwappedTraits::LhsProgress/4); - const Index endk = (depth/spk)*spk; - const Index endk4 = (depth/(spk*4))*(spk*4); - - Index k=0; - for(; k=8,typename unpacket_traits::half,SResPacket>::type SResPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SLhsPacket>::type SLhsPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SRhsPacket>::type SRhsPacketHalf; - typedef typename conditional=8,typename unpacket_traits::half,SAccPacket>::type SAccPacketHalf; - - SResPacketHalf R = res.template gatherPacket(i, j2); - SResPacketHalf alphav = pset1(alpha); - - if(depth-endk>0) - { - // We have to handle the last row of the rhs which corresponds to a half-packet - SLhsPacketHalf a0; - SRhsPacketHalf b0; - straits.loadLhsUnaligned(blB, a0); - straits.loadRhs(blA, b0); - SAccPacketHalf c0 = predux_downto4(C0); - straits.madd(a0,b0,c0,b0); - straits.acc(c0, alphav, R); - } - else - { - straits.acc(predux_downto4(C0), alphav, R); - } - res.scatterPacket(i, j2, R); - } - else - { - SResPacket R = res.template gatherPacket(i, j2); - SResPacket alphav = pset1(alpha); - straits.acc(C0, alphav, R); - res.scatterPacket(i, j2, R); - } - } - else // scalar path - { - // get a 1 x 4 res block as registers - ResScalar C0(0), C1(0), C2(0), C3(0); - - for(Index k=0; k -struct gemm_pack_lhs -{ - typedef typename DataMapper::LinearMapper LinearMapper; - EIGEN_DONT_INLINE void operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_lhs - ::operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride, Index offset) -{ - typedef typename packet_traits::type Packet; - enum { PacketSize = packet_traits::size }; - - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - eigen_assert( ((Pack1%PacketSize)==0 && Pack1<=4*PacketSize) || (Pack1<=4) ); - conj_if::IsComplex && Conjugate> cj; - Index count = 0; - - const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; - const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; - const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - const Index peeled_mc0 = Pack2>=1*PacketSize ? peeled_mc1 - : Pack2>1 ? (rows/Pack2)*Pack2 : 0; - - Index i=0; - - // Pack 3 packets - if(Pack1>=3*PacketSize) - { - for(; i=2*PacketSize) - { - for(; i=1*PacketSize) - { - for(; i1) - { - for(; i -struct gemm_pack_lhs -{ - typedef typename DataMapper::LinearMapper LinearMapper; - EIGEN_DONT_INLINE void operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_lhs - ::operator()(Scalar* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride, Index offset) -{ - typedef typename packet_traits::type Packet; - enum { PacketSize = packet_traits::size }; - - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK LHS"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index count = 0; - -// const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; -// const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; -// const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - - int pack = Pack1; - Index i = 0; - while(pack>0) - { - Index remaining_rows = rows-i; - Index peeled_mc = i+(remaining_rows/pack)*pack; - for(; i=PacketSize) - { - for(; k kernel; - for (int p = 0; p < PacketSize; ++p) kernel.packet[p] = lhs.loadPacket(i+p+m, k); - ptranspose(kernel); - for (int p = 0; p < PacketSize; ++p) pstore(blockA+count+m+(pack)*p, cj.pconj(kernel.packet[p])); - } - count += PacketSize*pack; - } - } - for(; k -struct gemm_pack_rhs -{ - typedef typename packet_traits::type Packet; - typedef typename DataMapper::LinearMapper LinearMapper; - enum { PacketSize = packet_traits::size }; - EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_rhs - ::operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride, Index offset) -{ - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS COLMAJOR"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - Index count = 0; - const Index peeled_k = (depth/PacketSize)*PacketSize; -// if(nr>=8) -// { -// for(Index j2=0; j2 kernel; -// for (int p = 0; p < PacketSize; ++p) { -// kernel.packet[p] = ploadu(&rhs[(j2+p)*rhsStride+k]); -// } -// ptranspose(kernel); -// for (int p = 0; p < PacketSize; ++p) { -// pstoreu(blockB+count, cj.pconj(kernel.packet[p])); -// count+=PacketSize; -// } -// } -// } -// for(; k=4) - { - for(Index j2=packet_cols8; j2 kernel; - kernel.packet[0] = dm0.loadPacket(k); - kernel.packet[1%PacketSize] = dm1.loadPacket(k); - kernel.packet[2%PacketSize] = dm2.loadPacket(k); - kernel.packet[3%PacketSize] = dm3.loadPacket(k); - ptranspose(kernel); - pstoreu(blockB+count+0*PacketSize, cj.pconj(kernel.packet[0])); - pstoreu(blockB+count+1*PacketSize, cj.pconj(kernel.packet[1%PacketSize])); - pstoreu(blockB+count+2*PacketSize, cj.pconj(kernel.packet[2%PacketSize])); - pstoreu(blockB+count+3*PacketSize, cj.pconj(kernel.packet[3%PacketSize])); - count+=4*PacketSize; - } - } - for(; k -struct gemm_pack_rhs -{ - typedef typename packet_traits::type Packet; - typedef typename DataMapper::LinearMapper LinearMapper; - enum { PacketSize = packet_traits::size }; - EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride=0, Index offset=0); -}; - -template -EIGEN_DONT_INLINE void gemm_pack_rhs - ::operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride, Index offset) -{ - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); - EIGEN_UNUSED_VARIABLE(stride); - EIGEN_UNUSED_VARIABLE(offset); - eigen_assert(((!PanelMode) && stride==0 && offset==0) || (PanelMode && stride>=depth && offset<=stride)); - conj_if::IsComplex && Conjugate> cj; - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - Index count = 0; - -// if(nr>=8) -// { -// for(Index j2=0; j2(&rhs[k*rhsStride + j2]); -// pstoreu(blockB+count, cj.pconj(A)); -// } else if (PacketSize==4) { -// Packet A = ploadu(&rhs[k*rhsStride + j2]); -// Packet B = ploadu(&rhs[k*rhsStride + j2 + PacketSize]); -// pstoreu(blockB+count, cj.pconj(A)); -// pstoreu(blockB+count+PacketSize, cj.pconj(B)); -// } else { -// const Scalar* b0 = &rhs[k*rhsStride + j2]; -// blockB[count+0] = cj(b0[0]); -// blockB[count+1] = cj(b0[1]); -// blockB[count+2] = cj(b0[2]); -// blockB[count+3] = cj(b0[3]); -// blockB[count+4] = cj(b0[4]); -// blockB[count+5] = cj(b0[5]); -// blockB[count+6] = cj(b0[6]); -// blockB[count+7] = cj(b0[7]); -// } -// count += 8; -// } -// // skip what we have after -// if(PanelMode) count += 8 * (stride-offset-depth); -// } -// } - if(nr>=4) - { - for(Index j2=packet_cols8; j2 -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_H -#define EIGEN_GENERAL_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -template class level3_blocking; - -/* Specialization for a row-major destination matrix => simple transposition of the product */ -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResInnerStride> -struct general_matrix_matrix_product -{ - typedef gebp_traits Traits; - - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, Index depth, - const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, - ResScalar* res, Index resIncr, Index resStride, - ResScalar alpha, - level3_blocking& blocking, - GemmParallelInfo* info = 0) - { - // transpose the product such that the result is column major - general_matrix_matrix_product - ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking,info); - } -}; - -/* Specialization for a col-major destination matrix - * => Blocking algorithm following Goto's paper */ -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResInnerStride> -struct general_matrix_matrix_product -{ - -typedef gebp_traits Traits; - -typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; -static void run(Index rows, Index cols, Index depth, - const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, - ResScalar alpha, - level3_blocking& blocking, - GemmParallelInfo* info = 0) -{ - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs, lhsStride); - RhsMapper rhs(_rhs, rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - Index nc = (std::min)(cols,blocking.nc()); // cache block size along the N direction - - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gebp_kernel gebp; - -#ifdef EIGEN_HAS_OPENMP - if(info) - { - // this is the parallel version! - int tid = omp_get_thread_num(); - int threads = omp_get_num_threads(); - - LhsScalar* blockA = blocking.blockA(); - eigen_internal_assert(blockA!=0); - - std::size_t sizeB = kc*nc; - ei_declare_aligned_stack_constructed_variable(RhsScalar, blockB, sizeB, 0); - - // For each horizontal panel of the rhs, and corresponding vertical panel of the lhs... - for(Index k=0; k rows of B', and cols of the A' - - // In order to reduce the chance that a thread has to wait for the other, - // let's start by packing B'. - pack_rhs(blockB, rhs.getSubMapper(k,0), actual_kc, nc); - - // Pack A_k to A' in a parallel fashion: - // each thread packs the sub block A_k,i to A'_i where i is the thread id. - - // However, before copying to A'_i, we have to make sure that no other thread is still using it, - // i.e., we test that info[tid].users equals 0. - // Then, we set info[tid].users to the number of threads to mark that all other threads are going to use it. - while(info[tid].users!=0) {} - info[tid].users += threads; - - pack_lhs(blockA+info[tid].lhs_start*actual_kc, lhs.getSubMapper(info[tid].lhs_start,k), actual_kc, info[tid].lhs_length); - - // Notify the other threads that the part A'_i is ready to go. - info[tid].sync = k; - - // Computes C_i += A' * B' per A'_i - for(int shift=0; shift0) { - while(info[i].sync!=k) { - } - } - - gebp(res.getSubMapper(info[i].lhs_start, 0), blockA+info[i].lhs_start*actual_kc, blockB, info[i].lhs_length, actual_kc, nc, alpha); - } - - // Then keep going as usual with the remaining B' - for(Index j=nc; j Pack lhs's panel into a sequential chunk of memory (L2/L3 caching) - // Note that this panel will be read as many times as the number of blocks in the rhs's - // horizontal panel which is, in practice, a very low number. - pack_lhs(blockA, lhs.getSubMapper(i2,k2), actual_kc, actual_mc); - - // For each kc x nc block of the rhs's horizontal panel... - for(Index j2=0; j2 -struct gemm_functor -{ - gemm_functor(const Lhs& lhs, const Rhs& rhs, Dest& dest, const Scalar& actualAlpha, BlockingType& blocking) - : m_lhs(lhs), m_rhs(rhs), m_dest(dest), m_actualAlpha(actualAlpha), m_blocking(blocking) - {} - - void initParallelSession(Index num_threads) const - { - m_blocking.initParallel(m_lhs.rows(), m_rhs.cols(), m_lhs.cols(), num_threads); - m_blocking.allocateA(); - } - - void operator() (Index row, Index rows, Index col=0, Index cols=-1, GemmParallelInfo* info=0) const - { - if(cols==-1) - cols = m_rhs.cols(); - - Gemm::run(rows, cols, m_lhs.cols(), - &m_lhs.coeffRef(row,0), m_lhs.outerStride(), - &m_rhs.coeffRef(0,col), m_rhs.outerStride(), - (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.innerStride(), m_dest.outerStride(), - m_actualAlpha, m_blocking, info); - } - - typedef typename Gemm::Traits Traits; - - protected: - const Lhs& m_lhs; - const Rhs& m_rhs; - Dest& m_dest; - Scalar m_actualAlpha; - BlockingType& m_blocking; -}; - -template class gemm_blocking_space; - -template -class level3_blocking -{ - typedef _LhsScalar LhsScalar; - typedef _RhsScalar RhsScalar; - - protected: - LhsScalar* m_blockA; - RhsScalar* m_blockB; - - Index m_mc; - Index m_nc; - Index m_kc; - - public: - - level3_blocking() - : m_blockA(0), m_blockB(0), m_mc(0), m_nc(0), m_kc(0) - {} - - inline Index mc() const { return m_mc; } - inline Index nc() const { return m_nc; } - inline Index kc() const { return m_kc; } - - inline LhsScalar* blockA() { return m_blockA; } - inline RhsScalar* blockB() { return m_blockB; } -}; - -template -class gemm_blocking_space - : public level3_blocking< - typename conditional::type, - typename conditional::type> -{ - enum { - Transpose = StorageOrder==RowMajor, - ActualRows = Transpose ? MaxCols : MaxRows, - ActualCols = Transpose ? MaxRows : MaxCols - }; - typedef typename conditional::type LhsScalar; - typedef typename conditional::type RhsScalar; - typedef gebp_traits Traits; - enum { - SizeA = ActualRows * MaxDepth, - SizeB = ActualCols * MaxDepth - }; - -#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES - EIGEN_ALIGN_MAX LhsScalar m_staticA[SizeA]; - EIGEN_ALIGN_MAX RhsScalar m_staticB[SizeB]; -#else - EIGEN_ALIGN_MAX char m_staticA[SizeA * sizeof(LhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1]; - EIGEN_ALIGN_MAX char m_staticB[SizeB * sizeof(RhsScalar) + EIGEN_DEFAULT_ALIGN_BYTES-1]; -#endif - - public: - - gemm_blocking_space(Index /*rows*/, Index /*cols*/, Index /*depth*/, Index /*num_threads*/, bool /*full_rows = false*/) - { - this->m_mc = ActualRows; - this->m_nc = ActualCols; - this->m_kc = MaxDepth; -#if EIGEN_MAX_STATIC_ALIGN_BYTES >= EIGEN_DEFAULT_ALIGN_BYTES - this->m_blockA = m_staticA; - this->m_blockB = m_staticB; -#else - this->m_blockA = reinterpret_cast((internal::UIntPtr(m_staticA) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1)); - this->m_blockB = reinterpret_cast((internal::UIntPtr(m_staticB) + (EIGEN_DEFAULT_ALIGN_BYTES-1)) & ~std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1)); -#endif - } - - void initParallel(Index, Index, Index, Index) - {} - - inline void allocateA() {} - inline void allocateB() {} - inline void allocateAll() {} -}; - -template -class gemm_blocking_space - : public level3_blocking< - typename conditional::type, - typename conditional::type> -{ - enum { - Transpose = StorageOrder==RowMajor - }; - typedef typename conditional::type LhsScalar; - typedef typename conditional::type RhsScalar; - typedef gebp_traits Traits; - - Index m_sizeA; - Index m_sizeB; - - public: - - gemm_blocking_space(Index rows, Index cols, Index depth, Index num_threads, bool l3_blocking) - { - this->m_mc = Transpose ? cols : rows; - this->m_nc = Transpose ? rows : cols; - this->m_kc = depth; - - if(l3_blocking) - { - computeProductBlockingSizes(this->m_kc, this->m_mc, this->m_nc, num_threads); - } - else // no l3 blocking - { - Index n = this->m_nc; - computeProductBlockingSizes(this->m_kc, this->m_mc, n, num_threads); - } - - m_sizeA = this->m_mc * this->m_kc; - m_sizeB = this->m_kc * this->m_nc; - } - - void initParallel(Index rows, Index cols, Index depth, Index num_threads) - { - this->m_mc = Transpose ? cols : rows; - this->m_nc = Transpose ? rows : cols; - this->m_kc = depth; - - eigen_internal_assert(this->m_blockA==0 && this->m_blockB==0); - Index m = this->m_mc; - computeProductBlockingSizes(this->m_kc, m, this->m_nc, num_threads); - m_sizeA = this->m_mc * this->m_kc; - m_sizeB = this->m_kc * this->m_nc; - } - - void allocateA() - { - if(this->m_blockA==0) - this->m_blockA = aligned_new(m_sizeA); - } - - void allocateB() - { - if(this->m_blockB==0) - this->m_blockB = aligned_new(m_sizeB); - } - - void allocateAll() - { - allocateA(); - allocateB(); - } - - ~gemm_blocking_space() - { - aligned_delete(this->m_blockA, m_sizeA); - aligned_delete(this->m_blockB, m_sizeB); - } -}; - -} // end namespace internal - -namespace internal { - -template -struct generic_product_impl - : generic_product_impl_base > -{ - typedef typename Product::Scalar Scalar; - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - enum { - MaxDepthAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(Lhs::MaxColsAtCompileTime,Rhs::MaxRowsAtCompileTime) - }; - - typedef generic_product_impl lazyproduct; - - template - static void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::assign_op()); - else - { - dst.setZero(); - scaleAndAddTo(dst, lhs, rhs, Scalar(1)); - } - } - - template - static void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::add_assign_op()); - else - scaleAndAddTo(dst,lhs, rhs, Scalar(1)); - } - - template - static void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) - { - if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::sub_assign_op()); - else - scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); - } - - template - static void scaleAndAddTo(Dest& dst, const Lhs& a_lhs, const Rhs& a_rhs, const Scalar& alpha) - { - eigen_assert(dst.rows()==a_lhs.rows() && dst.cols()==a_rhs.cols()); - if(a_lhs.cols()==0 || a_lhs.rows()==0 || a_rhs.cols()==0) - return; - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,LhsScalar,RhsScalar, - Dest::MaxRowsAtCompileTime,Dest::MaxColsAtCompileTime,MaxDepthAtCompileTime> BlockingType; - - typedef internal::gemm_functor< - Scalar, Index, - internal::general_matrix_matrix_product< - Index, - LhsScalar, (ActualLhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(LhsBlasTraits::NeedToConjugate), - RhsScalar, (ActualRhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(RhsBlasTraits::NeedToConjugate), - (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime>, - ActualLhsTypeCleaned, ActualRhsTypeCleaned, Dest, BlockingType> GemmFunctor; - - BlockingType blocking(dst.rows(), dst.cols(), lhs.cols(), 1, true); - internal::parallelize_gemm<(Dest::MaxRowsAtCompileTime>32 || Dest::MaxRowsAtCompileTime==Dynamic)> - (GemmFunctor(lhs, rhs, dst, actualAlpha, blocking), a_lhs.rows(), a_rhs.cols(), a_lhs.cols(), Dest::Flags&RowMajorBit); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h deleted file mode 100644 index d68d2f96572..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h +++ /dev/null @@ -1,317 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H -#define EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H - -namespace Eigen { - -template -struct selfadjoint_rank1_update; - -namespace internal { - -/********************************************************************** -* This file implements a general A * B product while -* evaluating only one triangular part of the product. -* This is a more general version of self adjoint product (C += A A^T) -* as the level 3 SYRK Blas routine. -**********************************************************************/ - -// forward declarations (defined at the end of this file) -template -struct tribb_kernel; - -/* Optimized matrix-matrix product evaluating only one triangular half */ -template -struct general_matrix_matrix_triangular_product; - -// as usual if the result is row major => we transpose the product -template -struct general_matrix_matrix_triangular_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resIncr, Index resStride, - const ResScalar& alpha, level3_blocking& blocking) - { - general_matrix_matrix_triangular_product - ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking); - } -}; - -template -struct general_matrix_matrix_triangular_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, - const ResScalar& alpha, level3_blocking& blocking) - { - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); - Index mc = (std::min)(size,blocking.mc()); - - // !!! mc must be a multiple of nr: - if(mc > Traits::nr) - mc = (mc/Traits::nr)*Traits::nr; - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*size; - - ei_declare_aligned_stack_constructed_variable(LhsScalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(RhsScalar, blockB, sizeB, blocking.blockB()); - - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gebp_kernel gebp; - tribb_kernel sybb; - - for(Index k2=0; k2 processed with gebp or skipped - // 2 - the actual_mc x actual_mc symmetric block => processed with a special kernel - // 3 - after the diagonal => processed with gebp or skipped - if (UpLo==Lower) - gebp(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, - (std::min)(size,i2), alpha, -1, -1, 0, 0); - - sybb(_res+resStride*i2 + resIncr*i2, resIncr, resStride, blockA, blockB + actual_kc*i2, actual_mc, actual_kc, alpha); - - if (UpLo==Upper) - { - Index j2 = i2+actual_mc; - gebp(res.getSubMapper(i2, j2), blockA, blockB+actual_kc*j2, actual_mc, - actual_kc, (std::max)(Index(0), size-j2), alpha, -1, -1, 0, 0); - } - } - } - } -}; - -// Optimized packed Block * packed Block product kernel evaluating only one given triangular part -// This kernel is built on top of the gebp kernel: -// - the current destination block is processed per panel of actual_mc x BlockSize -// where BlockSize is set to the minimal value allowing gebp to be as fast as possible -// - then, as usual, each panel is split into three parts along the diagonal, -// the sub blocks above and below the diagonal are processed as usual, -// while the triangular block overlapping the diagonal is evaluated into a -// small temporary buffer which is then accumulated into the result using a -// triangular traversal. -template -struct tribb_kernel -{ - typedef gebp_traits Traits; - typedef typename Traits::ResScalar ResScalar; - - enum { - BlockSize = meta_least_common_multiple::ret - }; - void operator()(ResScalar* _res, Index resIncr, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) - { - typedef blas_data_mapper ResMapper; - typedef blas_data_mapper BufferMapper; - ResMapper res(_res, resStride, resIncr); - gebp_kernel gebp_kernel1; - gebp_kernel gebp_kernel2; - - Matrix buffer((internal::constructor_without_unaligned_array_assert())); - - // let's process the block per panel of actual_mc x BlockSize, - // again, each is split into three parts, etc. - for (Index j=0; j(BlockSize,size - j); - const RhsScalar* actual_b = blockB+j*depth; - - if(UpLo==Upper) - gebp_kernel1(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - - // selfadjoint micro block - { - Index i = j; - buffer.setZero(); - // 1 - apply the kernel on the temporary buffer - gebp_kernel2(BufferMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - - // 2 - triangular accumulation - for(Index j1=0; j1 -struct general_product_to_triangular_selector; - - -template -struct general_product_to_triangular_selector -{ - static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha, bool beta) - { - typedef typename MatrixType::Scalar Scalar; - - typedef typename internal::remove_all::type Lhs; - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs; - typedef typename internal::remove_all::type _ActualLhs; - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(prod.lhs()); - - typedef typename internal::remove_all::type Rhs; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs; - typedef typename internal::remove_all::type _ActualRhs; - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(prod.rhs()); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived()); - - if(!beta) - mat.template triangularView().setZero(); - - enum { - StorageOrder = (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, - UseLhsDirectly = _ActualLhs::InnerStrideAtCompileTime==1, - UseRhsDirectly = _ActualRhs::InnerStrideAtCompileTime==1 - }; - - internal::gemv_static_vector_if static_lhs; - ei_declare_aligned_stack_constructed_variable(Scalar, actualLhsPtr, actualLhs.size(), - (UseLhsDirectly ? const_cast(actualLhs.data()) : static_lhs.data())); - if(!UseLhsDirectly) Map(actualLhsPtr, actualLhs.size()) = actualLhs; - - internal::gemv_static_vector_if static_rhs; - ei_declare_aligned_stack_constructed_variable(Scalar, actualRhsPtr, actualRhs.size(), - (UseRhsDirectly ? const_cast(actualRhs.data()) : static_rhs.data())); - if(!UseRhsDirectly) Map(actualRhsPtr, actualRhs.size()) = actualRhs; - - - selfadjoint_rank1_update::IsComplex, - RhsBlasTraits::NeedToConjugate && NumTraits::IsComplex> - ::run(actualLhs.size(), mat.data(), mat.outerStride(), actualLhsPtr, actualRhsPtr, actualAlpha); - } -}; - -template -struct general_product_to_triangular_selector -{ - static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha, bool beta) - { - typedef typename internal::remove_all::type Lhs; - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs; - typedef typename internal::remove_all::type _ActualLhs; - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(prod.lhs()); - - typedef typename internal::remove_all::type Rhs; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs; - typedef typename internal::remove_all::type _ActualRhs; - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(prod.rhs()); - - typename ProductType::Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs().derived()) * RhsBlasTraits::extractScalarFactor(prod.rhs().derived()); - - if(!beta) - mat.template triangularView().setZero(); - - enum { - IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0, - LhsIsRowMajor = _ActualLhs::Flags&RowMajorBit ? 1 : 0, - RhsIsRowMajor = _ActualRhs::Flags&RowMajorBit ? 1 : 0, - SkipDiag = (UpLo&(UnitDiag|ZeroDiag))!=0 - }; - - Index size = mat.cols(); - if(SkipDiag) - size--; - Index depth = actualLhs.cols(); - - typedef internal::gemm_blocking_space BlockingType; - - BlockingType blocking(size, size, depth, 1, false); - - internal::general_matrix_matrix_triangular_product - ::run(size, depth, - &actualLhs.coeffRef(SkipDiag&&(UpLo&Lower)==Lower ? 1 : 0,0), actualLhs.outerStride(), - &actualRhs.coeffRef(0,SkipDiag&&(UpLo&Upper)==Upper ? 1 : 0), actualRhs.outerStride(), - mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? mat.innerStride() : mat.outerStride() ) : 0), - mat.innerStride(), mat.outerStride(), actualAlpha, blocking); - } -}; - -template -template -TriangularView& TriangularViewImpl::_assignProduct(const ProductType& prod, const Scalar& alpha, bool beta) -{ - EIGEN_STATIC_ASSERT((UpLo&UnitDiag)==0, WRITING_TO_TRIANGULAR_PART_WITH_UNIT_DIAGONAL_IS_NOT_SUPPORTED); - eigen_assert(derived().nestedExpression().rows() == prod.rows() && derived().cols() == prod.cols()); - - general_product_to_triangular_selector::InnerSize==1>::run(derived().nestedExpression().const_cast_derived(), prod, alpha, beta); - - return derived(); -} - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h deleted file mode 100644 index 691f95d6979..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h +++ /dev/null @@ -1,145 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Level 3 BLAS SYRK/HERK implementation. - ******************************************************************************** -*/ - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H -#define EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H - -namespace Eigen { - -namespace internal { - -template -struct general_matrix_matrix_rankupdate : - general_matrix_matrix_triangular_product< - Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,1,UpLo,BuiltIn> {}; - - -// try to go to BLAS specialization -#define EIGEN_BLAS_RANKUPDATE_SPECIALIZE(Scalar) \ -template \ -struct general_matrix_matrix_triangular_product { \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const Scalar* lhs, Index lhsStride, \ - const Scalar* rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) \ - { \ - if ( lhs==rhs && ((UpLo&(Lower|Upper))==UpLo) ) { \ - general_matrix_matrix_rankupdate \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resStride,alpha,blocking); \ - } else { \ - general_matrix_matrix_triangular_product \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resIncr,resStride,alpha,blocking); \ - } \ - } \ -}; - -EIGEN_BLAS_RANKUPDATE_SPECIALIZE(double) -EIGEN_BLAS_RANKUPDATE_SPECIALIZE(float) -// TODO handle complex cases -// EIGEN_BLAS_RANKUPDATE_SPECIALIZE(dcomplex) -// EIGEN_BLAS_RANKUPDATE_SPECIALIZE(scomplex) - -// SYRK for float/double -#define EIGEN_BLAS_RANKUPDATE_R(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct general_matrix_matrix_rankupdate { \ - enum { \ - IsLower = (UpLo&Lower) == Lower, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((AStorageOrder==ColMajor) && ConjugateA) ? 1 : 0 \ - }; \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* /*rhs*/, Index /*rhsStride*/, EIGTYPE* res, Index resStride, EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - /* typedef Matrix MatrixRhs;*/ \ -\ - BlasIndex lda=convert_index(lhsStride), ldc=convert_index(resStride), n=convert_index(size), k=convert_index(depth); \ - char uplo=((IsLower) ? 'L' : 'U'), trans=((AStorageOrder==RowMajor) ? 'T':'N'); \ - EIGTYPE beta(1); \ - BLASFUNC(&uplo, &trans, &n, &k, (const BLASTYPE*)&numext::real_ref(alpha), lhs, &lda, (const BLASTYPE*)&numext::real_ref(beta), res, &ldc); \ - } \ -}; - -// HERK for complex data -#define EIGEN_BLAS_RANKUPDATE_C(EIGTYPE, BLASTYPE, RTYPE, BLASFUNC) \ -template \ -struct general_matrix_matrix_rankupdate { \ - enum { \ - IsLower = (UpLo&Lower) == Lower, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = (((AStorageOrder==ColMajor) && ConjugateA) || ((AStorageOrder==RowMajor) && !ConjugateA)) ? 1 : 0 \ - }; \ - static EIGEN_STRONG_INLINE void run(Index size, Index depth,const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* /*rhs*/, Index /*rhsStride*/, EIGTYPE* res, Index resStride, EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - typedef Matrix MatrixType; \ -\ - BlasIndex lda=convert_index(lhsStride), ldc=convert_index(resStride), n=convert_index(size), k=convert_index(depth); \ - char uplo=((IsLower) ? 'L' : 'U'), trans=((AStorageOrder==RowMajor) ? 'C':'N'); \ - RTYPE alpha_, beta_; \ - const EIGTYPE* a_ptr; \ -\ - alpha_ = alpha.real(); \ - beta_ = 1.0; \ -/* Copy with conjugation in some cases*/ \ - MatrixType a; \ - if (conjA) { \ - Map > mapA(lhs,n,k,OuterStride<>(lhsStride)); \ - a = mapA.conjugate(); \ - lda = a.outerStride(); \ - a_ptr = a.data(); \ - } else a_ptr=lhs; \ - BLASFUNC(&uplo, &trans, &n, &k, &alpha_, (BLASTYPE*)a_ptr, &lda, &beta_, (BLASTYPE*)res, &ldc); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_RANKUPDATE_R(double, double, dsyrk) -EIGEN_BLAS_RANKUPDATE_R(float, float, ssyrk) -#else -EIGEN_BLAS_RANKUPDATE_R(double, double, dsyrk_) -EIGEN_BLAS_RANKUPDATE_R(float, float, ssyrk_) -#endif - -// TODO hanlde complex cases -// EIGEN_BLAS_RANKUPDATE_C(dcomplex, double, double, zherk_) -// EIGEN_BLAS_RANKUPDATE_C(scomplex, float, float, cherk_) - - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h deleted file mode 100644 index 71abf4013d1..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h +++ /dev/null @@ -1,124 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * General matrix-matrix product functionality based on ?GEMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H -#define EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements general matrix-matrix multiplication using BLAS -* gemm function via partial specialization of -* general_matrix_matrix_product::run(..) method for float, double, -* std::complex and std::complex types -**********************************************************************/ - -// gemm specialization - -#define GEMM_SPECIALIZATION(EIGTYPE, EIGPREFIX, BLASTYPE, BLASFUNC) \ -template< \ - typename Index, \ - int LhsStorageOrder, bool ConjugateLhs, \ - int RhsStorageOrder, bool ConjugateRhs> \ -struct general_matrix_matrix_product \ -{ \ -typedef gebp_traits Traits; \ -\ -static void run(Index rows, Index cols, Index depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, \ - level3_blocking& /*blocking*/, \ - GemmParallelInfo* /*info = 0*/) \ -{ \ - using std::conj; \ -\ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char transa, transb; \ - BlasIndex m, n, k, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX a_tmp, b_tmp; \ -\ -/* Set transpose options */ \ - transa = (LhsStorageOrder==RowMajor) ? ((ConjugateLhs) ? 'C' : 'T') : 'N'; \ - transb = (RhsStorageOrder==RowMajor) ? ((ConjugateRhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ - k = convert_index(depth); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if ((LhsStorageOrder==ColMajor) && (ConjugateLhs)) { \ - Map > lhs(_lhs,m,k,OuterStride<>(lhsStride)); \ - a_tmp = lhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _lhs; \ -\ - if ((RhsStorageOrder==ColMajor) && (ConjugateRhs)) { \ - Map > rhs(_rhs,k,n,OuterStride<>(rhsStride)); \ - b_tmp = rhs.conjugate(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _rhs; \ -\ - BLASFUNC(&transa, &transb, &m, &n, &k, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -}}; - -#ifdef EIGEN_USE_MKL -GEMM_SPECIALIZATION(double, d, double, dgemm) -GEMM_SPECIALIZATION(float, f, float, sgemm) -GEMM_SPECIALIZATION(dcomplex, cd, MKL_Complex16, zgemm) -GEMM_SPECIALIZATION(scomplex, cf, MKL_Complex8, cgemm) -#else -GEMM_SPECIALIZATION(double, d, double, dgemm_) -GEMM_SPECIALIZATION(float, f, float, sgemm_) -GEMM_SPECIALIZATION(dcomplex, cd, double, zgemm_) -GEMM_SPECIALIZATION(scomplex, cf, float, cgemm_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h deleted file mode 100644 index a597c1f4ee6..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/GeneralMatrixVector.h +++ /dev/null @@ -1,619 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_GENERAL_MATRIX_VECTOR_H -#define EIGEN_GENERAL_MATRIX_VECTOR_H - -namespace Eigen { - -namespace internal { - -/* Optimized col-major matrix * vector product: - * This algorithm processes 4 columns at onces that allows to both reduce - * the number of load/stores of the result by a factor 4 and to reduce - * the instruction dependency. Moreover, we know that all bands have the - * same alignment pattern. - * - * Mixing type logic: C += alpha * A * B - * | A | B |alpha| comments - * |real |cplx |cplx | no vectorization - * |real |cplx |real | alpha is converted to a cplx when calling the run function, no vectorization - * |cplx |real |cplx | invalid, the caller has to do tmp: = A * B; C += alpha*tmp - * |cplx |real |real | optimal case, vectorization possible via real-cplx mul - * - * Accesses to the matrix coefficients follow the following logic: - * - * - if all columns have the same alignment then - * - if the columns have the same alignment as the result vector, then easy! (-> AllAligned case) - * - otherwise perform unaligned loads only (-> NoneAligned case) - * - otherwise - * - if even columns have the same alignment then - * // odd columns are guaranteed to have the same alignment too - * - if even or odd columns have the same alignment as the result, then - * // for a register size of 2 scalars, this is guarantee to be the case (e.g., SSE with double) - * - perform half aligned and half unaligned loads (-> EvenAligned case) - * - otherwise perform unaligned loads only (-> NoneAligned case) - * - otherwise, if the register size is 4 scalars (e.g., SSE with float) then - * - one over 4 consecutive columns is guaranteed to be aligned with the result vector, - * perform simple aligned loads for this column and aligned loads plus re-alignment for the other. (-> FirstAligned case) - * // this re-alignment is done by the palign function implemented for SSE in Eigen/src/Core/arch/SSE/PacketMath.h - * - otherwise, - * // if we get here, this means the register size is greater than 4 (e.g., AVX with floats), - * // we currently fall back to the NoneAligned case - * - * The same reasoning apply for the transposed case. - * - * The last case (PacketSize>4) could probably be improved by generalizing the FirstAligned case, but since we do not support AVX yet... - * One might also wonder why in the EvenAligned case we perform unaligned loads instead of using the aligned-loads plus re-alignment - * strategy as in the FirstAligned case. The reason is that we observed that unaligned loads on a 8 byte boundary are not too slow - * compared to unaligned loads on a 4 byte boundary. - * - */ -template -struct general_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - -enum { - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable - && int(packet_traits::size)==int(packet_traits::size), - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1 -}; - -typedef typename packet_traits::type _LhsPacket; -typedef typename packet_traits::type _RhsPacket; -typedef typename packet_traits::type _ResPacket; - -typedef typename conditional::type LhsPacket; -typedef typename conditional::type RhsPacket; -typedef typename conditional::type ResPacket; - -EIGEN_DONT_INLINE static void run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - RhsScalar alpha); -}; - -template -EIGEN_DONT_INLINE void general_matrix_vector_product::run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - RhsScalar alpha) -{ - EIGEN_UNUSED_VARIABLE(resIncr); - eigen_internal_assert(resIncr==1); - #ifdef _EIGEN_ACCUMULATE_PACKETS - #error _EIGEN_ACCUMULATE_PACKETS has already been defined - #endif - #define _EIGEN_ACCUMULATE_PACKETS(Alignment0,Alignment13,Alignment2) \ - pstore(&res[j], \ - padd(pload(&res[j]), \ - padd( \ - padd(pcj.pmul(lhs0.template load(j), ptmp0), \ - pcj.pmul(lhs1.template load(j), ptmp1)), \ - padd(pcj.pmul(lhs2.template load(j), ptmp2), \ - pcj.pmul(lhs3.template load(j), ptmp3)) ))) - - typedef typename LhsMapper::VectorMapper LhsScalars; - - conj_helper cj; - conj_helper pcj; - if(ConjugateRhs) - alpha = numext::conj(alpha); - - enum { AllAligned = 0, EvenAligned, FirstAligned, NoneAligned }; - const Index columnsAtOnce = 4; - const Index peels = 2; - const Index LhsPacketAlignedMask = LhsPacketSize-1; - const Index ResPacketAlignedMask = ResPacketSize-1; -// const Index PeelAlignedMask = ResPacketSize*peels-1; - const Index size = rows; - - const Index lhsStride = lhs.stride(); - - // How many coeffs of the result do we have to skip to be aligned. - // Here we assume data are at least aligned on the base scalar type. - Index alignedStart = internal::first_default_aligned(res,size); - Index alignedSize = ResPacketSize>1 ? alignedStart + ((size-alignedStart) & ~ResPacketAlignedMask) : 0; - const Index peeledSize = alignedSize - RhsPacketSize*peels - RhsPacketSize + 1; - - const Index alignmentStep = LhsPacketSize>1 ? (LhsPacketSize - lhsStride % LhsPacketSize) & LhsPacketAlignedMask : 0; - Index alignmentPattern = alignmentStep==0 ? AllAligned - : alignmentStep==(LhsPacketSize/2) ? EvenAligned - : FirstAligned; - - // we cannot assume the first element is aligned because of sub-matrices - const Index lhsAlignmentOffset = lhs.firstAligned(size); - - // find how many columns do we have to skip to be aligned with the result (if possible) - Index skipColumns = 0; - // if the data cannot be aligned (TODO add some compile time tests when possible, e.g. for floats) - if( (lhsAlignmentOffset < 0) || (lhsAlignmentOffset == size) || (UIntPtr(res)%sizeof(ResScalar)) ) - { - alignedSize = 0; - alignedStart = 0; - alignmentPattern = NoneAligned; - } - else if(LhsPacketSize > 4) - { - // TODO: extend the code to support aligned loads whenever possible when LhsPacketSize > 4. - // Currently, it seems to be better to perform unaligned loads anyway - alignmentPattern = NoneAligned; - } - else if (LhsPacketSize>1) - { - // eigen_internal_assert(size_t(firstLhs+lhsAlignmentOffset)%sizeof(LhsPacket)==0 || size= cols) - || LhsPacketSize > size - || (size_t(firstLhs+alignedStart+lhsStride*skipColumns)%sizeof(LhsPacket))==0);*/ - } - else if(Vectorizable) - { - alignedStart = 0; - alignedSize = size; - alignmentPattern = AllAligned; - } - - const Index offset1 = (alignmentPattern==FirstAligned && alignmentStep==1)?3:1; - const Index offset3 = (alignmentPattern==FirstAligned && alignmentStep==1)?1:3; - - Index columnBound = ((cols-skipColumns)/columnsAtOnce)*columnsAtOnce + skipColumns; - for (Index i=skipColumns; i(alpha*rhs(i, 0)), - ptmp1 = pset1(alpha*rhs(i+offset1, 0)), - ptmp2 = pset1(alpha*rhs(i+2, 0)), - ptmp3 = pset1(alpha*rhs(i+offset3, 0)); - - // this helps a lot generating better binary code - const LhsScalars lhs0 = lhs.getVectorMapper(0, i+0), lhs1 = lhs.getVectorMapper(0, i+offset1), - lhs2 = lhs.getVectorMapper(0, i+2), lhs3 = lhs.getVectorMapper(0, i+offset3); - - if (Vectorizable) - { - /* explicit vectorization */ - // process initial unaligned coeffs - for (Index j=0; jalignedStart) - { - switch(alignmentPattern) - { - case AllAligned: - for (Index j = alignedStart; j1) - { - LhsPacket A00, A01, A02, A03, A10, A11, A12, A13; - ResPacket T0, T1; - - A01 = lhs1.template load(alignedStart-1); - A02 = lhs2.template load(alignedStart-2); - A03 = lhs3.template load(alignedStart-3); - - for (; j(j-1+LhsPacketSize); palign<1>(A01,A11); - A12 = lhs2.template load(j-2+LhsPacketSize); palign<2>(A02,A12); - A13 = lhs3.template load(j-3+LhsPacketSize); palign<3>(A03,A13); - - A00 = lhs0.template load(j); - A10 = lhs0.template load(j+LhsPacketSize); - T0 = pcj.pmadd(A00, ptmp0, pload(&res[j])); - T1 = pcj.pmadd(A10, ptmp0, pload(&res[j+ResPacketSize])); - - T0 = pcj.pmadd(A01, ptmp1, T0); - A01 = lhs1.template load(j-1+2*LhsPacketSize); palign<1>(A11,A01); - T0 = pcj.pmadd(A02, ptmp2, T0); - A02 = lhs2.template load(j-2+2*LhsPacketSize); palign<2>(A12,A02); - T0 = pcj.pmadd(A03, ptmp3, T0); - pstore(&res[j],T0); - A03 = lhs3.template load(j-3+2*LhsPacketSize); palign<3>(A13,A03); - T1 = pcj.pmadd(A11, ptmp1, T1); - T1 = pcj.pmadd(A12, ptmp2, T1); - T1 = pcj.pmadd(A13, ptmp3, T1); - pstore(&res[j+ResPacketSize],T1); - } - } - for (; j(alpha*rhs(k, 0)); - const LhsScalars lhs0 = lhs.getVectorMapper(0, k); - - if (Vectorizable) - { - /* explicit vectorization */ - // process first unaligned result's coeffs - for (Index j=0; j(alignedStart)) - for (Index i = alignedStart;i(i), ptmp0, pload(&res[i]))); - else - for (Index i = alignedStart;i(i), ptmp0, pload(&res[i]))); - } - - // process remaining scalars (or all if no explicit vectorization) - for (Index i=alignedSize; i -struct general_matrix_vector_product -{ -typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - -enum { - Vectorizable = packet_traits::Vectorizable && packet_traits::Vectorizable - && int(packet_traits::size)==int(packet_traits::size), - LhsPacketSize = Vectorizable ? packet_traits::size : 1, - RhsPacketSize = Vectorizable ? packet_traits::size : 1, - ResPacketSize = Vectorizable ? packet_traits::size : 1 -}; - -typedef typename packet_traits::type _LhsPacket; -typedef typename packet_traits::type _RhsPacket; -typedef typename packet_traits::type _ResPacket; - -typedef typename conditional::type LhsPacket; -typedef typename conditional::type RhsPacket; -typedef typename conditional::type ResPacket; - -EIGEN_DONT_INLINE static void run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - ResScalar alpha); -}; - -template -EIGEN_DONT_INLINE void general_matrix_vector_product::run( - Index rows, Index cols, - const LhsMapper& lhs, - const RhsMapper& rhs, - ResScalar* res, Index resIncr, - ResScalar alpha) -{ - eigen_internal_assert(rhs.stride()==1); - - #ifdef _EIGEN_ACCUMULATE_PACKETS - #error _EIGEN_ACCUMULATE_PACKETS has already been defined - #endif - - #define _EIGEN_ACCUMULATE_PACKETS(Alignment0,Alignment13,Alignment2) {\ - RhsPacket b = rhs.getVectorMapper(j, 0).template load(0); \ - ptmp0 = pcj.pmadd(lhs0.template load(j), b, ptmp0); \ - ptmp1 = pcj.pmadd(lhs1.template load(j), b, ptmp1); \ - ptmp2 = pcj.pmadd(lhs2.template load(j), b, ptmp2); \ - ptmp3 = pcj.pmadd(lhs3.template load(j), b, ptmp3); } - - conj_helper cj; - conj_helper pcj; - - typedef typename LhsMapper::VectorMapper LhsScalars; - - enum { AllAligned=0, EvenAligned=1, FirstAligned=2, NoneAligned=3 }; - const Index rowsAtOnce = 4; - const Index peels = 2; - const Index RhsPacketAlignedMask = RhsPacketSize-1; - const Index LhsPacketAlignedMask = LhsPacketSize-1; - const Index depth = cols; - const Index lhsStride = lhs.stride(); - - // How many coeffs of the result do we have to skip to be aligned. - // Here we assume data are at least aligned on the base scalar type - // if that's not the case then vectorization is discarded, see below. - Index alignedStart = rhs.firstAligned(depth); - Index alignedSize = RhsPacketSize>1 ? alignedStart + ((depth-alignedStart) & ~RhsPacketAlignedMask) : 0; - const Index peeledSize = alignedSize - RhsPacketSize*peels - RhsPacketSize + 1; - - const Index alignmentStep = LhsPacketSize>1 ? (LhsPacketSize - lhsStride % LhsPacketSize) & LhsPacketAlignedMask : 0; - Index alignmentPattern = alignmentStep==0 ? AllAligned - : alignmentStep==(LhsPacketSize/2) ? EvenAligned - : FirstAligned; - - // we cannot assume the first element is aligned because of sub-matrices - const Index lhsAlignmentOffset = lhs.firstAligned(depth); - const Index rhsAlignmentOffset = rhs.firstAligned(rows); - - // find how many rows do we have to skip to be aligned with rhs (if possible) - Index skipRows = 0; - // if the data cannot be aligned (TODO add some compile time tests when possible, e.g. for floats) - if( (sizeof(LhsScalar)!=sizeof(RhsScalar)) || - (lhsAlignmentOffset < 0) || (lhsAlignmentOffset == depth) || - (rhsAlignmentOffset < 0) || (rhsAlignmentOffset == rows) ) - { - alignedSize = 0; - alignedStart = 0; - alignmentPattern = NoneAligned; - } - else if(LhsPacketSize > 4) - { - // TODO: extend the code to support aligned loads whenever possible when LhsPacketSize > 4. - alignmentPattern = NoneAligned; - } - else if (LhsPacketSize>1) - { - // eigen_internal_assert(size_t(firstLhs+lhsAlignmentOffset)%sizeof(LhsPacket)==0 || depth= rows) - || LhsPacketSize > depth - || (size_t(firstLhs+alignedStart+lhsStride*skipRows)%sizeof(LhsPacket))==0);*/ - } - else if(Vectorizable) - { - alignedStart = 0; - alignedSize = depth; - alignmentPattern = AllAligned; - } - - const Index offset1 = (alignmentPattern==FirstAligned && alignmentStep==1)?3:1; - const Index offset3 = (alignmentPattern==FirstAligned && alignmentStep==1)?1:3; - - Index rowBound = ((rows-skipRows)/rowsAtOnce)*rowsAtOnce + skipRows; - for (Index i=skipRows; i(ResScalar(0)), ptmp1 = pset1(ResScalar(0)), - ptmp2 = pset1(ResScalar(0)), ptmp3 = pset1(ResScalar(0)); - - // process initial unaligned coeffs - // FIXME this loop get vectorized by the compiler ! - for (Index j=0; jalignedStart) - { - switch(alignmentPattern) - { - case AllAligned: - for (Index j = alignedStart; j1) - { - /* Here we proccess 4 rows with with two peeled iterations to hide - * the overhead of unaligned loads. Moreover unaligned loads are handled - * using special shift/move operations between the two aligned packets - * overlaping the desired unaligned packet. This is *much* more efficient - * than basic unaligned loads. - */ - LhsPacket A01, A02, A03, A11, A12, A13; - A01 = lhs1.template load(alignedStart-1); - A02 = lhs2.template load(alignedStart-2); - A03 = lhs3.template load(alignedStart-3); - - for (; j(0); - A11 = lhs1.template load(j-1+LhsPacketSize); palign<1>(A01,A11); - A12 = lhs2.template load(j-2+LhsPacketSize); palign<2>(A02,A12); - A13 = lhs3.template load(j-3+LhsPacketSize); palign<3>(A03,A13); - - ptmp0 = pcj.pmadd(lhs0.template load(j), b, ptmp0); - ptmp1 = pcj.pmadd(A01, b, ptmp1); - A01 = lhs1.template load(j-1+2*LhsPacketSize); palign<1>(A11,A01); - ptmp2 = pcj.pmadd(A02, b, ptmp2); - A02 = lhs2.template load(j-2+2*LhsPacketSize); palign<2>(A12,A02); - ptmp3 = pcj.pmadd(A03, b, ptmp3); - A03 = lhs3.template load(j-3+2*LhsPacketSize); palign<3>(A13,A03); - - b = rhs.getVectorMapper(j+RhsPacketSize, 0).template load(0); - ptmp0 = pcj.pmadd(lhs0.template load(j+LhsPacketSize), b, ptmp0); - ptmp1 = pcj.pmadd(A11, b, ptmp1); - ptmp2 = pcj.pmadd(A12, b, ptmp2); - ptmp3 = pcj.pmadd(A13, b, ptmp3); - } - } - for (; j(tmp0); - const LhsScalars lhs0 = lhs.getVectorMapper(i, 0); - // process first unaligned result's coeffs - // FIXME this loop get vectorized by the compiler ! - for (Index j=0; jalignedStart) - { - // process aligned rhs coeffs - if (lhs0.template aligned(alignedStart)) - for (Index j = alignedStart;j(j), rhs.getVectorMapper(j, 0).template load(0), ptmp0); - else - for (Index j = alignedStart;j(j), rhs.getVectorMapper(j, 0).template load(0), ptmp0); - tmp0 += predux(ptmp0); - } - - // process remaining scalars - // FIXME this loop get vectorized by the compiler ! - for (Index j=alignedSize; j and std::complex types -**********************************************************************/ - -// gemv specialization - -template -struct general_matrix_vector_product_gemv; - -#define EIGEN_BLAS_GEMV_SPECIALIZE(Scalar) \ -template \ -struct general_matrix_vector_product,ColMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,Specialized> { \ -static void run( \ - Index rows, Index cols, \ - const const_blas_data_mapper &lhs, \ - const const_blas_data_mapper &rhs, \ - Scalar* res, Index resIncr, Scalar alpha) \ -{ \ - if (ConjugateLhs) { \ - general_matrix_vector_product,ColMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,BuiltIn>::run( \ - rows, cols, lhs, rhs, res, resIncr, alpha); \ - } else { \ - general_matrix_vector_product_gemv::run( \ - rows, cols, lhs.data(), lhs.stride(), rhs.data(), rhs.stride(), res, resIncr, alpha); \ - } \ -} \ -}; \ -template \ -struct general_matrix_vector_product,RowMajor,ConjugateLhs,Scalar,const_blas_data_mapper,ConjugateRhs,Specialized> { \ -static void run( \ - Index rows, Index cols, \ - const const_blas_data_mapper &lhs, \ - const const_blas_data_mapper &rhs, \ - Scalar* res, Index resIncr, Scalar alpha) \ -{ \ - general_matrix_vector_product_gemv::run( \ - rows, cols, lhs.data(), lhs.stride(), rhs.data(), rhs.stride(), res, resIncr, alpha); \ -} \ -}; \ - -EIGEN_BLAS_GEMV_SPECIALIZE(double) -EIGEN_BLAS_GEMV_SPECIALIZE(float) -EIGEN_BLAS_GEMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_GEMV_SPECIALIZE(scomplex) - -#define EIGEN_BLAS_GEMV_SPECIALIZATION(EIGTYPE,BLASTYPE,BLASFUNC) \ -template \ -struct general_matrix_vector_product_gemv \ -{ \ -typedef Matrix GEMVVector;\ -\ -static void run( \ - Index rows, Index cols, \ - const EIGTYPE* lhs, Index lhsStride, \ - const EIGTYPE* rhs, Index rhsIncr, \ - EIGTYPE* res, Index resIncr, EIGTYPE alpha) \ -{ \ - BlasIndex m=convert_index(rows), n=convert_index(cols), \ - lda=convert_index(lhsStride), incx=convert_index(rhsIncr), incy=convert_index(resIncr); \ - const EIGTYPE beta(1); \ - const EIGTYPE *x_ptr; \ - char trans=(LhsStorageOrder==ColMajor) ? 'N' : (ConjugateLhs) ? 'C' : 'T'; \ - if (LhsStorageOrder==RowMajor) { \ - m = convert_index(cols); \ - n = convert_index(rows); \ - }\ - GEMVVector x_tmp; \ - if (ConjugateRhs) { \ - Map > map_x(rhs,cols,1,InnerStride<>(incx)); \ - x_tmp=map_x.conjugate(); \ - x_ptr=x_tmp.data(); \ - incx=1; \ - } else x_ptr=rhs; \ - BLASFUNC(&trans, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)lhs, &lda, (const BLASTYPE*)x_ptr, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &incy); \ -}\ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_GEMV_SPECIALIZATION(double, double, dgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(float, float, sgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(dcomplex, MKL_Complex16, zgemv) -EIGEN_BLAS_GEMV_SPECIALIZATION(scomplex, MKL_Complex8 , cgemv) -#else -EIGEN_BLAS_GEMV_SPECIALIZATION(double, double, dgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(float, float, sgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(dcomplex, double, zgemv_) -EIGEN_BLAS_GEMV_SPECIALIZATION(scomplex, float, cgemv_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_GENERAL_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h deleted file mode 100644 index a3cc05b77bb..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/Parallelizer.h +++ /dev/null @@ -1,166 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARALLELIZER_H -#define EIGEN_PARALLELIZER_H - -namespace Eigen { - -namespace internal { - -/** \internal */ -inline void manage_multi_threading(Action action, int* v) -{ - static int m_maxThreads = -1; - EIGEN_UNUSED_VARIABLE(m_maxThreads); - - if(action==SetAction) - { - eigen_internal_assert(v!=0); - m_maxThreads = *v; - } - else if(action==GetAction) - { - eigen_internal_assert(v!=0); - #ifdef EIGEN_HAS_OPENMP - if(m_maxThreads>0) - *v = m_maxThreads; - else - *v = omp_get_max_threads(); - #else - *v = 1; - #endif - } - else - { - eigen_internal_assert(false); - } -} - -} - -/** Must be call first when calling Eigen from multiple threads */ -inline void initParallel() -{ - int nbt; - internal::manage_multi_threading(GetAction, &nbt); - std::ptrdiff_t l1, l2, l3; - internal::manage_caching_sizes(GetAction, &l1, &l2, &l3); -} - -/** \returns the max number of threads reserved for Eigen - * \sa setNbThreads */ -inline int nbThreads() -{ - int ret; - internal::manage_multi_threading(GetAction, &ret); - return ret; -} - -/** Sets the max number of threads reserved for Eigen - * \sa nbThreads */ -inline void setNbThreads(int v) -{ - internal::manage_multi_threading(SetAction, &v); -} - -namespace internal { - -template struct GemmParallelInfo -{ - GemmParallelInfo() : sync(-1), users(0), lhs_start(0), lhs_length(0) {} - - Index volatile sync; - int volatile users; - - Index lhs_start; - Index lhs_length; -}; - -template -void parallelize_gemm(const Functor& func, Index rows, Index cols, Index depth, bool transpose) -{ - // TODO when EIGEN_USE_BLAS is defined, - // we should still enable OMP for other scalar types -#if !(defined (EIGEN_HAS_OPENMP)) || defined (EIGEN_USE_BLAS) - // FIXME the transpose variable is only needed to properly split - // the matrix product when multithreading is enabled. This is a temporary - // fix to support row-major destination matrices. This whole - // parallelizer mechanism has to be redisigned anyway. - EIGEN_UNUSED_VARIABLE(depth); - EIGEN_UNUSED_VARIABLE(transpose); - func(0,rows, 0,cols); -#else - - // Dynamically check whether we should enable or disable OpenMP. - // The conditions are: - // - the max number of threads we can create is greater than 1 - // - we are not already in a parallel code - // - the sizes are large enough - - // compute the maximal number of threads from the size of the product: - // This first heuristic takes into account that the product kernel is fully optimized when working with nr columns at once. - Index size = transpose ? rows : cols; - Index pb_max_threads = std::max(1,size / Functor::Traits::nr); - - // compute the maximal number of threads from the total amount of work: - double work = static_cast(rows) * static_cast(cols) * - static_cast(depth); - double kMinTaskSize = 50000; // FIXME improve this heuristic. - pb_max_threads = std::max(1, std::min(pb_max_threads, work / kMinTaskSize)); - - // compute the number of threads we are going to use - Index threads = std::min(nbThreads(), pb_max_threads); - - // if multi-threading is explicitely disabled, not useful, or if we already are in a parallel session, - // then abort multi-threading - // FIXME omp_get_num_threads()>1 only works for openmp, what if the user does not use openmp? - if((!Condition) || (threads==1) || (omp_get_num_threads()>1)) - return func(0,rows, 0,cols); - - Eigen::initParallel(); - func.initParallelSession(threads); - - if(transpose) - std::swap(rows,cols); - - ei_declare_aligned_stack_constructed_variable(GemmParallelInfo,info,threads,0); - - #pragma omp parallel num_threads(threads) - { - Index i = omp_get_thread_num(); - // Note that the actual number of threads might be lower than the number of request ones. - Index actual_threads = omp_get_num_threads(); - - Index blockCols = (cols / actual_threads) & ~Index(0x3); - Index blockRows = (rows / actual_threads); - blockRows = (blockRows/Functor::Traits::mr)*Functor::Traits::mr; - - Index r0 = i*blockRows; - Index actualBlockRows = (i+1==actual_threads) ? rows-r0 : blockRows; - - Index c0 = i*blockCols; - Index actualBlockCols = (i+1==actual_threads) ? cols-c0 : blockCols; - - info[i].lhs_start = r0; - info[i].lhs_length = actualBlockRows; - - if(transpose) - func(c0, actualBlockCols, 0, rows, info); - else - func(0, rows, c0, actualBlockCols, info); - } -#endif -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PARALLELIZER_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h deleted file mode 100644 index 04c933480bd..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix.h +++ /dev/null @@ -1,527 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_MATRIX_MATRIX_H -#define EIGEN_SELFADJOINT_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -// pack a selfadjoint block diagonal for use with the gebp_kernel -template -struct symm_pack_lhs -{ - template inline - void pack(Scalar* blockA, const const_blas_data_mapper& lhs, Index cols, Index i, Index& count) - { - // normal copy - for(Index k=0; k::size }; - const_blas_data_mapper lhs(_lhs,lhsStride); - Index count = 0; - //Index peeled_mc3 = (rows/Pack1)*Pack1; - - const Index peeled_mc3 = Pack1>=3*PacketSize ? (rows/(3*PacketSize))*(3*PacketSize) : 0; - const Index peeled_mc2 = Pack1>=2*PacketSize ? peeled_mc3+((rows-peeled_mc3)/(2*PacketSize))*(2*PacketSize) : 0; - const Index peeled_mc1 = Pack1>=1*PacketSize ? (rows/(1*PacketSize))*(1*PacketSize) : 0; - - if(Pack1>=3*PacketSize) - for(Index i=0; i(blockA, lhs, cols, i, count); - - if(Pack1>=2*PacketSize) - for(Index i=peeled_mc3; i(blockA, lhs, cols, i, count); - - if(Pack1>=1*PacketSize) - for(Index i=peeled_mc2; i(blockA, lhs, cols, i, count); - - // do the same with mr==1 - for(Index i=peeled_mc1; i -struct symm_pack_rhs -{ - enum { PacketSize = packet_traits::size }; - void operator()(Scalar* blockB, const Scalar* _rhs, Index rhsStride, Index rows, Index cols, Index k2) - { - Index end_k = k2 + rows; - Index count = 0; - const_blas_data_mapper rhs(_rhs,rhsStride); - Index packet_cols8 = nr>=8 ? (cols/8) * 8 : 0; - Index packet_cols4 = nr>=4 ? (cols/4) * 4 : 0; - - // first part: normal case - for(Index j2=0; j2=4) - { - blockB[count+2] = rhs(k,j2+2); - blockB[count+3] = rhs(k,j2+3); - } - if (nr>=8) - { - blockB[count+4] = rhs(k,j2+4); - blockB[count+5] = rhs(k,j2+5); - blockB[count+6] = rhs(k,j2+6); - blockB[count+7] = rhs(k,j2+7); - } - count += nr; - } - } - - // second part: diagonal block - Index end8 = nr>=8 ? (std::min)(k2+rows,packet_cols8) : k2; - if(nr>=8) - { - for(Index j2=k2; j2=4) - { - for(Index j2=end8; j2<(std::min)(k2+rows,packet_cols4); j2+=4) - { - // again we can split vertically in three different parts (transpose, symmetric, normal) - // transpose - for(Index k=k2; k=8) - { - for(Index j2=k2+rows; j2=4) - { - for(Index j2=(std::max)(packet_cols8,k2+rows); j2 the same with nr==1) - for(Index j2=packet_cols4; j2 -struct product_selfadjoint_matrix; - -template -struct product_selfadjoint_matrix -{ - - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - product_selfadjoint_matrix::IsComplex && EIGEN_LOGICAL_XOR(RhsSelfAdjoint,ConjugateRhs), - EIGEN_LOGICAL_XOR(LhsSelfAdjoint,LhsStorageOrder==RowMajor) ? ColMajor : RowMajor, - LhsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsSelfAdjoint,ConjugateLhs), - ColMajor,ResInnerStride> - ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); - } -}; - -template -struct product_selfadjoint_matrix -{ - - static EIGEN_DONT_INLINE void run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - Index size = rows; - - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper LhsTransposeMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - LhsTransposeMapper lhs_transpose(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - // kc must be smaller than mc - kc = (std::min)(kc,mc); - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - gebp_kernel gebp_kernel; - symm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gemm_pack_lhs pack_lhs_transposed; - - for(Index k2=0; k2 transposed packed copy - // 2 - the diagonal block => special packed copy - // 3 - the panel below the diagonal block => generic packed copy - for(Index i2=0; i2() - (blockA, lhs.getSubMapper(i2, k2), actual_kc, actual_mc); - - gebp_kernel(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, cols, alpha); - } - } - } - -// matrix * selfadjoint product -template -struct product_selfadjoint_matrix -{ - - static EIGEN_DONT_INLINE void run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( - Index rows, Index cols, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - Index size = cols; - - typedef gebp_traits Traits; - - typedef const_blas_data_mapper LhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - ResMapper res(_res,resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - symm_pack_rhs pack_rhs; - - for(Index k2=0; k2 GEPP - for(Index i2=0; i2 -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - - enum { - LhsIsUpper = (LhsMode&(Upper|Lower))==Upper, - LhsIsSelfAdjoint = (LhsMode&SelfAdjoint)==SelfAdjoint, - RhsIsUpper = (RhsMode&(Upper|Lower))==Upper, - RhsIsSelfAdjoint = (RhsMode&SelfAdjoint)==SelfAdjoint - }; - - template - static void run(Dest &dst, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - eigen_assert(dst.rows()==a_lhs.rows() && dst.cols()==a_rhs.cols()); - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Lhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxColsAtCompileTime,1> BlockingType; - - BlockingType blocking(lhs.rows(), rhs.cols(), lhs.cols(), 1, false); - - internal::product_selfadjoint_matrix::Flags &RowMajorBit) ? RowMajor : ColMajor, LhsIsSelfAdjoint, - NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsIsUpper,bool(LhsBlasTraits::NeedToConjugate)), - EIGEN_LOGICAL_XOR(RhsIsUpper,internal::traits::Flags &RowMajorBit) ? RowMajor : ColMajor, RhsIsSelfAdjoint, - NumTraits::IsComplex && EIGEN_LOGICAL_XOR(RhsIsUpper,bool(RhsBlasTraits::NeedToConjugate)), - internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime> - ::run( - lhs.rows(), rhs.cols(), // sizes - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info - actualAlpha, blocking // alpha - ); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h deleted file mode 100644 index 61396dbdf6f..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h +++ /dev/null @@ -1,295 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -// - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Self adjoint matrix * matrix product functionality based on ?SYMM/?HEMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H -#define EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - - -/* Optimized selfadjoint matrix * matrix (?SYMM/?HEMM) product */ - -#define EIGEN_BLAS_SYMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ -\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='L', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ -\ -/* Set transpose options */ \ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (LhsStorageOrder==RowMajor) uplo='U'; \ - a = _lhs; \ -\ - if (RhsStorageOrder==RowMajor) { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.adjoint(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _rhs; \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - - -#define EIGEN_BLAS_HEMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='L', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ - Matrix a_tmp; \ -\ -/* Set transpose options */ \ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(lhsStride); \ - ldb = convert_index(rhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (((LhsStorageOrder==ColMajor) && ConjugateLhs) || ((LhsStorageOrder==RowMajor) && (!ConjugateLhs))) { \ - Map, 0, OuterStride<> > lhs(_lhs,m,m,OuterStride<>(lhsStride)); \ - a_tmp = lhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _lhs; \ - if (LhsStorageOrder==RowMajor) uplo='U'; \ -\ - if (RhsStorageOrder==ColMajor && (!ConjugateRhs)) { \ - b = _rhs; } \ - else { \ - if (RhsStorageOrder==ColMajor && ConjugateRhs) { \ - Map > rhs(_rhs,m,n,OuterStride<>(rhsStride)); \ - b_tmp = rhs.conjugate(); \ - } else \ - if (ConjugateRhs) { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.adjoint(); \ - } else { \ - Map > rhs(_rhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = rhs.transpose(); \ - } \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMM_L(double, double, d, dsymm) -EIGEN_BLAS_SYMM_L(float, float, f, ssymm) -EIGEN_BLAS_HEMM_L(dcomplex, MKL_Complex16, cd, zhemm) -EIGEN_BLAS_HEMM_L(scomplex, MKL_Complex8, cf, chemm) -#else -EIGEN_BLAS_SYMM_L(double, double, d, dsymm_) -EIGEN_BLAS_SYMM_L(float, float, f, ssymm_) -EIGEN_BLAS_HEMM_L(dcomplex, double, cd, zhemm_) -EIGEN_BLAS_HEMM_L(scomplex, float, cf, chemm_) -#endif - -/* Optimized matrix * selfadjoint matrix (?SYMM/?HEMM) product */ - -#define EIGEN_BLAS_SYMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ -\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='R', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(rhsStride); \ - ldb = convert_index(lhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (RhsStorageOrder==RowMajor) uplo='U'; \ - a = _rhs; \ -\ - if (LhsStorageOrder==RowMajor) { \ - Map > lhs(_lhs,n,m,OuterStride<>(rhsStride)); \ - b_tmp = lhs.adjoint(); \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } else b = _lhs; \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ -\ - } \ -}; - - -#define EIGEN_BLAS_HEMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_selfadjoint_matrix \ -{\ - static void run( \ - Index rows, Index cols, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ - EIGTYPE alpha, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - char side='R', uplo='L'; \ - BlasIndex m, n, lda, ldb, ldc; \ - const EIGTYPE *a, *b; \ - EIGTYPE beta(1); \ - MatrixX##EIGPREFIX b_tmp; \ - Matrix a_tmp; \ -\ -/* Set m, n, k */ \ - m = convert_index(rows); \ - n = convert_index(cols); \ -\ -/* Set lda, ldb, ldc */ \ - lda = convert_index(rhsStride); \ - ldb = convert_index(lhsStride); \ - ldc = convert_index(resStride); \ -\ -/* Set a, b, c */ \ - if (((RhsStorageOrder==ColMajor) && ConjugateRhs) || ((RhsStorageOrder==RowMajor) && (!ConjugateRhs))) { \ - Map, 0, OuterStride<> > rhs(_rhs,n,n,OuterStride<>(rhsStride)); \ - a_tmp = rhs.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else a = _rhs; \ - if (RhsStorageOrder==RowMajor) uplo='U'; \ -\ - if (LhsStorageOrder==ColMajor && (!ConjugateLhs)) { \ - b = _lhs; } \ - else { \ - if (LhsStorageOrder==ColMajor && ConjugateLhs) { \ - Map > lhs(_lhs,m,n,OuterStride<>(lhsStride)); \ - b_tmp = lhs.conjugate(); \ - } else \ - if (ConjugateLhs) { \ - Map > lhs(_lhs,n,m,OuterStride<>(lhsStride)); \ - b_tmp = lhs.adjoint(); \ - } else { \ - Map > lhs(_lhs,n,m,OuterStride<>(lhsStride)); \ - b_tmp = lhs.transpose(); \ - } \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ - } \ -\ - BLASFUNC(&side, &uplo, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)b, &ldb, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &ldc); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMM_R(double, double, d, dsymm) -EIGEN_BLAS_SYMM_R(float, float, f, ssymm) -EIGEN_BLAS_HEMM_R(dcomplex, MKL_Complex16, cd, zhemm) -EIGEN_BLAS_HEMM_R(scomplex, MKL_Complex8, cf, chemm) -#else -EIGEN_BLAS_SYMM_R(double, double, d, dsymm_) -EIGEN_BLAS_SYMM_R(float, float, f, ssymm_) -EIGEN_BLAS_HEMM_R(dcomplex, double, cd, zhemm_) -EIGEN_BLAS_HEMM_R(scomplex, float, cf, chemm_) -#endif -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h deleted file mode 100644 index 3fd180e6c0b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector.h +++ /dev/null @@ -1,260 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_MATRIX_VECTOR_H -#define EIGEN_SELFADJOINT_MATRIX_VECTOR_H - -namespace Eigen { - -namespace internal { - -/* Optimized selfadjoint matrix * vector product: - * This algorithm processes 2 columns at onces that allows to both reduce - * the number of load/stores of the result by a factor 2 and to reduce - * the instruction dependency. - */ - -template -struct selfadjoint_matrix_vector_product; - -template -struct selfadjoint_matrix_vector_product - -{ -static EIGEN_DONT_INLINE void run( - Index size, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, - Scalar* res, - Scalar alpha); -}; - -template -EIGEN_DONT_INLINE void selfadjoint_matrix_vector_product::run( - Index size, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, - Scalar* res, - Scalar alpha) -{ - typedef typename packet_traits::type Packet; - typedef typename NumTraits::Real RealScalar; - const Index PacketSize = sizeof(Packet)/sizeof(Scalar); - - enum { - IsRowMajor = StorageOrder==RowMajor ? 1 : 0, - IsLower = UpLo == Lower ? 1 : 0, - FirstTriangular = IsRowMajor == IsLower - }; - - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, IsRowMajor), ConjugateRhs> cj0; - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, !IsRowMajor), ConjugateRhs> cj1; - conj_helper cjd; - - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, IsRowMajor), ConjugateRhs> pcj0; - conj_helper::IsComplex && EIGEN_LOGICAL_XOR(ConjugateLhs, !IsRowMajor), ConjugateRhs> pcj1; - - Scalar cjAlpha = ConjugateRhs ? numext::conj(alpha) : alpha; - - - Index bound = (std::max)(Index(0),size-8) & 0xfffffffe; - if (FirstTriangular) - bound = size - bound; - - for (Index j=FirstTriangular ? bound : 0; - j<(FirstTriangular ? size : bound);j+=2) - { - const Scalar* EIGEN_RESTRICT A0 = lhs + j*lhsStride; - const Scalar* EIGEN_RESTRICT A1 = lhs + (j+1)*lhsStride; - - Scalar t0 = cjAlpha * rhs[j]; - Packet ptmp0 = pset1(t0); - Scalar t1 = cjAlpha * rhs[j+1]; - Packet ptmp1 = pset1(t1); - - Scalar t2(0); - Packet ptmp2 = pset1(t2); - Scalar t3(0); - Packet ptmp3 = pset1(t3); - - Index starti = FirstTriangular ? 0 : j+2; - Index endi = FirstTriangular ? j : size; - Index alignedStart = (starti) + internal::first_default_aligned(&res[starti], endi-starti); - Index alignedEnd = alignedStart + ((endi-alignedStart)/(PacketSize))*(PacketSize); - - res[j] += cjd.pmul(numext::real(A0[j]), t0); - res[j+1] += cjd.pmul(numext::real(A1[j+1]), t1); - if(FirstTriangular) - { - res[j] += cj0.pmul(A1[j], t1); - t3 += cj1.pmul(A1[j], rhs[j]); - } - else - { - res[j+1] += cj0.pmul(A0[j+1],t0); - t2 += cj1.pmul(A0[j+1], rhs[j+1]); - } - - for (Index i=starti; i huge speed up) - // gcc 4.2 does this optimization automatically. - const Scalar* EIGEN_RESTRICT a0It = A0 + alignedStart; - const Scalar* EIGEN_RESTRICT a1It = A1 + alignedStart; - const Scalar* EIGEN_RESTRICT rhsIt = rhs + alignedStart; - Scalar* EIGEN_RESTRICT resIt = res + alignedStart; - for (Index i=alignedStart; i(a0It); a0It += PacketSize; - Packet A1i = ploadu(a1It); a1It += PacketSize; - Packet Bi = ploadu(rhsIt); rhsIt += PacketSize; // FIXME should be aligned in most cases - Packet Xi = pload (resIt); - - Xi = pcj0.pmadd(A0i,ptmp0, pcj0.pmadd(A1i,ptmp1,Xi)); - ptmp2 = pcj1.pmadd(A0i, Bi, ptmp2); - ptmp3 = pcj1.pmadd(A1i, Bi, ptmp3); - pstore(resIt,Xi); resIt += PacketSize; - } - for (Index i=alignedEnd; i -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - enum { LhsUpLo = LhsMode&(Upper|Lower) }; - - template - static void run(Dest& dest, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - typedef typename Dest::Scalar ResScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; - - eigen_assert(dest.rows()==a_lhs.rows() && dest.cols()==a_rhs.cols()); - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(a_lhs) - * RhsBlasTraits::extractScalarFactor(a_rhs); - - enum { - EvalToDest = (Dest::InnerStrideAtCompileTime==1), - UseRhs = (ActualRhsTypeCleaned::InnerStrideAtCompileTime==1) - }; - - internal::gemv_static_vector_if static_dest; - internal::gemv_static_vector_if static_rhs; - - ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), - EvalToDest ? dest.data() : static_dest.data()); - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,rhs.size(), - UseRhs ? const_cast(rhs.data()) : static_rhs.data()); - - if(!EvalToDest) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = dest.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - MappedDest(actualDestPtr, dest.size()) = dest; - } - - if(!UseRhs) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = rhs.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - Map(actualRhsPtr, rhs.size()) = rhs; - } - - - internal::selfadjoint_matrix_vector_product::Flags&RowMajorBit) ? RowMajor : ColMajor, - int(LhsUpLo), bool(LhsBlasTraits::NeedToConjugate), bool(RhsBlasTraits::NeedToConjugate)>::run - ( - lhs.rows(), // size - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - actualRhsPtr, // rhs info - actualDestPtr, // result info - actualAlpha // scale factor - ); - - if(!EvalToDest) - dest = MappedDest(actualDestPtr, dest.size()); - } -}; - -template -struct selfadjoint_product_impl -{ - typedef typename Product::Scalar Scalar; - enum { RhsUpLo = RhsMode&(Upper|Lower) }; - - template - static void run(Dest& dest, const Lhs &a_lhs, const Rhs &a_rhs, const Scalar& alpha) - { - // let's simply transpose the product - Transpose destT(dest); - selfadjoint_product_impl, int(RhsUpLo)==Upper ? Lower : Upper, false, - Transpose, 0, true>::run(destT, a_rhs.transpose(), a_lhs.transpose(), alpha); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_VECTOR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h deleted file mode 100644 index 1238345e3fa..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointMatrixVector_BLAS.h +++ /dev/null @@ -1,118 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Selfadjoint matrix-vector product functionality based on ?SYMV/HEMV. - ******************************************************************************** -*/ - -#ifndef EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H -#define EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements selfadjoint matrix-vector multiplication using BLAS -**********************************************************************/ - -// symv/hemv specialization - -template -struct selfadjoint_matrix_vector_product_symv : - selfadjoint_matrix_vector_product {}; - -#define EIGEN_BLAS_SYMV_SPECIALIZE(Scalar) \ -template \ -struct selfadjoint_matrix_vector_product { \ -static void run( \ - Index size, const Scalar* lhs, Index lhsStride, \ - const Scalar* _rhs, Scalar* res, Scalar alpha) { \ - enum {\ - IsColMajor = StorageOrder==ColMajor \ - }; \ - if (IsColMajor == ConjugateLhs) {\ - selfadjoint_matrix_vector_product::run( \ - size, lhs, lhsStride, _rhs, res, alpha); \ - } else {\ - selfadjoint_matrix_vector_product_symv::run( \ - size, lhs, lhsStride, _rhs, res, alpha); \ - }\ - } \ -}; \ - -EIGEN_BLAS_SYMV_SPECIALIZE(double) -EIGEN_BLAS_SYMV_SPECIALIZE(float) -EIGEN_BLAS_SYMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_SYMV_SPECIALIZE(scomplex) - -#define EIGEN_BLAS_SYMV_SPECIALIZATION(EIGTYPE,BLASTYPE,BLASFUNC) \ -template \ -struct selfadjoint_matrix_vector_product_symv \ -{ \ -typedef Matrix SYMVVector;\ -\ -static void run( \ -Index size, const EIGTYPE* lhs, Index lhsStride, \ -const EIGTYPE* _rhs, EIGTYPE* res, EIGTYPE alpha) \ -{ \ - enum {\ - IsRowMajor = StorageOrder==RowMajor ? 1 : 0, \ - IsLower = UpLo == Lower ? 1 : 0 \ - }; \ - BlasIndex n=convert_index(size), lda=convert_index(lhsStride), incx=1, incy=1; \ - EIGTYPE beta(1); \ - const EIGTYPE *x_ptr; \ - char uplo=(IsRowMajor) ? (IsLower ? 'U' : 'L') : (IsLower ? 'L' : 'U'); \ - SYMVVector x_tmp; \ - if (ConjugateRhs) { \ - Map map_x(_rhs,size,1); \ - x_tmp=map_x.conjugate(); \ - x_ptr=x_tmp.data(); \ - } else x_ptr=_rhs; \ - BLASFUNC(&uplo, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)lhs, &lda, (const BLASTYPE*)x_ptr, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)res, &incy); \ -}\ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_SYMV_SPECIALIZATION(double, double, dsymv) -EIGEN_BLAS_SYMV_SPECIALIZATION(float, float, ssymv) -EIGEN_BLAS_SYMV_SPECIALIZATION(dcomplex, MKL_Complex16, zhemv) -EIGEN_BLAS_SYMV_SPECIALIZATION(scomplex, MKL_Complex8, chemv) -#else -EIGEN_BLAS_SYMV_SPECIALIZATION(double, double, dsymv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(float, float, ssymv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(dcomplex, double, zhemv_) -EIGEN_BLAS_SYMV_SPECIALIZATION(scomplex, float, chemv_) -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h deleted file mode 100644 index ef12c98f6cb..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointProduct.h +++ /dev/null @@ -1,133 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINT_PRODUCT_H -#define EIGEN_SELFADJOINT_PRODUCT_H - -/********************************************************************** -* This file implements a self adjoint product: C += A A^T updating only -* half of the selfadjoint matrix C. -* It corresponds to the level 3 SYRK and level 2 SYR Blas routines. -**********************************************************************/ - -namespace Eigen { - - -template -struct selfadjoint_rank1_update -{ - static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha) - { - internal::conj_if cj; - typedef Map > OtherMap; - typedef typename internal::conditional::type ConjLhsType; - for (Index i=0; i >(mat+stride*i+(UpLo==Lower ? i : 0), (UpLo==Lower ? size-i : (i+1))) - += (alpha * cj(vecY[i])) * ConjLhsType(OtherMap(vecX+(UpLo==Lower ? i : 0),UpLo==Lower ? size-i : (i+1))); - } - } -}; - -template -struct selfadjoint_rank1_update -{ - static void run(Index size, Scalar* mat, Index stride, const Scalar* vecX, const Scalar* vecY, const Scalar& alpha) - { - selfadjoint_rank1_update::run(size,mat,stride,vecY,vecX,alpha); - } -}; - -template -struct selfadjoint_product_selector; - -template -struct selfadjoint_product_selector -{ - static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha) - { - typedef typename MatrixType::Scalar Scalar; - typedef internal::blas_traits OtherBlasTraits; - typedef typename OtherBlasTraits::DirectLinearAccessType ActualOtherType; - typedef typename internal::remove_all::type _ActualOtherType; - typename internal::add_const_on_value_type::type actualOther = OtherBlasTraits::extract(other.derived()); - - Scalar actualAlpha = alpha * OtherBlasTraits::extractScalarFactor(other.derived()); - - enum { - StorageOrder = (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, - UseOtherDirectly = _ActualOtherType::InnerStrideAtCompileTime==1 - }; - internal::gemv_static_vector_if static_other; - - ei_declare_aligned_stack_constructed_variable(Scalar, actualOtherPtr, other.size(), - (UseOtherDirectly ? const_cast(actualOther.data()) : static_other.data())); - - if(!UseOtherDirectly) - Map(actualOtherPtr, actualOther.size()) = actualOther; - - selfadjoint_rank1_update::IsComplex, - (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex> - ::run(other.size(), mat.data(), mat.outerStride(), actualOtherPtr, actualOtherPtr, actualAlpha); - } -}; - -template -struct selfadjoint_product_selector -{ - static void run(MatrixType& mat, const OtherType& other, const typename MatrixType::Scalar& alpha) - { - typedef typename MatrixType::Scalar Scalar; - typedef internal::blas_traits OtherBlasTraits; - typedef typename OtherBlasTraits::DirectLinearAccessType ActualOtherType; - typedef typename internal::remove_all::type _ActualOtherType; - typename internal::add_const_on_value_type::type actualOther = OtherBlasTraits::extract(other.derived()); - - Scalar actualAlpha = alpha * OtherBlasTraits::extractScalarFactor(other.derived()); - - enum { - IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0, - OtherIsRowMajor = _ActualOtherType::Flags&RowMajorBit ? 1 : 0 - }; - - Index size = mat.cols(); - Index depth = actualOther.cols(); - - typedef internal::gemm_blocking_space BlockingType; - - BlockingType blocking(size, size, depth, 1, false); - - - internal::general_matrix_matrix_triangular_product::IsComplex, - Scalar, OtherIsRowMajor ? ColMajor : RowMajor, (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex, - IsRowMajor ? RowMajor : ColMajor, MatrixType::InnerStrideAtCompileTime, UpLo> - ::run(size, depth, - &actualOther.coeffRef(0,0), actualOther.outerStride(), &actualOther.coeffRef(0,0), actualOther.outerStride(), - mat.data(), mat.innerStride(), mat.outerStride(), actualAlpha, blocking); - } -}; - -// high level API - -template -template -SelfAdjointView& SelfAdjointView -::rankUpdate(const MatrixBase& u, const Scalar& alpha) -{ - selfadjoint_product_selector::run(_expression().const_cast_derived(), u.derived(), alpha); - - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINT_PRODUCT_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h deleted file mode 100644 index 2ae3641111c..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/SelfadjointRank2Update.h +++ /dev/null @@ -1,93 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_SELFADJOINTRANK2UPTADE_H -#define EIGEN_SELFADJOINTRANK2UPTADE_H - -namespace Eigen { - -namespace internal { - -/* Optimized selfadjoint matrix += alpha * uv' + conj(alpha)*vu' - * It corresponds to the Level2 syr2 BLAS routine - */ - -template -struct selfadjoint_rank2_update_selector; - -template -struct selfadjoint_rank2_update_selector -{ - static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha) - { - const Index size = u.size(); - for (Index i=0; i >(mat+stride*i+i, size-i) += - (numext::conj(alpha) * numext::conj(u.coeff(i))) * v.tail(size-i) - + (alpha * numext::conj(v.coeff(i))) * u.tail(size-i); - } - } -}; - -template -struct selfadjoint_rank2_update_selector -{ - static void run(Scalar* mat, Index stride, const UType& u, const VType& v, const Scalar& alpha) - { - const Index size = u.size(); - for (Index i=0; i >(mat+stride*i, i+1) += - (numext::conj(alpha) * numext::conj(u.coeff(i))) * v.head(i+1) - + (alpha * numext::conj(v.coeff(i))) * u.head(i+1); - } -}; - -template struct conj_expr_if - : conditional::Scalar>,T> > {}; - -} // end namespace internal - -template -template -SelfAdjointView& SelfAdjointView -::rankUpdate(const MatrixBase& u, const MatrixBase& v, const Scalar& alpha) -{ - typedef internal::blas_traits UBlasTraits; - typedef typename UBlasTraits::DirectLinearAccessType ActualUType; - typedef typename internal::remove_all::type _ActualUType; - typename internal::add_const_on_value_type::type actualU = UBlasTraits::extract(u.derived()); - - typedef internal::blas_traits VBlasTraits; - typedef typename VBlasTraits::DirectLinearAccessType ActualVType; - typedef typename internal::remove_all::type _ActualVType; - typename internal::add_const_on_value_type::type actualV = VBlasTraits::extract(v.derived()); - - // If MatrixType is row major, then we use the routine for lower triangular in the upper triangular case and - // vice versa, and take the complex conjugate of all coefficients and vector entries. - - enum { IsRowMajor = (internal::traits::Flags&RowMajorBit) ? 1 : 0 }; - Scalar actualAlpha = alpha * UBlasTraits::extractScalarFactor(u.derived()) - * numext::conj(VBlasTraits::extractScalarFactor(v.derived())); - if (IsRowMajor) - actualAlpha = numext::conj(actualAlpha); - - typedef typename internal::remove_all::type>::type UType; - typedef typename internal::remove_all::type>::type VType; - internal::selfadjoint_rank2_update_selector - ::run(_expression().const_cast_derived().data(),_expression().outerStride(),UType(actualU),VType(actualV),actualAlpha); - - return *this; -} - -} // end namespace Eigen - -#endif // EIGEN_SELFADJOINTRANK2UPTADE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h deleted file mode 100644 index 2fb408d1d73..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix.h +++ /dev/null @@ -1,472 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_MATRIX_MATRIX_H -#define EIGEN_TRIANGULAR_MATRIX_MATRIX_H - -namespace Eigen { - -namespace internal { - -// template -// struct gemm_pack_lhs_triangular -// { -// Matrix::IsComplex && Conjugate> cj; -// const_blas_data_mapper lhs(_lhs,lhsStride); -// int count = 0; -// const int peeled_mc = (rows/mr)*mr; -// for(int i=0; i -struct product_triangular_matrix_matrix; - -template -struct product_triangular_matrix_matrix -{ - static EIGEN_STRONG_INLINE void run( - Index rows, Index cols, Index depth, - const Scalar* lhs, Index lhsStride, - const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - product_triangular_matrix_matrix - ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); - } -}; - -// implements col-major += alpha * op(triangular) * op(general) -template -struct product_triangular_matrix_matrix -{ - - typedef gebp_traits Traits; - enum { - SmallPanelWidth = 2 * EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower, - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1 - }; - - static EIGEN_DONT_INLINE void run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - // strip zeros - Index diagSize = (std::min)(_rows,_depth); - Index rows = IsLower ? _rows : diagSize; - Index depth = IsLower ? diagSize : _depth; - Index cols = _cols; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - // The small panel size must not be larger than blocking size. - // Usually this should never be the case because SmallPanelWidth^2 is very small - // compared to L2 cache size, but let's be safe: - Index panelWidth = (std::min)(Index(SmallPanelWidth),(std::min)(kc,mc)); - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - // To work around an "error: member reference base type 'Matrix<...> - // (Eigen::internal::constructor_without_unaligned_array_assert (*)())' is - // not a structure or union" compilation error in nvcc (tested V8.0.61), - // create a dummy internal::constructor_without_unaligned_array_assert - // object to pass to the Matrix constructor. - internal::constructor_without_unaligned_array_assert a; - Matrix triangularBuffer(a); - triangularBuffer.setZero(); - if((Mode&ZeroDiag)==ZeroDiag) - triangularBuffer.diagonal().setZero(); - else - triangularBuffer.diagonal().setOnes(); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - - for(Index k2=IsLower ? depth : 0; - IsLower ? k2>0 : k2rows)) - { - actual_kc = rows-k2; - k2 = k2+actual_kc-kc; - } - - pack_rhs(blockB, rhs.getSubMapper(actual_k2,0), actual_kc, cols); - - // the selected lhs's panel has to be split in three different parts: - // 1 - the part which is zero => skip it - // 2 - the diagonal block => special kernel - // 3 - the dense panel below (lower case) or above (upper case) the diagonal block => GEPP - - // the block diagonal, if any: - if(IsLower || actual_k2(actual_kc-k1, panelWidth); - Index lengthTarget = IsLower ? actual_kc-k1-actualPanelWidth : k1; - Index startBlock = actual_k2+k1; - Index blockBOffset = k1; - - // => GEBP with the micro triangular block - // The trick is to pack this micro block while filling the opposite triangular part with zeros. - // To this end we do an extra triangular copy to a small temporary buffer - for (Index k=0;k0) - { - Index startTarget = IsLower ? actual_k2+k1+actualPanelWidth : actual_k2; - - pack_lhs(blockA, lhs.getSubMapper(startTarget,startBlock), actualPanelWidth, lengthTarget); - - gebp_kernel(res.getSubMapper(startTarget, 0), blockA, blockB, - lengthTarget, actualPanelWidth, cols, alpha, - actualPanelWidth, actual_kc, 0, blockBOffset); - } - } - } - // the part below (lower case) or above (upper case) the diagonal => GEPP - { - Index start = IsLower ? k2 : 0; - Index end = IsLower ? rows : (std::min)(actual_k2,rows); - for(Index i2=start; i2() - (blockA, lhs.getSubMapper(i2, actual_k2), actual_kc, actual_mc); - - gebp_kernel(res.getSubMapper(i2, 0), blockA, blockB, actual_mc, - actual_kc, cols, alpha, -1, -1, 0, 0); - } - } - } - } - -// implements col-major += alpha * op(general) * op(triangular) -template -struct product_triangular_matrix_matrix -{ - typedef gebp_traits Traits; - enum { - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower, - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1 - }; - - static EIGEN_DONT_INLINE void run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking); -}; - -template -EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( - Index _rows, Index _cols, Index _depth, - const Scalar* _lhs, Index lhsStride, - const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, - const Scalar& alpha, level3_blocking& blocking) - { - const Index PacketBytes = packet_traits::size*sizeof(Scalar); - // strip zeros - Index diagSize = (std::min)(_cols,_depth); - Index rows = _rows; - Index depth = IsLower ? _depth : diagSize; - Index cols = IsLower ? diagSize : _cols; - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols+EIGEN_MAX_ALIGN_BYTES/sizeof(Scalar); - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - internal::constructor_without_unaligned_array_assert a; - Matrix triangularBuffer(a); - triangularBuffer.setZero(); - if((Mode&ZeroDiag)==ZeroDiag) - triangularBuffer.diagonal().setZero(); - else - triangularBuffer.diagonal().setOnes(); - - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - gemm_pack_rhs pack_rhs_panel; - - for(Index k2=IsLower ? 0 : depth; - IsLower ? k20; - IsLower ? k2+=kc : k2-=kc) - { - Index actual_kc = (std::min)(IsLower ? depth-k2 : k2, kc); - Index actual_k2 = IsLower ? k2 : k2-actual_kc; - - // align blocks with the end of the triangular part for trapezoidal rhs - if(IsLower && (k2cols)) - { - actual_kc = cols-k2; - k2 = actual_k2 + actual_kc - kc; - } - - // remaining size - Index rs = IsLower ? (std::min)(cols,actual_k2) : cols - k2; - // size of the triangular part - Index ts = (IsLower && actual_k2>=cols) ? 0 : actual_kc; - - Scalar* geb = blockB+ts*ts; - geb = geb + internal::first_aligned(geb,PacketBytes/sizeof(Scalar)); - - pack_rhs(geb, rhs.getSubMapper(actual_k2,IsLower ? 0 : k2), actual_kc, rs); - - // pack the triangular part of the rhs padding the unrolled blocks with zeros - if(ts>0) - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index actual_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc-j2-actualPanelWidth : j2; - // general part - pack_rhs_panel(blockB+j2*actual_kc, - rhs.getSubMapper(actual_k2+panelOffset, actual_j2), - panelLength, actualPanelWidth, - actual_kc, panelOffset); - - // append the triangular part via a temporary buffer - for (Index j=0;j0) - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index panelLength = IsLower ? actual_kc-j2 : j2+actualPanelWidth; - Index blockOffset = IsLower ? j2 : 0; - - gebp_kernel(res.getSubMapper(i2, actual_k2 + j2), - blockA, blockB+j2*actual_kc, - actual_mc, panelLength, actualPanelWidth, - alpha, - actual_kc, actual_kc, // strides - blockOffset, blockOffset);// offsets - } - } - gebp_kernel(res.getSubMapper(i2, IsLower ? 0 : k2), - blockA, geb, actual_mc, actual_kc, rs, - alpha, - -1, -1, 0, 0); - } - } - } - -/*************************************************************************** -* Wrapper to product_triangular_matrix_matrix -***************************************************************************/ - -} // end namespace internal - -namespace internal { -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &a_lhs, const Rhs &a_rhs, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar Scalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef typename internal::remove_all::type ActualLhsTypeCleaned; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - typename internal::add_const_on_value_type::type lhs = LhsBlasTraits::extract(a_lhs); - typename internal::add_const_on_value_type::type rhs = RhsBlasTraits::extract(a_rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(a_lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(a_rhs); - Scalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - typedef internal::gemm_blocking_space<(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor,Scalar,Scalar, - Lhs::MaxRowsAtCompileTime, Rhs::MaxColsAtCompileTime, Lhs::MaxColsAtCompileTime,4> BlockingType; - - enum { IsLower = (Mode&Lower) == Lower }; - Index stripedRows = ((!LhsIsTriangular) || (IsLower)) ? lhs.rows() : (std::min)(lhs.rows(),lhs.cols()); - Index stripedCols = ((LhsIsTriangular) || (!IsLower)) ? rhs.cols() : (std::min)(rhs.cols(),rhs.rows()); - Index stripedDepth = LhsIsTriangular ? ((!IsLower) ? lhs.cols() : (std::min)(lhs.cols(),lhs.rows())) - : ((IsLower) ? rhs.rows() : (std::min)(rhs.rows(),rhs.cols())); - - BlockingType blocking(stripedRows, stripedCols, stripedDepth, 1, false); - - internal::product_triangular_matrix_matrix::Flags&RowMajorBit) ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, Dest::InnerStrideAtCompileTime> - ::run( - stripedRows, stripedCols, stripedDepth, // sizes - &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info - &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info - actualAlpha, blocking - ); - - // Apply correction if the diagonal is unit and a scalar factor was nested: - if ((Mode&UnitDiag)==UnitDiag) - { - if (LhsIsTriangular && lhs_alpha!=LhsScalar(1)) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dst.topRows(diagSize) -= ((lhs_alpha-LhsScalar(1))*a_rhs).topRows(diagSize); - } - else if ((!LhsIsTriangular) && rhs_alpha!=RhsScalar(1)) - { - Index diagSize = (std::min)(rhs.rows(),rhs.cols()); - dst.leftCols(diagSize) -= (rhs_alpha-RhsScalar(1))*a_lhs.leftCols(diagSize); - } - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h deleted file mode 100644 index a98d12e4ae1..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h +++ /dev/null @@ -1,317 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix * matrix product functionality based on ?TRMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H -#define EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - - -template -struct product_triangular_matrix_matrix_trmm : - product_triangular_matrix_matrix {}; - - -// try to go to BLAS specialization -#define EIGEN_BLAS_TRMM_SPECIALIZE(Scalar, LhsIsTriangular) \ -template \ -struct product_triangular_matrix_matrix { \ - static inline void run(Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride,\ - const Scalar* _rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ - product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ - } \ -}; - -EIGEN_BLAS_TRMM_SPECIALIZE(double, true) -EIGEN_BLAS_TRMM_SPECIALIZE(double, false) -EIGEN_BLAS_TRMM_SPECIALIZE(dcomplex, true) -EIGEN_BLAS_TRMM_SPECIALIZE(dcomplex, false) -EIGEN_BLAS_TRMM_SPECIALIZE(float, true) -EIGEN_BLAS_TRMM_SPECIALIZE(float, false) -EIGEN_BLAS_TRMM_SPECIALIZE(scomplex, true) -EIGEN_BLAS_TRMM_SPECIALIZE(scomplex, false) - -// implements col-major += alpha * op(triangular) * op(general) -#define EIGEN_BLAS_TRMM_L(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_triangular_matrix_matrix_trmm \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((LhsStorageOrder==ColMajor) && ConjugateLhs) ? 1 : 0 \ - }; \ -\ - static void run( \ - Index _rows, Index _cols, Index _depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ - EIGTYPE alpha, level3_blocking& blocking) \ - { \ - Index diagSize = (std::min)(_rows,_depth); \ - Index rows = IsLower ? _rows : diagSize; \ - Index depth = IsLower ? diagSize : _depth; \ - Index cols = _cols; \ -\ - typedef Matrix MatrixLhs; \ - typedef Matrix MatrixRhs; \ -\ -/* Non-square case - doesn't fit to BLAS ?TRMM. Fall to default triangular product or call BLAS ?GEMM*/ \ - if (rows != depth) { \ -\ - /* FIXME handle mkl_domain_get_max_threads */ \ - /*int nthr = mkl_domain_get_max_threads(EIGEN_BLAS_DOMAIN_BLAS);*/ int nthr = 1;\ -\ - if (((nthr==1) && (((std::max)(rows,depth)-diagSize)/(double)diagSize < 0.5))) { \ - /* Most likely no benefit to call TRMM or GEMM from BLAS */ \ - product_triangular_matrix_matrix::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ - /*std::cout << "TRMM_L: A is not square! Go to Eigen TRMM implementation!\n";*/ \ - } else { \ - /* Make sense to call GEMM */ \ - Map > lhsMap(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixLhs aa_tmp=lhsMap.template triangularView(); \ - BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ - gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, 1, resStride, alpha, gemm_blocking, 0); \ -\ - /*std::cout << "TRMM_L: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ - } \ - return; \ - } \ - char side = 'L', transa, uplo, diag = 'N'; \ - EIGTYPE *b; \ - const EIGTYPE *a; \ - BlasIndex m, n, lda, ldb; \ -\ -/* Set m, n */ \ - m = convert_index(diagSize); \ - n = convert_index(cols); \ -\ -/* Set trans */ \ - transa = (LhsStorageOrder==RowMajor) ? ((ConjugateLhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set b, ldb */ \ - Map > rhs(_rhs,depth,cols,OuterStride<>(rhsStride)); \ - MatrixX##EIGPREFIX b_tmp; \ -\ - if (ConjugateRhs) b_tmp = rhs.conjugate(); else b_tmp = rhs; \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ -\ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (LhsStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - Map > lhs(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixLhs a_tmp; \ -\ - if ((conjA!=0) || (SetDiag==0)) { \ - if (conjA) a_tmp = lhs.conjugate(); else a_tmp = lhs; \ - if (IsZeroDiag) \ - a_tmp.diagonal().setZero(); \ - else if (IsUnitDiag) \ - a_tmp.diagonal().setOnes();\ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _lhs; \ - lda = convert_index(lhsStride); \ - } \ - /*std::cout << "TRMM_L: A is square! Go to BLAS TRMM implementation! \n";*/ \ -/* call ?trmm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)b, &ldb); \ -\ -/* Add op(a_triangular)*b into res*/ \ - Map > res_tmp(res,rows,cols,OuterStride<>(resStride)); \ - res_tmp=res_tmp+b_tmp; \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMM_L(double, double, d, dtrmm) -EIGEN_BLAS_TRMM_L(dcomplex, MKL_Complex16, cd, ztrmm) -EIGEN_BLAS_TRMM_L(float, float, f, strmm) -EIGEN_BLAS_TRMM_L(scomplex, MKL_Complex8, cf, ctrmm) -#else -EIGEN_BLAS_TRMM_L(double, double, d, dtrmm_) -EIGEN_BLAS_TRMM_L(dcomplex, double, cd, ztrmm_) -EIGEN_BLAS_TRMM_L(float, float, f, strmm_) -EIGEN_BLAS_TRMM_L(scomplex, float, cf, ctrmm_) -#endif - -// implements col-major += alpha * op(general) * op(triangular) -#define EIGEN_BLAS_TRMM_R(EIGTYPE, BLASTYPE, EIGPREFIX, BLASFUNC) \ -template \ -struct product_triangular_matrix_matrix_trmm \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper, \ - conjA = ((RhsStorageOrder==ColMajor) && ConjugateRhs) ? 1 : 0 \ - }; \ -\ - static void run( \ - Index _rows, Index _cols, Index _depth, \ - const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ - EIGTYPE alpha, level3_blocking& blocking) \ - { \ - Index diagSize = (std::min)(_cols,_depth); \ - Index rows = _rows; \ - Index depth = IsLower ? _depth : diagSize; \ - Index cols = IsLower ? diagSize : _cols; \ -\ - typedef Matrix MatrixLhs; \ - typedef Matrix MatrixRhs; \ -\ -/* Non-square case - doesn't fit to BLAS ?TRMM. Fall to default triangular product or call BLAS ?GEMM*/ \ - if (cols != depth) { \ -\ - int nthr = 1 /*mkl_domain_get_max_threads(EIGEN_BLAS_DOMAIN_BLAS)*/; \ -\ - if ((nthr==1) && (((std::max)(cols,depth)-diagSize)/(double)diagSize < 0.5)) { \ - /* Most likely no benefit to call TRMM or GEMM from BLAS*/ \ - product_triangular_matrix_matrix::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ - /*std::cout << "TRMM_R: A is not square! Go to Eigen TRMM implementation!\n";*/ \ - } else { \ - /* Make sense to call GEMM */ \ - Map > rhsMap(_rhs,depth,cols, OuterStride<>(rhsStride)); \ - MatrixRhs aa_tmp=rhsMap.template triangularView(); \ - BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ - gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, 1, resStride, alpha, gemm_blocking, 0); \ -\ - /*std::cout << "TRMM_R: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ - } \ - return; \ - } \ - char side = 'R', transa, uplo, diag = 'N'; \ - EIGTYPE *b; \ - const EIGTYPE *a; \ - BlasIndex m, n, lda, ldb; \ -\ -/* Set m, n */ \ - m = convert_index(rows); \ - n = convert_index(diagSize); \ -\ -/* Set trans */ \ - transa = (RhsStorageOrder==RowMajor) ? ((ConjugateRhs) ? 'C' : 'T') : 'N'; \ -\ -/* Set b, ldb */ \ - Map > lhs(_lhs,rows,depth,OuterStride<>(lhsStride)); \ - MatrixX##EIGPREFIX b_tmp; \ -\ - if (ConjugateLhs) b_tmp = lhs.conjugate(); else b_tmp = lhs; \ - b = b_tmp.data(); \ - ldb = convert_index(b_tmp.outerStride()); \ -\ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (RhsStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - Map > rhs(_rhs,depth,cols, OuterStride<>(rhsStride)); \ - MatrixRhs a_tmp; \ -\ - if ((conjA!=0) || (SetDiag==0)) { \ - if (conjA) a_tmp = rhs.conjugate(); else a_tmp = rhs; \ - if (IsZeroDiag) \ - a_tmp.diagonal().setZero(); \ - else if (IsUnitDiag) \ - a_tmp.diagonal().setOnes();\ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _rhs; \ - lda = convert_index(rhsStride); \ - } \ - /*std::cout << "TRMM_R: A is square! Go to BLAS TRMM implementation! \n";*/ \ -/* call ?trmm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)b, &ldb); \ -\ -/* Add op(a_triangular)*b into res*/ \ - Map > res_tmp(res,rows,cols,OuterStride<>(resStride)); \ - res_tmp=res_tmp+b_tmp; \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMM_R(double, double, d, dtrmm) -EIGEN_BLAS_TRMM_R(dcomplex, MKL_Complex16, cd, ztrmm) -EIGEN_BLAS_TRMM_R(float, float, f, strmm) -EIGEN_BLAS_TRMM_R(scomplex, MKL_Complex8, cf, ctrmm) -#else -EIGEN_BLAS_TRMM_R(double, double, d, dtrmm_) -EIGEN_BLAS_TRMM_R(dcomplex, double, cd, ztrmm_) -EIGEN_BLAS_TRMM_R(float, float, f, strmm_) -EIGEN_BLAS_TRMM_R(scomplex, float, cf, ctrmm_) -#endif -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h deleted file mode 100644 index 76bfa159ced..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector.h +++ /dev/null @@ -1,350 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULARMATRIXVECTOR_H -#define EIGEN_TRIANGULARMATRIXVECTOR_H - -namespace Eigen { - -namespace internal { - -template -struct triangular_matrix_vector_product; - -template -struct triangular_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - enum { - IsLower = ((Mode&Lower)==Lower), - HasUnitDiag = (Mode & UnitDiag)==UnitDiag, - HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag - }; - static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const RhsScalar& alpha); -}; - -template -EIGEN_DONT_INLINE void triangular_matrix_vector_product - ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const RhsScalar& alpha) - { - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - Index size = (std::min)(_rows,_cols); - Index rows = IsLower ? _rows : (std::min)(_rows,_cols); - Index cols = IsLower ? (std::min)(_rows,_cols) : _cols; - - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,rows,cols,OuterStride<>(lhsStride)); - typename conj_expr_if::type cjLhs(lhs); - - typedef Map, 0, InnerStride<> > RhsMap; - const RhsMap rhs(_rhs,cols,InnerStride<>(rhsIncr)); - typename conj_expr_if::type cjRhs(rhs); - - typedef Map > ResMap; - ResMap res(_res,rows); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - for (Index pi=0; pi0) - res.segment(s,r) += (alpha * cjRhs.coeff(i)) * cjLhs.col(i).segment(s,r); - if (HasUnitDiag) - res.coeffRef(i) += alpha * cjRhs.coeff(i); - } - Index r = IsLower ? rows - pi - actualPanelWidth : pi; - if (r>0) - { - Index s = IsLower ? pi+actualPanelWidth : 0; - general_matrix_vector_product::run( - r, actualPanelWidth, - LhsMapper(&lhs.coeffRef(s,pi), lhsStride), - RhsMapper(&rhs.coeffRef(pi), rhsIncr), - &res.coeffRef(s), resIncr, alpha); - } - } - if((!IsLower) && cols>size) - { - general_matrix_vector_product::run( - rows, cols-size, - LhsMapper(&lhs.coeffRef(0,size), lhsStride), - RhsMapper(&rhs.coeffRef(size), rhsIncr), - _res, resIncr, alpha); - } - } - -template -struct triangular_matrix_vector_product -{ - typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; - enum { - IsLower = ((Mode&Lower)==Lower), - HasUnitDiag = (Mode & UnitDiag)==UnitDiag, - HasZeroDiag = (Mode & ZeroDiag)==ZeroDiag - }; - static EIGEN_DONT_INLINE void run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha); -}; - -template -EIGEN_DONT_INLINE void triangular_matrix_vector_product - ::run(Index _rows, Index _cols, const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsIncr, ResScalar* _res, Index resIncr, const ResScalar& alpha) - { - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - Index diagSize = (std::min)(_rows,_cols); - Index rows = IsLower ? _rows : diagSize; - Index cols = IsLower ? diagSize : _cols; - - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,rows,cols,OuterStride<>(lhsStride)); - typename conj_expr_if::type cjLhs(lhs); - - typedef Map > RhsMap; - const RhsMap rhs(_rhs,cols); - typename conj_expr_if::type cjRhs(rhs); - - typedef Map, 0, InnerStride<> > ResMap; - ResMap res(_res,rows,InnerStride<>(resIncr)); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - for (Index pi=0; pi0) - res.coeffRef(i) += alpha * (cjLhs.row(i).segment(s,r).cwiseProduct(cjRhs.segment(s,r).transpose())).sum(); - if (HasUnitDiag) - res.coeffRef(i) += alpha * cjRhs.coeff(i); - } - Index r = IsLower ? pi : cols - pi - actualPanelWidth; - if (r>0) - { - Index s = IsLower ? 0 : pi + actualPanelWidth; - general_matrix_vector_product::run( - actualPanelWidth, r, - LhsMapper(&lhs.coeffRef(pi,s), lhsStride), - RhsMapper(&rhs.coeffRef(s), rhsIncr), - &res.coeffRef(pi), resIncr, alpha); - } - } - if(IsLower && rows>diagSize) - { - general_matrix_vector_product::run( - rows-diagSize, cols, - LhsMapper(&lhs.coeffRef(diagSize,0), lhsStride), - RhsMapper(&rhs.coeffRef(0), rhsIncr), - &res.coeffRef(diagSize), resIncr, alpha); - } - } - -/*************************************************************************** -* Wrapper to product_triangular_vector -***************************************************************************/ - -template -struct trmv_selector; - -} // end namespace internal - -namespace internal { - -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &lhs, const Rhs &rhs, const typename Dest::Scalar& alpha) - { - eigen_assert(dst.rows()==lhs.rows() && dst.cols()==rhs.cols()); - - internal::trmv_selector::Flags)&RowMajorBit) ? RowMajor : ColMajor>::run(lhs, rhs, dst, alpha); - } -}; - -template -struct triangular_product_impl -{ - template static void run(Dest& dst, const Lhs &lhs, const Rhs &rhs, const typename Dest::Scalar& alpha) - { - eigen_assert(dst.rows()==lhs.rows() && dst.cols()==rhs.cols()); - - Transpose dstT(dst); - internal::trmv_selector<(Mode & (UnitDiag|ZeroDiag)) | ((Mode & Lower) ? Upper : Lower), - (int(internal::traits::Flags)&RowMajorBit) ? ColMajor : RowMajor> - ::run(rhs.transpose(),lhs.transpose(), dstT, alpha); - } -}; - -} // end namespace internal - -namespace internal { - -// TODO: find a way to factorize this piece of code with gemv_selector since the logic is exactly the same. - -template struct trmv_selector -{ - template - static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar ResScalar; - typedef typename Dest::RealScalar RealScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - - typedef Map, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits::size)> MappedDest; - - typename internal::add_const_on_value_type::type actualLhs = LhsBlasTraits::extract(lhs); - typename internal::add_const_on_value_type::type actualRhs = RhsBlasTraits::extract(rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(rhs); - ResScalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - enum { - // FIXME find a way to allow an inner stride on the result if packet_traits::size==1 - // on, the other hand it is good for the cache to pack the vector anyways... - EvalToDestAtCompileTime = Dest::InnerStrideAtCompileTime==1, - ComplexByReal = (NumTraits::IsComplex) && (!NumTraits::IsComplex), - MightCannotUseDest = (Dest::InnerStrideAtCompileTime!=1) || ComplexByReal - }; - - gemv_static_vector_if static_dest; - - bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0)); - bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible; - - RhsScalar compatibleAlpha = get_factor::run(actualAlpha); - - ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(), - evalToDest ? dest.data() : static_dest.data()); - - if(!evalToDest) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = dest.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - if(!alphaIsCompatible) - { - MappedDest(actualDestPtr, dest.size()).setZero(); - compatibleAlpha = RhsScalar(1); - } - else - MappedDest(actualDestPtr, dest.size()) = dest; - } - - internal::triangular_matrix_vector_product - - ::run(actualLhs.rows(),actualLhs.cols(), - actualLhs.data(),actualLhs.outerStride(), - actualRhs.data(),actualRhs.innerStride(), - actualDestPtr,1,compatibleAlpha); - - if (!evalToDest) - { - if(!alphaIsCompatible) - dest += actualAlpha * MappedDest(actualDestPtr, dest.size()); - else - dest = MappedDest(actualDestPtr, dest.size()); - } - - if ( ((Mode&UnitDiag)==UnitDiag) && (lhs_alpha!=LhsScalar(1)) ) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dest.head(diagSize) -= (lhs_alpha-LhsScalar(1))*rhs.head(diagSize); - } - } -}; - -template struct trmv_selector -{ - template - static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha) - { - typedef typename Lhs::Scalar LhsScalar; - typedef typename Rhs::Scalar RhsScalar; - typedef typename Dest::Scalar ResScalar; - - typedef internal::blas_traits LhsBlasTraits; - typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType; - typedef internal::blas_traits RhsBlasTraits; - typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType; - typedef typename internal::remove_all::type ActualRhsTypeCleaned; - - typename add_const::type actualLhs = LhsBlasTraits::extract(lhs); - typename add_const::type actualRhs = RhsBlasTraits::extract(rhs); - - LhsScalar lhs_alpha = LhsBlasTraits::extractScalarFactor(lhs); - RhsScalar rhs_alpha = RhsBlasTraits::extractScalarFactor(rhs); - ResScalar actualAlpha = alpha * lhs_alpha * rhs_alpha; - - enum { - DirectlyUseRhs = ActualRhsTypeCleaned::InnerStrideAtCompileTime==1 - }; - - gemv_static_vector_if static_rhs; - - ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,actualRhs.size(), - DirectlyUseRhs ? const_cast(actualRhs.data()) : static_rhs.data()); - - if(!DirectlyUseRhs) - { - #ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN - Index size = actualRhs.size(); - EIGEN_DENSE_STORAGE_CTOR_PLUGIN - #endif - Map(actualRhsPtr, actualRhs.size()) = actualRhs; - } - - internal::triangular_matrix_vector_product - - ::run(actualLhs.rows(),actualLhs.cols(), - actualLhs.data(),actualLhs.outerStride(), - actualRhsPtr,1, - dest.data(),dest.innerStride(), - actualAlpha); - - if ( ((Mode&UnitDiag)==UnitDiag) && (lhs_alpha!=LhsScalar(1)) ) - { - Index diagSize = (std::min)(lhs.rows(),lhs.cols()); - dest.head(diagSize) -= (lhs_alpha-LhsScalar(1))*rhs.head(diagSize); - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULARMATRIXVECTOR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h deleted file mode 100644 index 3d47a2b94cb..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularMatrixVector_BLAS.h +++ /dev/null @@ -1,255 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix-vector product functionality based on ?TRMV. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H -#define EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H - -namespace Eigen { - -namespace internal { - -/********************************************************************** -* This file implements triangular matrix-vector multiplication using BLAS -**********************************************************************/ - -// trmv/hemv specialization - -template -struct triangular_matrix_vector_product_trmv : - triangular_matrix_vector_product {}; - -#define EIGEN_BLAS_TRMV_SPECIALIZE(Scalar) \ -template \ -struct triangular_matrix_vector_product { \ - static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \ - const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \ - triangular_matrix_vector_product_trmv::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - } \ -}; \ -template \ -struct triangular_matrix_vector_product { \ - static void run(Index _rows, Index _cols, const Scalar* _lhs, Index lhsStride, \ - const Scalar* _rhs, Index rhsIncr, Scalar* _res, Index resIncr, Scalar alpha) { \ - triangular_matrix_vector_product_trmv::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - } \ -}; - -EIGEN_BLAS_TRMV_SPECIALIZE(double) -EIGEN_BLAS_TRMV_SPECIALIZE(float) -EIGEN_BLAS_TRMV_SPECIALIZE(dcomplex) -EIGEN_BLAS_TRMV_SPECIALIZE(scomplex) - -// implements col-major: res += alpha * op(triangular) * vector -#define EIGEN_BLAS_TRMV_CM(EIGTYPE, BLASTYPE, EIGPREFIX, BLASPREFIX, BLASPOSTFIX) \ -template \ -struct triangular_matrix_vector_product_trmv { \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper \ - }; \ - static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \ - { \ - if (ConjLhs || IsZeroDiag) { \ - triangular_matrix_vector_product::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - return; \ - }\ - Index size = (std::min)(_rows,_cols); \ - Index rows = IsLower ? _rows : size; \ - Index cols = IsLower ? size : _cols; \ -\ - typedef VectorX##EIGPREFIX VectorRhs; \ - EIGTYPE *x, *y;\ -\ -/* Set x*/ \ - Map > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \ - VectorRhs x_tmp; \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ -\ -/* Square part handling */\ -\ - char trans, uplo, diag; \ - BlasIndex m, n, lda, incx, incy; \ - EIGTYPE const *a; \ - EIGTYPE beta(1); \ -\ -/* Set m, n */ \ - n = convert_index(size); \ - lda = convert_index(lhsStride); \ - incx = 1; \ - incy = convert_index(resIncr); \ -\ -/* Set uplo, trans and diag*/ \ - trans = 'N'; \ - uplo = IsLower ? 'L' : 'U'; \ - diag = IsUnitDiag ? 'U' : 'N'; \ -\ -/* call ?TRMV*/ \ - BLASPREFIX##trmv##BLASPOSTFIX(&uplo, &trans, &diag, &n, (const BLASTYPE*)_lhs, &lda, (BLASTYPE*)x, &incx); \ -\ -/* Add op(a_tr)rhs into res*/ \ - BLASPREFIX##axpy##BLASPOSTFIX(&n, (const BLASTYPE*)&numext::real_ref(alpha),(const BLASTYPE*)x, &incx, (BLASTYPE*)_res, &incy); \ -/* Non-square case - doesn't fit to BLAS ?TRMV. Fall to default triangular product*/ \ - if (size<(std::max)(rows,cols)) { \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ - if (size(rows-size); \ - n = convert_index(size); \ - } \ - else { \ - x += size; \ - y = _res; \ - a = _lhs + size*lda; \ - m = convert_index(size); \ - n = convert_index(cols-size); \ - } \ - BLASPREFIX##gemv##BLASPOSTFIX(&trans, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)x, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)y, &incy); \ - } \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMV_CM(double, double, d, d,) -EIGEN_BLAS_TRMV_CM(dcomplex, MKL_Complex16, cd, z,) -EIGEN_BLAS_TRMV_CM(float, float, f, s,) -EIGEN_BLAS_TRMV_CM(scomplex, MKL_Complex8, cf, c,) -#else -EIGEN_BLAS_TRMV_CM(double, double, d, d, _) -EIGEN_BLAS_TRMV_CM(dcomplex, double, cd, z, _) -EIGEN_BLAS_TRMV_CM(float, float, f, s, _) -EIGEN_BLAS_TRMV_CM(scomplex, float, cf, c, _) -#endif - -// implements row-major: res += alpha * op(triangular) * vector -#define EIGEN_BLAS_TRMV_RM(EIGTYPE, BLASTYPE, EIGPREFIX, BLASPREFIX, BLASPOSTFIX) \ -template \ -struct triangular_matrix_vector_product_trmv { \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - SetDiag = (Mode&(ZeroDiag|UnitDiag)) ? 0 : 1, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - LowUp = IsLower ? Lower : Upper \ - }; \ - static void run(Index _rows, Index _cols, const EIGTYPE* _lhs, Index lhsStride, \ - const EIGTYPE* _rhs, Index rhsIncr, EIGTYPE* _res, Index resIncr, EIGTYPE alpha) \ - { \ - if (IsZeroDiag) { \ - triangular_matrix_vector_product::run( \ - _rows, _cols, _lhs, lhsStride, _rhs, rhsIncr, _res, resIncr, alpha); \ - return; \ - }\ - Index size = (std::min)(_rows,_cols); \ - Index rows = IsLower ? _rows : size; \ - Index cols = IsLower ? size : _cols; \ -\ - typedef VectorX##EIGPREFIX VectorRhs; \ - EIGTYPE *x, *y;\ -\ -/* Set x*/ \ - Map > rhs(_rhs,cols,InnerStride<>(rhsIncr)); \ - VectorRhs x_tmp; \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ -\ -/* Square part handling */\ -\ - char trans, uplo, diag; \ - BlasIndex m, n, lda, incx, incy; \ - EIGTYPE const *a; \ - EIGTYPE beta(1); \ -\ -/* Set m, n */ \ - n = convert_index(size); \ - lda = convert_index(lhsStride); \ - incx = 1; \ - incy = convert_index(resIncr); \ -\ -/* Set uplo, trans and diag*/ \ - trans = ConjLhs ? 'C' : 'T'; \ - uplo = IsLower ? 'U' : 'L'; \ - diag = IsUnitDiag ? 'U' : 'N'; \ -\ -/* call ?TRMV*/ \ - BLASPREFIX##trmv##BLASPOSTFIX(&uplo, &trans, &diag, &n, (const BLASTYPE*)_lhs, &lda, (BLASTYPE*)x, &incx); \ -\ -/* Add op(a_tr)rhs into res*/ \ - BLASPREFIX##axpy##BLASPOSTFIX(&n, (const BLASTYPE*)&numext::real_ref(alpha),(const BLASTYPE*)x, &incx, (BLASTYPE*)_res, &incy); \ -/* Non-square case - doesn't fit to BLAS ?TRMV. Fall to default triangular product*/ \ - if (size<(std::max)(rows,cols)) { \ - if (ConjRhs) x_tmp = rhs.conjugate(); else x_tmp = rhs; \ - x = x_tmp.data(); \ - if (size(rows-size); \ - n = convert_index(size); \ - } \ - else { \ - x += size; \ - y = _res; \ - a = _lhs + size; \ - m = convert_index(size); \ - n = convert_index(cols-size); \ - } \ - BLASPREFIX##gemv##BLASPOSTFIX(&trans, &n, &m, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (const BLASTYPE*)x, &incx, (const BLASTYPE*)&numext::real_ref(beta), (BLASTYPE*)y, &incy); \ - } \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRMV_RM(double, double, d, d,) -EIGEN_BLAS_TRMV_RM(dcomplex, MKL_Complex16, cd, z,) -EIGEN_BLAS_TRMV_RM(float, float, f, s,) -EIGEN_BLAS_TRMV_RM(scomplex, MKL_Complex8, cf, c,) -#else -EIGEN_BLAS_TRMV_RM(double, double, d, d,_) -EIGEN_BLAS_TRMV_RM(dcomplex, double, cd, z,_) -EIGEN_BLAS_TRMV_RM(float, float, f, s,_) -EIGEN_BLAS_TRMV_RM(scomplex, float, cf, c,_) -#endif - -} // end namespase internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_MATRIX_VECTOR_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h deleted file mode 100644 index e3ed2cd19ea..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix.h +++ /dev/null @@ -1,335 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_SOLVER_MATRIX_H -#define EIGEN_TRIANGULAR_SOLVER_MATRIX_H - -namespace Eigen { - -namespace internal { - -// if the rhs is row major, let's transpose the product -template -struct triangular_solve_matrix -{ - static void run( - Index size, Index cols, - const Scalar* tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - triangular_solve_matrix< - Scalar, Index, Side==OnTheLeft?OnTheRight:OnTheLeft, - (Mode&UnitDiag) | ((Mode&Upper) ? Lower : Upper), - NumTraits::IsComplex && Conjugate, - TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor, OtherInnerStride> - ::run(size, cols, tri, triStride, _other, otherIncr, otherStride, blocking); - } -}; - -/* Optimized triangular solver with multiple right hand side and the triangular matrix on the left - */ -template -struct triangular_solve_matrix -{ - static EIGEN_DONT_INLINE void run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking); -}; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - Index cols = otherSize; - - typedef const_blas_data_mapper TriMapper; - typedef blas_data_mapper OtherMapper; - TriMapper tri(_tri, triStride); - OtherMapper other(_other, otherStride, otherIncr); - - typedef gebp_traits Traits; - - enum { - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower - }; - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(size,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*cols; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - conj_if conj; - gebp_kernel gebp_kernel; - gemm_pack_lhs pack_lhs; - gemm_pack_rhs pack_rhs; - - // the goal here is to subdivise the Rhs panels such that we keep some cache - // coherence when accessing the rhs elements - std::ptrdiff_t l1, l2, l3; - manage_caching_sizes(GetAction, &l1, &l2, &l3); - Index subcols = cols>0 ? l2/(4 * sizeof(Scalar) * std::max(otherStride,size)) : 0; - subcols = std::max((subcols/Traits::nr)*Traits::nr, Traits::nr); - - for(Index k2=IsLower ? 0 : size; - IsLower ? k20; - IsLower ? k2+=kc : k2-=kc) - { - const Index actual_kc = (std::min)(IsLower ? size-k2 : k2, kc); - - // We have selected and packed a big horizontal panel R1 of rhs. Let B be the packed copy of this panel, - // and R2 the remaining part of rhs. The corresponding vertical panel of lhs is split into - // A11 (the triangular part) and A21 the remaining rectangular part. - // Then the high level algorithm is: - // - B = R1 => general block copy (done during the next step) - // - R1 = A11^-1 B => tricky part - // - update B from the new R1 => actually this has to be performed continuously during the above step - // - R2 -= A21 * B => GEPP - - // The tricky part: compute R1 = A11^-1 B while updating B from R1 - // The idea is to split A11 into multiple small vertical panels. - // Each panel can be split into a small triangular part T1k which is processed without optimization, - // and the remaining small part T2k which is processed using gebp with appropriate block strides - for(Index j2=0; j2(actual_kc-k1, SmallPanelWidth); - // tr solve - for (Index k=0; k0) - { - Index startTarget = IsLower ? k2+k1+actualPanelWidth : k2-actual_kc; - - pack_lhs(blockA, tri.getSubMapper(startTarget,startBlock), actualPanelWidth, lengthTarget); - - gebp_kernel(other.getSubMapper(startTarget,j2), blockA, blockB+actual_kc*j2, lengthTarget, actualPanelWidth, actual_cols, Scalar(-1), - actualPanelWidth, actual_kc, 0, blockBOffset); - } - } - } - - // R2 -= A21 * B => GEPP - { - Index start = IsLower ? k2+kc : 0; - Index end = IsLower ? size : k2-kc; - for(Index i2=start; i20) - { - pack_lhs(blockA, tri.getSubMapper(i2, IsLower ? k2 : k2-kc), actual_kc, actual_mc); - - gebp_kernel(other.getSubMapper(i2, 0), blockA, blockB, actual_mc, actual_kc, cols, Scalar(-1), -1, -1, 0, 0); - } - } - } - } - } - -/* Optimized triangular solver with multiple left hand sides and the triangular matrix on the right - */ -template -struct triangular_solve_matrix -{ - static EIGEN_DONT_INLINE void run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking); -}; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( - Index size, Index otherSize, - const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, - level3_blocking& blocking) - { - Index rows = otherSize; - typedef typename NumTraits::Real RealScalar; - - typedef blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - LhsMapper lhs(_other, otherStride, otherIncr); - RhsMapper rhs(_tri, triStride); - - typedef gebp_traits Traits; - enum { - RhsStorageOrder = TriStorageOrder, - SmallPanelWidth = EIGEN_PLAIN_ENUM_MAX(Traits::mr,Traits::nr), - IsLower = (Mode&Lower) == Lower - }; - - Index kc = blocking.kc(); // cache block size along the K direction - Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction - - std::size_t sizeA = kc*mc; - std::size_t sizeB = kc*size; - - ei_declare_aligned_stack_constructed_variable(Scalar, blockA, sizeA, blocking.blockA()); - ei_declare_aligned_stack_constructed_variable(Scalar, blockB, sizeB, blocking.blockB()); - - conj_if conj; - gebp_kernel gebp_kernel; - gemm_pack_rhs pack_rhs; - gemm_pack_rhs pack_rhs_panel; - gemm_pack_lhs pack_lhs_panel; - - for(Index k2=IsLower ? size : 0; - IsLower ? k2>0 : k20) pack_rhs(geb, rhs.getSubMapper(actual_k2,startPanel), actual_kc, rs); - - // triangular packing (we only pack the panels off the diagonal, - // neglecting the blocks overlapping the diagonal - { - for (Index j2=0; j2(actual_kc-j2, SmallPanelWidth); - Index actual_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc-j2-actualPanelWidth : j2; - - if (panelLength>0) - pack_rhs_panel(blockB+j2*actual_kc, - rhs.getSubMapper(actual_k2+panelOffset, actual_j2), - panelLength, actualPanelWidth, - actual_kc, panelOffset); - } - } - - for(Index i2=0; i2 vertical panels of rhs) - for (Index j2 = IsLower - ? (actual_kc - ((actual_kc%SmallPanelWidth) ? Index(actual_kc%SmallPanelWidth) - : Index(SmallPanelWidth))) - : 0; - IsLower ? j2>=0 : j2(actual_kc-j2, SmallPanelWidth); - Index absolute_j2 = actual_k2 + j2; - Index panelOffset = IsLower ? j2+actualPanelWidth : 0; - Index panelLength = IsLower ? actual_kc - j2 - actualPanelWidth : j2; - - // GEBP - if(panelLength>0) - { - gebp_kernel(lhs.getSubMapper(i2,absolute_j2), - blockA, blockB+j2*actual_kc, - actual_mc, panelLength, actualPanelWidth, - Scalar(-1), - actual_kc, actual_kc, // strides - panelOffset, panelOffset); // offsets - } - - // unblocked triangular solve - for (Index k=0; k0) - gebp_kernel(lhs.getSubMapper(i2, startPanel), blockA, geb, - actual_mc, actual_kc, rs, Scalar(-1), - -1, -1, 0, 0); - } - } - } - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_MATRIX_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h deleted file mode 100644 index 621194ce6ca..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverMatrix_BLAS.h +++ /dev/null @@ -1,167 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to BLAS F77 - * Triangular matrix * matrix product functionality based on ?TRMM. - ******************************************************************************** -*/ - -#ifndef EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H -#define EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H - -namespace Eigen { - -namespace internal { - -// implements LeftSide op(triangular)^-1 * general -#define EIGEN_BLAS_TRSM_L(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct triangular_solve_matrix \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - conjA = ((TriStorageOrder==ColMajor) && Conjugate) ? 1 : 0 \ - }; \ - static void run( \ - Index size, Index otherSize, \ - const EIGTYPE* _tri, Index triStride, \ - EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ - BlasIndex m = convert_index(size), n = convert_index(otherSize), lda, ldb; \ - char side = 'L', uplo, diag='N', transa; \ - /* Set alpha_ */ \ - EIGTYPE alpha(1); \ - ldb = convert_index(otherStride);\ -\ - const EIGTYPE *a; \ -/* Set trans */ \ - transa = (TriStorageOrder==RowMajor) ? ((Conjugate) ? 'C' : 'T') : 'N'; \ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (TriStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - typedef Matrix MatrixTri; \ - Map > tri(_tri,size,size,OuterStride<>(triStride)); \ - MatrixTri a_tmp; \ -\ - if (conjA) { \ - a_tmp = tri.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _tri; \ - lda = convert_index(triStride); \ - } \ - if (IsUnitDiag) diag='U'; \ -/* call ?trsm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)_other, &ldb); \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRSM_L(double, double, dtrsm) -EIGEN_BLAS_TRSM_L(dcomplex, MKL_Complex16, ztrsm) -EIGEN_BLAS_TRSM_L(float, float, strsm) -EIGEN_BLAS_TRSM_L(scomplex, MKL_Complex8, ctrsm) -#else -EIGEN_BLAS_TRSM_L(double, double, dtrsm_) -EIGEN_BLAS_TRSM_L(dcomplex, double, ztrsm_) -EIGEN_BLAS_TRSM_L(float, float, strsm_) -EIGEN_BLAS_TRSM_L(scomplex, float, ctrsm_) -#endif - -// implements RightSide general * op(triangular)^-1 -#define EIGEN_BLAS_TRSM_R(EIGTYPE, BLASTYPE, BLASFUNC) \ -template \ -struct triangular_solve_matrix \ -{ \ - enum { \ - IsLower = (Mode&Lower) == Lower, \ - IsUnitDiag = (Mode&UnitDiag) ? 1 : 0, \ - IsZeroDiag = (Mode&ZeroDiag) ? 1 : 0, \ - conjA = ((TriStorageOrder==ColMajor) && Conjugate) ? 1 : 0 \ - }; \ - static void run( \ - Index size, Index otherSize, \ - const EIGTYPE* _tri, Index triStride, \ - EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ - { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ - BlasIndex m = convert_index(otherSize), n = convert_index(size), lda, ldb; \ - char side = 'R', uplo, diag='N', transa; \ - /* Set alpha_ */ \ - EIGTYPE alpha(1); \ - ldb = convert_index(otherStride);\ -\ - const EIGTYPE *a; \ -/* Set trans */ \ - transa = (TriStorageOrder==RowMajor) ? ((Conjugate) ? 'C' : 'T') : 'N'; \ -/* Set uplo */ \ - uplo = IsLower ? 'L' : 'U'; \ - if (TriStorageOrder==RowMajor) uplo = (uplo == 'L') ? 'U' : 'L'; \ -/* Set a, lda */ \ - typedef Matrix MatrixTri; \ - Map > tri(_tri,size,size,OuterStride<>(triStride)); \ - MatrixTri a_tmp; \ -\ - if (conjA) { \ - a_tmp = tri.conjugate(); \ - a = a_tmp.data(); \ - lda = convert_index(a_tmp.outerStride()); \ - } else { \ - a = _tri; \ - lda = convert_index(triStride); \ - } \ - if (IsUnitDiag) diag='U'; \ -/* call ?trsm*/ \ - BLASFUNC(&side, &uplo, &transa, &diag, &m, &n, (const BLASTYPE*)&numext::real_ref(alpha), (const BLASTYPE*)a, &lda, (BLASTYPE*)_other, &ldb); \ - /*std::cout << "TRMS_L specialization!\n";*/ \ - } \ -}; - -#ifdef EIGEN_USE_MKL -EIGEN_BLAS_TRSM_R(double, double, dtrsm) -EIGEN_BLAS_TRSM_R(dcomplex, MKL_Complex16, ztrsm) -EIGEN_BLAS_TRSM_R(float, float, strsm) -EIGEN_BLAS_TRSM_R(scomplex, MKL_Complex8, ctrsm) -#else -EIGEN_BLAS_TRSM_R(double, double, dtrsm_) -EIGEN_BLAS_TRSM_R(dcomplex, double, ztrsm_) -EIGEN_BLAS_TRSM_R(float, float, strsm_) -EIGEN_BLAS_TRSM_R(scomplex, float, ctrsm_) -#endif - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_MATRIX_BLAS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h deleted file mode 100644 index b994759b268..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/products/TriangularSolverVector.h +++ /dev/null @@ -1,145 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_TRIANGULAR_SOLVER_VECTOR_H -#define EIGEN_TRIANGULAR_SOLVER_VECTOR_H - -namespace Eigen { - -namespace internal { - -template -struct triangular_solve_vector -{ - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - triangular_solve_vector::run(size, _lhs, lhsStride, rhs); - } -}; - -// forward and backward substitution, row-major, rhs is a vector -template -struct triangular_solve_vector -{ - enum { - IsLower = ((Mode&Lower)==Lower) - }; - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,size,size,OuterStride<>(lhsStride)); - - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - - typename internal::conditional< - Conjugate, - const CwiseUnaryOp,LhsMap>, - const LhsMap&> - ::type cjLhs(lhs); - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - for(Index pi=IsLower ? 0 : size; - IsLower ? pi0; - IsLower ? pi+=PanelWidth : pi-=PanelWidth) - { - Index actualPanelWidth = (std::min)(IsLower ? size - pi : pi, PanelWidth); - - Index r = IsLower ? pi : size - pi; // remaining size - if (r > 0) - { - // let's directly call the low level product function because: - // 1 - it is faster to compile - // 2 - it is slighlty faster at runtime - Index startRow = IsLower ? pi : pi-actualPanelWidth; - Index startCol = IsLower ? 0 : pi; - - general_matrix_vector_product::run( - actualPanelWidth, r, - LhsMapper(&lhs.coeffRef(startRow,startCol), lhsStride), - RhsMapper(rhs + startCol, 1), - rhs + startRow, 1, - RhsScalar(-1)); - } - - for(Index k=0; k0) - rhs[i] -= (cjLhs.row(i).segment(s,k).transpose().cwiseProduct(Map >(rhs+s,k))).sum(); - - if(!(Mode & UnitDiag)) - rhs[i] /= cjLhs(i,i); - } - } - } -}; - -// forward and backward substitution, column-major, rhs is a vector -template -struct triangular_solve_vector -{ - enum { - IsLower = ((Mode&Lower)==Lower) - }; - static void run(Index size, const LhsScalar* _lhs, Index lhsStride, RhsScalar* rhs) - { - typedef Map, 0, OuterStride<> > LhsMap; - const LhsMap lhs(_lhs,size,size,OuterStride<>(lhsStride)); - typedef const_blas_data_mapper LhsMapper; - typedef const_blas_data_mapper RhsMapper; - typename internal::conditional,LhsMap>, - const LhsMap& - >::type cjLhs(lhs); - static const Index PanelWidth = EIGEN_TUNE_TRIANGULAR_PANEL_WIDTH; - - for(Index pi=IsLower ? 0 : size; - IsLower ? pi0; - IsLower ? pi+=PanelWidth : pi-=PanelWidth) - { - Index actualPanelWidth = (std::min)(IsLower ? size - pi : pi, PanelWidth); - Index startBlock = IsLower ? pi : pi-actualPanelWidth; - Index endBlock = IsLower ? pi + actualPanelWidth : 0; - - for(Index k=0; k0) - Map >(rhs+s,r) -= rhs[i] * cjLhs.col(i).segment(s,r); - } - Index r = IsLower ? size - endBlock : startBlock; // remaining size - if (r > 0) - { - // let's directly call the low level product function because: - // 1 - it is faster to compile - // 2 - it is slighlty faster at runtime - general_matrix_vector_product::run( - r, actualPanelWidth, - LhsMapper(&lhs.coeffRef(endBlock,startBlock), lhsStride), - RhsMapper(rhs+startBlock, 1), - rhs+endBlock, 1, RhsScalar(-1)); - } - } - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_TRIANGULAR_SOLVER_VECTOR_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h deleted file mode 100644 index 3dff9bc9b33..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/BlasUtil.h +++ /dev/null @@ -1,499 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009-2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BLASUTIL_H -#define EIGEN_BLASUTIL_H - -// This file contains many lightweight helper classes used to -// implement and control fast level 2 and level 3 BLAS-like routines. - -namespace Eigen { - -namespace internal { - -// forward declarations -template -struct gebp_kernel; - -template -struct gemm_pack_rhs; - -template -struct gemm_pack_lhs; - -template< - typename Index, - typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResStorageOrder, int ResInnerStride> -struct general_matrix_matrix_product; - -template -struct general_matrix_vector_product; - - -template struct conj_if; - -template<> struct conj_if { - template - inline T operator()(const T& x) const { return numext::conj(x); } - template - inline T pconj(const T& x) const { return internal::pconj(x); } -}; - -template<> struct conj_if { - template - inline const T& operator()(const T& x) const { return x; } - template - inline const T& pconj(const T& x) const { return x; } -}; - -// Generic implementation for custom complex types. -template -struct conj_helper -{ - typedef typename ScalarBinaryOpTraits::ReturnType Scalar; - - EIGEN_STRONG_INLINE Scalar pmadd(const LhsScalar& x, const RhsScalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - - EIGEN_STRONG_INLINE Scalar pmul(const LhsScalar& x, const RhsScalar& y) const - { return conj_if()(x) * conj_if()(y); } -}; - -template struct conj_helper -{ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const { return internal::pmadd(x,y,c); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const { return internal::pmul(x,y); } -}; - -template struct conj_helper, std::complex, false,true> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) + numext::imag(x)*numext::imag(y), numext::imag(x)*numext::real(y) - numext::real(x)*numext::imag(y)); } -}; - -template struct conj_helper, std::complex, true,false> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) + numext::imag(x)*numext::imag(y), numext::real(x)*numext::imag(y) - numext::imag(x)*numext::real(y)); } -}; - -template struct conj_helper, std::complex, true,true> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const Scalar& y, const Scalar& c) const - { return c + pmul(x,y); } - - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const Scalar& y) const - { return Scalar(numext::real(x)*numext::real(y) - numext::imag(x)*numext::imag(y), - numext::real(x)*numext::imag(y) - numext::imag(x)*numext::real(y)); } -}; - -template struct conj_helper, RealScalar, Conj,false> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const Scalar& x, const RealScalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - EIGEN_STRONG_INLINE Scalar pmul(const Scalar& x, const RealScalar& y) const - { return conj_if()(x)*y; } -}; - -template struct conj_helper, false,Conj> -{ - typedef std::complex Scalar; - EIGEN_STRONG_INLINE Scalar pmadd(const RealScalar& x, const Scalar& y, const Scalar& c) const - { return padd(c, pmul(x,y)); } - EIGEN_STRONG_INLINE Scalar pmul(const RealScalar& x, const Scalar& y) const - { return x*conj_if()(y); } -}; - -template struct get_factor { - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE To run(const From& x) { return To(x); } -}; - -template struct get_factor::Real> { - EIGEN_DEVICE_FUNC - static EIGEN_STRONG_INLINE typename NumTraits::Real run(const Scalar& x) { return numext::real(x); } -}; - - -template -class BlasVectorMapper { - public: - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasVectorMapper(Scalar *data) : m_data(data) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar operator()(Index i) const { - return m_data[i]; - } - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet load(Index i) const { - return ploadt(m_data + i); - } - - template - EIGEN_DEVICE_FUNC bool aligned(Index i) const { - return (UIntPtr(m_data+i)%sizeof(Packet))==0; - } - - protected: - Scalar* m_data; -}; - -template -class BlasLinearMapper; - -template -class BlasLinearMapper { - public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data, Index incr=1) - : m_data(data) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { - internal::prefetch(&operator()(i)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { - return m_data[i]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { - return ploadt(m_data + i); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE HalfPacket loadHalfPacket(Index i) const { - return ploadt(m_data + i); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const Packet &p) const { - pstoret(m_data + i, p); - } - - protected: - Scalar *m_data; -}; - -// Lightweight helper class to access matrix coefficients. -template -class blas_data_mapper; - -template -class blas_data_mapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - typedef BlasLinearMapper LinearMapper; - typedef BlasVectorMapper VectorMapper; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr=1) - : m_data(data), m_stride(stride) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper - getSubMapper(Index i, Index j) const { - return blas_data_mapper(&operator()(i, j), m_stride); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { - return LinearMapper(&operator()(i, j)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE VectorMapper getVectorMapper(Index i, Index j) const { - return VectorMapper(&operator()(i, j)); - } - - - EIGEN_DEVICE_FUNC - EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { - return m_data[StorageOrder==RowMajor ? j + i*m_stride : i + j*m_stride]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { - return ploadt(&operator()(i, j)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE HalfPacket loadHalfPacket(Index i, Index j) const { - return ploadt(&operator()(i, j)); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { - pscatter(&operator()(i, j), p, m_stride); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { - return pgather(&operator()(i, j), m_stride); - } - - EIGEN_DEVICE_FUNC const Index stride() const { return m_stride; } - EIGEN_DEVICE_FUNC const Scalar* data() const { return m_data; } - - EIGEN_DEVICE_FUNC Index firstAligned(Index size) const { - if (UIntPtr(m_data)%sizeof(Scalar)) { - return -1; - } - return internal::first_default_aligned(m_data, size); - } - - protected: - Scalar* EIGEN_RESTRICT m_data; - const Index m_stride; -}; - -// Implementation of non-natural increment (i.e. inner-stride != 1) -// The exposed API is not complete yet compared to the Incr==1 case -// because some features makes less sense in this case. -template -class BlasLinearMapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data,Index incr) : m_data(data), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { - internal::prefetch(&operator()(i)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { - return m_data[i*m_incr.value()]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { - return pgather(m_data + i*m_incr.value(), m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const PacketType &p) const { - pscatter(m_data + i*m_incr.value(), p, m_incr.value()); - } - -protected: - Scalar *m_data; - const internal::variable_if_dynamic m_incr; -}; - -template -class blas_data_mapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - typedef BlasLinearMapper LinearMapper; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr) : m_data(data), m_stride(stride), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper - getSubMapper(Index i, Index j) const { - return blas_data_mapper(&operator()(i, j), m_stride, m_incr.value()); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { - return LinearMapper(&operator()(i, j), m_incr.value()); - } - - EIGEN_DEVICE_FUNC - EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { - return m_data[StorageOrder==RowMajor ? j*m_incr.value() + i*m_stride : i*m_incr.value() + j*m_stride]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE PacketT load(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { - pscatter(&operator()(i, j), p, m_stride); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { - return pgather(&operator()(i, j), m_stride); - } - -protected: - Scalar* EIGEN_RESTRICT m_data; - const Index m_stride; - const internal::variable_if_dynamic m_incr; -}; - -// lightweight helper class to access matrix coefficients (const version) -template -class const_blas_data_mapper : public blas_data_mapper { - public: - EIGEN_ALWAYS_INLINE const_blas_data_mapper(const Scalar *data, Index stride) : blas_data_mapper(data, stride) {} - - EIGEN_ALWAYS_INLINE const_blas_data_mapper getSubMapper(Index i, Index j) const { - return const_blas_data_mapper(&(this->operator()(i, j)), this->m_stride); - } -}; - - -/* Helper class to analyze the factors of a Product expression. - * In particular it allows to pop out operator-, scalar multiples, - * and conjugate */ -template struct blas_traits -{ - typedef typename traits::Scalar Scalar; - typedef const XprType& ExtractType; - typedef XprType _ExtractType; - enum { - IsComplex = NumTraits::IsComplex, - IsTransposed = false, - NeedToConjugate = false, - HasUsableDirectAccess = ( (int(XprType::Flags)&DirectAccessBit) - && ( bool(XprType::IsVectorAtCompileTime) - || int(inner_stride_at_compile_time::ret) == 1) - ) ? 1 : 0 - }; - typedef typename conditional::type DirectLinearAccessType; - static inline ExtractType extract(const XprType& x) { return x; } - static inline const Scalar extractScalarFactor(const XprType&) { return Scalar(1); } -}; - -// pop conjugate -template -struct blas_traits, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseUnaryOp, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - - enum { - IsComplex = NumTraits::IsComplex, - NeedToConjugate = Base::NeedToConjugate ? 0 : IsComplex - }; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } - static inline Scalar extractScalarFactor(const XprType& x) { return conj(Base::extractScalarFactor(x.nestedExpression())); } -}; - -// pop scalar multiple -template -struct blas_traits, const CwiseNullaryOp,Plain>, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseBinaryOp, const CwiseNullaryOp,Plain>, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.rhs()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return x.lhs().functor().m_other * Base::extractScalarFactor(x.rhs()); } -}; -template -struct blas_traits, NestedXpr, const CwiseNullaryOp,Plain> > > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseBinaryOp, NestedXpr, const CwiseNullaryOp,Plain> > XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.lhs()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return Base::extractScalarFactor(x.lhs()) * x.rhs().functor().m_other; } -}; -template -struct blas_traits, const CwiseNullaryOp,Plain1>, - const CwiseNullaryOp,Plain2> > > - : blas_traits,Plain1> > -{}; - -// pop opposite -template -struct blas_traits, NestedXpr> > - : blas_traits -{ - typedef blas_traits Base; - typedef CwiseUnaryOp, NestedXpr> XprType; - typedef typename Base::ExtractType ExtractType; - static inline ExtractType extract(const XprType& x) { return Base::extract(x.nestedExpression()); } - static inline Scalar extractScalarFactor(const XprType& x) - { return - Base::extractScalarFactor(x.nestedExpression()); } -}; - -// pop/push transpose -template -struct blas_traits > - : blas_traits -{ - typedef typename NestedXpr::Scalar Scalar; - typedef blas_traits Base; - typedef Transpose XprType; - typedef Transpose ExtractType; // const to get rid of a compile error; anyway blas traits are only used on the RHS - typedef Transpose _ExtractType; - typedef typename conditional::type DirectLinearAccessType; - enum { - IsTransposed = Base::IsTransposed ? 0 : 1 - }; - static inline ExtractType extract(const XprType& x) { return ExtractType(Base::extract(x.nestedExpression())); } - static inline Scalar extractScalarFactor(const XprType& x) { return Base::extractScalarFactor(x.nestedExpression()); } -}; - -template -struct blas_traits - : blas_traits -{}; - -template::HasUsableDirectAccess> -struct extract_data_selector { - static const typename T::Scalar* run(const T& m) - { - return blas_traits::extract(m).data(); - } -}; - -template -struct extract_data_selector { - static typename T::Scalar* run(const T&) { return 0; } -}; - -template const typename T::Scalar* extract_data(const T& m) -{ - return extract_data_selector::run(m); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BLASUTIL_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h deleted file mode 100644 index 7587d684243..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Constants.h +++ /dev/null @@ -1,547 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2007-2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_CONSTANTS_H -#define EIGEN_CONSTANTS_H - -namespace Eigen { - -/** This value means that a positive quantity (e.g., a size) is not known at compile-time, and that instead the value is - * stored in some runtime variable. - * - * Changing the value of Dynamic breaks the ABI, as Dynamic is often used as a template parameter for Matrix. - */ -const int Dynamic = -1; - -/** This value means that a signed quantity (e.g., a signed index) is not known at compile-time, and that instead its value - * has to be specified at runtime. - */ -const int DynamicIndex = 0xffffff; - -/** This value means +Infinity; it is currently used only as the p parameter to MatrixBase::lpNorm(). - * The value Infinity there means the L-infinity norm. - */ -const int Infinity = -1; - -/** This value means that the cost to evaluate an expression coefficient is either very expensive or - * cannot be known at compile time. - * - * This value has to be positive to (1) simplify cost computation, and (2) allow to distinguish between a very expensive and very very expensive expressions. - * It thus must also be large enough to make sure unrolling won't happen and that sub expressions will be evaluated, but not too large to avoid overflow. - */ -const int HugeCost = 10000; - -/** \defgroup flags Flags - * \ingroup Core_Module - * - * These are the possible bits which can be OR'ed to constitute the flags of a matrix or - * expression. - * - * It is important to note that these flags are a purely compile-time notion. They are a compile-time property of - * an expression type, implemented as enum's. They are not stored in memory at runtime, and they do not incur any - * runtime overhead. - * - * \sa MatrixBase::Flags - */ - -/** \ingroup flags - * - * for a matrix, this means that the storage order is row-major. - * If this bit is not set, the storage order is column-major. - * For an expression, this determines the storage order of - * the matrix created by evaluation of that expression. - * \sa \blank \ref TopicStorageOrders */ -const unsigned int RowMajorBit = 0x1; - -/** \ingroup flags - * means the expression should be evaluated by the calling expression */ -const unsigned int EvalBeforeNestingBit = 0x2; - -/** \ingroup flags - * \deprecated - * means the expression should be evaluated before any assignment */ -EIGEN_DEPRECATED -const unsigned int EvalBeforeAssigningBit = 0x4; // FIXME deprecated - -/** \ingroup flags - * - * Short version: means the expression might be vectorized - * - * Long version: means that the coefficients can be handled by packets - * and start at a memory location whose alignment meets the requirements - * of the present CPU architecture for optimized packet access. In the fixed-size - * case, there is the additional condition that it be possible to access all the - * coefficients by packets (this implies the requirement that the size be a multiple of 16 bytes, - * and that any nontrivial strides don't break the alignment). In the dynamic-size case, - * there is no such condition on the total size and strides, so it might not be possible to access - * all coeffs by packets. - * - * \note This bit can be set regardless of whether vectorization is actually enabled. - * To check for actual vectorizability, see \a ActualPacketAccessBit. - */ -const unsigned int PacketAccessBit = 0x8; - -#ifdef EIGEN_VECTORIZE -/** \ingroup flags - * - * If vectorization is enabled (EIGEN_VECTORIZE is defined) this constant - * is set to the value \a PacketAccessBit. - * - * If vectorization is not enabled (EIGEN_VECTORIZE is not defined) this constant - * is set to the value 0. - */ -const unsigned int ActualPacketAccessBit = PacketAccessBit; -#else -const unsigned int ActualPacketAccessBit = 0x0; -#endif - -/** \ingroup flags - * - * Short version: means the expression can be seen as 1D vector. - * - * Long version: means that one can access the coefficients - * of this expression by coeff(int), and coeffRef(int) in the case of a lvalue expression. These - * index-based access methods are guaranteed - * to not have to do any runtime computation of a (row, col)-pair from the index, so that it - * is guaranteed that whenever it is available, index-based access is at least as fast as - * (row,col)-based access. Expressions for which that isn't possible don't have the LinearAccessBit. - * - * If both PacketAccessBit and LinearAccessBit are set, then the - * packets of this expression can be accessed by packet(int), and writePacket(int) in the case of a - * lvalue expression. - * - * Typically, all vector expressions have the LinearAccessBit, but there is one exception: - * Product expressions don't have it, because it would be troublesome for vectorization, even when the - * Product is a vector expression. Thus, vector Product expressions allow index-based coefficient access but - * not index-based packet access, so they don't have the LinearAccessBit. - */ -const unsigned int LinearAccessBit = 0x10; - -/** \ingroup flags - * - * Means the expression has a coeffRef() method, i.e. is writable as its individual coefficients are directly addressable. - * This rules out read-only expressions. - * - * Note that DirectAccessBit and LvalueBit are mutually orthogonal, as there are examples of expression having one but note - * the other: - * \li writable expressions that don't have a very simple memory layout as a strided array, have LvalueBit but not DirectAccessBit - * \li Map-to-const expressions, for example Map, have DirectAccessBit but not LvalueBit - * - * Expressions having LvalueBit also have their coeff() method returning a const reference instead of returning a new value. - */ -const unsigned int LvalueBit = 0x20; - -/** \ingroup flags - * - * Means that the underlying array of coefficients can be directly accessed as a plain strided array. The memory layout - * of the array of coefficients must be exactly the natural one suggested by rows(), cols(), - * outerStride(), innerStride(), and the RowMajorBit. This rules out expressions such as Diagonal, whose coefficients, - * though referencable, do not have such a regular memory layout. - * - * See the comment on LvalueBit for an explanation of how LvalueBit and DirectAccessBit are mutually orthogonal. - */ -const unsigned int DirectAccessBit = 0x40; - -/** \deprecated \ingroup flags - * - * means the first coefficient packet is guaranteed to be aligned. - * An expression cannot has the AlignedBit without the PacketAccessBit flag. - * In other words, this means we are allow to perform an aligned packet access to the first element regardless - * of the expression kind: - * \code - * expression.packet(0); - * \endcode - */ -EIGEN_DEPRECATED const unsigned int AlignedBit = 0x80; - -const unsigned int NestByRefBit = 0x100; - -/** \ingroup flags - * - * for an expression, this means that the storage order - * can be either row-major or column-major. - * The precise choice will be decided at evaluation time or when - * combined with other expressions. - * \sa \blank \ref RowMajorBit, \ref TopicStorageOrders */ -const unsigned int NoPreferredStorageOrderBit = 0x200; - -/** \ingroup flags - * - * Means that the underlying coefficients can be accessed through pointers to the sparse (un)compressed storage format, - * that is, the expression provides: - * \code - inline const Scalar* valuePtr() const; - inline const Index* innerIndexPtr() const; - inline const Index* outerIndexPtr() const; - inline const Index* innerNonZeroPtr() const; - \endcode - */ -const unsigned int CompressedAccessBit = 0x400; - - -// list of flags that are inherited by default -const unsigned int HereditaryBits = RowMajorBit - | EvalBeforeNestingBit; - -/** \defgroup enums Enumerations - * \ingroup Core_Module - * - * Various enumerations used in %Eigen. Many of these are used as template parameters. - */ - -/** \ingroup enums - * Enum containing possible values for the \c Mode or \c UpLo parameter of - * MatrixBase::selfadjointView() and MatrixBase::triangularView(), and selfadjoint solvers. */ -enum UpLoType { - /** View matrix as a lower triangular matrix. */ - Lower=0x1, - /** View matrix as an upper triangular matrix. */ - Upper=0x2, - /** %Matrix has ones on the diagonal; to be used in combination with #Lower or #Upper. */ - UnitDiag=0x4, - /** %Matrix has zeros on the diagonal; to be used in combination with #Lower or #Upper. */ - ZeroDiag=0x8, - /** View matrix as a lower triangular matrix with ones on the diagonal. */ - UnitLower=UnitDiag|Lower, - /** View matrix as an upper triangular matrix with ones on the diagonal. */ - UnitUpper=UnitDiag|Upper, - /** View matrix as a lower triangular matrix with zeros on the diagonal. */ - StrictlyLower=ZeroDiag|Lower, - /** View matrix as an upper triangular matrix with zeros on the diagonal. */ - StrictlyUpper=ZeroDiag|Upper, - /** Used in BandMatrix and SelfAdjointView to indicate that the matrix is self-adjoint. */ - SelfAdjoint=0x10, - /** Used to support symmetric, non-selfadjoint, complex matrices. */ - Symmetric=0x20 -}; - -/** \ingroup enums - * Enum for indicating whether a buffer is aligned or not. */ -enum AlignmentType { - Unaligned=0, /**< Data pointer has no specific alignment. */ - Aligned8=8, /**< Data pointer is aligned on a 8 bytes boundary. */ - Aligned16=16, /**< Data pointer is aligned on a 16 bytes boundary. */ - Aligned32=32, /**< Data pointer is aligned on a 32 bytes boundary. */ - Aligned64=64, /**< Data pointer is aligned on a 64 bytes boundary. */ - Aligned128=128, /**< Data pointer is aligned on a 128 bytes boundary. */ - AlignedMask=255, - Aligned=16, /**< \deprecated Synonym for Aligned16. */ -#if EIGEN_MAX_ALIGN_BYTES==128 - AlignedMax = Aligned128 -#elif EIGEN_MAX_ALIGN_BYTES==64 - AlignedMax = Aligned64 -#elif EIGEN_MAX_ALIGN_BYTES==32 - AlignedMax = Aligned32 -#elif EIGEN_MAX_ALIGN_BYTES==16 - AlignedMax = Aligned16 -#elif EIGEN_MAX_ALIGN_BYTES==8 - AlignedMax = Aligned8 -#elif EIGEN_MAX_ALIGN_BYTES==0 - AlignedMax = Unaligned -#else -#error Invalid value for EIGEN_MAX_ALIGN_BYTES -#endif -}; - -/** \ingroup enums - * Enum used by DenseBase::corner() in Eigen2 compatibility mode. */ -// FIXME after the corner() API change, this was not needed anymore, except by AlignedBox -// TODO: find out what to do with that. Adapt the AlignedBox API ? -enum CornerType { TopLeft, TopRight, BottomLeft, BottomRight }; - -/** \ingroup enums - * Enum containing possible values for the \p Direction parameter of - * Reverse, PartialReduxExpr and VectorwiseOp. */ -enum DirectionType { - /** For Reverse, all columns are reversed; - * for PartialReduxExpr and VectorwiseOp, act on columns. */ - Vertical, - /** For Reverse, all rows are reversed; - * for PartialReduxExpr and VectorwiseOp, act on rows. */ - Horizontal, - /** For Reverse, both rows and columns are reversed; - * not used for PartialReduxExpr and VectorwiseOp. */ - BothDirections -}; - -/** \internal \ingroup enums - * Enum to specify how to traverse the entries of a matrix. */ -enum TraversalType { - /** \internal Default traversal, no vectorization, no index-based access */ - DefaultTraversal, - /** \internal No vectorization, use index-based access to have only one for loop instead of 2 nested loops */ - LinearTraversal, - /** \internal Equivalent to a slice vectorization for fixed-size matrices having good alignment - * and good size */ - InnerVectorizedTraversal, - /** \internal Vectorization path using a single loop plus scalar loops for the - * unaligned boundaries */ - LinearVectorizedTraversal, - /** \internal Generic vectorization path using one vectorized loop per row/column with some - * scalar loops to handle the unaligned boundaries */ - SliceVectorizedTraversal, - /** \internal Special case to properly handle incompatible scalar types or other defecting cases*/ - InvalidTraversal, - /** \internal Evaluate all entries at once */ - AllAtOnceTraversal -}; - -/** \internal \ingroup enums - * Enum to specify whether to unroll loops when traversing over the entries of a matrix. */ -enum UnrollingType { - /** \internal Do not unroll loops. */ - NoUnrolling, - /** \internal Unroll only the inner loop, but not the outer loop. */ - InnerUnrolling, - /** \internal Unroll both the inner and the outer loop. If there is only one loop, - * because linear traversal is used, then unroll that loop. */ - CompleteUnrolling -}; - -/** \internal \ingroup enums - * Enum to specify whether to use the default (built-in) implementation or the specialization. */ -enum SpecializedType { - Specialized, - BuiltIn -}; - -/** \ingroup enums - * Enum containing possible values for the \p _Options template parameter of - * Matrix, Array and BandMatrix. */ -enum StorageOptions { - /** Storage order is column major (see \ref TopicStorageOrders). */ - ColMajor = 0, - /** Storage order is row major (see \ref TopicStorageOrders). */ - RowMajor = 0x1, // it is only a coincidence that this is equal to RowMajorBit -- don't rely on that - /** Align the matrix itself if it is vectorizable fixed-size */ - AutoAlign = 0, - /** Don't require alignment for the matrix itself (the array of coefficients, if dynamically allocated, may still be requested to be aligned) */ // FIXME --- clarify the situation - DontAlign = 0x2 -}; - -/** \ingroup enums - * Enum for specifying whether to apply or solve on the left or right. */ -enum SideType { - /** Apply transformation on the left. */ - OnTheLeft = 1, - /** Apply transformation on the right. */ - OnTheRight = 2 -}; - -/* the following used to be written as: - * - * struct NoChange_t {}; - * namespace { - * EIGEN_UNUSED NoChange_t NoChange; - * } - * - * on the ground that it feels dangerous to disambiguate overloaded functions on enum/integer types. - * However, this leads to "variable declared but never referenced" warnings on Intel Composer XE, - * and we do not know how to get rid of them (bug 450). - */ - -enum NoChange_t { NoChange }; -enum Sequential_t { Sequential }; -enum Default_t { Default }; - -/** \internal \ingroup enums - * Used in AmbiVector. */ -enum AmbiVectorMode { - IsDense = 0, - IsSparse -}; - -/** \ingroup enums - * Used as template parameter in DenseCoeffBase and MapBase to indicate - * which accessors should be provided. */ -enum AccessorLevels { - /** Read-only access via a member function. */ - ReadOnlyAccessors, - /** Read/write access via member functions. */ - WriteAccessors, - /** Direct read-only access to the coefficients. */ - DirectAccessors, - /** Direct read/write access to the coefficients. */ - DirectWriteAccessors -}; - -/** \ingroup enums - * Enum with options to give to various decompositions. */ -enum DecompositionOptions { - /** \internal Not used (meant for LDLT?). */ - Pivoting = 0x01, - /** \internal Not used (meant for LDLT?). */ - NoPivoting = 0x02, - /** Used in JacobiSVD to indicate that the square matrix U is to be computed. */ - ComputeFullU = 0x04, - /** Used in JacobiSVD to indicate that the thin matrix U is to be computed. */ - ComputeThinU = 0x08, - /** Used in JacobiSVD to indicate that the square matrix V is to be computed. */ - ComputeFullV = 0x10, - /** Used in JacobiSVD to indicate that the thin matrix V is to be computed. */ - ComputeThinV = 0x20, - /** Used in SelfAdjointEigenSolver and GeneralizedSelfAdjointEigenSolver to specify - * that only the eigenvalues are to be computed and not the eigenvectors. */ - EigenvaluesOnly = 0x40, - /** Used in SelfAdjointEigenSolver and GeneralizedSelfAdjointEigenSolver to specify - * that both the eigenvalues and the eigenvectors are to be computed. */ - ComputeEigenvectors = 0x80, - /** \internal */ - EigVecMask = EigenvaluesOnly | ComputeEigenvectors, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ Ax = \lambda B x \f$. */ - Ax_lBx = 0x100, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ ABx = \lambda x \f$. */ - ABx_lx = 0x200, - /** Used in GeneralizedSelfAdjointEigenSolver to indicate that it should - * solve the generalized eigenproblem \f$ BAx = \lambda x \f$. */ - BAx_lx = 0x400, - /** \internal */ - GenEigMask = Ax_lBx | ABx_lx | BAx_lx -}; - -/** \ingroup enums - * Possible values for the \p QRPreconditioner template parameter of JacobiSVD. */ -enum QRPreconditioners { - /** Do not specify what is to be done if the SVD of a non-square matrix is asked for. */ - NoQRPreconditioner, - /** Use a QR decomposition without pivoting as the first step. */ - HouseholderQRPreconditioner, - /** Use a QR decomposition with column pivoting as the first step. */ - ColPivHouseholderQRPreconditioner, - /** Use a QR decomposition with full pivoting as the first step. */ - FullPivHouseholderQRPreconditioner -}; - -#ifdef Success -#error The preprocessor symbol 'Success' is defined, possibly by the X11 header file X.h -#endif - -/** \ingroup enums - * Enum for reporting the status of a computation. */ -enum ComputationInfo { - /** Computation was successful. */ - Success = 0, - /** The provided data did not satisfy the prerequisites. */ - NumericalIssue = 1, - /** Iterative procedure did not converge. */ - NoConvergence = 2, - /** The inputs are invalid, or the algorithm has been improperly called. - * When assertions are enabled, such errors trigger an assert. */ - InvalidInput = 3 -}; - -/** \ingroup enums - * Enum used to specify how a particular transformation is stored in a matrix. - * \sa Transform, Hyperplane::transform(). */ -enum TransformTraits { - /** Transformation is an isometry. */ - Isometry = 0x1, - /** Transformation is an affine transformation stored as a (Dim+1)^2 matrix whose last row is - * assumed to be [0 ... 0 1]. */ - Affine = 0x2, - /** Transformation is an affine transformation stored as a (Dim) x (Dim+1) matrix. */ - AffineCompact = 0x10 | Affine, - /** Transformation is a general projective transformation stored as a (Dim+1)^2 matrix. */ - Projective = 0x20 -}; - -/** \internal \ingroup enums - * Enum used to choose between implementation depending on the computer architecture. */ -namespace Architecture -{ - enum Type { - Generic = 0x0, - SSE = 0x1, - AltiVec = 0x2, - VSX = 0x3, - NEON = 0x4, -#if defined EIGEN_VECTORIZE_SSE - Target = SSE -#elif defined EIGEN_VECTORIZE_ALTIVEC - Target = AltiVec -#elif defined EIGEN_VECTORIZE_VSX - Target = VSX -#elif defined EIGEN_VECTORIZE_NEON - Target = NEON -#else - Target = Generic -#endif - }; -} - -/** \internal \ingroup enums - * Enum used as template parameter in Product and product evaluators. */ -enum ProductImplType -{ DefaultProduct=0, LazyProduct, AliasFreeProduct, CoeffBasedProductMode, LazyCoeffBasedProductMode, OuterProduct, InnerProduct, GemvProduct, GemmProduct }; - -/** \internal \ingroup enums - * Enum used in experimental parallel implementation. */ -enum Action {GetAction, SetAction}; - -/** The type used to identify a dense storage. */ -struct Dense {}; - -/** The type used to identify a general sparse storage. */ -struct Sparse {}; - -/** The type used to identify a general solver (factored) storage. */ -struct SolverStorage {}; - -/** The type used to identify a permutation storage. */ -struct PermutationStorage {}; - -/** The type used to identify a permutation storage. */ -struct TranspositionsStorage {}; - -/** The type used to identify a matrix expression */ -struct MatrixXpr {}; - -/** The type used to identify an array expression */ -struct ArrayXpr {}; - -// An evaluator must define its shape. By default, it can be one of the following: -struct DenseShape { static std::string debugName() { return "DenseShape"; } }; -struct SolverShape { static std::string debugName() { return "SolverShape"; } }; -struct HomogeneousShape { static std::string debugName() { return "HomogeneousShape"; } }; -struct DiagonalShape { static std::string debugName() { return "DiagonalShape"; } }; -struct BandShape { static std::string debugName() { return "BandShape"; } }; -struct TriangularShape { static std::string debugName() { return "TriangularShape"; } }; -struct SelfAdjointShape { static std::string debugName() { return "SelfAdjointShape"; } }; -struct PermutationShape { static std::string debugName() { return "PermutationShape"; } }; -struct TranspositionsShape { static std::string debugName() { return "TranspositionsShape"; } }; -struct SparseShape { static std::string debugName() { return "SparseShape"; } }; - -namespace internal { - - // random access iterators based on coeff*() accessors. -struct IndexBased {}; - -// evaluator based on iterators to access coefficients. -struct IteratorBased {}; - -/** \internal - * Constants for comparison functors - */ -enum ComparisonName { - cmp_EQ = 0, - cmp_LT = 1, - cmp_LE = 2, - cmp_UNORD = 3, - cmp_NEQ = 4, - cmp_GT = 5, - cmp_GE = 6 -}; -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_CONSTANTS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h deleted file mode 100644 index 74f74cc42b8..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/DisableStupidWarnings.h +++ /dev/null @@ -1,94 +0,0 @@ -#ifndef EIGEN_WARNINGS_DISABLED -#define EIGEN_WARNINGS_DISABLED - -#ifdef _MSC_VER - // 4100 - unreferenced formal parameter (occurred e.g. in aligned_allocator::destroy(pointer p)) - // 4101 - unreferenced local variable - // 4127 - conditional expression is constant - // 4181 - qualifier applied to reference type ignored - // 4211 - nonstandard extension used : redefined extern to static - // 4244 - 'argument' : conversion from 'type1' to 'type2', possible loss of data - // 4273 - QtAlignedMalloc, inconsistent DLL linkage - // 4324 - structure was padded due to declspec(align()) - // 4503 - decorated name length exceeded, name was truncated - // 4512 - assignment operator could not be generated - // 4522 - 'class' : multiple assignment operators specified - // 4700 - uninitialized local variable 'xyz' used - // 4714 - function marked as __forceinline not inlined - // 4717 - 'function' : recursive on all control paths, function will cause runtime stack overflow - // 4800 - 'type' : forcing value to bool 'true' or 'false' (performance warning) - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma warning( push ) - #endif - #pragma warning( disable : 4100 4101 4127 4181 4211 4244 4273 4324 4503 4512 4522 4700 4714 4717 4800) - -#elif defined __INTEL_COMPILER - // 2196 - routine is both "inline" and "noinline" ("noinline" assumed) - // ICC 12 generates this warning even without any inline keyword, when defining class methods 'inline' i.e. inside of class body - // typedef that may be a reference type. - // 279 - controlling expression is constant - // ICC 12 generates this warning on assert(constant_expression_depending_on_template_params) and frankly this is a legitimate use case. - // 1684 - conversion from pointer to same-sized integral type (potential portability problem) - // 2259 - non-pointer conversion from "Eigen::Index={ptrdiff_t={long}}" to "int" may lose significant bits - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma warning push - #endif - #pragma warning disable 2196 279 1684 2259 - -#elif defined __clang__ - // -Wconstant-logical-operand - warning: use of logical && with constant operand; switch to bitwise & or remove constant - // this is really a stupid warning as it warns on compile-time expressions involving enums - #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #pragma clang diagnostic push - #endif - #pragma clang diagnostic ignored "-Wconstant-logical-operand" - -#elif defined __GNUC__ - - #if (!defined(EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS)) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) - #pragma GCC diagnostic push - #endif - // g++ warns about local variables shadowing member functions, which is too strict - #pragma GCC diagnostic ignored "-Wshadow" - #if __GNUC__ == 4 && __GNUC_MINOR__ < 8 - // Until g++-4.7 there are warnings when comparing unsigned int vs 0, even in templated functions: - #pragma GCC diagnostic ignored "-Wtype-limits" - #endif - #if __GNUC__>=6 - #pragma GCC diagnostic ignored "-Wignored-attributes" - #endif - #if __GNUC__==7 - // See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89325 - #pragma GCC diagnostic ignored "-Wattributes" - #endif -#endif - -#if defined __NVCC__ - // Disable the "statement is unreachable" message - #pragma diag_suppress code_is_unreachable - // Disable the "dynamic initialization in unreachable code" message - #pragma diag_suppress initialization_not_reachable - // Disable the "invalid error number" message that we get with older versions of nvcc - #pragma diag_suppress 1222 - // Disable the "calling a __host__ function from a __host__ __device__ function is not allowed" messages (yes, there are many of them and they seem to change with every version of the compiler) - #pragma diag_suppress 2527 - #pragma diag_suppress 2529 - #pragma diag_suppress 2651 - #pragma diag_suppress 2653 - #pragma diag_suppress 2668 - #pragma diag_suppress 2669 - #pragma diag_suppress 2670 - #pragma diag_suppress 2671 - #pragma diag_suppress 2735 - #pragma diag_suppress 2737 -#endif - -#else -// warnings already disabled: -# ifndef EIGEN_WARNINGS_DISABLED_2 -# define EIGEN_WARNINGS_DISABLED_2 -# elif defined(EIGEN_INTERNAL_DEBUGGING) -# error "Do not include \"DisableStupidWarnings.h\" recursively more than twice!" -# endif - -#endif // not EIGEN_WARNINGS_DISABLED diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h deleted file mode 100644 index 134544f9643..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ForwardDeclarations.h +++ /dev/null @@ -1,298 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2007-2010 Benoit Jacob -// Copyright (C) 2008-2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_FORWARDDECLARATIONS_H -#define EIGEN_FORWARDDECLARATIONS_H - -namespace Eigen { -namespace internal { - -template struct traits; - -// here we say once and for all that traits == traits -// When constness must affect traits, it has to be constness on template parameters on which T itself depends. -// For example, traits > != traits >, but -// traits > == traits > -template struct traits : traits {}; - -template struct has_direct_access -{ - enum { ret = (traits::Flags & DirectAccessBit) ? 1 : 0 }; -}; - -template struct accessors_level -{ - enum { has_direct_access = (traits::Flags & DirectAccessBit) ? 1 : 0, - has_write_access = (traits::Flags & LvalueBit) ? 1 : 0, - value = has_direct_access ? (has_write_access ? DirectWriteAccessors : DirectAccessors) - : (has_write_access ? WriteAccessors : ReadOnlyAccessors) - }; -}; - -template struct evaluator_traits; - -template< typename T> struct evaluator; - -} // end namespace internal - -template struct NumTraits; - -template struct EigenBase; -template class DenseBase; -template class PlainObjectBase; -template class DenseCoeffsBase; - -template class Matrix; - -template class MatrixBase; -template class ArrayBase; - -template class Flagged; -template class StorageBase > class NoAlias; -template class NestByValue; -template class ForceAlignedAccess; -template class SwapWrapper; - -template class Block; - -template class VectorBlock; -template class Transpose; -template class Conjugate; -template class CwiseNullaryOp; -template class CwiseUnaryOp; -template class CwiseUnaryView; -template class CwiseBinaryOp; -template class CwiseTernaryOp; -template class Solve; -template class Inverse; - -template class Product; - -template class DiagonalBase; -template class DiagonalWrapper; -template class DiagonalMatrix; -template class DiagonalProduct; -template class Diagonal; -template class PermutationMatrix; -template class Transpositions; -template class PermutationBase; -template class TranspositionsBase; -template class PermutationWrapper; -template class TranspositionsWrapper; - -template::has_write_access ? WriteAccessors : ReadOnlyAccessors -> class MapBase; -template class Stride; -template class InnerStride; -template class OuterStride; -template > class Map; -template class RefBase; -template,OuterStride<> >::type > class Ref; - -template class TriangularBase; -template class TriangularView; -template class SelfAdjointView; -template class SparseView; -template class WithFormat; -template struct CommaInitializer; -template class ReturnByValue; -template class ArrayWrapper; -template class MatrixWrapper; -template class SolverBase; -template class InnerIterator; - -namespace internal { -template struct kernel_retval_base; -template struct kernel_retval; -template struct image_retval_base; -template struct image_retval; -} // end namespace internal - -namespace internal { -template class BandMatrix; -} - -namespace internal { -template struct product_type; - -template struct EnableIf; - -/** \internal - * \class product_evaluator - * Products need their own evaluator with more template arguments allowing for - * easier partial template specializations. - */ -template< typename T, - int ProductTag = internal::product_type::ret, - typename LhsShape = typename evaluator_traits::Shape, - typename RhsShape = typename evaluator_traits::Shape, - typename LhsScalar = typename traits::Scalar, - typename RhsScalar = typename traits::Scalar - > struct product_evaluator; -} - -template::value> -struct ProductReturnType; - -// this is a workaround for sun CC -template struct LazyProductReturnType; - -namespace internal { - -// Provides scalar/packet-wise product and product with accumulation -// with optional conjugation of the arguments. -template struct conj_helper; - -template struct scalar_sum_op; -template struct scalar_difference_op; -template struct scalar_conj_product_op; -template struct scalar_min_op; -template struct scalar_max_op; -template struct scalar_opposite_op; -template struct scalar_conjugate_op; -template struct scalar_real_op; -template struct scalar_imag_op; -template struct scalar_abs_op; -template struct scalar_abs2_op; -template struct scalar_sqrt_op; -template struct scalar_rsqrt_op; -template struct scalar_exp_op; -template struct scalar_log_op; -template struct scalar_cos_op; -template struct scalar_sin_op; -template struct scalar_acos_op; -template struct scalar_asin_op; -template struct scalar_tan_op; -template struct scalar_inverse_op; -template struct scalar_square_op; -template struct scalar_cube_op; -template struct scalar_cast_op; -template struct scalar_random_op; -template struct scalar_constant_op; -template struct scalar_identity_op; -template struct scalar_sign_op; -template struct scalar_pow_op; -template struct scalar_hypot_op; -template struct scalar_product_op; -template struct scalar_quotient_op; - -// SpecialFunctions module -template struct scalar_lgamma_op; -template struct scalar_digamma_op; -template struct scalar_erf_op; -template struct scalar_erfc_op; -template struct scalar_igamma_op; -template struct scalar_igammac_op; -template struct scalar_zeta_op; -template struct scalar_betainc_op; - -} // end namespace internal - -struct IOFormat; - -// Array module -template class Array; -template class Select; -template class PartialReduxExpr; -template class VectorwiseOp; -template class Replicate; -template class Reverse; - -template class FullPivLU; -template class PartialPivLU; -namespace internal { -template struct inverse_impl; -} -template class HouseholderQR; -template class ColPivHouseholderQR; -template class FullPivHouseholderQR; -template class CompleteOrthogonalDecomposition; -template class JacobiSVD; -template class BDCSVD; -template class LLT; -template class LDLT; -template class HouseholderSequence; -template class JacobiRotation; - -// Geometry module: -template class RotationBase; -template class Cross; -template class QuaternionBase; -template class Rotation2D; -template class AngleAxis; -template class Translation; -template class AlignedBox; -template class Quaternion; -template class Transform; -template class ParametrizedLine; -template class Hyperplane; -template class UniformScaling; -template class Homogeneous; - -// Sparse module: -template class SparseMatrixBase; - -// MatrixFunctions module -template struct MatrixExponentialReturnValue; -template class MatrixFunctionReturnValue; -template class MatrixSquareRootReturnValue; -template class MatrixLogarithmReturnValue; -template class MatrixPowerReturnValue; -template class MatrixComplexPowerReturnValue; - -namespace internal { -template -struct stem_function -{ - typedef std::complex::Real> ComplexScalar; - typedef ComplexScalar type(ComplexScalar, int); -}; -} - -} // end namespace Eigen - -#endif // EIGEN_FORWARDDECLARATIONS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h deleted file mode 100644 index b7d6ecc76e8..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/MKL_support.h +++ /dev/null @@ -1,130 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to Intel(R) MKL - * Include file with common MKL declarations - ******************************************************************************** -*/ - -#ifndef EIGEN_MKL_SUPPORT_H -#define EIGEN_MKL_SUPPORT_H - -#ifdef EIGEN_USE_MKL_ALL - #ifndef EIGEN_USE_BLAS - #define EIGEN_USE_BLAS - #endif - #ifndef EIGEN_USE_LAPACKE - #define EIGEN_USE_LAPACKE - #endif - #ifndef EIGEN_USE_MKL_VML - #define EIGEN_USE_MKL_VML - #endif -#endif - -#ifdef EIGEN_USE_LAPACKE_STRICT - #define EIGEN_USE_LAPACKE -#endif - -#if defined(EIGEN_USE_MKL_VML) && !defined(EIGEN_USE_MKL) - #define EIGEN_USE_MKL -#endif - - -#if defined EIGEN_USE_MKL -# include -/*Check IMKL version for compatibility: < 10.3 is not usable with Eigen*/ -# ifndef INTEL_MKL_VERSION -# undef EIGEN_USE_MKL /* INTEL_MKL_VERSION is not even defined on older versions */ -# elif INTEL_MKL_VERSION < 100305 /* the intel-mkl-103-release-notes say this was when the lapacke.h interface was added*/ -# undef EIGEN_USE_MKL -# endif -# ifndef EIGEN_USE_MKL - /*If the MKL version is too old, undef everything*/ -# undef EIGEN_USE_MKL_ALL -# undef EIGEN_USE_LAPACKE -# undef EIGEN_USE_MKL_VML -# undef EIGEN_USE_LAPACKE_STRICT -# undef EIGEN_USE_LAPACKE -# endif -#endif - -#if defined EIGEN_USE_MKL - -#define EIGEN_MKL_VML_THRESHOLD 128 - -/* MKL_DOMAIN_BLAS, etc are defined only in 10.3 update 7 */ -/* MKL_BLAS, etc are not defined in 11.2 */ -#ifdef MKL_DOMAIN_ALL -#define EIGEN_MKL_DOMAIN_ALL MKL_DOMAIN_ALL -#else -#define EIGEN_MKL_DOMAIN_ALL MKL_ALL -#endif - -#ifdef MKL_DOMAIN_BLAS -#define EIGEN_MKL_DOMAIN_BLAS MKL_DOMAIN_BLAS -#else -#define EIGEN_MKL_DOMAIN_BLAS MKL_BLAS -#endif - -#ifdef MKL_DOMAIN_FFT -#define EIGEN_MKL_DOMAIN_FFT MKL_DOMAIN_FFT -#else -#define EIGEN_MKL_DOMAIN_FFT MKL_FFT -#endif - -#ifdef MKL_DOMAIN_VML -#define EIGEN_MKL_DOMAIN_VML MKL_DOMAIN_VML -#else -#define EIGEN_MKL_DOMAIN_VML MKL_VML -#endif - -#ifdef MKL_DOMAIN_PARDISO -#define EIGEN_MKL_DOMAIN_PARDISO MKL_DOMAIN_PARDISO -#else -#define EIGEN_MKL_DOMAIN_PARDISO MKL_PARDISO -#endif -#endif - -#if defined(EIGEN_USE_BLAS) && !defined(EIGEN_USE_MKL) -#include "../../misc/blas.h" -#endif - -namespace Eigen { - -typedef std::complex dcomplex; -typedef std::complex scomplex; - -#if defined(EIGEN_USE_MKL) -typedef MKL_INT BlasIndex; -#else -typedef int BlasIndex; -#endif - -} // end namespace Eigen - - -#endif // EIGEN_MKL_SUPPORT_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h deleted file mode 100644 index 6b0399ebcc5..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Macros.h +++ /dev/null @@ -1,1053 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MACROS_H -#define EIGEN_MACROS_H - -#define EIGEN_WORLD_VERSION 3 -#define EIGEN_MAJOR_VERSION 3 -#define EIGEN_MINOR_VERSION 9 - -#define EIGEN_VERSION_AT_LEAST(x,y,z) (EIGEN_WORLD_VERSION>x || (EIGEN_WORLD_VERSION>=x && \ - (EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \ - EIGEN_MINOR_VERSION>=z)))) - -// Compiler identification, EIGEN_COMP_* - -/// \internal EIGEN_COMP_GNUC set to 1 for all compilers compatible with GCC -#ifdef __GNUC__ - #define EIGEN_COMP_GNUC 1 -#else - #define EIGEN_COMP_GNUC 0 -#endif - -/// \internal EIGEN_COMP_CLANG set to major+minor version (e.g., 307 for clang 3.7) if the compiler is clang -#if defined(__clang__) - #define EIGEN_COMP_CLANG (__clang_major__*100+__clang_minor__) -#else - #define EIGEN_COMP_CLANG 0 -#endif - - -/// \internal EIGEN_COMP_LLVM set to 1 if the compiler backend is llvm -#if defined(__llvm__) - #define EIGEN_COMP_LLVM 1 -#else - #define EIGEN_COMP_LLVM 0 -#endif - -/// \internal EIGEN_COMP_ICC set to __INTEL_COMPILER if the compiler is Intel compiler, 0 otherwise -#if defined(__INTEL_COMPILER) - #define EIGEN_COMP_ICC __INTEL_COMPILER -#else - #define EIGEN_COMP_ICC 0 -#endif - -/// \internal EIGEN_COMP_MINGW set to 1 if the compiler is mingw -#if defined(__MINGW32__) - #define EIGEN_COMP_MINGW 1 -#else - #define EIGEN_COMP_MINGW 0 -#endif - -/// \internal EIGEN_COMP_SUNCC set to 1 if the compiler is Solaris Studio -#if defined(__SUNPRO_CC) - #define EIGEN_COMP_SUNCC 1 -#else - #define EIGEN_COMP_SUNCC 0 -#endif - -/// \internal EIGEN_COMP_MSVC set to _MSC_VER if the compiler is Microsoft Visual C++, 0 otherwise. -#if defined(_MSC_VER) - #define EIGEN_COMP_MSVC _MSC_VER -#else - #define EIGEN_COMP_MSVC 0 -#endif - -// For the record, here is a table summarizing the possible values for EIGEN_COMP_MSVC: -// name ver MSC_VER -// 2008 9 1500 -// 2010 10 1600 -// 2012 11 1700 -// 2013 12 1800 -// 2015 14 1900 -// "15" 15 1900 - -/// \internal EIGEN_COMP_MSVC_STRICT set to 1 if the compiler is really Microsoft Visual C++ and not ,e.g., ICC or clang-cl -#if EIGEN_COMP_MSVC && !(EIGEN_COMP_ICC || EIGEN_COMP_LLVM || EIGEN_COMP_CLANG) - #define EIGEN_COMP_MSVC_STRICT _MSC_VER -#else - #define EIGEN_COMP_MSVC_STRICT 0 -#endif - -/// \internal EIGEN_COMP_IBM set to 1 if the compiler is IBM XL C++ -#if defined(__IBMCPP__) || defined(__xlc__) - #define EIGEN_COMP_IBM 1 -#else - #define EIGEN_COMP_IBM 0 -#endif - -/// \internal EIGEN_COMP_PGI set to 1 if the compiler is Portland Group Compiler -#if defined(__PGI) - #define EIGEN_COMP_PGI 1 -#else - #define EIGEN_COMP_PGI 0 -#endif - -/// \internal EIGEN_COMP_ARM set to 1 if the compiler is ARM Compiler -#if defined(__CC_ARM) || defined(__ARMCC_VERSION) - #define EIGEN_COMP_ARM 1 -#else - #define EIGEN_COMP_ARM 0 -#endif - -/// \internal EIGEN_COMP_ARM set to 1 if the compiler is ARM Compiler -#if defined(__EMSCRIPTEN__) - #define EIGEN_COMP_EMSCRIPTEN 1 -#else - #define EIGEN_COMP_EMSCRIPTEN 0 -#endif - - -/// \internal EIGEN_GNUC_STRICT set to 1 if the compiler is really GCC and not a compatible compiler (e.g., ICC, clang, mingw, etc.) -#if EIGEN_COMP_GNUC && !(EIGEN_COMP_CLANG || EIGEN_COMP_ICC || EIGEN_COMP_MINGW || EIGEN_COMP_PGI || EIGEN_COMP_IBM || EIGEN_COMP_ARM || EIGEN_COMP_EMSCRIPTEN) - #define EIGEN_COMP_GNUC_STRICT 1 -#else - #define EIGEN_COMP_GNUC_STRICT 0 -#endif - - -#if EIGEN_COMP_GNUC - #define EIGEN_GNUC_AT_LEAST(x,y) ((__GNUC__==x && __GNUC_MINOR__>=y) || __GNUC__>x) - #define EIGEN_GNUC_AT_MOST(x,y) ((__GNUC__==x && __GNUC_MINOR__<=y) || __GNUC__=11 && (defined(__cplusplus) && (__cplusplus >= 201103L) || EIGEN_COMP_MSVC >= 1900) -#define EIGEN_HAS_CXX11 1 -#else -#define EIGEN_HAS_CXX11 0 -#endif - - -// Do we support r-value references? -#ifndef EIGEN_HAS_RVALUE_REFERENCES -#if EIGEN_MAX_CPP_VER>=11 && \ - (__has_feature(cxx_rvalue_references) || \ - (defined(__cplusplus) && __cplusplus >= 201103L) || \ - (EIGEN_COMP_MSVC >= 1600)) - #define EIGEN_HAS_RVALUE_REFERENCES 1 -#else - #define EIGEN_HAS_RVALUE_REFERENCES 0 -#endif -#endif - -// Does the compiler support C99? -#ifndef EIGEN_HAS_C99_MATH -#if EIGEN_MAX_CPP_VER>=11 && \ - ((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)) \ - || (defined(__GNUC__) && defined(_GLIBCXX_USE_C99)) \ - || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER)) \ - || (EIGEN_COMP_MSVC >= 1900) ) - #define EIGEN_HAS_C99_MATH 1 -#else - #define EIGEN_HAS_C99_MATH 0 -#endif -#endif - -// Does the compiler support result_of? -#ifndef EIGEN_HAS_STD_RESULT_OF -#if EIGEN_MAX_CPP_VER>=11 && ((__has_feature(cxx_lambdas) || (defined(__cplusplus) && __cplusplus >= 201103L))) -#define EIGEN_HAS_STD_RESULT_OF 1 -#else -#define EIGEN_HAS_STD_RESULT_OF 0 -#endif -#endif - -// Does the compiler support type_traits? -// - full support of type traits was added only to GCC 5.1.0. -// - 20150626 corresponds to the last release of 4.x libstdc++ -#ifndef EIGEN_HAS_TYPE_TRAITS -#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_HAS_CXX11 || EIGEN_COMP_MSVC >= 1700) \ - && ((!EIGEN_COMP_GNUC_STRICT) || EIGEN_GNUC_AT_LEAST(5, 1)) \ - && ((!defined(__GLIBCXX__)) || __GLIBCXX__ > 20150626) -#define EIGEN_HAS_TYPE_TRAITS 1 -#define EIGEN_INCLUDE_TYPE_TRAITS -#else -#define EIGEN_HAS_TYPE_TRAITS 0 -#endif -#endif - -// Does the compiler support variadic templates? -#ifndef EIGEN_HAS_VARIADIC_TEMPLATES -#if EIGEN_MAX_CPP_VER>=11 && (__cplusplus > 199711L || EIGEN_COMP_MSVC >= 1900) \ - && (!defined(__NVCC__) || !EIGEN_ARCH_ARM_OR_ARM64 || (EIGEN_CUDACC_VER >= 80000) ) - // ^^ Disable the use of variadic templates when compiling with versions of nvcc older than 8.0 on ARM devices: - // this prevents nvcc from crashing when compiling Eigen on Tegra X1 -#define EIGEN_HAS_VARIADIC_TEMPLATES 1 -#else -#define EIGEN_HAS_VARIADIC_TEMPLATES 0 -#endif -#endif - -// Does the compiler fully support const expressions? (as in c++14) -#ifndef EIGEN_HAS_CONSTEXPR - -#ifdef __CUDACC__ -// Const expressions are supported provided that c++11 is enabled and we're using either clang or nvcc 7.5 or above -#if EIGEN_MAX_CPP_VER>=14 && (__cplusplus > 199711L && (EIGEN_COMP_CLANG || EIGEN_CUDACC_VER >= 70500)) - #define EIGEN_HAS_CONSTEXPR 1 -#endif -#elif EIGEN_MAX_CPP_VER>=14 && (__has_feature(cxx_relaxed_constexpr) || (defined(__cplusplus) && __cplusplus >= 201402L) || \ - (EIGEN_GNUC_AT_LEAST(4,8) && (__cplusplus > 199711L))) -#define EIGEN_HAS_CONSTEXPR 1 -#endif - -#ifndef EIGEN_HAS_CONSTEXPR -#define EIGEN_HAS_CONSTEXPR 0 -#endif - -#endif - -// Does the compiler support C++11 math? -// Let's be conservative and enable the default C++11 implementation only if we are sure it exists -#ifndef EIGEN_HAS_CXX11_MATH - #if EIGEN_MAX_CPP_VER>=11 && ((__cplusplus > 201103L) || (__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC) \ - && (EIGEN_ARCH_i386_OR_x86_64) && (EIGEN_OS_GNULINUX || EIGEN_OS_WIN_STRICT || EIGEN_OS_MAC)) - #define EIGEN_HAS_CXX11_MATH 1 - #else - #define EIGEN_HAS_CXX11_MATH 0 - #endif -#endif - -// Does the compiler support proper C++11 containers? -#ifndef EIGEN_HAS_CXX11_CONTAINERS - #if EIGEN_MAX_CPP_VER>=11 && \ - ((__cplusplus > 201103L) \ - || ((__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_ICC>=1400)) \ - || EIGEN_COMP_MSVC >= 1900) - #define EIGEN_HAS_CXX11_CONTAINERS 1 - #else - #define EIGEN_HAS_CXX11_CONTAINERS 0 - #endif -#endif - -// Does the compiler support C++11 noexcept? -#ifndef EIGEN_HAS_CXX11_NOEXCEPT - #if EIGEN_MAX_CPP_VER>=11 && \ - (__has_feature(cxx_noexcept) \ - || (__cplusplus > 201103L) \ - || ((__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_ICC>=1400)) \ - || EIGEN_COMP_MSVC >= 1900) - #define EIGEN_HAS_CXX11_NOEXCEPT 1 - #else - #define EIGEN_HAS_CXX11_NOEXCEPT 0 - #endif -#endif - -/** Allows to disable some optimizations which might affect the accuracy of the result. - * Such optimization are enabled by default, and set EIGEN_FAST_MATH to 0 to disable them. - * They currently include: - * - single precision ArrayBase::sin() and ArrayBase::cos() for SSE and AVX vectorization. - */ -#ifndef EIGEN_FAST_MATH -#define EIGEN_FAST_MATH 1 -#endif - -#define EIGEN_DEBUG_VAR(x) std::cerr << #x << " = " << x << std::endl; - -// concatenate two tokens -#define EIGEN_CAT2(a,b) a ## b -#define EIGEN_CAT(a,b) EIGEN_CAT2(a,b) - -#define EIGEN_COMMA , - -// convert a token to a string -#define EIGEN_MAKESTRING2(a) #a -#define EIGEN_MAKESTRING(a) EIGEN_MAKESTRING2(a) - -// EIGEN_STRONG_INLINE is a stronger version of the inline, using __forceinline on MSVC, -// but it still doesn't use GCC's always_inline. This is useful in (common) situations where MSVC needs forceinline -// but GCC is still doing fine with just inline. -#ifndef EIGEN_STRONG_INLINE -#if EIGEN_COMP_MSVC || EIGEN_COMP_ICC -#define EIGEN_STRONG_INLINE __forceinline -#else -#define EIGEN_STRONG_INLINE inline -#endif -#endif - -// EIGEN_ALWAYS_INLINE is the stronget, it has the effect of making the function inline and adding every possible -// attribute to maximize inlining. This should only be used when really necessary: in particular, -// it uses __attribute__((always_inline)) on GCC, which most of the time is useless and can severely harm compile times. -// FIXME with the always_inline attribute, -// gcc 3.4.x and 4.1 reports the following compilation error: -// Eval.h:91: sorry, unimplemented: inlining failed in call to 'const Eigen::Eval Eigen::MatrixBase::eval() const' -// : function body not available -// See also bug 1367 -#if EIGEN_GNUC_AT_LEAST(4,2) -#define EIGEN_ALWAYS_INLINE __attribute__((always_inline)) inline -#else -#define EIGEN_ALWAYS_INLINE EIGEN_STRONG_INLINE -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_DONT_INLINE __attribute__((noinline)) -#elif EIGEN_COMP_MSVC -#define EIGEN_DONT_INLINE __declspec(noinline) -#else -#define EIGEN_DONT_INLINE -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_PERMISSIVE_EXPR __extension__ -#else -#define EIGEN_PERMISSIVE_EXPR -#endif - -// this macro allows to get rid of linking errors about multiply defined functions. -// - static is not very good because it prevents definitions from different object files to be merged. -// So static causes the resulting linked executable to be bloated with multiple copies of the same function. -// - inline is not perfect either as it unwantedly hints the compiler toward inlining the function. -#define EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS -#define EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS inline - -#ifdef NDEBUG -# ifndef EIGEN_NO_DEBUG -# define EIGEN_NO_DEBUG -# endif -#endif - -// eigen_plain_assert is where we implement the workaround for the assert() bug in GCC <= 4.3, see bug 89 -#ifdef EIGEN_NO_DEBUG - #define eigen_plain_assert(x) -#else - #if EIGEN_SAFE_TO_USE_STANDARD_ASSERT_MACRO - namespace Eigen { - namespace internal { - inline bool copy_bool(bool b) { return b; } - } - } - #define eigen_plain_assert(x) assert(x) - #else - // work around bug 89 - #include // for abort - #include // for std::cerr - - namespace Eigen { - namespace internal { - // trivial function copying a bool. Must be EIGEN_DONT_INLINE, so we implement it after including Eigen headers. - // see bug 89. - namespace { - EIGEN_DONT_INLINE bool copy_bool(bool b) { return b; } - } - inline void assert_fail(const char *condition, const char *function, const char *file, int line) - { - std::cerr << "assertion failed: " << condition << " in function " << function << " at " << file << ":" << line << std::endl; - abort(); - } - } - } - #define eigen_plain_assert(x) \ - do { \ - if(!Eigen::internal::copy_bool(x)) \ - Eigen::internal::assert_fail(EIGEN_MAKESTRING(x), __PRETTY_FUNCTION__, __FILE__, __LINE__); \ - } while(false) - #endif -#endif - -// eigen_assert can be overridden -#ifndef eigen_assert -#define eigen_assert(x) eigen_plain_assert(x) -#endif - -#ifdef EIGEN_INTERNAL_DEBUGGING -#define eigen_internal_assert(x) eigen_assert(x) -#else -#define eigen_internal_assert(x) -#endif - -#ifdef EIGEN_NO_DEBUG -#define EIGEN_ONLY_USED_FOR_DEBUG(x) EIGEN_UNUSED_VARIABLE(x) -#else -#define EIGEN_ONLY_USED_FOR_DEBUG(x) -#endif - -#ifndef EIGEN_NO_DEPRECATED_WARNING - #if EIGEN_COMP_GNUC - #define EIGEN_DEPRECATED __attribute__((deprecated)) - #elif EIGEN_COMP_MSVC - #define EIGEN_DEPRECATED __declspec(deprecated) - #else - #define EIGEN_DEPRECATED - #endif -#else - #define EIGEN_DEPRECATED -#endif - -#if EIGEN_COMP_GNUC -#define EIGEN_UNUSED __attribute__((unused)) -#else -#define EIGEN_UNUSED -#endif - -// Suppresses 'unused variable' warnings. -namespace Eigen { - namespace internal { - template EIGEN_DEVICE_FUNC void ignore_unused_variable(const T&) {} - } -} -#define EIGEN_UNUSED_VARIABLE(var) Eigen::internal::ignore_unused_variable(var); - -#if !defined(EIGEN_ASM_COMMENT) - #if EIGEN_COMP_GNUC && (EIGEN_ARCH_i386_OR_x86_64 || EIGEN_ARCH_ARM_OR_ARM64) - #define EIGEN_ASM_COMMENT(X) __asm__("#" X) - #else - #define EIGEN_ASM_COMMENT(X) - #endif -#endif - - -//------------------------------------------------------------------------------------------ -// Static and dynamic alignment control -// -// The main purpose of this section is to define EIGEN_MAX_ALIGN_BYTES and EIGEN_MAX_STATIC_ALIGN_BYTES -// as the maximal boundary in bytes on which dynamically and statically allocated data may be alignment respectively. -// The values of EIGEN_MAX_ALIGN_BYTES and EIGEN_MAX_STATIC_ALIGN_BYTES can be specified by the user. If not, -// a default value is automatically computed based on architecture, compiler, and OS. -// -// This section also defines macros EIGEN_ALIGN_TO_BOUNDARY(N) and the shortcuts EIGEN_ALIGN{8,16,32,_MAX} -// to be used to declare statically aligned buffers. -//------------------------------------------------------------------------------------------ - - -/* EIGEN_ALIGN_TO_BOUNDARY(n) forces data to be n-byte aligned. This is used to satisfy SIMD requirements. - * However, we do that EVEN if vectorization (EIGEN_VECTORIZE) is disabled, - * so that vectorization doesn't affect binary compatibility. - * - * If we made alignment depend on whether or not EIGEN_VECTORIZE is defined, it would be impossible to link - * vectorized and non-vectorized code. - */ -#if (defined __CUDACC__) - #define EIGEN_ALIGN_TO_BOUNDARY(n) __align__(n) -#elif EIGEN_COMP_GNUC || EIGEN_COMP_PGI || EIGEN_COMP_IBM || EIGEN_COMP_ARM - #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n))) -#elif EIGEN_COMP_MSVC - #define EIGEN_ALIGN_TO_BOUNDARY(n) __declspec(align(n)) -#elif EIGEN_COMP_SUNCC - // FIXME not sure about this one: - #define EIGEN_ALIGN_TO_BOUNDARY(n) __attribute__((aligned(n))) -#else - #error Please tell me what is the equivalent of __attribute__((aligned(n))) for your compiler -#endif - -// If the user explicitly disable vectorization, then we also disable alignment -#if defined(EIGEN_DONT_VECTORIZE) - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 0 -#elif defined(EIGEN_VECTORIZE_AVX512) - // 64 bytes static alignmeent is preferred only if really required - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 64 -#elif defined(__AVX__) - // 32 bytes static alignmeent is preferred only if really required - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 32 -#else - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 16 -#endif - - -// EIGEN_MIN_ALIGN_BYTES defines the minimal value for which the notion of explicit alignment makes sense -#define EIGEN_MIN_ALIGN_BYTES 16 - -// Defined the boundary (in bytes) on which the data needs to be aligned. Note -// that unless EIGEN_ALIGN is defined and not equal to 0, the data may not be -// aligned at all regardless of the value of this #define. - -#if (defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN)) && defined(EIGEN_MAX_STATIC_ALIGN_BYTES) && EIGEN_MAX_STATIC_ALIGN_BYTES>0 -#error EIGEN_MAX_STATIC_ALIGN_BYTES and EIGEN_DONT_ALIGN[_STATICALLY] are both defined with EIGEN_MAX_STATIC_ALIGN_BYTES!=0. Use EIGEN_MAX_STATIC_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN_STATICALLY. -#endif - -// EIGEN_DONT_ALIGN_STATICALLY and EIGEN_DONT_ALIGN are deprectated -// They imply EIGEN_MAX_STATIC_ALIGN_BYTES=0 -#if defined(EIGEN_DONT_ALIGN_STATICALLY) || defined(EIGEN_DONT_ALIGN) - #ifdef EIGEN_MAX_STATIC_ALIGN_BYTES - #undef EIGEN_MAX_STATIC_ALIGN_BYTES - #endif - #define EIGEN_MAX_STATIC_ALIGN_BYTES 0 -#endif - -#ifndef EIGEN_MAX_STATIC_ALIGN_BYTES - - // Try to automatically guess what is the best default value for EIGEN_MAX_STATIC_ALIGN_BYTES - - // 16 byte alignment is only useful for vectorization. Since it affects the ABI, we need to enable - // 16 byte alignment on all platforms where vectorization might be enabled. In theory we could always - // enable alignment, but it can be a cause of problems on some platforms, so we just disable it in - // certain common platform (compiler+architecture combinations) to avoid these problems. - // Only static alignment is really problematic (relies on nonstandard compiler extensions), - // try to keep heap alignment even when we have to disable static alignment. - #if EIGEN_COMP_GNUC && !(EIGEN_ARCH_i386_OR_x86_64 || EIGEN_ARCH_ARM_OR_ARM64 || EIGEN_ARCH_PPC || EIGEN_ARCH_IA64) - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1 - #elif EIGEN_ARCH_ARM_OR_ARM64 && EIGEN_COMP_GNUC_STRICT && EIGEN_GNUC_AT_MOST(4, 6) - // Old versions of GCC on ARM, at least 4.4, were once seen to have buggy static alignment support. - // Not sure which version fixed it, hopefully it doesn't affect 4.7, which is still somewhat in use. - // 4.8 and newer seem definitely unaffected. - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 1 - #else - #define EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT 0 - #endif - - // static alignment is completely disabled with GCC 3, Sun Studio, and QCC/QNX - #if !EIGEN_GCC_AND_ARCH_DOESNT_WANT_STACK_ALIGNMENT \ - && !EIGEN_GCC3_OR_OLDER \ - && !EIGEN_COMP_SUNCC \ - && !EIGEN_OS_QNX - #define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 1 - #else - #define EIGEN_ARCH_WANTS_STACK_ALIGNMENT 0 - #endif - - #if EIGEN_ARCH_WANTS_STACK_ALIGNMENT - #define EIGEN_MAX_STATIC_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES - #else - #define EIGEN_MAX_STATIC_ALIGN_BYTES 0 - #endif - -#endif - -// If EIGEN_MAX_ALIGN_BYTES is defined, then it is considered as an upper bound for EIGEN_MAX_ALIGN_BYTES -#if defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES0 is the true test whether we want to align arrays on the stack or not. -// It takes into account both the user choice to explicitly enable/disable alignment (by settting EIGEN_MAX_STATIC_ALIGN_BYTES) -// and the architecture config (EIGEN_ARCH_WANTS_STACK_ALIGNMENT). -// Henceforth, only EIGEN_MAX_STATIC_ALIGN_BYTES should be used. - - -// Shortcuts to EIGEN_ALIGN_TO_BOUNDARY -#define EIGEN_ALIGN8 EIGEN_ALIGN_TO_BOUNDARY(8) -#define EIGEN_ALIGN16 EIGEN_ALIGN_TO_BOUNDARY(16) -#define EIGEN_ALIGN32 EIGEN_ALIGN_TO_BOUNDARY(32) -#define EIGEN_ALIGN64 EIGEN_ALIGN_TO_BOUNDARY(64) -#if EIGEN_MAX_STATIC_ALIGN_BYTES>0 -#define EIGEN_ALIGN_MAX EIGEN_ALIGN_TO_BOUNDARY(EIGEN_MAX_STATIC_ALIGN_BYTES) -#else -#define EIGEN_ALIGN_MAX -#endif - - -// Dynamic alignment control - -#if defined(EIGEN_DONT_ALIGN) && defined(EIGEN_MAX_ALIGN_BYTES) && EIGEN_MAX_ALIGN_BYTES>0 -#error EIGEN_MAX_ALIGN_BYTES and EIGEN_DONT_ALIGN are both defined with EIGEN_MAX_ALIGN_BYTES!=0. Use EIGEN_MAX_ALIGN_BYTES=0 as a synonym of EIGEN_DONT_ALIGN. -#endif - -#ifdef EIGEN_DONT_ALIGN - #ifdef EIGEN_MAX_ALIGN_BYTES - #undef EIGEN_MAX_ALIGN_BYTES - #endif - #define EIGEN_MAX_ALIGN_BYTES 0 -#elif !defined(EIGEN_MAX_ALIGN_BYTES) - #define EIGEN_MAX_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES -#endif - -#if EIGEN_IDEAL_MAX_ALIGN_BYTES > EIGEN_MAX_ALIGN_BYTES -#define EIGEN_DEFAULT_ALIGN_BYTES EIGEN_IDEAL_MAX_ALIGN_BYTES -#else -#define EIGEN_DEFAULT_ALIGN_BYTES EIGEN_MAX_ALIGN_BYTES -#endif - - -#ifndef EIGEN_UNALIGNED_VECTORIZE -#define EIGEN_UNALIGNED_VECTORIZE 1 -#endif - -//---------------------------------------------------------------------- - - -#ifdef EIGEN_DONT_USE_RESTRICT_KEYWORD - #define EIGEN_RESTRICT -#endif -#ifndef EIGEN_RESTRICT - #define EIGEN_RESTRICT __restrict -#endif - -#ifndef EIGEN_STACK_ALLOCATION_LIMIT -// 131072 == 128 KB -#define EIGEN_STACK_ALLOCATION_LIMIT 131072 -#endif - -#ifndef EIGEN_DEFAULT_IO_FORMAT -#ifdef EIGEN_MAKING_DOCS -// format used in Eigen's documentation -// needed to define it here as escaping characters in CMake add_definition's argument seems very problematic. -#define EIGEN_DEFAULT_IO_FORMAT Eigen::IOFormat(3, 0, " ", "\n", "", "") -#else -#define EIGEN_DEFAULT_IO_FORMAT Eigen::IOFormat() -#endif -#endif - -// just an empty macro ! -#define EIGEN_EMPTY - -#if EIGEN_COMP_MSVC_STRICT && (EIGEN_COMP_MSVC < 1900 || EIGEN_CUDACC_VER>0) - // for older MSVC versions, as well as 1900 && CUDA 8, using the base operator is sufficient (cf Bugs 1000, 1324) - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; -#elif EIGEN_COMP_CLANG // workaround clang bug (see http://forum.kde.org/viewtopic.php?f=74&t=102653) - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const Derived& other) { Base::operator=(other); return *this; } \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const DenseBase& other) { Base::operator=(other.derived()); return *this; } -#else - #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - using Base::operator =; \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& operator=(const Derived& other) \ - { \ - Base::operator=(other); \ - return *this; \ - } -#endif - - -/** - * \internal - * \brief Macro to explicitly define the default copy constructor. - * This is necessary, because the implicit definition is deprecated if the copy-assignment is overridden. - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) EIGEN_DEVICE_FUNC CLASS(const CLASS&) = default; -#else -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) -#endif - - - -/** \internal - * \brief Macro to manually inherit assignment operators. - * This is necessary, because the implicitly defined assignment operator gets deleted when a custom operator= is defined. - * With C++11 or later this also default-implements the copy-constructor - */ -#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) \ - EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - EIGEN_DEFAULT_COPY_CONSTRUCTOR(Derived) - -/** \internal - * \brief Macro to manually define default constructors and destructors. - * This is necessary when the copy constructor is re-defined. - * For empty helper classes this should usually be protected, to avoid accidentally creating empty objects. - * - * Hiding the default destructor lead to problems in C++03 mode together with boost::multiprecision - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() = default; \ - EIGEN_DEVICE_FUNC ~Derived() = default; -#else -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() {}; \ - /* EIGEN_DEVICE_FUNC ~Derived() {}; */ -#endif - - - - - -/** -* Just a side note. Commenting within defines works only by documenting -* behind the object (via '!<'). Comments cannot be multi-line and thus -* we have these extra long lines. What is confusing doxygen over here is -* that we use '\' and basically have a bunch of typedefs with their -* documentation in a single line. -**/ - -#define EIGEN_GENERIC_PUBLIC_INTERFACE(Derived) \ - typedef typename Eigen::internal::traits::Scalar Scalar; /*!< \brief Numeric type, e.g. float, double, int or std::complex. */ \ - typedef typename Eigen::NumTraits::Real RealScalar; /*!< \brief The underlying numeric type for composed scalar types. \details In cases where Scalar is e.g. std::complex, T were corresponding to RealScalar. */ \ - typedef typename Base::CoeffReturnType CoeffReturnType; /*!< \brief The return type for coefficient access. \details Depending on whether the object allows direct coefficient access (e.g. for a MatrixXd), this type is either 'const Scalar&' or simply 'Scalar' for objects that do not allow direct coefficient access. */ \ - typedef typename Eigen::internal::ref_selector::type Nested; \ - typedef typename Eigen::internal::traits::StorageKind StorageKind; \ - typedef typename Eigen::internal::traits::StorageIndex StorageIndex; \ - enum { RowsAtCompileTime = Eigen::internal::traits::RowsAtCompileTime, \ - ColsAtCompileTime = Eigen::internal::traits::ColsAtCompileTime, \ - Flags = Eigen::internal::traits::Flags, \ - SizeAtCompileTime = Base::SizeAtCompileTime, \ - MaxSizeAtCompileTime = Base::MaxSizeAtCompileTime, \ - IsVectorAtCompileTime = Base::IsVectorAtCompileTime }; \ - using Base::derived; \ - using Base::const_cast_derived; - - -// FIXME Maybe the EIGEN_DENSE_PUBLIC_INTERFACE could be removed as importing PacketScalar is rarely needed -#define EIGEN_DENSE_PUBLIC_INTERFACE(Derived) \ - EIGEN_GENERIC_PUBLIC_INTERFACE(Derived) \ - typedef typename Base::PacketScalar PacketScalar; - - -#define EIGEN_PLAIN_ENUM_MIN(a,b) (((int)a <= (int)b) ? (int)a : (int)b) -#define EIGEN_PLAIN_ENUM_MAX(a,b) (((int)a >= (int)b) ? (int)a : (int)b) - -// EIGEN_SIZE_MIN_PREFER_DYNAMIC gives the min between compile-time sizes. 0 has absolute priority, followed by 1, -// followed by Dynamic, followed by other finite values. The reason for giving Dynamic the priority over -// finite values is that min(3, Dynamic) should be Dynamic, since that could be anything between 0 and 3. -#define EIGEN_SIZE_MIN_PREFER_DYNAMIC(a,b) (((int)a == 0 || (int)b == 0) ? 0 \ - : ((int)a == 1 || (int)b == 1) ? 1 \ - : ((int)a == Dynamic || (int)b == Dynamic) ? Dynamic \ - : ((int)a <= (int)b) ? (int)a : (int)b) - -// EIGEN_SIZE_MIN_PREFER_FIXED is a variant of EIGEN_SIZE_MIN_PREFER_DYNAMIC comparing MaxSizes. The difference is that finite values -// now have priority over Dynamic, so that min(3, Dynamic) gives 3. Indeed, whatever the actual value is -// (between 0 and 3), it is not more than 3. -#define EIGEN_SIZE_MIN_PREFER_FIXED(a,b) (((int)a == 0 || (int)b == 0) ? 0 \ - : ((int)a == 1 || (int)b == 1) ? 1 \ - : ((int)a == Dynamic && (int)b == Dynamic) ? Dynamic \ - : ((int)a == Dynamic) ? (int)b \ - : ((int)b == Dynamic) ? (int)a \ - : ((int)a <= (int)b) ? (int)a : (int)b) - -// see EIGEN_SIZE_MIN_PREFER_DYNAMIC. No need for a separate variant for MaxSizes here. -#define EIGEN_SIZE_MAX(a,b) (((int)a == Dynamic || (int)b == Dynamic) ? Dynamic \ - : ((int)a >= (int)b) ? (int)a : (int)b) - -#define EIGEN_LOGICAL_XOR(a,b) (((a) || (b)) && !((a) && (b))) - -#define EIGEN_IMPLIES(a,b) (!(a) || (b)) - -// the expression type of a standard coefficient wise binary operation -#define EIGEN_CWISE_BINARY_RETURN_TYPE(LHS,RHS,OPNAME) \ - CwiseBinaryOp< \ - EIGEN_CAT(EIGEN_CAT(internal::scalar_,OPNAME),_op)< \ - typename internal::traits::Scalar, \ - typename internal::traits::Scalar \ - >, \ - const LHS, \ - const RHS \ - > - -#define EIGEN_MAKE_CWISE_BINARY_OP(METHOD,OPNAME) \ - template \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,OPNAME) \ - (METHOD)(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ - { \ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,OPNAME)(derived(), other.derived()); \ - } - -#define EIGEN_SCALAR_BINARY_SUPPORTED(OPNAME,TYPEA,TYPEB) \ - (Eigen::internal::has_ReturnType > >::value) - -#define EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(EXPR,SCALAR,OPNAME) \ - CwiseBinaryOp::Scalar,SCALAR>, const EXPR, \ - const typename internal::plain_constant_type::type> - -#define EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(SCALAR,EXPR,OPNAME) \ - CwiseBinaryOp::Scalar>, \ - const typename internal::plain_constant_type::type, const EXPR> - -// Workaround for MSVC 2010 (see ML thread "patch with compile for for MSVC 2010") -#if EIGEN_COMP_MSVC_STRICT<=1600 -#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) typename internal::enable_if::type -#else -#define EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(X) X -#endif - -#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME) \ - template EIGEN_DEVICE_FUNC inline \ - EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename internal::promote_scalar_arg::type,OPNAME))\ - (METHOD)(const T& scalar) const { \ - typedef typename internal::promote_scalar_arg::type PromotedT; \ - return EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,PromotedT,OPNAME)(derived(), \ - typename internal::plain_constant_type::type(derived().rows(), derived().cols(), internal::scalar_constant_op(scalar))); \ - } - -#define EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(METHOD,OPNAME) \ - template EIGEN_DEVICE_FUNC inline friend \ - EIGEN_MSVC10_WORKAROUND_BINARYOP_RETURN_TYPE(const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename internal::promote_scalar_arg::type,Derived,OPNAME)) \ - (METHOD)(const T& scalar, const StorageBaseType& matrix) { \ - typedef typename internal::promote_scalar_arg::type PromotedT; \ - return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(PromotedT,Derived,OPNAME)( \ - typename internal::plain_constant_type::type(matrix.derived().rows(), matrix.derived().cols(), internal::scalar_constant_op(scalar)), matrix.derived()); \ - } - -#define EIGEN_MAKE_SCALAR_BINARY_OP(METHOD,OPNAME) \ - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(METHOD,OPNAME) \ - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(METHOD,OPNAME) - - -#ifdef EIGEN_EXCEPTIONS -# define EIGEN_THROW_X(X) throw X -# define EIGEN_THROW throw -# define EIGEN_TRY try -# define EIGEN_CATCH(X) catch (X) -#else -# ifdef __CUDA_ARCH__ -# define EIGEN_THROW_X(X) asm("trap;") -# define EIGEN_THROW asm("trap;") -# else -# define EIGEN_THROW_X(X) std::abort() -# define EIGEN_THROW std::abort() -# endif -# define EIGEN_TRY if (true) -# define EIGEN_CATCH(X) else -#endif - - -#if EIGEN_HAS_CXX11_NOEXCEPT -# define EIGEN_INCLUDE_TYPE_TRAITS -# define EIGEN_NOEXCEPT noexcept -# define EIGEN_NOEXCEPT_IF(x) noexcept(x) -# define EIGEN_NO_THROW noexcept(true) -# define EIGEN_EXCEPTION_SPEC(X) noexcept(false) -#else -# define EIGEN_NOEXCEPT -# define EIGEN_NOEXCEPT_IF(x) -# define EIGEN_NO_THROW throw() -# if EIGEN_COMP_MSVC - // MSVC does not support exception specifications (warning C4290), - // and they are deprecated in c++11 anyway. -# define EIGEN_EXCEPTION_SPEC(X) throw() -# else -# define EIGEN_EXCEPTION_SPEC(X) throw(X) -# endif -#endif - -#endif // EIGEN_MACROS_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h deleted file mode 100644 index 291383c581a..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Memory.h +++ /dev/null @@ -1,993 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2008-2009 Benoit Jacob -// Copyright (C) 2009 Kenneth Riddile -// Copyright (C) 2010 Hauke Heibel -// Copyright (C) 2010 Thomas Capricelli -// Copyright (C) 2013 Pavel Holoborodko -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - - -/***************************************************************************** -*** Platform checks for aligned malloc functions *** -*****************************************************************************/ - -#ifndef EIGEN_MEMORY_H -#define EIGEN_MEMORY_H - -#ifndef EIGEN_MALLOC_ALREADY_ALIGNED - -// Try to determine automatically if malloc is already aligned. - -// On 64-bit systems, glibc's malloc returns 16-byte-aligned pointers, see: -// http://www.gnu.org/s/libc/manual/html_node/Aligned-Memory-Blocks.html -// This is true at least since glibc 2.8. -// This leaves the question how to detect 64-bit. According to this document, -// http://gcc.fyxm.net/summit/2003/Porting%20to%2064%20bit.pdf -// page 114, "[The] LP64 model [...] is used by all 64-bit UNIX ports" so it's indeed -// quite safe, at least within the context of glibc, to equate 64-bit with LP64. -#if defined(__GLIBC__) && ((__GLIBC__>=2 && __GLIBC_MINOR__ >= 8) || __GLIBC__>2) \ - && defined(__LP64__) && ! defined( __SANITIZE_ADDRESS__ ) && (EIGEN_DEFAULT_ALIGN_BYTES == 16) - #define EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED 0 -#endif - -// FreeBSD 6 seems to have 16-byte aligned malloc -// See http://svn.freebsd.org/viewvc/base/stable/6/lib/libc/stdlib/malloc.c?view=markup -// FreeBSD 7 seems to have 16-byte aligned malloc except on ARM and MIPS architectures -// See http://svn.freebsd.org/viewvc/base/stable/7/lib/libc/stdlib/malloc.c?view=markup -#if defined(__FreeBSD__) && !(EIGEN_ARCH_ARM || EIGEN_ARCH_MIPS) && (EIGEN_DEFAULT_ALIGN_BYTES == 16) - #define EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED 0 -#endif - -#if (EIGEN_OS_MAC && (EIGEN_DEFAULT_ALIGN_BYTES == 16)) \ - || (EIGEN_OS_WIN64 && (EIGEN_DEFAULT_ALIGN_BYTES == 16)) \ - || EIGEN_GLIBC_MALLOC_ALREADY_ALIGNED \ - || EIGEN_FREEBSD_MALLOC_ALREADY_ALIGNED - #define EIGEN_MALLOC_ALREADY_ALIGNED 1 -#else - #define EIGEN_MALLOC_ALREADY_ALIGNED 0 -#endif - -#endif - -namespace Eigen { - -namespace internal { - -EIGEN_DEVICE_FUNC -inline void throw_std_bad_alloc() -{ - #ifdef EIGEN_EXCEPTIONS - throw std::bad_alloc(); - #else - std::size_t huge = static_cast(-1); - ::operator new(huge); - #endif -} - -/***************************************************************************** -*** Implementation of handmade aligned functions *** -*****************************************************************************/ - -/* ----- Hand made implementations of aligned malloc/free and realloc ----- */ - -/** \internal Like malloc, but the returned pointer is guaranteed to be 16-byte aligned. - * Fast, but wastes 16 additional bytes of memory. Does not throw any exception. - */ -inline void* handmade_aligned_malloc(std::size_t size) -{ - void *original = std::malloc(size+EIGEN_DEFAULT_ALIGN_BYTES); - if (original == 0) return 0; - void *aligned = reinterpret_cast((reinterpret_cast(original) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) + EIGEN_DEFAULT_ALIGN_BYTES); - *(reinterpret_cast(aligned) - 1) = original; - return aligned; -} - -/** \internal Frees memory allocated with handmade_aligned_malloc */ -inline void handmade_aligned_free(void *ptr) -{ - if (ptr) std::free(*(reinterpret_cast(ptr) - 1)); -} - -/** \internal - * \brief Reallocates aligned memory. - * Since we know that our handmade version is based on std::malloc - * we can use std::realloc to implement efficient reallocation. - */ -inline void* handmade_aligned_realloc(void* ptr, std::size_t size, std::size_t = 0) -{ - if (ptr == 0) return handmade_aligned_malloc(size); - void *original = *(reinterpret_cast(ptr) - 1); - std::ptrdiff_t previous_offset = static_cast(ptr)-static_cast(original); - original = std::realloc(original,size+EIGEN_DEFAULT_ALIGN_BYTES); - if (original == 0) return 0; - void *aligned = reinterpret_cast((reinterpret_cast(original) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) + EIGEN_DEFAULT_ALIGN_BYTES); - void *previous_aligned = static_cast(original)+previous_offset; - if(aligned!=previous_aligned) - std::memmove(aligned, previous_aligned, size); - - *(reinterpret_cast(aligned) - 1) = original; - return aligned; -} - -/***************************************************************************** -*** Implementation of portable aligned versions of malloc/free/realloc *** -*****************************************************************************/ - -#ifdef EIGEN_NO_MALLOC -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{ - eigen_assert(false && "heap allocation is forbidden (EIGEN_NO_MALLOC is defined)"); -} -#elif defined EIGEN_RUNTIME_NO_MALLOC -EIGEN_DEVICE_FUNC inline bool is_malloc_allowed_impl(bool update, bool new_value = false) -{ - static bool value = true; - if (update == 1) - value = new_value; - return value; -} -EIGEN_DEVICE_FUNC inline bool is_malloc_allowed() { return is_malloc_allowed_impl(false); } -EIGEN_DEVICE_FUNC inline bool set_is_malloc_allowed(bool new_value) { return is_malloc_allowed_impl(true, new_value); } -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{ - eigen_assert(is_malloc_allowed() && "heap allocation is forbidden (EIGEN_RUNTIME_NO_MALLOC is defined and g_is_malloc_allowed is false)"); -} -#else -EIGEN_DEVICE_FUNC inline void check_that_malloc_is_allowed() -{} -#endif - -/** \internal Allocates \a size bytes. The returned pointer is guaranteed to have 16 or 32 bytes alignment depending on the requirements. - * On allocation error, the returned pointer is null, and std::bad_alloc is thrown. - */ -EIGEN_DEVICE_FUNC inline void* aligned_malloc(std::size_t size) -{ - check_that_malloc_is_allowed(); - - void *result; - #if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - result = std::malloc(size); - #if EIGEN_DEFAULT_ALIGN_BYTES==16 - eigen_assert((size<16 || (std::size_t(result)%16)==0) && "System's malloc returned an unaligned pointer. Compile with EIGEN_MALLOC_ALREADY_ALIGNED=0 to fallback to handmade alignd memory allocator."); - #endif - #else - result = handmade_aligned_malloc(size); - #endif - - if(!result && size) - throw_std_bad_alloc(); - - return result; -} - -/** \internal Frees memory allocated with aligned_malloc. */ -EIGEN_DEVICE_FUNC inline void aligned_free(void *ptr) -{ - #if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - std::free(ptr); - #else - handmade_aligned_free(ptr); - #endif -} - -/** - * \internal - * \brief Reallocates an aligned block of memory. - * \throws std::bad_alloc on allocation failure - */ -inline void* aligned_realloc(void *ptr, std::size_t new_size, std::size_t old_size) -{ - EIGEN_UNUSED_VARIABLE(old_size); - - void *result; -#if (EIGEN_DEFAULT_ALIGN_BYTES==0) || EIGEN_MALLOC_ALREADY_ALIGNED - result = std::realloc(ptr,new_size); -#else - result = handmade_aligned_realloc(ptr,new_size,old_size); -#endif - - if (!result && new_size) - throw_std_bad_alloc(); - - return result; -} - -/***************************************************************************** -*** Implementation of conditionally aligned functions *** -*****************************************************************************/ - -/** \internal Allocates \a size bytes. If Align is true, then the returned ptr is 16-byte-aligned. - * On allocation error, the returned pointer is null, and a std::bad_alloc is thrown. - */ -template EIGEN_DEVICE_FUNC inline void* conditional_aligned_malloc(std::size_t size) -{ - return aligned_malloc(size); -} - -template<> EIGEN_DEVICE_FUNC inline void* conditional_aligned_malloc(std::size_t size) -{ - check_that_malloc_is_allowed(); - - void *result = std::malloc(size); - if(!result && size) - throw_std_bad_alloc(); - return result; -} - -/** \internal Frees memory allocated with conditional_aligned_malloc */ -template EIGEN_DEVICE_FUNC inline void conditional_aligned_free(void *ptr) -{ - aligned_free(ptr); -} - -template<> EIGEN_DEVICE_FUNC inline void conditional_aligned_free(void *ptr) -{ - std::free(ptr); -} - -template inline void* conditional_aligned_realloc(void* ptr, std::size_t new_size, std::size_t old_size) -{ - return aligned_realloc(ptr, new_size, old_size); -} - -template<> inline void* conditional_aligned_realloc(void* ptr, std::size_t new_size, std::size_t) -{ - return std::realloc(ptr, new_size); -} - -/***************************************************************************** -*** Construction/destruction of array elements *** -*****************************************************************************/ - -/** \internal Destructs the elements of an array. - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void destruct_elements_of_array(T *ptr, std::size_t size) -{ - // always destruct an array starting from the end. - if(ptr) - while(size) ptr[--size].~T(); -} - -/** \internal Constructs the elements of an array. - * The \a size parameter tells on how many objects to call the constructor of T. - */ -template EIGEN_DEVICE_FUNC inline T* construct_elements_of_array(T *ptr, std::size_t size) -{ - std::size_t i; - EIGEN_TRY - { - for (i = 0; i < size; ++i) ::new (ptr + i) T; - return ptr; - } - EIGEN_CATCH(...) - { - destruct_elements_of_array(ptr, i); - EIGEN_THROW; - } - return NULL; -} - -/***************************************************************************** -*** Implementation of aligned new/delete-like functions *** -*****************************************************************************/ - -template -EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void check_size_for_overflow(std::size_t size) -{ - if(size > std::size_t(-1) / sizeof(T)) - throw_std_bad_alloc(); -} - -/** \internal Allocates \a size objects of type T. The returned pointer is guaranteed to have 16 bytes alignment. - * On allocation error, the returned pointer is undefined, but a std::bad_alloc is thrown. - * The default constructor of T is called. - */ -template EIGEN_DEVICE_FUNC inline T* aligned_new(std::size_t size) -{ - check_size_for_overflow(size); - T *result = reinterpret_cast(aligned_malloc(sizeof(T)*size)); - EIGEN_TRY - { - return construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - aligned_free(result); - EIGEN_THROW; - } - return result; -} - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_new(std::size_t size) -{ - check_size_for_overflow(size); - T *result = reinterpret_cast(conditional_aligned_malloc(sizeof(T)*size)); - EIGEN_TRY - { - return construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - return result; -} - -/** \internal Deletes objects constructed with aligned_new - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void aligned_delete(T *ptr, std::size_t size) -{ - destruct_elements_of_array(ptr, size); - aligned_free(ptr); -} - -/** \internal Deletes objects constructed with conditional_aligned_new - * The \a size parameters tells on how many objects to call the destructor of T. - */ -template EIGEN_DEVICE_FUNC inline void conditional_aligned_delete(T *ptr, std::size_t size) -{ - destruct_elements_of_array(ptr, size); - conditional_aligned_free(ptr); -} - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_realloc_new(T* pts, std::size_t new_size, std::size_t old_size) -{ - check_size_for_overflow(new_size); - check_size_for_overflow(old_size); - if(new_size < old_size) - destruct_elements_of_array(pts+new_size, old_size-new_size); - T *result = reinterpret_cast(conditional_aligned_realloc(reinterpret_cast(pts), sizeof(T)*new_size, sizeof(T)*old_size)); - if(new_size > old_size) - { - EIGEN_TRY - { - construct_elements_of_array(result+old_size, new_size-old_size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - - -template EIGEN_DEVICE_FUNC inline T* conditional_aligned_new_auto(std::size_t size) -{ - if(size==0) - return 0; // short-cut. Also fixes Bug 884 - check_size_for_overflow(size); - T *result = reinterpret_cast(conditional_aligned_malloc(sizeof(T)*size)); - if(NumTraits::RequireInitialization) - { - EIGEN_TRY - { - construct_elements_of_array(result, size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - -template inline T* conditional_aligned_realloc_new_auto(T* pts, std::size_t new_size, std::size_t old_size) -{ - check_size_for_overflow(new_size); - check_size_for_overflow(old_size); - if(NumTraits::RequireInitialization && (new_size < old_size)) - destruct_elements_of_array(pts+new_size, old_size-new_size); - T *result = reinterpret_cast(conditional_aligned_realloc(reinterpret_cast(pts), sizeof(T)*new_size, sizeof(T)*old_size)); - if(NumTraits::RequireInitialization && (new_size > old_size)) - { - EIGEN_TRY - { - construct_elements_of_array(result+old_size, new_size-old_size); - } - EIGEN_CATCH(...) - { - conditional_aligned_free(result); - EIGEN_THROW; - } - } - return result; -} - -template EIGEN_DEVICE_FUNC inline void conditional_aligned_delete_auto(T *ptr, std::size_t size) -{ - if(NumTraits::RequireInitialization) - destruct_elements_of_array(ptr, size); - conditional_aligned_free(ptr); -} - -/****************************************************************************/ - -/** \internal Returns the index of the first element of the array that is well aligned with respect to the requested \a Alignment. - * - * \tparam Alignment requested alignment in Bytes. - * \param array the address of the start of the array - * \param size the size of the array - * - * \note If no element of the array is well aligned or the requested alignment is not a multiple of a scalar, - * the size of the array is returned. For example with SSE, the requested alignment is typically 16-bytes. If - * packet size for the given scalar type is 1, then everything is considered well-aligned. - * - * \note Otherwise, if the Alignment is larger that the scalar size, we rely on the assumptions that sizeof(Scalar) is a - * power of 2. On the other hand, we do not assume that the array address is a multiple of sizeof(Scalar), as that fails for - * example with Scalar=double on certain 32-bit platforms, see bug #79. - * - * There is also the variant first_aligned(const MatrixBase&) defined in DenseCoeffsBase.h. - * \sa first_default_aligned() - */ -template -EIGEN_DEVICE_FUNC inline Index first_aligned(const Scalar* array, Index size) -{ - const Index ScalarSize = sizeof(Scalar); - const Index AlignmentSize = Alignment / ScalarSize; - const Index AlignmentMask = AlignmentSize-1; - - if(AlignmentSize<=1) - { - // Either the requested alignment if smaller than a scalar, or it exactly match a 1 scalar - // so that all elements of the array have the same alignment. - return 0; - } - else if( (UIntPtr(array) & (sizeof(Scalar)-1)) || (Alignment%ScalarSize)!=0) - { - // The array is not aligned to the size of a single scalar, or the requested alignment is not a multiple of the scalar size. - // Consequently, no element of the array is well aligned. - return size; - } - else - { - Index first = (AlignmentSize - (Index((UIntPtr(array)/sizeof(Scalar))) & AlignmentMask)) & AlignmentMask; - return (first < size) ? first : size; - } -} - -/** \internal Returns the index of the first element of the array that is well aligned with respect the largest packet requirement. - * \sa first_aligned(Scalar*,Index) and first_default_aligned(DenseBase) */ -template -EIGEN_DEVICE_FUNC inline Index first_default_aligned(const Scalar* array, Index size) -{ - typedef typename packet_traits::type DefaultPacketType; - return first_aligned::alignment>(array, size); -} - -/** \internal Returns the smallest integer multiple of \a base and greater or equal to \a size - */ -template -inline Index first_multiple(Index size, Index base) -{ - return ((size+base-1)/base)*base; -} - -// std::copy is much slower than memcpy, so let's introduce a smart_copy which -// use memcpy on trivial types, i.e., on types that does not require an initialization ctor. -template struct smart_copy_helper; - -template EIGEN_DEVICE_FUNC void smart_copy(const T* start, const T* end, T* target) -{ - smart_copy_helper::RequireInitialization>::run(start, end, target); -} - -template struct smart_copy_helper { - EIGEN_DEVICE_FUNC static inline void run(const T* start, const T* end, T* target) - { - IntPtr size = IntPtr(end)-IntPtr(start); - if(size==0) return; - eigen_internal_assert(start!=0 && end!=0 && target!=0); - std::memcpy(target, start, size); - } -}; - -template struct smart_copy_helper { - EIGEN_DEVICE_FUNC static inline void run(const T* start, const T* end, T* target) - { std::copy(start, end, target); } -}; - -// intelligent memmove. falls back to std::memmove for POD types, uses std::copy otherwise. -template struct smart_memmove_helper; - -template void smart_memmove(const T* start, const T* end, T* target) -{ - smart_memmove_helper::RequireInitialization>::run(start, end, target); -} - -template struct smart_memmove_helper { - static inline void run(const T* start, const T* end, T* target) - { - IntPtr size = IntPtr(end)-IntPtr(start); - if(size==0) return; - eigen_internal_assert(start!=0 && end!=0 && target!=0); - std::memmove(target, start, size); - } -}; - -template struct smart_memmove_helper { - static inline void run(const T* start, const T* end, T* target) - { - if (UIntPtr(target) < UIntPtr(start)) - { - std::copy(start, end, target); - } - else - { - std::ptrdiff_t count = (std::ptrdiff_t(end)-std::ptrdiff_t(start)) / sizeof(T); - std::copy_backward(start, end, target + count); - } - } -}; - - -/***************************************************************************** -*** Implementation of runtime stack allocation (falling back to malloc) *** -*****************************************************************************/ - -// you can overwrite Eigen's default behavior regarding alloca by defining EIGEN_ALLOCA -// to the appropriate stack allocation function -#ifndef EIGEN_ALLOCA - #if EIGEN_OS_LINUX || EIGEN_OS_MAC || (defined alloca) - #define EIGEN_ALLOCA alloca - #elif EIGEN_COMP_MSVC - #define EIGEN_ALLOCA _alloca - #endif -#endif - -// This helper class construct the allocated memory, and takes care of destructing and freeing the handled data -// at destruction time. In practice this helper class is mainly useful to avoid memory leak in case of exceptions. -template class aligned_stack_memory_handler : noncopyable -{ - public: - /* Creates a stack_memory_handler responsible for the buffer \a ptr of size \a size. - * Note that \a ptr can be 0 regardless of the other parameters. - * This constructor takes care of constructing/initializing the elements of the buffer if required by the scalar type T (see NumTraits::RequireInitialization). - * In this case, the buffer elements will also be destructed when this handler will be destructed. - * Finally, if \a dealloc is true, then the pointer \a ptr is freed. - **/ - aligned_stack_memory_handler(T* ptr, std::size_t size, bool dealloc) - : m_ptr(ptr), m_size(size), m_deallocate(dealloc) - { - if(NumTraits::RequireInitialization && m_ptr) - Eigen::internal::construct_elements_of_array(m_ptr, size); - } - ~aligned_stack_memory_handler() - { - if(NumTraits::RequireInitialization && m_ptr) - Eigen::internal::destruct_elements_of_array(m_ptr, m_size); - if(m_deallocate) - Eigen::internal::aligned_free(m_ptr); - } - protected: - T* m_ptr; - std::size_t m_size; - bool m_deallocate; -}; - -template class scoped_array : noncopyable -{ - T* m_ptr; -public: - explicit scoped_array(std::ptrdiff_t size) - { - m_ptr = new T[size]; - } - ~scoped_array() - { - delete[] m_ptr; - } - T& operator[](std::ptrdiff_t i) { return m_ptr[i]; } - const T& operator[](std::ptrdiff_t i) const { return m_ptr[i]; } - T* &ptr() { return m_ptr; } - const T* ptr() const { return m_ptr; } - operator const T*() const { return m_ptr; } -}; - -template void swap(scoped_array &a,scoped_array &b) -{ - std::swap(a.ptr(),b.ptr()); -} - -} // end namespace internal - -/** \internal - * Declares, allocates and construct an aligned buffer named NAME of SIZE elements of type TYPE on the stack - * if SIZE is smaller than EIGEN_STACK_ALLOCATION_LIMIT, and if stack allocation is supported by the platform - * (currently, this is Linux and Visual Studio only). Otherwise the memory is allocated on the heap. - * The allocated buffer is automatically deleted when exiting the scope of this declaration. - * If BUFFER is non null, then the declared variable is simply an alias for BUFFER, and no allocation/deletion occurs. - * Here is an example: - * \code - * { - * ei_declare_aligned_stack_constructed_variable(float,data,size,0); - * // use data[0] to data[size-1] - * } - * \endcode - * The underlying stack allocation function can controlled with the EIGEN_ALLOCA preprocessor token. - */ -#ifdef EIGEN_ALLOCA - - #if EIGEN_DEFAULT_ALIGN_BYTES>0 - // We always manually re-align the result of EIGEN_ALLOCA. - // If alloca is already aligned, the compiler should be smart enough to optimize away the re-alignment. - #define EIGEN_ALIGNED_ALLOCA(SIZE) reinterpret_cast((internal::UIntPtr(EIGEN_ALLOCA(SIZE+EIGEN_DEFAULT_ALIGN_BYTES-1)) + EIGEN_DEFAULT_ALIGN_BYTES-1) & ~(std::size_t(EIGEN_DEFAULT_ALIGN_BYTES-1))) - #else - #define EIGEN_ALIGNED_ALLOCA(SIZE) EIGEN_ALLOCA(SIZE) - #endif - - #define ei_declare_aligned_stack_constructed_variable(TYPE,NAME,SIZE,BUFFER) \ - Eigen::internal::check_size_for_overflow(SIZE); \ - TYPE* NAME = (BUFFER)!=0 ? (BUFFER) \ - : reinterpret_cast( \ - (sizeof(TYPE)*SIZE<=EIGEN_STACK_ALLOCATION_LIMIT) ? EIGEN_ALIGNED_ALLOCA(sizeof(TYPE)*SIZE) \ - : Eigen::internal::aligned_malloc(sizeof(TYPE)*SIZE) ); \ - Eigen::internal::aligned_stack_memory_handler EIGEN_CAT(NAME,_stack_memory_destructor)((BUFFER)==0 ? NAME : 0,SIZE,sizeof(TYPE)*SIZE>EIGEN_STACK_ALLOCATION_LIMIT) - -#else - - #define ei_declare_aligned_stack_constructed_variable(TYPE,NAME,SIZE,BUFFER) \ - Eigen::internal::check_size_for_overflow(SIZE); \ - TYPE* NAME = (BUFFER)!=0 ? BUFFER : reinterpret_cast(Eigen::internal::aligned_malloc(sizeof(TYPE)*SIZE)); \ - Eigen::internal::aligned_stack_memory_handler EIGEN_CAT(NAME,_stack_memory_destructor)((BUFFER)==0 ? NAME : 0,SIZE,true) - -#endif - - -/***************************************************************************** -*** Implementation of EIGEN_MAKE_ALIGNED_OPERATOR_NEW [_IF] *** -*****************************************************************************/ - -#if EIGEN_MAX_ALIGN_BYTES!=0 - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_NOTHROW(NeedsToAlign) \ - void* operator new(std::size_t size, const std::nothrow_t&) EIGEN_NO_THROW { \ - EIGEN_TRY { return Eigen::internal::conditional_aligned_malloc(size); } \ - EIGEN_CATCH (...) { return 0; } \ - } - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) \ - void *operator new(std::size_t size) { \ - return Eigen::internal::conditional_aligned_malloc(size); \ - } \ - void *operator new[](std::size_t size) { \ - return Eigen::internal::conditional_aligned_malloc(size); \ - } \ - void operator delete(void * ptr) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete[](void * ptr) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete(void * ptr, std::size_t /* sz */) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - void operator delete[](void * ptr, std::size_t /* sz */) EIGEN_NO_THROW { Eigen::internal::conditional_aligned_free(ptr); } \ - /* in-place new and delete. since (at least afaik) there is no actual */ \ - /* memory allocated we can safely let the default implementation handle */ \ - /* this particular case. */ \ - static void *operator new(std::size_t size, void *ptr) { return ::operator new(size,ptr); } \ - static void *operator new[](std::size_t size, void* ptr) { return ::operator new[](size,ptr); } \ - void operator delete(void * memory, void *ptr) EIGEN_NO_THROW { return ::operator delete(memory,ptr); } \ - void operator delete[](void * memory, void *ptr) EIGEN_NO_THROW { return ::operator delete[](memory,ptr); } \ - /* nothrow-new (returns zero instead of std::bad_alloc) */ \ - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_NOTHROW(NeedsToAlign) \ - void operator delete(void *ptr, const std::nothrow_t&) EIGEN_NO_THROW { \ - Eigen::internal::conditional_aligned_free(ptr); \ - } \ - typedef void eigen_aligned_operator_new_marker_type; -#else - #define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign) -#endif - -#define EIGEN_MAKE_ALIGNED_OPERATOR_NEW EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(true) -#define EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(Scalar,Size) \ - EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(bool(((Size)!=Eigen::Dynamic) && ((sizeof(Scalar)*(Size))%EIGEN_MAX_ALIGN_BYTES==0))) - -/****************************************************************************/ - -/** \class aligned_allocator -* \ingroup Core_Module -* -* \brief STL compatible allocator to use with types requiring a non standrad alignment. -* -* The memory is aligned as for dynamically aligned matrix/array types such as MatrixXd. -* By default, it will thus provide at least 16 bytes alignment and more in following cases: -* - 32 bytes alignment if AVX is enabled. -* - 64 bytes alignment if AVX512 is enabled. -* -* This can be controled using the \c EIGEN_MAX_ALIGN_BYTES macro as documented -* \link TopicPreprocessorDirectivesPerformance there \endlink. -* -* Example: -* \code -* // Matrix4f requires 16 bytes alignment: -* std::map< int, Matrix4f, std::less, -* aligned_allocator > > my_map_mat4; -* // Vector3f does not require 16 bytes alignment, no need to use Eigen's allocator: -* std::map< int, Vector3f > my_map_vec3; -* \endcode -* -* \sa \blank \ref TopicStlContainers. -*/ -template -class aligned_allocator : public std::allocator -{ -public: - typedef std::size_t size_type; - typedef std::ptrdiff_t difference_type; - typedef T* pointer; - typedef const T* const_pointer; - typedef T& reference; - typedef const T& const_reference; - typedef T value_type; - - template - struct rebind - { - typedef aligned_allocator other; - }; - - aligned_allocator() : std::allocator() {} - - aligned_allocator(const aligned_allocator& other) : std::allocator(other) {} - - template - aligned_allocator(const aligned_allocator& other) : std::allocator(other) {} - - ~aligned_allocator() {} - - pointer allocate(size_type num, const void* /*hint*/ = 0) - { - internal::check_size_for_overflow(num); - size_type size = num * sizeof(T); -#if EIGEN_COMP_GNUC_STRICT && EIGEN_GNUC_AT_LEAST(7,0) - // workaround gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87544 - // It triggered eigen/Eigen/src/Core/util/Memory.h:189:12: warning: argument 1 value '18446744073709551612' exceeds maximum object size 9223372036854775807 - if(size>=std::size_t((std::numeric_limits::max)())) - return 0; - else -#endif - return static_cast( internal::aligned_malloc(size) ); - } - - void deallocate(pointer p, size_type /*num*/) - { - internal::aligned_free(p); - } -}; - -//---------- Cache sizes ---------- - -#if !defined(EIGEN_NO_CPUID) -# if EIGEN_COMP_GNUC && EIGEN_ARCH_i386_OR_x86_64 -# if defined(__PIC__) && EIGEN_ARCH_i386 - // Case for x86 with PIC -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("xchgl %%ebx, %k1;cpuid; xchgl %%ebx,%k1": "=a" (abcd[0]), "=&r" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "a" (func), "c" (id)); -# elif defined(__PIC__) && EIGEN_ARCH_x86_64 - // Case for x64 with PIC. In theory this is only a problem with recent gcc and with medium or large code model, not with the default small code model. - // However, we cannot detect which code model is used, and the xchg overhead is negligible anyway. -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("xchg{q}\t{%%}rbx, %q1; cpuid; xchg{q}\t{%%}rbx, %q1": "=a" (abcd[0]), "=&r" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "0" (func), "2" (id)); -# else - // Case for x86_64 or x86 w/o PIC -# define EIGEN_CPUID(abcd,func,id) \ - __asm__ __volatile__ ("cpuid": "=a" (abcd[0]), "=b" (abcd[1]), "=c" (abcd[2]), "=d" (abcd[3]) : "0" (func), "2" (id) ); -# endif -# elif EIGEN_COMP_MSVC -# if (EIGEN_COMP_MSVC > 1500) && EIGEN_ARCH_i386_OR_x86_64 -# define EIGEN_CPUID(abcd,func,id) __cpuidex((int*)abcd,func,id) -# endif -# endif -#endif - -namespace internal { - -#ifdef EIGEN_CPUID - -inline bool cpuid_is_vendor(int abcd[4], const int vendor[3]) -{ - return abcd[1]==vendor[0] && abcd[3]==vendor[1] && abcd[2]==vendor[2]; -} - -inline void queryCacheSizes_intel_direct(int& l1, int& l2, int& l3) -{ - int abcd[4]; - l1 = l2 = l3 = 0; - int cache_id = 0; - int cache_type = 0; - do { - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x4,cache_id); - cache_type = (abcd[0] & 0x0F) >> 0; - if(cache_type==1||cache_type==3) // data or unified cache - { - int cache_level = (abcd[0] & 0xE0) >> 5; // A[7:5] - int ways = (abcd[1] & 0xFFC00000) >> 22; // B[31:22] - int partitions = (abcd[1] & 0x003FF000) >> 12; // B[21:12] - int line_size = (abcd[1] & 0x00000FFF) >> 0; // B[11:0] - int sets = (abcd[2]); // C[31:0] - - int cache_size = (ways+1) * (partitions+1) * (line_size+1) * (sets+1); - - switch(cache_level) - { - case 1: l1 = cache_size; break; - case 2: l2 = cache_size; break; - case 3: l3 = cache_size; break; - default: break; - } - } - cache_id++; - } while(cache_type>0 && cache_id<16); -} - -inline void queryCacheSizes_intel_codes(int& l1, int& l2, int& l3) -{ - int abcd[4]; - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - l1 = l2 = l3 = 0; - EIGEN_CPUID(abcd,0x00000002,0); - unsigned char * bytes = reinterpret_cast(abcd)+2; - bool check_for_p2_core2 = false; - for(int i=0; i<14; ++i) - { - switch(bytes[i]) - { - case 0x0A: l1 = 8; break; // 0Ah data L1 cache, 8 KB, 2 ways, 32 byte lines - case 0x0C: l1 = 16; break; // 0Ch data L1 cache, 16 KB, 4 ways, 32 byte lines - case 0x0E: l1 = 24; break; // 0Eh data L1 cache, 24 KB, 6 ways, 64 byte lines - case 0x10: l1 = 16; break; // 10h data L1 cache, 16 KB, 4 ways, 32 byte lines (IA-64) - case 0x15: l1 = 16; break; // 15h code L1 cache, 16 KB, 4 ways, 32 byte lines (IA-64) - case 0x2C: l1 = 32; break; // 2Ch data L1 cache, 32 KB, 8 ways, 64 byte lines - case 0x30: l1 = 32; break; // 30h code L1 cache, 32 KB, 8 ways, 64 byte lines - case 0x60: l1 = 16; break; // 60h data L1 cache, 16 KB, 8 ways, 64 byte lines, sectored - case 0x66: l1 = 8; break; // 66h data L1 cache, 8 KB, 4 ways, 64 byte lines, sectored - case 0x67: l1 = 16; break; // 67h data L1 cache, 16 KB, 4 ways, 64 byte lines, sectored - case 0x68: l1 = 32; break; // 68h data L1 cache, 32 KB, 4 ways, 64 byte lines, sectored - case 0x1A: l2 = 96; break; // code and data L2 cache, 96 KB, 6 ways, 64 byte lines (IA-64) - case 0x22: l3 = 512; break; // code and data L3 cache, 512 KB, 4 ways (!), 64 byte lines, dual-sectored - case 0x23: l3 = 1024; break; // code and data L3 cache, 1024 KB, 8 ways, 64 byte lines, dual-sectored - case 0x25: l3 = 2048; break; // code and data L3 cache, 2048 KB, 8 ways, 64 byte lines, dual-sectored - case 0x29: l3 = 4096; break; // code and data L3 cache, 4096 KB, 8 ways, 64 byte lines, dual-sectored - case 0x39: l2 = 128; break; // code and data L2 cache, 128 KB, 4 ways, 64 byte lines, sectored - case 0x3A: l2 = 192; break; // code and data L2 cache, 192 KB, 6 ways, 64 byte lines, sectored - case 0x3B: l2 = 128; break; // code and data L2 cache, 128 KB, 2 ways, 64 byte lines, sectored - case 0x3C: l2 = 256; break; // code and data L2 cache, 256 KB, 4 ways, 64 byte lines, sectored - case 0x3D: l2 = 384; break; // code and data L2 cache, 384 KB, 6 ways, 64 byte lines, sectored - case 0x3E: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 64 byte lines, sectored - case 0x40: l2 = 0; break; // no integrated L2 cache (P6 core) or L3 cache (P4 core) - case 0x41: l2 = 128; break; // code and data L2 cache, 128 KB, 4 ways, 32 byte lines - case 0x42: l2 = 256; break; // code and data L2 cache, 256 KB, 4 ways, 32 byte lines - case 0x43: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 32 byte lines - case 0x44: l2 = 1024; break; // code and data L2 cache, 1024 KB, 4 ways, 32 byte lines - case 0x45: l2 = 2048; break; // code and data L2 cache, 2048 KB, 4 ways, 32 byte lines - case 0x46: l3 = 4096; break; // code and data L3 cache, 4096 KB, 4 ways, 64 byte lines - case 0x47: l3 = 8192; break; // code and data L3 cache, 8192 KB, 8 ways, 64 byte lines - case 0x48: l2 = 3072; break; // code and data L2 cache, 3072 KB, 12 ways, 64 byte lines - case 0x49: if(l2!=0) l3 = 4096; else {check_for_p2_core2=true; l3 = l2 = 4096;} break;// code and data L3 cache, 4096 KB, 16 ways, 64 byte lines (P4) or L2 for core2 - case 0x4A: l3 = 6144; break; // code and data L3 cache, 6144 KB, 12 ways, 64 byte lines - case 0x4B: l3 = 8192; break; // code and data L3 cache, 8192 KB, 16 ways, 64 byte lines - case 0x4C: l3 = 12288; break; // code and data L3 cache, 12288 KB, 12 ways, 64 byte lines - case 0x4D: l3 = 16384; break; // code and data L3 cache, 16384 KB, 16 ways, 64 byte lines - case 0x4E: l2 = 6144; break; // code and data L2 cache, 6144 KB, 24 ways, 64 byte lines - case 0x78: l2 = 1024; break; // code and data L2 cache, 1024 KB, 4 ways, 64 byte lines - case 0x79: l2 = 128; break; // code and data L2 cache, 128 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7A: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7B: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7C: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 64 byte lines, dual-sectored - case 0x7D: l2 = 2048; break; // code and data L2 cache, 2048 KB, 8 ways, 64 byte lines - case 0x7E: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 128 byte lines, sect. (IA-64) - case 0x7F: l2 = 512; break; // code and data L2 cache, 512 KB, 2 ways, 64 byte lines - case 0x80: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 64 byte lines - case 0x81: l2 = 128; break; // code and data L2 cache, 128 KB, 8 ways, 32 byte lines - case 0x82: l2 = 256; break; // code and data L2 cache, 256 KB, 8 ways, 32 byte lines - case 0x83: l2 = 512; break; // code and data L2 cache, 512 KB, 8 ways, 32 byte lines - case 0x84: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 32 byte lines - case 0x85: l2 = 2048; break; // code and data L2 cache, 2048 KB, 8 ways, 32 byte lines - case 0x86: l2 = 512; break; // code and data L2 cache, 512 KB, 4 ways, 64 byte lines - case 0x87: l2 = 1024; break; // code and data L2 cache, 1024 KB, 8 ways, 64 byte lines - case 0x88: l3 = 2048; break; // code and data L3 cache, 2048 KB, 4 ways, 64 byte lines (IA-64) - case 0x89: l3 = 4096; break; // code and data L3 cache, 4096 KB, 4 ways, 64 byte lines (IA-64) - case 0x8A: l3 = 8192; break; // code and data L3 cache, 8192 KB, 4 ways, 64 byte lines (IA-64) - case 0x8D: l3 = 3072; break; // code and data L3 cache, 3072 KB, 12 ways, 128 byte lines (IA-64) - - default: break; - } - } - if(check_for_p2_core2 && l2 == l3) - l3 = 0; - l1 *= 1024; - l2 *= 1024; - l3 *= 1024; -} - -inline void queryCacheSizes_intel(int& l1, int& l2, int& l3, int max_std_funcs) -{ - if(max_std_funcs>=4) - queryCacheSizes_intel_direct(l1,l2,l3); - else - queryCacheSizes_intel_codes(l1,l2,l3); -} - -inline void queryCacheSizes_amd(int& l1, int& l2, int& l3) -{ - int abcd[4]; - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x80000005,0); - l1 = (abcd[2] >> 24) * 1024; // C[31:24] = L1 size in KB - abcd[0] = abcd[1] = abcd[2] = abcd[3] = 0; - EIGEN_CPUID(abcd,0x80000006,0); - l2 = (abcd[2] >> 16) * 1024; // C[31;16] = l2 cache size in KB - l3 = ((abcd[3] & 0xFFFC000) >> 18) * 512 * 1024; // D[31;18] = l3 cache size in 512KB -} -#endif - -/** \internal - * Queries and returns the cache sizes in Bytes of the L1, L2, and L3 data caches respectively */ -inline void queryCacheSizes(int& l1, int& l2, int& l3) -{ - #ifdef EIGEN_CPUID - int abcd[4]; - const int GenuineIntel[] = {0x756e6547, 0x49656e69, 0x6c65746e}; - const int AuthenticAMD[] = {0x68747541, 0x69746e65, 0x444d4163}; - const int AMDisbetter_[] = {0x69444d41, 0x74656273, 0x21726574}; // "AMDisbetter!" - - // identify the CPU vendor - EIGEN_CPUID(abcd,0x0,0); - int max_std_funcs = abcd[1]; - if(cpuid_is_vendor(abcd,GenuineIntel)) - queryCacheSizes_intel(l1,l2,l3,max_std_funcs); - else if(cpuid_is_vendor(abcd,AuthenticAMD) || cpuid_is_vendor(abcd,AMDisbetter_)) - queryCacheSizes_amd(l1,l2,l3); - else - // by default let's use Intel's API - queryCacheSizes_intel(l1,l2,l3,max_std_funcs); - - // here is the list of other vendors: -// ||cpuid_is_vendor(abcd,"VIA VIA VIA ") -// ||cpuid_is_vendor(abcd,"CyrixInstead") -// ||cpuid_is_vendor(abcd,"CentaurHauls") -// ||cpuid_is_vendor(abcd,"GenuineTMx86") -// ||cpuid_is_vendor(abcd,"TransmetaCPU") -// ||cpuid_is_vendor(abcd,"RiseRiseRise") -// ||cpuid_is_vendor(abcd,"Geode by NSC") -// ||cpuid_is_vendor(abcd,"SiS SiS SiS ") -// ||cpuid_is_vendor(abcd,"UMC UMC UMC ") -// ||cpuid_is_vendor(abcd,"NexGenDriven") - #else - l1 = l2 = l3 = -1; - #endif -} - -/** \internal - * \returns the size in Bytes of the L1 data cache */ -inline int queryL1CacheSize() -{ - int l1(-1), l2, l3; - queryCacheSizes(l1,l2,l3); - return l1; -} - -/** \internal - * \returns the size in Bytes of the L2 or L3 cache if this later is present */ -inline int queryTopLevelCacheSize() -{ - int l1, l2(-1), l3(-1); - queryCacheSizes(l1,l2,l3); - return (std::max)(l2,l3); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_MEMORY_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h deleted file mode 100644 index 9b61ff037aa..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/Meta.h +++ /dev/null @@ -1,568 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2015 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_META_H -#define EIGEN_META_H - -#if defined(__CUDA_ARCH__) -#include -#include -#endif - -#if EIGEN_COMP_ICC>=1600 && __cplusplus >= 201103L -#include -#endif - -namespace Eigen { - -typedef EIGEN_DEFAULT_DENSE_INDEX_TYPE DenseIndex; - -/** - * \brief The Index type as used for the API. - * \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE. - * \sa \blank \ref TopicPreprocessorDirectives, StorageIndex. - */ - -typedef EIGEN_DEFAULT_DENSE_INDEX_TYPE Index; - -namespace internal { - -/** \internal - * \file Meta.h - * This file contains generic metaprogramming classes which are not specifically related to Eigen. - * \note In case you wonder, yes we're aware that Boost already provides all these features, - * we however don't want to add a dependency to Boost. - */ - -// Only recent versions of ICC complain about using ptrdiff_t to hold pointers, -// and older versions do not provide *intptr_t types. -#if EIGEN_COMP_ICC>=1600 && __cplusplus >= 201103L -typedef std::intptr_t IntPtr; -typedef std::uintptr_t UIntPtr; -#else -typedef std::ptrdiff_t IntPtr; -typedef std::size_t UIntPtr; -#endif - -struct true_type { enum { value = 1 }; }; -struct false_type { enum { value = 0 }; }; - -template -struct conditional { typedef Then type; }; - -template -struct conditional { typedef Else type; }; - -template struct is_same { enum { value = 0 }; }; -template struct is_same { enum { value = 1 }; }; - -template struct remove_reference { typedef T type; }; -template struct remove_reference { typedef T type; }; - -template struct remove_pointer { typedef T type; }; -template struct remove_pointer { typedef T type; }; -template struct remove_pointer { typedef T type; }; - -template struct remove_const { typedef T type; }; -template struct remove_const { typedef T type; }; -template struct remove_const { typedef T type[]; }; -template struct remove_const { typedef T type[Size]; }; - -template struct remove_all { typedef T type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; -template struct remove_all { typedef typename remove_all::type type; }; - -template struct is_arithmetic { enum { value = false }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic{ enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; -template<> struct is_arithmetic { enum { value = true }; }; - -#if EIGEN_HAS_CXX11 -using std::is_integral; -#else -template struct is_integral { enum { value = false }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral { enum { value = true }; }; -#if EIGEN_COMP_MSVC -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral{ enum { value = true }; }; -#endif -#endif - -#if EIGEN_HAS_CXX11 -using std::make_unsigned; -#else -// TODO: Possibly improve this implementation of make_unsigned. -// It is currently used only by -// template struct random_default_impl. -template struct make_unsigned; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned char type; }; -template<> struct make_unsigned { typedef unsigned short type; }; -template<> struct make_unsigned { typedef unsigned short type; }; -template<> struct make_unsigned { typedef unsigned int type; }; -template<> struct make_unsigned { typedef unsigned int type; }; -template<> struct make_unsigned { typedef unsigned long type; }; -template<> struct make_unsigned { typedef unsigned long type; }; -#if EIGEN_COMP_MSVC -template<> struct make_unsigned { typedef unsigned __int64 type; }; -template<> struct make_unsigned { typedef unsigned __int64 type; }; -#endif -#endif - -template struct add_const { typedef const T type; }; -template struct add_const { typedef T& type; }; - -template struct is_const { enum { value = 0 }; }; -template struct is_const { enum { value = 1 }; }; - -template struct add_const_on_value_type { typedef const T type; }; -template struct add_const_on_value_type { typedef T const& type; }; -template struct add_const_on_value_type { typedef T const* type; }; -template struct add_const_on_value_type { typedef T const* const type; }; -template struct add_const_on_value_type { typedef T const* const type; }; - - -template -struct is_convertible_impl -{ -private: - struct any_conversion - { - template any_conversion(const volatile T&); - template any_conversion(T&); - }; - struct yes {int a[1];}; - struct no {int a[2];}; - - static yes test(const To&, int); - static no test(any_conversion, ...); - -public: - static From ms_from; -#ifdef __INTEL_COMPILER - #pragma warning push - #pragma warning ( disable : 2259 ) -#endif - enum { value = sizeof(test(ms_from, 0))==sizeof(yes) }; -#ifdef __INTEL_COMPILER - #pragma warning pop -#endif -}; - -template -struct is_convertible -{ - enum { value = is_convertible_impl::type, - typename remove_all::type>::value }; -}; - -/** \internal Allows to enable/disable an overload - * according to a compile time condition. - */ -template struct enable_if; - -template struct enable_if -{ typedef T type; }; - -#if defined(__CUDA_ARCH__) -#if !defined(__FLT_EPSILON__) -#define __FLT_EPSILON__ FLT_EPSILON -#define __DBL_EPSILON__ DBL_EPSILON -#endif - -namespace device { - -template struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static T epsilon() { return 0; } - static T (max)() { assert(false && "Highest not supported for this type"); } - static T (min)() { assert(false && "Lowest not supported for this type"); } - static T infinity() { assert(false && "Infinity not supported for this type"); } - static T quiet_NaN() { assert(false && "quiet_NaN not supported for this type"); } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static float epsilon() { return __FLT_EPSILON__; } - EIGEN_DEVICE_FUNC - static float (max)() { return CUDART_MAX_NORMAL_F; } - EIGEN_DEVICE_FUNC - static float (min)() { return FLT_MIN; } - EIGEN_DEVICE_FUNC - static float infinity() { return CUDART_INF_F; } - EIGEN_DEVICE_FUNC - static float quiet_NaN() { return CUDART_NAN_F; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static double epsilon() { return __DBL_EPSILON__; } - EIGEN_DEVICE_FUNC - static double (max)() { return DBL_MAX; } - EIGEN_DEVICE_FUNC - static double (min)() { return DBL_MIN; } - EIGEN_DEVICE_FUNC - static double infinity() { return CUDART_INF; } - EIGEN_DEVICE_FUNC - static double quiet_NaN() { return CUDART_NAN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static int epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static int (max)() { return INT_MAX; } - EIGEN_DEVICE_FUNC - static int (min)() { return INT_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned int epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned int (max)() { return UINT_MAX; } - EIGEN_DEVICE_FUNC - static unsigned int (min)() { return 0; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static long (max)() { return LONG_MAX; } - EIGEN_DEVICE_FUNC - static long (min)() { return LONG_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned long (max)() { return ULONG_MAX; } - EIGEN_DEVICE_FUNC - static unsigned long (min)() { return 0; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static long long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static long long (max)() { return LLONG_MAX; } - EIGEN_DEVICE_FUNC - static long long (min)() { return LLONG_MIN; } -}; -template<> struct numeric_limits -{ - EIGEN_DEVICE_FUNC - static unsigned long long epsilon() { return 0; } - EIGEN_DEVICE_FUNC - static unsigned long long (max)() { return ULLONG_MAX; } - EIGEN_DEVICE_FUNC - static unsigned long long (min)() { return 0; } -}; - -} - -#endif - -/** \internal - * A base class do disable default copy ctor and copy assignement operator. - */ -class noncopyable -{ - EIGEN_DEVICE_FUNC noncopyable(const noncopyable&); - EIGEN_DEVICE_FUNC const noncopyable& operator=(const noncopyable&); -protected: - EIGEN_DEVICE_FUNC noncopyable() {} - EIGEN_DEVICE_FUNC ~noncopyable() {} -}; - -/** \internal - * Convenient struct to get the result type of a unary or binary functor. - * - * It supports both the current STL mechanism (using the result_type member) as well as - * upcoming next STL generation (using a templated result member). - * If none of these members is provided, then the type of the first argument is returned. FIXME, that behavior is a pretty bad hack. - */ -#if EIGEN_HAS_STD_RESULT_OF -template struct result_of { - typedef typename std::result_of::type type1; - typedef typename remove_all::type type; -}; -#else -template struct result_of { }; - -struct has_none {int a[1];}; -struct has_std_result_type {int a[2];}; -struct has_tr1_result {int a[3];}; - -template -struct unary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct unary_result_of_select {typedef typename Func::result_type type;}; - -template -struct unary_result_of_select {typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename unary_result_of_select::type type; -}; - -template -struct binary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct binary_result_of_select -{typedef typename Func::result_type type;}; - -template -struct binary_result_of_select -{typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename binary_result_of_select::type type; -}; - -template -struct ternary_result_of_select {typedef typename internal::remove_all::type type;}; - -template -struct ternary_result_of_select -{typedef typename Func::result_type type;}; - -template -struct ternary_result_of_select -{typedef typename Func::template result::type type;}; - -template -struct result_of { - template - static has_std_result_type testFunctor(T const *, typename T::result_type const * = 0); - template - static has_tr1_result testFunctor(T const *, typename T::template result::type const * = 0); - static has_none testFunctor(...); - - // note that the following indirection is needed for gcc-3.3 - enum {FunctorType = sizeof(testFunctor(static_cast(0)))}; - typedef typename ternary_result_of_select::type type; -}; -#endif - -struct meta_yes { char a[1]; }; -struct meta_no { char a[2]; }; - -// Check whether T::ReturnType does exist -template -struct has_ReturnType -{ - template static meta_yes testFunctor(typename C::ReturnType const *); - template static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(0)) == sizeof(meta_yes) }; -}; - -template const T* return_ptr(); - -template -struct has_nullary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()())>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -template -struct has_unary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()(IndexType(0)))>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -template -struct has_binary_operator -{ - template static meta_yes testFunctor(C const *,typename enable_if<(sizeof(return_ptr()->operator()(IndexType(0),IndexType(0)))>0)>::type * = 0); - static meta_no testFunctor(...); - - enum { value = sizeof(testFunctor(static_cast(0))) == sizeof(meta_yes) }; -}; - -/** \internal In short, it computes int(sqrt(\a Y)) with \a Y an integer. - * Usage example: \code meta_sqrt<1023>::ret \endcode - */ -template Y))) > - // use ?: instead of || just to shut up a stupid gcc 4.3 warning -class meta_sqrt -{ - enum { - MidX = (InfX+SupX)/2, - TakeInf = MidX*MidX > Y ? 1 : 0, - NewInf = int(TakeInf) ? InfX : int(MidX), - NewSup = int(TakeInf) ? int(MidX) : SupX - }; - public: - enum { ret = meta_sqrt::ret }; -}; - -template -class meta_sqrt { public: enum { ret = (SupX*SupX <= Y) ? SupX : InfX }; }; - - -/** \internal Computes the least common multiple of two positive integer A and B - * at compile-time. It implements a naive algorithm testing all multiples of A. - * It thus works better if A>=B. - */ -template -struct meta_least_common_multiple -{ - enum { ret = meta_least_common_multiple::ret }; -}; -template -struct meta_least_common_multiple -{ - enum { ret = A*K }; -}; - -/** \internal determines whether the product of two numeric types is allowed and what the return type is */ -template struct scalar_product_traits -{ - enum { Defined = 0 }; -}; - -// FIXME quick workaround around current limitation of result_of -// template -// struct result_of(ArgType0,ArgType1)> { -// typedef typename scalar_product_traits::type, typename remove_all::type>::ReturnType type; -// }; - -} // end namespace internal - -namespace numext { - -#if defined(__CUDA_ARCH__) -template EIGEN_DEVICE_FUNC void swap(T &a, T &b) { T tmp = b; b = a; a = tmp; } -#else -template EIGEN_STRONG_INLINE void swap(T &a, T &b) { std::swap(a,b); } -#endif - -#if defined(__CUDA_ARCH__) -using internal::device::numeric_limits; -#else -using std::numeric_limits; -#endif - -// Integer division with rounding up. -// T is assumed to be an integer type with a>=0, and b>0 -template -T div_ceil(const T &a, const T &b) -{ - return (a+b-1) / b; -} - -// The aim of the following functions is to bypass -Wfloat-equal warnings -// when we really want a strict equality comparison on floating points. -template EIGEN_STRONG_INLINE -bool equal_strict(const X& x,const Y& y) { return x == y; } - -template<> EIGEN_STRONG_INLINE -bool equal_strict(const float& x,const float& y) { return std::equal_to()(x,y); } - -template<> EIGEN_STRONG_INLINE -bool equal_strict(const double& x,const double& y) { return std::equal_to()(x,y); } - -template EIGEN_STRONG_INLINE -bool not_equal_strict(const X& x,const Y& y) { return x != y; } - -template<> EIGEN_STRONG_INLINE -bool not_equal_strict(const float& x,const float& y) { return std::not_equal_to()(x,y); } - -template<> EIGEN_STRONG_INLINE -bool not_equal_strict(const double& x,const double& y) { return std::not_equal_to()(x,y); } - -} // end namespace numext - -} // end namespace Eigen - -// Define portable (u)int{32,64} types -#if EIGEN_HAS_CXX11 -#include -namespace Eigen { -namespace numext { -typedef std::uint32_t uint32_t; -typedef std::int32_t int32_t; -typedef std::uint64_t uint64_t; -typedef std::int64_t int64_t; -} -} -#else -// Without c++11, all compilers able to compile Eigen also -// provides the C99 stdint.h header file. -#include -namespace Eigen { -namespace numext { -typedef ::uint32_t uint32_t; -typedef ::int32_t int32_t; -typedef ::uint64_t uint64_t; -typedef ::int64_t int64_t; -} -} -#endif - - -#endif // EIGEN_META_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h deleted file mode 100644 index 1ce6fd1b009..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/ReenableStupidWarnings.h +++ /dev/null @@ -1,31 +0,0 @@ -#ifdef EIGEN_WARNINGS_DISABLED_2 -// "DisableStupidWarnings.h" was included twice recursively: Do not reenable warnings yet! -# undef EIGEN_WARNINGS_DISABLED_2 - -#elif defined(EIGEN_WARNINGS_DISABLED) -#undef EIGEN_WARNINGS_DISABLED - -#ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS - #ifdef _MSC_VER - #pragma warning( pop ) - #elif defined __INTEL_COMPILER - #pragma warning pop - #elif defined __clang__ - #pragma clang diagnostic pop - #elif defined __GNUC__ && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) - #pragma GCC diagnostic pop - #endif - - #if defined __NVCC__ -// Don't reenable the diagnostic messages, as it turns out these messages need -// to be disabled at the point of the template instantiation (i.e the user code) -// otherwise they'll be triggered by nvcc. -// #pragma diag_default code_is_unreachable -// #pragma diag_default initialization_not_reachable -// #pragma diag_default 2651 -// #pragma diag_default 2653 - #endif - -#endif - -#endif // EIGEN_WARNINGS_DISABLED diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h deleted file mode 100644 index 500e47792a4..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/StaticAssert.h +++ /dev/null @@ -1,218 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_STATIC_ASSERT_H -#define EIGEN_STATIC_ASSERT_H - -/* Some notes on Eigen's static assertion mechanism: - * - * - in EIGEN_STATIC_ASSERT(CONDITION,MSG) the parameter CONDITION must be a compile time boolean - * expression, and MSG an enum listed in struct internal::static_assertion - * - * - define EIGEN_NO_STATIC_ASSERT to disable them (and save compilation time) - * in that case, the static assertion is converted to the following runtime assert: - * eigen_assert(CONDITION && "MSG") - * - * - currently EIGEN_STATIC_ASSERT can only be used in function scope - * - */ - -#ifndef EIGEN_STATIC_ASSERT -#ifndef EIGEN_NO_STATIC_ASSERT - - #if EIGEN_MAX_CPP_VER>=11 && (__has_feature(cxx_static_assert) || (defined(__cplusplus) && __cplusplus >= 201103L) || (EIGEN_COMP_MSVC >= 1600)) - - // if native static_assert is enabled, let's use it - #define EIGEN_STATIC_ASSERT(X,MSG) static_assert(X,#MSG); - - #else // not CXX0X - - namespace Eigen { - - namespace internal { - - template - struct static_assertion {}; - - template<> - struct static_assertion - { - enum { - YOU_TRIED_CALLING_A_VECTOR_METHOD_ON_A_MATRIX=1, - YOU_MIXED_VECTORS_OF_DIFFERENT_SIZES=1, - YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES=1, - THIS_METHOD_IS_ONLY_FOR_VECTORS_OF_A_SPECIFIC_SIZE=1, - THIS_METHOD_IS_ONLY_FOR_MATRICES_OF_A_SPECIFIC_SIZE=1, - THIS_METHOD_IS_ONLY_FOR_OBJECTS_OF_A_SPECIFIC_SIZE=1, - OUT_OF_RANGE_ACCESS=1, - YOU_MADE_A_PROGRAMMING_MISTAKE=1, - EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT=1, - EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE=1, - YOU_CALLED_A_FIXED_SIZE_METHOD_ON_A_DYNAMIC_SIZE_MATRIX_OR_VECTOR=1, - YOU_CALLED_A_DYNAMIC_SIZE_METHOD_ON_A_FIXED_SIZE_MATRIX_OR_VECTOR=1, - UNALIGNED_LOAD_AND_STORE_OPERATIONS_UNIMPLEMENTED_ON_ALTIVEC=1, - THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES=1, - FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED=1, - NUMERIC_TYPE_MUST_BE_REAL=1, - COEFFICIENT_WRITE_ACCESS_TO_SELFADJOINT_NOT_SUPPORTED=1, - WRITING_TO_TRIANGULAR_PART_WITH_UNIT_DIAGONAL_IS_NOT_SUPPORTED=1, - THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE=1, - INVALID_MATRIX_PRODUCT=1, - INVALID_VECTOR_VECTOR_PRODUCT__IF_YOU_WANTED_A_DOT_OR_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTIONS=1, - INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION=1, - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY=1, - THIS_METHOD_IS_ONLY_FOR_COLUMN_MAJOR_MATRICES=1, - THIS_METHOD_IS_ONLY_FOR_ROW_MAJOR_MATRICES=1, - INVALID_MATRIX_TEMPLATE_PARAMETERS=1, - INVALID_MATRIXBASE_TEMPLATE_PARAMETERS=1, - BOTH_MATRICES_MUST_HAVE_THE_SAME_STORAGE_ORDER=1, - THIS_METHOD_IS_ONLY_FOR_DIAGONAL_MATRIX=1, - THE_MATRIX_OR_EXPRESSION_THAT_YOU_PASSED_DOES_NOT_HAVE_THE_EXPECTED_TYPE=1, - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_WITH_DIRECT_MEMORY_ACCESS_SUCH_AS_MAP_OR_PLAIN_MATRICES=1, - YOU_ALREADY_SPECIFIED_THIS_STRIDE=1, - INVALID_STORAGE_ORDER_FOR_THIS_VECTOR_EXPRESSION=1, - THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD=1, - PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1=1, - THIS_METHOD_IS_ONLY_FOR_SPECIFIC_TRANSFORMATIONS=1, - YOU_CANNOT_MIX_ARRAYS_AND_MATRICES=1, - YOU_PERFORMED_AN_INVALID_TRANSFORMATION_CONVERSION=1, - THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY=1, - YOU_ARE_TRYING_TO_USE_AN_INDEX_BASED_ACCESSOR_ON_AN_EXPRESSION_THAT_DOES_NOT_SUPPORT_THAT=1, - THIS_METHOD_IS_ONLY_FOR_1x1_EXPRESSIONS=1, - THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS=1, - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL=1, - THIS_METHOD_IS_ONLY_FOR_ARRAYS_NOT_MATRICES=1, - YOU_PASSED_A_ROW_VECTOR_BUT_A_COLUMN_VECTOR_WAS_EXPECTED=1, - YOU_PASSED_A_COLUMN_VECTOR_BUT_A_ROW_VECTOR_WAS_EXPECTED=1, - THE_INDEX_TYPE_MUST_BE_A_SIGNED_TYPE=1, - THE_STORAGE_ORDER_OF_BOTH_SIDES_MUST_MATCH=1, - OBJECT_ALLOCATED_ON_STACK_IS_TOO_BIG=1, - IMPLICIT_CONVERSION_TO_SCALAR_IS_FOR_INNER_PRODUCT_ONLY=1, - STORAGE_LAYOUT_DOES_NOT_MATCH=1, - EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT__INVALID_COST_VALUE=1, - THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS=1, - MATRIX_FREE_CONJUGATE_GRADIENT_IS_COMPATIBLE_WITH_UPPER_UNION_LOWER_MODE_ONLY=1, - THIS_TYPE_IS_NOT_SUPPORTED=1, - STORAGE_KIND_MUST_MATCH=1, - STORAGE_INDEX_MUST_MATCH=1, - CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY=1, - SELFADJOINTVIEW_ACCEPTS_UPPER_AND_LOWER_MODE_ONLY=1 - }; - }; - - } // end namespace internal - - } // end namespace Eigen - - // Specialized implementation for MSVC to avoid "conditional - // expression is constant" warnings. This implementation doesn't - // appear to work under GCC, hence the multiple implementations. - #if EIGEN_COMP_MSVC - - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) \ - {Eigen::internal::static_assertion::MSG;} - - #else - // In some cases clang interprets bool(CONDITION) as function declaration - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) \ - if (Eigen::internal::static_assertion(CONDITION)>::MSG) {} - - #endif - - #endif // not CXX0X - -#else // EIGEN_NO_STATIC_ASSERT - - #define EIGEN_STATIC_ASSERT(CONDITION,MSG) eigen_assert((CONDITION) && #MSG); - -#endif // EIGEN_NO_STATIC_ASSERT -#endif // EIGEN_STATIC_ASSERT - -// static assertion failing if the type \a TYPE is not a vector type -#define EIGEN_STATIC_ASSERT_VECTOR_ONLY(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::IsVectorAtCompileTime, \ - YOU_TRIED_CALLING_A_VECTOR_METHOD_ON_A_MATRIX) - -// static assertion failing if the type \a TYPE is not fixed-size -#define EIGEN_STATIC_ASSERT_FIXED_SIZE(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::SizeAtCompileTime!=Eigen::Dynamic, \ - YOU_CALLED_A_FIXED_SIZE_METHOD_ON_A_DYNAMIC_SIZE_MATRIX_OR_VECTOR) - -// static assertion failing if the type \a TYPE is not dynamic-size -#define EIGEN_STATIC_ASSERT_DYNAMIC_SIZE(TYPE) \ - EIGEN_STATIC_ASSERT(TYPE::SizeAtCompileTime==Eigen::Dynamic, \ - YOU_CALLED_A_DYNAMIC_SIZE_METHOD_ON_A_FIXED_SIZE_MATRIX_OR_VECTOR) - -// static assertion failing if the type \a TYPE is not a vector type of the given size -#define EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(TYPE, SIZE) \ - EIGEN_STATIC_ASSERT(TYPE::IsVectorAtCompileTime && TYPE::SizeAtCompileTime==SIZE, \ - THIS_METHOD_IS_ONLY_FOR_VECTORS_OF_A_SPECIFIC_SIZE) - -// static assertion failing if the type \a TYPE is not a vector type of the given size -#define EIGEN_STATIC_ASSERT_MATRIX_SPECIFIC_SIZE(TYPE, ROWS, COLS) \ - EIGEN_STATIC_ASSERT(TYPE::RowsAtCompileTime==ROWS && TYPE::ColsAtCompileTime==COLS, \ - THIS_METHOD_IS_ONLY_FOR_MATRICES_OF_A_SPECIFIC_SIZE) - -// static assertion failing if the two vector expression types are not compatible (same fixed-size or dynamic size) -#define EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(TYPE0,TYPE1) \ - EIGEN_STATIC_ASSERT( \ - (int(TYPE0::SizeAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::SizeAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::SizeAtCompileTime)==int(TYPE1::SizeAtCompileTime)),\ - YOU_MIXED_VECTORS_OF_DIFFERENT_SIZES) - -#define EIGEN_PREDICATE_SAME_MATRIX_SIZE(TYPE0,TYPE1) \ - ( \ - (int(Eigen::internal::size_of_xpr_at_compile_time::ret)==0 && int(Eigen::internal::size_of_xpr_at_compile_time::ret)==0) \ - || (\ - (int(TYPE0::RowsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::RowsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::RowsAtCompileTime)==int(TYPE1::RowsAtCompileTime)) \ - && (int(TYPE0::ColsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE1::ColsAtCompileTime)==Eigen::Dynamic \ - || int(TYPE0::ColsAtCompileTime)==int(TYPE1::ColsAtCompileTime))\ - ) \ - ) - -#define EIGEN_STATIC_ASSERT_NON_INTEGER(TYPE) \ - EIGEN_STATIC_ASSERT(!NumTraits::IsInteger, THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES) - - -// static assertion failing if it is guaranteed at compile-time that the two matrix expression types have different sizes -#define EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(TYPE0,TYPE1) \ - EIGEN_STATIC_ASSERT( \ - EIGEN_PREDICATE_SAME_MATRIX_SIZE(TYPE0,TYPE1),\ - YOU_MIXED_MATRICES_OF_DIFFERENT_SIZES) - -#define EIGEN_STATIC_ASSERT_SIZE_1x1(TYPE) \ - EIGEN_STATIC_ASSERT((TYPE::RowsAtCompileTime == 1 || TYPE::RowsAtCompileTime == Dynamic) && \ - (TYPE::ColsAtCompileTime == 1 || TYPE::ColsAtCompileTime == Dynamic), \ - THIS_METHOD_IS_ONLY_FOR_1x1_EXPRESSIONS) - -#define EIGEN_STATIC_ASSERT_LVALUE(Derived) \ - EIGEN_STATIC_ASSERT(Eigen::internal::is_lvalue::value, \ - THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY) - -#define EIGEN_STATIC_ASSERT_ARRAYXPR(Derived) \ - EIGEN_STATIC_ASSERT((Eigen::internal::is_same::XprKind, ArrayXpr>::value), \ - THIS_METHOD_IS_ONLY_FOR_ARRAYS_NOT_MATRICES) - -#define EIGEN_STATIC_ASSERT_SAME_XPR_KIND(Derived1, Derived2) \ - EIGEN_STATIC_ASSERT((Eigen::internal::is_same::XprKind, \ - typename Eigen::internal::traits::XprKind \ - >::value), \ - YOU_CANNOT_MIX_ARRAYS_AND_MATRICES) - -// Check that a cost value is positive, and that is stay within a reasonable range -// TODO this check could be enabled for internal debugging only -#define EIGEN_INTERNAL_CHECK_COST_VALUE(C) \ - EIGEN_STATIC_ASSERT((C)>=0 && (C)<=HugeCost*HugeCost, EIGEN_INTERNAL_ERROR_PLEASE_FILE_A_BUG_REPORT__INVALID_COST_VALUE); - -#endif // EIGEN_STATIC_ASSERT_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h deleted file mode 100644 index 6bb4970828e..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Core/util/XprHelper.h +++ /dev/null @@ -1,838 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_XPRHELPER_H -#define EIGEN_XPRHELPER_H - -// just a workaround because GCC seems to not really like empty structs -// FIXME: gcc 4.3 generates bad code when strict-aliasing is enabled -// so currently we simply disable this optimization for gcc 4.3 -#if EIGEN_COMP_GNUC && !EIGEN_GNUC_AT(4,3) - #define EIGEN_EMPTY_STRUCT_CTOR(X) \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE X() {} \ - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE X(const X& ) {} -#else - #define EIGEN_EMPTY_STRUCT_CTOR(X) -#endif - -namespace Eigen { - -namespace internal { - -template -EIGEN_DEVICE_FUNC -inline IndexDest convert_index(const IndexSrc& idx) { - // for sizeof(IndexDest)>=sizeof(IndexSrc) compilers should be able to optimize this away: - eigen_internal_assert(idx <= NumTraits::highest() && "Index value to big for target type"); - return IndexDest(idx); -} - -// true if T can be considered as an integral index (i.e., and integral type or enum) -template struct is_valid_index_type -{ - enum { value = -#if EIGEN_HAS_TYPE_TRAITS - internal::is_integral::value || std::is_enum::value -#elif EIGEN_COMP_MSVC - internal::is_integral::value || __is_enum(T) -#else - // without C++11, we use is_convertible to Index instead of is_integral in order to treat enums as Index. - internal::is_convertible::value && !internal::is_same::value && !is_same::value -#endif - }; -}; - -// promote_scalar_arg is an helper used in operation between an expression and a scalar, like: -// expression * scalar -// Its role is to determine how the type T of the scalar operand should be promoted given the scalar type ExprScalar of the given expression. -// The IsSupported template parameter must be provided by the caller as: internal::has_ReturnType >::value using the proper order for ExprScalar and T. -// Then the logic is as follows: -// - if the operation is natively supported as defined by IsSupported, then the scalar type is not promoted, and T is returned. -// - otherwise, NumTraits::Literal is returned if T is implicitly convertible to NumTraits::Literal AND that this does not imply a float to integer conversion. -// - otherwise, ExprScalar is returned if T is implicitly convertible to ExprScalar AND that this does not imply a float to integer conversion. -// - In all other cases, the promoted type is not defined, and the respective operation is thus invalid and not available (SFINAE). -template -struct promote_scalar_arg; - -template -struct promote_scalar_arg -{ - typedef T type; -}; - -// Recursively check safe conversion to PromotedType, and then ExprScalar if they are different. -template::value, - bool IsSafe = NumTraits::IsInteger || !NumTraits::IsInteger> -struct promote_scalar_arg_unsupported; - -// Start recursion with NumTraits::Literal -template -struct promote_scalar_arg : promote_scalar_arg_unsupported::Literal> {}; - -// We found a match! -template -struct promote_scalar_arg_unsupported -{ - typedef PromotedType type; -}; - -// No match, but no real-to-integer issues, and ExprScalar and current PromotedType are different, -// so let's try to promote to ExprScalar -template -struct promote_scalar_arg_unsupported - : promote_scalar_arg_unsupported -{}; - -// Unsafe real-to-integer, let's stop. -template -struct promote_scalar_arg_unsupported {}; - -// T is not even convertible to ExprScalar, let's stop. -template -struct promote_scalar_arg_unsupported {}; - -//classes inheriting no_assignment_operator don't generate a default operator=. -class no_assignment_operator -{ - private: - no_assignment_operator& operator=(const no_assignment_operator&); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(no_assignment_operator) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(no_assignment_operator) -}; - -/** \internal return the index type with the largest number of bits */ -template -struct promote_index_type -{ - typedef typename conditional<(sizeof(I1)::type type; -}; - -/** \internal If the template parameter Value is Dynamic, this class is just a wrapper around a T variable that - * can be accessed using value() and setValue(). - * Otherwise, this class is an empty structure and value() just returns the template parameter Value. - */ -template class variable_if_dynamic -{ - public: - EIGEN_EMPTY_STRUCT_CTOR(variable_if_dynamic) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamic(T v) { EIGEN_ONLY_USED_FOR_DEBUG(v); eigen_assert(v == T(Value)); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE T value() { return T(Value); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T) {} -}; - -template class variable_if_dynamic -{ - T m_value; - EIGEN_DEVICE_FUNC variable_if_dynamic() { eigen_assert(false); } - public: - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamic(T value) : m_value(value) {} - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE T value() const { return m_value; } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T value) { m_value = value; } -}; - -/** \internal like variable_if_dynamic but for DynamicIndex - */ -template class variable_if_dynamicindex -{ - public: - EIGEN_EMPTY_STRUCT_CTOR(variable_if_dynamicindex) - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamicindex(T v) { EIGEN_ONLY_USED_FOR_DEBUG(v); eigen_assert(v == T(Value)); } - EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE T value() { return T(Value); } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T) {} -}; - -template class variable_if_dynamicindex -{ - T m_value; - EIGEN_DEVICE_FUNC variable_if_dynamicindex() { eigen_assert(false); } - public: - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit variable_if_dynamicindex(T value) : m_value(value) {} - EIGEN_DEVICE_FUNC T EIGEN_STRONG_INLINE value() const { return m_value; } - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void setValue(T value) { m_value = value; } -}; - -template struct functor_traits -{ - enum - { - Cost = 10, - PacketAccess = false, - IsRepeatable = false - }; -}; - -template struct packet_traits; - -template struct unpacket_traits -{ - typedef T type; - typedef T half; - enum - { - size = 1, - alignment = 1 - }; -}; - -template::size)==0 || is_same::half>::value> -struct find_best_packet_helper; - -template< int Size, typename PacketType> -struct find_best_packet_helper -{ - typedef PacketType type; -}; - -template -struct find_best_packet_helper -{ - typedef typename find_best_packet_helper::half>::type type; -}; - -template -struct find_best_packet -{ - typedef typename find_best_packet_helper::type>::type type; -}; - -#if EIGEN_MAX_STATIC_ALIGN_BYTES>0 -template -struct compute_default_alignment_helper -{ - enum { value = 0 }; -}; - -template -struct compute_default_alignment_helper // Match -{ - enum { value = AlignmentBytes }; -}; - -template -struct compute_default_alignment_helper // Try-half -{ - // current packet too large, try with an half-packet - enum { value = compute_default_alignment_helper::value }; -}; -#else -// If static alignment is disabled, no need to bother. -// This also avoids a division by zero in "bool Match = bool((ArrayBytes%AlignmentBytes)==0)" -template -struct compute_default_alignment_helper -{ - enum { value = 0 }; -}; -#endif - -template struct compute_default_alignment { - enum { value = compute_default_alignment_helper::value }; -}; - -template struct compute_default_alignment { - enum { value = EIGEN_MAX_ALIGN_BYTES }; -}; - -template class make_proper_matrix_type -{ - enum { - IsColVector = _Cols==1 && _Rows!=1, - IsRowVector = _Rows==1 && _Cols!=1, - Options = IsColVector ? (_Options | ColMajor) & ~RowMajor - : IsRowVector ? (_Options | RowMajor) & ~ColMajor - : _Options - }; - public: - typedef Matrix<_Scalar, _Rows, _Cols, Options, _MaxRows, _MaxCols> type; -}; - -template -class compute_matrix_flags -{ - enum { row_major_bit = Options&RowMajor ? RowMajorBit : 0 }; - public: - // FIXME currently we still have to handle DirectAccessBit at the expression level to handle DenseCoeffsBase<> - // and then propagate this information to the evaluator's flags. - // However, I (Gael) think that DirectAccessBit should only matter at the evaluation stage. - enum { ret = DirectAccessBit | LvalueBit | NestByRefBit | row_major_bit }; -}; - -template struct size_at_compile_time -{ - enum { ret = (_Rows==Dynamic || _Cols==Dynamic) ? Dynamic : _Rows * _Cols }; -}; - -template struct size_of_xpr_at_compile_time -{ - enum { ret = size_at_compile_time::RowsAtCompileTime,traits::ColsAtCompileTime>::ret }; -}; - -/* plain_matrix_type : the difference from eval is that plain_matrix_type is always a plain matrix type, - * whereas eval is a const reference in the case of a matrix - */ - -template::StorageKind> struct plain_matrix_type; -template struct plain_matrix_type_dense; -template struct plain_matrix_type -{ - typedef typename plain_matrix_type_dense::XprKind, traits::Flags>::type type; -}; -template struct plain_matrix_type -{ - typedef typename T::PlainObject type; -}; - -template struct plain_matrix_type_dense -{ - typedef Matrix::Scalar, - traits::RowsAtCompileTime, - traits::ColsAtCompileTime, - AutoAlign | (Flags&RowMajorBit ? RowMajor : ColMajor), - traits::MaxRowsAtCompileTime, - traits::MaxColsAtCompileTime - > type; -}; - -template struct plain_matrix_type_dense -{ - typedef Array::Scalar, - traits::RowsAtCompileTime, - traits::ColsAtCompileTime, - AutoAlign | (Flags&RowMajorBit ? RowMajor : ColMajor), - traits::MaxRowsAtCompileTime, - traits::MaxColsAtCompileTime - > type; -}; - -/* eval : the return type of eval(). For matrices, this is just a const reference - * in order to avoid a useless copy - */ - -template::StorageKind> struct eval; - -template struct eval -{ - typedef typename plain_matrix_type::type type; -// typedef typename T::PlainObject type; -// typedef T::Matrix::Scalar, -// traits::RowsAtCompileTime, -// traits::ColsAtCompileTime, -// AutoAlign | (traits::Flags&RowMajorBit ? RowMajor : ColMajor), -// traits::MaxRowsAtCompileTime, -// traits::MaxColsAtCompileTime -// > type; -}; - -template struct eval -{ - typedef typename plain_matrix_type::type type; -}; - -// for matrices, no need to evaluate, just use a const reference to avoid a useless copy -template -struct eval, Dense> -{ - typedef const Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& type; -}; - -template -struct eval, Dense> -{ - typedef const Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& type; -}; - - -/* similar to plain_matrix_type, but using the evaluator's Flags */ -template::StorageKind> struct plain_object_eval; - -template -struct plain_object_eval -{ - typedef typename plain_matrix_type_dense::XprKind, evaluator::Flags>::type type; -}; - - -/* plain_matrix_type_column_major : same as plain_matrix_type but guaranteed to be column-major - */ -template struct plain_matrix_type_column_major -{ - enum { Rows = traits::RowsAtCompileTime, - Cols = traits::ColsAtCompileTime, - MaxRows = traits::MaxRowsAtCompileTime, - MaxCols = traits::MaxColsAtCompileTime - }; - typedef Matrix::Scalar, - Rows, - Cols, - (MaxRows==1&&MaxCols!=1) ? RowMajor : ColMajor, - MaxRows, - MaxCols - > type; -}; - -/* plain_matrix_type_row_major : same as plain_matrix_type but guaranteed to be row-major - */ -template struct plain_matrix_type_row_major -{ - enum { Rows = traits::RowsAtCompileTime, - Cols = traits::ColsAtCompileTime, - MaxRows = traits::MaxRowsAtCompileTime, - MaxCols = traits::MaxColsAtCompileTime - }; - typedef Matrix::Scalar, - Rows, - Cols, - (MaxCols==1&&MaxRows!=1) ? RowMajor : ColMajor, - MaxRows, - MaxCols - > type; -}; - -/** \internal The reference selector for template expressions. The idea is that we don't - * need to use references for expressions since they are light weight proxy - * objects which should generate no copying overhead. */ -template -struct ref_selector -{ - typedef typename conditional< - bool(traits::Flags & NestByRefBit), - T const&, - const T - >::type type; - - typedef typename conditional< - bool(traits::Flags & NestByRefBit), - T &, - T - >::type non_const_type; -}; - -/** \internal Adds the const qualifier on the value-type of T2 if and only if T1 is a const type */ -template -struct transfer_constness -{ - typedef typename conditional< - bool(internal::is_const::value), - typename internal::add_const_on_value_type::type, - T2 - >::type type; -}; - - -// However, we still need a mechanism to detect whether an expression which is evaluated multiple time -// has to be evaluated into a temporary. -// That's the purpose of this new nested_eval helper: -/** \internal Determines how a given expression should be nested when evaluated multiple times. - * For example, when you do a * (b+c), Eigen will determine how the expression b+c should be - * evaluated into the bigger product expression. The choice is between nesting the expression b+c as-is, or - * evaluating that expression b+c into a temporary variable d, and nest d so that the resulting expression is - * a*d. Evaluating can be beneficial for example if every coefficient access in the resulting expression causes - * many coefficient accesses in the nested expressions -- as is the case with matrix product for example. - * - * \tparam T the type of the expression being nested. - * \tparam n the number of coefficient accesses in the nested expression for each coefficient access in the bigger expression. - * \tparam PlainObject the type of the temporary if needed. - */ -template::type> struct nested_eval -{ - enum { - ScalarReadCost = NumTraits::Scalar>::ReadCost, - CoeffReadCost = evaluator::CoeffReadCost, // NOTE What if an evaluator evaluate itself into a tempory? - // Then CoeffReadCost will be small (e.g., 1) but we still have to evaluate, especially if n>1. - // This situation is already taken care by the EvalBeforeNestingBit flag, which is turned ON - // for all evaluator creating a temporary. This flag is then propagated by the parent evaluators. - // Another solution could be to count the number of temps? - NAsInteger = n == Dynamic ? HugeCost : n, - CostEval = (NAsInteger+1) * ScalarReadCost + CoeffReadCost, - CostNoEval = NAsInteger * CoeffReadCost, - Evaluate = (int(evaluator::Flags) & EvalBeforeNestingBit) || (int(CostEval) < int(CostNoEval)) - }; - - typedef typename conditional::type>::type type; -}; - -template -EIGEN_DEVICE_FUNC -inline T* const_cast_ptr(const T* ptr) -{ - return const_cast(ptr); -} - -template::XprKind> -struct dense_xpr_base -{ - /* dense_xpr_base should only ever be used on dense expressions, thus falling either into the MatrixXpr or into the ArrayXpr cases */ -}; - -template -struct dense_xpr_base -{ - typedef MatrixBase type; -}; - -template -struct dense_xpr_base -{ - typedef ArrayBase type; -}; - -template::XprKind, typename StorageKind = typename traits::StorageKind> -struct generic_xpr_base; - -template -struct generic_xpr_base -{ - typedef typename dense_xpr_base::type type; -}; - -template struct cast_return_type -{ - typedef typename XprType::Scalar CurrentScalarType; - typedef typename remove_all::type _CastType; - typedef typename _CastType::Scalar NewScalarType; - typedef typename conditional::value, - const XprType&,CastType>::type type; -}; - -template struct promote_storage_type; - -template struct promote_storage_type -{ - typedef A ret; -}; -template struct promote_storage_type -{ - typedef A ret; -}; -template struct promote_storage_type -{ - typedef A ret; -}; - -/** \internal Specify the "storage kind" of applying a coefficient-wise - * binary operations between two expressions of kinds A and B respectively. - * The template parameter Functor permits to specialize the resulting storage kind wrt to - * the functor. - * The default rules are as follows: - * \code - * A op A -> A - * A op dense -> dense - * dense op B -> dense - * sparse op dense -> sparse - * dense op sparse -> sparse - * \endcode - */ -template struct cwise_promote_storage_type; - -template struct cwise_promote_storage_type { typedef A ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Dense ret; }; -template struct cwise_promote_storage_type { typedef Sparse ret; }; -template struct cwise_promote_storage_type { typedef Sparse ret; }; - -template struct cwise_promote_storage_order { - enum { value = LhsOrder }; -}; - -template struct cwise_promote_storage_order { enum { value = RhsOrder }; }; -template struct cwise_promote_storage_order { enum { value = LhsOrder }; }; -template struct cwise_promote_storage_order { enum { value = Order }; }; - - -/** \internal Specify the "storage kind" of multiplying an expression of kind A with kind B. - * The template parameter ProductTag permits to specialize the resulting storage kind wrt to - * some compile-time properties of the product: GemmProduct, GemvProduct, OuterProduct, InnerProduct. - * The default rules are as follows: - * \code - * K * K -> K - * dense * K -> dense - * K * dense -> dense - * diag * K -> K - * K * diag -> K - * Perm * K -> K - * K * Perm -> K - * \endcode - */ -template struct product_promote_storage_type; - -template struct product_promote_storage_type { typedef A ret;}; -template struct product_promote_storage_type { typedef Dense ret;}; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -template struct product_promote_storage_type { typedef A ret; }; -template struct product_promote_storage_type { typedef B ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -template struct product_promote_storage_type { typedef A ret; }; -template struct product_promote_storage_type { typedef B ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; -template struct product_promote_storage_type { typedef Dense ret; }; - -/** \internal gives the plain matrix or array type to store a row/column/diagonal of a matrix type. - * \tparam Scalar optional parameter allowing to pass a different scalar type than the one of the MatrixType. - */ -template -struct plain_row_type -{ - typedef Matrix MatrixRowType; - typedef Array ArrayRowType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixRowType, - ArrayRowType - >::type type; -}; - -template -struct plain_col_type -{ - typedef Matrix MatrixColType; - typedef Array ArrayColType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixColType, - ArrayColType - >::type type; -}; - -template -struct plain_diag_type -{ - enum { diag_size = EIGEN_SIZE_MIN_PREFER_DYNAMIC(ExpressionType::RowsAtCompileTime, ExpressionType::ColsAtCompileTime), - max_diag_size = EIGEN_SIZE_MIN_PREFER_FIXED(ExpressionType::MaxRowsAtCompileTime, ExpressionType::MaxColsAtCompileTime) - }; - typedef Matrix MatrixDiagType; - typedef Array ArrayDiagType; - - typedef typename conditional< - is_same< typename traits::XprKind, MatrixXpr >::value, - MatrixDiagType, - ArrayDiagType - >::type type; -}; - -template -struct plain_constant_type -{ - enum { Options = (traits::Flags&RowMajorBit)?RowMajor:0 }; - - typedef Array::RowsAtCompileTime, traits::ColsAtCompileTime, - Options, traits::MaxRowsAtCompileTime,traits::MaxColsAtCompileTime> array_type; - - typedef Matrix::RowsAtCompileTime, traits::ColsAtCompileTime, - Options, traits::MaxRowsAtCompileTime,traits::MaxColsAtCompileTime> matrix_type; - - typedef CwiseNullaryOp, const typename conditional::XprKind, MatrixXpr >::value, matrix_type, array_type>::type > type; -}; - -template -struct is_lvalue -{ - enum { value = (!bool(is_const::value)) && - bool(traits::Flags & LvalueBit) }; -}; - -template struct is_diagonal -{ enum { ret = false }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct is_diagonal > -{ enum { ret = true }; }; - -template struct glue_shapes; -template<> struct glue_shapes { typedef TriangularShape type; }; - -template -bool is_same_dense(const T1 &mat1, const T2 &mat2, typename enable_if::ret&&has_direct_access::ret, T1>::type * = 0) -{ - return (mat1.data()==mat2.data()) && (mat1.innerStride()==mat2.innerStride()) && (mat1.outerStride()==mat2.outerStride()); -} - -template -bool is_same_dense(const T1 &, const T2 &, typename enable_if::ret&&has_direct_access::ret), T1>::type * = 0) -{ - return false; -} - -// Internal helper defining the cost of a scalar division for the type T. -// The default heuristic can be specialized for each scalar type and architecture. -template -struct scalar_div_cost { - enum { value = 8*NumTraits::MulCost }; -}; - -template -struct scalar_div_cost, Vectorized> { - enum { value = 2*scalar_div_cost::value - + 6*NumTraits::MulCost - + 3*NumTraits::AddCost - }; -}; - - -template -struct scalar_div_cost::type> { enum { value = 24 }; }; -template -struct scalar_div_cost::type> { enum { value = 21 }; }; - - -#ifdef EIGEN_DEBUG_ASSIGN -std::string demangle_traversal(int t) -{ - if(t==DefaultTraversal) return "DefaultTraversal"; - if(t==LinearTraversal) return "LinearTraversal"; - if(t==InnerVectorizedTraversal) return "InnerVectorizedTraversal"; - if(t==LinearVectorizedTraversal) return "LinearVectorizedTraversal"; - if(t==SliceVectorizedTraversal) return "SliceVectorizedTraversal"; - return "?"; -} -std::string demangle_unrolling(int t) -{ - if(t==NoUnrolling) return "NoUnrolling"; - if(t==InnerUnrolling) return "InnerUnrolling"; - if(t==CompleteUnrolling) return "CompleteUnrolling"; - return "?"; -} -std::string demangle_flags(int f) -{ - std::string res; - if(f&RowMajorBit) res += " | RowMajor"; - if(f&PacketAccessBit) res += " | Packet"; - if(f&LinearAccessBit) res += " | Linear"; - if(f&LvalueBit) res += " | Lvalue"; - if(f&DirectAccessBit) res += " | Direct"; - if(f&NestByRefBit) res += " | NestByRef"; - if(f&NoPreferredStorageOrderBit) res += " | NoPreferredStorageOrderBit"; - - return res; -} -#endif - -} // end namespace internal - - -/** \class ScalarBinaryOpTraits - * \ingroup Core_Module - * - * \brief Determines whether the given binary operation of two numeric types is allowed and what the scalar return type is. - * - * This class permits to control the scalar return type of any binary operation performed on two different scalar types through (partial) template specializations. - * - * For instance, let \c U1, \c U2 and \c U3 be three user defined scalar types for which most operations between instances of \c U1 and \c U2 returns an \c U3. - * You can let %Eigen knows that by defining: - \code - template - struct ScalarBinaryOpTraits { typedef U3 ReturnType; }; - template - struct ScalarBinaryOpTraits { typedef U3 ReturnType; }; - \endcode - * You can then explicitly disable some particular operations to get more explicit error messages: - \code - template<> - struct ScalarBinaryOpTraits > {}; - \endcode - * Or customize the return type for individual operation: - \code - template<> - struct ScalarBinaryOpTraits > { typedef U1 ReturnType; }; - \endcode - * - * By default, the following generic combinations are supported: - - - - - -
ScalarAScalarBBinaryOpReturnTypeNote
\c T \c T \c * \c T
\c NumTraits::Real \c T \c * \c T Only if \c NumTraits::IsComplex
\c T \c NumTraits::Real \c * \c T Only if \c NumTraits::IsComplex
- * - * \sa CwiseBinaryOp - */ -template > -struct ScalarBinaryOpTraits -#ifndef EIGEN_PARSED_BY_DOXYGEN - // for backward compatibility, use the hints given by the (deprecated) internal::scalar_product_traits class. - : internal::scalar_product_traits -#endif // EIGEN_PARSED_BY_DOXYGEN -{}; - -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -template -struct ScalarBinaryOpTraits::IsComplex,T>::type>::Real, BinaryOp> -{ - typedef T ReturnType; -}; -template -struct ScalarBinaryOpTraits::IsComplex,T>::type>::Real, T, BinaryOp> -{ - typedef T ReturnType; -}; - -// For Matrix * Permutation -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -// For Permutation * Matrix -template -struct ScalarBinaryOpTraits -{ - typedef T ReturnType; -}; - -// for Permutation*Permutation -template -struct ScalarBinaryOpTraits -{ - typedef void ReturnType; -}; - -// We require Lhs and Rhs to have "compatible" scalar types. -// It is tempting to always allow mixing different types but remember that this is often impossible in the vectorized paths. -// So allowing mixing different types gives very unexpected errors when enabling vectorization, when the user tries to -// add together a float matrix and a double matrix. -#define EIGEN_CHECK_BINARY_COMPATIBILIY(BINOP,LHS,RHS) \ - EIGEN_STATIC_ASSERT((Eigen::internal::has_ReturnType >::value), \ - YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY) - -} // end namespace Eigen - -#endif // EIGEN_XPRHELPER_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h deleted file mode 100644 index 01a7ed1884d..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/BlockHouseholder.h +++ /dev/null @@ -1,103 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Vincent Lejeune -// Copyright (C) 2010 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_BLOCK_HOUSEHOLDER_H -#define EIGEN_BLOCK_HOUSEHOLDER_H - -// This file contains some helper function to deal with block householder reflectors - -namespace Eigen { - -namespace internal { - -/** \internal */ -// template -// void make_block_householder_triangular_factor(TriangularFactorType& triFactor, const VectorsType& vectors, const CoeffsType& hCoeffs) -// { -// typedef typename VectorsType::Scalar Scalar; -// const Index nbVecs = vectors.cols(); -// eigen_assert(triFactor.rows() == nbVecs && triFactor.cols() == nbVecs && vectors.rows()>=nbVecs); -// -// for(Index i = 0; i < nbVecs; i++) -// { -// Index rs = vectors.rows() - i; -// // Warning, note that hCoeffs may alias with vectors. -// // It is then necessary to copy it before modifying vectors(i,i). -// typename CoeffsType::Scalar h = hCoeffs(i); -// // This hack permits to pass trough nested Block<> and Transpose<> expressions. -// Scalar *Vii_ptr = const_cast(vectors.data() + vectors.outerStride()*i + vectors.innerStride()*i); -// Scalar Vii = *Vii_ptr; -// *Vii_ptr = Scalar(1); -// triFactor.col(i).head(i).noalias() = -h * vectors.block(i, 0, rs, i).adjoint() -// * vectors.col(i).tail(rs); -// *Vii_ptr = Vii; -// // FIXME add .noalias() once the triangular product can work inplace -// triFactor.col(i).head(i) = triFactor.block(0,0,i,i).template triangularView() -// * triFactor.col(i).head(i); -// triFactor(i,i) = hCoeffs(i); -// } -// } - -/** \internal */ -// This variant avoid modifications in vectors -template -void make_block_householder_triangular_factor(TriangularFactorType& triFactor, const VectorsType& vectors, const CoeffsType& hCoeffs) -{ - const Index nbVecs = vectors.cols(); - eigen_assert(triFactor.rows() == nbVecs && triFactor.cols() == nbVecs && vectors.rows()>=nbVecs); - - for(Index i = nbVecs-1; i >=0 ; --i) - { - Index rs = vectors.rows() - i - 1; - Index rt = nbVecs-i-1; - - if(rt>0) - { - triFactor.row(i).tail(rt).noalias() = -hCoeffs(i) * vectors.col(i).tail(rs).adjoint() - * vectors.bottomRightCorner(rs, rt).template triangularView(); - - // FIXME add .noalias() once the triangular product can work inplace - triFactor.row(i).tail(rt) = triFactor.row(i).tail(rt) * triFactor.bottomRightCorner(rt,rt).template triangularView(); - - } - triFactor(i,i) = hCoeffs(i); - } -} - -/** \internal - * if forward then perform mat = H0 * H1 * H2 * mat - * otherwise perform mat = H2 * H1 * H0 * mat - */ -template -void apply_block_householder_on_the_left(MatrixType& mat, const VectorsType& vectors, const CoeffsType& hCoeffs, bool forward) -{ - enum { TFactorSize = MatrixType::ColsAtCompileTime }; - Index nbVecs = vectors.cols(); - Matrix T(nbVecs,nbVecs); - - if(forward) make_block_householder_triangular_factor(T, vectors, hCoeffs); - else make_block_householder_triangular_factor(T, vectors, hCoeffs.conjugate()); - const TriangularView V(vectors); - - // A -= V T V^* A - Matrix tmp = V.adjoint() * mat; - // FIXME add .noalias() once the triangular product can work inplace - if(forward) tmp = T.template triangularView() * tmp; - else tmp = T.template triangularView().adjoint() * tmp; - mat.noalias() -= V * tmp; -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_BLOCK_HOUSEHOLDER_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h deleted file mode 100644 index 80de2c3052c..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/Householder.h +++ /dev/null @@ -1,172 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2010 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HOUSEHOLDER_H -#define EIGEN_HOUSEHOLDER_H - -namespace Eigen { - -namespace internal { -template struct decrement_size -{ - enum { - ret = n==Dynamic ? n : n-1 - }; -}; -} - -/** Computes the elementary reflector H such that: - * \f$ H *this = [ beta 0 ... 0]^T \f$ - * where the transformation H is: - * \f$ H = I - tau v v^*\f$ - * and the vector v is: - * \f$ v^T = [1 essential^T] \f$ - * - * The essential part of the vector \c v is stored in *this. - * - * On output: - * \param tau the scaling factor of the Householder transformation - * \param beta the result of H * \c *this - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::applyHouseholderOnTheLeft(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -void MatrixBase::makeHouseholderInPlace(Scalar& tau, RealScalar& beta) -{ - VectorBlock::ret> essentialPart(derived(), 1, size()-1); - makeHouseholder(essentialPart, tau, beta); -} - -/** Computes the elementary reflector H such that: - * \f$ H *this = [ beta 0 ... 0]^T \f$ - * where the transformation H is: - * \f$ H = I - tau v v^*\f$ - * and the vector v is: - * \f$ v^T = [1 essential^T] \f$ - * - * On output: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param beta the result of H * \c *this - * - * \sa MatrixBase::makeHouseholderInPlace(), MatrixBase::applyHouseholderOnTheLeft(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -template -void MatrixBase::makeHouseholder( - EssentialPart& essential, - Scalar& tau, - RealScalar& beta) const -{ - using std::sqrt; - using numext::conj; - - EIGEN_STATIC_ASSERT_VECTOR_ONLY(EssentialPart) - VectorBlock tail(derived(), 1, size()-1); - - RealScalar tailSqNorm = size()==1 ? RealScalar(0) : tail.squaredNorm(); - Scalar c0 = coeff(0); - const RealScalar tol = (std::numeric_limits::min)(); - - if(tailSqNorm <= tol && numext::abs2(numext::imag(c0))<=tol) - { - tau = RealScalar(0); - beta = numext::real(c0); - essential.setZero(); - } - else - { - beta = sqrt(numext::abs2(c0) + tailSqNorm); - if (numext::real(c0)>=RealScalar(0)) - beta = -beta; - essential = tail / (c0 - beta); - tau = conj((beta - c0) / beta); - } -} - -/** Apply the elementary reflector H given by - * \f$ H = I - tau v v^*\f$ - * with - * \f$ v^T = [1 essential^T] \f$ - * from the left to a vector or matrix. - * - * On input: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param workspace a pointer to working space with at least - * this->cols() * essential.size() entries - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::makeHouseholderInPlace(), - * MatrixBase::applyHouseholderOnTheRight() - */ -template -template -void MatrixBase::applyHouseholderOnTheLeft( - const EssentialPart& essential, - const Scalar& tau, - Scalar* workspace) -{ - if(rows() == 1) - { - *this *= Scalar(1)-tau; - } - else if(tau!=Scalar(0)) - { - Map::type> tmp(workspace,cols()); - Block bottom(derived(), 1, 0, rows()-1, cols()); - tmp.noalias() = essential.adjoint() * bottom; - tmp += this->row(0); - this->row(0) -= tau * tmp; - bottom.noalias() -= tau * essential * tmp; - } -} - -/** Apply the elementary reflector H given by - * \f$ H = I - tau v v^*\f$ - * with - * \f$ v^T = [1 essential^T] \f$ - * from the right to a vector or matrix. - * - * On input: - * \param essential the essential part of the vector \c v - * \param tau the scaling factor of the Householder transformation - * \param workspace a pointer to working space with at least - * this->cols() * essential.size() entries - * - * \sa MatrixBase::makeHouseholder(), MatrixBase::makeHouseholderInPlace(), - * MatrixBase::applyHouseholderOnTheLeft() - */ -template -template -void MatrixBase::applyHouseholderOnTheRight( - const EssentialPart& essential, - const Scalar& tau, - Scalar* workspace) -{ - if(cols() == 1) - { - *this *= Scalar(1)-tau; - } - else if(tau!=Scalar(0)) - { - Map::type> tmp(workspace,rows()); - Block right(derived(), 0, 1, rows(), cols()-1); - tmp.noalias() = right * essential.conjugate(); - tmp += this->col(0); - this->col(0) -= tau * tmp; - right.noalias() -= tau * tmp * essential.transpose(); - } -} - -} // end namespace Eigen - -#endif // EIGEN_HOUSEHOLDER_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h deleted file mode 100644 index 3ce0a693d9b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Householder/HouseholderSequence.h +++ /dev/null @@ -1,470 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Gael Guennebaud -// Copyright (C) 2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_HOUSEHOLDER_SEQUENCE_H -#define EIGEN_HOUSEHOLDER_SEQUENCE_H - -namespace Eigen { - -/** \ingroup Householder_Module - * \householder_module - * \class HouseholderSequence - * \brief Sequence of Householder reflections acting on subspaces with decreasing size - * \tparam VectorsType type of matrix containing the Householder vectors - * \tparam CoeffsType type of vector containing the Householder coefficients - * \tparam Side either OnTheLeft (the default) or OnTheRight - * - * This class represents a product sequence of Householder reflections where the first Householder reflection - * acts on the whole space, the second Householder reflection leaves the one-dimensional subspace spanned by - * the first unit vector invariant, the third Householder reflection leaves the two-dimensional subspace - * spanned by the first two unit vectors invariant, and so on up to the last reflection which leaves all but - * one dimensions invariant and acts only on the last dimension. Such sequences of Householder reflections - * are used in several algorithms to zero out certain parts of a matrix. Indeed, the methods - * HessenbergDecomposition::matrixQ(), Tridiagonalization::matrixQ(), HouseholderQR::householderQ(), - * and ColPivHouseholderQR::householderQ() all return a %HouseholderSequence. - * - * More precisely, the class %HouseholderSequence represents an \f$ n \times n \f$ matrix \f$ H \f$ of the - * form \f$ H = \prod_{i=0}^{n-1} H_i \f$ where the i-th Householder reflection is \f$ H_i = I - h_i v_i - * v_i^* \f$. The i-th Householder coefficient \f$ h_i \f$ is a scalar and the i-th Householder vector \f$ - * v_i \f$ is a vector of the form - * \f[ - * v_i = [\underbrace{0, \ldots, 0}_{i-1\mbox{ zeros}}, 1, \underbrace{*, \ldots,*}_{n-i\mbox{ arbitrary entries}} ]. - * \f] - * The last \f$ n-i \f$ entries of \f$ v_i \f$ are called the essential part of the Householder vector. - * - * Typical usages are listed below, where H is a HouseholderSequence: - * \code - * A.applyOnTheRight(H); // A = A * H - * A.applyOnTheLeft(H); // A = H * A - * A.applyOnTheRight(H.adjoint()); // A = A * H^* - * A.applyOnTheLeft(H.adjoint()); // A = H^* * A - * MatrixXd Q = H; // conversion to a dense matrix - * \endcode - * In addition to the adjoint, you can also apply the inverse (=adjoint), the transpose, and the conjugate operators. - * - * See the documentation for HouseholderSequence(const VectorsType&, const CoeffsType&) for an example. - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ - -namespace internal { - -template -struct traits > -{ - typedef typename VectorsType::Scalar Scalar; - typedef typename VectorsType::StorageIndex StorageIndex; - typedef typename VectorsType::StorageKind StorageKind; - enum { - RowsAtCompileTime = Side==OnTheLeft ? traits::RowsAtCompileTime - : traits::ColsAtCompileTime, - ColsAtCompileTime = RowsAtCompileTime, - MaxRowsAtCompileTime = Side==OnTheLeft ? traits::MaxRowsAtCompileTime - : traits::MaxColsAtCompileTime, - MaxColsAtCompileTime = MaxRowsAtCompileTime, - Flags = 0 - }; -}; - -struct HouseholderSequenceShape {}; - -template -struct evaluator_traits > - : public evaluator_traits_base > -{ - typedef HouseholderSequenceShape Shape; -}; - -template -struct hseq_side_dependent_impl -{ - typedef Block EssentialVectorType; - typedef HouseholderSequence HouseholderSequenceType; - static inline const EssentialVectorType essentialVector(const HouseholderSequenceType& h, Index k) - { - Index start = k+1+h.m_shift; - return Block(h.m_vectors, start, k, h.rows()-start, 1); - } -}; - -template -struct hseq_side_dependent_impl -{ - typedef Transpose > EssentialVectorType; - typedef HouseholderSequence HouseholderSequenceType; - static inline const EssentialVectorType essentialVector(const HouseholderSequenceType& h, Index k) - { - Index start = k+1+h.m_shift; - return Block(h.m_vectors, k, start, 1, h.rows()-start).transpose(); - } -}; - -template struct matrix_type_times_scalar_type -{ - typedef typename ScalarBinaryOpTraits::ReturnType - ResultScalar; - typedef Matrix Type; -}; - -} // end namespace internal - -template class HouseholderSequence - : public EigenBase > -{ - typedef typename internal::hseq_side_dependent_impl::EssentialVectorType EssentialVectorType; - - public: - enum { - RowsAtCompileTime = internal::traits::RowsAtCompileTime, - ColsAtCompileTime = internal::traits::ColsAtCompileTime, - MaxRowsAtCompileTime = internal::traits::MaxRowsAtCompileTime, - MaxColsAtCompileTime = internal::traits::MaxColsAtCompileTime - }; - typedef typename internal::traits::Scalar Scalar; - - typedef HouseholderSequence< - typename internal::conditional::IsComplex, - typename internal::remove_all::type, - VectorsType>::type, - typename internal::conditional::IsComplex, - typename internal::remove_all::type, - CoeffsType>::type, - Side - > ConjugateReturnType; - - /** \brief Constructor. - * \param[in] v %Matrix containing the essential parts of the Householder vectors - * \param[in] h Vector containing the Householder coefficients - * - * Constructs the Householder sequence with coefficients given by \p h and vectors given by \p v. The - * i-th Householder coefficient \f$ h_i \f$ is given by \p h(i) and the essential part of the i-th - * Householder vector \f$ v_i \f$ is given by \p v(k,i) with \p k > \p i (the subdiagonal part of the - * i-th column). If \p v has fewer columns than rows, then the Householder sequence contains as many - * Householder reflections as there are columns. - * - * \note The %HouseholderSequence object stores \p v and \p h by reference. - * - * Example: \include HouseholderSequence_HouseholderSequence.cpp - * Output: \verbinclude HouseholderSequence_HouseholderSequence.out - * - * \sa setLength(), setShift() - */ - HouseholderSequence(const VectorsType& v, const CoeffsType& h) - : m_vectors(v), m_coeffs(h), m_trans(false), m_length(v.diagonalSize()), - m_shift(0) - { - } - - /** \brief Copy constructor. */ - HouseholderSequence(const HouseholderSequence& other) - : m_vectors(other.m_vectors), - m_coeffs(other.m_coeffs), - m_trans(other.m_trans), - m_length(other.m_length), - m_shift(other.m_shift) - { - } - - /** \brief Number of rows of transformation viewed as a matrix. - * \returns Number of rows - * \details This equals the dimension of the space that the transformation acts on. - */ - Index rows() const { return Side==OnTheLeft ? m_vectors.rows() : m_vectors.cols(); } - - /** \brief Number of columns of transformation viewed as a matrix. - * \returns Number of columns - * \details This equals the dimension of the space that the transformation acts on. - */ - Index cols() const { return rows(); } - - /** \brief Essential part of a Householder vector. - * \param[in] k Index of Householder reflection - * \returns Vector containing non-trivial entries of k-th Householder vector - * - * This function returns the essential part of the Householder vector \f$ v_i \f$. This is a vector of - * length \f$ n-i \f$ containing the last \f$ n-i \f$ entries of the vector - * \f[ - * v_i = [\underbrace{0, \ldots, 0}_{i-1\mbox{ zeros}}, 1, \underbrace{*, \ldots,*}_{n-i\mbox{ arbitrary entries}} ]. - * \f] - * The index \f$ i \f$ equals \p k + shift(), corresponding to the k-th column of the matrix \p v - * passed to the constructor. - * - * \sa setShift(), shift() - */ - const EssentialVectorType essentialVector(Index k) const - { - eigen_assert(k >= 0 && k < m_length); - return internal::hseq_side_dependent_impl::essentialVector(*this, k); - } - - /** \brief %Transpose of the Householder sequence. */ - HouseholderSequence transpose() const - { - return HouseholderSequence(*this).setTrans(!m_trans); - } - - /** \brief Complex conjugate of the Householder sequence. */ - ConjugateReturnType conjugate() const - { - return ConjugateReturnType(m_vectors.conjugate(), m_coeffs.conjugate()) - .setTrans(m_trans) - .setLength(m_length) - .setShift(m_shift); - } - - /** \brief Adjoint (conjugate transpose) of the Householder sequence. */ - ConjugateReturnType adjoint() const - { - return conjugate().setTrans(!m_trans); - } - - /** \brief Inverse of the Householder sequence (equals the adjoint). */ - ConjugateReturnType inverse() const { return adjoint(); } - - /** \internal */ - template inline void evalTo(DestType& dst) const - { - Matrix workspace(rows()); - evalTo(dst, workspace); - } - - /** \internal */ - template - void evalTo(Dest& dst, Workspace& workspace) const - { - workspace.resize(rows()); - Index vecs = m_length; - if(internal::is_same_dense(dst,m_vectors)) - { - // in-place - dst.diagonal().setOnes(); - dst.template triangularView().setZero(); - for(Index k = vecs-1; k >= 0; --k) - { - Index cornerSize = rows() - k - m_shift; - if(m_trans) - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheRight(essentialVector(k), m_coeffs.coeff(k), workspace.data()); - else - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheLeft(essentialVector(k), m_coeffs.coeff(k), workspace.data()); - - // clear the off diagonal vector - dst.col(k).tail(rows()-k-1).setZero(); - } - // clear the remaining columns if needed - for(Index k = 0; k= 0; --k) - { - Index cornerSize = rows() - k - m_shift; - if(m_trans) - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheRight(essentialVector(k), m_coeffs.coeff(k), &workspace.coeffRef(0)); - else - dst.bottomRightCorner(cornerSize, cornerSize) - .applyHouseholderOnTheLeft(essentialVector(k), m_coeffs.coeff(k), &workspace.coeffRef(0)); - } - } - } - - /** \internal */ - template inline void applyThisOnTheRight(Dest& dst) const - { - Matrix workspace(dst.rows()); - applyThisOnTheRight(dst, workspace); - } - - /** \internal */ - template - inline void applyThisOnTheRight(Dest& dst, Workspace& workspace) const - { - workspace.resize(dst.rows()); - for(Index k = 0; k < m_length; ++k) - { - Index actual_k = m_trans ? m_length-k-1 : k; - dst.rightCols(rows()-m_shift-actual_k) - .applyHouseholderOnTheRight(essentialVector(actual_k), m_coeffs.coeff(actual_k), workspace.data()); - } - } - - /** \internal */ - template inline void applyThisOnTheLeft(Dest& dst) const - { - Matrix workspace; - applyThisOnTheLeft(dst, workspace); - } - - /** \internal */ - template - inline void applyThisOnTheLeft(Dest& dst, Workspace& workspace) const - { - const Index BlockSize = 48; - // if the entries are large enough, then apply the reflectors by block - if(m_length>=BlockSize && dst.cols()>1) - { - for(Index i = 0; i < m_length; i+=BlockSize) - { - Index end = m_trans ? (std::min)(m_length,i+BlockSize) : m_length-i; - Index k = m_trans ? i : (std::max)(Index(0),end-BlockSize); - Index bs = end-k; - Index start = k + m_shift; - - typedef Block::type,Dynamic,Dynamic> SubVectorsType; - SubVectorsType sub_vecs1(m_vectors.const_cast_derived(), Side==OnTheRight ? k : start, - Side==OnTheRight ? start : k, - Side==OnTheRight ? bs : m_vectors.rows()-start, - Side==OnTheRight ? m_vectors.cols()-start : bs); - typename internal::conditional, SubVectorsType&>::type sub_vecs(sub_vecs1); - Block sub_dst(dst,dst.rows()-rows()+m_shift+k,0, rows()-m_shift-k,dst.cols()); - apply_block_householder_on_the_left(sub_dst, sub_vecs, m_coeffs.segment(k, bs), !m_trans); - } - } - else - { - workspace.resize(dst.cols()); - for(Index k = 0; k < m_length; ++k) - { - Index actual_k = m_trans ? k : m_length-k-1; - dst.bottomRows(rows()-m_shift-actual_k) - .applyHouseholderOnTheLeft(essentialVector(actual_k), m_coeffs.coeff(actual_k), workspace.data()); - } - } - } - - /** \brief Computes the product of a Householder sequence with a matrix. - * \param[in] other %Matrix being multiplied. - * \returns Expression object representing the product. - * - * This function computes \f$ HM \f$ where \f$ H \f$ is the Householder sequence represented by \p *this - * and \f$ M \f$ is the matrix \p other. - */ - template - typename internal::matrix_type_times_scalar_type::Type operator*(const MatrixBase& other) const - { - typename internal::matrix_type_times_scalar_type::Type - res(other.template cast::ResultScalar>()); - applyThisOnTheLeft(res); - return res; - } - - template friend struct internal::hseq_side_dependent_impl; - - /** \brief Sets the length of the Householder sequence. - * \param [in] length New value for the length. - * - * By default, the length \f$ n \f$ of the Householder sequence \f$ H = H_0 H_1 \ldots H_{n-1} \f$ is set - * to the number of columns of the matrix \p v passed to the constructor, or the number of rows if that - * is smaller. After this function is called, the length equals \p length. - * - * \sa length() - */ - HouseholderSequence& setLength(Index length) - { - m_length = length; - return *this; - } - - /** \brief Sets the shift of the Householder sequence. - * \param [in] shift New value for the shift. - * - * By default, a %HouseholderSequence object represents \f$ H = H_0 H_1 \ldots H_{n-1} \f$ and the i-th - * column of the matrix \p v passed to the constructor corresponds to the i-th Householder - * reflection. After this function is called, the object represents \f$ H = H_{\mathrm{shift}} - * H_{\mathrm{shift}+1} \ldots H_{n-1} \f$ and the i-th column of \p v corresponds to the (shift+i)-th - * Householder reflection. - * - * \sa shift() - */ - HouseholderSequence& setShift(Index shift) - { - m_shift = shift; - return *this; - } - - Index length() const { return m_length; } /**< \brief Returns the length of the Householder sequence. */ - Index shift() const { return m_shift; } /**< \brief Returns the shift of the Householder sequence. */ - - /* Necessary for .adjoint() and .conjugate() */ - template friend class HouseholderSequence; - - protected: - - /** \brief Sets the transpose flag. - * \param [in] trans New value of the transpose flag. - * - * By default, the transpose flag is not set. If the transpose flag is set, then this object represents - * \f$ H^T = H_{n-1}^T \ldots H_1^T H_0^T \f$ instead of \f$ H = H_0 H_1 \ldots H_{n-1} \f$. - * - * \sa trans() - */ - HouseholderSequence& setTrans(bool trans) - { - m_trans = trans; - return *this; - } - - bool trans() const { return m_trans; } /**< \brief Returns the transpose flag. */ - - typename VectorsType::Nested m_vectors; - typename CoeffsType::Nested m_coeffs; - bool m_trans; - Index m_length; - Index m_shift; -}; - -/** \brief Computes the product of a matrix with a Householder sequence. - * \param[in] other %Matrix being multiplied. - * \param[in] h %HouseholderSequence being multiplied. - * \returns Expression object representing the product. - * - * This function computes \f$ MH \f$ where \f$ M \f$ is the matrix \p other and \f$ H \f$ is the - * Householder sequence represented by \p h. - */ -template -typename internal::matrix_type_times_scalar_type::Type operator*(const MatrixBase& other, const HouseholderSequence& h) -{ - typename internal::matrix_type_times_scalar_type::Type - res(other.template cast::ResultScalar>()); - h.applyThisOnTheRight(res); - return res; -} - -/** \ingroup Householder_Module \householder_module - * \brief Convenience function for constructing a Householder sequence. - * \returns A HouseholderSequence constructed from the specified arguments. - */ -template -HouseholderSequence householderSequence(const VectorsType& v, const CoeffsType& h) -{ - return HouseholderSequence(v, h); -} - -/** \ingroup Householder_Module \householder_module - * \brief Convenience function for constructing a Householder sequence. - * \returns A HouseholderSequence constructed from the specified arguments. - * \details This function differs from householderSequence() in that the template argument \p OnTheSide of - * the constructed HouseholderSequence is set to OnTheRight, instead of the default OnTheLeft. - */ -template -HouseholderSequence rightHouseholderSequence(const VectorsType& v, const CoeffsType& h) -{ - return HouseholderSequence(v, h); -} - -} // end namespace Eigen - -#endif // EIGEN_HOUSEHOLDER_SEQUENCE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h deleted file mode 100644 index 1998c632274..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/Jacobi/Jacobi.h +++ /dev/null @@ -1,462 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_JACOBI_H -#define EIGEN_JACOBI_H - -namespace Eigen { - -/** \ingroup Jacobi_Module - * \jacobi_module - * \class JacobiRotation - * \brief Rotation given by a cosine-sine pair. - * - * This class represents a Jacobi or Givens rotation. - * This is a 2D rotation in the plane \c J of angle \f$ \theta \f$ defined by - * its cosine \c c and sine \c s as follow: - * \f$ J = \left ( \begin{array}{cc} c & \overline s \\ -s & \overline c \end{array} \right ) \f$ - * - * You can apply the respective counter-clockwise rotation to a column vector \c v by - * applying its adjoint on the left: \f$ v = J^* v \f$ that translates to the following Eigen code: - * \code - * v.applyOnTheLeft(J.adjoint()); - * \endcode - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template class JacobiRotation -{ - public: - typedef typename NumTraits::Real RealScalar; - - /** Default constructor without any initialization. */ - JacobiRotation() {} - - /** Construct a planar rotation from a cosine-sine pair (\a c, \c s). */ - JacobiRotation(const Scalar& c, const Scalar& s) : m_c(c), m_s(s) {} - - Scalar& c() { return m_c; } - Scalar c() const { return m_c; } - Scalar& s() { return m_s; } - Scalar s() const { return m_s; } - - /** Concatenates two planar rotation */ - JacobiRotation operator*(const JacobiRotation& other) - { - using numext::conj; - return JacobiRotation(m_c * other.m_c - conj(m_s) * other.m_s, - conj(m_c * conj(other.m_s) + conj(m_s) * conj(other.m_c))); - } - - /** Returns the transposed transformation */ - JacobiRotation transpose() const { using numext::conj; return JacobiRotation(m_c, -conj(m_s)); } - - /** Returns the adjoint transformation */ - JacobiRotation adjoint() const { using numext::conj; return JacobiRotation(conj(m_c), -m_s); } - - template - bool makeJacobi(const MatrixBase&, Index p, Index q); - bool makeJacobi(const RealScalar& x, const Scalar& y, const RealScalar& z); - - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r=0); - - protected: - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::true_type); - void makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::false_type); - - Scalar m_c, m_s; -}; - -/** Makes \c *this as a Jacobi rotation \a J such that applying \a J on both the right and left sides of the selfadjoint 2x2 matrix - * \f$ B = \left ( \begin{array}{cc} x & y \\ \overline y & z \end{array} \right )\f$ yields a diagonal matrix \f$ A = J^* B J \f$ - * - * \sa MatrixBase::makeJacobi(const MatrixBase&, Index, Index), MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -bool JacobiRotation::makeJacobi(const RealScalar& x, const Scalar& y, const RealScalar& z) -{ - using std::sqrt; - using std::abs; - RealScalar deno = RealScalar(2)*abs(y); - if(deno < (std::numeric_limits::min)()) - { - m_c = Scalar(1); - m_s = Scalar(0); - return false; - } - else - { - RealScalar tau = (x-z)/deno; - RealScalar w = sqrt(numext::abs2(tau) + RealScalar(1)); - RealScalar t; - if(tau>RealScalar(0)) - { - t = RealScalar(1) / (tau + w); - } - else - { - t = RealScalar(1) / (tau - w); - } - RealScalar sign_t = t > RealScalar(0) ? RealScalar(1) : RealScalar(-1); - RealScalar n = RealScalar(1) / sqrt(numext::abs2(t)+RealScalar(1)); - m_s = - sign_t * (numext::conj(y) / abs(y)) * abs(t) * n; - m_c = n; - return true; - } -} - -/** Makes \c *this as a Jacobi rotation \c J such that applying \a J on both the right and left sides of the 2x2 selfadjoint matrix - * \f$ B = \left ( \begin{array}{cc} \text{this}_{pp} & \text{this}_{pq} \\ (\text{this}_{pq})^* & \text{this}_{qq} \end{array} \right )\f$ yields - * a diagonal matrix \f$ A = J^* B J \f$ - * - * Example: \include Jacobi_makeJacobi.cpp - * Output: \verbinclude Jacobi_makeJacobi.out - * - * \sa JacobiRotation::makeJacobi(RealScalar, Scalar, RealScalar), MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -template -inline bool JacobiRotation::makeJacobi(const MatrixBase& m, Index p, Index q) -{ - return makeJacobi(numext::real(m.coeff(p,p)), m.coeff(p,q), numext::real(m.coeff(q,q))); -} - -/** Makes \c *this as a Givens rotation \c G such that applying \f$ G^* \f$ to the left of the vector - * \f$ V = \left ( \begin{array}{c} p \\ q \end{array} \right )\f$ yields: - * \f$ G^* V = \left ( \begin{array}{c} r \\ 0 \end{array} \right )\f$. - * - * The value of \a r is returned if \a r is not null (the default is null). - * Also note that G is built such that the cosine is always real. - * - * Example: \include Jacobi_makeGivens.cpp - * Output: \verbinclude Jacobi_makeGivens.out - * - * This function implements the continuous Givens rotation generation algorithm - * found in Anderson (2000), Discontinuous Plane Rotations and the Symmetric Eigenvalue Problem. - * LAPACK Working Note 150, University of Tennessee, UT-CS-00-454, December 4, 2000. - * - * \sa MatrixBase::applyOnTheLeft(), MatrixBase::applyOnTheRight() - */ -template -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r) -{ - makeGivens(p, q, r, typename internal::conditional::IsComplex, internal::true_type, internal::false_type>::type()); -} - - -// specialization for complexes -template -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::true_type) -{ - using std::sqrt; - using std::abs; - using numext::conj; - - if(q==Scalar(0)) - { - m_c = numext::real(p)<0 ? Scalar(-1) : Scalar(1); - m_s = 0; - if(r) *r = m_c * p; - } - else if(p==Scalar(0)) - { - m_c = 0; - m_s = -q/abs(q); - if(r) *r = abs(q); - } - else - { - RealScalar p1 = numext::norm1(p); - RealScalar q1 = numext::norm1(q); - if(p1>=q1) - { - Scalar ps = p / p1; - RealScalar p2 = numext::abs2(ps); - Scalar qs = q / p1; - RealScalar q2 = numext::abs2(qs); - - RealScalar u = sqrt(RealScalar(1) + q2/p2); - if(numext::real(p) -void JacobiRotation::makeGivens(const Scalar& p, const Scalar& q, Scalar* r, internal::false_type) -{ - using std::sqrt; - using std::abs; - if(q==Scalar(0)) - { - m_c = p abs(q)) - { - Scalar t = q/p; - Scalar u = sqrt(Scalar(1) + numext::abs2(t)); - if(p -void apply_rotation_in_the_plane(DenseBase& xpr_x, DenseBase& xpr_y, const JacobiRotation& j); -} - -/** \jacobi_module - * Applies the rotation in the plane \a j to the rows \a p and \a q of \c *this, i.e., it computes B = J * B, - * with \f$ B = \left ( \begin{array}{cc} \text{*this.row}(p) \\ \text{*this.row}(q) \end{array} \right ) \f$. - * - * \sa class JacobiRotation, MatrixBase::applyOnTheRight(), internal::apply_rotation_in_the_plane() - */ -template -template -inline void MatrixBase::applyOnTheLeft(Index p, Index q, const JacobiRotation& j) -{ - RowXpr x(this->row(p)); - RowXpr y(this->row(q)); - internal::apply_rotation_in_the_plane(x, y, j); -} - -/** \ingroup Jacobi_Module - * Applies the rotation in the plane \a j to the columns \a p and \a q of \c *this, i.e., it computes B = B * J - * with \f$ B = \left ( \begin{array}{cc} \text{*this.col}(p) & \text{*this.col}(q) \end{array} \right ) \f$. - * - * \sa class JacobiRotation, MatrixBase::applyOnTheLeft(), internal::apply_rotation_in_the_plane() - */ -template -template -inline void MatrixBase::applyOnTheRight(Index p, Index q, const JacobiRotation& j) -{ - ColXpr x(this->col(p)); - ColXpr y(this->col(q)); - internal::apply_rotation_in_the_plane(x, y, j.transpose()); -} - -namespace internal { - -template -struct apply_rotation_in_the_plane_selector -{ - static inline void run(Scalar *x, Index incrx, Scalar *y, Index incry, Index size, OtherScalar c, OtherScalar s) - { - for(Index i=0; i -struct apply_rotation_in_the_plane_selector -{ - static inline void run(Scalar *x, Index incrx, Scalar *y, Index incry, Index size, OtherScalar c, OtherScalar s) - { - enum { - PacketSize = packet_traits::size, - OtherPacketSize = packet_traits::size - }; - typedef typename packet_traits::type Packet; - typedef typename packet_traits::type OtherPacket; - - /*** dynamic-size vectorized paths ***/ - if(SizeAtCompileTime == Dynamic && ((incrx==1 && incry==1) || PacketSize == 1)) - { - // both vectors are sequentially stored in memory => vectorization - enum { Peeling = 2 }; - - Index alignedStart = internal::first_default_aligned(y, size); - Index alignedEnd = alignedStart + ((size-alignedStart)/PacketSize)*PacketSize; - - const OtherPacket pc = pset1(c); - const OtherPacket ps = pset1(s); - conj_helper::IsComplex,false> pcj; - conj_helper pm; - - for(Index i=0; i(px); - Packet yi = pload(py); - pstore(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore(py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - px += PacketSize; - py += PacketSize; - } - } - else - { - Index peelingEnd = alignedStart + ((size-alignedStart)/(Peeling*PacketSize))*(Peeling*PacketSize); - for(Index i=alignedStart; i(px); - Packet xi1 = ploadu(px+PacketSize); - Packet yi = pload (py); - Packet yi1 = pload (py+PacketSize); - pstoreu(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstoreu(px+PacketSize, padd(pm.pmul(pc,xi1),pcj.pmul(ps,yi1))); - pstore (py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - pstore (py+PacketSize, psub(pcj.pmul(pc,yi1),pm.pmul(ps,xi1))); - px += Peeling*PacketSize; - py += Peeling*PacketSize; - } - if(alignedEnd!=peelingEnd) - { - Packet xi = ploadu(x+peelingEnd); - Packet yi = pload (y+peelingEnd); - pstoreu(x+peelingEnd, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore (y+peelingEnd, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - } - } - - for(Index i=alignedEnd; i0) // FIXME should be compared to the required alignment - { - const OtherPacket pc = pset1(c); - const OtherPacket ps = pset1(s); - conj_helper::IsComplex,false> pcj; - conj_helper pm; - Scalar* EIGEN_RESTRICT px = x; - Scalar* EIGEN_RESTRICT py = y; - for(Index i=0; i(px); - Packet yi = pload(py); - pstore(px, padd(pm.pmul(pc,xi),pcj.pmul(ps,yi))); - pstore(py, psub(pcj.pmul(pc,yi),pm.pmul(ps,xi))); - px += PacketSize; - py += PacketSize; - } - } - - /*** non-vectorized path ***/ - else - { - apply_rotation_in_the_plane_selector::run(x,incrx,y,incry,size,c,s); - } - } -}; - -template -void /*EIGEN_DONT_INLINE*/ apply_rotation_in_the_plane(DenseBase& xpr_x, DenseBase& xpr_y, const JacobiRotation& j) -{ - typedef typename VectorX::Scalar Scalar; - const bool Vectorizable = (VectorX::Flags & VectorY::Flags & PacketAccessBit) - && (int(packet_traits::size) == int(packet_traits::size)); - - eigen_assert(xpr_x.size() == xpr_y.size()); - Index size = xpr_x.size(); - Index incrx = xpr_x.derived().innerStride(); - Index incry = xpr_y.derived().innerStride(); - - Scalar* EIGEN_RESTRICT x = &xpr_x.derived().coeffRef(0); - Scalar* EIGEN_RESTRICT y = &xpr_y.derived().coeffRef(0); - - OtherScalar c = j.c(); - OtherScalar s = j.s(); - if (c==OtherScalar(1) && s==OtherScalar(0)) - return; - - apply_rotation_in_the_plane_selector< - Scalar,OtherScalar, - VectorX::SizeAtCompileTime, - EIGEN_PLAIN_ENUM_MIN(evaluator::Alignment, evaluator::Alignment), - Vectorizable>::run(x,incrx,y,incry,size,c,s); -} - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_JACOBI_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h deleted file mode 100644 index d6a3c1e5a54..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/Determinant.h +++ /dev/null @@ -1,101 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_DETERMINANT_H -#define EIGEN_DETERMINANT_H - -namespace Eigen { - -namespace internal { - -template -inline const typename Derived::Scalar bruteforce_det3_helper -(const MatrixBase& matrix, int a, int b, int c) -{ - return matrix.coeff(0,a) - * (matrix.coeff(1,b) * matrix.coeff(2,c) - matrix.coeff(1,c) * matrix.coeff(2,b)); -} - -template -const typename Derived::Scalar bruteforce_det4_helper -(const MatrixBase& matrix, int j, int k, int m, int n) -{ - return (matrix.coeff(j,0) * matrix.coeff(k,1) - matrix.coeff(k,0) * matrix.coeff(j,1)) - * (matrix.coeff(m,2) * matrix.coeff(n,3) - matrix.coeff(n,2) * matrix.coeff(m,3)); -} - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - if(Derived::ColsAtCompileTime==Dynamic && m.rows()==0) - return typename traits::Scalar(1); - return m.partialPivLu().determinant(); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return m.coeff(0,0); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return m.coeff(0,0) * m.coeff(1,1) - m.coeff(1,0) * m.coeff(0,1); - } -}; - -template struct determinant_impl -{ - static inline typename traits::Scalar run(const Derived& m) - { - return bruteforce_det3_helper(m,0,1,2) - - bruteforce_det3_helper(m,1,0,2) - + bruteforce_det3_helper(m,2,0,1); - } -}; - -template struct determinant_impl -{ - static typename traits::Scalar run(const Derived& m) - { - // trick by Martin Costabel to compute 4x4 det with only 30 muls - return bruteforce_det4_helper(m,0,1,2,3) - - bruteforce_det4_helper(m,0,2,1,3) - + bruteforce_det4_helper(m,0,3,1,2) - + bruteforce_det4_helper(m,1,2,0,3) - - bruteforce_det4_helper(m,1,3,0,2) - + bruteforce_det4_helper(m,2,3,0,1); - } -}; - -} // end namespace internal - -/** \lu_module - * - * \returns the determinant of this matrix - */ -template -inline typename internal::traits::Scalar MatrixBase::determinant() const -{ - eigen_assert(rows() == cols()); - typedef typename internal::nested_eval::type Nested; - return internal::determinant_impl::type>::run(derived()); -} - -} // end namespace Eigen - -#endif // EIGEN_DETERMINANT_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h deleted file mode 100644 index 03b6af70613..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/FullPivLU.h +++ /dev/null @@ -1,891 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_LU_H -#define EIGEN_LU_H - -namespace Eigen { - -namespace internal { -template struct traits > - : traits<_MatrixType> -{ - typedef MatrixXpr XprKind; - typedef SolverStorage StorageKind; - enum { Flags = 0 }; -}; - -} // end namespace internal - -/** \ingroup LU_Module - * - * \class FullPivLU - * - * \brief LU decomposition of a matrix with complete pivoting, and related features - * - * \tparam _MatrixType the type of the matrix of which we are computing the LU decomposition - * - * This class represents a LU decomposition of any matrix, with complete pivoting: the matrix A is - * decomposed as \f$ A = P^{-1} L U Q^{-1} \f$ where L is unit-lower-triangular, U is - * upper-triangular, and P and Q are permutation matrices. This is a rank-revealing LU - * decomposition. The eigenvalues (diagonal coefficients) of U are sorted in such a way that any - * zeros are at the end. - * - * This decomposition provides the generic approach to solving systems of linear equations, computing - * the rank, invertibility, inverse, kernel, and determinant. - * - * This LU decomposition is very stable and well tested with large matrices. However there are use cases where the SVD - * decomposition is inherently more stable and/or flexible. For example, when computing the kernel of a matrix, - * working with the SVD allows to select the smallest singular values of the matrix, something that - * the LU decomposition doesn't see. - * - * The data of the LU decomposition can be directly accessed through the methods matrixLU(), - * permutationP(), permutationQ(). - * - * As an exemple, here is how the original matrix can be retrieved: - * \include class_FullPivLU.cpp - * Output: \verbinclude class_FullPivLU.out - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::fullPivLu(), MatrixBase::determinant(), MatrixBase::inverse() - */ -template class FullPivLU - : public SolverBase > -{ - public: - typedef _MatrixType MatrixType; - typedef SolverBase Base; - - EIGEN_GENERIC_PUBLIC_INTERFACE(FullPivLU) - // FIXME StorageIndex defined in EIGEN_GENERIC_PUBLIC_INTERFACE should be int - enum { - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef typename internal::plain_row_type::type IntRowVectorType; - typedef typename internal::plain_col_type::type IntColVectorType; - typedef PermutationMatrix PermutationQType; - typedef PermutationMatrix PermutationPType; - typedef typename MatrixType::PlainObject PlainObject; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via LU::compute(const MatrixType&). - */ - FullPivLU(); - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa FullPivLU() - */ - FullPivLU(Index rows, Index cols); - - /** Constructor. - * - * \param matrix the matrix of which to compute the LU decomposition. - * It is required to be nonzero. - */ - template - explicit FullPivLU(const EigenBase& matrix); - - /** \brief Constructs a LU factorization from a given matrix - * - * This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref. - * - * \sa FullPivLU(const EigenBase&) - */ - template - explicit FullPivLU(EigenBase& matrix); - - /** Computes the LU decomposition of the given matrix. - * - * \param matrix the matrix of which to compute the LU decomposition. - * It is required to be nonzero. - * - * \returns a reference to *this - */ - template - FullPivLU& compute(const EigenBase& matrix) { - m_lu = matrix.derived(); - computeInPlace(); - return *this; - } - - /** \returns the LU decomposition matrix: the upper-triangular part is U, the - * unit-lower-triangular part is L (at least for square matrices; in the non-square - * case, special care is needed, see the documentation of class FullPivLU). - * - * \sa matrixL(), matrixU() - */ - inline const MatrixType& matrixLU() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_lu; - } - - /** \returns the number of nonzero pivots in the LU decomposition. - * Here nonzero is meant in the exact sense, not in a fuzzy sense. - * So that notion isn't really intrinsically interesting, but it is - * still useful when implementing algorithms. - * - * \sa rank() - */ - inline Index nonzeroPivots() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_nonzero_pivots; - } - - /** \returns the absolute value of the biggest pivot, i.e. the biggest - * diagonal coefficient of U. - */ - RealScalar maxPivot() const { return m_maxpivot; } - - /** \returns the permutation matrix P - * - * \sa permutationQ() - */ - EIGEN_DEVICE_FUNC inline const PermutationPType& permutationP() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_p; - } - - /** \returns the permutation matrix Q - * - * \sa permutationP() - */ - inline const PermutationQType& permutationQ() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return m_q; - } - - /** \returns the kernel of the matrix, also called its null-space. The columns of the returned matrix - * will form a basis of the kernel. - * - * \note If the kernel has dimension zero, then the returned matrix is a column-vector filled with zeros. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - * - * Example: \include FullPivLU_kernel.cpp - * Output: \verbinclude FullPivLU_kernel.out - * - * \sa image() - */ - inline const internal::kernel_retval kernel() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return internal::kernel_retval(*this); - } - - /** \returns the image of the matrix, also called its column-space. The columns of the returned matrix - * will form a basis of the image (column-space). - * - * \param originalMatrix the original matrix, of which *this is the LU decomposition. - * The reason why it is needed to pass it here, is that this allows - * a large optimization, as otherwise this method would need to reconstruct it - * from the LU decomposition. - * - * \note If the image has dimension zero, then the returned matrix is a column-vector filled with zeros. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - * - * Example: \include FullPivLU_image.cpp - * Output: \verbinclude FullPivLU_image.out - * - * \sa kernel() - */ - inline const internal::image_retval - image(const MatrixType& originalMatrix) const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return internal::image_retval(*this, originalMatrix); - } - - /** \return a solution x to the equation Ax=b, where A is the matrix of which - * *this is the LU decomposition. - * - * \param b the right-hand-side of the equation to solve. Can be a vector or a matrix, - * the only requirement in order for the equation to make sense is that - * b.rows()==A.rows(), where A is the matrix of which *this is the LU decomposition. - * - * \returns a solution. - * - * \note_about_checking_solutions - * - * \note_about_arbitrary_choice_of_solution - * \note_about_using_kernel_to_study_multiple_solutions - * - * Example: \include FullPivLU_solve.cpp - * Output: \verbinclude FullPivLU_solve.out - * - * \sa TriangularView::solve(), kernel(), inverse() - */ - // FIXME this is a copy-paste of the base-class member to add the isInitialized assertion. - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return Solve(*this, b.derived()); - } - - /** \returns an estimate of the reciprocal condition number of the matrix of which \c *this is - the LU decomposition. - */ - inline RealScalar rcond() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return internal::rcond_estimate_helper(m_l1_norm, *this); - } - - /** \returns the determinant of the matrix of which - * *this is the LU decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the LU decomposition has already been computed. - * - * \note This is only for square matrices. - * - * \note For fixed-size matrices of size up to 4, MatrixBase::determinant() offers - * optimized paths. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * - * \sa MatrixBase::determinant() - */ - typename internal::traits::Scalar determinant() const; - - /** Allows to prescribe a threshold to be used by certain methods, such as rank(), - * who need to determine when pivots are to be considered nonzero. This is not used for the - * LU decomposition itself. - * - * When it needs to get the threshold value, Eigen calls threshold(). By default, this - * uses a formula to automatically determine a reasonable threshold. - * Once you have called the present method setThreshold(const RealScalar&), - * your value is used instead. - * - * \param threshold The new value to use as the threshold. - * - * A pivot will be considered nonzero if its absolute value is strictly greater than - * \f$ \vert pivot \vert \leqslant threshold \times \vert maxpivot \vert \f$ - * where maxpivot is the biggest pivot. - * - * If you want to come back to the default behavior, call setThreshold(Default_t) - */ - FullPivLU& setThreshold(const RealScalar& threshold) - { - m_usePrescribedThreshold = true; - m_prescribedThreshold = threshold; - return *this; - } - - /** Allows to come back to the default behavior, letting Eigen use its default formula for - * determining the threshold. - * - * You should pass the special object Eigen::Default as parameter here. - * \code lu.setThreshold(Eigen::Default); \endcode - * - * See the documentation of setThreshold(const RealScalar&). - */ - FullPivLU& setThreshold(Default_t) - { - m_usePrescribedThreshold = false; - return *this; - } - - /** Returns the threshold that will be used by certain methods such as rank(). - * - * See the documentation of setThreshold(const RealScalar&). - */ - RealScalar threshold() const - { - eigen_assert(m_isInitialized || m_usePrescribedThreshold); - return m_usePrescribedThreshold ? m_prescribedThreshold - // this formula comes from experimenting (see "LU precision tuning" thread on the list) - // and turns out to be identical to Higham's formula used already in LDLt. - : NumTraits::epsilon() * m_lu.diagonalSize(); - } - - /** \returns the rank of the matrix of which *this is the LU decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index rank() const - { - using std::abs; - eigen_assert(m_isInitialized && "LU is not initialized."); - RealScalar premultiplied_threshold = abs(m_maxpivot) * threshold(); - Index result = 0; - for(Index i = 0; i < m_nonzero_pivots; ++i) - result += (abs(m_lu.coeff(i,i)) > premultiplied_threshold); - return result; - } - - /** \returns the dimension of the kernel of the matrix of which *this is the LU decomposition. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline Index dimensionOfKernel() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return cols() - rank(); - } - - /** \returns true if the matrix of which *this is the LU decomposition represents an injective - * linear map, i.e. has trivial kernel; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInjective() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return rank() == cols(); - } - - /** \returns true if the matrix of which *this is the LU decomposition represents a surjective - * linear map; false otherwise. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isSurjective() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return rank() == rows(); - } - - /** \returns true if the matrix of which *this is the LU decomposition is invertible. - * - * \note This method has to determine which pivots should be considered nonzero. - * For that, it uses the threshold value that you can control by calling - * setThreshold(const RealScalar&). - */ - inline bool isInvertible() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - return isInjective() && (m_lu.rows() == m_lu.cols()); - } - - /** \returns the inverse of the matrix of which *this is the LU decomposition. - * - * \note If this matrix is not invertible, the returned matrix has undefined coefficients. - * Use isInvertible() to first determine whether this matrix is invertible. - * - * \sa MatrixBase::inverse() - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "LU is not initialized."); - eigen_assert(m_lu.rows() == m_lu.cols() && "You can't take the inverse of a non-square matrix!"); - return Inverse(*this); - } - - MatrixType reconstructedMatrix() const; - - EIGEN_DEVICE_FUNC inline Index rows() const { return m_lu.rows(); } - EIGEN_DEVICE_FUNC inline Index cols() const { return m_lu.cols(); } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const; - - template - EIGEN_DEVICE_FUNC - void _solve_impl_transposed(const RhsType &rhs, DstType &dst) const; - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void computeInPlace(); - - MatrixType m_lu; - PermutationPType m_p; - PermutationQType m_q; - IntColVectorType m_rowsTranspositions; - IntRowVectorType m_colsTranspositions; - Index m_nonzero_pivots; - RealScalar m_l1_norm; - RealScalar m_maxpivot, m_prescribedThreshold; - signed char m_det_pq; - bool m_isInitialized, m_usePrescribedThreshold; -}; - -template -FullPivLU::FullPivLU() - : m_isInitialized(false), m_usePrescribedThreshold(false) -{ -} - -template -FullPivLU::FullPivLU(Index rows, Index cols) - : m_lu(rows, cols), - m_p(rows), - m_q(cols), - m_rowsTranspositions(rows), - m_colsTranspositions(cols), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ -} - -template -template -FullPivLU::FullPivLU(const EigenBase& matrix) - : m_lu(matrix.rows(), matrix.cols()), - m_p(matrix.rows()), - m_q(matrix.cols()), - m_rowsTranspositions(matrix.rows()), - m_colsTranspositions(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ - compute(matrix.derived()); -} - -template -template -FullPivLU::FullPivLU(EigenBase& matrix) - : m_lu(matrix.derived()), - m_p(matrix.rows()), - m_q(matrix.cols()), - m_rowsTranspositions(matrix.rows()), - m_colsTranspositions(matrix.cols()), - m_isInitialized(false), - m_usePrescribedThreshold(false) -{ - computeInPlace(); -} - -template -void FullPivLU::computeInPlace() -{ - check_template_parameters(); - - // the permutations are stored as int indices, so just to be sure: - eigen_assert(m_lu.rows()<=NumTraits::highest() && m_lu.cols()<=NumTraits::highest()); - - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); - - const Index size = m_lu.diagonalSize(); - const Index rows = m_lu.rows(); - const Index cols = m_lu.cols(); - - // will store the transpositions, before we accumulate them at the end. - // can't accumulate on-the-fly because that will be done in reverse order for the rows. - m_rowsTranspositions.resize(m_lu.rows()); - m_colsTranspositions.resize(m_lu.cols()); - Index number_of_transpositions = 0; // number of NONTRIVIAL transpositions, i.e. m_rowsTranspositions[i]!=i - - m_nonzero_pivots = size; // the generic case is that in which all pivots are nonzero (invertible case) - m_maxpivot = RealScalar(0); - - for(Index k = 0; k < size; ++k) - { - // First, we need to find the pivot. - - // biggest coefficient in the remaining bottom-right corner (starting at row k, col k) - Index row_of_biggest_in_corner, col_of_biggest_in_corner; - typedef internal::scalar_score_coeff_op Scoring; - typedef typename Scoring::result_type Score; - Score biggest_in_corner; - biggest_in_corner = m_lu.bottomRightCorner(rows-k, cols-k) - .unaryExpr(Scoring()) - .maxCoeff(&row_of_biggest_in_corner, &col_of_biggest_in_corner); - row_of_biggest_in_corner += k; // correct the values! since they were computed in the corner, - col_of_biggest_in_corner += k; // need to add k to them. - - if(biggest_in_corner==Score(0)) - { - // before exiting, make sure to initialize the still uninitialized transpositions - // in a sane state without destroying what we already have. - m_nonzero_pivots = k; - for(Index i = k; i < size; ++i) - { - m_rowsTranspositions.coeffRef(i) = i; - m_colsTranspositions.coeffRef(i) = i; - } - break; - } - - RealScalar abs_pivot = internal::abs_knowing_score()(m_lu(row_of_biggest_in_corner, col_of_biggest_in_corner), biggest_in_corner); - if(abs_pivot > m_maxpivot) m_maxpivot = abs_pivot; - - // Now that we've found the pivot, we need to apply the row/col swaps to - // bring it to the location (k,k). - - m_rowsTranspositions.coeffRef(k) = row_of_biggest_in_corner; - m_colsTranspositions.coeffRef(k) = col_of_biggest_in_corner; - if(k != row_of_biggest_in_corner) { - m_lu.row(k).swap(m_lu.row(row_of_biggest_in_corner)); - ++number_of_transpositions; - } - if(k != col_of_biggest_in_corner) { - m_lu.col(k).swap(m_lu.col(col_of_biggest_in_corner)); - ++number_of_transpositions; - } - - // Now that the pivot is at the right location, we update the remaining - // bottom-right corner by Gaussian elimination. - - if(k= 0; --k) - m_p.applyTranspositionOnTheRight(k, m_rowsTranspositions.coeff(k)); - - m_q.setIdentity(cols); - for(Index k = 0; k < size; ++k) - m_q.applyTranspositionOnTheRight(k, m_colsTranspositions.coeff(k)); - - m_det_pq = (number_of_transpositions%2) ? -1 : 1; - - m_isInitialized = true; -} - -template -typename internal::traits::Scalar FullPivLU::determinant() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - eigen_assert(m_lu.rows() == m_lu.cols() && "You can't take the determinant of a non-square matrix!"); - return Scalar(m_det_pq) * Scalar(m_lu.diagonal().prod()); -} - -/** \returns the matrix represented by the decomposition, - * i.e., it returns the product: \f$ P^{-1} L U Q^{-1} \f$. - * This function is provided for debug purposes. */ -template -MatrixType FullPivLU::reconstructedMatrix() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - const Index smalldim = (std::min)(m_lu.rows(), m_lu.cols()); - // LU - MatrixType res(m_lu.rows(),m_lu.cols()); - // FIXME the .toDenseMatrix() should not be needed... - res = m_lu.leftCols(smalldim) - .template triangularView().toDenseMatrix() - * m_lu.topRows(smalldim) - .template triangularView().toDenseMatrix(); - - // P^{-1}(LU) - res = m_p.inverse() * res; - - // (P^{-1}LU)Q^{-1} - res = res * m_q.inverse(); - - return res; -} - -/********* Implementation of kernel() **************************************************/ - -namespace internal { -template -struct kernel_retval > - : kernel_retval_base > -{ - EIGEN_MAKE_KERNEL_HELPERS(FullPivLU<_MatrixType>) - - enum { MaxSmallDimAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED( - MatrixType::MaxColsAtCompileTime, - MatrixType::MaxRowsAtCompileTime) - }; - - template void evalTo(Dest& dst) const - { - using std::abs; - const Index cols = dec().matrixLU().cols(), dimker = cols - rank(); - if(dimker == 0) - { - // The Kernel is just {0}, so it doesn't have a basis properly speaking, but let's - // avoid crashing/asserting as that depends on floating point calculations. Let's - // just return a single column vector filled with zeros. - dst.setZero(); - return; - } - - /* Let us use the following lemma: - * - * Lemma: If the matrix A has the LU decomposition PAQ = LU, - * then Ker A = Q(Ker U). - * - * Proof: trivial: just keep in mind that P, Q, L are invertible. - */ - - /* Thus, all we need to do is to compute Ker U, and then apply Q. - * - * U is upper triangular, with eigenvalues sorted so that any zeros appear at the end. - * Thus, the diagonal of U ends with exactly - * dimKer zero's. Let us use that to construct dimKer linearly - * independent vectors in Ker U. - */ - - Matrix pivots(rank()); - RealScalar premultiplied_threshold = dec().maxPivot() * dec().threshold(); - Index p = 0; - for(Index i = 0; i < dec().nonzeroPivots(); ++i) - if(abs(dec().matrixLU().coeff(i,i)) > premultiplied_threshold) - pivots.coeffRef(p++) = i; - eigen_internal_assert(p == rank()); - - // we construct a temporaty trapezoid matrix m, by taking the U matrix and - // permuting the rows and cols to bring the nonnegligible pivots to the top of - // the main diagonal. We need that to be able to apply our triangular solvers. - // FIXME when we get triangularView-for-rectangular-matrices, this can be simplified - Matrix - m(dec().matrixLU().block(0, 0, rank(), cols)); - for(Index i = 0; i < rank(); ++i) - { - if(i) m.row(i).head(i).setZero(); - m.row(i).tail(cols-i) = dec().matrixLU().row(pivots.coeff(i)).tail(cols-i); - } - m.block(0, 0, rank(), rank()); - m.block(0, 0, rank(), rank()).template triangularView().setZero(); - for(Index i = 0; i < rank(); ++i) - m.col(i).swap(m.col(pivots.coeff(i))); - - // ok, we have our trapezoid matrix, we can apply the triangular solver. - // notice that the math behind this suggests that we should apply this to the - // negative of the RHS, but for performance we just put the negative sign elsewhere, see below. - m.topLeftCorner(rank(), rank()) - .template triangularView().solveInPlace( - m.topRightCorner(rank(), dimker) - ); - - // now we must undo the column permutation that we had applied! - for(Index i = rank()-1; i >= 0; --i) - m.col(i).swap(m.col(pivots.coeff(i))); - - // see the negative sign in the next line, that's what we were talking about above. - for(Index i = 0; i < rank(); ++i) dst.row(dec().permutationQ().indices().coeff(i)) = -m.row(i).tail(dimker); - for(Index i = rank(); i < cols; ++i) dst.row(dec().permutationQ().indices().coeff(i)).setZero(); - for(Index k = 0; k < dimker; ++k) dst.coeffRef(dec().permutationQ().indices().coeff(rank()+k), k) = Scalar(1); - } -}; - -/***** Implementation of image() *****************************************************/ - -template -struct image_retval > - : image_retval_base > -{ - EIGEN_MAKE_IMAGE_HELPERS(FullPivLU<_MatrixType>) - - enum { MaxSmallDimAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED( - MatrixType::MaxColsAtCompileTime, - MatrixType::MaxRowsAtCompileTime) - }; - - template void evalTo(Dest& dst) const - { - using std::abs; - if(rank() == 0) - { - // The Image is just {0}, so it doesn't have a basis properly speaking, but let's - // avoid crashing/asserting as that depends on floating point calculations. Let's - // just return a single column vector filled with zeros. - dst.setZero(); - return; - } - - Matrix pivots(rank()); - RealScalar premultiplied_threshold = dec().maxPivot() * dec().threshold(); - Index p = 0; - for(Index i = 0; i < dec().nonzeroPivots(); ++i) - if(abs(dec().matrixLU().coeff(i,i)) > premultiplied_threshold) - pivots.coeffRef(p++) = i; - eigen_internal_assert(p == rank()); - - for(Index i = 0; i < rank(); ++i) - dst.col(i) = originalMatrix().col(dec().permutationQ().indices().coeff(pivots.coeff(i))); - } -}; - -/***** Implementation of solve() *****************************************************/ - -} // end namespace internal - -#ifndef EIGEN_PARSED_BY_DOXYGEN -template -template -void FullPivLU<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const -{ - /* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1}. - * So we proceed as follows: - * Step 1: compute c = P * rhs. - * Step 2: replace c by the solution x to Lx = c. Exists because L is invertible. - * Step 3: replace c by the solution x to Ux = c. May or may not exist. - * Step 4: result = Q * c; - */ - - const Index rows = this->rows(), - cols = this->cols(), - nonzero_pivots = this->rank(); - eigen_assert(rhs.rows() == rows); - const Index smalldim = (std::min)(rows, cols); - - if(nonzero_pivots == 0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs.rows(), rhs.cols()); - - // Step 1 - c = permutationP() * rhs; - - // Step 2 - m_lu.topLeftCorner(smalldim,smalldim) - .template triangularView() - .solveInPlace(c.topRows(smalldim)); - if(rows>cols) - c.bottomRows(rows-cols) -= m_lu.bottomRows(rows-cols) * c.topRows(cols); - - // Step 3 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .solveInPlace(c.topRows(nonzero_pivots)); - - // Step 4 - for(Index i = 0; i < nonzero_pivots; ++i) - dst.row(permutationQ().indices().coeff(i)) = c.row(i); - for(Index i = nonzero_pivots; i < m_lu.cols(); ++i) - dst.row(permutationQ().indices().coeff(i)).setZero(); -} - -template -template -void FullPivLU<_MatrixType>::_solve_impl_transposed(const RhsType &rhs, DstType &dst) const -{ - /* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1}, - * and since permutations are real and unitary, we can write this - * as A^T = Q U^T L^T P, - * So we proceed as follows: - * Step 1: compute c = Q^T rhs. - * Step 2: replace c by the solution x to U^T x = c. May or may not exist. - * Step 3: replace c by the solution x to L^T x = c. - * Step 4: result = P^T c. - * If Conjugate is true, replace "^T" by "^*" above. - */ - - const Index rows = this->rows(), cols = this->cols(), - nonzero_pivots = this->rank(); - eigen_assert(rhs.rows() == cols); - const Index smalldim = (std::min)(rows, cols); - - if(nonzero_pivots == 0) - { - dst.setZero(); - return; - } - - typename RhsType::PlainObject c(rhs.rows(), rhs.cols()); - - // Step 1 - c = permutationQ().inverse() * rhs; - - if (Conjugate) { - // Step 2 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .adjoint() - .solveInPlace(c.topRows(nonzero_pivots)); - // Step 3 - m_lu.topLeftCorner(smalldim, smalldim) - .template triangularView() - .adjoint() - .solveInPlace(c.topRows(smalldim)); - } else { - // Step 2 - m_lu.topLeftCorner(nonzero_pivots, nonzero_pivots) - .template triangularView() - .transpose() - .solveInPlace(c.topRows(nonzero_pivots)); - // Step 3 - m_lu.topLeftCorner(smalldim, smalldim) - .template triangularView() - .transpose() - .solveInPlace(c.topRows(smalldim)); - } - - // Step 4 - PermutationPType invp = permutationP().inverse().eval(); - for(Index i = 0; i < smalldim; ++i) - dst.row(invp.indices().coeff(i)) = c.row(i); - for(Index i = smalldim; i < rows; ++i) - dst.row(invp.indices().coeff(i)).setZero(); -} - -#endif - -namespace internal { - - -/***** Implementation of inverse() *****************************************************/ -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef FullPivLU LuType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; -} // end namespace internal - -/******* MatrixBase methods *****************************************************************/ - -/** \lu_module - * - * \return the full-pivoting LU decomposition of \c *this. - * - * \sa class FullPivLU - */ -template -inline const FullPivLU::PlainObject> -MatrixBase::fullPivLu() const -{ - return FullPivLU(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_LU_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h deleted file mode 100644 index f49f2336004..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/InverseImpl.h +++ /dev/null @@ -1,415 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Benoit Jacob -// Copyright (C) 2014 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_INVERSE_IMPL_H -#define EIGEN_INVERSE_IMPL_H - -namespace Eigen { - -namespace internal { - -/********************************** -*** General case implementation *** -**********************************/ - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - result = matrix.partialPivLu().inverse(); - } -}; - -template -struct compute_inverse_and_det_with_check { /* nothing! general case not supported. */ }; - -/**************************** -*** Size 1 implementation *** -****************************/ - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename MatrixType::Scalar Scalar; - internal::evaluator matrixEval(matrix); - result.coeffRef(0,0) = Scalar(1) / matrixEval.coeff(0,0); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& result, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - determinant = matrix.coeff(0,0); - invertible = abs(determinant) > absDeterminantThreshold; - if(invertible) result.coeffRef(0,0) = typename ResultType::Scalar(1) / determinant; - } -}; - -/**************************** -*** Size 2 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline void compute_inverse_size2_helper( - const MatrixType& matrix, const typename ResultType::Scalar& invdet, - ResultType& result) -{ - result.coeffRef(0,0) = matrix.coeff(1,1) * invdet; - result.coeffRef(1,0) = -matrix.coeff(1,0) * invdet; - result.coeffRef(0,1) = -matrix.coeff(0,1) * invdet; - result.coeffRef(1,1) = matrix.coeff(0,0) * invdet; -} - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename ResultType::Scalar Scalar; - const Scalar invdet = typename MatrixType::Scalar(1) / matrix.determinant(); - compute_inverse_size2_helper(matrix, invdet, result); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - typedef typename ResultType::Scalar Scalar; - determinant = matrix.determinant(); - invertible = abs(determinant) > absDeterminantThreshold; - if(!invertible) return; - const Scalar invdet = Scalar(1) / determinant; - compute_inverse_size2_helper(matrix, invdet, inverse); - } -}; - -/**************************** -*** Size 3 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline typename MatrixType::Scalar cofactor_3x3(const MatrixType& m) -{ - enum { - i1 = (i+1) % 3, - i2 = (i+2) % 3, - j1 = (j+1) % 3, - j2 = (j+2) % 3 - }; - return m.coeff(i1, j1) * m.coeff(i2, j2) - - m.coeff(i1, j2) * m.coeff(i2, j1); -} - -template -EIGEN_DEVICE_FUNC -inline void compute_inverse_size3_helper( - const MatrixType& matrix, - const typename ResultType::Scalar& invdet, - const Matrix& cofactors_col0, - ResultType& result) -{ - result.row(0) = cofactors_col0 * invdet; - result.coeffRef(1,0) = cofactor_3x3(matrix) * invdet; - result.coeffRef(1,1) = cofactor_3x3(matrix) * invdet; - result.coeffRef(1,2) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,0) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,1) = cofactor_3x3(matrix) * invdet; - result.coeffRef(2,2) = cofactor_3x3(matrix) * invdet; -} - -template -struct compute_inverse -{ - EIGEN_DEVICE_FUNC - static inline void run(const MatrixType& matrix, ResultType& result) - { - typedef typename ResultType::Scalar Scalar; - Matrix cofactors_col0; - cofactors_col0.coeffRef(0) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(1) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(2) = cofactor_3x3(matrix); - const Scalar det = (cofactors_col0.cwiseProduct(matrix.col(0))).sum(); - const Scalar invdet = Scalar(1) / det; - compute_inverse_size3_helper(matrix, invdet, cofactors_col0, result); - } -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - typedef typename ResultType::Scalar Scalar; - Matrix cofactors_col0; - cofactors_col0.coeffRef(0) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(1) = cofactor_3x3(matrix); - cofactors_col0.coeffRef(2) = cofactor_3x3(matrix); - determinant = (cofactors_col0.cwiseProduct(matrix.col(0))).sum(); - invertible = abs(determinant) > absDeterminantThreshold; - if(!invertible) return; - const Scalar invdet = Scalar(1) / determinant; - compute_inverse_size3_helper(matrix, invdet, cofactors_col0, inverse); - } -}; - -/**************************** -*** Size 4 implementation *** -****************************/ - -template -EIGEN_DEVICE_FUNC -inline const typename Derived::Scalar general_det3_helper -(const MatrixBase& matrix, int i1, int i2, int i3, int j1, int j2, int j3) -{ - return matrix.coeff(i1,j1) - * (matrix.coeff(i2,j2) * matrix.coeff(i3,j3) - matrix.coeff(i2,j3) * matrix.coeff(i3,j2)); -} - -template -EIGEN_DEVICE_FUNC -inline typename MatrixType::Scalar cofactor_4x4(const MatrixType& matrix) -{ - enum { - i1 = (i+1) % 4, - i2 = (i+2) % 4, - i3 = (i+3) % 4, - j1 = (j+1) % 4, - j2 = (j+2) % 4, - j3 = (j+3) % 4 - }; - return general_det3_helper(matrix, i1, i2, i3, j1, j2, j3) - + general_det3_helper(matrix, i2, i3, i1, j1, j2, j3) - + general_det3_helper(matrix, i3, i1, i2, j1, j2, j3); -} - -template -struct compute_inverse_size4 -{ - EIGEN_DEVICE_FUNC - static void run(const MatrixType& matrix, ResultType& result) - { - result.coeffRef(0,0) = cofactor_4x4(matrix); - result.coeffRef(1,0) = -cofactor_4x4(matrix); - result.coeffRef(2,0) = cofactor_4x4(matrix); - result.coeffRef(3,0) = -cofactor_4x4(matrix); - result.coeffRef(0,2) = cofactor_4x4(matrix); - result.coeffRef(1,2) = -cofactor_4x4(matrix); - result.coeffRef(2,2) = cofactor_4x4(matrix); - result.coeffRef(3,2) = -cofactor_4x4(matrix); - result.coeffRef(0,1) = -cofactor_4x4(matrix); - result.coeffRef(1,1) = cofactor_4x4(matrix); - result.coeffRef(2,1) = -cofactor_4x4(matrix); - result.coeffRef(3,1) = cofactor_4x4(matrix); - result.coeffRef(0,3) = -cofactor_4x4(matrix); - result.coeffRef(1,3) = cofactor_4x4(matrix); - result.coeffRef(2,3) = -cofactor_4x4(matrix); - result.coeffRef(3,3) = cofactor_4x4(matrix); - result /= (matrix.col(0).cwiseProduct(result.row(0).transpose())).sum(); - } -}; - -template -struct compute_inverse - : compute_inverse_size4 -{ -}; - -template -struct compute_inverse_and_det_with_check -{ - EIGEN_DEVICE_FUNC - static inline void run( - const MatrixType& matrix, - const typename MatrixType::RealScalar& absDeterminantThreshold, - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible - ) - { - using std::abs; - determinant = matrix.determinant(); - invertible = abs(determinant) > absDeterminantThreshold; - if(invertible) compute_inverse::run(matrix, inverse); - } -}; - -/************************* -*** MatrixBase methods *** -*************************/ - -} // end namespace internal - -namespace internal { - -// Specialization for "dense = dense_xpr.inverse()" -template -struct Assignment, internal::assign_op, Dense2Dense> -{ - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - Index dstRows = src.rows(); - Index dstCols = src.cols(); - if((dst.rows()!=dstRows) || (dst.cols()!=dstCols)) - dst.resize(dstRows, dstCols); - - const int Size = EIGEN_PLAIN_ENUM_MIN(XprType::ColsAtCompileTime,DstXprType::ColsAtCompileTime); - EIGEN_ONLY_USED_FOR_DEBUG(Size); - eigen_assert(( (Size<=1) || (Size>4) || (extract_data(src.nestedExpression())!=extract_data(dst))) - && "Aliasing problem detected in inverse(), you need to do inverse().eval() here."); - - typedef typename internal::nested_eval::type ActualXprType; - typedef typename internal::remove_all::type ActualXprTypeCleanded; - - ActualXprType actual_xpr(src.nestedExpression()); - - compute_inverse::run(actual_xpr, dst); - } -}; - - -} // end namespace internal - -/** \lu_module - * - * \returns the matrix inverse of this matrix. - * - * For small fixed sizes up to 4x4, this method uses cofactors. - * In the general case, this method uses class PartialPivLU. - * - * \note This matrix must be invertible, otherwise the result is undefined. If you need an - * invertibility check, do the following: - * \li for fixed sizes up to 4x4, use computeInverseAndDetWithCheck(). - * \li for the general case, use class FullPivLU. - * - * Example: \include MatrixBase_inverse.cpp - * Output: \verbinclude MatrixBase_inverse.out - * - * \sa computeInverseAndDetWithCheck() - */ -template -inline const Inverse MatrixBase::inverse() const -{ - EIGEN_STATIC_ASSERT(!NumTraits::IsInteger,THIS_FUNCTION_IS_NOT_FOR_INTEGER_NUMERIC_TYPES) - eigen_assert(rows() == cols()); - return Inverse(derived()); -} - -/** \lu_module - * - * Computation of matrix inverse and determinant, with invertibility check. - * - * This is only for fixed-size square matrices of size up to 4x4. - * - * \param inverse Reference to the matrix in which to store the inverse. - * \param determinant Reference to the variable in which to store the determinant. - * \param invertible Reference to the bool variable in which to store whether the matrix is invertible. - * \param absDeterminantThreshold Optional parameter controlling the invertibility check. - * The matrix will be declared invertible if the absolute value of its - * determinant is greater than this threshold. - * - * Example: \include MatrixBase_computeInverseAndDetWithCheck.cpp - * Output: \verbinclude MatrixBase_computeInverseAndDetWithCheck.out - * - * \sa inverse(), computeInverseWithCheck() - */ -template -template -inline void MatrixBase::computeInverseAndDetWithCheck( - ResultType& inverse, - typename ResultType::Scalar& determinant, - bool& invertible, - const RealScalar& absDeterminantThreshold - ) const -{ - // i'd love to put some static assertions there, but SFINAE means that they have no effect... - eigen_assert(rows() == cols()); - // for 2x2, it's worth giving a chance to avoid evaluating. - // for larger sizes, evaluating has negligible cost and limits code size. - typedef typename internal::conditional< - RowsAtCompileTime == 2, - typename internal::remove_all::type>::type, - PlainObject - >::type MatrixType; - internal::compute_inverse_and_det_with_check::run - (derived(), absDeterminantThreshold, inverse, determinant, invertible); -} - -/** \lu_module - * - * Computation of matrix inverse, with invertibility check. - * - * This is only for fixed-size square matrices of size up to 4x4. - * - * \param inverse Reference to the matrix in which to store the inverse. - * \param invertible Reference to the bool variable in which to store whether the matrix is invertible. - * \param absDeterminantThreshold Optional parameter controlling the invertibility check. - * The matrix will be declared invertible if the absolute value of its - * determinant is greater than this threshold. - * - * Example: \include MatrixBase_computeInverseWithCheck.cpp - * Output: \verbinclude MatrixBase_computeInverseWithCheck.out - * - * \sa inverse(), computeInverseAndDetWithCheck() - */ -template -template -inline void MatrixBase::computeInverseWithCheck( - ResultType& inverse, - bool& invertible, - const RealScalar& absDeterminantThreshold - ) const -{ - Scalar determinant; - // i'd love to put some static assertions there, but SFINAE means that they have no effect... - eigen_assert(rows() == cols()); - computeInverseAndDetWithCheck(inverse,determinant,invertible,absDeterminantThreshold); -} - -} // end namespace Eigen - -#endif // EIGEN_INVERSE_IMPL_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h deleted file mode 100644 index 6b10f39fab4..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU.h +++ /dev/null @@ -1,614 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2006-2009 Benoit Jacob -// Copyright (C) 2009 Gael Guennebaud -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARTIALLU_H -#define EIGEN_PARTIALLU_H - -namespace Eigen { - -namespace internal { -template struct traits > - : traits<_MatrixType> -{ - typedef MatrixXpr XprKind; - typedef SolverStorage StorageKind; - typedef traits<_MatrixType> BaseTraits; - enum { - Flags = BaseTraits::Flags & RowMajorBit, - CoeffReadCost = Dynamic - }; -}; - -template -struct enable_if_ref; -// { -// typedef Derived type; -// }; - -template -struct enable_if_ref,Derived> { - typedef Derived type; -}; - -} // end namespace internal - -/** \ingroup LU_Module - * - * \class PartialPivLU - * - * \brief LU decomposition of a matrix with partial pivoting, and related features - * - * \tparam _MatrixType the type of the matrix of which we are computing the LU decomposition - * - * This class represents a LU decomposition of a \b square \b invertible matrix, with partial pivoting: the matrix A - * is decomposed as A = PLU where L is unit-lower-triangular, U is upper-triangular, and P - * is a permutation matrix. - * - * Typically, partial pivoting LU decomposition is only considered numerically stable for square invertible - * matrices. Thus LAPACK's dgesv and dgesvx require the matrix to be square and invertible. The present class - * does the same. It will assert that the matrix is square, but it won't (actually it can't) check that the - * matrix is invertible: it is your task to check that you only use this decomposition on invertible matrices. - * - * The guaranteed safe alternative, working for all matrices, is the full pivoting LU decomposition, provided - * by class FullPivLU. - * - * This is \b not a rank-revealing LU decomposition. Many features are intentionally absent from this class, - * such as rank computation. If you need these features, use class FullPivLU. - * - * This LU decomposition is suitable to invert invertible matrices. It is what MatrixBase::inverse() uses - * in the general case. - * On the other hand, it is \b not suitable to determine whether a given matrix is invertible. - * - * The data of the LU decomposition can be directly accessed through the methods matrixLU(), permutationP(). - * - * This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism. - * - * \sa MatrixBase::partialPivLu(), MatrixBase::determinant(), MatrixBase::inverse(), MatrixBase::computeInverse(), class FullPivLU - */ -template class PartialPivLU - : public SolverBase > -{ - public: - - typedef _MatrixType MatrixType; - typedef SolverBase Base; - EIGEN_GENERIC_PUBLIC_INTERFACE(PartialPivLU) - // FIXME StorageIndex defined in EIGEN_GENERIC_PUBLIC_INTERFACE should be int - enum { - MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime, - MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime - }; - typedef PermutationMatrix PermutationType; - typedef Transpositions TranspositionType; - typedef typename MatrixType::PlainObject PlainObject; - - /** - * \brief Default Constructor. - * - * The default constructor is useful in cases in which the user intends to - * perform decompositions via PartialPivLU::compute(const MatrixType&). - */ - PartialPivLU(); - - /** \brief Default Constructor with memory preallocation - * - * Like the default constructor but with preallocation of the internal data - * according to the specified problem \a size. - * \sa PartialPivLU() - */ - explicit PartialPivLU(Index size); - - /** Constructor. - * - * \param matrix the matrix of which to compute the LU decomposition. - * - * \warning The matrix should have full rank (e.g. if it's square, it should be invertible). - * If you need to deal with non-full rank, use class FullPivLU instead. - */ - template - explicit PartialPivLU(const EigenBase& matrix); - - /** Constructor for \link InplaceDecomposition inplace decomposition \endlink - * - * \param matrix the matrix of which to compute the LU decomposition. - * - * \warning The matrix should have full rank (e.g. if it's square, it should be invertible). - * If you need to deal with non-full rank, use class FullPivLU instead. - */ - template - explicit PartialPivLU(EigenBase& matrix); - - template - PartialPivLU& compute(const EigenBase& matrix) { - m_lu = matrix.derived(); - compute(); - return *this; - } - - /** \returns the LU decomposition matrix: the upper-triangular part is U, the - * unit-lower-triangular part is L (at least for square matrices; in the non-square - * case, special care is needed, see the documentation of class FullPivLU). - * - * \sa matrixL(), matrixU() - */ - inline const MatrixType& matrixLU() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return m_lu; - } - - /** \returns the permutation matrix P. - */ - inline const PermutationType& permutationP() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return m_p; - } - - /** This method returns the solution x to the equation Ax=b, where A is the matrix of which - * *this is the LU decomposition. - * - * \param b the right-hand-side of the equation to solve. Can be a vector or a matrix, - * the only requirement in order for the equation to make sense is that - * b.rows()==A.rows(), where A is the matrix of which *this is the LU decomposition. - * - * \returns the solution. - * - * Example: \include PartialPivLU_solve.cpp - * Output: \verbinclude PartialPivLU_solve.out - * - * Since this PartialPivLU class assumes anyway that the matrix A is invertible, the solution - * theoretically exists and is unique regardless of b. - * - * \sa TriangularView::solve(), inverse(), computeInverse() - */ - // FIXME this is a copy-paste of the base-class member to add the isInitialized assertion. - template - inline const Solve - solve(const MatrixBase& b) const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Solve(*this, b.derived()); - } - - /** \returns an estimate of the reciprocal condition number of the matrix of which \c *this is - the LU decomposition. - */ - inline RealScalar rcond() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return internal::rcond_estimate_helper(m_l1_norm, *this); - } - - /** \returns the inverse of the matrix of which *this is the LU decomposition. - * - * \warning The matrix being decomposed here is assumed to be invertible. If you need to check for - * invertibility, use class FullPivLU instead. - * - * \sa MatrixBase::inverse(), LU::inverse() - */ - inline const Inverse inverse() const - { - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Inverse(*this); - } - - /** \returns the determinant of the matrix of which - * *this is the LU decomposition. It has only linear complexity - * (that is, O(n) where n is the dimension of the square matrix) - * as the LU decomposition has already been computed. - * - * \note For fixed-size matrices of size up to 4, MatrixBase::determinant() offers - * optimized paths. - * - * \warning a determinant can be very big or small, so for matrices - * of large enough dimension, there is a risk of overflow/underflow. - * - * \sa MatrixBase::determinant() - */ - Scalar determinant() const; - - MatrixType reconstructedMatrix() const; - - inline Index rows() const { return m_lu.rows(); } - inline Index cols() const { return m_lu.cols(); } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - template - EIGEN_DEVICE_FUNC - void _solve_impl(const RhsType &rhs, DstType &dst) const { - /* The decomposition PA = LU can be rewritten as A = P^{-1} L U. - * So we proceed as follows: - * Step 1: compute c = Pb. - * Step 2: replace c by the solution x to Lx = c. - * Step 3: replace c by the solution x to Ux = c. - */ - - eigen_assert(rhs.rows() == m_lu.rows()); - - // Step 1 - dst = permutationP() * rhs; - - // Step 2 - m_lu.template triangularView().solveInPlace(dst); - - // Step 3 - m_lu.template triangularView().solveInPlace(dst); - } - - template - EIGEN_DEVICE_FUNC - void _solve_impl_transposed(const RhsType &rhs, DstType &dst) const { - /* The decomposition PA = LU can be rewritten as A = P^{-1} L U. - * So we proceed as follows: - * Step 1: compute c = Pb. - * Step 2: replace c by the solution x to Lx = c. - * Step 3: replace c by the solution x to Ux = c. - */ - - eigen_assert(rhs.rows() == m_lu.cols()); - - if (Conjugate) { - // Step 1 - dst = m_lu.template triangularView().adjoint().solve(rhs); - // Step 2 - m_lu.template triangularView().adjoint().solveInPlace(dst); - } else { - // Step 1 - dst = m_lu.template triangularView().transpose().solve(rhs); - // Step 2 - m_lu.template triangularView().transpose().solveInPlace(dst); - } - // Step 3 - dst = permutationP().transpose() * dst; - } - #endif - - protected: - - static void check_template_parameters() - { - EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); - } - - void compute(); - - MatrixType m_lu; - PermutationType m_p; - TranspositionType m_rowsTranspositions; - RealScalar m_l1_norm; - signed char m_det_p; - bool m_isInitialized; -}; - -template -PartialPivLU::PartialPivLU() - : m_lu(), - m_p(), - m_rowsTranspositions(), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ -} - -template -PartialPivLU::PartialPivLU(Index size) - : m_lu(size, size), - m_p(size), - m_rowsTranspositions(size), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ -} - -template -template -PartialPivLU::PartialPivLU(const EigenBase& matrix) - : m_lu(matrix.rows(),matrix.cols()), - m_p(matrix.rows()), - m_rowsTranspositions(matrix.rows()), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ - compute(matrix.derived()); -} - -template -template -PartialPivLU::PartialPivLU(EigenBase& matrix) - : m_lu(matrix.derived()), - m_p(matrix.rows()), - m_rowsTranspositions(matrix.rows()), - m_l1_norm(0), - m_det_p(0), - m_isInitialized(false) -{ - compute(); -} - -namespace internal { - -/** \internal This is the blocked version of fullpivlu_unblocked() */ -template -struct partial_lu_impl -{ - // FIXME add a stride to Map, so that the following mapping becomes easier, - // another option would be to create an expression being able to automatically - // warp any Map, Matrix, and Block expressions as a unique type, but since that's exactly - // a Map + stride, why not adding a stride to Map, and convenient ctors from a Matrix, - // and Block. - typedef Map > MapLU; - typedef Block MatrixType; - typedef Block BlockType; - typedef typename MatrixType::RealScalar RealScalar; - - /** \internal performs the LU decomposition in-place of the matrix \a lu - * using an unblocked algorithm. - * - * In addition, this function returns the row transpositions in the - * vector \a row_transpositions which must have a size equal to the number - * of columns of the matrix \a lu, and an integer \a nb_transpositions - * which returns the actual number of transpositions. - * - * \returns The index of the first pivot which is exactly zero if any, or a negative number otherwise. - */ - static Index unblocked_lu(MatrixType& lu, PivIndex* row_transpositions, PivIndex& nb_transpositions) - { - typedef scalar_score_coeff_op Scoring; - typedef typename Scoring::result_type Score; - const Index rows = lu.rows(); - const Index cols = lu.cols(); - const Index size = (std::min)(rows,cols); - nb_transpositions = 0; - Index first_zero_pivot = -1; - for(Index k = 0; k < size; ++k) - { - Index rrows = rows-k-1; - Index rcols = cols-k-1; - - Index row_of_biggest_in_col; - Score biggest_in_corner - = lu.col(k).tail(rows-k).unaryExpr(Scoring()).maxCoeff(&row_of_biggest_in_col); - row_of_biggest_in_col += k; - - row_transpositions[k] = PivIndex(row_of_biggest_in_col); - - if(biggest_in_corner != Score(0)) - { - if(k != row_of_biggest_in_col) - { - lu.row(k).swap(lu.row(row_of_biggest_in_col)); - ++nb_transpositions; - } - - // FIXME shall we introduce a safe quotient expression in cas 1/lu.coeff(k,k) - // overflow but not the actual quotient? - lu.col(k).tail(rrows) /= lu.coeff(k,k); - } - else if(first_zero_pivot==-1) - { - // the pivot is exactly zero, we record the index of the first pivot which is exactly 0, - // and continue the factorization such we still have A = PLU - first_zero_pivot = k; - } - - if(k > > - */ - static Index blocked_lu(Index rows, Index cols, Scalar* lu_data, Index luStride, PivIndex* row_transpositions, PivIndex& nb_transpositions, Index maxBlockSize=256) - { - MapLU lu1(lu_data,StorageOrder==RowMajor?rows:luStride,StorageOrder==RowMajor?luStride:cols); - MatrixType lu(lu1,0,0,rows,cols); - - const Index size = (std::min)(rows,cols); - - // if the matrix is too small, no blocking: - if(size<=16) - { - return unblocked_lu(lu, row_transpositions, nb_transpositions); - } - - // automatically adjust the number of subdivisions to the size - // of the matrix so that there is enough sub blocks: - Index blockSize; - { - blockSize = size/8; - blockSize = (blockSize/16)*16; - blockSize = (std::min)((std::max)(blockSize,Index(8)), maxBlockSize); - } - - nb_transpositions = 0; - Index first_zero_pivot = -1; - for(Index k = 0; k < size; k+=blockSize) - { - Index bs = (std::min)(size-k,blockSize); // actual size of the block - Index trows = rows - k - bs; // trailing rows - Index tsize = size - k - bs; // trailing size - - // partition the matrix: - // A00 | A01 | A02 - // lu = A_0 | A_1 | A_2 = A10 | A11 | A12 - // A20 | A21 | A22 - BlockType A_0(lu,0,0,rows,k); - BlockType A_2(lu,0,k+bs,rows,tsize); - BlockType A11(lu,k,k,bs,bs); - BlockType A12(lu,k,k+bs,bs,tsize); - BlockType A21(lu,k+bs,k,trows,bs); - BlockType A22(lu,k+bs,k+bs,trows,tsize); - - PivIndex nb_transpositions_in_panel; - // recursively call the blocked LU algorithm on [A11^T A21^T]^T - // with a very small blocking size: - Index ret = blocked_lu(trows+bs, bs, &lu.coeffRef(k,k), luStride, - row_transpositions+k, nb_transpositions_in_panel, 16); - if(ret>=0 && first_zero_pivot==-1) - first_zero_pivot = k+ret; - - nb_transpositions += nb_transpositions_in_panel; - // update permutations and apply them to A_0 - for(Index i=k; i(k)); - A_0.row(i).swap(A_0.row(piv)); - } - - if(trows) - { - // apply permutations to A_2 - for(Index i=k;i().solveInPlace(A12); - - A22.noalias() -= A21 * A12; - } - } - return first_zero_pivot; - } -}; - -/** \internal performs the LU decomposition with partial pivoting in-place. - */ -template -void partial_lu_inplace(MatrixType& lu, TranspositionType& row_transpositions, typename TranspositionType::StorageIndex& nb_transpositions) -{ - eigen_assert(lu.cols() == row_transpositions.size()); - eigen_assert((&row_transpositions.coeffRef(1)-&row_transpositions.coeffRef(0)) == 1); - - partial_lu_impl - - ::blocked_lu(lu.rows(), lu.cols(), &lu.coeffRef(0,0), lu.outerStride(), &row_transpositions.coeffRef(0), nb_transpositions); -} - -} // end namespace internal - -template -void PartialPivLU::compute() -{ - check_template_parameters(); - - // the row permutation is stored as int indices, so just to be sure: - eigen_assert(m_lu.rows()::highest()); - - if(m_lu.cols()>0) - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); - else - m_l1_norm = RealScalar(0); - - eigen_assert(m_lu.rows() == m_lu.cols() && "PartialPivLU is only for square (and moreover invertible) matrices"); - const Index size = m_lu.rows(); - - m_rowsTranspositions.resize(size); - - typename TranspositionType::StorageIndex nb_transpositions; - internal::partial_lu_inplace(m_lu, m_rowsTranspositions, nb_transpositions); - m_det_p = (nb_transpositions%2) ? -1 : 1; - - m_p = m_rowsTranspositions; - - m_isInitialized = true; -} - -template -typename PartialPivLU::Scalar PartialPivLU::determinant() const -{ - eigen_assert(m_isInitialized && "PartialPivLU is not initialized."); - return Scalar(m_det_p) * m_lu.diagonal().prod(); -} - -/** \returns the matrix represented by the decomposition, - * i.e., it returns the product: P^{-1} L U. - * This function is provided for debug purpose. */ -template -MatrixType PartialPivLU::reconstructedMatrix() const -{ - eigen_assert(m_isInitialized && "LU is not initialized."); - // LU - MatrixType res = m_lu.template triangularView().toDenseMatrix() - * m_lu.template triangularView(); - - // P^{-1}(LU) - res = m_p.inverse() * res; - - return res; -} - -/***** Implementation details *****************************************************/ - -namespace internal { - -/***** Implementation of inverse() *****************************************************/ -template -struct Assignment >, internal::assign_op::Scalar>, Dense2Dense> -{ - typedef PartialPivLU LuType; - typedef Inverse SrcXprType; - static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op &) - { - dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols())); - } -}; -} // end namespace internal - -/******** MatrixBase methods *******/ - -/** \lu_module - * - * \return the partial-pivoting LU decomposition of \c *this. - * - * \sa class PartialPivLU - */ -template -inline const PartialPivLU::PlainObject> -MatrixBase::partialPivLu() const -{ - return PartialPivLU(eval()); -} - -/** \lu_module - * - * Synonym of partialPivLu(). - * - * \return the partial-pivoting LU decomposition of \c *this. - * - * \sa class PartialPivLU - */ -template -inline const PartialPivLU::PlainObject> -MatrixBase::lu() const -{ - return PartialPivLU(eval()); -} - -} // end namespace Eigen - -#endif // EIGEN_PARTIALLU_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h deleted file mode 100644 index 755168a9460..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/PartialPivLU_LAPACKE.h +++ /dev/null @@ -1,83 +0,0 @@ -/* - Copyright (c) 2011, Intel Corporation. All rights reserved. - - Redistribution and use in source and binary forms, with or without modification, - are permitted provided that the following conditions are met: - - * Redistributions of source code must retain the above copyright notice, this - list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above copyright notice, - this list of conditions and the following disclaimer in the documentation - and/or other materials provided with the distribution. - * Neither the name of Intel Corporation nor the names of its contributors may - be used to endorse or promote products derived from this software without - specific prior written permission. - - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND - ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED - WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE - DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR - ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES - (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; - LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON - ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - - ******************************************************************************** - * Content : Eigen bindings to LAPACKe - * LU decomposition with partial pivoting based on LAPACKE_?getrf function. - ******************************************************************************** -*/ - -#ifndef EIGEN_PARTIALLU_LAPACK_H -#define EIGEN_PARTIALLU_LAPACK_H - -namespace Eigen { - -namespace internal { - -/** \internal Specialization for the data types supported by LAPACKe */ - -#define EIGEN_LAPACKE_LU_PARTPIV(EIGTYPE, LAPACKE_TYPE, LAPACKE_PREFIX) \ -template \ -struct partial_lu_impl \ -{ \ - /* \internal performs the LU decomposition in-place of the matrix represented */ \ - static lapack_int blocked_lu(Index rows, Index cols, EIGTYPE* lu_data, Index luStride, lapack_int* row_transpositions, lapack_int& nb_transpositions, lapack_int maxBlockSize=256) \ - { \ - EIGEN_UNUSED_VARIABLE(maxBlockSize);\ - lapack_int matrix_order, first_zero_pivot; \ - lapack_int m, n, lda, *ipiv, info; \ - EIGTYPE* a; \ -/* Set up parameters for ?getrf */ \ - matrix_order = StorageOrder==RowMajor ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \ - lda = convert_index(luStride); \ - a = lu_data; \ - ipiv = row_transpositions; \ - m = convert_index(rows); \ - n = convert_index(cols); \ - nb_transpositions = 0; \ -\ - info = LAPACKE_##LAPACKE_PREFIX##getrf( matrix_order, m, n, (LAPACKE_TYPE*)a, lda, ipiv ); \ -\ - for(int i=0;i= 0); \ -/* something should be done with nb_transpositions */ \ -\ - first_zero_pivot = info; \ - return first_zero_pivot; \ - } \ -}; - -EIGEN_LAPACKE_LU_PARTPIV(double, double, d) -EIGEN_LAPACKE_LU_PARTPIV(float, float, s) -EIGEN_LAPACKE_LU_PARTPIV(dcomplex, lapack_complex_double, z) -EIGEN_LAPACKE_LU_PARTPIV(scomplex, lapack_complex_float, c) - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_PARTIALLU_LAPACK_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h deleted file mode 100644 index 4dce2ef20ee..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/LU/arch/Inverse_SSE.h +++ /dev/null @@ -1,338 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2001 Intel Corporation -// Copyright (C) 2010 Gael Guennebaud -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// The SSE code for the 4x4 float and double matrix inverse in this file -// comes from the following Intel's library: -// http://software.intel.com/en-us/articles/optimized-matrix-library-for-use-with-the-intel-pentiumr-4-processors-sse2-instructions/ -// -// Here is the respective copyright and license statement: -// -// Copyright (c) 2001 Intel Corporation. -// -// Permition is granted to use, copy, distribute and prepare derivative works -// of this library for any purpose and without fee, provided, that the above -// copyright notice and this statement appear in all copies. -// Intel makes no representations about the suitability of this software for -// any purpose, and specifically disclaims all warranties. -// See LEGAL.TXT for all the legal information. - -#ifndef EIGEN_INVERSE_SSE_H -#define EIGEN_INVERSE_SSE_H - -namespace Eigen { - -namespace internal { - -template -struct compute_inverse_size4 -{ - enum { - MatrixAlignment = traits::Alignment, - ResultAlignment = traits::Alignment, - StorageOrdersMatch = (MatrixType::Flags&RowMajorBit) == (ResultType::Flags&RowMajorBit) - }; - typedef typename conditional<(MatrixType::Flags&LinearAccessBit),MatrixType const &,typename MatrixType::PlainObject>::type ActualMatrixType; - - static void run(const MatrixType& mat, ResultType& result) - { - ActualMatrixType matrix(mat); - const Packet4f p4f_sign_PNNP = _mm_castsi128_ps(_mm_set_epi32(0x00000000, 0x80000000, 0x80000000, 0x00000000)); - - // Load the full matrix into registers - __m128 _L1 = matrix.template packet( 0); - __m128 _L2 = matrix.template packet( 4); - __m128 _L3 = matrix.template packet( 8); - __m128 _L4 = matrix.template packet(12); - - // The inverse is calculated using "Divide and Conquer" technique. The - // original matrix is divide into four 2x2 sub-matrices. Since each - // register holds four matrix element, the smaller matrices are - // represented as a registers. Hence we get a better locality of the - // calculations. - - __m128 A, B, C, D; // the four sub-matrices - if(!StorageOrdersMatch) - { - A = _mm_unpacklo_ps(_L1, _L2); - B = _mm_unpacklo_ps(_L3, _L4); - C = _mm_unpackhi_ps(_L1, _L2); - D = _mm_unpackhi_ps(_L3, _L4); - } - else - { - A = _mm_movelh_ps(_L1, _L2); - B = _mm_movehl_ps(_L2, _L1); - C = _mm_movelh_ps(_L3, _L4); - D = _mm_movehl_ps(_L4, _L3); - } - - __m128 iA, iB, iC, iD, // partial inverse of the sub-matrices - DC, AB; - __m128 dA, dB, dC, dD; // determinant of the sub-matrices - __m128 det, d, d1, d2; - __m128 rd; // reciprocal of the determinant - - // AB = A# * B - AB = _mm_mul_ps(_mm_shuffle_ps(A,A,0x0F), B); - AB = _mm_sub_ps(AB,_mm_mul_ps(_mm_shuffle_ps(A,A,0xA5), _mm_shuffle_ps(B,B,0x4E))); - // DC = D# * C - DC = _mm_mul_ps(_mm_shuffle_ps(D,D,0x0F), C); - DC = _mm_sub_ps(DC,_mm_mul_ps(_mm_shuffle_ps(D,D,0xA5), _mm_shuffle_ps(C,C,0x4E))); - - // dA = |A| - dA = _mm_mul_ps(_mm_shuffle_ps(A, A, 0x5F),A); - dA = _mm_sub_ss(dA, _mm_movehl_ps(dA,dA)); - // dB = |B| - dB = _mm_mul_ps(_mm_shuffle_ps(B, B, 0x5F),B); - dB = _mm_sub_ss(dB, _mm_movehl_ps(dB,dB)); - - // dC = |C| - dC = _mm_mul_ps(_mm_shuffle_ps(C, C, 0x5F),C); - dC = _mm_sub_ss(dC, _mm_movehl_ps(dC,dC)); - // dD = |D| - dD = _mm_mul_ps(_mm_shuffle_ps(D, D, 0x5F),D); - dD = _mm_sub_ss(dD, _mm_movehl_ps(dD,dD)); - - // d = trace(AB*DC) = trace(A#*B*D#*C) - d = _mm_mul_ps(_mm_shuffle_ps(DC,DC,0xD8),AB); - - // iD = C*A#*B - iD = _mm_mul_ps(_mm_shuffle_ps(C,C,0xA0), _mm_movelh_ps(AB,AB)); - iD = _mm_add_ps(iD,_mm_mul_ps(_mm_shuffle_ps(C,C,0xF5), _mm_movehl_ps(AB,AB))); - // iA = B*D#*C - iA = _mm_mul_ps(_mm_shuffle_ps(B,B,0xA0), _mm_movelh_ps(DC,DC)); - iA = _mm_add_ps(iA,_mm_mul_ps(_mm_shuffle_ps(B,B,0xF5), _mm_movehl_ps(DC,DC))); - - // d = trace(AB*DC) = trace(A#*B*D#*C) [continue] - d = _mm_add_ps(d, _mm_movehl_ps(d, d)); - d = _mm_add_ss(d, _mm_shuffle_ps(d, d, 1)); - d1 = _mm_mul_ss(dA,dD); - d2 = _mm_mul_ss(dB,dC); - - // iD = D*|A| - C*A#*B - iD = _mm_sub_ps(_mm_mul_ps(D,_mm_shuffle_ps(dA,dA,0)), iD); - - // iA = A*|D| - B*D#*C; - iA = _mm_sub_ps(_mm_mul_ps(A,_mm_shuffle_ps(dD,dD,0)), iA); - - // det = |A|*|D| + |B|*|C| - trace(A#*B*D#*C) - det = _mm_sub_ss(_mm_add_ss(d1,d2),d); - rd = _mm_div_ss(_mm_set_ss(1.0f), det); - -// #ifdef ZERO_SINGULAR -// rd = _mm_and_ps(_mm_cmpneq_ss(det,_mm_setzero_ps()), rd); -// #endif - - // iB = D * (A#B)# = D*B#*A - iB = _mm_mul_ps(D, _mm_shuffle_ps(AB,AB,0x33)); - iB = _mm_sub_ps(iB, _mm_mul_ps(_mm_shuffle_ps(D,D,0xB1), _mm_shuffle_ps(AB,AB,0x66))); - // iC = A * (D#C)# = A*C#*D - iC = _mm_mul_ps(A, _mm_shuffle_ps(DC,DC,0x33)); - iC = _mm_sub_ps(iC, _mm_mul_ps(_mm_shuffle_ps(A,A,0xB1), _mm_shuffle_ps(DC,DC,0x66))); - - rd = _mm_shuffle_ps(rd,rd,0); - rd = _mm_xor_ps(rd, p4f_sign_PNNP); - - // iB = C*|B| - D*B#*A - iB = _mm_sub_ps(_mm_mul_ps(C,_mm_shuffle_ps(dB,dB,0)), iB); - - // iC = B*|C| - A*C#*D; - iC = _mm_sub_ps(_mm_mul_ps(B,_mm_shuffle_ps(dC,dC,0)), iC); - - // iX = iX / det - iA = _mm_mul_ps(rd,iA); - iB = _mm_mul_ps(rd,iB); - iC = _mm_mul_ps(rd,iC); - iD = _mm_mul_ps(rd,iD); - - Index res_stride = result.outerStride(); - float* res = result.data(); - pstoret(res+0, _mm_shuffle_ps(iA,iB,0x77)); - pstoret(res+res_stride, _mm_shuffle_ps(iA,iB,0x22)); - pstoret(res+2*res_stride, _mm_shuffle_ps(iC,iD,0x77)); - pstoret(res+3*res_stride, _mm_shuffle_ps(iC,iD,0x22)); - } - -}; - -template -struct compute_inverse_size4 -{ - enum { - MatrixAlignment = traits::Alignment, - ResultAlignment = traits::Alignment, - StorageOrdersMatch = (MatrixType::Flags&RowMajorBit) == (ResultType::Flags&RowMajorBit) - }; - typedef typename conditional<(MatrixType::Flags&LinearAccessBit),MatrixType const &,typename MatrixType::PlainObject>::type ActualMatrixType; - - static void run(const MatrixType& mat, ResultType& result) - { - ActualMatrixType matrix(mat); - const __m128d _Sign_NP = _mm_castsi128_pd(_mm_set_epi32(0x0,0x0,0x80000000,0x0)); - const __m128d _Sign_PN = _mm_castsi128_pd(_mm_set_epi32(0x80000000,0x0,0x0,0x0)); - - // The inverse is calculated using "Divide and Conquer" technique. The - // original matrix is divide into four 2x2 sub-matrices. Since each - // register of the matrix holds two elements, the smaller matrices are - // consisted of two registers. Hence we get a better locality of the - // calculations. - - // the four sub-matrices - __m128d A1, A2, B1, B2, C1, C2, D1, D2; - - if(StorageOrdersMatch) - { - A1 = matrix.template packet( 0); B1 = matrix.template packet( 2); - A2 = matrix.template packet( 4); B2 = matrix.template packet( 6); - C1 = matrix.template packet( 8); D1 = matrix.template packet(10); - C2 = matrix.template packet(12); D2 = matrix.template packet(14); - } - else - { - __m128d tmp; - A1 = matrix.template packet( 0); C1 = matrix.template packet( 2); - A2 = matrix.template packet( 4); C2 = matrix.template packet( 6); - tmp = A1; - A1 = _mm_unpacklo_pd(A1,A2); - A2 = _mm_unpackhi_pd(tmp,A2); - tmp = C1; - C1 = _mm_unpacklo_pd(C1,C2); - C2 = _mm_unpackhi_pd(tmp,C2); - - B1 = matrix.template packet( 8); D1 = matrix.template packet(10); - B2 = matrix.template packet(12); D2 = matrix.template packet(14); - tmp = B1; - B1 = _mm_unpacklo_pd(B1,B2); - B2 = _mm_unpackhi_pd(tmp,B2); - tmp = D1; - D1 = _mm_unpacklo_pd(D1,D2); - D2 = _mm_unpackhi_pd(tmp,D2); - } - - __m128d iA1, iA2, iB1, iB2, iC1, iC2, iD1, iD2, // partial invese of the sub-matrices - DC1, DC2, AB1, AB2; - __m128d dA, dB, dC, dD; // determinant of the sub-matrices - __m128d det, d1, d2, rd; - - // dA = |A| - dA = _mm_shuffle_pd(A2, A2, 1); - dA = _mm_mul_pd(A1, dA); - dA = _mm_sub_sd(dA, _mm_shuffle_pd(dA,dA,3)); - // dB = |B| - dB = _mm_shuffle_pd(B2, B2, 1); - dB = _mm_mul_pd(B1, dB); - dB = _mm_sub_sd(dB, _mm_shuffle_pd(dB,dB,3)); - - // AB = A# * B - AB1 = _mm_mul_pd(B1, _mm_shuffle_pd(A2,A2,3)); - AB2 = _mm_mul_pd(B2, _mm_shuffle_pd(A1,A1,0)); - AB1 = _mm_sub_pd(AB1, _mm_mul_pd(B2, _mm_shuffle_pd(A1,A1,3))); - AB2 = _mm_sub_pd(AB2, _mm_mul_pd(B1, _mm_shuffle_pd(A2,A2,0))); - - // dC = |C| - dC = _mm_shuffle_pd(C2, C2, 1); - dC = _mm_mul_pd(C1, dC); - dC = _mm_sub_sd(dC, _mm_shuffle_pd(dC,dC,3)); - // dD = |D| - dD = _mm_shuffle_pd(D2, D2, 1); - dD = _mm_mul_pd(D1, dD); - dD = _mm_sub_sd(dD, _mm_shuffle_pd(dD,dD,3)); - - // DC = D# * C - DC1 = _mm_mul_pd(C1, _mm_shuffle_pd(D2,D2,3)); - DC2 = _mm_mul_pd(C2, _mm_shuffle_pd(D1,D1,0)); - DC1 = _mm_sub_pd(DC1, _mm_mul_pd(C2, _mm_shuffle_pd(D1,D1,3))); - DC2 = _mm_sub_pd(DC2, _mm_mul_pd(C1, _mm_shuffle_pd(D2,D2,0))); - - // rd = trace(AB*DC) = trace(A#*B*D#*C) - d1 = _mm_mul_pd(AB1, _mm_shuffle_pd(DC1, DC2, 0)); - d2 = _mm_mul_pd(AB2, _mm_shuffle_pd(DC1, DC2, 3)); - rd = _mm_add_pd(d1, d2); - rd = _mm_add_sd(rd, _mm_shuffle_pd(rd, rd,3)); - - // iD = C*A#*B - iD1 = _mm_mul_pd(AB1, _mm_shuffle_pd(C1,C1,0)); - iD2 = _mm_mul_pd(AB1, _mm_shuffle_pd(C2,C2,0)); - iD1 = _mm_add_pd(iD1, _mm_mul_pd(AB2, _mm_shuffle_pd(C1,C1,3))); - iD2 = _mm_add_pd(iD2, _mm_mul_pd(AB2, _mm_shuffle_pd(C2,C2,3))); - - // iA = B*D#*C - iA1 = _mm_mul_pd(DC1, _mm_shuffle_pd(B1,B1,0)); - iA2 = _mm_mul_pd(DC1, _mm_shuffle_pd(B2,B2,0)); - iA1 = _mm_add_pd(iA1, _mm_mul_pd(DC2, _mm_shuffle_pd(B1,B1,3))); - iA2 = _mm_add_pd(iA2, _mm_mul_pd(DC2, _mm_shuffle_pd(B2,B2,3))); - - // iD = D*|A| - C*A#*B - dA = _mm_shuffle_pd(dA,dA,0); - iD1 = _mm_sub_pd(_mm_mul_pd(D1, dA), iD1); - iD2 = _mm_sub_pd(_mm_mul_pd(D2, dA), iD2); - - // iA = A*|D| - B*D#*C; - dD = _mm_shuffle_pd(dD,dD,0); - iA1 = _mm_sub_pd(_mm_mul_pd(A1, dD), iA1); - iA2 = _mm_sub_pd(_mm_mul_pd(A2, dD), iA2); - - d1 = _mm_mul_sd(dA, dD); - d2 = _mm_mul_sd(dB, dC); - - // iB = D * (A#B)# = D*B#*A - iB1 = _mm_mul_pd(D1, _mm_shuffle_pd(AB2,AB1,1)); - iB2 = _mm_mul_pd(D2, _mm_shuffle_pd(AB2,AB1,1)); - iB1 = _mm_sub_pd(iB1, _mm_mul_pd(_mm_shuffle_pd(D1,D1,1), _mm_shuffle_pd(AB2,AB1,2))); - iB2 = _mm_sub_pd(iB2, _mm_mul_pd(_mm_shuffle_pd(D2,D2,1), _mm_shuffle_pd(AB2,AB1,2))); - - // det = |A|*|D| + |B|*|C| - trace(A#*B*D#*C) - det = _mm_add_sd(d1, d2); - det = _mm_sub_sd(det, rd); - - // iC = A * (D#C)# = A*C#*D - iC1 = _mm_mul_pd(A1, _mm_shuffle_pd(DC2,DC1,1)); - iC2 = _mm_mul_pd(A2, _mm_shuffle_pd(DC2,DC1,1)); - iC1 = _mm_sub_pd(iC1, _mm_mul_pd(_mm_shuffle_pd(A1,A1,1), _mm_shuffle_pd(DC2,DC1,2))); - iC2 = _mm_sub_pd(iC2, _mm_mul_pd(_mm_shuffle_pd(A2,A2,1), _mm_shuffle_pd(DC2,DC1,2))); - - rd = _mm_div_sd(_mm_set_sd(1.0), det); -// #ifdef ZERO_SINGULAR -// rd = _mm_and_pd(_mm_cmpneq_sd(det,_mm_setzero_pd()), rd); -// #endif - rd = _mm_shuffle_pd(rd,rd,0); - - // iB = C*|B| - D*B#*A - dB = _mm_shuffle_pd(dB,dB,0); - iB1 = _mm_sub_pd(_mm_mul_pd(C1, dB), iB1); - iB2 = _mm_sub_pd(_mm_mul_pd(C2, dB), iB2); - - d1 = _mm_xor_pd(rd, _Sign_PN); - d2 = _mm_xor_pd(rd, _Sign_NP); - - // iC = B*|C| - A*C#*D; - dC = _mm_shuffle_pd(dC,dC,0); - iC1 = _mm_sub_pd(_mm_mul_pd(B1, dC), iC1); - iC2 = _mm_sub_pd(_mm_mul_pd(B2, dC), iC2); - - Index res_stride = result.outerStride(); - double* res = result.data(); - pstoret(res+0, _mm_mul_pd(_mm_shuffle_pd(iA2, iA1, 3), d1)); - pstoret(res+res_stride, _mm_mul_pd(_mm_shuffle_pd(iA2, iA1, 0), d2)); - pstoret(res+2, _mm_mul_pd(_mm_shuffle_pd(iB2, iB1, 3), d1)); - pstoret(res+res_stride+2, _mm_mul_pd(_mm_shuffle_pd(iB2, iB1, 0), d2)); - pstoret(res+2*res_stride, _mm_mul_pd(_mm_shuffle_pd(iC2, iC1, 3), d1)); - pstoret(res+3*res_stride, _mm_mul_pd(_mm_shuffle_pd(iC2, iC1, 0), d2)); - pstoret(res+2*res_stride+2,_mm_mul_pd(_mm_shuffle_pd(iD2, iD1, 3), d1)); - pstoret(res+3*res_stride+2,_mm_mul_pd(_mm_shuffle_pd(iD2, iD1, 0), d2)); - } -}; - -} // end namespace internal - -} // end namespace Eigen - -#endif // EIGEN_INVERSE_SSE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Image.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Image.h deleted file mode 100644 index b8b8a045529..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Image.h +++ /dev/null @@ -1,82 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MISC_IMAGE_H -#define EIGEN_MISC_IMAGE_H - -namespace Eigen { - -namespace internal { - -/** \class image_retval_base - * - */ -template -struct traits > -{ - typedef typename DecompositionType::MatrixType MatrixType; - typedef Matrix< - typename MatrixType::Scalar, - MatrixType::RowsAtCompileTime, // the image is a subspace of the destination space, whose - // dimension is the number of rows of the original matrix - Dynamic, // we don't know at compile time the dimension of the image (the rank) - MatrixType::Options, - MatrixType::MaxRowsAtCompileTime, // the image matrix will consist of columns from the original matrix, - MatrixType::MaxColsAtCompileTime // so it has the same number of rows and at most as many columns. - > ReturnType; -}; - -template struct image_retval_base - : public ReturnByValue > -{ - typedef _DecompositionType DecompositionType; - typedef typename DecompositionType::MatrixType MatrixType; - typedef ReturnByValue Base; - - image_retval_base(const DecompositionType& dec, const MatrixType& originalMatrix) - : m_dec(dec), m_rank(dec.rank()), - m_cols(m_rank == 0 ? 1 : m_rank), - m_originalMatrix(originalMatrix) - {} - - inline Index rows() const { return m_dec.rows(); } - inline Index cols() const { return m_cols; } - inline Index rank() const { return m_rank; } - inline const DecompositionType& dec() const { return m_dec; } - inline const MatrixType& originalMatrix() const { return m_originalMatrix; } - - template inline void evalTo(Dest& dst) const - { - static_cast*>(this)->evalTo(dst); - } - - protected: - const DecompositionType& m_dec; - Index m_rank, m_cols; - const MatrixType& m_originalMatrix; -}; - -} // end namespace internal - -#define EIGEN_MAKE_IMAGE_HELPERS(DecompositionType) \ - typedef typename DecompositionType::MatrixType MatrixType; \ - typedef typename MatrixType::Scalar Scalar; \ - typedef typename MatrixType::RealScalar RealScalar; \ - typedef Eigen::internal::image_retval_base Base; \ - using Base::dec; \ - using Base::originalMatrix; \ - using Base::rank; \ - using Base::rows; \ - using Base::cols; \ - image_retval(const DecompositionType& dec, const MatrixType& originalMatrix) \ - : Base(dec, originalMatrix) {} - -} // end namespace Eigen - -#endif // EIGEN_MISC_IMAGE_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h deleted file mode 100644 index bef5d6ff583..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/misc/Kernel.h +++ /dev/null @@ -1,79 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2009 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_MISC_KERNEL_H -#define EIGEN_MISC_KERNEL_H - -namespace Eigen { - -namespace internal { - -/** \class kernel_retval_base - * - */ -template -struct traits > -{ - typedef typename DecompositionType::MatrixType MatrixType; - typedef Matrix< - typename MatrixType::Scalar, - MatrixType::ColsAtCompileTime, // the number of rows in the "kernel matrix" - // is the number of cols of the original matrix - // so that the product "matrix * kernel = zero" makes sense - Dynamic, // we don't know at compile-time the dimension of the kernel - MatrixType::Options, - MatrixType::MaxColsAtCompileTime, // see explanation for 2nd template parameter - MatrixType::MaxColsAtCompileTime // the kernel is a subspace of the domain space, - // whose dimension is the number of columns of the original matrix - > ReturnType; -}; - -template struct kernel_retval_base - : public ReturnByValue > -{ - typedef _DecompositionType DecompositionType; - typedef ReturnByValue Base; - - explicit kernel_retval_base(const DecompositionType& dec) - : m_dec(dec), - m_rank(dec.rank()), - m_cols(m_rank==dec.cols() ? 1 : dec.cols() - m_rank) - {} - - inline Index rows() const { return m_dec.cols(); } - inline Index cols() const { return m_cols; } - inline Index rank() const { return m_rank; } - inline const DecompositionType& dec() const { return m_dec; } - - template inline void evalTo(Dest& dst) const - { - static_cast*>(this)->evalTo(dst); - } - - protected: - const DecompositionType& m_dec; - Index m_rank, m_cols; -}; - -} // end namespace internal - -#define EIGEN_MAKE_KERNEL_HELPERS(DecompositionType) \ - typedef typename DecompositionType::MatrixType MatrixType; \ - typedef typename MatrixType::Scalar Scalar; \ - typedef typename MatrixType::RealScalar RealScalar; \ - typedef Eigen::internal::kernel_retval_base Base; \ - using Base::dec; \ - using Base::rank; \ - using Base::rows; \ - using Base::cols; \ - kernel_retval(const DecompositionType& dec) : Base(dec) {} - -} // end namespace Eigen - -#endif // EIGEN_MISC_KERNEL_H diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h deleted file mode 100644 index 05a7449bc9e..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseBinaryOps.h +++ /dev/null @@ -1,332 +0,0 @@ - -/** \returns an expression of the coefficient wise product of \c *this and \a other - * - * \sa MatrixBase::cwiseProduct - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product) -operator*(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product)(derived(), other.derived()); -} - -/** \returns an expression of the coefficient wise quotient of \c *this and \a other - * - * \sa MatrixBase::cwiseQuotient - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -operator/(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of \c *this and \a other - * - * Example: \include Cwise_min.cpp - * Output: \verbinclude Cwise_min.out - * - * \sa max() - */ -EIGEN_MAKE_CWISE_BINARY_OP(min,min) - -/** \returns an expression of the coefficient-wise min of \c *this and scalar \a other - * - * \sa max() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, - const CwiseNullaryOp, PlainObject> > -#ifdef EIGEN_PARSED_BY_DOXYGEN -min -#else -(min) -#endif -(const Scalar &other) const -{ - return (min)(Derived::PlainObject::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise max of \c *this and \a other - * - * Example: \include Cwise_max.cpp - * Output: \verbinclude Cwise_max.out - * - * \sa min() - */ -EIGEN_MAKE_CWISE_BINARY_OP(max,max) - -/** \returns an expression of the coefficient-wise max of \c *this and scalar \a other - * - * \sa min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, - const CwiseNullaryOp, PlainObject> > -#ifdef EIGEN_PARSED_BY_DOXYGEN -max -#else -(max) -#endif -(const Scalar &other) const -{ - return (max)(Derived::PlainObject::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise power of \c *this to the given array of \a exponents. - * - * This function computes the coefficient-wise power. - * - * Example: \include Cwise_array_power_array.cpp - * Output: \verbinclude Cwise_array_power_array.out - */ -EIGEN_MAKE_CWISE_BINARY_OP(pow,pow) - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(pow,pow) -#else -/** \returns an expression of the coefficients of \c *this rasied to the constant power \a exponent - * - * \tparam T is the scalar type of \a exponent. It must be compatible with the scalar type of the given expression. - * - * This function computes the coefficient-wise power. The function MatrixBase::pow() in the - * unsupported module MatrixFunctions computes the matrix power. - * - * Example: \include Cwise_pow.cpp - * Output: \verbinclude Cwise_pow.out - * - * \sa ArrayBase::pow(ArrayBase), square(), cube(), exp(), log() - */ -template -const CwiseBinaryOp,Derived,Constant > pow(const T& exponent) const; -#endif - - -// TODO code generating macros could be moved to Macros.h and could include generation of documentation -#define EIGEN_MAKE_CWISE_COMP_OP(OP, COMPARATOR) \ -template \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> \ -OP(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ -{ \ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); \ -}\ -typedef CwiseBinaryOp, const Derived, const CwiseNullaryOp, PlainObject> > Cmp ## COMPARATOR ## ReturnType; \ -typedef CwiseBinaryOp, const CwiseNullaryOp, PlainObject>, const Derived > RCmp ## COMPARATOR ## ReturnType; \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Cmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s) const { \ - return this->OP(Derived::PlainObject::Constant(rows(), cols(), s)); \ -} \ -EIGEN_DEVICE_FUNC friend EIGEN_STRONG_INLINE const RCmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s, const EIGEN_CURRENT_STORAGE_BASE_CLASS& d) { \ - return Derived::PlainObject::Constant(d.rows(), d.cols(), s).OP(d); \ -} - -#define EIGEN_MAKE_CWISE_COMP_R_OP(OP, R_OP, RCOMPARATOR) \ -template \ -EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseBinaryOp, const OtherDerived, const Derived> \ -OP(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const \ -{ \ - return CwiseBinaryOp, const OtherDerived, const Derived>(other.derived(), derived()); \ -} \ -EIGEN_DEVICE_FUNC \ -inline const RCmp ## RCOMPARATOR ## ReturnType \ -OP(const Scalar& s) const { \ - return Derived::PlainObject::Constant(rows(), cols(), s).R_OP(*this); \ -} \ -friend inline const Cmp ## RCOMPARATOR ## ReturnType \ -OP(const Scalar& s, const Derived& d) { \ - return d.R_OP(Derived::PlainObject::Constant(d.rows(), d.cols(), s)); \ -} - - - -/** \returns an expression of the coefficient-wise \< operator of *this and \a other - * - * Example: \include Cwise_less.cpp - * Output: \verbinclude Cwise_less.out - * - * \sa all(), any(), operator>(), operator<=() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator<, LT) - -/** \returns an expression of the coefficient-wise \<= operator of *this and \a other - * - * Example: \include Cwise_less_equal.cpp - * Output: \verbinclude Cwise_less_equal.out - * - * \sa all(), any(), operator>=(), operator<() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator<=, LE) - -/** \returns an expression of the coefficient-wise \> operator of *this and \a other - * - * Example: \include Cwise_greater.cpp - * Output: \verbinclude Cwise_greater.out - * - * \sa all(), any(), operator>=(), operator<() - */ -EIGEN_MAKE_CWISE_COMP_R_OP(operator>, operator<, LT) - -/** \returns an expression of the coefficient-wise \>= operator of *this and \a other - * - * Example: \include Cwise_greater_equal.cpp - * Output: \verbinclude Cwise_greater_equal.out - * - * \sa all(), any(), operator>(), operator<=() - */ -EIGEN_MAKE_CWISE_COMP_R_OP(operator>=, operator<=, LE) - -/** \returns an expression of the coefficient-wise == operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include Cwise_equal_equal.cpp - * Output: \verbinclude Cwise_equal_equal.out - * - * \sa all(), any(), isApprox(), isMuchSmallerThan() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator==, EQ) - -/** \returns an expression of the coefficient-wise != operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include Cwise_not_equal.cpp - * Output: \verbinclude Cwise_not_equal.out - * - * \sa all(), any(), isApprox(), isMuchSmallerThan() - */ -EIGEN_MAKE_CWISE_COMP_OP(operator!=, NEQ) - - -#undef EIGEN_MAKE_CWISE_COMP_OP -#undef EIGEN_MAKE_CWISE_COMP_R_OP - -// scalar addition -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator+,sum) -#else -/** \returns an expression of \c *this with each coeff incremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - * - * Example: \include Cwise_plus.cpp - * Output: \verbinclude Cwise_plus.out - * - * \sa operator+=(), operator-() - */ -template -const CwiseBinaryOp,Derived,Constant > operator+(const T& scalar) const; -/** \returns an expression of \a expr with each coeff incremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator+(const T& scalar, const StorageBaseType& expr); -#endif - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator-,difference) -#else -/** \returns an expression of \c *this with each coeff decremented by the constant \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - * - * Example: \include Cwise_minus.cpp - * Output: \verbinclude Cwise_minus.out - * - * \sa operator+=(), operator-() - */ -template -const CwiseBinaryOp,Derived,Constant > operator-(const T& scalar) const; -/** \returns an expression of the constant matrix of value \a scalar decremented by the coefficients of \a expr - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator-(const T& scalar, const StorageBaseType& expr); -#endif - - -#ifndef EIGEN_PARSED_BY_DOXYGEN - EIGEN_MAKE_SCALAR_BINARY_OP_ONTHELEFT(operator/,quotient) -#else - /** - * \brief Component-wise division of the scalar \a s by array elements of \a a. - * - * \tparam Scalar is the scalar type of \a x. It must be compatible with the scalar type of the given array expression (\c Derived::Scalar). - */ - template friend - inline const CwiseBinaryOp,Constant,Derived> - operator/(const T& s,const StorageBaseType& a); -#endif - -/** \returns an expression of the coefficient-wise ^ operator of *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_xor.cpp - * Output: \verbinclude Cwise_boolean_xor.out - * - * \sa operator&&(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator^(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} - -// NOTE disabled until we agree on argument order -#if 0 -/** \cpp11 \returns an expression of the coefficient-wise polygamma function. - * - * \specialfunctions_module - * - * It returns the \a n -th derivative of the digamma(psi) evaluated at \c *this. - * - * \warning Be careful with the order of the parameters: x.polygamma(n) is equivalent to polygamma(n,x) - * - * \sa Eigen::polygamma() - */ -template -inline const CwiseBinaryOp, const DerivedN, const Derived> -polygamma(const EIGEN_CURRENT_STORAGE_BASE_CLASS &n) const -{ - return CwiseBinaryOp, const DerivedN, const Derived>(n.derived(), this->derived()); -} -#endif - -/** \returns an expression of the coefficient-wise zeta function. - * - * \specialfunctions_module - * - * It returns the Riemann zeta function of two arguments \c *this and \a q: - * - * \param *this is the exposent, it must be > 1 - * \param q is the shift, it must be > 0 - * - * \note This function supports only float and double scalar types. To support other scalar types, the user has - * to provide implementations of zeta(T,T) for any scalar type T to be supported. - * - * This method is an alias for zeta(*this,q); - * - * \sa Eigen::zeta() - */ -template -inline const CwiseBinaryOp, const Derived, const DerivedQ> -zeta(const EIGEN_CURRENT_STORAGE_BASE_CLASS &q) const -{ - return CwiseBinaryOp, const Derived, const DerivedQ>(this->derived(), q.derived()); -} diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h deleted file mode 100644 index ebaa3f192b1..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/ArrayCwiseUnaryOps.h +++ /dev/null @@ -1,552 +0,0 @@ - - -typedef CwiseUnaryOp, const Derived> AbsReturnType; -typedef CwiseUnaryOp, const Derived> ArgReturnType; -typedef CwiseUnaryOp, const Derived> Abs2ReturnType; -typedef CwiseUnaryOp, const Derived> SqrtReturnType; -typedef CwiseUnaryOp, const Derived> RsqrtReturnType; -typedef CwiseUnaryOp, const Derived> SignReturnType; -typedef CwiseUnaryOp, const Derived> InverseReturnType; -typedef CwiseUnaryOp, const Derived> BooleanNotReturnType; - -typedef CwiseUnaryOp, const Derived> ExpReturnType; -typedef CwiseUnaryOp, const Derived> LogReturnType; -typedef CwiseUnaryOp, const Derived> Log1pReturnType; -typedef CwiseUnaryOp, const Derived> Log10ReturnType; -typedef CwiseUnaryOp, const Derived> CosReturnType; -typedef CwiseUnaryOp, const Derived> SinReturnType; -typedef CwiseUnaryOp, const Derived> TanReturnType; -typedef CwiseUnaryOp, const Derived> AcosReturnType; -typedef CwiseUnaryOp, const Derived> AsinReturnType; -typedef CwiseUnaryOp, const Derived> AtanReturnType; -typedef CwiseUnaryOp, const Derived> TanhReturnType; -typedef CwiseUnaryOp, const Derived> SinhReturnType; -typedef CwiseUnaryOp, const Derived> CoshReturnType; -typedef CwiseUnaryOp, const Derived> SquareReturnType; -typedef CwiseUnaryOp, const Derived> CubeReturnType; -typedef CwiseUnaryOp, const Derived> RoundReturnType; -typedef CwiseUnaryOp, const Derived> FloorReturnType; -typedef CwiseUnaryOp, const Derived> CeilReturnType; -typedef CwiseUnaryOp, const Derived> IsNaNReturnType; -typedef CwiseUnaryOp, const Derived> IsInfReturnType; -typedef CwiseUnaryOp, const Derived> IsFiniteReturnType; - -/** \returns an expression of the coefficient-wise absolute value of \c *this - * - * Example: \include Cwise_abs.cpp - * Output: \verbinclude Cwise_abs.out - * - * \sa Math functions, abs2() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const AbsReturnType -abs() const -{ - return AbsReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise phase angle of \c *this - * - * Example: \include Cwise_arg.cpp - * Output: \verbinclude Cwise_arg.out - * - * \sa abs() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const ArgReturnType -arg() const -{ - return ArgReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise squared absolute value of \c *this - * - * Example: \include Cwise_abs2.cpp - * Output: \verbinclude Cwise_abs2.out - * - * \sa Math functions, abs(), square() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const Abs2ReturnType -abs2() const -{ - return Abs2ReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise exponential of *this. - * - * This function computes the coefficient-wise exponential. The function MatrixBase::exp() in the - * unsupported module MatrixFunctions computes the matrix exponential. - * - * Example: \include Cwise_exp.cpp - * Output: \verbinclude Cwise_exp.out - * - * \sa Math functions, pow(), log(), sin(), cos() - */ -EIGEN_DEVICE_FUNC -inline const ExpReturnType -exp() const -{ - return ExpReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise logarithm of *this. - * - * This function computes the coefficient-wise logarithm. The function MatrixBase::log() in the - * unsupported module MatrixFunctions computes the matrix logarithm. - * - * Example: \include Cwise_log.cpp - * Output: \verbinclude Cwise_log.out - * - * \sa Math functions, exp() - */ -EIGEN_DEVICE_FUNC -inline const LogReturnType -log() const -{ - return LogReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise logarithm of 1 plus \c *this. - * - * In exact arithmetic, \c x.log() is equivalent to \c (x+1).log(), - * however, with finite precision, this function is much more accurate when \c x is close to zero. - * - * \sa Math functions, log() - */ -EIGEN_DEVICE_FUNC -inline const Log1pReturnType -log1p() const -{ - return Log1pReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise base-10 logarithm of *this. - * - * This function computes the coefficient-wise base-10 logarithm. - * - * Example: \include Cwise_log10.cpp - * Output: \verbinclude Cwise_log10.out - * - * \sa Math functions, log() - */ -EIGEN_DEVICE_FUNC -inline const Log10ReturnType -log10() const -{ - return Log10ReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise square root of *this. - * - * This function computes the coefficient-wise square root. The function MatrixBase::sqrt() in the - * unsupported module MatrixFunctions computes the matrix square root. - * - * Example: \include Cwise_sqrt.cpp - * Output: \verbinclude Cwise_sqrt.out - * - * \sa Math functions, pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const SqrtReturnType -sqrt() const -{ - return SqrtReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise inverse square root of *this. - * - * This function computes the coefficient-wise inverse square root. - * - * Example: \include Cwise_sqrt.cpp - * Output: \verbinclude Cwise_sqrt.out - * - * \sa pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const RsqrtReturnType -rsqrt() const -{ - return RsqrtReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise signum of *this. - * - * This function computes the coefficient-wise signum. - * - * Example: \include Cwise_sign.cpp - * Output: \verbinclude Cwise_sign.out - * - * \sa pow(), square() - */ -EIGEN_DEVICE_FUNC -inline const SignReturnType -sign() const -{ - return SignReturnType(derived()); -} - - -/** \returns an expression of the coefficient-wise cosine of *this. - * - * This function computes the coefficient-wise cosine. The function MatrixBase::cos() in the - * unsupported module MatrixFunctions computes the matrix cosine. - * - * Example: \include Cwise_cos.cpp - * Output: \verbinclude Cwise_cos.out - * - * \sa Math functions, sin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const CosReturnType -cos() const -{ - return CosReturnType(derived()); -} - - -/** \returns an expression of the coefficient-wise sine of *this. - * - * This function computes the coefficient-wise sine. The function MatrixBase::sin() in the - * unsupported module MatrixFunctions computes the matrix sine. - * - * Example: \include Cwise_sin.cpp - * Output: \verbinclude Cwise_sin.out - * - * \sa Math functions, cos(), asin() - */ -EIGEN_DEVICE_FUNC -inline const SinReturnType -sin() const -{ - return SinReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise tan of *this. - * - * Example: \include Cwise_tan.cpp - * Output: \verbinclude Cwise_tan.out - * - * \sa Math functions, cos(), sin() - */ -EIGEN_DEVICE_FUNC -inline const TanReturnType -tan() const -{ - return TanReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc tan of *this. - * - * Example: \include Cwise_atan.cpp - * Output: \verbinclude Cwise_atan.out - * - * \sa Math functions, tan(), asin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const AtanReturnType -atan() const -{ - return AtanReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc cosine of *this. - * - * Example: \include Cwise_acos.cpp - * Output: \verbinclude Cwise_acos.out - * - * \sa Math functions, cos(), asin() - */ -EIGEN_DEVICE_FUNC -inline const AcosReturnType -acos() const -{ - return AcosReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise arc sine of *this. - * - * Example: \include Cwise_asin.cpp - * Output: \verbinclude Cwise_asin.out - * - * \sa Math functions, sin(), acos() - */ -EIGEN_DEVICE_FUNC -inline const AsinReturnType -asin() const -{ - return AsinReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic tan of *this. - * - * Example: \include Cwise_tanh.cpp - * Output: \verbinclude Cwise_tanh.out - * - * \sa Math functions, tan(), sinh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const TanhReturnType -tanh() const -{ - return TanhReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic sin of *this. - * - * Example: \include Cwise_sinh.cpp - * Output: \verbinclude Cwise_sinh.out - * - * \sa Math functions, sin(), tanh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const SinhReturnType -sinh() const -{ - return SinhReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise hyperbolic cos of *this. - * - * Example: \include Cwise_cosh.cpp - * Output: \verbinclude Cwise_cosh.out - * - * \sa Math functions, tan(), sinh(), cosh() - */ -EIGEN_DEVICE_FUNC -inline const CoshReturnType -cosh() const -{ - return CoshReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise inverse of *this. - * - * Example: \include Cwise_inverse.cpp - * Output: \verbinclude Cwise_inverse.out - * - * \sa operator/(), operator*() - */ -EIGEN_DEVICE_FUNC -inline const InverseReturnType -inverse() const -{ - return InverseReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise square of *this. - * - * Example: \include Cwise_square.cpp - * Output: \verbinclude Cwise_square.out - * - * \sa Math functions, abs2(), cube(), pow() - */ -EIGEN_DEVICE_FUNC -inline const SquareReturnType -square() const -{ - return SquareReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise cube of *this. - * - * Example: \include Cwise_cube.cpp - * Output: \verbinclude Cwise_cube.out - * - * \sa Math functions, square(), pow() - */ -EIGEN_DEVICE_FUNC -inline const CubeReturnType -cube() const -{ - return CubeReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise round of *this. - * - * Example: \include Cwise_round.cpp - * Output: \verbinclude Cwise_round.out - * - * \sa Math functions, ceil(), floor() - */ -EIGEN_DEVICE_FUNC -inline const RoundReturnType -round() const -{ - return RoundReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise floor of *this. - * - * Example: \include Cwise_floor.cpp - * Output: \verbinclude Cwise_floor.out - * - * \sa Math functions, ceil(), round() - */ -EIGEN_DEVICE_FUNC -inline const FloorReturnType -floor() const -{ - return FloorReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise ceil of *this. - * - * Example: \include Cwise_ceil.cpp - * Output: \verbinclude Cwise_ceil.out - * - * \sa Math functions, floor(), round() - */ -EIGEN_DEVICE_FUNC -inline const CeilReturnType -ceil() const -{ - return CeilReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isnan of *this. - * - * Example: \include Cwise_isNaN.cpp - * Output: \verbinclude Cwise_isNaN.out - * - * \sa isfinite(), isinf() - */ -EIGEN_DEVICE_FUNC -inline const IsNaNReturnType -isNaN() const -{ - return IsNaNReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isinf of *this. - * - * Example: \include Cwise_isInf.cpp - * Output: \verbinclude Cwise_isInf.out - * - * \sa isnan(), isfinite() - */ -EIGEN_DEVICE_FUNC -inline const IsInfReturnType -isInf() const -{ - return IsInfReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise isfinite of *this. - * - * Example: \include Cwise_isFinite.cpp - * Output: \verbinclude Cwise_isFinite.out - * - * \sa isnan(), isinf() - */ -EIGEN_DEVICE_FUNC -inline const IsFiniteReturnType -isFinite() const -{ - return IsFiniteReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise ! operator of *this - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_not.cpp - * Output: \verbinclude Cwise_boolean_not.out - * - * \sa operator!=() - */ -EIGEN_DEVICE_FUNC -inline const BooleanNotReturnType -operator!() const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return BooleanNotReturnType(derived()); -} - - -// --- SpecialFunctions module --- - -typedef CwiseUnaryOp, const Derived> LgammaReturnType; -typedef CwiseUnaryOp, const Derived> DigammaReturnType; -typedef CwiseUnaryOp, const Derived> ErfReturnType; -typedef CwiseUnaryOp, const Derived> ErfcReturnType; - -/** \cpp11 \returns an expression of the coefficient-wise ln(|gamma(*this)|). - * - * \specialfunctions_module - * - * Example: \include Cwise_lgamma.cpp - * Output: \verbinclude Cwise_lgamma.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of lgamma(T) for any scalar - * type T to be supported. - * - * \sa Math functions, digamma() - */ -EIGEN_DEVICE_FUNC -inline const LgammaReturnType -lgamma() const -{ - return LgammaReturnType(derived()); -} - -/** \returns an expression of the coefficient-wise digamma (psi, derivative of lgamma). - * - * \specialfunctions_module - * - * \note This function supports only float and double scalar types. To support other scalar types, - * the user has to provide implementations of digamma(T) for any scalar - * type T to be supported. - * - * \sa Math functions, Eigen::digamma(), Eigen::polygamma(), lgamma() - */ -EIGEN_DEVICE_FUNC -inline const DigammaReturnType -digamma() const -{ - return DigammaReturnType(derived()); -} - -/** \cpp11 \returns an expression of the coefficient-wise Gauss error - * function of *this. - * - * \specialfunctions_module - * - * Example: \include Cwise_erf.cpp - * Output: \verbinclude Cwise_erf.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of erf(T) for any scalar - * type T to be supported. - * - * \sa Math functions, erfc() - */ -EIGEN_DEVICE_FUNC -inline const ErfReturnType -erf() const -{ - return ErfReturnType(derived()); -} - -/** \cpp11 \returns an expression of the coefficient-wise Complementary error - * function of *this. - * - * \specialfunctions_module - * - * Example: \include Cwise_erfc.cpp - * Output: \verbinclude Cwise_erfc.out - * - * \note This function supports only float and double scalar types in c++11 mode. To support other scalar types, - * or float/double in non c++11 mode, the user has to provide implementations of erfc(T) for any scalar - * type T to be supported. - * - * \sa Math functions, erf() - */ -EIGEN_DEVICE_FUNC -inline const ErfcReturnType -erfc() const -{ - return ErfcReturnType(derived()); -} diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h deleted file mode 100644 index ac35a0086cf..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/BlockMethods.h +++ /dev/null @@ -1,1058 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2010 Gael Guennebaud -// Copyright (C) 2006-2010 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -#ifndef EIGEN_PARSED_BY_DOXYGEN - -/// \internal expression type of a column */ -typedef Block::RowsAtCompileTime, 1, !IsRowMajor> ColXpr; -typedef const Block::RowsAtCompileTime, 1, !IsRowMajor> ConstColXpr; -/// \internal expression type of a row */ -typedef Block::ColsAtCompileTime, IsRowMajor> RowXpr; -typedef const Block::ColsAtCompileTime, IsRowMajor> ConstRowXpr; -/// \internal expression type of a block of whole columns */ -typedef Block::RowsAtCompileTime, Dynamic, !IsRowMajor> ColsBlockXpr; -typedef const Block::RowsAtCompileTime, Dynamic, !IsRowMajor> ConstColsBlockXpr; -/// \internal expression type of a block of whole rows */ -typedef Block::ColsAtCompileTime, IsRowMajor> RowsBlockXpr; -typedef const Block::ColsAtCompileTime, IsRowMajor> ConstRowsBlockXpr; -/// \internal expression type of a block of whole columns */ -template struct NColsBlockXpr { typedef Block::RowsAtCompileTime, N, !IsRowMajor> Type; }; -template struct ConstNColsBlockXpr { typedef const Block::RowsAtCompileTime, N, !IsRowMajor> Type; }; -/// \internal expression type of a block of whole rows */ -template struct NRowsBlockXpr { typedef Block::ColsAtCompileTime, IsRowMajor> Type; }; -template struct ConstNRowsBlockXpr { typedef const Block::ColsAtCompileTime, IsRowMajor> Type; }; -/// \internal expression of a block */ -typedef Block BlockXpr; -typedef const Block ConstBlockXpr; -/// \internal expression of a block of fixed sizes */ -template struct FixedBlockXpr { typedef Block Type; }; -template struct ConstFixedBlockXpr { typedef Block Type; }; - -typedef VectorBlock SegmentReturnType; -typedef const VectorBlock ConstSegmentReturnType; -template struct FixedSegmentReturnType { typedef VectorBlock Type; }; -template struct ConstFixedSegmentReturnType { typedef const VectorBlock Type; }; - -#endif // not EIGEN_PARSED_BY_DOXYGEN - -/// \returns a dynamic-size expression of a block in *this. -/// -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// \param blockRows the number of rows in the block -/// \param blockCols the number of columns in the block -/// -/// Example: \include MatrixBase_block_int_int_int_int.cpp -/// Output: \verbinclude MatrixBase_block_int_int_int_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size matrix, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr block(Index startRow, Index startCol, Index blockRows, Index blockCols) -{ - return BlockXpr(derived(), startRow, startCol, blockRows, blockCols); -} - -/// This is the const version of block(Index,Index,Index,Index). */ -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr block(Index startRow, Index startCol, Index blockRows, Index blockCols) const -{ - return ConstBlockXpr(derived(), startRow, startCol, blockRows, blockCols); -} - - - - -/// \returns a dynamic-size expression of a top-right corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_topRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_topRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr topRightCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// This is the const version of topRightCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr topRightCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// \returns an expression of a fixed-size top-right corner of *this. -/// -/// \tparam CRows the number of rows in the corner -/// \tparam CCols the number of columns in the corner -/// -/// Example: \include MatrixBase_template_int_int_topRightCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topRightCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type topRightCorner() -{ - return typename FixedBlockXpr::Type(derived(), 0, cols() - CCols); -} - -/// This is the const version of topRightCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type topRightCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, cols() - CCols); -} - -/// \returns an expression of a top-right corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_topRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type topRightCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), 0, cols() - cCols, cRows, cCols); -} - -/// This is the const version of topRightCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type topRightCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, cols() - cCols, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a top-left corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_topLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_topLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr topLeftCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), 0, 0, cRows, cCols); -} - -/// This is the const version of topLeftCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr topLeftCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), 0, 0, cRows, cCols); -} - -/// \returns an expression of a fixed-size top-left corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_topLeftCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topLeftCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type topLeftCorner() -{ - return typename FixedBlockXpr::Type(derived(), 0, 0); -} - -/// This is the const version of topLeftCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type topLeftCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, 0); -} - -/// \returns an expression of a top-left corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_topLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_topLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type topLeftCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), 0, 0, cRows, cCols); -} - -/// This is the const version of topLeftCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type topLeftCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), 0, 0, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a bottom-right corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_bottomRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_bottomRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr bottomRightCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// This is the const version of bottomRightCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr bottomRightCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// \returns an expression of a fixed-size bottom-right corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_bottomRightCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomRightCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type bottomRightCorner() -{ - return typename FixedBlockXpr::Type(derived(), rows() - CRows, cols() - CCols); -} - -/// This is the const version of bottomRightCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type bottomRightCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - CRows, cols() - CCols); -} - -/// \returns an expression of a bottom-right corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_bottomRightCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomRightCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type bottomRightCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - -/// This is the const version of bottomRightCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type bottomRightCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - cRows, cols() - cCols, cRows, cCols); -} - - - -/// \returns a dynamic-size expression of a bottom-left corner of *this. -/// -/// \param cRows the number of rows in the corner -/// \param cCols the number of columns in the corner -/// -/// Example: \include MatrixBase_bottomLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_bottomLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline BlockXpr bottomLeftCorner(Index cRows, Index cCols) -{ - return BlockXpr(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// This is the const version of bottomLeftCorner(Index, Index). -EIGEN_DEVICE_FUNC -inline const ConstBlockXpr bottomLeftCorner(Index cRows, Index cCols) const -{ - return ConstBlockXpr(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// \returns an expression of a fixed-size bottom-left corner of *this. -/// -/// The template parameters CRows and CCols are the number of rows and columns in the corner. -/// -/// Example: \include MatrixBase_template_int_int_bottomLeftCorner.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomLeftCorner.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type bottomLeftCorner() -{ - return typename FixedBlockXpr::Type(derived(), rows() - CRows, 0); -} - -/// This is the const version of bottomLeftCorner(). -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type bottomLeftCorner() const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - CRows, 0); -} - -/// \returns an expression of a bottom-left corner of *this. -/// -/// \tparam CRows number of rows in corner as specified at compile-time -/// \tparam CCols number of columns in corner as specified at compile-time -/// \param cRows number of rows in corner as specified at run-time -/// \param cCols number of columns in corner as specified at run-time -/// -/// This function is mainly useful for corners where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a cRows should equal \a CRows unless -/// \a CRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_bottomLeftCorner_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_bottomLeftCorner_int_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block -/// -template -inline typename FixedBlockXpr::Type bottomLeftCorner(Index cRows, Index cCols) -{ - return typename FixedBlockXpr::Type(derived(), rows() - cRows, 0, cRows, cCols); -} - -/// This is the const version of bottomLeftCorner(Index, Index). -template -inline const typename ConstFixedBlockXpr::Type bottomLeftCorner(Index cRows, Index cCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), rows() - cRows, 0, cRows, cCols); -} - - - -/// \returns a block consisting of the top rows of *this. -/// -/// \param n the number of rows in the block -/// -/// Example: \include MatrixBase_topRows_int.cpp -/// Output: \verbinclude MatrixBase_topRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr topRows(Index n) -{ - return RowsBlockXpr(derived(), 0, 0, n, cols()); -} - -/// This is the const version of topRows(Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr topRows(Index n) const -{ - return ConstRowsBlockXpr(derived(), 0, 0, n, cols()); -} - -/// \returns a block consisting of the top rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_topRows.cpp -/// Output: \verbinclude MatrixBase_template_int_topRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type topRows(Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), 0, 0, n, cols()); -} - -/// This is the const version of topRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type topRows(Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), 0, 0, n, cols()); -} - - - -/// \returns a block consisting of the bottom rows of *this. -/// -/// \param n the number of rows in the block -/// -/// Example: \include MatrixBase_bottomRows_int.cpp -/// Output: \verbinclude MatrixBase_bottomRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr bottomRows(Index n) -{ - return RowsBlockXpr(derived(), rows() - n, 0, n, cols()); -} - -/// This is the const version of bottomRows(Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr bottomRows(Index n) const -{ - return ConstRowsBlockXpr(derived(), rows() - n, 0, n, cols()); -} - -/// \returns a block consisting of the bottom rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_bottomRows.cpp -/// Output: \verbinclude MatrixBase_template_int_bottomRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type bottomRows(Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), rows() - n, 0, n, cols()); -} - -/// This is the const version of bottomRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type bottomRows(Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), rows() - n, 0, n, cols()); -} - - - -/// \returns a block consisting of a range of rows of *this. -/// -/// \param startRow the index of the first row in the block -/// \param n the number of rows in the block -/// -/// Example: \include DenseBase_middleRows_int.cpp -/// Output: \verbinclude DenseBase_middleRows_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline RowsBlockXpr middleRows(Index startRow, Index n) -{ - return RowsBlockXpr(derived(), startRow, 0, n, cols()); -} - -/// This is the const version of middleRows(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstRowsBlockXpr middleRows(Index startRow, Index n) const -{ - return ConstRowsBlockXpr(derived(), startRow, 0, n, cols()); -} - -/// \returns a block consisting of a range of rows of *this. -/// -/// \tparam N the number of rows in the block as specified at compile-time -/// \param startRow the index of the first row in the block -/// \param n the number of rows in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include DenseBase_template_int_middleRows.cpp -/// Output: \verbinclude DenseBase_template_int_middleRows.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NRowsBlockXpr::Type middleRows(Index startRow, Index n = N) -{ - return typename NRowsBlockXpr::Type(derived(), startRow, 0, n, cols()); -} - -/// This is the const version of middleRows(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNRowsBlockXpr::Type middleRows(Index startRow, Index n = N) const -{ - return typename ConstNRowsBlockXpr::Type(derived(), startRow, 0, n, cols()); -} - - - -/// \returns a block consisting of the left columns of *this. -/// -/// \param n the number of columns in the block -/// -/// Example: \include MatrixBase_leftCols_int.cpp -/// Output: \verbinclude MatrixBase_leftCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr leftCols(Index n) -{ - return ColsBlockXpr(derived(), 0, 0, rows(), n); -} - -/// This is the const version of leftCols(Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr leftCols(Index n) const -{ - return ConstColsBlockXpr(derived(), 0, 0, rows(), n); -} - -/// \returns a block consisting of the left columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_leftCols.cpp -/// Output: \verbinclude MatrixBase_template_int_leftCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type leftCols(Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, 0, rows(), n); -} - -/// This is the const version of leftCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type leftCols(Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, 0, rows(), n); -} - - - -/// \returns a block consisting of the right columns of *this. -/// -/// \param n the number of columns in the block -/// -/// Example: \include MatrixBase_rightCols_int.cpp -/// Output: \verbinclude MatrixBase_rightCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr rightCols(Index n) -{ - return ColsBlockXpr(derived(), 0, cols() - n, rows(), n); -} - -/// This is the const version of rightCols(Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr rightCols(Index n) const -{ - return ConstColsBlockXpr(derived(), 0, cols() - n, rows(), n); -} - -/// \returns a block consisting of the right columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_rightCols.cpp -/// Output: \verbinclude MatrixBase_template_int_rightCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type rightCols(Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, cols() - n, rows(), n); -} - -/// This is the const version of rightCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type rightCols(Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, cols() - n, rows(), n); -} - - - -/// \returns a block consisting of a range of columns of *this. -/// -/// \param startCol the index of the first column in the block -/// \param numCols the number of columns in the block -/// -/// Example: \include DenseBase_middleCols_int.cpp -/// Output: \verbinclude DenseBase_middleCols_int.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -EIGEN_DEVICE_FUNC -inline ColsBlockXpr middleCols(Index startCol, Index numCols) -{ - return ColsBlockXpr(derived(), 0, startCol, rows(), numCols); -} - -/// This is the const version of middleCols(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstColsBlockXpr middleCols(Index startCol, Index numCols) const -{ - return ConstColsBlockXpr(derived(), 0, startCol, rows(), numCols); -} - -/// \returns a block consisting of a range of columns of *this. -/// -/// \tparam N the number of columns in the block as specified at compile-time -/// \param startCol the index of the first column in the block -/// \param n the number of columns in the block as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include DenseBase_template_int_middleCols.cpp -/// Output: \verbinclude DenseBase_template_int_middleCols.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename NColsBlockXpr::Type middleCols(Index startCol, Index n = N) -{ - return typename NColsBlockXpr::Type(derived(), 0, startCol, rows(), n); -} - -/// This is the const version of middleCols(). -template -EIGEN_DEVICE_FUNC -inline typename ConstNColsBlockXpr::Type middleCols(Index startCol, Index n = N) const -{ - return typename ConstNColsBlockXpr::Type(derived(), 0, startCol, rows(), n); -} - - - -/// \returns a fixed-size expression of a block in *this. -/// -/// The template parameters \a NRows and \a NCols are the number of -/// rows and columns in the block. -/// -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// -/// Example: \include MatrixBase_block_int_int.cpp -/// Output: \verbinclude MatrixBase_block_int_int.out -/// -/// \note since block is a templated member, the keyword template has to be used -/// if the matrix type is also a template parameter: \code m.template block<3,3>(1,1); \endcode -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedBlockXpr::Type block(Index startRow, Index startCol) -{ - return typename FixedBlockXpr::Type(derived(), startRow, startCol); -} - -/// This is the const version of block<>(Index, Index). */ -template -EIGEN_DEVICE_FUNC -inline const typename ConstFixedBlockXpr::Type block(Index startRow, Index startCol) const -{ - return typename ConstFixedBlockXpr::Type(derived(), startRow, startCol); -} - -/// \returns an expression of a block in *this. -/// -/// \tparam NRows number of rows in block as specified at compile-time -/// \tparam NCols number of columns in block as specified at compile-time -/// \param startRow the first row in the block -/// \param startCol the first column in the block -/// \param blockRows number of rows in block as specified at run-time -/// \param blockCols number of columns in block as specified at run-time -/// -/// This function is mainly useful for blocks where the number of rows is specified at compile-time -/// and the number of columns is specified at run-time, or vice versa. The compile-time and run-time -/// information should not contradict. In other words, \a blockRows should equal \a NRows unless -/// \a NRows is \a Dynamic, and the same for the number of columns. -/// -/// Example: \include MatrixBase_template_int_int_block_int_int_int_int.cpp -/// Output: \verbinclude MatrixBase_template_int_int_block_int_int_int_int.cpp -/// -EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL -/// -/// \sa class Block, block(Index,Index,Index,Index) -/// -template -inline typename FixedBlockXpr::Type block(Index startRow, Index startCol, - Index blockRows, Index blockCols) -{ - return typename FixedBlockXpr::Type(derived(), startRow, startCol, blockRows, blockCols); -} - -/// This is the const version of block<>(Index, Index, Index, Index). -template -inline const typename ConstFixedBlockXpr::Type block(Index startRow, Index startCol, - Index blockRows, Index blockCols) const -{ - return typename ConstFixedBlockXpr::Type(derived(), startRow, startCol, blockRows, blockCols); -} - -/// \returns an expression of the \a i-th column of *this. Note that the numbering starts at 0. -/// -/// Example: \include MatrixBase_col.cpp -/// Output: \verbinclude MatrixBase_col.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(column-major) -/** - * \sa row(), class Block */ -EIGEN_DEVICE_FUNC -inline ColXpr col(Index i) -{ - return ColXpr(derived(), i); -} - -/// This is the const version of col(). -EIGEN_DEVICE_FUNC -inline ConstColXpr col(Index i) const -{ - return ConstColXpr(derived(), i); -} - -/// \returns an expression of the \a i-th row of *this. Note that the numbering starts at 0. -/// -/// Example: \include MatrixBase_row.cpp -/// Output: \verbinclude MatrixBase_row.out -/// -EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(row-major) -/** - * \sa col(), class Block */ -EIGEN_DEVICE_FUNC -inline RowXpr row(Index i) -{ - return RowXpr(derived(), i); -} - -/// This is the const version of row(). */ -EIGEN_DEVICE_FUNC -inline ConstRowXpr row(Index i) const -{ - return ConstRowXpr(derived(), i); -} - -/// \returns a dynamic-size expression of a segment (i.e. a vector block) in *this. -/// -/// \only_for_vectors -/// -/// \param start the first coefficient in the segment -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_segment_int_int.cpp -/// Output: \verbinclude MatrixBase_segment_int_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, segment(Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType segment(Index start, Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), start, n); -} - - -/// This is the const version of segment(Index,Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType segment(Index start, Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), start, n); -} - -/// \returns a dynamic-size expression of the first coefficients of *this. -/// -/// \only_for_vectors -/// -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_start_int.cpp -/// Output: \verbinclude MatrixBase_start_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType head(Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), 0, n); -} - -/// This is the const version of head(Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType head(Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), 0, n); -} - -/// \returns a dynamic-size expression of the last coefficients of *this. -/// -/// \only_for_vectors -/// -/// \param n the number of coefficients in the segment -/// -/// Example: \include MatrixBase_end_int.cpp -/// Output: \verbinclude MatrixBase_end_int.out -/// -/// \note Even though the returned expression has dynamic size, in the case -/// when it is applied to a fixed-size vector, it inherits a fixed maximal size, -/// which means that evaluating it does not cause a dynamic memory allocation. -/// -/// \sa class Block, block(Index,Index) -/// -EIGEN_DEVICE_FUNC -inline SegmentReturnType tail(Index n) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return SegmentReturnType(derived(), this->size() - n, n); -} - -/// This is the const version of tail(Index). -EIGEN_DEVICE_FUNC -inline ConstSegmentReturnType tail(Index n) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return ConstSegmentReturnType(derived(), this->size() - n, n); -} - -/// \returns a fixed-size expression of a segment (i.e. a vector block) in \c *this -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param start the index of the first element in the segment -/// \param n the number of coefficients in the segment as specified at compile-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_segment.cpp -/// Output: \verbinclude MatrixBase_template_int_segment.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type segment(Index start, Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), start, n); -} - -/// This is the const version of segment(Index). -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type segment(Index start, Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), start, n); -} - -/// \returns a fixed-size expression of the first coefficients of *this. -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param n the number of coefficients in the segment as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_start.cpp -/// Output: \verbinclude MatrixBase_template_int_start.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type head(Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), 0, n); -} - -/// This is the const version of head(). -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type head(Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), 0, n); -} - -/// \returns a fixed-size expression of the last coefficients of *this. -/// -/// \only_for_vectors -/// -/// \tparam N the number of coefficients in the segment as specified at compile-time -/// \param n the number of coefficients in the segment as specified at run-time -/// -/// The compile-time and run-time information should not contradict. In other words, -/// \a n should equal \a N unless \a N is \a Dynamic. -/// -/// Example: \include MatrixBase_template_int_end.cpp -/// Output: \verbinclude MatrixBase_template_int_end.out -/// -/// \sa class Block -/// -template -EIGEN_DEVICE_FUNC -inline typename FixedSegmentReturnType::Type tail(Index n = N) -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename FixedSegmentReturnType::Type(derived(), size() - n); -} - -/// This is the const version of tail. -template -EIGEN_DEVICE_FUNC -inline typename ConstFixedSegmentReturnType::Type tail(Index n = N) const -{ - EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived) - return typename ConstFixedSegmentReturnType::Type(derived(), size() - n); -} diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h deleted file mode 100644 index 8b6730ede02..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseBinaryOps.h +++ /dev/null @@ -1,115 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2016 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing common coefficient wise functions. - -/** \returns an expression of the difference of \c *this and \a other - * - * \note If you want to substract a given scalar from all coefficients, see Cwise::operator-(). - * - * \sa class CwiseBinaryOp, operator-=() - */ -EIGEN_MAKE_CWISE_BINARY_OP(operator-,difference) - -/** \returns an expression of the sum of \c *this and \a other - * - * \note If you want to add a given scalar to all coefficients, see Cwise::operator+(). - * - * \sa class CwiseBinaryOp, operator+=() - */ -EIGEN_MAKE_CWISE_BINARY_OP(operator+,sum) - -/** \returns an expression of a custom coefficient-wise operator \a func of *this and \a other - * - * The template parameter \a CustomBinaryOp is the type of the functor - * of the custom operator (see class CwiseBinaryOp for an example) - * - * Here is an example illustrating the use of custom functors: - * \include class_CwiseBinaryOp.cpp - * Output: \verbinclude class_CwiseBinaryOp.out - * - * \sa class CwiseBinaryOp, operator+(), operator-(), cwiseProduct() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp -binaryExpr(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other, const CustomBinaryOp& func = CustomBinaryOp()) const -{ - return CwiseBinaryOp(derived(), other.derived(), func); -} - - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP(operator*,product) -#else -/** \returns an expression of \c *this scaled by the scalar factor \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template -const CwiseBinaryOp,Derived,Constant > operator*(const T& scalar) const; -/** \returns an expression of \a expr scaled by the scalar factor \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template friend -const CwiseBinaryOp,Constant,Derived> operator*(const T& scalar, const StorageBaseType& expr); -#endif - - - -#ifndef EIGEN_PARSED_BY_DOXYGEN -EIGEN_MAKE_SCALAR_BINARY_OP_ONTHERIGHT(operator/,quotient) -#else -/** \returns an expression of \c *this divided by the scalar value \a scalar - * - * \tparam T is the scalar type of \a scalar. It must be compatible with the scalar type of the given expression. - */ -template -const CwiseBinaryOp,Derived,Constant > operator/(const T& scalar) const; -#endif - -/** \returns an expression of the coefficient-wise boolean \b and operator of \c *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_and.cpp - * Output: \verbinclude Cwise_boolean_and.out - * - * \sa operator||(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator&&(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} - -/** \returns an expression of the coefficient-wise boolean \b or operator of \c *this and \a other - * - * \warning this operator is for expression of bool only. - * - * Example: \include Cwise_boolean_or.cpp - * Output: \verbinclude Cwise_boolean_or.out - * - * \sa operator&&(), select() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp -operator||(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - EIGEN_STATIC_ASSERT((internal::is_same::value && internal::is_same::value), - THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_OF_BOOL); - return CwiseBinaryOp(derived(),other.derived()); -} diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h deleted file mode 100644 index 89f4faaac6b..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/CommonCwiseUnaryOps.h +++ /dev/null @@ -1,163 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing common coefficient wise functions. - -#ifndef EIGEN_PARSED_BY_DOXYGEN - -/** \internal the return type of conjugate() */ -typedef typename internal::conditional::IsComplex, - const CwiseUnaryOp, const Derived>, - const Derived& - >::type ConjugateReturnType; -/** \internal the return type of real() const */ -typedef typename internal::conditional::IsComplex, - const CwiseUnaryOp, const Derived>, - const Derived& - >::type RealReturnType; -/** \internal the return type of real() */ -typedef typename internal::conditional::IsComplex, - CwiseUnaryView, Derived>, - Derived& - >::type NonConstRealReturnType; -/** \internal the return type of imag() const */ -typedef CwiseUnaryOp, const Derived> ImagReturnType; -/** \internal the return type of imag() */ -typedef CwiseUnaryView, Derived> NonConstImagReturnType; - -typedef CwiseUnaryOp, const Derived> NegativeReturnType; - -#endif // not EIGEN_PARSED_BY_DOXYGEN - -/// \returns an expression of the opposite of \c *this -/// -EIGEN_DOC_UNARY_ADDONS(operator-,opposite) -/// -EIGEN_DEVICE_FUNC -inline const NegativeReturnType -operator-() const { return NegativeReturnType(derived()); } - - -template struct CastXpr { typedef typename internal::cast_return_type, const Derived> >::type Type; }; - -/// \returns an expression of \c *this with the \a Scalar type casted to -/// \a NewScalar. -/// -/// The template parameter \a NewScalar is the type we are casting the scalars to. -/// -EIGEN_DOC_UNARY_ADDONS(cast,conversion function) -/// -/// \sa class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -typename CastXpr::Type -cast() const -{ - return typename CastXpr::Type(derived()); -} - -/// \returns an expression of the complex conjugate of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(conjugate,complex conjugate) -/// -/// \sa Math functions, MatrixBase::adjoint() -EIGEN_DEVICE_FUNC -inline ConjugateReturnType -conjugate() const -{ - return ConjugateReturnType(derived()); -} - -/// \returns a read-only expression of the real part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(real,real part function) -/// -/// \sa imag() -EIGEN_DEVICE_FUNC -inline RealReturnType -real() const { return RealReturnType(derived()); } - -/// \returns an read-only expression of the imaginary part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(imag,imaginary part function) -/// -/// \sa real() -EIGEN_DEVICE_FUNC -inline const ImagReturnType -imag() const { return ImagReturnType(derived()); } - -/// \brief Apply a unary operator coefficient-wise -/// \param[in] func Functor implementing the unary operator -/// \tparam CustomUnaryOp Type of \a func -/// \returns An expression of a custom coefficient-wise unary operator \a func of *this -/// -/// The function \c ptr_fun() from the C++ standard library can be used to make functors out of normal functions. -/// -/// Example: -/// \include class_CwiseUnaryOp_ptrfun.cpp -/// Output: \verbinclude class_CwiseUnaryOp_ptrfun.out -/// -/// Genuine functors allow for more possibilities, for instance it may contain a state. -/// -/// Example: -/// \include class_CwiseUnaryOp.cpp -/// Output: \verbinclude class_CwiseUnaryOp.out -/// -EIGEN_DOC_UNARY_ADDONS(unaryExpr,unary function) -/// -/// \sa unaryViewExpr, binaryExpr, class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -inline const CwiseUnaryOp -unaryExpr(const CustomUnaryOp& func = CustomUnaryOp()) const -{ - return CwiseUnaryOp(derived(), func); -} - -/// \returns an expression of a custom coefficient-wise unary operator \a func of *this -/// -/// The template parameter \a CustomUnaryOp is the type of the functor -/// of the custom unary operator. -/// -/// Example: -/// \include class_CwiseUnaryOp.cpp -/// Output: \verbinclude class_CwiseUnaryOp.out -/// -EIGEN_DOC_UNARY_ADDONS(unaryViewExpr,unary function) -/// -/// \sa unaryExpr, binaryExpr class CwiseUnaryOp -/// -template -EIGEN_DEVICE_FUNC -inline const CwiseUnaryView -unaryViewExpr(const CustomViewOp& func = CustomViewOp()) const -{ - return CwiseUnaryView(derived(), func); -} - -/// \returns a non const expression of the real part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(real,real part function) -/// -/// \sa imag() -EIGEN_DEVICE_FUNC -inline NonConstRealReturnType -real() { return NonConstRealReturnType(derived()); } - -/// \returns a non const expression of the imaginary part of \c *this. -/// -EIGEN_DOC_UNARY_ADDONS(imag,imaginary part function) -/// -/// \sa real() -EIGEN_DEVICE_FUNC -inline NonConstImagReturnType -imag() { return NonConstImagReturnType(derived()); } diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h deleted file mode 100644 index f1084abefbc..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseBinaryOps.h +++ /dev/null @@ -1,152 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is a base class plugin containing matrix specifics coefficient wise functions. - -/** \returns an expression of the Schur product (coefficient wise product) of *this and \a other - * - * Example: \include MatrixBase_cwiseProduct.cpp - * Output: \verbinclude MatrixBase_cwiseProduct.out - * - * \sa class CwiseBinaryOp, cwiseAbs2 - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product) -cwiseProduct(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return EIGEN_CWISE_BINARY_RETURN_TYPE(Derived,OtherDerived,product)(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise == operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include MatrixBase_cwiseEqual.cpp - * Output: \verbinclude MatrixBase_cwiseEqual.out - * - * \sa cwiseNotEqual(), isApprox(), isMuchSmallerThan() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseEqual(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise != operator of *this and \a other - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * Example: \include MatrixBase_cwiseNotEqual.cpp - * Output: \verbinclude MatrixBase_cwiseNotEqual.out - * - * \sa cwiseEqual(), isApprox(), isMuchSmallerThan() - */ -template -EIGEN_DEVICE_FUNC -inline const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseNotEqual(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of *this and \a other - * - * Example: \include MatrixBase_cwiseMin.cpp - * Output: \verbinclude MatrixBase_cwiseMin.out - * - * \sa class CwiseBinaryOp, max() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseMin(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise min of *this and scalar \a other - * - * \sa class CwiseBinaryOp, min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const ConstantReturnType> -cwiseMin(const Scalar &other) const -{ - return cwiseMin(Derived::Constant(rows(), cols(), other)); -} - -/** \returns an expression of the coefficient-wise max of *this and \a other - * - * Example: \include MatrixBase_cwiseMax.cpp - * Output: \verbinclude MatrixBase_cwiseMax.out - * - * \sa class CwiseBinaryOp, min() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseMax(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -/** \returns an expression of the coefficient-wise max of *this and scalar \a other - * - * \sa class CwiseBinaryOp, min() - */ -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const ConstantReturnType> -cwiseMax(const Scalar &other) const -{ - return cwiseMax(Derived::Constant(rows(), cols(), other)); -} - - -/** \returns an expression of the coefficient-wise quotient of *this and \a other - * - * Example: \include MatrixBase_cwiseQuotient.cpp - * Output: \verbinclude MatrixBase_cwiseQuotient.out - * - * \sa class CwiseBinaryOp, cwiseProduct(), cwiseInverse() - */ -template -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseBinaryOp, const Derived, const OtherDerived> -cwiseQuotient(const EIGEN_CURRENT_STORAGE_BASE_CLASS &other) const -{ - return CwiseBinaryOp, const Derived, const OtherDerived>(derived(), other.derived()); -} - -typedef CwiseBinaryOp, const Derived, const ConstantReturnType> CwiseScalarEqualReturnType; - -/** \returns an expression of the coefficient-wise == operator of \c *this and a scalar \a s - * - * \warning this performs an exact comparison, which is generally a bad idea with floating-point types. - * In order to check for equality between two vectors or matrices with floating-point coefficients, it is - * generally a far better idea to use a fuzzy comparison as provided by isApprox() and - * isMuchSmallerThan(). - * - * \sa cwiseEqual(const MatrixBase &) const - */ -EIGEN_DEVICE_FUNC -inline const CwiseScalarEqualReturnType -cwiseEqual(const Scalar& s) const -{ - return CwiseScalarEqualReturnType(derived(), Derived::Constant(rows(), cols(), s), internal::scalar_cmp_op()); -} diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h deleted file mode 100644 index b1be3d566c6..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/Eigen/src/plugins/MatrixCwiseUnaryOps.h +++ /dev/null @@ -1,85 +0,0 @@ -// This file is part of Eigen, a lightweight C++ template library -// for linear algebra. -// -// Copyright (C) 2008-2009 Gael Guennebaud -// Copyright (C) 2006-2008 Benoit Jacob -// -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. - -// This file is included into the body of the base classes supporting matrix specific coefficient-wise functions. -// This include MatrixBase and SparseMatrixBase. - - -typedef CwiseUnaryOp, const Derived> CwiseAbsReturnType; -typedef CwiseUnaryOp, const Derived> CwiseAbs2ReturnType; -typedef CwiseUnaryOp, const Derived> CwiseSqrtReturnType; -typedef CwiseUnaryOp, const Derived> CwiseSignReturnType; -typedef CwiseUnaryOp, const Derived> CwiseInverseReturnType; - -/// \returns an expression of the coefficient-wise absolute value of \c *this -/// -/// Example: \include MatrixBase_cwiseAbs.cpp -/// Output: \verbinclude MatrixBase_cwiseAbs.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseAbs,absolute value) -/// -/// \sa cwiseAbs2() -/// -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseAbsReturnType -cwiseAbs() const { return CwiseAbsReturnType(derived()); } - -/// \returns an expression of the coefficient-wise squared absolute value of \c *this -/// -/// Example: \include MatrixBase_cwiseAbs2.cpp -/// Output: \verbinclude MatrixBase_cwiseAbs2.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseAbs2,squared absolute value) -/// -/// \sa cwiseAbs() -/// -EIGEN_DEVICE_FUNC -EIGEN_STRONG_INLINE const CwiseAbs2ReturnType -cwiseAbs2() const { return CwiseAbs2ReturnType(derived()); } - -/// \returns an expression of the coefficient-wise square root of *this. -/// -/// Example: \include MatrixBase_cwiseSqrt.cpp -/// Output: \verbinclude MatrixBase_cwiseSqrt.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseSqrt,square-root) -/// -/// \sa cwisePow(), cwiseSquare() -/// -EIGEN_DEVICE_FUNC -inline const CwiseSqrtReturnType -cwiseSqrt() const { return CwiseSqrtReturnType(derived()); } - -/// \returns an expression of the coefficient-wise signum of *this. -/// -/// Example: \include MatrixBase_cwiseSign.cpp -/// Output: \verbinclude MatrixBase_cwiseSign.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseSign,sign function) -/// -EIGEN_DEVICE_FUNC -inline const CwiseSignReturnType -cwiseSign() const { return CwiseSignReturnType(derived()); } - - -/// \returns an expression of the coefficient-wise inverse of *this. -/// -/// Example: \include MatrixBase_cwiseInverse.cpp -/// Output: \verbinclude MatrixBase_cwiseInverse.out -/// -EIGEN_DOC_UNARY_ADDONS(cwiseInverse,inverse) -/// -/// \sa cwiseProduct() -/// -EIGEN_DEVICE_FUNC -inline const CwiseInverseReturnType -cwiseInverse() const { return CwiseInverseReturnType(derived()); } - - diff --git a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/README.md b/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/README.md deleted file mode 100644 index 4654a81c372..00000000000 --- a/third_party/cpgfunctionEP-0.4/third_party/eigen-3.3.9/README.md +++ /dev/null @@ -1,3 +0,0 @@ -**Eigen is a C++ template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms.** - -For more information go to http://eigen.tuxfamily.org/. From 68261e1c29e4d364267aae6713a9014d6f3ef437 Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Fri, 30 Jul 2021 19:09:19 -0500 Subject: [PATCH 67/77] Make EP cmake link to bumped cpgfunctionEP folder --- CMakeLists.txt | 4 ++-- third_party/CMakeLists.txt | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/CMakeLists.txt b/CMakeLists.txt index 5fc851f5f7d..d65d9f2a1f0 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -265,8 +265,8 @@ target_include_directories(project_options SYSTEM INTERFACE "${kiva_SOURCE_DIR}/ add_subdirectory(scripts/dev/generate_embeddable_epJSON_schema) set_target_properties(generate_embeddedable_epJSON_schema PROPERTIES FOLDER "Internal") -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.3.2) -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.3.2/include) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4/include) if(OPENGL_FOUND) set(BUILD_PENUMBRA_TESTING diff --git a/third_party/CMakeLists.txt b/third_party/CMakeLists.txt index 333b1308e21..895a89d7685 100644 --- a/third_party/CMakeLists.txt +++ b/third_party/CMakeLists.txt @@ -124,8 +124,8 @@ add_subdirectory(re2) set_target_properties(re2 PROPERTIES FOLDER ThirdParty/re2) add_subdirectory(fmt-6.1.2) set_target_properties(fmt PROPERTIES FOLDER ThirdParty/fmt) -add_subdirectory(cpgfunctionEP-0.3.2) -set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.3.2) +add_subdirectory(cpgfunctionEP-0.4) +set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.4) # Kiva include(cmake/kiva.cmake) From 33a28cd8b81991fadc680f34157ae704aea3b8ec Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Mon, 16 Aug 2021 17:27:39 -0500 Subject: [PATCH 68/77] Bump Eigen library to 3.3.9 --- third_party/eigen/CMakeLists.txt | 53 +- third_party/eigen/COPYING.MINPACK | 104 +- third_party/eigen/CTestConfig.cmake | 6 +- third_party/eigen/Eigen/Core | 7 +- third_party/eigen/Eigen/Eigenvalues | 4 +- third_party/eigen/Eigen/Geometry | 4 +- third_party/eigen/Eigen/QR | 4 +- third_party/eigen/Eigen/SparseQR | 1 - third_party/eigen/Eigen/src/Core/ArrayBase.h | 4 +- .../eigen/Eigen/src/Core/CwiseUnaryView.h | 2 + third_party/eigen/Eigen/src/Core/DenseBase.h | 7 +- .../eigen/Eigen/src/Core/DenseStorage.h | 6 +- .../eigen/Eigen/src/Core/GenericPacketMath.h | 9 +- third_party/eigen/Eigen/src/Core/MapBase.h | 5 + .../eigen/Eigen/src/Core/MathFunctions.h | 16 +- third_party/eigen/Eigen/src/Core/MatrixBase.h | 3 +- .../eigen/Eigen/src/Core/PermutationMatrix.h | 28 - .../eigen/Eigen/src/Core/PlainObjectBase.h | 12 +- .../eigen/Eigen/src/Core/ProductEvaluators.h | 28 +- third_party/eigen/Eigen/src/Core/Ref.h | 5 +- .../eigen/Eigen/src/Core/SolveTriangular.h | 6 +- third_party/eigen/Eigen/src/Core/Transpose.h | 2 + .../eigen/Eigen/src/Core/Transpositions.h | 39 - .../eigen/Eigen/src/Core/TriangularMatrix.h | 8 +- .../src/Core/arch/AVX512/MathFunctions.h | 70 +- .../Eigen/src/Core/arch/AVX512/PacketMath.h | 717 +- .../eigen/Eigen/src/Core/arch/CUDA/Half.h | 1 + .../Eigen/src/Core/arch/CUDA/PacketMathHalf.h | 4 +- .../Eigen/src/Core/functors/UnaryFunctors.h | 2 +- .../Core/products/GeneralBlockPanelKernel.h | 7 +- .../src/Core/products/GeneralMatrixMatrix.h | 37 +- .../products/GeneralMatrixMatrixTriangular.h | 68 +- .../GeneralMatrixMatrixTriangular_BLAS.h | 10 +- .../Core/products/GeneralMatrixMatrix_BLAS.h | 6 +- .../Eigen/src/Core/products/Parallelizer.h | 9 +- .../Core/products/SelfadjointMatrixMatrix.h | 54 +- .../products/SelfadjointMatrixMatrix_BLAS.h | 24 +- .../src/Core/products/SelfadjointProduct.h | 4 +- .../Core/products/TriangularMatrixMatrix.h | 54 +- .../products/TriangularMatrixMatrix_BLAS.h | 26 +- .../Core/products/TriangularSolverMatrix.h | 62 +- .../products/TriangularSolverMatrix_BLAS.h | 12 +- .../eigen/Eigen/src/Core/util/BlasUtil.h | 115 +- .../src/Core/util/DisableStupidWarnings.h | 13 +- .../Eigen/src/Core/util/ForwardDeclarations.h | 6 +- .../eigen/Eigen/src/Core/util/Macros.h | 58 +- third_party/eigen/Eigen/src/Core/util/Meta.h | 34 + .../src/Core/util/ReenableStupidWarnings.h | 6 +- .../eigen/Eigen/src/Core/util/XprHelper.h | 17 + .../Eigen/src/Eigenvalues/ComplexSchur.h | 9 +- .../eigen/Eigen/src/Eigenvalues/RealSchur.h | 15 +- .../src/Eigenvalues/SelfAdjointEigenSolver.h | 7 +- .../eigen/Eigen/src/Geometry/Quaternion.h | 22 +- .../eigen/Eigen/src/Geometry/Scaling.h | 2 +- .../eigen/Eigen/src/Geometry/Transform.h | 4 +- .../eigen/Eigen/src/Geometry/Translation.h | 6 - .../eigen/Eigen/src/Geometry/Umeyama.h | 2 +- third_party/eigen/Eigen/src/LU/PartialPivLU.h | 5 +- .../eigen/Eigen/src/LU/arch/Inverse_SSE.h | 4 +- .../Eigen/src/PardisoSupport/PardisoSupport.h | 3 +- third_party/eigen/Eigen/src/SVD/BDCSVD.h | 61 +- third_party/eigen/Eigen/src/SVD/SVDBase.h | 2 +- .../src/SparseCholesky/SimplicialCholesky.h | 2 +- .../SparseCholesky/SimplicialCholesky_impl.h | 2 +- .../eigen/Eigen/src/SparseCore/AmbiVector.h | 7 +- .../Eigen/src/SparseCore/SparseCwiseUnaryOp.h | 2 + .../eigen/Eigen/src/SparseCore/SparseMatrix.h | 3 +- .../src/SparseCore/SparseSelfAdjointView.h | 4 +- .../eigen/Eigen/src/SparseCore/SparseView.h | 1 + .../eigen/Eigen/src/SparseLU/SparseLU.h | 4 +- .../eigen/Eigen/src/StlSupport/StdDeque.h | 6 +- .../Eigen/src/plugins/ArrayCwiseBinaryOps.h | 2 +- third_party/eigen/bench/bench_gemm.cpp | 5 +- third_party/eigen/blas/level3_impl.h | 182 +- third_party/eigen/cmake/EigenTesting.cmake | 2 + .../eigen/cmake/FindStandardMathLibrary.cmake | 7 +- third_party/eigen/debug/msvc/eigen.natvis | 470 +- .../eigen/debug/msvc/eigen_autoexp_part.dat | 590 +- third_party/eigen/doc/CMakeLists.txt | 9 +- .../doc/CustomizingEigen_CustomScalar.dox | 2 +- third_party/eigen/doc/Doxyfile.in | 17 +- third_party/eigen/doc/Pitfalls.dox | 84 +- .../eigen/doc/SparseQuickReference.dox | 2 +- third_party/eigen/doc/TopicLazyEvaluation.dox | 76 +- third_party/eigen/doc/TopicMultithreading.dox | 1 + third_party/eigen/doc/TutorialGeometry.dox | 4 +- third_party/eigen/doc/eigen_navtree_hacks.js | 5 +- .../eigen/doc/eigendoxy_footer.html.in | 13 - .../eigen/doc/eigendoxy_header.html.in | 3 + ...orial_BlockOperations_block_assignment.cpp | 2 +- third_party/eigen/lapack/CMakeLists.txt | 4 +- third_party/eigen/scripts/CMakeLists.txt | 8 +- third_party/eigen/test/CMakeLists.txt | 4 +- third_party/eigen/test/bdcsvd.cpp | 8 +- third_party/eigen/test/constructor.cpp | 14 + third_party/eigen/test/ctorleak.cpp | 20 +- .../eigen/test/eigensolver_generic.cpp | 2 +- third_party/eigen/test/exceptions.cpp | 4 +- third_party/eigen/test/fastmath.cpp | 34 +- third_party/eigen/test/geo_alignedbox.cpp | 3 +- third_party/eigen/test/geo_quaternion.cpp | 8 + .../eigen/test/geo_transformations.cpp | 61 +- third_party/eigen/test/inverse.cpp | 17 + third_party/eigen/test/main.h | 5 + third_party/eigen/test/numext.cpp | 5 +- third_party/eigen/test/packetmath.cpp | 17 +- third_party/eigen/test/product.h | 26 + third_party/eigen/test/product_large.cpp | 2 + third_party/eigen/test/product_mmtr.cpp | 10 + third_party/eigen/test/product_symm.cpp | 20 +- third_party/eigen/test/product_syrk.cpp | 11 + third_party/eigen/test/product_trmm.cpp | 12 +- third_party/eigen/test/product_trsolve.cpp | 26 + third_party/eigen/test/ref.cpp | 12 +- third_party/eigen/test/rvalue_types.cpp | 74 +- third_party/eigen/test/sparse_basic.cpp | 8 + third_party/eigen/test/stddeque.cpp | 16 +- third_party/eigen/test/stddeque_overload.cpp | 14 +- third_party/eigen/test/stdlist.cpp | 16 +- third_party/eigen/test/stdlist_overload.cpp | 14 +- third_party/eigen/test/stdvector.cpp | 6 +- third_party/eigen/test/stdvector_overload.cpp | 6 +- .../eigen/test/vectorization_logic.cpp | 8 + .../eigen/unsupported/Eigen/ArpackSupport | 4 +- .../src/Tensor/TensorContractionMapper.h | 2 + .../CXX11/src/Tensor/TensorDeviceThreadPool.h | 2 +- .../Eigen/CXX11/src/ThreadPool/EventCount.h | 2 +- .../Eigen/CXX11/src/ThreadPool/RunQueue.h | 2 +- .../eigen/unsupported/Eigen/Polynomials | 4 +- .../Eigen/src/AutoDiff/AutoDiffScalar.h | 28 +- .../ArpackSelfAdjointEigenSolver.h | 21 +- .../unsupported/Eigen/src/FFT/ei_fftw_impl.h | 2 + .../src/MatrixFunctions/MatrixExponential.h | 2 +- .../src/MatrixFunctions/MatrixSquareRoot.h | 18 +- .../Eigen/src/Polynomials/Companion.h | 57 +- .../Eigen/src/Polynomials/PolynomialSolver.h | 46 +- .../unsupported/Eigen/src/Splines/Spline.h | 2 +- .../test/NonLinearOptimization.cpp | 119 +- .../eigen/unsupported/test/autodiff.cpp | 16 + .../unsupported/test/matrix_function.cpp | 38 + .../eigen/unsupported/test/mpreal/mpreal.h | 6208 ++++++++--------- .../unsupported/test/polynomialsolver.cpp | 54 +- 142 files changed, 5666 insertions(+), 4884 deletions(-) diff --git a/third_party/eigen/CMakeLists.txt b/third_party/eigen/CMakeLists.txt index a3ed81c9dde..dbb9bcf220f 100644 --- a/third_party/eigen/CMakeLists.txt +++ b/third_party/eigen/CMakeLists.txt @@ -4,7 +4,7 @@ cmake_minimum_required(VERSION 2.8.5) # guard against in-source builds -if(${PROJECT_SOURCE_DIR} STREQUAL ${CMAKE_CURRENT_BINARY_DIR}) +if(${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_BINARY_DIR}) message(FATAL_ERROR "In-source builds not allowed. Please make a new directory (called a build directory) and run CMake from there. You may need to remove CMakeCache.txt. ") endif() @@ -42,11 +42,11 @@ set(EIGEN_MINOR_VERSION "${CMAKE_MATCH_1}") set(EIGEN_VERSION_NUMBER ${EIGEN_WORLD_VERSION}.${EIGEN_MAJOR_VERSION}.${EIGEN_MINOR_VERSION}) # if we are not in a mercurial clone -if(IS_DIRECTORY ${PROJECT_SOURCE_DIR}/.hg) +if(IS_DIRECTORY ${CMAKE_SOURCE_DIR}/.hg) # if the mercurial program is absent or this will leave the EIGEN_HG_CHANGESET string empty, # but won't stop CMake. - execute_process(COMMAND hg tip -R ${PROJECT_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_HGTIP_OUTPUT) - execute_process(COMMAND hg branch -R ${PROJECT_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_BRANCH_OUTPUT) + execute_process(COMMAND hg tip -R ${CMAKE_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_HGTIP_OUTPUT) + execute_process(COMMAND hg branch -R ${CMAKE_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_BRANCH_OUTPUT) endif() # if this is the default (aka development) branch, extract the mercurial changeset number from the hg tip output... @@ -156,9 +156,9 @@ if(NOT MSVC) endif() ei_add_cxx_compiler_flag("-pedantic") ei_add_cxx_compiler_flag("-Wall") - #ei_add_cxx_compiler_flag("-Wextra") + ei_add_cxx_compiler_flag("-Wextra") #ei_add_cxx_compiler_flag("-Weverything") # clang - + ei_add_cxx_compiler_flag("-Wundef") ei_add_cxx_compiler_flag("-Wcast-align") ei_add_cxx_compiler_flag("-Wchar-subscripts") @@ -173,29 +173,29 @@ if(NOT MSVC) ei_add_cxx_compiler_flag("-Wc++11-extensions") ei_add_cxx_compiler_flag("-Wdouble-promotion") # ei_add_cxx_compiler_flag("-Wconversion") - + # -Wshadow is insanely too strict with gcc, hopefully it will become usable with gcc 6 # if(NOT CMAKE_COMPILER_IS_GNUCXX OR (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER "5.0.0")) if(NOT CMAKE_COMPILER_IS_GNUCXX) ei_add_cxx_compiler_flag("-Wshadow") endif() - + ei_add_cxx_compiler_flag("-Wno-psabi") ei_add_cxx_compiler_flag("-Wno-variadic-macros") ei_add_cxx_compiler_flag("-Wno-long-long") - + ei_add_cxx_compiler_flag("-fno-check-new") ei_add_cxx_compiler_flag("-fno-common") ei_add_cxx_compiler_flag("-fstrict-aliasing") ei_add_cxx_compiler_flag("-wd981") # disable ICC's "operands are evaluated in unspecified order" remark ei_add_cxx_compiler_flag("-wd2304") # disable ICC's "warning #2304: non-explicit constructor with single argument may cause implicit type conversion" produced by -Wnon-virtual-dtor - - + + # The -ansi flag must be added last, otherwise it is also used as a linker flag by check_cxx_compiler_flag making it fails # Moreover we should not set both -strict-ansi and -ansi check_cxx_compiler_flag("-strict-ansi" COMPILER_SUPPORT_STRICTANSI) ei_add_cxx_compiler_flag("-Qunused-arguments") # disable clang warning: argument unused during compilation: '-ansi' - + if(COMPILER_SUPPORT_STRICTANSI) set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -strict-ansi") else() @@ -206,7 +206,7 @@ if(NOT MSVC) ei_add_cxx_compiler_flag("-pie") ei_add_cxx_compiler_flag("-fPIE") endif() - + set(CMAKE_REQUIRED_FLAGS "") option(EIGEN_TEST_SSE2 "Enable/Disable SSE2 in tests/examples" OFF) @@ -391,22 +391,27 @@ endif() if(EIGEN_INCLUDE_INSTALL_DIR AND NOT INCLUDE_INSTALL_DIR) set(INCLUDE_INSTALL_DIR ${EIGEN_INCLUDE_INSTALL_DIR} - CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed") + CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed") else() set(INCLUDE_INSTALL_DIR "${CMAKE_INSTALL_INCLUDEDIR}/eigen3" - CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed" + CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed" ) endif() set(CMAKEPACKAGE_INSTALL_DIR "${CMAKE_INSTALL_DATADIR}/eigen3/cmake" - CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen3Config.cmake is installed" + CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen3Config.cmake is installed" ) set(PKGCONFIG_INSTALL_DIR "${CMAKE_INSTALL_DATADIR}/pkgconfig" - CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where eigen3.pc is installed" + CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where eigen3.pc is installed" ) +foreach(var INCLUDE_INSTALL_DIR CMAKEPACKAGE_INSTALL_DIR PKGCONFIG_INSTALL_DIR) + if(IS_ABSOLUTE "${${var}}") + message(FATAL_ERROR "${var} must be relative to CMAKE_PREFIX_PATH. Got: ${${var}}") + endif() +endforeach() # similar to set_target_properties but append the property instead of overwriting it macro(ei_add_target_property target prop value) @@ -435,8 +440,8 @@ add_subdirectory(Eigen) add_subdirectory(doc EXCLUDE_FROM_ALL) -option(Eigen_BUILD_TESTING "Enable creation of Eigen tests." ON) -if(Eigen_BUILD_TESTING) +option(BUILD_TESTING "Enable creation of Eigen tests." ON) +if(BUILD_TESTING) include(EigenConfigureTesting) if(EIGEN_LEAVE_TEST_IN_ALL_TARGET) @@ -479,7 +484,7 @@ endif(NOT WIN32) configure_file(scripts/cdashtesting.cmake.in cdashtesting.cmake @ONLY) -if(Eigen_BUILD_TESTING) +if(BUILD_TESTING) ei_testing_print_summary() endif() @@ -576,11 +581,11 @@ if (NOT CMAKE_VERSION VERSION_LESS 3.0) else (NOT CMAKE_VERSION VERSION_LESS 3.0) # Fallback to legacy Eigen3Config.cmake without the imported target - + # If CMakePackageConfigHelpers module is available (CMake >= 2.8.8) - # create a relocatable Config file, otherwise leave the hardcoded paths + # create a relocatable Config file, otherwise leave the hardcoded paths include(CMakePackageConfigHelpers OPTIONAL RESULT_VARIABLE CPCH_PATH) - + if(CPCH_PATH) configure_package_config_file ( ${CMAKE_CURRENT_SOURCE_DIR}/cmake/Eigen3ConfigLegacy.cmake.in @@ -589,7 +594,7 @@ else (NOT CMAKE_VERSION VERSION_LESS 3.0) INSTALL_DESTINATION ${CMAKEPACKAGE_INSTALL_DIR} NO_CHECK_REQUIRED_COMPONENTS_MACRO # Eigen does not provide components ) - else() + else() # The PACKAGE_* variables are defined by the configure_package_config_file # but without it we define them manually to the hardcoded paths set(PACKAGE_INIT "") diff --git a/third_party/eigen/COPYING.MINPACK b/third_party/eigen/COPYING.MINPACK index ae7984daec9..11d8a9a6c34 100644 --- a/third_party/eigen/COPYING.MINPACK +++ b/third_party/eigen/COPYING.MINPACK @@ -1,52 +1,52 @@ -Minpack Copyright Notice (1999) University of Chicago. All rights reserved - -Redistribution and use in source and binary forms, with or -without modification, are permitted provided that the -following conditions are met: - -1. Redistributions of source code must retain the above -copyright notice, this list of conditions and the following -disclaimer. - -2. Redistributions in binary form must reproduce the above -copyright notice, this list of conditions and the following -disclaimer in the documentation and/or other materials -provided with the distribution. - -3. The end-user documentation included with the -redistribution, if any, must include the following -acknowledgment: - - "This product includes software developed by the - University of Chicago, as Operator of Argonne National - Laboratory. - -Alternately, this acknowledgment may appear in the software -itself, if and wherever such third-party acknowledgments -normally appear. - -4. WARRANTY DISCLAIMER. THE SOFTWARE IS SUPPLIED "AS IS" -WITHOUT WARRANTY OF ANY KIND. THE COPYRIGHT HOLDER, THE -UNITED STATES, THE UNITED STATES DEPARTMENT OF ENERGY, AND -THEIR EMPLOYEES: (1) DISCLAIM ANY WARRANTIES, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO ANY IMPLIED WARRANTIES -OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE -OR NON-INFRINGEMENT, (2) DO NOT ASSUME ANY LEGAL LIABILITY -OR RESPONSIBILITY FOR THE ACCURACY, COMPLETENESS, OR -USEFULNESS OF THE SOFTWARE, (3) DO NOT REPRESENT THAT USE OF -THE SOFTWARE WOULD NOT INFRINGE PRIVATELY OWNED RIGHTS, (4) -DO NOT WARRANT THAT THE SOFTWARE WILL FUNCTION -UNINTERRUPTED, THAT IT IS ERROR-FREE OR THAT ANY ERRORS WILL -BE CORRECTED. - -5. LIMITATION OF LIABILITY. IN NO EVENT WILL THE COPYRIGHT -HOLDER, THE UNITED STATES, THE UNITED STATES DEPARTMENT OF -ENERGY, OR THEIR EMPLOYEES: BE LIABLE FOR ANY INDIRECT, -INCIDENTAL, CONSEQUENTIAL, SPECIAL OR PUNITIVE DAMAGES OF -ANY KIND OR NATURE, INCLUDING BUT NOT LIMITED TO LOSS OF -PROFITS OR LOSS OF DATA, FOR ANY REASON WHATSOEVER, WHETHER -SUCH LIABILITY IS ASSERTED ON THE BASIS OF CONTRACT, TORT -(INCLUDING NEGLIGENCE OR STRICT LIABILITY), OR OTHERWISE, -EVEN IF ANY OF SAID PARTIES HAS BEEN WARNED OF THE -POSSIBILITY OF SUCH LOSS OR DAMAGES. - +Minpack Copyright Notice (1999) University of Chicago. All rights reserved + +Redistribution and use in source and binary forms, with or +without modification, are permitted provided that the +following conditions are met: + +1. Redistributions of source code must retain the above +copyright notice, this list of conditions and the following +disclaimer. + +2. Redistributions in binary form must reproduce the above +copyright notice, this list of conditions and the following +disclaimer in the documentation and/or other materials +provided with the distribution. + +3. The end-user documentation included with the +redistribution, if any, must include the following +acknowledgment: + + "This product includes software developed by the + University of Chicago, as Operator of Argonne National + Laboratory. + +Alternately, this acknowledgment may appear in the software +itself, if and wherever such third-party acknowledgments +normally appear. + +4. WARRANTY DISCLAIMER. THE SOFTWARE IS SUPPLIED "AS IS" +WITHOUT WARRANTY OF ANY KIND. THE COPYRIGHT HOLDER, THE +UNITED STATES, THE UNITED STATES DEPARTMENT OF ENERGY, AND +THEIR EMPLOYEES: (1) DISCLAIM ANY WARRANTIES, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO ANY IMPLIED WARRANTIES +OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE +OR NON-INFRINGEMENT, (2) DO NOT ASSUME ANY LEGAL LIABILITY +OR RESPONSIBILITY FOR THE ACCURACY, COMPLETENESS, OR +USEFULNESS OF THE SOFTWARE, (3) DO NOT REPRESENT THAT USE OF +THE SOFTWARE WOULD NOT INFRINGE PRIVATELY OWNED RIGHTS, (4) +DO NOT WARRANT THAT THE SOFTWARE WILL FUNCTION +UNINTERRUPTED, THAT IT IS ERROR-FREE OR THAT ANY ERRORS WILL +BE CORRECTED. + +5. LIMITATION OF LIABILITY. IN NO EVENT WILL THE COPYRIGHT +HOLDER, THE UNITED STATES, THE UNITED STATES DEPARTMENT OF +ENERGY, OR THEIR EMPLOYEES: BE LIABLE FOR ANY INDIRECT, +INCIDENTAL, CONSEQUENTIAL, SPECIAL OR PUNITIVE DAMAGES OF +ANY KIND OR NATURE, INCLUDING BUT NOT LIMITED TO LOSS OF +PROFITS OR LOSS OF DATA, FOR ANY REASON WHATSOEVER, WHETHER +SUCH LIABILITY IS ASSERTED ON THE BASIS OF CONTRACT, TORT +(INCLUDING NEGLIGENCE OR STRICT LIABILITY), OR OTHERWISE, +EVEN IF ANY OF SAID PARTIES HAS BEEN WARNED OF THE +POSSIBILITY OF SUCH LOSS OR DAMAGES. + diff --git a/third_party/eigen/CTestConfig.cmake b/third_party/eigen/CTestConfig.cmake index 0039bf8acd8..45de0e6fcaf 100644 --- a/third_party/eigen/CTestConfig.cmake +++ b/third_party/eigen/CTestConfig.cmake @@ -4,10 +4,10 @@ ## # The following are required to uses Dart and the Cdash dashboard ## ENABLE_TESTING() ## INCLUDE(CTest) -set(CTEST_PROJECT_NAME "Eigen 3.3") +set(CTEST_PROJECT_NAME "Eigen") set(CTEST_NIGHTLY_START_TIME "00:00:00 UTC") set(CTEST_DROP_METHOD "http") -set(CTEST_DROP_SITE "manao.inria.fr") -set(CTEST_DROP_LOCATION "/CDash/submit.php?project=Eigen+3.3") +set(CTEST_DROP_SITE "my.cdash.org") +set(CTEST_DROP_LOCATION "/submit.php?project=Eigen") set(CTEST_DROP_SITE_CDASH TRUE) diff --git a/third_party/eigen/Eigen/Core b/third_party/eigen/Eigen/Core index b923b8c000c..ac7c5b3004f 100644 --- a/third_party/eigen/Eigen/Core +++ b/third_party/eigen/Eigen/Core @@ -279,7 +279,10 @@ #include #include #include -#include +#include +#ifndef EIGEN_NO_IO + #include +#endif #include #include #include @@ -375,7 +378,9 @@ using std::ptrdiff_t; #if defined EIGEN_VECTORIZE_AVX512 #include "src/Core/arch/SSE/PacketMath.h" + #include "src/Core/arch/SSE/MathFunctions.h" #include "src/Core/arch/AVX/PacketMath.h" + #include "src/Core/arch/AVX/MathFunctions.h" #include "src/Core/arch/AVX512/PacketMath.h" #include "src/Core/arch/AVX512/MathFunctions.h" #elif defined EIGEN_VECTORIZE_AVX diff --git a/third_party/eigen/Eigen/Eigenvalues b/third_party/eigen/Eigen/Eigenvalues index f3f661b074b..7d6ac787bed 100644 --- a/third_party/eigen/Eigen/Eigenvalues +++ b/third_party/eigen/Eigen/Eigenvalues @@ -10,14 +10,14 @@ #include "Core" -#include "src/Core/util/DisableStupidWarnings.h" - #include "Cholesky" #include "Jacobi" #include "Householder" #include "LU" #include "Geometry" +#include "src/Core/util/DisableStupidWarnings.h" + /** \defgroup Eigenvalues_Module Eigenvalues module * * diff --git a/third_party/eigen/Eigen/Geometry b/third_party/eigen/Eigen/Geometry index 716d529529a..da88c03bbf5 100644 --- a/third_party/eigen/Eigen/Geometry +++ b/third_party/eigen/Eigen/Geometry @@ -10,12 +10,12 @@ #include "Core" -#include "src/Core/util/DisableStupidWarnings.h" - #include "SVD" #include "LU" #include +#include "src/Core/util/DisableStupidWarnings.h" + /** \defgroup Geometry_Module Geometry module * * This module provides support for: diff --git a/third_party/eigen/Eigen/QR b/third_party/eigen/Eigen/QR index c7e91446994..1be1863a1d5 100644 --- a/third_party/eigen/Eigen/QR +++ b/third_party/eigen/Eigen/QR @@ -10,12 +10,12 @@ #include "Core" -#include "src/Core/util/DisableStupidWarnings.h" - #include "Cholesky" #include "Jacobi" #include "Householder" +#include "src/Core/util/DisableStupidWarnings.h" + /** \defgroup QR_Module QR module * * diff --git a/third_party/eigen/Eigen/SparseQR b/third_party/eigen/Eigen/SparseQR index a6f3b7f7d75..f5fc5fa7fee 100644 --- a/third_party/eigen/Eigen/SparseQR +++ b/third_party/eigen/Eigen/SparseQR @@ -28,7 +28,6 @@ * */ -#include "OrderingMethods" #include "src/SparseCore/SparseColEtree.h" #include "src/SparseQR/SparseQR.h" diff --git a/third_party/eigen/Eigen/src/Core/ArrayBase.h b/third_party/eigen/Eigen/src/Core/ArrayBase.h index 3dbc7084cda..33f644e2132 100644 --- a/third_party/eigen/Eigen/src/Core/ArrayBase.h +++ b/third_party/eigen/Eigen/src/Core/ArrayBase.h @@ -153,8 +153,8 @@ template class ArrayBase // inline void evalTo(Dest& dst) const { dst = matrix(); } protected: - EIGEN_DEVICE_FUNC - ArrayBase() : Base() {} + EIGEN_DEFAULT_COPY_CONSTRUCTOR(ArrayBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(ArrayBase) private: explicit ArrayBase(Index); diff --git a/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h b/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h index 27103305629..5a30fa8df18 100644 --- a/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h +++ b/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h @@ -121,6 +121,8 @@ class CwiseUnaryViewImpl { return derived().nestedExpression().outerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); } + protected: + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(CwiseUnaryViewImpl) }; } // end namespace Eigen diff --git a/third_party/eigen/Eigen/src/Core/DenseBase.h b/third_party/eigen/Eigen/src/Core/DenseBase.h index 90066ae73fd..c55a68230cd 100644 --- a/third_party/eigen/Eigen/src/Core/DenseBase.h +++ b/third_party/eigen/Eigen/src/Core/DenseBase.h @@ -40,7 +40,7 @@ static inline void check_DenseIndex_is_signed() { */ template class DenseBase #ifndef EIGEN_PARSED_BY_DOXYGEN - : public DenseCoeffsBase + : public DenseCoeffsBase::value> #else : public DenseCoeffsBase #endif // not EIGEN_PARSED_BY_DOXYGEN @@ -71,7 +71,7 @@ template class DenseBase typedef Scalar value_type; typedef typename NumTraits::Real RealScalar; - typedef DenseCoeffsBase Base; + typedef DenseCoeffsBase::value> Base; using Base::derived; using Base::const_cast_derived; @@ -587,11 +587,12 @@ template class DenseBase } protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(DenseBase) /** Default constructor. Do nothing. */ EIGEN_DEVICE_FUNC DenseBase() { /* Just checks for self-consistency of the flags. - * Only do it when debugging Eigen, as this borders on paranoiac and could slow compilation down + * Only do it when debugging Eigen, as this borders on paranoia and could slow compilation down */ #ifdef EIGEN_INTERNAL_DEBUGGING EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, int(IsRowMajor)) diff --git a/third_party/eigen/Eigen/src/Core/DenseStorage.h b/third_party/eigen/Eigen/src/Core/DenseStorage.h index 7958feeb9c0..7d6d4e66d4b 100644 --- a/third_party/eigen/Eigen/src/Core/DenseStorage.h +++ b/third_party/eigen/Eigen/src/Core/DenseStorage.h @@ -404,7 +404,7 @@ template class DenseStorage(m_data, m_rows*m_cols); - if (size) + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; @@ -479,7 +479,7 @@ template class DenseStorage(m_data, _Rows*m_cols); - if (size) + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; @@ -553,7 +553,7 @@ template class DenseStorage(m_data, _Cols*m_rows); - if (size) + if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; diff --git a/third_party/eigen/Eigen/src/Core/GenericPacketMath.h b/third_party/eigen/Eigen/src/Core/GenericPacketMath.h index 029f8ac36fb..e594437791b 100644 --- a/third_party/eigen/Eigen/src/Core/GenericPacketMath.h +++ b/third_party/eigen/Eigen/src/Core/GenericPacketMath.h @@ -351,10 +351,7 @@ template EIGEN_DEVICE_FUNC inline Packet preverse(const Packet& /** \internal \returns \a a with real and imaginary part flipped (for complex type only) */ template EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet& a) { - // FIXME: uncomment the following in case we drop the internal imag and real functions. -// using std::imag; -// using std::real; - return Packet(imag(a),real(a)); + return Packet(a.imag(),a.real()); } /************************** @@ -524,10 +521,10 @@ inline void palign(PacketType& first, const PacketType& second) #ifndef __CUDACC__ template<> inline std::complex pmul(const std::complex& a, const std::complex& b) -{ return std::complex(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); } +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } template<> inline std::complex pmul(const std::complex& a, const std::complex& b) -{ return std::complex(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); } +{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } #endif diff --git a/third_party/eigen/Eigen/src/Core/MapBase.h b/third_party/eigen/Eigen/src/Core/MapBase.h index 668922ffccb..92c3b2818a4 100644 --- a/third_party/eigen/Eigen/src/Core/MapBase.h +++ b/third_party/eigen/Eigen/src/Core/MapBase.h @@ -182,6 +182,8 @@ template class MapBase #endif protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) template EIGEN_DEVICE_FUNC @@ -294,6 +296,9 @@ template class MapBase // In theory we could simply refer to Base:Base::operator=, but MSVC does not like Base::Base, // see bugs 821 and 920. using ReadOnlyMapBase::Base::operator=; + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) }; #undef EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS diff --git a/third_party/eigen/Eigen/src/Core/MathFunctions.h b/third_party/eigen/Eigen/src/Core/MathFunctions.h index b249ce0c8b0..01736c2a060 100644 --- a/third_party/eigen/Eigen/src/Core/MathFunctions.h +++ b/third_party/eigen/Eigen/src/Core/MathFunctions.h @@ -287,7 +287,7 @@ struct abs2_impl_default // IsComplex EIGEN_DEVICE_FUNC static inline RealScalar run(const Scalar& x) { - return real(x)*real(x) + imag(x)*imag(x); + return x.real()*x.real() + x.imag()*x.imag(); } }; @@ -313,14 +313,17 @@ struct abs2_retval ****************************************************************************/ template -struct norm1_default_impl +struct norm1_default_impl; + +template +struct norm1_default_impl { typedef typename NumTraits::Real RealScalar; EIGEN_DEVICE_FUNC static inline RealScalar run(const Scalar& x) { EIGEN_USING_STD_MATH(abs); - return abs(real(x)) + abs(imag(x)); + return abs(x.real()) + abs(x.imag()); } }; @@ -662,8 +665,8 @@ struct random_default_impl { static inline Scalar run(const Scalar& x, const Scalar& y) { - return Scalar(random(real(x), real(y)), - random(imag(x), imag(y))); + return Scalar(random(x.real(), y.real()), + random(x.imag(), y.imag())); } static inline Scalar run() { @@ -916,6 +919,9 @@ inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x) return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x); } +EIGEN_DEVICE_FUNC +inline bool abs2(bool x) { return x; } + template EIGEN_DEVICE_FUNC inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x) diff --git a/third_party/eigen/Eigen/src/Core/MatrixBase.h b/third_party/eigen/Eigen/src/Core/MatrixBase.h index e6c35907c39..f8bcc8c6f56 100644 --- a/third_party/eigen/Eigen/src/Core/MatrixBase.h +++ b/third_party/eigen/Eigen/src/Core/MatrixBase.h @@ -464,7 +464,8 @@ template class MatrixBase EIGEN_MATRIX_FUNCTION_1(MatrixComplexPowerReturnValue, pow, power to \c p, const std::complex& p) protected: - EIGEN_DEVICE_FUNC MatrixBase() : Base() {} + EIGEN_DEFAULT_COPY_CONSTRUCTOR(MatrixBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MatrixBase) private: EIGEN_DEVICE_FUNC explicit MatrixBase(int); diff --git a/third_party/eigen/Eigen/src/Core/PermutationMatrix.h b/third_party/eigen/Eigen/src/Core/PermutationMatrix.h index b1fb455b98c..47c06ba7706 100644 --- a/third_party/eigen/Eigen/src/Core/PermutationMatrix.h +++ b/third_party/eigen/Eigen/src/Core/PermutationMatrix.h @@ -87,17 +87,6 @@ class PermutationBase : public EigenBase return derived(); } - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** This is a special case of the templated operator=. Its purpose is to - * prevent a default operator= from hiding the templated operator=. - */ - Derived& operator=(const PermutationBase& other) - { - indices() = other.indices(); - return derived(); - } - #endif - /** \returns the number of rows */ inline Index rows() const { return Index(indices().size()); } @@ -333,12 +322,6 @@ class PermutationMatrix : public PermutationBase& other) : m_indices(other.indices()) {} - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** Standard copy constructor. Defined only to prevent a default copy constructor - * from hiding the other templated constructor */ - inline PermutationMatrix(const PermutationMatrix& other) : m_indices(other.indices()) {} - #endif - /** Generic constructor from expression of the indices. The indices * array has the meaning that the permutations sends each integer i to indices[i]. * @@ -373,17 +356,6 @@ class PermutationMatrix : public PermutationBase::type EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if::type* = 0) { - EIGEN_STATIC_ASSERT(bool(NumTraits::IsInteger) && - bool(NumTraits::IsInteger), + const bool t0_is_integer_alike = internal::is_valid_index_type::value; + const bool t1_is_integer_alike = internal::is_valid_index_type::value; + EIGEN_STATIC_ASSERT(t0_is_integer_alike && + t1_is_integer_alike, FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) resize(rows,cols); } @@ -773,9 +775,9 @@ class PlainObjectBase : public internal::dense_xpr_base::type && ((!internal::is_same::XprKind,ArrayXpr>::value || Base::SizeAtCompileTime==Dynamic)),T>::type* = 0) { // NOTE MSVC 2008 complains if we directly put bool(NumTraits::IsInteger) as the EIGEN_STATIC_ASSERT argument. - const bool is_integer = NumTraits::IsInteger; - EIGEN_UNUSED_VARIABLE(is_integer); - EIGEN_STATIC_ASSERT(is_integer, + const bool is_integer_alike = internal::is_valid_index_type::value; + EIGEN_UNUSED_VARIABLE(is_integer_alike); + EIGEN_STATIC_ASSERT(is_integer_alike, FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) resize(size); } diff --git a/third_party/eigen/Eigen/src/Core/ProductEvaluators.h b/third_party/eigen/Eigen/src/Core/ProductEvaluators.h index 9b99bd76962..bce1310c96f 100644 --- a/third_party/eigen/Eigen/src/Core/ProductEvaluators.h +++ b/third_party/eigen/Eigen/src/Core/ProductEvaluators.h @@ -396,7 +396,7 @@ struct generic_product_impl // but easier on the compiler side call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::assign_op()); } - + template static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { @@ -410,6 +410,32 @@ struct generic_product_impl // dst.noalias() -= lhs.lazyProduct(rhs); call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::sub_assign_op()); } + + // Catch "dst {,+,-}= (s*A)*B" and evaluate it lazily by moving out the scalar factor: + // dst {,+,-}= s * (A.lazyProduct(B)) + // This is a huge benefit for heap-allocated matrix types as it save one costly allocation. + // For them, this strategy is also faster than simply by-passing the heap allocation through + // stack allocation. + // For fixed sizes matrices, this is less obvious, it is sometimes x2 faster, but sometimes x3 slower, + // and the behavior depends also a lot on the compiler... so let's be conservative and enable them for dynamic-size only, + // that is when coming from generic_product_impl<...,GemmProduct> in file GeneralMatrixMatrix.h + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const CwiseBinaryOp, + const CwiseNullaryOp, Plain1>, Xpr2>& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lhs().functor().m_other * lhs.rhs().lazyProduct(rhs), func); + } + + // Here, we we always have LhsT==Lhs, but we need to make it a template type to make the above + // overload more specialized. + template + static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE + void eval_dynamic(Dst& dst, const LhsT& lhs, const Rhs& rhs, const Func &func) + { + call_assignment_no_alias(dst, lhs.lazyProduct(rhs), func); + } + // template // static inline void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) diff --git a/third_party/eigen/Eigen/src/Core/Ref.h b/third_party/eigen/Eigen/src/Core/Ref.h index 9c6e3c5d9b0..17a1496b842 100644 --- a/third_party/eigen/Eigen/src/Core/Ref.h +++ b/third_party/eigen/Eigen/src/Core/Ref.h @@ -28,12 +28,13 @@ struct traits > template struct match { enum { + IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime, HasDirectAccess = internal::has_direct_access::ret, - StorageOrderMatch = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), + StorageOrderMatch = IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), InnerStrideMatch = int(StrideType::InnerStrideAtCompileTime)==int(Dynamic) || int(StrideType::InnerStrideAtCompileTime)==int(Derived::InnerStrideAtCompileTime) || (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1), - OuterStrideMatch = Derived::IsVectorAtCompileTime + OuterStrideMatch = IsVectorAtCompileTime || int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime), // NOTE, this indirection of evaluator::Alignment is needed // to workaround a very strange bug in MSVC related to the instantiation diff --git a/third_party/eigen/Eigen/src/Core/SolveTriangular.h b/third_party/eigen/Eigen/src/Core/SolveTriangular.h index 4652e2e19ff..fd0acb1a58e 100644 --- a/third_party/eigen/Eigen/src/Core/SolveTriangular.h +++ b/third_party/eigen/Eigen/src/Core/SolveTriangular.h @@ -19,7 +19,7 @@ namespace internal { template struct triangular_solve_vector; -template +template struct triangular_solve_matrix; // small helper struct extracting some traits on the underlying solver operation @@ -98,8 +98,8 @@ struct triangular_solver_selector BlockingType blocking(rhs.rows(), rhs.cols(), size, 1, false); triangular_solve_matrix - ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.outerStride(), blocking); + (Rhs::Flags&RowMajorBit) ? RowMajor : ColMajor, Rhs::InnerStrideAtCompileTime> + ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.innerStride(), rhs.outerStride(), blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/Transpose.h b/third_party/eigen/Eigen/src/Core/Transpose.h index 79b767bccae..960dc4510c7 100644 --- a/third_party/eigen/Eigen/src/Core/Transpose.h +++ b/third_party/eigen/Eigen/src/Core/Transpose.h @@ -146,6 +146,8 @@ template class TransposeImpl { return derived().nestedExpression().coeffRef(index); } + protected: + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TransposeImpl) }; /** \returns an expression of the transpose of *this. diff --git a/third_party/eigen/Eigen/src/Core/Transpositions.h b/third_party/eigen/Eigen/src/Core/Transpositions.h index 86da5af5936..7718625e80b 100644 --- a/third_party/eigen/Eigen/src/Core/Transpositions.h +++ b/third_party/eigen/Eigen/src/Core/Transpositions.h @@ -33,17 +33,6 @@ class TranspositionsBase indices() = other.indices(); return derived(); } - - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** This is a special case of the templated operator=. Its purpose is to - * prevent a default operator= from hiding the templated operator=. - */ - Derived& operator=(const TranspositionsBase& other) - { - indices() = other.indices(); - return derived(); - } - #endif /** \returns the number of transpositions */ Index size() const { return indices().size(); } @@ -171,12 +160,6 @@ class Transpositions : public TranspositionsBase& other) : m_indices(other.indices()) {} - #ifndef EIGEN_PARSED_BY_DOXYGEN - /** Standard copy constructor. Defined only to prevent a default copy constructor - * from hiding the other templated constructor */ - inline Transpositions(const Transpositions& other) : m_indices(other.indices()) {} - #endif - /** Generic constructor from expression of the transposition indices. */ template explicit inline Transpositions(const MatrixBase& indices) : m_indices(indices) @@ -189,17 +172,6 @@ class Transpositions : public TranspositionsBase class TriangularView explicit inline TriangularView(MatrixType& matrix) : m_matrix(matrix) {} - using Base::operator=; - TriangularView& operator=(const TriangularView &other) - { return Base::operator=(other); } + EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TriangularView) /** \copydoc EigenBase::rows() */ EIGEN_DEVICE_FUNC @@ -544,6 +542,10 @@ template class TriangularViewImpl<_Mat template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TriangularViewType& _assignProduct(const ProductType& prod, const Scalar& alpha, bool beta); + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(TriangularViewImpl) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TriangularViewImpl) + }; /*************************************************************************** diff --git a/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h b/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h index 9c1717f76dd..b259c1e1f92 100644 --- a/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h +++ b/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h @@ -29,6 +29,7 @@ namespace internal { #define _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(NAME, X) \ const Packet8d p8d_##NAME = _mm512_castsi512_pd(_mm512_set1_epi64(X)) + // Natural logarithm // Computes log(x) as log(2^e * m) = C*e + log(m), where the constant C =log(2) // and m is in the range [sqrt(1/2),sqrt(2)). In this range, the logarithm can @@ -47,6 +48,7 @@ plog(const Packet16f& _x) { // The smallest non denormalized float number. _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(min_norm_pos, 0x00800000); _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(minus_inf, 0xff800000); + _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(pos_inf, 0x7f800000); _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(nan, 0x7fc00000); // Polynomial coefficients. @@ -64,11 +66,9 @@ plog(const Packet16f& _x) { _EIGEN_DECLARE_CONST_Packet16f(cephes_log_q2, 0.693359375f); // invalid_mask is set to true when x is NaN - __mmask16 invalid_mask = - _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_NGE_UQ); - __mmask16 iszero_mask = - _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_EQ_UQ); - + __mmask16 invalid_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_NGE_UQ); + __mmask16 iszero_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_EQ_OQ); + // Truncate input values to the minimum positive normal. x = pmax(x, p16f_min_norm_pos); @@ -118,11 +118,18 @@ plog(const Packet16f& _x) { x = padd(x, y); x = padd(x, y2); - // Filter out invalid inputs, i.e. negative arg will be NAN, 0 will be -INF. + __mmask16 pos_inf_mask = _mm512_cmp_ps_mask(_x,p16f_pos_inf,_CMP_EQ_OQ); + // Filter out invalid inputs, i.e.: + // - negative arg will be NAN, + // - 0 will be -INF. + // - +INF will be +INF return _mm512_mask_blend_ps(iszero_mask, - _mm512_mask_blend_ps(invalid_mask, x, p16f_nan), - p16f_minus_inf); + _mm512_mask_blend_ps(invalid_mask, + _mm512_mask_blend_ps(pos_inf_mask,x,p16f_pos_inf), + p16f_nan), + p16f_minus_inf); } + #endif // Exponential function. Works by writing "x = m*log(2) + r" where @@ -258,48 +265,39 @@ pexp(const Packet8d& _x) { template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f psqrt(const Packet16f& _x) { - _EIGEN_DECLARE_CONST_Packet16f(one_point_five, 1.5f); - _EIGEN_DECLARE_CONST_Packet16f(minus_half, -0.5f); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(flt_min, 0x00800000); + Packet16f neg_half = pmul(_x, pset1(-.5f)); + __mmask16 denormal_mask = _mm512_kand( + _mm512_cmp_ps_mask(_x, pset1((std::numeric_limits::min)()), + _CMP_LT_OQ), + _mm512_cmp_ps_mask(_x, _mm512_setzero_ps(), _CMP_GE_OQ)); - Packet16f neg_half = pmul(_x, p16f_minus_half); - - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - __mmask16 non_zero_mask = _mm512_cmp_ps_mask(_x, p16f_flt_min, _CMP_GE_OQ); - Packet16f x = _mm512_mask_blend_ps(non_zero_mask, _mm512_setzero_ps(), _mm512_rsqrt14_ps(_x)); + Packet16f x = _mm512_rsqrt14_ps(_x); // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p16f_one_point_five)); + x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5f))); - // Multiply the original _x by it's reciprocal square root to extract the - // square root. - return pmul(_x, x); + // Flush results for denormals to zero. + return _mm512_mask_blend_ps(denormal_mask, pmul(_x,x), _mm512_setzero_ps()); } template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8d psqrt(const Packet8d& _x) { - _EIGEN_DECLARE_CONST_Packet8d(one_point_five, 1.5); - _EIGEN_DECLARE_CONST_Packet8d(minus_half, -0.5); - _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(dbl_min, 0x0010000000000000LL); + Packet8d neg_half = pmul(_x, pset1(-.5)); + __mmask16 denormal_mask = _mm512_kand( + _mm512_cmp_pd_mask(_x, pset1((std::numeric_limits::min)()), + _CMP_LT_OQ), + _mm512_cmp_pd_mask(_x, _mm512_setzero_pd(), _CMP_GE_OQ)); - Packet8d neg_half = pmul(_x, p8d_minus_half); + Packet8d x = _mm512_rsqrt14_pd(_x); - // select only the inverse sqrt of positive normal inputs (denormals are - // flushed to zero and cause infs as well). - __mmask8 non_zero_mask = _mm512_cmp_pd_mask(_x, p8d_dbl_min, _CMP_GE_OQ); - Packet8d x = _mm512_mask_blend_pd(non_zero_mask, _mm512_setzero_pd(), _mm512_rsqrt14_pd(_x)); - - // Do a first step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); + // Do a single step of Newton's iteration. + x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); // Do a second step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); + x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); - // Multiply the original _x by it's reciprocal square root to extract the - // square root. - return pmul(_x, x); + return _mm512_mask_blend_pd(denormal_mask, pmul(_x,x), _mm512_setzero_pd()); } #else template <> diff --git a/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h b/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h index 5adddc7aec4..000b7762ff2 100644 --- a/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h +++ b/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h @@ -19,10 +19,10 @@ namespace internal { #endif #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS (2*sizeof(void*)) +#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32 #endif -#ifdef __FMA__ +#ifdef EIGEN_VECTORIZE_FMA #ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD #endif @@ -54,13 +54,14 @@ template<> struct packet_traits : default_packet_traits AlignedOnScalar = 1, size = 16, HasHalfPacket = 1, -#if EIGEN_GNUC_AT_LEAST(5, 3) + HasBlend = 0, +#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) #ifdef EIGEN_VECTORIZE_AVX512DQ HasLog = 1, #endif HasExp = 1, - HasSqrt = 1, - HasRsqrt = 1, + HasSqrt = EIGEN_FAST_MATH, + HasRsqrt = EIGEN_FAST_MATH, #endif HasDiv = 1 }; @@ -74,8 +75,8 @@ template<> struct packet_traits : default_packet_traits AlignedOnScalar = 1, size = 8, HasHalfPacket = 1, -#if EIGEN_GNUC_AT_LEAST(5, 3) - HasSqrt = 1, +#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) + HasSqrt = EIGEN_FAST_MATH, HasRsqrt = EIGEN_FAST_MATH, #endif HasDiv = 1 @@ -98,6 +99,7 @@ template <> struct unpacket_traits { typedef float type; typedef Packet8f half; + typedef Packet16i integer_packet; enum { size = 16, alignment=Aligned64 }; }; template <> @@ -132,7 +134,7 @@ EIGEN_STRONG_INLINE Packet16f pload1(const float* from) { } template <> EIGEN_STRONG_INLINE Packet8d pload1(const double* from) { - return _mm512_broadcastsd_pd(_mm_load_pd1(from)); + return _mm512_set1_pd(*from); } template <> @@ -158,6 +160,11 @@ EIGEN_STRONG_INLINE Packet8d padd(const Packet8d& a, const Packet8d& b) { return _mm512_add_pd(a, b); } +template <> +EIGEN_STRONG_INLINE Packet16i padd(const Packet16i& a, + const Packet16i& b) { + return _mm512_add_epi32(a, b); +} template <> EIGEN_STRONG_INLINE Packet16f psub(const Packet16f& a, @@ -169,6 +176,11 @@ EIGEN_STRONG_INLINE Packet8d psub(const Packet8d& a, const Packet8d& b) { return _mm512_sub_pd(a, b); } +template <> +EIGEN_STRONG_INLINE Packet16i psub(const Packet16i& a, + const Packet16i& b) { + return _mm512_sub_epi32(a, b); +} template <> EIGEN_STRONG_INLINE Packet16f pnegate(const Packet16f& a) { @@ -202,6 +214,11 @@ EIGEN_STRONG_INLINE Packet8d pmul(const Packet8d& a, const Packet8d& b) { return _mm512_mul_pd(a, b); } +template <> +EIGEN_STRONG_INLINE Packet16i pmul(const Packet16i& a, + const Packet16i& b) { + return _mm512_mul_epi32(a, b); +} template <> EIGEN_STRONG_INLINE Packet16f pdiv(const Packet16f& a, @@ -214,7 +231,7 @@ EIGEN_STRONG_INLINE Packet8d pdiv(const Packet8d& a, return _mm512_div_pd(a, b); } -#ifdef __FMA__ +#ifdef EIGEN_VECTORIZE_FMA template <> EIGEN_STRONG_INLINE Packet16f pmadd(const Packet16f& a, const Packet16f& b, const Packet16f& c) { @@ -230,49 +247,82 @@ EIGEN_STRONG_INLINE Packet8d pmadd(const Packet8d& a, const Packet8d& b, template <> EIGEN_STRONG_INLINE Packet16f pmin(const Packet16f& a, const Packet16f& b) { - return _mm512_min_ps(a, b); + // Arguments are reversed to match NaN propagation behavior of std::min. + return _mm512_min_ps(b, a); } template <> EIGEN_STRONG_INLINE Packet8d pmin(const Packet8d& a, const Packet8d& b) { - return _mm512_min_pd(a, b); + // Arguments are reversed to match NaN propagation behavior of std::min. + return _mm512_min_pd(b, a); } template <> EIGEN_STRONG_INLINE Packet16f pmax(const Packet16f& a, const Packet16f& b) { - return _mm512_max_ps(a, b); + // Arguments are reversed to match NaN propagation behavior of std::max. + return _mm512_max_ps(b, a); } template <> EIGEN_STRONG_INLINE Packet8d pmax(const Packet8d& a, const Packet8d& b) { - return _mm512_max_pd(a, b); + // Arguments are reversed to match NaN propagation behavior of std::max. + return _mm512_max_pd(b, a); } -template <> -EIGEN_STRONG_INLINE Packet16f pand(const Packet16f& a, - const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_and_ps(a, b); +template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { return _mm512_extractf32x8_ps(x,I_); } +template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { return _mm512_extractf64x2_pd(x,I_); } +EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { return _mm512_insertf32x8(_mm512_castps256_ps512(a),b,1); } #else - Packet16f res = _mm512_undefined_ps(); - Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); - Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); - res = _mm512_insertf32x4(res, _mm_and_ps(lane0_a, lane0_b), 0); +// AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 +template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { + return _mm256_castsi256_ps(_mm512_extracti64x4_epi64( _mm512_castps_si512(x),I_)); +} - Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); - Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); - res = _mm512_insertf32x4(res, _mm_and_ps(lane1_a, lane1_b), 1); +// AVX512F does not define _mm512_extractf64x2_pd to extract _m128 from _m512 +template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { + return _mm_castsi128_pd(_mm512_extracti32x4_epi32( _mm512_castpd_si512(x),I_)); +} - Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); - Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); - res = _mm512_insertf32x4(res, _mm_and_ps(lane2_a, lane2_b), 2); +EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { + return _mm512_castsi512_ps(_mm512_inserti64x4(_mm512_castsi256_si512(_mm256_castps_si256(a)), + _mm256_castps_si256(b),1)); +} +#endif - Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); - Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); - res = _mm512_insertf32x4(res, _mm_and_ps(lane3_a, lane3_b), 3); +// Helper function for bit packing snippet of low precision comparison. +// It packs the flags from 32x16 to 16x16. +EIGEN_STRONG_INLINE __m256i Pack32To16(Packet16f rf) { + // Split data into small pieces and handle with AVX instructions + // to guarantee internal order of vector. + // Operation: + // dst[15:0] := Saturate16(rf[31:0]) + // dst[31:16] := Saturate16(rf[63:32]) + // ... + // dst[255:240] := Saturate16(rf[255:224]) + __m256i lo = _mm256_castps_si256(extract256<0>(rf)); + __m256i hi = _mm256_castps_si256(extract256<1>(rf)); + __m128i result_lo = _mm_packs_epi32(_mm256_extractf128_si256(lo, 0), + _mm256_extractf128_si256(lo, 1)); + __m128i result_hi = _mm_packs_epi32(_mm256_extractf128_si256(hi, 0), + _mm256_extractf128_si256(hi, 1)); + return _mm256_insertf128_si256(_mm256_castsi128_si256(result_lo), result_hi, 1); +} - return res; +template <> +EIGEN_STRONG_INLINE Packet16i pand(const Packet16i& a, + const Packet16i& b) { + return _mm512_and_si512(a,b); +} + +template <> +EIGEN_STRONG_INLINE Packet16f pand(const Packet16f& a, + const Packet16f& b) { +#ifdef EIGEN_VECTORIZE_AVX512DQ + return _mm512_and_ps(a, b); +#else + return _mm512_castsi512_ps(pand(_mm512_castps_si512(a),_mm512_castps_si512(b))); #endif } template <> @@ -288,35 +338,21 @@ EIGEN_STRONG_INLINE Packet8d pand(const Packet8d& a, Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - res = _mm512_insertf64x4(res, _mm256_and_pd(lane1_a, lane1_b), 1); - - return res; + return _mm512_insertf64x4(res, _mm256_and_pd(lane1_a, lane1_b), 1); #endif } + +template <> +EIGEN_STRONG_INLINE Packet16i por(const Packet16i& a, const Packet16i& b) { + return _mm512_or_si512(a, b); +} + template <> -EIGEN_STRONG_INLINE Packet16f por(const Packet16f& a, - const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16f por(const Packet16f& a, const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_or_ps(a, b); #else - Packet16f res = _mm512_undefined_ps(); - Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); - Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); - res = _mm512_insertf32x4(res, _mm_or_ps(lane0_a, lane0_b), 0); - - Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); - Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); - res = _mm512_insertf32x4(res, _mm_or_ps(lane1_a, lane1_b), 1); - - Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); - Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); - res = _mm512_insertf32x4(res, _mm_or_ps(lane2_a, lane2_b), 2); - - Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); - Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); - res = _mm512_insertf32x4(res, _mm_or_ps(lane3_a, lane3_b), 3); - - return res; + return _mm512_castsi512_ps(por(_mm512_castps_si512(a),_mm512_castps_si512(b))); #endif } @@ -326,107 +362,65 @@ EIGEN_STRONG_INLINE Packet8d por(const Packet8d& a, #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_or_pd(a, b); #else - Packet8d res = _mm512_undefined_pd(); - Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); - Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); - res = _mm512_insertf64x4(res, _mm256_or_pd(lane0_a, lane0_b), 0); - - Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); - Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - res = _mm512_insertf64x4(res, _mm256_or_pd(lane1_a, lane1_b), 1); - - return res; + return _mm512_castsi512_pd(por(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); #endif } template <> -EIGEN_STRONG_INLINE Packet16f pxor(const Packet16f& a, - const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16i pxor(const Packet16i& a, const Packet16i& b) { + return _mm512_xor_si512(a, b); +} + +template <> +EIGEN_STRONG_INLINE Packet16f pxor(const Packet16f& a, const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_xor_ps(a, b); #else - Packet16f res = _mm512_undefined_ps(); - Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); - Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); - res = _mm512_insertf32x4(res, _mm_xor_ps(lane0_a, lane0_b), 0); - - Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); - Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); - res = _mm512_insertf32x4(res, _mm_xor_ps(lane1_a, lane1_b), 1); - - Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); - Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); - res = _mm512_insertf32x4(res, _mm_xor_ps(lane2_a, lane2_b), 2); - - Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); - Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); - res = _mm512_insertf32x4(res, _mm_xor_ps(lane3_a, lane3_b), 3); - - return res; + return _mm512_castsi512_ps(pxor(_mm512_castps_si512(a),_mm512_castps_si512(b))); #endif } + template <> -EIGEN_STRONG_INLINE Packet8d pxor(const Packet8d& a, - const Packet8d& b) { +EIGEN_STRONG_INLINE Packet8d pxor(const Packet8d& a, const Packet8d& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_xor_pd(a, b); #else - Packet8d res = _mm512_undefined_pd(); - Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); - Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); - res = _mm512_insertf64x4(res, _mm256_xor_pd(lane0_a, lane0_b), 0); - - Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); - Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - res = _mm512_insertf64x4(res, _mm256_xor_pd(lane1_a, lane1_b), 1); - - return res; + return _mm512_castsi512_pd(pxor(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); #endif } template <> -EIGEN_STRONG_INLINE Packet16f pandnot(const Packet16f& a, - const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16i pandnot(const Packet16i& a, const Packet16i& b) { + return _mm512_andnot_si512(b, a); +} + +template <> +EIGEN_STRONG_INLINE Packet16f pandnot(const Packet16f& a, const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_ps(a, b); + return _mm512_andnot_ps(b, a); #else - Packet16f res = _mm512_undefined_ps(); - Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); - Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); - res = _mm512_insertf32x4(res, _mm_andnot_ps(lane0_a, lane0_b), 0); - - Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); - Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); - res = _mm512_insertf32x4(res, _mm_andnot_ps(lane1_a, lane1_b), 1); - - Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); - Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); - res = _mm512_insertf32x4(res, _mm_andnot_ps(lane2_a, lane2_b), 2); - - Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); - Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); - res = _mm512_insertf32x4(res, _mm_andnot_ps(lane3_a, lane3_b), 3); - - return res; + return _mm512_castsi512_ps(pandnot(_mm512_castps_si512(a),_mm512_castps_si512(b))); #endif } template <> -EIGEN_STRONG_INLINE Packet8d pandnot(const Packet8d& a, - const Packet8d& b) { +EIGEN_STRONG_INLINE Packet8d pandnot(const Packet8d& a,const Packet8d& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_pd(a, b); + return _mm512_andnot_pd(b, a); #else - Packet8d res = _mm512_undefined_pd(); - Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); - Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); - res = _mm512_insertf64x4(res, _mm256_andnot_pd(lane0_a, lane0_b), 0); + return _mm512_castsi512_pd(pandnot(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); +#endif +} - Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); - Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - res = _mm512_insertf64x4(res, _mm256_andnot_pd(lane1_a, lane1_b), 1); +template EIGEN_STRONG_INLINE Packet16i parithmetic_shift_right(Packet16i a) { + return _mm512_srai_epi32(a, N); +} - return res; -#endif +template EIGEN_STRONG_INLINE Packet16i plogical_shift_right(Packet16i a) { + return _mm512_srli_epi32(a, N); +} + +template EIGEN_STRONG_INLINE Packet16i plogical_shift_left(Packet16i a) { + return _mm512_slli_epi32(a, N); } template <> @@ -461,75 +455,55 @@ EIGEN_STRONG_INLINE Packet16i ploadu(const int* from) { // {a0, a0 a1, a1, a2, a2, a3, a3, a4, a4, a5, a5, a6, a6, a7, a7} template <> EIGEN_STRONG_INLINE Packet16f ploaddup(const float* from) { - Packet8f lane0 = _mm256_broadcast_ps((const __m128*)(const void*)from); - // mimic an "inplace" permutation of the lower 128bits using a blend - lane0 = _mm256_blend_ps( - lane0, _mm256_castps128_ps256(_mm_permute_ps( - _mm256_castps256_ps128(lane0), _MM_SHUFFLE(1, 0, 1, 0))), - 15); - // then we can perform a consistent permutation on the global register to get - // everything in shape: - lane0 = _mm256_permute_ps(lane0, _MM_SHUFFLE(3, 3, 2, 2)); - - Packet8f lane1 = _mm256_broadcast_ps((const __m128*)(const void*)(from + 4)); - // mimic an "inplace" permutation of the lower 128bits using a blend - lane1 = _mm256_blend_ps( - lane1, _mm256_castps128_ps256(_mm_permute_ps( - _mm256_castps256_ps128(lane1), _MM_SHUFFLE(1, 0, 1, 0))), - 15); - // then we can perform a consistent permutation on the global register to get - // everything in shape: - lane1 = _mm256_permute_ps(lane1, _MM_SHUFFLE(3, 3, 2, 2)); + // an unaligned load is required here as there is no requirement + // on the alignment of input pointer 'from' + __m256i low_half = _mm256_loadu_si256(reinterpret_cast(from)); + __m512 even_elements = _mm512_castsi512_ps(_mm512_cvtepu32_epi64(low_half)); + __m512 pairs = _mm512_permute_ps(even_elements, _MM_SHUFFLE(2, 2, 0, 0)); + return pairs; +} #ifdef EIGEN_VECTORIZE_AVX512DQ - Packet16f res = _mm512_undefined_ps(); - return _mm512_insertf32x8(res, lane0, 0); - return _mm512_insertf32x8(res, lane1, 1); - return res; -#else - Packet16f res = _mm512_undefined_ps(); - res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane0, 0), 0); - res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane0, 1), 1); - res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane1, 0), 2); - res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane1, 1), 3); - return res; -#endif -} +// FIXME: this does not look optimal, better load a Packet4d and shuffle... // Loads 4 doubles from memory a returns the packet {a0, a0 a1, a1, a2, a2, a3, // a3} template <> EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { - Packet4d lane0 = _mm256_broadcast_pd((const __m128d*)(const void*)from); - lane0 = _mm256_permute_pd(lane0, 3 << 2); - - Packet4d lane1 = _mm256_broadcast_pd((const __m128d*)(const void*)(from + 2)); - lane1 = _mm256_permute_pd(lane1, 3 << 2); - - Packet8d res = _mm512_undefined_pd(); - res = _mm512_insertf64x4(res, lane0, 0); - return _mm512_insertf64x4(res, lane1, 1); + __m512d x = _mm512_setzero_pd(); + x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[0]), 0); + x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[1]), 1); + x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[2]), 2); + x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[3]), 3); + return x; } +#else +template <> +EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { + __m512d x = _mm512_setzero_pd(); + x = _mm512_mask_broadcastsd_pd(x, 0x3<<0, _mm_load_sd(from+0)); + x = _mm512_mask_broadcastsd_pd(x, 0x3<<2, _mm_load_sd(from+1)); + x = _mm512_mask_broadcastsd_pd(x, 0x3<<4, _mm_load_sd(from+2)); + x = _mm512_mask_broadcastsd_pd(x, 0x3<<6, _mm_load_sd(from+3)); + return x; +} +#endif // Loads 4 floats from memory a returns the packet // {a0, a0 a0, a0, a1, a1, a1, a1, a2, a2, a2, a2, a3, a3, a3, a3} template <> EIGEN_STRONG_INLINE Packet16f ploadquad(const float* from) { - Packet16f tmp = _mm512_undefined_ps(); - tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from), 0); - tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 1), 1); - tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 2), 2); - tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 3), 3); - return tmp; + Packet16f tmp = _mm512_castps128_ps512(ploadu(from)); + const Packet16i scatter_mask = _mm512_set_epi32(3,3,3,3, 2,2,2,2, 1,1,1,1, 0,0,0,0); + return _mm512_permutexvar_ps(scatter_mask, tmp); } + // Loads 2 doubles from memory a returns the packet // {a0, a0 a0, a0, a1, a1, a1, a1} template <> EIGEN_STRONG_INLINE Packet8d ploadquad(const double* from) { - Packet8d tmp = _mm512_undefined_pd(); - Packet2d tmp0 = _mm_load_pd1(from); - Packet2d tmp1 = _mm_load_pd1(from + 1); - Packet4d lane0 = _mm256_broadcastsd_pd(tmp0); - Packet4d lane1 = _mm256_broadcastsd_pd(tmp1); + __m256d lane0 = _mm256_set1_pd(*from); + __m256d lane1 = _mm256_set1_pd(*(from+1)); + __m512d tmp = _mm512_undefined_pd(); tmp = _mm512_insertf64x4(tmp, lane0, 0); return _mm512_insertf64x4(tmp, lane1, 1); } @@ -565,7 +539,7 @@ EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet16i& from) { template <> EIGEN_DEVICE_FUNC inline Packet16f pgather(const float* from, Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(stride); + Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); Packet16i stride_multiplier = _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); @@ -575,7 +549,7 @@ EIGEN_DEVICE_FUNC inline Packet16f pgather(const float* from, template <> EIGEN_DEVICE_FUNC inline Packet8d pgather(const double* from, Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(stride); + Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); @@ -586,7 +560,7 @@ template <> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet16f& from, Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(stride); + Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); Packet16i stride_multiplier = _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); @@ -596,7 +570,7 @@ template <> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet8d& from, Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(stride); + Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); _mm512_i32scatter_pd(to, indices, from, 8); @@ -660,8 +634,8 @@ EIGEN_STRONG_INLINE Packet8d pabs(const Packet8d& a) { #ifdef EIGEN_VECTORIZE_AVX512DQ // AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 #define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ - __m256 OUTPUT##_0 = _mm512_extractf32x8_ps(INPUT, 0) __m256 OUTPUT##_1 = \ - _mm512_extractf32x8_ps(INPUT, 1) + __m256 OUTPUT##_0 = _mm512_extractf32x8_ps(INPUT, 0); \ + __m256 OUTPUT##_1 = _mm512_extractf32x8_ps(INPUT, 1) #else #define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ __m256 OUTPUT##_0 = _mm256_insertf128_ps( \ @@ -674,17 +648,136 @@ EIGEN_STRONG_INLINE Packet8d pabs(const Packet8d& a) { #ifdef EIGEN_VECTORIZE_AVX512DQ #define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ - OUTPUT = _mm512_insertf32x8(OUTPUT, INPUTA, 0); \ - OUTPUT = _mm512_insertf32x8(OUTPUT, INPUTB, 1); + OUTPUT = _mm512_insertf32x8(_mm512_castps256_ps512(INPUTA), INPUTB, 1); #else #define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ + OUTPUT = _mm512_undefined_ps(); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 0), 0); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 1), 1); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 0), 2); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 1), 3); #endif -template<> EIGEN_STRONG_INLINE Packet16f preduxp(const Packet16f* -vecs) + +template <> +EIGEN_STRONG_INLINE float predux(const Packet16f& a) { +#ifdef EIGEN_VECTORIZE_AVX512DQ + __m256 lane0 = _mm512_extractf32x8_ps(a, 0); + __m256 lane1 = _mm512_extractf32x8_ps(a, 1); + Packet8f x = _mm256_add_ps(lane0, lane1); + return predux(x); +#else + __m128 lane0 = _mm512_extractf32x4_ps(a, 0); + __m128 lane1 = _mm512_extractf32x4_ps(a, 1); + __m128 lane2 = _mm512_extractf32x4_ps(a, 2); + __m128 lane3 = _mm512_extractf32x4_ps(a, 3); + __m128 sum = _mm_add_ps(_mm_add_ps(lane0, lane1), _mm_add_ps(lane2, lane3)); + sum = _mm_hadd_ps(sum, sum); + sum = _mm_hadd_ps(sum, _mm_permute_ps(sum, 1)); + return _mm_cvtss_f32(sum); +#endif +} +template <> +EIGEN_STRONG_INLINE double predux(const Packet8d& a) { + __m256d lane0 = _mm512_extractf64x4_pd(a, 0); + __m256d lane1 = _mm512_extractf64x4_pd(a, 1); + __m256d sum = _mm256_add_pd(lane0, lane1); + __m256d tmp0 = _mm256_hadd_pd(sum, _mm256_permute2f128_pd(sum, sum, 1)); + return _mm_cvtsd_f64(_mm256_castpd256_pd128(_mm256_hadd_pd(tmp0, tmp0))); +} + +template <> +EIGEN_STRONG_INLINE Packet8f predux_downto4(const Packet16f& a) { +#ifdef EIGEN_VECTORIZE_AVX512DQ + Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); + Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); + return padd(lane0, lane1); +#else + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f sum0 = padd(lane0, lane2); + Packet4f sum1 = padd(lane1, lane3); + return _mm256_insertf128_ps(_mm256_castps128_ps256(sum0), sum1, 1); +#endif +} +template <> +EIGEN_STRONG_INLINE Packet4d predux_downto4(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d res = padd(lane0, lane1); + return res; +} + +template <> +EIGEN_STRONG_INLINE float predux_mul(const Packet16f& a) { +//#ifdef EIGEN_VECTORIZE_AVX512DQ +#if 0 + Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); + Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); + Packet8f res = pmul(lane0, lane1); + res = pmul(res, _mm256_permute2f128_ps(res, res, 1)); + res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +#else + __m128 lane0 = _mm512_extractf32x4_ps(a, 0); + __m128 lane1 = _mm512_extractf32x4_ps(a, 1); + __m128 lane2 = _mm512_extractf32x4_ps(a, 2); + __m128 lane3 = _mm512_extractf32x4_ps(a, 3); + __m128 res = pmul(pmul(lane0, lane1), pmul(lane2, lane3)); + res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +#endif +} +template <> +EIGEN_STRONG_INLINE double predux_mul(const Packet8d& a) { + __m256d lane0 = _mm512_extractf64x4_pd(a, 0); + __m256d lane1 = _mm512_extractf64x4_pd(a, 1); + __m256d res = pmul(lane0, lane1); + res = pmul(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(pmul(res, _mm256_shuffle_pd(res, res, 1))); +} + +template <> +EIGEN_STRONG_INLINE float predux_min(const Packet16f& a) { + __m128 lane0 = _mm512_extractf32x4_ps(a, 0); + __m128 lane1 = _mm512_extractf32x4_ps(a, 1); + __m128 lane2 = _mm512_extractf32x4_ps(a, 2); + __m128 lane3 = _mm512_extractf32x4_ps(a, 3); + __m128 res = _mm_min_ps(_mm_min_ps(lane0, lane1), _mm_min_ps(lane2, lane3)); + res = _mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(_mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +} +template <> +EIGEN_STRONG_INLINE double predux_min(const Packet8d& a) { + __m256d lane0 = _mm512_extractf64x4_pd(a, 0); + __m256d lane1 = _mm512_extractf64x4_pd(a, 1); + __m256d res = _mm256_min_pd(lane0, lane1); + res = _mm256_min_pd(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(_mm256_min_pd(res, _mm256_shuffle_pd(res, res, 1))); +} + +template <> +EIGEN_STRONG_INLINE float predux_max(const Packet16f& a) { + __m128 lane0 = _mm512_extractf32x4_ps(a, 0); + __m128 lane1 = _mm512_extractf32x4_ps(a, 1); + __m128 lane2 = _mm512_extractf32x4_ps(a, 2); + __m128 lane3 = _mm512_extractf32x4_ps(a, 3); + __m128 res = _mm_max_ps(_mm_max_ps(lane0, lane1), _mm_max_ps(lane2, lane3)); + res = _mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(_mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +} + +template <> +EIGEN_STRONG_INLINE double predux_max(const Packet8d& a) { + __m256d lane0 = _mm512_extractf64x4_pd(a, 0); + __m256d lane1 = _mm512_extractf64x4_pd(a, 1); + __m256d res = _mm256_max_pd(lane0, lane1); + res = _mm256_max_pd(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(_mm256_max_pd(res, _mm256_shuffle_pd(res, res, 1))); +} + +template<> EIGEN_STRONG_INLINE Packet16f preduxp(const Packet16f* vecs) { EIGEN_EXTRACT_8f_FROM_16f(vecs[0], vecs0); EIGEN_EXTRACT_8f_FROM_16f(vecs[1], vecs1); @@ -873,174 +966,7 @@ template<> EIGEN_STRONG_INLINE Packet8d preduxp(const Packet8d* vecs) return _mm512_insertf64x4(final_output, final_1, 1); } - -template <> -EIGEN_STRONG_INLINE float predux(const Packet16f& a) { - //#ifdef EIGEN_VECTORIZE_AVX512DQ -#if 0 - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f sum = padd(lane0, lane1); - Packet8f tmp0 = _mm256_hadd_ps(sum, _mm256_permute2f128_ps(a, a, 1)); - tmp0 = _mm256_hadd_ps(tmp0, tmp0); - return pfirst(_mm256_hadd_ps(tmp0, tmp0)); -#else - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f sum = padd(padd(lane0, lane1), padd(lane2, lane3)); - sum = _mm_hadd_ps(sum, sum); - sum = _mm_hadd_ps(sum, _mm_permute_ps(sum, 1)); - return pfirst(sum); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d sum = padd(lane0, lane1); - Packet4d tmp0 = _mm256_hadd_pd(sum, _mm256_permute2f128_pd(sum, sum, 1)); - return pfirst(_mm256_hadd_pd(tmp0, tmp0)); -} - -template <> -EIGEN_STRONG_INLINE Packet8f predux_downto4(const Packet16f& a) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - return padd(lane0, lane1); -#else - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f sum0 = padd(lane0, lane2); - Packet4f sum1 = padd(lane1, lane3); - return _mm256_insertf128_ps(_mm256_castps128_ps256(sum0), sum1, 1); -#endif -} -template <> -EIGEN_STRONG_INLINE Packet4d predux_downto4(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = padd(lane0, lane1); - return res; -} - -template <> -EIGEN_STRONG_INLINE float predux_mul(const Packet16f& a) { -//#ifdef EIGEN_VECTORIZE_AVX512DQ -#if 0 - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_ps(res, res, 1)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#else - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f res = pmul(pmul(lane0, lane1), pmul(lane2, lane3)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux_mul(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(pmul(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_min(const Packet16f& a) { - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f res = _mm_min_ps(_mm_min_ps(lane0, lane1), _mm_min_ps(lane2, lane3)); - res = _mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} -template <> -EIGEN_STRONG_INLINE double predux_min(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = _mm256_min_pd(lane0, lane1); - res = _mm256_min_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_min_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_max(const Packet16f& a) { - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f res = _mm_max_ps(_mm_max_ps(lane0, lane1), _mm_max_ps(lane2, lane3)); - res = _mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} -template <> -EIGEN_STRONG_INLINE double predux_max(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = _mm256_max_pd(lane0, lane1); - res = _mm256_max_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_max_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet16f& first, - const Packet16f& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - Offset + 15, Offset + 14, Offset + 13, Offset + 12, Offset + 11, - Offset + 10, Offset + 9, Offset + 8, Offset + 7, Offset + 6, - Offset + 5, Offset + 4, Offset + 3, Offset + 2, Offset + 1, Offset); - - __m512i second_idx = - _mm512_set_epi32(Offset - 1, Offset - 2, Offset - 3, Offset - 4, - Offset - 5, Offset - 6, Offset - 7, Offset - 8, - Offset - 9, Offset - 10, Offset - 11, Offset - 12, - Offset - 13, Offset - 14, Offset - 15, Offset - 16); - - unsigned short mask = 0xFFFF; - mask <<= (16 - Offset); - - first = _mm512_permutexvar_ps(first_idx, first); - Packet16f tmp = _mm512_permutexvar_ps(second_idx, second); - first = _mm512_mask_blend_ps(mask, first, tmp); - } - } -}; -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet8d& first, const Packet8d& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - 0, Offset + 7, 0, Offset + 6, 0, Offset + 5, 0, Offset + 4, 0, - Offset + 3, 0, Offset + 2, 0, Offset + 1, 0, Offset); - - __m512i second_idx = _mm512_set_epi32( - 0, Offset - 1, 0, Offset - 2, 0, Offset - 3, 0, Offset - 4, 0, - Offset - 5, 0, Offset - 6, 0, Offset - 7, 0, Offset - 8); - - unsigned char mask = 0xFF; - mask <<= (8 - Offset); - - first = _mm512_permutexvar_pd(first_idx, first); - Packet8d tmp = _mm512_permutexvar_pd(second_idx, second); - first = _mm512_mask_blend_pd(mask, first, tmp); - } - } -}; + #define PACK_OUTPUT(OUTPUT, INPUT, INDEX, STRIDE) \ @@ -1302,13 +1228,76 @@ EIGEN_STRONG_INLINE Packet16f pblend(const Selector<16>& /*ifPacket*/, return Packet16f(); } template <> -EIGEN_STRONG_INLINE Packet8d pblend(const Selector<8>& /*ifPacket*/, - const Packet8d& /*thenPacket*/, - const Packet8d& /*elsePacket*/) { - assert(false && "To be implemented"); - return Packet8d(); +EIGEN_STRONG_INLINE Packet8d pblend(const Selector<8>& ifPacket, + const Packet8d& thenPacket, + const Packet8d& elsePacket) { + __mmask8 m = (ifPacket.select[0] ) + | (ifPacket.select[1]<<1) + | (ifPacket.select[2]<<2) + | (ifPacket.select[3]<<3) + | (ifPacket.select[4]<<4) + | (ifPacket.select[5]<<5) + | (ifPacket.select[6]<<6) + | (ifPacket.select[7]<<7); + return _mm512_mask_blend_pd(m, elsePacket, thenPacket); +} + +template<> EIGEN_STRONG_INLINE Packet16i pcast(const Packet16f& a) { + return _mm512_cvttps_epi32(a); +} + +template<> EIGEN_STRONG_INLINE Packet16f pcast(const Packet16i& a) { + return _mm512_cvtepi32_ps(a); } +template +struct palign_impl { + static EIGEN_STRONG_INLINE void run(Packet16f& first, + const Packet16f& second) { + if (Offset != 0) { + __m512i first_idx = _mm512_set_epi32( + Offset + 15, Offset + 14, Offset + 13, Offset + 12, Offset + 11, + Offset + 10, Offset + 9, Offset + 8, Offset + 7, Offset + 6, + Offset + 5, Offset + 4, Offset + 3, Offset + 2, Offset + 1, Offset); + + __m512i second_idx = + _mm512_set_epi32(Offset - 1, Offset - 2, Offset - 3, Offset - 4, + Offset - 5, Offset - 6, Offset - 7, Offset - 8, + Offset - 9, Offset - 10, Offset - 11, Offset - 12, + Offset - 13, Offset - 14, Offset - 15, Offset - 16); + + unsigned short mask = 0xFFFF; + mask <<= (16 - Offset); + + first = _mm512_permutexvar_ps(first_idx, first); + Packet16f tmp = _mm512_permutexvar_ps(second_idx, second); + first = _mm512_mask_blend_ps(mask, first, tmp); + } + } +}; +template +struct palign_impl { + static EIGEN_STRONG_INLINE void run(Packet8d& first, const Packet8d& second) { + if (Offset != 0) { + __m512i first_idx = _mm512_set_epi32( + 0, Offset + 7, 0, Offset + 6, 0, Offset + 5, 0, Offset + 4, 0, + Offset + 3, 0, Offset + 2, 0, Offset + 1, 0, Offset); + + __m512i second_idx = _mm512_set_epi32( + 0, Offset - 1, 0, Offset - 2, 0, Offset - 3, 0, Offset - 4, 0, + Offset - 5, 0, Offset - 6, 0, Offset - 7, 0, Offset - 8); + + unsigned char mask = 0xFF; + mask <<= (8 - Offset); + + first = _mm512_permutexvar_pd(first_idx, first); + Packet8d tmp = _mm512_permutexvar_pd(second_idx, second); + first = _mm512_mask_blend_pd(mask, first, tmp); + } + } +}; + + } // end namespace internal } // end namespace Eigen diff --git a/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h b/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h index 755e6209d1e..59717b4fe6c 100644 --- a/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h +++ b/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h @@ -42,6 +42,7 @@ #define EIGEN_EXPLICIT_CAST(tgt_type) operator tgt_type() #endif +#include namespace Eigen { diff --git a/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h b/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h index c66d38469f8..f749c573ff6 100644 --- a/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +++ b/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h @@ -230,7 +230,7 @@ template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux(const half2& #else float a1 = __low2float(a); float a2 = __high2float(a); - return Eigen::half(half_impl::raw_uint16_to_half(__float2half_rn(a1 + a2))); + return Eigen::half(__float2half_rn(a1 + a2)); #endif } @@ -264,7 +264,7 @@ template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_mul(const ha #else float a1 = __low2float(a); float a2 = __high2float(a); - return Eigen::half(half_impl::raw_uint16_to_half(__float2half_rn(a1 * a2))); + return Eigen::half(__float2half_rn(a1 * a2)); #endif } diff --git a/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h b/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h index 2e6a00ffd1b..b56e7afd2c5 100644 --- a/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h +++ b/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h @@ -768,7 +768,7 @@ struct scalar_sign_op { if (aa==real_type(0)) return Scalar(0); aa = real_type(1)/aa; - return Scalar(real(a)*aa, imag(a)*aa ); + return Scalar(a.real()*aa, a.imag()*aa ); } //TODO //template diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h index e3980f6ffd4..681451cc309 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -115,7 +115,8 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // registers. However once the latency is hidden there is no point in // increasing the value of k, so we'll cap it at 320 (value determined // experimentally). - const Index k_cache = (numext::mini)((l1-ksub)/kdiv, 320); + // To avoid that k vanishes, we make k_cache at least as big as kr + const Index k_cache = numext::maxi(kr, (numext::mini)((l1-ksub)/kdiv, 320)); if (k_cache < k) { k = k_cache - (k_cache % kr); eigen_internal_assert(k > 0); @@ -648,8 +649,8 @@ class gebp_traits, std::complex, _ConjLhs, // Vectorized path EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, DoublePacketType& dest) const { - dest.first = pset1(real(*b)); - dest.second = pset1(imag(*b)); + dest.first = pset1(numext::real(*b)); + dest.second = pset1(numext::imag(*b)); } EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, ResPacket& dest) const diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h index 6440e1d09c7..ed6234c378e 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h @@ -20,8 +20,9 @@ template class level3_blocking; template< typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs> -struct general_matrix_matrix_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +struct general_matrix_matrix_product { typedef gebp_traits Traits; @@ -30,7 +31,7 @@ struct general_matrix_matrix_product& blocking, GemmParallelInfo* info = 0) @@ -39,8 +40,8 @@ struct general_matrix_matrix_product - ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resStride,alpha,blocking,info); + ColMajor,ResInnerStride> + ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking,info); } }; @@ -49,8 +50,9 @@ struct general_matrix_matrix_product -struct general_matrix_matrix_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +struct general_matrix_matrix_product { typedef gebp_traits Traits; @@ -59,17 +61,17 @@ typedef typename ScalarBinaryOpTraits::ReturnType ResScala static void run(Index rows, Index cols, Index depth, const LhsScalar* _lhs, Index lhsStride, const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resStride, + ResScalar* _res, Index resIncr, Index resStride, ResScalar alpha, level3_blocking& blocking, GemmParallelInfo* info = 0) { typedef const_blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs,lhsStride); - RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride); + typedef blas_data_mapper ResMapper; + LhsMapper lhs(_lhs, lhsStride); + RhsMapper rhs(_rhs, rhsStride); + ResMapper res(_res, resStride, resIncr); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -226,7 +228,7 @@ struct gemm_functor Gemm::run(rows, cols, m_lhs.cols(), &m_lhs.coeffRef(row,0), m_lhs.outerStride(), &m_rhs.coeffRef(0,col), m_rhs.outerStride(), - (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.outerStride(), + (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.innerStride(), m_dest.outerStride(), m_actualAlpha, m_blocking, info); } @@ -428,7 +430,7 @@ struct generic_product_impl static void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::evalTo(dst, lhs, rhs); + lazyproduct::eval_dynamic(dst, lhs, rhs, internal::assign_op()); else { dst.setZero(); @@ -440,7 +442,7 @@ struct generic_product_impl static void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::addTo(dst, lhs, rhs); + lazyproduct::eval_dynamic(dst, lhs, rhs, internal::add_assign_op()); else scaleAndAddTo(dst,lhs, rhs, Scalar(1)); } @@ -449,7 +451,7 @@ struct generic_product_impl static void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::subTo(dst, lhs, rhs); + lazyproduct::eval_dynamic(dst, lhs, rhs, internal::sub_assign_op()); else scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); } @@ -476,7 +478,8 @@ struct generic_product_impl Index, LhsScalar, (ActualLhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(LhsBlasTraits::NeedToConjugate), RhsScalar, (ActualRhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(RhsBlasTraits::NeedToConjugate), - (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor>, + (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor, + Dest::InnerStrideAtCompileTime>, ActualLhsTypeCleaned, ActualRhsTypeCleaned, Dest, BlockingType> GemmFunctor; BlockingType blocking(dst.rows(), dst.cols(), lhs.cols(), 1, true); diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h index e844e37d16b..d68d2f96572 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h @@ -25,51 +25,54 @@ namespace internal { **********************************************************************/ // forward declarations (defined at the end of this file) -template +template struct tribb_kernel; /* Optimized matrix-matrix product evaluating only one triangular half */ template + int ResStorageOrder, int ResInnerStride, int UpLo, int Version = Specialized> struct general_matrix_matrix_triangular_product; // as usual if the result is row major => we transpose the product template -struct general_matrix_matrix_triangular_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int UpLo, int Version> +struct general_matrix_matrix_triangular_product { typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resStride, + const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resIncr, Index resStride, const ResScalar& alpha, level3_blocking& blocking) { general_matrix_matrix_triangular_product - ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resStride,alpha,blocking); + ColMajor, ResInnerStride, UpLo==Lower?Upper:Lower> + ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking); } }; template -struct general_matrix_matrix_triangular_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int UpLo, int Version> +struct general_matrix_matrix_triangular_product { typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, ResScalar* _res, Index resStride, + const RhsScalar* _rhs, Index rhsStride, + ResScalar* _res, Index resIncr, Index resStride, const ResScalar& alpha, level3_blocking& blocking) { typedef gebp_traits Traits; typedef const_blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride); + ResMapper res(_res, resStride, resIncr); Index kc = blocking.kc(); Index mc = (std::min)(size,blocking.mc()); @@ -87,7 +90,7 @@ struct general_matrix_matrix_triangular_product pack_lhs; gemm_pack_rhs pack_rhs; gebp_kernel gebp; - tribb_kernel sybb; + tribb_kernel sybb; for(Index k2=0; k2 +template struct tribb_kernel { typedef gebp_traits Traits; @@ -142,11 +144,13 @@ struct tribb_kernel enum { BlockSize = meta_least_common_multiple::ret }; - void operator()(ResScalar* _res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) + void operator()(ResScalar* _res, Index resIncr, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) { - typedef blas_data_mapper ResMapper; - ResMapper res(_res, resStride); - gebp_kernel gebp_kernel; + typedef blas_data_mapper ResMapper; + typedef blas_data_mapper BufferMapper; + ResMapper res(_res, resStride, resIncr); + gebp_kernel gebp_kernel1; + gebp_kernel gebp_kernel2; Matrix buffer((internal::constructor_without_unaligned_array_assert())); @@ -158,31 +162,32 @@ struct tribb_kernel const RhsScalar* actual_b = blockB+j*depth; if(UpLo==Upper) - gebp_kernel(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - + gebp_kernel1(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, + -1, -1, 0, 0); + // selfadjoint micro block { Index i = j; buffer.setZero(); // 1 - apply the kernel on the temporary buffer - gebp_kernel(ResMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, - -1, -1, 0, 0); + gebp_kernel2(BufferMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, + -1, -1, 0, 0); + // 2 - triangular accumulation for(Index j1=0; j1 internal::general_matrix_matrix_triangular_product + IsRowMajor ? RowMajor : ColMajor, MatrixType::InnerStrideAtCompileTime, UpLo&(Lower|Upper)> ::run(size, depth, &actualLhs.coeffRef(SkipDiag&&(UpLo&Lower)==Lower ? 1 : 0,0), actualLhs.outerStride(), &actualRhs.coeffRef(0,SkipDiag&&(UpLo&Upper)==Upper ? 1 : 0), actualRhs.outerStride(), - mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? 1 : mat.outerStride() ) : 0), mat.outerStride(), actualAlpha, blocking); + mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? mat.innerStride() : mat.outerStride() ) : 0), + mat.innerStride(), mat.outerStride(), actualAlpha, blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h index f6f9ebecae7..691f95d6979 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h @@ -40,7 +40,7 @@ namespace internal { template struct general_matrix_matrix_rankupdate : general_matrix_matrix_triangular_product< - Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,UpLo,BuiltIn> {}; + Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,1,UpLo,BuiltIn> {}; // try to go to BLAS specialization @@ -48,9 +48,9 @@ struct general_matrix_matrix_rankupdate : template \ struct general_matrix_matrix_triangular_product { \ + Scalar,RhsStorageOrder,ConjugateRhs,ColMajor,1,UpLo,Specialized> { \ static EIGEN_STRONG_INLINE void run(Index size, Index depth,const Scalar* lhs, Index lhsStride, \ - const Scalar* rhs, Index rhsStride, Scalar* res, Index resStride, Scalar alpha, level3_blocking& blocking) \ + const Scalar* rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) \ { \ if ( lhs==rhs && ((UpLo&(Lower|Upper))==UpLo) ) { \ general_matrix_matrix_rankupdate \ @@ -59,8 +59,8 @@ struct general_matrix_matrix_triangular_product \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resStride,alpha,blocking); \ + ColMajor, 1, UpLo, BuiltIn> \ + ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resIncr,resStride,alpha,blocking); \ } \ } \ }; diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h index b0f6b0d5b9e..71abf4013d1 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h @@ -51,20 +51,22 @@ template< \ typename Index, \ int LhsStorageOrder, bool ConjugateLhs, \ int RhsStorageOrder, bool ConjugateRhs> \ -struct general_matrix_matrix_product \ +struct general_matrix_matrix_product \ { \ typedef gebp_traits Traits; \ \ static void run(Index rows, Index cols, Index depth, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ + EIGTYPE* res, Index resIncr, Index resStride, \ EIGTYPE alpha, \ level3_blocking& /*blocking*/, \ GemmParallelInfo* /*info = 0*/) \ { \ using std::conj; \ \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ char transa, transb; \ BlasIndex m, n, k, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ diff --git a/third_party/eigen/Eigen/src/Core/products/Parallelizer.h b/third_party/eigen/Eigen/src/Core/products/Parallelizer.h index c2f084c82cc..a3cc05b77bb 100644 --- a/third_party/eigen/Eigen/src/Core/products/Parallelizer.h +++ b/third_party/eigen/Eigen/src/Core/products/Parallelizer.h @@ -17,7 +17,8 @@ namespace internal { /** \internal */ inline void manage_multi_threading(Action action, int* v) { - static EIGEN_UNUSED int m_maxThreads = -1; + static int m_maxThreads = -1; + EIGEN_UNUSED_VARIABLE(m_maxThreads); if(action==SetAction) { @@ -150,8 +151,10 @@ void parallelize_gemm(const Functor& func, Index rows, Index cols, Index depth, info[i].lhs_start = r0; info[i].lhs_length = actualBlockRows; - if(transpose) func(c0, actualBlockCols, 0, rows, info); - else func(0, rows, c0, actualBlockCols, info); + if(transpose) + func(c0, actualBlockCols, 0, rows, info); + else + func(0, rows, c0, actualBlockCols, info); } #endif } diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h index da6f82abcd7..04c933480bd 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h @@ -277,20 +277,21 @@ struct symm_pack_rhs template + int ResStorageOrder, int ResInnerStride> struct product_selfadjoint_matrix; template -struct product_selfadjoint_matrix + int RhsStorageOrder, bool RhsSelfAdjoint, bool ConjugateRhs, + int ResInnerStride> +struct product_selfadjoint_matrix { static EIGEN_STRONG_INLINE void run( Index rows, Index cols, const Scalar* lhs, Index lhsStride, const Scalar* rhs, Index rhsStride, - Scalar* res, Index resStride, + Scalar* res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { product_selfadjoint_matrix::IsComplex && EIGEN_LOGICAL_XOR(RhsSelfAdjoint,ConjugateRhs), EIGEN_LOGICAL_XOR(LhsSelfAdjoint,LhsStorageOrder==RowMajor) ? ColMajor : RowMajor, LhsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsSelfAdjoint,ConjugateLhs), - ColMajor> - ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resStride, alpha, blocking); + ColMajor,ResInnerStride> + ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); } }; template -struct product_selfadjoint_matrix + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +struct product_selfadjoint_matrix { static EIGEN_DONT_INLINE void run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resStride, + Scalar* res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking); }; template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resStride, + Scalar* _res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { Index size = rows; @@ -334,11 +337,11 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix LhsMapper; typedef const_blas_data_mapper LhsTransposeMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); LhsTransposeMapper lhs_transpose(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride); + ResMapper res(_res, resStride, resIncr); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -398,26 +401,28 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix -struct product_selfadjoint_matrix + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +struct product_selfadjoint_matrix { static EIGEN_DONT_INLINE void run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resStride, + Scalar* res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking); }; template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride> +EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resStride, + Scalar* _res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { Index size = cols; @@ -425,9 +430,9 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix Traits; typedef const_blas_data_mapper LhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); - ResMapper res(_res,resStride); + ResMapper res(_res,resStride, resIncr); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -503,12 +508,13 @@ struct selfadjoint_product_impl NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsIsUpper,bool(LhsBlasTraits::NeedToConjugate)), EIGEN_LOGICAL_XOR(RhsIsUpper,internal::traits::Flags &RowMajorBit) ? RowMajor : ColMajor, RhsIsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(RhsIsUpper,bool(RhsBlasTraits::NeedToConjugate)), - internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor> + internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor, + Dest::InnerStrideAtCompileTime> ::run( lhs.rows(), rhs.cols(), // sizes &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.outerStride(), // result info + &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info actualAlpha, blocking // alpha ); } diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h index 9a5318507af..61396dbdf6f 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h @@ -44,16 +44,18 @@ namespace internal { template \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ \ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ + EIGTYPE* res, Index resIncr, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ char side='L', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -91,15 +93,17 @@ struct product_selfadjoint_matrix \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ + EIGTYPE* res, Index resIncr, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ char side='L', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -167,16 +171,18 @@ EIGEN_BLAS_HEMM_L(scomplex, float, cf, chemm_) template \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ \ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ + EIGTYPE* res, Index resIncr, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ char side='R', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -213,15 +219,17 @@ struct product_selfadjoint_matrix \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resStride, \ + EIGTYPE* res, Index resIncr, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ char side='R', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h index f038d686f5d..ef12c98f6cb 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h @@ -109,10 +109,10 @@ struct selfadjoint_product_selector internal::general_matrix_matrix_triangular_product::IsComplex, Scalar, OtherIsRowMajor ? ColMajor : RowMajor, (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex, - IsRowMajor ? RowMajor : ColMajor, UpLo> + IsRowMajor ? RowMajor : ColMajor, MatrixType::InnerStrideAtCompileTime, UpLo> ::run(size, depth, &actualOther.coeffRef(0,0), actualOther.outerStride(), &actualOther.coeffRef(0,0), actualOther.outerStride(), - mat.data(), mat.outerStride(), actualAlpha, blocking); + mat.data(), mat.innerStride(), mat.outerStride(), actualAlpha, blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h index f784507e77f..2fb408d1d73 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h @@ -45,22 +45,24 @@ template + int ResStorageOrder, int ResInnerStride, + int Version = Specialized> struct product_triangular_matrix_matrix; template + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,RowMajor,ResInnerStride,Version> { static EIGEN_STRONG_INLINE void run( Index rows, Index cols, Index depth, const Scalar* lhs, Index lhsStride, const Scalar* rhs, Index rhsStride, - Scalar* res, Index resStride, + Scalar* res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { product_triangular_matrix_matrix - ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resStride, alpha, blocking); + ColMajor, ResInnerStride> + ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); } }; // implements col-major += alpha * op(triangular) * op(general) template + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,ColMajor,ResInnerStride,Version> { typedef gebp_traits Traits; @@ -95,20 +98,21 @@ struct product_triangular_matrix_matrix& blocking); }; template + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int Version> EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( + RhsStorageOrder,ConjugateRhs,ColMajor,ResInnerStride,Version>::run( Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resStride, + Scalar* _res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { // strip zeros @@ -119,10 +123,10 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride); + ResMapper res(_res, resStride, resIncr); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -235,10 +239,11 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,ColMajor,ResInnerStride,Version> { typedef gebp_traits Traits; enum { @@ -251,20 +256,21 @@ struct product_triangular_matrix_matrix& blocking); }; template + int RhsStorageOrder, bool ConjugateRhs, + int ResInnerStride, int Version> EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( + RhsStorageOrder,ConjugateRhs,ColMajor,ResInnerStride,Version>::run( Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resStride, + Scalar* _res, Index resIncr, Index resStride, const Scalar& alpha, level3_blocking& blocking) { const Index PacketBytes = packet_traits::size*sizeof(Scalar); @@ -276,10 +282,10 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride); + ResMapper res(_res, resStride, resIncr); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -433,12 +439,12 @@ struct triangular_product_impl Mode, LhsIsTriangular, (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate, (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor> + (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, Dest::InnerStrideAtCompileTime> ::run( stripedRows, stripedCols, stripedDepth, // sizes &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.outerStride(), // result info + &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info actualAlpha, blocking ); diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h index a25197ab017..a98d12e4ae1 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h @@ -46,7 +46,7 @@ template {}; + RhsStorageOrder, ConjugateRhs, ResStorageOrder, 1, BuiltIn> {}; // try to go to BLAS specialization @@ -55,13 +55,15 @@ template \ struct product_triangular_matrix_matrix { \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder,ConjugateRhs,ColMajor,1,Specialized> { \ static inline void run(Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride,\ - const Scalar* _rhs, Index rhsStride, Scalar* res, Index resStride, Scalar alpha, level3_blocking& blocking) { \ + const Scalar* _rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) { \ + EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ + eigen_assert(resIncr == 1); \ product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ } \ }; @@ -115,8 +117,8 @@ struct product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder, ConjugateRhs, ColMajor, 1, BuiltIn>::run( \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ /*std::cout << "TRMM_L: A is not square! Go to Eigen TRMM implementation!\n";*/ \ } else { \ /* Make sense to call GEMM */ \ @@ -124,8 +126,8 @@ struct product_triangular_matrix_matrix_trmm(); \ BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, resStride, alpha, gemm_blocking, 0); \ + general_matrix_matrix_product::run( \ + rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, 1, resStride, alpha, gemm_blocking, 0); \ \ /*std::cout << "TRMM_L: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ } \ @@ -232,8 +234,8 @@ struct product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder, ConjugateRhs, ColMajor, 1, BuiltIn>::run( \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ /*std::cout << "TRMM_R: A is not square! Go to Eigen TRMM implementation!\n";*/ \ } else { \ /* Make sense to call GEMM */ \ @@ -241,8 +243,8 @@ struct product_triangular_matrix_matrix_trmm(); \ BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, resStride, alpha, gemm_blocking, 0); \ + general_matrix_matrix_product::run( \ + rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, 1, resStride, alpha, gemm_blocking, 0); \ \ /*std::cout << "TRMM_R: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ } \ diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h b/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h index 223c38b8656..e3ed2cd19ea 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h @@ -15,48 +15,48 @@ namespace Eigen { namespace internal { // if the rhs is row major, let's transpose the product -template -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static void run( Index size, Index cols, const Scalar* tri, Index triStride, - Scalar* _other, Index otherStride, + Scalar* _other, Index otherIncr, Index otherStride, level3_blocking& blocking) { triangular_solve_matrix< Scalar, Index, Side==OnTheLeft?OnTheRight:OnTheLeft, (Mode&UnitDiag) | ((Mode&Upper) ? Lower : Upper), NumTraits::IsComplex && Conjugate, - TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor> - ::run(size, cols, tri, triStride, _other, otherStride, blocking); + TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor, OtherInnerStride> + ::run(size, cols, tri, triStride, _other, otherIncr, otherStride, blocking); } }; /* Optimized triangular solver with multiple right hand side and the triangular matrix on the left */ -template -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static EIGEN_DONT_INLINE void run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherStride, + Scalar* _other, Index otherIncr, Index otherStride, level3_blocking& blocking); }; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( +template +EIGEN_DONT_INLINE void triangular_solve_matrix::run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherStride, + Scalar* _other, Index otherIncr, Index otherStride, level3_blocking& blocking) { Index cols = otherSize; typedef const_blas_data_mapper TriMapper; - typedef blas_data_mapper OtherMapper; + typedef blas_data_mapper OtherMapper; TriMapper tri(_tri, triStride); - OtherMapper other(_other, otherStride); + OtherMapper other(_other, otherStride, otherIncr); typedef gebp_traits Traits; @@ -128,19 +128,19 @@ EIGEN_DONT_INLINE void triangular_solve_matrix -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static EIGEN_DONT_INLINE void run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherStride, + Scalar* _other, Index otherIncr, Index otherStride, level3_blocking& blocking); }; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( +template +EIGEN_DONT_INLINE void triangular_solve_matrix::run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherStride, + Scalar* _other, Index otherIncr, Index otherStride, level3_blocking& blocking) { Index rows = otherSize; typedef typename NumTraits::Real RealScalar; - typedef blas_data_mapper LhsMapper; + typedef blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - LhsMapper lhs(_other, otherStride); + LhsMapper lhs(_other, otherStride, otherIncr); RhsMapper rhs(_tri, triStride); typedef gebp_traits Traits; @@ -297,24 +297,24 @@ EIGEN_DONT_INLINE void triangular_solve_matrix \ -struct triangular_solve_matrix \ +struct triangular_solve_matrix \ { \ enum { \ IsLower = (Mode&Lower) == Lower, \ @@ -51,8 +51,10 @@ struct triangular_solve_matrix& /*blocking*/) \ + EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ + eigen_assert(otherIncr == 1); \ BlasIndex m = convert_index(size), n = convert_index(otherSize), lda, ldb; \ char side = 'L', uplo, diag='N', transa; \ /* Set alpha_ */ \ @@ -99,7 +101,7 @@ EIGEN_BLAS_TRSM_L(scomplex, float, ctrsm_) // implements RightSide general * op(triangular)^-1 #define EIGEN_BLAS_TRSM_R(EIGTYPE, BLASTYPE, BLASFUNC) \ template \ -struct triangular_solve_matrix \ +struct triangular_solve_matrix \ { \ enum { \ IsLower = (Mode&Lower) == Lower, \ @@ -110,8 +112,10 @@ struct triangular_solve_matrix& /*blocking*/) \ + EIGTYPE* _other, Index otherIncr, Index otherStride, level3_blocking& /*blocking*/) \ { \ + EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ + eigen_assert(otherIncr == 1); \ BlasIndex m = convert_index(otherSize), n = convert_index(size), lda, ldb; \ char side = 'R', uplo, diag='N', transa; \ /* Set alpha_ */ \ diff --git a/third_party/eigen/Eigen/src/Core/util/BlasUtil.h b/third_party/eigen/Eigen/src/Core/util/BlasUtil.h index 6e6ee119b64..3dff9bc9b33 100755 --- a/third_party/eigen/Eigen/src/Core/util/BlasUtil.h +++ b/third_party/eigen/Eigen/src/Core/util/BlasUtil.h @@ -31,7 +31,7 @@ template< typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResStorageOrder> + int ResStorageOrder, int ResInnerStride> struct general_matrix_matrix_product; template +class BlasLinearMapper; + template -class BlasLinearMapper { +class BlasLinearMapper { public: typedef typename packet_traits::type Packet; typedef typename packet_traits::half HalfPacket; - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data) : m_data(data) {} + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data, Index incr=1) + : m_data(data) + { + EIGEN_ONLY_USED_FOR_DEBUG(incr); + eigen_assert(incr==1); + } EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { internal::prefetch(&operator()(i)); @@ -188,16 +196,25 @@ class BlasLinearMapper { }; // Lightweight helper class to access matrix coefficients. -template -class blas_data_mapper { - public: +template +class blas_data_mapper; + +template +class blas_data_mapper +{ +public: typedef typename packet_traits::type Packet; typedef typename packet_traits::half HalfPacket; typedef BlasLinearMapper LinearMapper; typedef BlasVectorMapper VectorMapper; - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride) : m_data(data), m_stride(stride) {} + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr=1) + : m_data(data), m_stride(stride) + { + EIGEN_ONLY_USED_FOR_DEBUG(incr); + eigen_assert(incr==1); + } EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper getSubMapper(Index i, Index j) const { @@ -251,6 +268,90 @@ class blas_data_mapper { const Index m_stride; }; +// Implementation of non-natural increment (i.e. inner-stride != 1) +// The exposed API is not complete yet compared to the Incr==1 case +// because some features makes less sense in this case. +template +class BlasLinearMapper +{ +public: + typedef typename packet_traits::type Packet; + typedef typename packet_traits::half HalfPacket; + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data,Index incr) : m_data(data), m_incr(incr) {} + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { + internal::prefetch(&operator()(i)); + } + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { + return m_data[i*m_incr.value()]; + } + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { + return pgather(m_data + i*m_incr.value(), m_incr.value()); + } + + template + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const PacketType &p) const { + pscatter(m_data + i*m_incr.value(), p, m_incr.value()); + } + +protected: + Scalar *m_data; + const internal::variable_if_dynamic m_incr; +}; + +template +class blas_data_mapper +{ +public: + typedef typename packet_traits::type Packet; + typedef typename packet_traits::half HalfPacket; + + typedef BlasLinearMapper LinearMapper; + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr) : m_data(data), m_stride(stride), m_incr(incr) {} + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper + getSubMapper(Index i, Index j) const { + return blas_data_mapper(&operator()(i, j), m_stride, m_incr.value()); + } + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { + return LinearMapper(&operator()(i, j), m_incr.value()); + } + + EIGEN_DEVICE_FUNC + EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { + return m_data[StorageOrder==RowMajor ? j*m_incr.value() + i*m_stride : i*m_incr.value() + j*m_stride]; + } + + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { + return pgather(&operator()(i, j),m_incr.value()); + } + + template + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE PacketT load(Index i, Index j) const { + return pgather(&operator()(i, j),m_incr.value()); + } + + template + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { + pscatter(&operator()(i, j), p, m_stride); + } + + template + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { + return pgather(&operator()(i, j), m_stride); + } + +protected: + Scalar* EIGEN_RESTRICT m_data; + const Index m_stride; + const internal::variable_if_dynamic m_incr; +}; + // lightweight helper class to access matrix coefficients (const version) template class const_blas_data_mapper : public blas_data_mapper { diff --git a/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h b/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h index 351bd6c6000..74f74cc42b8 100755 --- a/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h +++ b/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h @@ -57,7 +57,10 @@ #if __GNUC__>=6 #pragma GCC diagnostic ignored "-Wignored-attributes" #endif - + #if __GNUC__==7 + // See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89325 + #pragma GCC diagnostic ignored "-Wattributes" + #endif #endif #if defined __NVCC__ @@ -80,4 +83,12 @@ #pragma diag_suppress 2737 #endif +#else +// warnings already disabled: +# ifndef EIGEN_WARNINGS_DISABLED_2 +# define EIGEN_WARNINGS_DISABLED_2 +# elif defined(EIGEN_INTERNAL_DEBUGGING) +# error "Do not include \"DisableStupidWarnings.h\" recursively more than twice!" +# endif + #endif // not EIGEN_WARNINGS_DISABLED diff --git a/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h b/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h index ea107393a7d..134544f9643 100644 --- a/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h +++ b/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h @@ -47,11 +47,7 @@ template struct NumTraits; template struct EigenBase; template class DenseBase; template class PlainObjectBase; - - -template::value > -class DenseCoeffsBase; +template class DenseCoeffsBase; templatex || (EIGEN_WORLD_VERSION>=x && \ (EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \ @@ -380,7 +380,8 @@ #if EIGEN_MAX_CPP_VER>=11 && \ ((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)) \ || (defined(__GNUC__) && defined(_GLIBCXX_USE_C99)) \ - || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER))) + || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER)) \ + || (EIGEN_COMP_MSVC >= 1900) ) #define EIGEN_HAS_C99_MATH 1 #else #define EIGEN_HAS_C99_MATH 0 @@ -396,6 +397,20 @@ #endif #endif +// Does the compiler support type_traits? +// - full support of type traits was added only to GCC 5.1.0. +// - 20150626 corresponds to the last release of 4.x libstdc++ +#ifndef EIGEN_HAS_TYPE_TRAITS +#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_HAS_CXX11 || EIGEN_COMP_MSVC >= 1700) \ + && ((!EIGEN_COMP_GNUC_STRICT) || EIGEN_GNUC_AT_LEAST(5, 1)) \ + && ((!defined(__GLIBCXX__)) || __GLIBCXX__ > 20150626) +#define EIGEN_HAS_TYPE_TRAITS 1 +#define EIGEN_INCLUDE_TYPE_TRAITS +#else +#define EIGEN_HAS_TYPE_TRAITS 0 +#endif +#endif + // Does the compiler support variadic templates? #ifndef EIGEN_HAS_VARIADIC_TEMPLATES #if EIGEN_MAX_CPP_VER>=11 && (__cplusplus > 199711L || EIGEN_COMP_MSVC >= 1900) \ @@ -835,11 +850,48 @@ namespace Eigen { #endif +/** + * \internal + * \brief Macro to explicitly define the default copy constructor. + * This is necessary, because the implicit definition is deprecated if the copy-assignment is overridden. + */ +#if EIGEN_HAS_CXX11 +#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) EIGEN_DEVICE_FUNC CLASS(const CLASS&) = default; +#else +#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) +#endif + + + /** \internal * \brief Macro to manually inherit assignment operators. * This is necessary, because the implicitly defined assignment operator gets deleted when a custom operator= is defined. + * With C++11 or later this also default-implements the copy-constructor */ -#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) +#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) \ + EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ + EIGEN_DEFAULT_COPY_CONSTRUCTOR(Derived) + +/** \internal + * \brief Macro to manually define default constructors and destructors. + * This is necessary when the copy constructor is re-defined. + * For empty helper classes this should usually be protected, to avoid accidentally creating empty objects. + * + * Hiding the default destructor lead to problems in C++03 mode together with boost::multiprecision + */ +#if EIGEN_HAS_CXX11 +#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ + EIGEN_DEVICE_FUNC Derived() = default; \ + EIGEN_DEVICE_FUNC ~Derived() = default; +#else +#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ + EIGEN_DEVICE_FUNC Derived() {}; \ + /* EIGEN_DEVICE_FUNC ~Derived() {}; */ +#endif + + + + /** * Just a side note. Commenting within defines works only by documenting diff --git a/third_party/eigen/Eigen/src/Core/util/Meta.h b/third_party/eigen/Eigen/src/Core/util/Meta.h index d31e9541122..9b61ff037aa 100755 --- a/third_party/eigen/Eigen/src/Core/util/Meta.h +++ b/third_party/eigen/Eigen/src/Core/util/Meta.h @@ -97,6 +97,9 @@ template<> struct is_arithmetic { enum { value = true }; }; template<> struct is_arithmetic { enum { value = true }; }; template<> struct is_arithmetic { enum { value = true }; }; +#if EIGEN_HAS_CXX11 +using std::is_integral; +#else template struct is_integral { enum { value = false }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; @@ -108,6 +111,11 @@ template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; +#if EIGEN_COMP_MSVC +template<> struct is_integral { enum { value = true }; }; +template<> struct is_integral{ enum { value = true }; }; +#endif +#endif #if EIGEN_HAS_CXX11 using std::make_unsigned; @@ -531,4 +539,30 @@ bool not_equal_strict(const double& x,const double& y) { return std::not_equal_t } // end namespace Eigen +// Define portable (u)int{32,64} types +#if EIGEN_HAS_CXX11 +#include +namespace Eigen { +namespace numext { +typedef std::uint32_t uint32_t; +typedef std::int32_t int32_t; +typedef std::uint64_t uint64_t; +typedef std::int64_t int64_t; +} +} +#else +// Without c++11, all compilers able to compile Eigen also +// provides the C99 stdint.h header file. +#include +namespace Eigen { +namespace numext { +typedef ::uint32_t uint32_t; +typedef ::int32_t int32_t; +typedef ::uint64_t uint64_t; +typedef ::int64_t int64_t; +} +} +#endif + + #endif // EIGEN_META_H diff --git a/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h b/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h index ecc82b7c8df..1ce6fd1b009 100644 --- a/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h +++ b/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h @@ -1,4 +1,8 @@ -#ifdef EIGEN_WARNINGS_DISABLED +#ifdef EIGEN_WARNINGS_DISABLED_2 +// "DisableStupidWarnings.h" was included twice recursively: Do not reenable warnings yet! +# undef EIGEN_WARNINGS_DISABLED_2 + +#elif defined(EIGEN_WARNINGS_DISABLED) #undef EIGEN_WARNINGS_DISABLED #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS diff --git a/third_party/eigen/Eigen/src/Core/util/XprHelper.h b/third_party/eigen/Eigen/src/Core/util/XprHelper.h index ba5bd186d29..6bb4970828e 100644 --- a/third_party/eigen/Eigen/src/Core/util/XprHelper.h +++ b/third_party/eigen/Eigen/src/Core/util/XprHelper.h @@ -34,6 +34,20 @@ inline IndexDest convert_index(const IndexSrc& idx) { return IndexDest(idx); } +// true if T can be considered as an integral index (i.e., and integral type or enum) +template struct is_valid_index_type +{ + enum { value = +#if EIGEN_HAS_TYPE_TRAITS + internal::is_integral::value || std::is_enum::value +#elif EIGEN_COMP_MSVC + internal::is_integral::value || __is_enum(T) +#else + // without C++11, we use is_convertible to Index instead of is_integral in order to treat enums as Index. + internal::is_convertible::value && !internal::is_same::value && !is_same::value +#endif + }; +}; // promote_scalar_arg is an helper used in operation between an expression and a scalar, like: // expression * scalar @@ -90,6 +104,9 @@ class no_assignment_operator { private: no_assignment_operator& operator=(const no_assignment_operator&); + protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(no_assignment_operator) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(no_assignment_operator) }; /** \internal return the index type with the largest number of bits */ diff --git a/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h b/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h index 7f38919f778..4354e4018fd 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h @@ -300,10 +300,13 @@ typename ComplexSchur::ComplexScalar ComplexSchur::compu ComplexScalar trace = t.coeff(0,0) + t.coeff(1,1); ComplexScalar eival1 = (trace + disc) / RealScalar(2); ComplexScalar eival2 = (trace - disc) / RealScalar(2); - - if(numext::norm1(eival1) > numext::norm1(eival2)) + RealScalar eival1_norm = numext::norm1(eival1); + RealScalar eival2_norm = numext::norm1(eival2); + // A division by zero can only occur if eival1==eival2==0. + // In this case, det==0, and all we have to do is checking that eival2_norm!=0 + if(eival1_norm > eival2_norm) eival2 = det / eival1; - else + else if(eival2_norm!=RealScalar(0)) eival1 = det / eival2; // choose the eigenvalue closest to the bottom entry of the diagonal diff --git a/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h b/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h index 17ea903f5f1..9191519abe9 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h @@ -236,7 +236,7 @@ template class RealSchur typedef Matrix Vector3s; Scalar computeNormOfT(); - Index findSmallSubdiagEntry(Index iu); + Index findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero); void splitOffTwoRows(Index iu, bool computeU, const Scalar& exshift); void computeShift(Index iu, Index iter, Scalar& exshift, Vector3s& shiftInfo); void initFrancisQRStep(Index il, Index iu, const Vector3s& shiftInfo, Index& im, Vector3s& firstHouseholderVector); @@ -302,12 +302,16 @@ RealSchur& RealSchur::computeFromHessenberg(const HessMa Index totalIter = 0; // iteration count for whole matrix Scalar exshift(0); // sum of exceptional shifts Scalar norm = computeNormOfT(); + // sub-diagonal entries smaller than considerAsZero will be treated as zero. + // We use eps^2 to enable more precision in small eigenvalues. + Scalar considerAsZero = numext::maxi( norm * numext::abs2(NumTraits::epsilon()), + (std::numeric_limits::min)() ); if(norm!=Scalar(0)) { while (iu >= 0) { - Index il = findSmallSubdiagEntry(iu); + Index il = findSmallSubdiagEntry(iu,considerAsZero); // Check for convergence if (il == iu) // One root found @@ -364,14 +368,17 @@ inline typename MatrixType::Scalar RealSchur::computeNormOfT() /** \internal Look for single small sub-diagonal element and returns its index */ template -inline Index RealSchur::findSmallSubdiagEntry(Index iu) +inline Index RealSchur::findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero) { using std::abs; Index res = iu; while (res > 0) { Scalar s = abs(m_matT.coeff(res-1,res-1)) + abs(m_matT.coeff(res,res)); - if (abs(m_matT.coeff(res,res-1)) <= NumTraits::epsilon() * s) + + s = numext::maxi(s * NumTraits::epsilon(), considerAsZero); + + if (abs(m_matT.coeff(res,res-1)) <= s) break; res--; } diff --git a/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h b/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h index 9ddd553f2f1..d37656fa207 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h @@ -605,7 +605,8 @@ template struct direct_selfadjoint_eigenvalues res, Ref representative) { - using std::abs; + EIGEN_USING_STD_MATH(sqrt) + EIGEN_USING_STD_MATH(abs) Index i0; // Find non-zero column i0 (by construction, there must exist a non zero coefficient on the diagonal): mat.diagonal().cwiseAbs().maxCoeff(&i0); @@ -616,8 +617,8 @@ template struct direct_selfadjoint_eigenvaluesn1) res = c0/std::sqrt(n0); - else res = c1/std::sqrt(n1); + if(n0>n1) res = c0/sqrt(n0); + else res = c1/sqrt(n1); return true; } diff --git a/third_party/eigen/Eigen/src/Geometry/Quaternion.h b/third_party/eigen/Eigen/src/Geometry/Quaternion.h index c3fd8c3e0f8..b818206568e 100644 --- a/third_party/eigen/Eigen/src/Geometry/Quaternion.h +++ b/third_party/eigen/Eigen/src/Geometry/Quaternion.h @@ -169,20 +169,38 @@ class QuaternionBase : public RotationBase /** return the result vector of \a v through the rotation*/ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Vector3 _transformVector(const Vector3& v) const; + #ifdef EIGEN_PARSED_BY_DOXYGEN /** \returns \c *this with scalar type casted to \a NewScalarType * * Note that if \a NewScalarType is equal to the current scalar type of \c *this * then this function smartly returns a const reference to \c *this. */ template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const + EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const; + + #else + + template + EIGEN_DEVICE_FUNC inline + typename internal::enable_if::value,const Derived&>::type cast() const { - return typename internal::cast_return_type >::type(derived()); + return derived(); } + template + EIGEN_DEVICE_FUNC inline + typename internal::enable_if::value,Quaternion >::type cast() const + { + return Quaternion(coeffs().template cast()); + } + #endif + #ifdef EIGEN_QUATERNIONBASE_PLUGIN # include EIGEN_QUATERNIONBASE_PLUGIN #endif +protected: + EIGEN_DEFAULT_COPY_CONSTRUCTOR(QuaternionBase) + EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(QuaternionBase) }; /*************************************************************************** diff --git a/third_party/eigen/Eigen/src/Geometry/Scaling.h b/third_party/eigen/Eigen/src/Geometry/Scaling.h index f58ca03d94a..33eabd81a4c 100755 --- a/third_party/eigen/Eigen/src/Geometry/Scaling.h +++ b/third_party/eigen/Eigen/src/Geometry/Scaling.h @@ -14,7 +14,7 @@ namespace Eigen { /** \geometry_module \ingroup Geometry_Module * - * \class Scaling + * \class UniformScaling * * \brief Represents a generic uniform scaling transformation * diff --git a/third_party/eigen/Eigen/src/Geometry/Transform.h b/third_party/eigen/Eigen/src/Geometry/Transform.h index 3f31ee45df9..c21d9e550e8 100644 --- a/third_party/eigen/Eigen/src/Geometry/Transform.h +++ b/third_party/eigen/Eigen/src/Geometry/Transform.h @@ -252,11 +252,11 @@ class Transform public: /** Default constructor without initialization of the meaningful coefficients. - * If Mode==Affine, then the last row is set to [0 ... 0 1] */ + * If Mode==Affine or Mode==Isometry, then the last row is set to [0 ... 0 1] */ EIGEN_DEVICE_FUNC inline Transform() { check_template_params(); - internal::transform_make_affine<(int(Mode)==Affine) ? Affine : AffineCompact>::run(m_matrix); + internal::transform_make_affine<(int(Mode)==Affine || int(Mode)==Isometry) ? Affine : AffineCompact>::run(m_matrix); } EIGEN_DEVICE_FUNC inline Transform(const Transform& other) diff --git a/third_party/eigen/Eigen/src/Geometry/Translation.h b/third_party/eigen/Eigen/src/Geometry/Translation.h index 51d9a82ebbd..0e99ce68e26 100644 --- a/third_party/eigen/Eigen/src/Geometry/Translation.h +++ b/third_party/eigen/Eigen/src/Geometry/Translation.h @@ -138,12 +138,6 @@ class Translation /** \returns the inverse translation (opposite) */ Translation inverse() const { return Translation(-m_coeffs); } - Translation& operator=(const Translation& other) - { - m_coeffs = other.m_coeffs; - return *this; - } - static const Translation Identity() { return Translation(VectorType::Zero()); } /** \returns \c *this with scalar type casted to \a NewScalarType diff --git a/third_party/eigen/Eigen/src/Geometry/Umeyama.h b/third_party/eigen/Eigen/src/Geometry/Umeyama.h index 7e933fca13e..6b755008fdc 100644 --- a/third_party/eigen/Eigen/src/Geometry/Umeyama.h +++ b/third_party/eigen/Eigen/src/Geometry/Umeyama.h @@ -87,7 +87,7 @@ struct umeyama_transform_matrix_type * \f{align*} * T = \begin{bmatrix} c\mathbf{R} & \mathbf{t} \\ \mathbf{0} & 1 \end{bmatrix} * \f} -* minimizing the resudiual above. This transformation is always returned as an +* minimizing the residual above. This transformation is always returned as an * Eigen::Matrix. */ template diff --git a/third_party/eigen/Eigen/src/LU/PartialPivLU.h b/third_party/eigen/Eigen/src/LU/PartialPivLU.h index d439618879a..6b10f39fab4 100644 --- a/third_party/eigen/Eigen/src/LU/PartialPivLU.h +++ b/third_party/eigen/Eigen/src/LU/PartialPivLU.h @@ -519,7 +519,10 @@ void PartialPivLU::compute() // the row permutation is stored as int indices, so just to be sure: eigen_assert(m_lu.rows()::highest()); - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); + if(m_lu.cols()>0) + m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); + else + m_l1_norm = RealScalar(0); eigen_assert(m_lu.rows() == m_lu.cols() && "PartialPivLU is only for square (and moreover invertible) matrices"); const Index size = m_lu.rows(); diff --git a/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h b/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h index ebb64a62b02..4dce2ef20ee 100644 --- a/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h +++ b/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h @@ -44,7 +44,7 @@ struct compute_inverse_size4 static void run(const MatrixType& mat, ResultType& result) { ActualMatrixType matrix(mat); - EIGEN_ALIGN16 const unsigned int _Sign_PNNP[4] = { 0x00000000, 0x80000000, 0x80000000, 0x00000000 }; + const Packet4f p4f_sign_PNNP = _mm_castsi128_ps(_mm_set_epi32(0x00000000, 0x80000000, 0x80000000, 0x00000000)); // Load the full matrix into registers __m128 _L1 = matrix.template packet( 0); @@ -139,7 +139,7 @@ struct compute_inverse_size4 iC = _mm_sub_ps(iC, _mm_mul_ps(_mm_shuffle_ps(A,A,0xB1), _mm_shuffle_ps(DC,DC,0x66))); rd = _mm_shuffle_ps(rd,rd,0); - rd = _mm_xor_ps(rd, _mm_load_ps((float*)_Sign_PNNP)); + rd = _mm_xor_ps(rd, p4f_sign_PNNP); // iB = C*|B| - D*B#*A iB = _mm_sub_ps(_mm_mul_ps(C,_mm_shuffle_ps(dB,dB,0)), iB); diff --git a/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h b/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h index 091c3970e81..98d0e3f21ab 100644 --- a/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h +++ b/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h @@ -192,7 +192,8 @@ class PardisoImpl : public SparseSolverBase void pardisoInit(int type) { m_type = type; - bool symmetric = std::abs(m_type) < 10; + EIGEN_USING_STD_MATH(abs); + bool symmetric = abs(m_type) < 10; m_iparm[0] = 1; // No solver default m_iparm[1] = 2; // use Metis for the ordering m_iparm[2] = 0; // Reserved. Set to zero. (??Numbers of processors, value of OMP_NUM_THREADS??) diff --git a/third_party/eigen/Eigen/src/SVD/BDCSVD.h b/third_party/eigen/Eigen/src/SVD/BDCSVD.h index 1134d66e7e9..a5b73f8f21f 100644 --- a/third_party/eigen/Eigen/src/SVD/BDCSVD.h +++ b/third_party/eigen/Eigen/src/SVD/BDCSVD.h @@ -768,6 +768,21 @@ void BDCSVD::computeSingVals(const ArrayRef& col0, const ArrayRef& d // measure everything relative to shift Map diagShifted(m_workspace.data()+4*n, n); diagShifted = diag - shift; + + if(k!=actual_n-1) + { + // check that after the shift, f(mid) is still negative: + RealScalar midShifted = (right - left) / RealScalar(2); + if(shift==right) + midShifted = -midShifted; + RealScalar fMidShifted = secularEq(midShifted, col0, diag, perm, diagShifted, shift); + if(fMidShifted>0) + { + // fMid was erroneous, fix it: + shift = fMidShifted > Literal(0) ? left : right; + diagShifted = diag - shift; + } + } // initial guess RealScalar muPrev, muCur; @@ -845,11 +860,13 @@ void BDCSVD::computeSingVals(const ArrayRef& col0, const ArrayRef& d } RealScalar fLeft = secularEq(leftShifted, col0, diag, perm, diagShifted, shift); + eigen_internal_assert(fLeft::computeSingVals(const ArrayRef& col0, const ArrayRef& d } #endif eigen_internal_assert(fLeft * fRight < Literal(0)); - - while (rightShifted - leftShifted > Literal(2) * NumTraits::epsilon() * numext::maxi(abs(leftShifted), abs(rightShifted))) + + if(fLeft Literal(2) * NumTraits::epsilon() * numext::maxi(abs(leftShifted), abs(rightShifted))) { - leftShifted = midShifted; - fLeft = fMid; + RealScalar midShifted = (leftShifted + rightShifted) / Literal(2); + fMid = secularEq(midShifted, col0, diag, perm, diagShifted, shift); + eigen_internal_assert((numext::isfinite)(fMid)); + + if (fLeft * fMid < Literal(0)) + { + rightShifted = midShifted; + } + else + { + leftShifted = midShifted; + fLeft = fMid; + } } + muCur = (leftShifted + rightShifted) / Literal(2); + } + else + { + // We have a problem as shifting on the left or right give either a positive or negative value + // at the middle of [left,right]... + // Instead fo abbording or entering an infinite loop, + // let's just use the middle as the estimated zero-crossing: + muCur = (right - left) * RealScalar(0.5); + if(shift == right) + muCur = -muCur; } - - muCur = (leftShifted + rightShifted) / Literal(2); } singVals[k] = shift + muCur; @@ -924,7 +955,7 @@ void BDCSVD::perturbCol0 Index j = i 0.9 ) + if(i!=k && numext::abs(((singVals(j)+dk)*(mus(j)+(shifts(j)-dk)))/((diag(i)+dk)*(diag(i)-dk)) - 1) > 0.9 ) std::cout << " " << ((singVals(j)+dk)*(mus(j)+(shifts(j)-dk)))/((diag(i)+dk)*(diag(i)-dk)) << " == (" << (singVals(j)+dk) << " * " << (mus(j)+(shifts(j)-dk)) << ") / (" << (diag(i)+dk) << " * " << (diag(i)-dk) << ")\n"; #endif @@ -934,7 +965,7 @@ void BDCSVD::perturbCol0 std::cout << "zhat(" << k << ") = sqrt( " << prod << ") ; " << (singVals(last) + dk) << " * " << mus(last) + shifts(last) << " - " << dk << "\n"; #endif RealScalar tmp = sqrt(prod); - zhat(k) = col0(k) > Literal(0) ? tmp : -tmp; + zhat(k) = col0(k) > Literal(0) ? RealScalar(tmp) : RealScalar(-tmp); } } } diff --git a/third_party/eigen/Eigen/src/SVD/SVDBase.h b/third_party/eigen/Eigen/src/SVD/SVDBase.h index 3d1ef373ea7..53da2848883 100644 --- a/third_party/eigen/Eigen/src/SVD/SVDBase.h +++ b/third_party/eigen/Eigen/src/SVD/SVDBase.h @@ -183,7 +183,7 @@ class SVDBase // this temporary is needed to workaround a MSVC issue Index diagSize = (std::max)(1,m_diagSize); return m_usePrescribedThreshold ? m_prescribedThreshold - : diagSize*NumTraits::epsilon(); + : RealScalar(diagSize)*NumTraits::epsilon(); } /** \returns true if \a U (full or thin) is asked for in this SVD decomposition */ diff --git a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h index 2907f65296b..369e6804afa 100644 --- a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h +++ b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h @@ -608,7 +608,7 @@ template } if(Base::m_diag.size()>0) - dest = Base::m_diag.asDiagonal().inverse() * dest; + dest = Base::m_diag.real().asDiagonal().inverse() * dest; if (Base::m_matrix.nonZeros()>0) // otherwise I==I { diff --git a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h index 31e06995b8a..7b6183d08b1 100644 --- a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h +++ b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h @@ -156,7 +156,7 @@ void SimplicialCholeskyBase::factorize_preordered(const CholMatrixType& /* the nonzero entry L(k,i) */ Scalar l_ki; if(DoLDLT) - l_ki = yi / m_diag[i]; + l_ki = yi / numext::real(m_diag[i]); else yi = l_ki = yi / Lx[Lp[i]]; diff --git a/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h b/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h index cec49d1841f..2cb7747cc99 100644 --- a/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h +++ b/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h @@ -28,7 +28,7 @@ class AmbiVector typedef typename NumTraits::Real RealScalar; explicit AmbiVector(Index size) - : m_buffer(0), m_zero(0), m_size(0), m_allocatedSize(0), m_allocatedElements(0), m_mode(-1) + : m_buffer(0), m_zero(0), m_size(0), m_end(0), m_allocatedSize(0), m_allocatedElements(0), m_mode(-1) { resize(size); } @@ -147,7 +147,8 @@ template void AmbiVector<_Scalar,_StorageIndex>::init(int mode) { m_mode = mode; - if (m_mode==IsSparse) + // This is only necessary in sparse mode, but we set these unconditionally to avoid some maybe-uninitialized warnings + // if (m_mode==IsSparse) { m_llSize = 0; m_llStart = -1; @@ -267,7 +268,7 @@ _Scalar& AmbiVector<_Scalar,_StorageIndex>::coeff(Index i) else { Index elid = m_llStart; - while (elid >= 0 && (llElements[elid].index)= 0 && llElements[elid].index class unary_evaluator, IteratorBased>::InnerIterator : public unary_evaluator, IteratorBased>::EvalIterator { + protected: typedef typename XprType::Scalar Scalar; typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; public: @@ -99,6 +100,7 @@ template class unary_evaluator, IteratorBased>::InnerIterator : public unary_evaluator, IteratorBased>::EvalIterator { + protected: typedef typename XprType::Scalar Scalar; typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; public: diff --git a/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h b/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h index 0a2490bcc3d..a5396538ba9 100644 --- a/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h +++ b/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h @@ -327,7 +327,8 @@ class SparseMatrix m_outerIndex[j] = newOuterIndex[j]; m_innerNonZeros[j] = innerNNZ; } - m_outerIndex[m_outerSize] = m_outerIndex[m_outerSize-1] + m_innerNonZeros[m_outerSize-1] + reserveSizes[m_outerSize-1]; + if(m_outerSize>0) + m_outerIndex[m_outerSize] = m_outerIndex[m_outerSize-1] + m_innerNonZeros[m_outerSize-1] + reserveSizes[m_outerSize-1]; m_data.resize(m_outerIndex[m_outerSize]); } diff --git a/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h b/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h index 65611b3d4ca..76117a0107e 100644 --- a/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h +++ b/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h @@ -453,7 +453,7 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix, IteratorBased> class InnerIterator : public EvalIterator { + protected: typedef typename XprType::Scalar Scalar; public: diff --git a/third_party/eigen/Eigen/src/SparseLU/SparseLU.h b/third_party/eigen/Eigen/src/SparseLU/SparseLU.h index 7104831c03b..87f0efe37ed 100644 --- a/third_party/eigen/Eigen/src/SparseLU/SparseLU.h +++ b/third_party/eigen/Eigen/src/SparseLU/SparseLU.h @@ -43,8 +43,8 @@ template struct SparseLUMatrixURetu * Simple example with key steps * \code * VectorXd x(n), b(n); - * SparseMatrix A; - * SparseLU, COLAMDOrdering > solver; + * SparseMatrix A; + * SparseLU, COLAMDOrdering > solver; * // fill A and b; * // Compute the ordering permutation vector from the structural pattern of A * solver.analyzePattern(A); diff --git a/third_party/eigen/Eigen/src/StlSupport/StdDeque.h b/third_party/eigen/Eigen/src/StlSupport/StdDeque.h index cf1fedf9273..af158f425d4 100644 --- a/third_party/eigen/Eigen/src/StlSupport/StdDeque.h +++ b/third_party/eigen/Eigen/src/StlSupport/StdDeque.h @@ -98,8 +98,10 @@ namespace std { { return deque_base::insert(position,x); } void insert(const_iterator position, size_type new_size, const value_type& x) { deque_base::insert(position, new_size, x); } -#elif defined(_GLIBCXX_DEQUE) && EIGEN_GNUC_AT_LEAST(4,2) +#elif defined(_GLIBCXX_DEQUE) && EIGEN_GNUC_AT_LEAST(4,2) && !EIGEN_GNUC_AT_LEAST(10, 1) // workaround GCC std::deque implementation + // GCC 10.1 doesn't let us access _Deque_impl _M_impl anymore and we have to + // fall-back to the default case void resize(size_type new_size, const value_type& x) { if (new_size < deque_base::size()) @@ -108,7 +110,7 @@ namespace std { deque_base::insert(deque_base::end(), new_size - deque_base::size(), x); } #else - // either GCC 4.1 or non-GCC + // either non-GCC or GCC between 4.1 and 10.1 // default implementation which should always work. void resize(size_type new_size, const value_type& x) { diff --git a/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h b/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h index 1f8a531af59..05a7449bc9e 100644 --- a/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h +++ b/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h @@ -119,7 +119,7 @@ OP(const Scalar& s) const { \ return this->OP(Derived::PlainObject::Constant(rows(), cols(), s)); \ } \ EIGEN_DEVICE_FUNC friend EIGEN_STRONG_INLINE const RCmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s, const Derived& d) { \ +OP(const Scalar& s, const EIGEN_CURRENT_STORAGE_BASE_CLASS& d) { \ return Derived::PlainObject::Constant(d.rows(), d.cols(), s).OP(d); \ } diff --git a/third_party/eigen/bench/bench_gemm.cpp b/third_party/eigen/bench/bench_gemm.cpp index 8528c558742..dccab96a8bb 100644 --- a/third_party/eigen/bench/bench_gemm.cpp +++ b/third_party/eigen/bench/bench_gemm.cpp @@ -112,6 +112,7 @@ void matlab_cplx_cplx(const M& ar, const M& ai, const M& br, const M& bi, M& cr, cr.noalias() -= ai * bi; ci.noalias() += ar * bi; ci.noalias() += ai * br; + // [cr ci] += [ar ai] * br + [-ai ar] * bi } void matlab_real_cplx(const M& a, const M& br, const M& bi, M& cr, M& ci) @@ -240,7 +241,7 @@ int main(int argc, char ** argv) blas_gemm(a,b,r); c.noalias() += a * b; if(!r.isApprox(c)) { - std::cout << r - c << "\n"; + std::cout << (r - c).norm() << "\n"; std::cerr << "Warning, your product is crap!\n\n"; } #else @@ -249,7 +250,7 @@ int main(int argc, char ** argv) gemm(a,b,c); r.noalias() += a.cast() .lazyProduct( b.cast() ); if(!r.isApprox(c)) { - std::cout << r - c << "\n"; + std::cout << (r - c).norm() << "\n"; std::cerr << "Warning, your product is crap!\n\n"; } } diff --git a/third_party/eigen/blas/level3_impl.h b/third_party/eigen/blas/level3_impl.h index 6c802cd5fdb..6dd6338b432 100644 --- a/third_party/eigen/blas/level3_impl.h +++ b/third_party/eigen/blas/level3_impl.h @@ -13,28 +13,28 @@ int EIGEN_BLAS_FUNC(gemm)(const char *opa, const char *opb, const int *m, const const RealScalar *pa, const int *lda, const RealScalar *pb, const int *ldb, const RealScalar *pbeta, RealScalar *pc, const int *ldc) { // std::cerr << "in gemm " << *opa << " " << *opb << " " << *m << " " << *n << " " << *k << " " << *lda << " " << *ldb << " " << *ldc << " " << *palpha << " " << *pbeta << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, Scalar, internal::level3_blocking&, Eigen::internal::GemmParallelInfo*); + typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, Scalar, internal::level3_blocking&, Eigen::internal::GemmParallelInfo*); static const functype func[12] = { // array index: NOTR | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0, // array index: NOTR | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0, // array index: NOTR | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0 }; @@ -71,7 +71,7 @@ int EIGEN_BLAS_FUNC(gemm)(const char *opa, const char *opb, const int *m, const internal::gemm_blocking_space blocking(*m,*n,*k,1,true); int code = OP(*opa) | (OP(*opb) << 2); - func[code](*m, *n, *k, a, *lda, b, *ldb, c, *ldc, alpha, blocking, 0); + func[code](*m, *n, *k, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking, 0); return 0; } @@ -79,63 +79,63 @@ int EIGEN_BLAS_FUNC(trsm)(const char *side, const char *uplo, const char *opa, c const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { // std::cerr << "in trsm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << "," << *n << " " << *palpha << " " << *lda << " " << *ldb<< "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, internal::level3_blocking&); static const functype func[32] = { // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run),\ + (internal::triangular_solve_matrix::run),\ 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0 }; @@ -163,12 +163,12 @@ int EIGEN_BLAS_FUNC(trsm)(const char *side, const char *uplo, const char *opa, c if(SIDE(*side)==LEFT) { internal::gemm_blocking_space blocking(*m,*n,*m,1,false); - func[code](*m, *n, a, *lda, b, *ldb, blocking); + func[code](*m, *n, a, *lda, b, 1, *ldb, blocking); } else { internal::gemm_blocking_space blocking(*m,*n,*n,1,false); - func[code](*n, *m, a, *lda, b, *ldb, blocking); + func[code](*n, *m, a, *lda, b, 1, *ldb, blocking); } if(alpha!=Scalar(1)) @@ -184,63 +184,63 @@ int EIGEN_BLAS_FUNC(trmm)(const char *side, const char *uplo, const char *opa, c const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { // std::cerr << "in trmm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << " " << *n << " " << *lda << " " << *ldb << " " << *palpha << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[32] = { // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0 }; @@ -272,12 +272,12 @@ int EIGEN_BLAS_FUNC(trmm)(const char *side, const char *uplo, const char *opa, c if(SIDE(*side)==LEFT) { internal::gemm_blocking_space blocking(*m,*n,*m,1,false); - func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, *ldb, alpha, blocking); + func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, 1, *ldb, alpha, blocking); } else { internal::gemm_blocking_space blocking(*m,*n,*n,1,false); - func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, *ldb, alpha, blocking); + func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, 1, *ldb, alpha, blocking); } return 1; } @@ -338,12 +338,12 @@ int EIGEN_BLAS_FUNC(symm)(const char *side, const char *uplo, const int *m, cons internal::gemm_blocking_space blocking(*m,*n,size,1,false); if(SIDE(*side)==LEFT) - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); else return 0; else if(SIDE(*side)==RIGHT) - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); else return 0; else return 0; @@ -359,21 +359,21 @@ int EIGEN_BLAS_FUNC(syrk)(const char *uplo, const char *op, const int *n, const { // std::cerr << "in syrk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " << *pbeta << " " << *ldc << "\n"; #if !ISCOMPLEX - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[8] = { // array index: NOTR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: TR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: ADJ | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: NOTR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: TR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: ADJ | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0 }; #endif @@ -426,7 +426,7 @@ int EIGEN_BLAS_FUNC(syrk)(const char *uplo, const char *op, const int *n, const internal::gemm_blocking_space blocking(*n,*n,*k,1,false); int code = OP(*op) | (UPLO(*uplo) << 2); - func[code](*n, *k, a, *lda, a, *lda, c, *ldc, alpha, blocking); + func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); #endif return 0; @@ -537,18 +537,18 @@ int EIGEN_BLAS_FUNC(hemm)(const char *side, const char *uplo, const int *m, cons if(SIDE(*side)==LEFT) { - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix - ::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix - ::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix + ::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix + ::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); else return 0; } else if(SIDE(*side)==RIGHT) { - if(UPLO(*uplo)==UP) matrix(c,*m,*n,*ldc) += alpha * matrix(b,*m,*n,*ldb) * matrix(a,*n,*n,*lda).selfadjointView();/*internal::product_selfadjoint_matrix - ::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking);*/ - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix - ::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) matrix(c,*m,*n,*ldc) += alpha * matrix(b,*m,*n,*ldb) * matrix(a,*n,*n,*lda).selfadjointView();/*internal::product_selfadjoint_matrix + ::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking);*/ + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix + ::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); else return 0; } else @@ -566,19 +566,19 @@ int EIGEN_BLAS_FUNC(herk)(const char *uplo, const char *op, const int *n, const { // std::cerr << "in herk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " << *pbeta << " " << *ldc << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[8] = { // array index: NOTR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: ADJ | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: NOTR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: ADJ | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0 }; @@ -620,7 +620,7 @@ int EIGEN_BLAS_FUNC(herk)(const char *uplo, const char *op, const int *n, const if(*k>0 && alpha!=RealScalar(0)) { internal::gemm_blocking_space blocking(*n,*n,*k,1,false); - func[code](*n, *k, a, *lda, a, *lda, c, *ldc, alpha, blocking); + func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); matrix(c, *n, *n, *ldc).diagonal().imag().setZero(); } return 0; diff --git a/third_party/eigen/cmake/EigenTesting.cmake b/third_party/eigen/cmake/EigenTesting.cmake index a92a2978b0d..3d0074c71ad 100644 --- a/third_party/eigen/cmake/EigenTesting.cmake +++ b/third_party/eigen/cmake/EigenTesting.cmake @@ -677,6 +677,8 @@ macro(ei_set_build_string) set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-cxx11) endif() + set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-v3.3) + if(EIGEN_BUILD_STRING_SUFFIX) set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-${EIGEN_BUILD_STRING_SUFFIX}) endif() diff --git a/third_party/eigen/cmake/FindStandardMathLibrary.cmake b/third_party/eigen/cmake/FindStandardMathLibrary.cmake index 711b0e4b4f1..337f1b30469 100644 --- a/third_party/eigen/cmake/FindStandardMathLibrary.cmake +++ b/third_party/eigen/cmake/FindStandardMathLibrary.cmake @@ -19,8 +19,11 @@ include(CheckCXXSourceCompiles) # notice the std:: is required on some platforms such as QNX set(find_standard_math_library_test_program -"#include -int main() { std::sin(0.0); std::log(0.0f); }") +" +#include +int main(int argc, char **){ + return int(std::sin(double(argc)) + std::log(double(argc))); +}") # first try compiling/linking the test program without any linker flags diff --git a/third_party/eigen/debug/msvc/eigen.natvis b/third_party/eigen/debug/msvc/eigen.natvis index 22cf346576d..da898571797 100644 --- a/third_party/eigen/debug/msvc/eigen.natvis +++ b/third_party/eigen/debug/msvc/eigen.natvis @@ -1,235 +1,235 @@ - - - - - - - - [{$T2}, {$T3}] (fixed matrix) - - - 2 - $i==0 ? $T2 : $T3 - m_storage.m_data.array - - - Backward - 2 - $i==0 ? $T2 : $T3 - m_storage.m_data.array - - - - - - - - [2, 2] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[2]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[3]}) - - - - - - - - [3, 3] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[3]}, {m_storage.m_data.array[6]}) - - - ({m_storage.m_data.array[3]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[7]}) - - - ({m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[8]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[8]}) - - - - - - - - [4, 4] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[8]}, {m_storage.m_data.array[12]}) - - - ({m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[13]}) - - - ({m_storage.m_data.array[8]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[11]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[14]}) - - - ({m_storage.m_data.array[12]}, {m_storage.m_data.array[13]}, {m_storage.m_data.array[14]}, {m_storage.m_data.array[15]}) - - - ({m_storage.m_data.array[3]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[11]}, {m_storage.m_data.array[15]}) - - - - - - - - empty - [{m_storage.m_rows}, {m_storage.m_cols}] (dynamic matrix) - - - 2 - $i==0 ? m_storage.m_rows : m_storage.m_cols - m_storage.m_data - - - Backward - 2 - $i==0 ? m_storage.m_rows : m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{$T2}, {m_storage.m_cols}] (dynamic column matrix) - - - 2 - $i==0 ? $T2 : m_storage.m_cols - m_storage.m_data - - - Backward - 2 - $i==0 ? $T2 : m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{m_storage.m_rows}, {$T2}] (dynamic row matrix) - - - 2 - $i==0 ? m_storage.m_rows : $T2 - m_storage.m_data - - - Backward - 2 - $i==0 ? m_storage.m_rows : $T2 - m_storage.m_data - - - - - - - - empty - [{m_storage.m_cols}] (dynamic column vector) - - m_storage.m_cols - - m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{m_storage.m_rows}] (dynamic row vector) - - m_storage.m_rows - - m_storage.m_rows - m_storage.m_data - - - - - - - - [1] ({m_storage.m_data.array[0]}) - - m_storage.m_data.array[0] - - - - - - - - [2] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - - - - - - - - [3] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - m_storage.m_data.array[2] - - - - - - - - [4] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - m_storage.m_data.array[2] - m_storage.m_data.array[3] - - - - + + + + + + + + [{$T2}, {$T3}] (fixed matrix) + + + 2 + $i==0 ? $T2 : $T3 + m_storage.m_data.array + + + Backward + 2 + $i==0 ? $T2 : $T3 + m_storage.m_data.array + + + + + + + + [2, 2] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[2]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[3]}) + + + + + + + + [3, 3] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[3]}, {m_storage.m_data.array[6]}) + + + ({m_storage.m_data.array[3]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[7]}) + + + ({m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[8]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[8]}) + + + + + + + + [4, 4] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[8]}, {m_storage.m_data.array[12]}) + + + ({m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[13]}) + + + ({m_storage.m_data.array[8]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[11]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[14]}) + + + ({m_storage.m_data.array[12]}, {m_storage.m_data.array[13]}, {m_storage.m_data.array[14]}, {m_storage.m_data.array[15]}) + + + ({m_storage.m_data.array[3]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[11]}, {m_storage.m_data.array[15]}) + + + + + + + + empty + [{m_storage.m_rows}, {m_storage.m_cols}] (dynamic matrix) + + + 2 + $i==0 ? m_storage.m_rows : m_storage.m_cols + m_storage.m_data + + + Backward + 2 + $i==0 ? m_storage.m_rows : m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{$T2}, {m_storage.m_cols}] (dynamic column matrix) + + + 2 + $i==0 ? $T2 : m_storage.m_cols + m_storage.m_data + + + Backward + 2 + $i==0 ? $T2 : m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{m_storage.m_rows}, {$T2}] (dynamic row matrix) + + + 2 + $i==0 ? m_storage.m_rows : $T2 + m_storage.m_data + + + Backward + 2 + $i==0 ? m_storage.m_rows : $T2 + m_storage.m_data + + + + + + + + empty + [{m_storage.m_cols}] (dynamic column vector) + + m_storage.m_cols + + m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{m_storage.m_rows}] (dynamic row vector) + + m_storage.m_rows + + m_storage.m_rows + m_storage.m_data + + + + + + + + [1] ({m_storage.m_data.array[0]}) + + m_storage.m_data.array[0] + + + + + + + + [2] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + + + + + + + + [3] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + m_storage.m_data.array[2] + + + + + + + + [4] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + m_storage.m_data.array[2] + m_storage.m_data.array[3] + + + + diff --git a/third_party/eigen/debug/msvc/eigen_autoexp_part.dat b/third_party/eigen/debug/msvc/eigen_autoexp_part.dat index ba7eefc8e8a..07aa43739e5 100644 --- a/third_party/eigen/debug/msvc/eigen_autoexp_part.dat +++ b/third_party/eigen/debug/msvc/eigen_autoexp_part.dat @@ -1,295 +1,295 @@ -; *************************************************************** -; * Eigen Visualizer -; * -; * Author: Hauke Heibel -; * -; * Support the enhanced debugging of the following Eigen -; * types (*: any, +:fixed dimension) : -; * -; * - Eigen::Matrix<*,4,1,*,*,*> and Eigen::Matrix<*,1,4,*,*,*> -; * - Eigen::Matrix<*,3,1,*,*,*> and Eigen::Matrix<*,1,3,*,*,*> -; * - Eigen::Matrix<*,2,1,*,*,*> and Eigen::Matrix<*,1,2,*,*,*> -; * - Eigen::Matrix<*,-1,-1,*,*,*> -; * - Eigen::Matrix<*,+,-1,*,*,*> -; * - Eigen::Matrix<*,-1,+,*,*,*> -; * - Eigen::Matrix<*,+,+,*,*,*> -; * -; * Matrices are displayed properly independantly of the memory -; * alignment (RowMajor vs. ColMajor). -; * -; * This file is distributed WITHOUT ANY WARRANTY. Please ensure -; * that your original autoexp.dat file is copied to a safe -; * place before proceeding with its modification. -; *************************************************************** - -[Visualizer] - -; Fixed size 4-vectors -Eigen::Matrix<*,4,1,*,*,*>|Eigen::Matrix<*,1,4,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1], - z : ($c.m_storage.m_data.array)[2], - w : ($c.m_storage.m_data.array)[3] - ) - ) - - preview - ( - #( - "[", - 4, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 4), - ")" - ) - ) -} - -; Fixed size 3-vectors -Eigen::Matrix<*,3,1,*,*,*>|Eigen::Matrix<*,1,3,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1], - z : ($c.m_storage.m_data.array)[2] - ) - ) - - preview - ( - #( - "[", - 3, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 3), - ")" - ) - ) -} - -; Fixed size 2-vectors -Eigen::Matrix<*,2,1,*,*,*>|Eigen::Matrix<*,1,2,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1] - ) - ) - - preview - ( - #( - "[", - 2, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 2), - ")" - ) - ) -} - -; Fixed size 1-vectors -Eigen::Matrix<*,1,1,*,*,*>|Eigen::Matrix<*,1,1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0] - ) - ) - - preview - ( - #( - "[", - 1, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 1), - ")" - ) - ) -} - -; Dynamic matrices (ColMajor and RowMajor support) -Eigen::Matrix<*,-1,-1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.m_storage.m_rows, - cols: $c.m_storage.m_cols, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.m_storage.m_cols + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.m_storage.m_rows, - ",", - $c.m_storage.m_cols, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.m_storage.m_rows*$c.m_storage.m_cols - ), - ")" - ) - ) -} - -; Fixed rows, dynamic columns matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,*,-1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.RowsAtCompileTime, - cols: $c.m_storage.m_cols, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.RowsAtCompileTime)*$c.m_storage.m_cols + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.RowsAtCompileTime, - ",", - $c.m_storage.m_cols, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.RowsAtCompileTime*$c.m_storage.m_cols - ), - ")" - ) - ) -} - -; Dynamic rows, fixed columns matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,-1,*,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.m_storage.m_rows, - cols: $c.ColsAtCompileTime, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.ColsAtCompileTime + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.m_storage.m_rows, - ",", - $c.ColsAtCompileTime, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.m_storage.m_rows*$c.ColsAtCompileTime - ), - ")" - ) - ) -} - -; Fixed size matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,*,*,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.RowsAtCompileTime, - cols: $c.ColsAtCompileTime, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data.array)[($i % $c.RowsAtCompileTime)*$c.ColsAtCompileTime + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data.array)[$i], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.RowsAtCompileTime, - ",", - $c.ColsAtCompileTime, - "](", - #array( - expr : [($c.m_storage.m_data.array)[$i],g], - size : $c.RowsAtCompileTime*$c.ColsAtCompileTime - ), - ")" - ) - ) -} +; *************************************************************** +; * Eigen Visualizer +; * +; * Author: Hauke Heibel +; * +; * Support the enhanced debugging of the following Eigen +; * types (*: any, +:fixed dimension) : +; * +; * - Eigen::Matrix<*,4,1,*,*,*> and Eigen::Matrix<*,1,4,*,*,*> +; * - Eigen::Matrix<*,3,1,*,*,*> and Eigen::Matrix<*,1,3,*,*,*> +; * - Eigen::Matrix<*,2,1,*,*,*> and Eigen::Matrix<*,1,2,*,*,*> +; * - Eigen::Matrix<*,-1,-1,*,*,*> +; * - Eigen::Matrix<*,+,-1,*,*,*> +; * - Eigen::Matrix<*,-1,+,*,*,*> +; * - Eigen::Matrix<*,+,+,*,*,*> +; * +; * Matrices are displayed properly independantly of the memory +; * alignment (RowMajor vs. ColMajor). +; * +; * This file is distributed WITHOUT ANY WARRANTY. Please ensure +; * that your original autoexp.dat file is copied to a safe +; * place before proceeding with its modification. +; *************************************************************** + +[Visualizer] + +; Fixed size 4-vectors +Eigen::Matrix<*,4,1,*,*,*>|Eigen::Matrix<*,1,4,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1], + z : ($c.m_storage.m_data.array)[2], + w : ($c.m_storage.m_data.array)[3] + ) + ) + + preview + ( + #( + "[", + 4, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 4), + ")" + ) + ) +} + +; Fixed size 3-vectors +Eigen::Matrix<*,3,1,*,*,*>|Eigen::Matrix<*,1,3,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1], + z : ($c.m_storage.m_data.array)[2] + ) + ) + + preview + ( + #( + "[", + 3, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 3), + ")" + ) + ) +} + +; Fixed size 2-vectors +Eigen::Matrix<*,2,1,*,*,*>|Eigen::Matrix<*,1,2,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1] + ) + ) + + preview + ( + #( + "[", + 2, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 2), + ")" + ) + ) +} + +; Fixed size 1-vectors +Eigen::Matrix<*,1,1,*,*,*>|Eigen::Matrix<*,1,1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0] + ) + ) + + preview + ( + #( + "[", + 1, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 1), + ")" + ) + ) +} + +; Dynamic matrices (ColMajor and RowMajor support) +Eigen::Matrix<*,-1,-1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.m_storage.m_rows, + cols: $c.m_storage.m_cols, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.m_storage.m_cols + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.m_storage.m_rows, + ",", + $c.m_storage.m_cols, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.m_storage.m_rows*$c.m_storage.m_cols + ), + ")" + ) + ) +} + +; Fixed rows, dynamic columns matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,*,-1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.RowsAtCompileTime, + cols: $c.m_storage.m_cols, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.RowsAtCompileTime)*$c.m_storage.m_cols + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.RowsAtCompileTime, + ",", + $c.m_storage.m_cols, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.RowsAtCompileTime*$c.m_storage.m_cols + ), + ")" + ) + ) +} + +; Dynamic rows, fixed columns matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,-1,*,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.m_storage.m_rows, + cols: $c.ColsAtCompileTime, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.ColsAtCompileTime + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.m_storage.m_rows, + ",", + $c.ColsAtCompileTime, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.m_storage.m_rows*$c.ColsAtCompileTime + ), + ")" + ) + ) +} + +; Fixed size matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,*,*,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.RowsAtCompileTime, + cols: $c.ColsAtCompileTime, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data.array)[($i % $c.RowsAtCompileTime)*$c.ColsAtCompileTime + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data.array)[$i], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.RowsAtCompileTime, + ",", + $c.ColsAtCompileTime, + "](", + #array( + expr : [($c.m_storage.m_data.array)[$i],g], + size : $c.RowsAtCompileTime*$c.ColsAtCompileTime + ), + ")" + ) + ) +} diff --git a/third_party/eigen/doc/CMakeLists.txt b/third_party/eigen/doc/CMakeLists.txt index 8ff7559885e..179824dd149 100644 --- a/third_party/eigen/doc/CMakeLists.txt +++ b/third_party/eigen/doc/CMakeLists.txt @@ -11,7 +11,7 @@ if(CMAKE_COMPILER_IS_GNUCXX) endif(CMAKE_COMPILER_IS_GNUCXX) option(EIGEN_INTERNAL_DOCUMENTATION "Build internal documentation" OFF) - +option(EIGEN_DOC_USE_MATHJAX "Use MathJax for rendering math in HTML docs" ON) # Set some Doxygen flags set(EIGEN_DOXY_PROJECT_NAME "Eigen") @@ -19,12 +19,19 @@ set(EIGEN_DOXY_OUTPUT_DIRECTORY_SUFFIX "") set(EIGEN_DOXY_INPUT "\"${Eigen_SOURCE_DIR}/Eigen\" \"${Eigen_SOURCE_DIR}/doc\"") set(EIGEN_DOXY_HTML_COLORSTYLE_HUE "220") set(EIGEN_DOXY_TAGFILES "") + if(EIGEN_INTERNAL_DOCUMENTATION) set(EIGEN_DOXY_INTERNAL "YES") else(EIGEN_INTERNAL_DOCUMENTATION) set(EIGEN_DOXY_INTERNAL "NO") endif(EIGEN_INTERNAL_DOCUMENTATION) +if (EIGEN_DOC_USE_MATHJAX) + set(EIGEN_DOXY_USE_MATHJAX "YES") +else () + set(EIGEN_DOXY_USE_MATHJAX "NO") +endif() + configure_file( ${CMAKE_CURRENT_SOURCE_DIR}/Doxyfile.in ${CMAKE_CURRENT_BINARY_DIR}/Doxyfile diff --git a/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox b/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox index 1ee78cbe5f8..24e5f563b8c 100644 --- a/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox +++ b/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox @@ -75,7 +75,7 @@ namespace Eigen { static inline Real epsilon() { return 0; } static inline Real dummy_precision() { return 0; } - static inline Real digits10() { return 0; } + static inline int digits10() { return 0; } enum { IsInteger = 0, diff --git a/third_party/eigen/doc/Doxyfile.in b/third_party/eigen/doc/Doxyfile.in index 37948a6122d..ac6eafcf927 100644 --- a/third_party/eigen/doc/Doxyfile.in +++ b/third_party/eigen/doc/Doxyfile.in @@ -736,6 +736,14 @@ EXCLUDE = "${Eigen_SOURCE_DIR}/Eigen/src/Core/products" \ "${Eigen_SOURCE_DIR}/unsupported/doc/examples" \ "${Eigen_SOURCE_DIR}/unsupported/doc/snippets" +# Forward declarations of class templates cause the title of the main page for +# the class template to not contain the template signature. This only happens +# when the \class command is used to document the class. Possibly caused +# by https://github.com/doxygen/doxygen/issues/7698. Confirmed fixed by +# doxygen release 1.8.19. + +EXCLUDE += "${Eigen_SOURCE_DIR}/Eigen/src/Core/util/ForwardDeclarations.h" + # The EXCLUDE_SYMLINKS tag can be used to select whether or not files or # directories that are symbolic links (a Unix file system feature) are excluded # from the input. @@ -1245,7 +1253,7 @@ FORMULA_TRANSPARENT = YES # output. When enabled you may also need to install MathJax separately and # configure the path to it using the MATHJAX_RELPATH option. -USE_MATHJAX = NO +USE_MATHJAX = @EIGEN_DOXY_USE_MATHJAX@ # When MathJax is enabled you need to specify the location relative to the # HTML output directory using the MATHJAX_RELPATH option. The destination @@ -1257,12 +1265,12 @@ USE_MATHJAX = NO # However, it is strongly recommended to install a local # copy of MathJax from http://www.mathjax.org before deployment. -MATHJAX_RELPATH = http://cdn.mathjax.org/mathjax/latest +MATHJAX_RELPATH = https://cdn.mathjax.org/mathjax/latest # The MATHJAX_EXTENSIONS tag can be used to specify one or MathJax extension # names that should be enabled during MathJax rendering. -MATHJAX_EXTENSIONS = +MATHJAX_EXTENSIONS = TeX/AMSmath TeX/AMSsymbols # When the SEARCHENGINE tag is enabled doxygen will generate a search box # for the HTML output. The underlying search engine uses javascript @@ -1609,6 +1617,9 @@ PREDEFINED = EIGEN_EMPTY_STRUCT \ EXPAND_AS_DEFINED = EIGEN_MAKE_TYPEDEFS \ EIGEN_MAKE_FIXED_TYPEDEFS \ EIGEN_MAKE_TYPEDEFS_ALL_SIZES \ + EIGEN_MAKE_ARRAY_TYPEDEFS \ + EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS \ + EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES \ EIGEN_CWISE_UNOP_RETURN_TYPE \ EIGEN_CWISE_BINOP_RETURN_TYPE \ EIGEN_CURRENT_STORAGE_BASE_CLASS \ diff --git a/third_party/eigen/doc/Pitfalls.dox b/third_party/eigen/doc/Pitfalls.dox index 3f395053d09..fda40257207 100644 --- a/third_party/eigen/doc/Pitfalls.dox +++ b/third_party/eigen/doc/Pitfalls.dox @@ -7,14 +7,30 @@ namespace Eigen { See this \link TopicTemplateKeyword page \endlink. + \section TopicPitfalls_aliasing Aliasing Don't miss this \link TopicAliasing page \endlink on aliasing, especially if you got wrong results in statements where the destination appears on the right hand side of the expression. + +\section TopicPitfalls_alignment_issue Alignment Issues (runtime assertion) + +%Eigen does explicit vectorization, and while that is appreciated by many users, that also leads to some issues in special situations where data alignment is compromised. +Indeed, since C++17, C++ does not have quite good enough support for explicit data alignment. +In that case your program hits an assertion failure (that is, a "controlled crash") with a message that tells you to consult this page: +\code +http://eigen.tuxfamily.org/dox/group__TopicUnalignedArrayAssert.html +\endcode +Have a look at \link TopicUnalignedArrayAssert it \endlink and see for yourself if that's something that you can cope with. +It contains detailed information about how to deal with each known cause for that issue. + +Now what if you don't care about vectorization and so don't want to be annoyed with these alignment issues? Then read \link getrid how to get rid of them \endlink. + + \section TopicPitfalls_auto_keyword C++11 and the auto keyword -In short: do not use the auto keywords with Eigen's expressions, unless you are 100% sure about what you are doing. In particular, do not use the auto keyword as a replacement for a Matrix<> type. Here is an example: +In short: do not use the auto keywords with %Eigen's expressions, unless you are 100% sure about what you are doing. In particular, do not use the auto keyword as a replacement for a \c Matrix<> type. Here is an example: \code MatrixXd A, B; @@ -22,23 +38,81 @@ auto C = A*B; for(...) { ... w = C * v; ...} \endcode -In this example, the type of C is not a MatrixXd but an abstract expression representing a matrix product and storing references to A and B. Therefore, the product of A*B will be carried out multiple times, once per iteration of the for loop. Moreover, if the coefficients of A or B change during the iteration, then C will evaluate to different values. +In this example, the type of C is not a \c MatrixXd but an abstract expression representing a matrix product and storing references to \c A and \c B. +Therefore, the product of \c A*B will be carried out multiple times, once per iteration of the for loop. +Moreover, if the coefficients of A or B change during the iteration, then C will evaluate to different values. Here is another example leading to a segfault: \code auto C = ((A+B).eval()).transpose(); // do something with C \endcode -The problem is that eval() returns a temporary object (in this case a MatrixXd) which is then referenced by the Transpose<> expression. However, this temporary is deleted right after the first line, and there the C expression reference a dead object. The same issue might occur when sub expressions are automatically evaluated by Eigen as in the following example: +The problem is that \c eval() returns a temporary object (in this case a \c MatrixXd) which is then referenced by the \c Transpose<> expression. +However, this temporary is deleted right after the first line, and then the \c C expression references a dead object. +One possible fix consists in applying \c eval() on the whole expression: +\code +auto C = (A+B).transpose().eval(); +\endcode + +The same issue might occur when sub expressions are automatically evaluated by %Eigen as in the following example: \code VectorXd u, v; auto C = u + (A*v).normalized(); // do something with C \endcode -where the normalized() method has to evaluate the expensive product A*v to avoid evaluating it twice. On the other hand, the following example is perfectly fine: +Here the \c normalized() method has to evaluate the expensive product \c A*v to avoid evaluating it twice. +Again, one possible fix is to call \c .eval() on the whole expression: \code auto C = (u + (A*v).normalized()).eval(); \endcode -In this case, C will be a regular VectorXd object. +In this case, \c C will be a regular \c VectorXd object. +Note that DenseBase::eval() is smart enough to avoid copies when the underlying expression is already a plain \c Matrix<>. + + +\section TopicPitfalls_header_issues Header Issues (failure to compile) + +With all libraries, one must check the documentation for which header to include. +The same is true with %Eigen, but slightly worse: with %Eigen, a method in a class may require an additional #include over what the class itself requires! +For example, if you want to use the \c cross() method on a vector (it computes a cross-product) then you need to: +\code +#include +\endcode +We try to always document this, but do tell us if we forgot an occurrence. + + +\section TopicPitfalls_ternary_operator Ternary operator + +In short: avoid the use of the ternary operator (COND ? THEN : ELSE) with %Eigen's expressions for the \c THEN and \c ELSE statements. +To see why, let's consider the following example: +\code +Vector3f A; +A << 1, 2, 3; +Vector3f B = ((1 < 0) ? (A.reverse()) : A); +\endcode +This example will return B = 3, 2, 1. Do you see why? +The reason is that in c++ the type of the \c ELSE statement is inferred from the type of the \c THEN expression such that both match. +Since \c THEN is a Reverse, the \c ELSE statement A is converted to a Reverse, and the compiler thus generates: +\code +Vector3f B = ((1 < 0) ? (A.reverse()) : Reverse(A)); +\endcode +In this very particular case, a workaround would be to call A.reverse().eval() for the \c THEN statement, but the safest and fastest is really to avoid this ternary operator with %Eigen's expressions and use a if/else construct. + + +\section TopicPitfalls_pass_by_value Pass-by-value + +If you don't know why passing-by-value is wrong with %Eigen, read this \link TopicPassingByValue page \endlink first. + +While you may be extremely careful and use care to make sure that all of your code that explicitly uses %Eigen types is pass-by-reference you have to watch out for templates which define the argument types at compile time. + +If a template has a function that takes arguments pass-by-value, and the relevant template parameter ends up being an %Eigen type, then you will of course have the same alignment problems that you would in an explicitly defined function passing %Eigen types by reference. + +Using %Eigen types with other third party libraries or even the STL can present the same problem. +boost::bind for example uses pass-by-value to store arguments in the returned functor. +This will of course be a problem. + +There are at least two ways around this: + - If the value you are passing is guaranteed to be around for the life of the functor, you can use boost::ref() to wrap the value as you pass it to boost::bind. Generally this is not a solution for values on the stack as if the functor ever gets passed to a lower or independent scope, the object may be gone by the time it's attempted to be used. + - The other option is to make your functions take a reference counted pointer like boost::shared_ptr as the argument. This avoids needing to worry about managing the lifetime of the object being passed. + */ } diff --git a/third_party/eigen/doc/SparseQuickReference.dox b/third_party/eigen/doc/SparseQuickReference.dox index a25622e800e..653bf33ef84 100644 --- a/third_party/eigen/doc/SparseQuickReference.dox +++ b/third_party/eigen/doc/SparseQuickReference.dox @@ -244,7 +244,7 @@ As stated earlier, for a read-write sub-matrix (RW), the evaluation can be done \code sm1.valuePtr(); // Pointer to the values -sm1.innerIndextr(); // Pointer to the indices. +sm1.innerIndexPtr(); // Pointer to the indices. sm1.outerIndexPtr(); // Pointer to the beginning of each inner vector \endcode diff --git a/third_party/eigen/doc/TopicLazyEvaluation.dox b/third_party/eigen/doc/TopicLazyEvaluation.dox index 101ef8c72c2..d2a704f132a 100644 --- a/third_party/eigen/doc/TopicLazyEvaluation.dox +++ b/third_party/eigen/doc/TopicLazyEvaluation.dox @@ -2,63 +2,95 @@ namespace Eigen { /** \page TopicLazyEvaluation Lazy Evaluation and Aliasing -Executive summary: Eigen has intelligent compile-time mechanisms to enable lazy evaluation and removing temporaries where appropriate. +Executive summary: %Eigen has intelligent compile-time mechanisms to enable lazy evaluation and removing temporaries where appropriate. It will handle aliasing automatically in most cases, for example with matrix products. The automatic behavior can be overridden manually by using the MatrixBase::eval() and MatrixBase::noalias() methods. When you write a line of code involving a complex expression such as -\code mat1 = mat2 + mat3 * (mat4 + mat5); \endcode +\code mat1 = mat2 + mat3 * (mat4 + mat5); +\endcode -Eigen determines automatically, for each sub-expression, whether to evaluate it into a temporary variable. Indeed, in certain cases it is better to evaluate immediately a sub-expression into a temporary variable, while in other cases it is better to avoid that. +%Eigen determines automatically, for each sub-expression, whether to evaluate it into a temporary variable. Indeed, in certain cases it is better to evaluate a sub-expression into a temporary variable, while in other cases it is better to avoid that. A traditional math library without expression templates always evaluates all sub-expressions into temporaries. So with this code, -\code vec1 = vec2 + vec3; \endcode +\code vec1 = vec2 + vec3; +\endcode a traditional library would evaluate \c vec2 + vec3 into a temporary \c vec4 and then copy \c vec4 into \c vec1. This is of course inefficient: the arrays are traversed twice, so there are a lot of useless load/store operations. -Expression-templates-based libraries can avoid evaluating sub-expressions into temporaries, which in many cases results in large speed improvements. This is called lazy evaluation as an expression is getting evaluated as late as possible, instead of immediately. However, most other expression-templates-based libraries always choose lazy evaluation. There are two problems with that: first, lazy evaluation is not always a good choice for performance; second, lazy evaluation can be very dangerous, for example with matrix products: doing matrix = matrix*matrix gives a wrong result if the matrix product is lazy-evaluated, because of the way matrix product works. +Expression-templates-based libraries can avoid evaluating sub-expressions into temporaries, which in many cases results in large speed improvements. +This is called lazy evaluation as an expression is getting evaluated as late as possible. +In %Eigen all expressions are lazy-evaluated. +More precisely, an expression starts to be evaluated once it is assigned to a matrix. +Until then nothing happens beyond constructing the abstract expression tree. +In contrast to most other expression-templates-based libraries, however, %Eigen might choose to evaluate some sub-expressions into temporaries. +There are two reasons for that: first, pure lazy evaluation is not always a good choice for performance; second, pure lazy evaluation can be very dangerous, for example with matrix products: doing mat = mat*mat gives a wrong result if the matrix product is directly evaluated within the destination matrix, because of the way matrix product works. -For these reasons, Eigen has intelligent compile-time mechanisms to determine automatically when to use lazy evaluation, and when on the contrary it should evaluate immediately into a temporary variable. +For these reasons, %Eigen has intelligent compile-time mechanisms to determine automatically which sub-expression should be evaluated into a temporary variable. So in the basic example, -\code matrix1 = matrix2 + matrix3; \endcode +\code mat1 = mat2 + mat3; +\endcode -Eigen chooses lazy evaluation. Thus the arrays are traversed only once, producing optimized code. If you really want to force immediate evaluation, use \link MatrixBase::eval() eval()\endlink: +%Eigen chooses not to introduce any temporary. Thus the arrays are traversed only once, producing optimized code. +If you really want to force immediate evaluation, use \link MatrixBase::eval() eval()\endlink: -\code matrix1 = (matrix2 + matrix3).eval(); \endcode +\code mat1 = (mat2 + mat3).eval(); +\endcode Here is now a more involved example: -\code matrix1 = -matrix2 + matrix3 + 5 * matrix4; \endcode +\code mat1 = -mat2 + mat3 + 5 * mat4; +\endcode -Eigen chooses lazy evaluation at every stage in that example, which is clearly the correct choice. In fact, lazy evaluation is the "default choice" and Eigen will choose it except in a few circumstances. +Here again %Eigen won't introduce any temporary, thus producing a single fused evaluation loop, which is clearly the correct choice. -The first circumstance in which Eigen chooses immediate evaluation, is when it sees an assignment a = b; and the expression \c b has the evaluate-before-assigning \link flags flag\endlink. The most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do +\section TopicLazyEvaluationWhichExpr Which sub-expressions are evaluated into temporaries? -\code matrix = matrix * matrix; \endcode +The default evaluation strategy is to fuse the operations in a single loop, and %Eigen will choose it except in a few circumstances. -Eigen first evaluates matrix * matrix into a temporary matrix, and then copies it into the original \c matrix. This guarantees a correct result as we saw above that lazy evaluation gives wrong results with matrix products. It also doesn't cost much, as the cost of the matrix product itself is much higher. +The first circumstance in which %Eigen chooses to evaluate a sub-expression is when it sees an assignment a = b; and the expression \c b has the evaluate-before-assigning \link flags flag\endlink. +The most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do + +\code mat = mat * mat; +\endcode + +%Eigen will evaluate mat * mat into a temporary matrix, and then copies it into the original \c mat. +This guarantees a correct result as we saw above that lazy evaluation gives wrong results with matrix products. +It also doesn't cost much, as the cost of the matrix product itself is much higher. +Note that this temporary is introduced at evaluation time only, that is, within operator= in this example. +The expression mat * mat still return a abstract product type. What if you know that the result does no alias the operand of the product and want to force lazy evaluation? Then use \link MatrixBase::noalias() .noalias()\endlink instead. Here is an example: -\code matrix1.noalias() = matrix2 * matrix2; \endcode +\code mat1.noalias() = mat2 * mat2; +\endcode -Here, since we know that matrix2 is not the same matrix as matrix1, we know that lazy evaluation is not dangerous, so we may force lazy evaluation. Concretely, the effect of noalias() here is to bypass the evaluate-before-assigning \link flags flag\endlink. +Here, since we know that mat2 is not the same matrix as mat1, we know that lazy evaluation is not dangerous, so we may force lazy evaluation. Concretely, the effect of noalias() here is to bypass the evaluate-before-assigning \link flags flag\endlink. -The second circumstance in which Eigen chooses immediate evaluation, is when it sees a nested expression such as a + b where \c b is already an expression having the evaluate-before-nesting \link flags flag\endlink. Again, the most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do +The second circumstance in which %Eigen chooses to evaluate a sub-expression, is when it sees a nested expression such as a + b where \c b is already an expression having the evaluate-before-nesting \link flags flag\endlink. +Again, the most important example of such an expression is the \link Product matrix product expression\endlink. +For example, when you do -\code matrix1 = matrix2 + matrix3 * matrix4; \endcode +\code mat1 = mat2 * mat3 + mat4 * mat5; +\endcode -the product matrix3 * matrix4 gets evaluated immediately into a temporary matrix. Indeed, experiments showed that it is often beneficial for performance to evaluate immediately matrix products when they are nested into bigger expressions. +the products mat2 * mat3 and mat4 * mat5 gets evaluated separately into temporary matrices before being summed up in mat1. +Indeed, to be efficient matrix products need to be evaluated within a destination matrix at hand, and not as simple "dot products". +For small matrices, however, you might want to enforce a "dot-product" based lazy evaluation with lazyProduct(). +Again, it is important to understand that those temporaries are created at evaluation time only, that is in operator =. +See TopicPitfalls_auto_keyword for common pitfalls regarding this remark. -The third circumstance in which Eigen chooses immediate evaluation, is when its cost model shows that the total cost of an operation is reduced if a sub-expression gets evaluated into a temporary. Indeed, in certain cases, an intermediate result is sufficiently costly to compute and is reused sufficiently many times, that is worth "caching". Here is an example: +The third circumstance in which %Eigen chooses to evaluate a sub-expression, is when its cost model shows that the total cost of an operation is reduced if a sub-expression gets evaluated into a temporary. +Indeed, in certain cases, an intermediate result is sufficiently costly to compute and is reused sufficiently many times, that is worth "caching". Here is an example: -\code matrix1 = matrix2 * (matrix3 + matrix4); \endcode +\code mat1 = mat2 * (mat3 + mat4); +\endcode -Here, provided the matrices have at least 2 rows and 2 columns, each coefficienct of the expression matrix3 + matrix4 is going to be used several times in the matrix product. Instead of computing the sum everytime, it is much better to compute it once and store it in a temporary variable. Eigen understands this and evaluates matrix3 + matrix4 into a temporary variable before evaluating the product. +Here, provided the matrices have at least 2 rows and 2 columns, each coefficient of the expression mat3 + mat4 is going to be used several times in the matrix product. Instead of computing the sum every time, it is much better to compute it once and store it in a temporary variable. %Eigen understands this and evaluates mat3 + mat4 into a temporary variable before evaluating the product. */ diff --git a/third_party/eigen/doc/TopicMultithreading.dox b/third_party/eigen/doc/TopicMultithreading.dox index 47c9b261f89..a2855745b60 100644 --- a/third_party/eigen/doc/TopicMultithreading.dox +++ b/third_party/eigen/doc/TopicMultithreading.dox @@ -49,6 +49,7 @@ int main(int argc, char** argv) In the case your application is parallelized with OpenMP, you might want to disable Eigen's own parallization as detailed in the previous section. +\warning Using OpenMP with custom scalar types that might throw exceptions can lead to unexpected behaviour in the event of throwing. */ } diff --git a/third_party/eigen/doc/TutorialGeometry.dox b/third_party/eigen/doc/TutorialGeometry.dox index 2e1420f98c8..723f4dbcee9 100644 --- a/third_party/eigen/doc/TutorialGeometry.dox +++ b/third_party/eigen/doc/TutorialGeometry.dox @@ -232,8 +232,8 @@ On the other hand, since there exist 24 different conventions, they are pretty c to create a rotation matrix according to the 2-1-2 convention.\code Matrix3f m; m = AngleAxisf(angle1, Vector3f::UnitZ()) - * AngleAxisf(angle2, Vector3f::UnitY()) - * AngleAxisf(angle3, Vector3f::UnitZ()); + * * AngleAxisf(angle2, Vector3f::UnitY()) + * * AngleAxisf(angle3, Vector3f::UnitZ()); \endcode diff --git a/third_party/eigen/doc/eigen_navtree_hacks.js b/third_party/eigen/doc/eigen_navtree_hacks.js index a6f8c342821..afb97edf5c7 100644 --- a/third_party/eigen/doc/eigen_navtree_hacks.js +++ b/third_party/eigen/doc/eigen_navtree_hacks.js @@ -5,6 +5,7 @@ function generate_autotoc() { if(headers.length > 1) { var toc = $("#side-nav").append('

'); toc = $("#nav-toc"); + var footer = $("#nav-path"); var footerHeight = footer.height(); toc = toc.append('
    '); toc = toc.find('ul'); @@ -137,7 +138,7 @@ function initNavTree(toroot,relpath) } }) - $(window).load(showRoot); + $(window).on("load", showRoot); } // return false if the the node has no children at all, or has only section/subsection children @@ -241,6 +242,6 @@ $(document).ready(function() { } })(); - $(window).load(resizeHeight); + $(window).on("load", resizeHeight); }); diff --git a/third_party/eigen/doc/eigendoxy_footer.html.in b/third_party/eigen/doc/eigendoxy_footer.html.in index 9ac0596cb0a..126653589d5 100644 --- a/third_party/eigen/doc/eigendoxy_footer.html.in +++ b/third_party/eigen/doc/eigendoxy_footer.html.in @@ -17,19 +17,6 @@ $generatedby   - - - - diff --git a/third_party/eigen/doc/eigendoxy_header.html.in b/third_party/eigen/doc/eigendoxy_header.html.in index bb149f8f0f1..a6b1c1d081a 100644 --- a/third_party/eigen/doc/eigendoxy_header.html.in +++ b/third_party/eigen/doc/eigendoxy_header.html.in @@ -20,6 +20,9 @@ $mathjax + + +
    diff --git a/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp b/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp index 76f49f2fbc6..0b87313a1cf 100644 --- a/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp +++ b/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp @@ -14,5 +14,5 @@ int main() a.block<2,2>(1,1) = m; cout << "Here is now a with m copied into its central 2x2 block:" << endl << a << endl << endl; a.block(0,0,2,3) = a.block(2,1,2,3); - cout << "Here is now a with bottom-right 2x3 block copied into top-left 2x2 block:" << endl << a << endl << endl; + cout << "Here is now a with bottom-right 2x3 block copied into top-left 2x3 block:" << endl << a << endl << endl; } diff --git a/third_party/eigen/lapack/CMakeLists.txt b/third_party/eigen/lapack/CMakeLists.txt index 6df1fa9584b..fbecd662498 100644 --- a/third_party/eigen/lapack/CMakeLists.txt +++ b/third_party/eigen/lapack/CMakeLists.txt @@ -133,12 +133,14 @@ if(EXISTS ${eigen_full_path_to_testing_lapack}) string(REGEX REPLACE "(.*)/STACK:(.*) (.*)" "\\1/STACK:900000000000000000 \\3" CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}") endif() + file(MAKE_DIRECTORY "${LAPACK_BINARY_DIR}/TESTING") add_subdirectory(testing/MATGEN) add_subdirectory(testing/LIN) add_subdirectory(testing/EIG) + cmake_policy(SET CMP0026 OLD) macro(add_lapack_test output input target) set(TEST_INPUT "${LAPACK_SOURCE_DIR}/testing/${input}") - set(TEST_OUTPUT "${LAPACK_BINARY_DIR}/testing/${output}") + set(TEST_OUTPUT "${LAPACK_BINARY_DIR}/TESTING/${output}") get_target_property(TEST_LOC ${target} LOCATION) string(REPLACE "." "_" input_name ${input}) set(testName "${target}_${input_name}") diff --git a/third_party/eigen/scripts/CMakeLists.txt b/third_party/eigen/scripts/CMakeLists.txt index 5f08da5b9f9..0d9a631a9dd 100644 --- a/third_party/eigen/scripts/CMakeLists.txt +++ b/third_party/eigen/scripts/CMakeLists.txt @@ -1,6 +1,6 @@ get_property(EIGEN_TESTS_LIST GLOBAL PROPERTY EIGEN_TESTS_LIST) -configure_file(buildtests.in ${CMAKE_CURRENT_BINARY_DIR}/buildtests.sh @ONLY) +configure_file(buildtests.in ${CMAKE_BINARY_DIR}/buildtests.sh @ONLY) -configure_file(check.in ${CMAKE_CURRENT_BINARY_DIR}/check.sh COPYONLY) -configure_file(debug.in ${CMAKE_CURRENT_BINARY_DIR}/debug.sh COPYONLY) -configure_file(release.in ${CMAKE_CURRENT_BINARY_DIR}/release.sh COPYONLY) +configure_file(check.in ${CMAKE_BINARY_DIR}/check.sh COPYONLY) +configure_file(debug.in ${CMAKE_BINARY_DIR}/debug.sh COPYONLY) +configure_file(release.in ${CMAKE_BINARY_DIR}/release.sh COPYONLY) diff --git a/third_party/eigen/test/CMakeLists.txt b/third_party/eigen/test/CMakeLists.txt index 0747aa6cb65..47e6fee4b27 100644 --- a/third_party/eigen/test/CMakeLists.txt +++ b/third_party/eigen/test/CMakeLists.txt @@ -163,7 +163,7 @@ ei_add_test(constructor) ei_add_test(linearstructure) ei_add_test(integer_types) ei_add_test(unalignedcount) -if(NOT EIGEN_TEST_NO_EXCEPTIONS) +if(NOT EIGEN_TEST_NO_EXCEPTIONS AND NOT EIGEN_TEST_OPENMP) ei_add_test(exceptions) endif() ei_add_test(redux) @@ -185,7 +185,7 @@ ei_add_test(smallvectors) ei_add_test(mapped_matrix) ei_add_test(mapstride) ei_add_test(mapstaticmethods) -ei_add_test(array) +ei_add_test(array_cwise) ei_add_test(array_for_matrix) ei_add_test(array_replicate) ei_add_test(array_reverse) diff --git a/third_party/eigen/test/bdcsvd.cpp b/third_party/eigen/test/bdcsvd.cpp index 6c7b09696cf..3ca273635c4 100644 --- a/third_party/eigen/test/bdcsvd.cpp +++ b/third_party/eigen/test/bdcsvd.cpp @@ -28,9 +28,13 @@ template void bdcsvd(const MatrixType& a = MatrixType(), bool pickrandom = true) { - MatrixType m = a; - if(pickrandom) + MatrixType m; + if(pickrandom) { + m.resizeLike(a); svd_fill_random(m); + } + else + m = a; CALL_SUBTEST(( svd_test_all_computation_options >(m, false) )); } diff --git a/third_party/eigen/test/constructor.cpp b/third_party/eigen/test/constructor.cpp index eec9e219296..9885399510a 100644 --- a/third_party/eigen/test/constructor.cpp +++ b/third_party/eigen/test/constructor.cpp @@ -20,6 +20,8 @@ template struct Wrapper inline operator MatrixType& () { return m_mat; } }; +enum my_sizes { M = 12, N = 7}; + template void ctor_init1(const MatrixType& m) { // Check logic in PlainObjectBase::_init1 @@ -81,4 +83,16 @@ void test_constructor() Array a(123); VERIFY_IS_EQUAL(a(4), 123.f); } + { + MatrixXi m1(M,N); + VERIFY_IS_EQUAL(m1.rows(),M); + VERIFY_IS_EQUAL(m1.cols(),N); + ArrayXXi a1(M,N); + VERIFY_IS_EQUAL(a1.rows(),M); + VERIFY_IS_EQUAL(a1.cols(),N); + VectorXi v1(M); + VERIFY_IS_EQUAL(v1.size(),M); + ArrayXi a2(M); + VERIFY_IS_EQUAL(a2.size(),M); + } } diff --git a/third_party/eigen/test/ctorleak.cpp b/third_party/eigen/test/ctorleak.cpp index c158f5e4ee4..d73fecfe2c7 100644 --- a/third_party/eigen/test/ctorleak.cpp +++ b/third_party/eigen/test/ctorleak.cpp @@ -8,7 +8,7 @@ struct Foo static Index object_limit; int dummy; - Foo() + Foo() : dummy(0) { #ifdef EIGEN_EXCEPTIONS // TODO: Is this the correct way to handle this? @@ -37,22 +37,33 @@ void test_ctorleak() { typedef Matrix MatrixX; typedef Matrix VectorX; + Foo::object_count = 0; for(int i = 0; i < g_repeat; i++) { Index rows = internal::random(2,EIGEN_TEST_MAX_SIZE), cols = internal::random(2,EIGEN_TEST_MAX_SIZE); - Foo::object_limit = internal::random(0, rows*cols - 2); + Foo::object_limit = rows*cols; + { + MatrixX r(rows, cols); + Foo::object_limit = r.size()+internal::random(0, rows*cols - 2); std::cout << "object_limit =" << Foo::object_limit << std::endl; #ifdef EIGEN_EXCEPTIONS try { #endif - std::cout << "\nMatrixX m(" << rows << ", " << cols << ");\n"; - MatrixX m(rows, cols); + if(internal::random()) { + std::cout << "\nMatrixX m(" << rows << ", " << cols << ");\n"; + MatrixX m(rows, cols); + } + else { + std::cout << "\nMatrixX m(r);\n"; + MatrixX m(r); + } #ifdef EIGEN_EXCEPTIONS VERIFY(false); // not reached if exceptions are enabled } catch (const Foo::Fail&) { /* ignore */ } #endif + } VERIFY_IS_EQUAL(Index(0), Foo::object_count); { @@ -66,4 +77,5 @@ void test_ctorleak() } VERIFY_IS_EQUAL(Index(0), Foo::object_count); } + std::cout << "\n"; } diff --git a/third_party/eigen/test/eigensolver_generic.cpp b/third_party/eigen/test/eigensolver_generic.cpp index 07bf65e0327..5c13175690b 100644 --- a/third_party/eigen/test/eigensolver_generic.cpp +++ b/third_party/eigen/test/eigensolver_generic.cpp @@ -67,7 +67,7 @@ template void eigensolver(const MatrixType& m) // Test matrix with NaN a(0,0) = std::numeric_limits::quiet_NaN(); EigenSolver eiNaN(a); - VERIFY_IS_EQUAL(eiNaN.info(), NoConvergence); + VERIFY_IS_NOT_EQUAL(eiNaN.info(), Success); } // regression test for bug 1098 diff --git a/third_party/eigen/test/exceptions.cpp b/third_party/eigen/test/exceptions.cpp index b83fb82ba60..015b9fd33f3 100644 --- a/third_party/eigen/test/exceptions.cpp +++ b/third_party/eigen/test/exceptions.cpp @@ -109,5 +109,7 @@ void memoryleak() void test_exceptions() { - CALL_SUBTEST( memoryleak() ); + EIGEN_TRY { + CALL_SUBTEST( memoryleak() ); + } EIGEN_CATCH(...) {} } diff --git a/third_party/eigen/test/fastmath.cpp b/third_party/eigen/test/fastmath.cpp index cc5db074637..e84bdc97232 100644 --- a/third_party/eigen/test/fastmath.cpp +++ b/third_party/eigen/test/fastmath.cpp @@ -43,11 +43,11 @@ void check_inf_nan(bool dryrun) { } else { - VERIFY( !(numext::isfinite)(m(3)) ); - VERIFY( !(numext::isinf)(m(3)) ); - VERIFY( (numext::isnan)(m(3)) ); - VERIFY( !m.allFinite() ); - VERIFY( m.hasNaN() ); + if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !(numext::isfinite)(m(3)) ); g_test_level=0; + if( (std::isinf) (m(3))) g_test_level=1; VERIFY( !(numext::isinf)(m(3)) ); g_test_level=0; + if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( (numext::isnan)(m(3)) ); g_test_level=0; + if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; + if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( m.hasNaN() ); g_test_level=0; } T hidden_zero = (std::numeric_limits::min)()*(std::numeric_limits::min)(); m(4) /= hidden_zero; @@ -62,29 +62,29 @@ void check_inf_nan(bool dryrun) { } else { - VERIFY( !(numext::isfinite)(m(4)) ); - VERIFY( (numext::isinf)(m(4)) ); - VERIFY( !(numext::isnan)(m(4)) ); - VERIFY( !m.allFinite() ); - VERIFY( m.hasNaN() ); + if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !(numext::isfinite)(m(4)) ); g_test_level=0; + if(!(std::isinf) (m(3))) g_test_level=1; VERIFY( (numext::isinf)(m(4)) ); g_test_level=0; + if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !(numext::isnan)(m(4)) ); g_test_level=0; + if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; + if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( m.hasNaN() ); g_test_level=0; } m(3) = 0; if(dryrun) { std::cout << "std::isfinite(" << m(3) << ") = "; check((std::isfinite)(m(3)),true); std::cout << " ; numext::isfinite = "; check((numext::isfinite)(m(3)), true); std::cout << "\n"; - std::cout << "std::isinf(" << m(3) << ") = "; check((std::isinf)(m(3)),false); std::cout << " ; numext::isinf = "; check((numext::isinf)(m(3)), false); std::cout << "\n"; - std::cout << "std::isnan(" << m(3) << ") = "; check((std::isnan)(m(3)),false); std::cout << " ; numext::isnan = "; check((numext::isnan)(m(3)), false); std::cout << "\n"; + std::cout << "std::isinf(" << m(3) << ") = "; check((std::isinf)(m(3)),false); std::cout << " ; numext::isinf = "; check((numext::isinf)(m(3)), false); std::cout << "\n"; + std::cout << "std::isnan(" << m(3) << ") = "; check((std::isnan)(m(3)),false); std::cout << " ; numext::isnan = "; check((numext::isnan)(m(3)), false); std::cout << "\n"; std::cout << "allFinite: "; check(m.allFinite(), 0); std::cout << "\n"; std::cout << "hasNaN: "; check(m.hasNaN(), 0); std::cout << "\n"; std::cout << "\n\n"; } else { - VERIFY( (numext::isfinite)(m(3)) ); - VERIFY( !(numext::isinf)(m(3)) ); - VERIFY( !(numext::isnan)(m(3)) ); - VERIFY( !m.allFinite() ); - VERIFY( !m.hasNaN() ); + if(!(std::isfinite)(m(3))) g_test_level=1; VERIFY( (numext::isfinite)(m(3)) ); g_test_level=0; + if( (std::isinf) (m(3))) g_test_level=1; VERIFY( !(numext::isinf)(m(3)) ); g_test_level=0; + if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !(numext::isnan)(m(3)) ); g_test_level=0; + if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; + if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !m.hasNaN() ); g_test_level=0; } } diff --git a/third_party/eigen/test/geo_alignedbox.cpp b/third_party/eigen/test/geo_alignedbox.cpp index b64ea3bdc47..4cf51aafb58 100644 --- a/third_party/eigen/test/geo_alignedbox.cpp +++ b/third_party/eigen/test/geo_alignedbox.cpp @@ -15,8 +15,9 @@ #include using namespace std; +// TODO not sure if this is actually still necessary anywhere ... template EIGEN_DONT_INLINE -void kill_extra_precision(T& x) { eigen_assert((void*)(&x) != (void*)0); } +void kill_extra_precision(T& ) { } template void alignedbox(const BoxType& _box) diff --git a/third_party/eigen/test/geo_quaternion.cpp b/third_party/eigen/test/geo_quaternion.cpp index 8ee8fdb27f1..87680f1ccac 100644 --- a/third_party/eigen/test/geo_quaternion.cpp +++ b/third_party/eigen/test/geo_quaternion.cpp @@ -244,6 +244,14 @@ template void mapQuaternion(void){ // is used to determine wether we can return a coeff by reference or not, which is not enough for Map. //const MCQuaternionUA& cmcq3(mcq3); //VERIFY( &cmcq3.x() == &mcq3.x() ); + + // test cast + { + Quaternion q1f = mq1.template cast(); + VERIFY_IS_APPROX(q1f.template cast(),mq1); + Quaternion q1d = mq1.template cast(); + VERIFY_IS_APPROX(q1d.template cast(),mq1); + } } template void quaternionAlignment(void){ diff --git a/third_party/eigen/test/geo_transformations.cpp b/third_party/eigen/test/geo_transformations.cpp index 278e527c252..8d064ddc3bc 100755 --- a/third_party/eigen/test/geo_transformations.cpp +++ b/third_party/eigen/test/geo_transformations.cpp @@ -612,6 +612,62 @@ template void transform_products() VERIFY_IS_APPROX((ac*p).matrix(), a_m*p_m); } +template void transformations_no_scale() +{ + /* this test covers the following files: + Cross.h Quaternion.h, Transform.h + */ + typedef Matrix Vector3; + typedef Matrix Vector4; + typedef Quaternion Quaternionx; + typedef AngleAxis AngleAxisx; + typedef Transform Transform3; + typedef Translation Translation3; + typedef Matrix Matrix4; + + Vector3 v0 = Vector3::Random(), + v1 = Vector3::Random(); + + Transform3 t0, t1, t2; + + Scalar a = internal::random(-Scalar(EIGEN_PI), Scalar(EIGEN_PI)); + + Quaternionx q1, q2; + + q1 = AngleAxisx(a, v0.normalized()); + + t0 = Transform3::Identity(); + VERIFY_IS_APPROX(t0.matrix(), Transform3::MatrixType::Identity()); + + t0.setIdentity(); + t1.setIdentity(); + v1 = Vector3::Ones(); + t0.linear() = q1.toRotationMatrix(); + t0.pretranslate(v0); + t1.linear() = q1.conjugate().toRotationMatrix(); + t1.translate(-v0); + + VERIFY((t0 * t1).matrix().isIdentity(test_precision())); + + t1.fromPositionOrientationScale(v0, q1, v1); + VERIFY_IS_APPROX(t1.matrix(), t0.matrix()); + VERIFY_IS_APPROX(t1*v1, t0*v1); + + // translation * vector + t0.setIdentity(); + t0.translate(v0); + VERIFY_IS_APPROX((t0 * v1).template head<3>(), Translation3(v0) * v1); + + // Conversion to matrix. + Transform3 t3; + t3.linear() = q1.toRotationMatrix(); + t3.translation() = v1; + Matrix4 m3 = t3.matrix(); + VERIFY((m3 * m3.inverse()).isIdentity(test_precision())); + // Verify implicit last row is initialized. + VERIFY_IS_APPROX(Vector4(m3.row(3)), Vector4(0.0, 0.0, 0.0, 1.0)); +} + void test_geo_transformations() { for(int i = 0; i < g_repeat; i++) { @@ -625,7 +681,7 @@ void test_geo_transformations() CALL_SUBTEST_3(( transformations() )); CALL_SUBTEST_3(( transformations() )); CALL_SUBTEST_3(( transform_alignment() )); - + CALL_SUBTEST_4(( transformations() )); CALL_SUBTEST_4(( non_projective_only() )); @@ -641,5 +697,8 @@ void test_geo_transformations() CALL_SUBTEST_8(( transform_associativity(Rotation2D(internal::random()*double(EIGEN_PI))) )); CALL_SUBTEST_8(( transform_associativity(Quaterniond::UnitRandom()) )); + + CALL_SUBTEST_9(( transformations_no_scale() )); + CALL_SUBTEST_9(( transformations_no_scale() )); } } diff --git a/third_party/eigen/test/inverse.cpp b/third_party/eigen/test/inverse.cpp index be607cc8bea..d81af26c1e6 100644 --- a/third_party/eigen/test/inverse.cpp +++ b/third_party/eigen/test/inverse.cpp @@ -92,6 +92,22 @@ template void inverse(const MatrixType& m) } } +template +void inverse_zerosized() +{ + Matrix A(0,0); + { + Matrix b, x; + x = A.inverse() * b; + } + { + Matrix b(0,1), x; + x = A.inverse() * b; + VERIFY_IS_EQUAL(x.rows(), 0); + VERIFY_IS_EQUAL(x.cols(), 1); + } +} + void test_inverse() { int s = 0; @@ -105,6 +121,7 @@ void test_inverse() s = internal::random(50,320); CALL_SUBTEST_5( inverse(MatrixXf(s,s)) ); TEST_SET_BUT_UNUSED_VARIABLE(s) + CALL_SUBTEST_5( inverse_zerosized() ); s = internal::random(25,100); CALL_SUBTEST_6( inverse(MatrixXcd(s,s)) ); diff --git a/third_party/eigen/test/main.h b/third_party/eigen/test/main.h index 8c868ee79e7..18bb5c82550 100644 --- a/third_party/eigen/test/main.h +++ b/third_party/eigen/test/main.h @@ -72,6 +72,11 @@ #define isnan(X) please_protect_your_isnan_with_parentheses #define isinf(X) please_protect_your_isinf_with_parentheses #define isfinite(X) please_protect_your_isfinite_with_parentheses + +// test possible conflicts +struct real {}; +struct imag {}; + #ifdef M_PI #undef M_PI #endif diff --git a/third_party/eigen/test/numext.cpp b/third_party/eigen/test/numext.cpp index 3de33e2f918..beba9e911b3 100644 --- a/third_party/eigen/test/numext.cpp +++ b/third_party/eigen/test/numext.cpp @@ -12,6 +12,7 @@ template void check_abs() { typedef typename NumTraits::Real Real; + Real zero(0); if(NumTraits::IsSigned) VERIFY_IS_EQUAL(numext::abs(-T(1)), T(1)); @@ -26,9 +27,9 @@ void check_abs() { if(NumTraits::IsSigned) { VERIFY_IS_EQUAL(numext::abs(x), numext::abs(-x)); - VERIFY( numext::abs(-x) >= Real(0)); + VERIFY( numext::abs(-x) >= zero ); } - VERIFY( numext::abs(x) >= Real(0)); + VERIFY( numext::abs(x) >= zero ); VERIFY_IS_APPROX( numext::abs2(x), numext::abs2(numext::abs(x)) ); } } diff --git a/third_party/eigen/test/packetmath.cpp b/third_party/eigen/test/packetmath.cpp index 7821a1738df..74ac435cf47 100644 --- a/third_party/eigen/test/packetmath.cpp +++ b/third_party/eigen/test/packetmath.cpp @@ -16,12 +16,6 @@ #endif // using namespace Eigen; -#ifdef EIGEN_VECTORIZE_SSE -const bool g_vectorize_sse = true; -#else -const bool g_vectorize_sse = false; -#endif - namespace Eigen { namespace internal { template T negate(const T& x) { return -x; } @@ -248,12 +242,13 @@ template void packetmath() VERIFY(isApproxAbs(ref[0], internal::predux(internal::pload(data1)), refvalue) && "internal::predux"); { - for (int i=0; i<4; ++i) + int newsize = PacketSize>4?PacketSize/2:PacketSize; + for (int i=0; i(data1))); - VERIFY(areApprox(ref, data2, PacketSize>4?PacketSize/2:PacketSize) && "internal::predux_downto4"); + VERIFY(areApprox(ref, data2, newsize) && "internal::predux_downto4"); } ref[0] = 1; @@ -304,7 +299,7 @@ template void packetmath() } } - if (PacketTraits::HasBlend || g_vectorize_sse) { + if (PacketTraits::HasBlend) { // pinsertfirst for (int i=0; i void packetmath() VERIFY(areApprox(ref, data2, PacketSize) && "internal::pinsertfirst"); } - if (PacketTraits::HasBlend || g_vectorize_sse) { + if (PacketTraits::HasBlend) { // pinsertlast for (int i=0; i void product(const MatrixType& m) vcres.noalias() -= m1.transpose() * v1; VERIFY_IS_APPROX(vcres, vc2 - m1.transpose() * v1); + // test scaled products + res = square; + res.noalias() = s1 * m1 * m2.transpose(); + VERIFY_IS_APPROX(res, ((s1*m1).eval() * m2.transpose())); + res = square; + res.noalias() += s1 * m1 * m2.transpose(); + VERIFY_IS_APPROX(res, square + ((s1*m1).eval() * m2.transpose())); + res = square; + res.noalias() -= s1 * m1 * m2.transpose(); + VERIFY_IS_APPROX(res, square - ((s1*m1).eval() * m2.transpose())); + // test d ?= a+b*c rules res.noalias() = square + m1 * m2.transpose(); VERIFY_IS_APPROX(res, square + m1 * m2.transpose()); @@ -228,4 +239,19 @@ template void product(const MatrixType& m) VERIFY_IS_APPROX(square * (square*square).conjugate(), square * square.conjugate() * square.conjugate()); } + // destination with a non-default inner-stride + // see bug 1741 + if(!MatrixType::IsRowMajor) + { + typedef Matrix MatrixX; + MatrixX buffer(2*rows,2*rows); + Map > map1(buffer.data(),rows,rows,Stride(2*rows,2)); + buffer.setZero(); + VERIFY_IS_APPROX(map1 = m1 * m2.transpose(), (m1 * m2.transpose()).eval()); + buffer.setZero(); + VERIFY_IS_APPROX(map1.noalias() = m1 * m2.transpose(), (m1 * m2.transpose()).eval()); + buffer.setZero(); + VERIFY_IS_APPROX(map1.noalias() += m1 * m2.transpose(), (m1 * m2.transpose()).eval()); + } + } diff --git a/third_party/eigen/test/product_large.cpp b/third_party/eigen/test/product_large.cpp index 845cd40ca10..14a4f739d48 100644 --- a/third_party/eigen/test/product_large.cpp +++ b/third_party/eigen/test/product_large.cpp @@ -35,6 +35,8 @@ void test_product_large() for(int i = 0; i < g_repeat; i++) { CALL_SUBTEST_1( product(MatrixXf(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); CALL_SUBTEST_2( product(MatrixXd(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); + CALL_SUBTEST_2( product(MatrixXd(internal::random(1,10), internal::random(1,10))) ); + CALL_SUBTEST_3( product(MatrixXi(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); CALL_SUBTEST_4( product(MatrixXcf(internal::random(1,EIGEN_TEST_MAX_SIZE/2), internal::random(1,EIGEN_TEST_MAX_SIZE/2))) ); CALL_SUBTEST_5( product(Matrix(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); diff --git a/third_party/eigen/test/product_mmtr.cpp b/third_party/eigen/test/product_mmtr.cpp index d3e24b012dd..35686460cdc 100644 --- a/third_party/eigen/test/product_mmtr.cpp +++ b/third_party/eigen/test/product_mmtr.cpp @@ -82,6 +82,16 @@ template void mmtr(int size) ref2.template triangularView() = ref1.template triangularView(); matc.template triangularView() = sqc * matc * sqc.adjoint(); VERIFY_IS_APPROX(matc, ref2); + + // destination with a non-default inner-stride + // see bug 1741 + { + typedef Matrix MatrixX; + MatrixX buffer(2*size,2*size); + Map > map1(buffer.data(),size,size,Stride(2*size,2)); + buffer.setZero(); + CHECK_MMTR(map1, Lower, = s*soc*sor.adjoint()); + } } void test_product_mmtr() diff --git a/third_party/eigen/test/product_symm.cpp b/third_party/eigen/test/product_symm.cpp index 7d1042a4f25..0ed027dff1a 100644 --- a/third_party/eigen/test/product_symm.cpp +++ b/third_party/eigen/test/product_symm.cpp @@ -75,12 +75,12 @@ template void symm(int size = Size, in rhs13 = (s1*m1.adjoint()) * (s2*rhs2.adjoint())); // test row major = <...> - m2 = m1.template triangularView(); rhs12.setRandom(); rhs13 = rhs12; - VERIFY_IS_APPROX(rhs12 -= (s1*m2).template selfadjointView() * (s2*rhs3), + m2 = m1.template triangularView(); rhs32.setRandom(); rhs13 = rhs32; + VERIFY_IS_APPROX(rhs32.noalias() -= (s1*m2).template selfadjointView() * (s2*rhs3), rhs13 -= (s1*m1) * (s2 * rhs3)); m2 = m1.template triangularView(); - VERIFY_IS_APPROX(rhs12 = (s1*m2.adjoint()).template selfadjointView() * (s2*rhs3).conjugate(), + VERIFY_IS_APPROX(rhs32.noalias() = (s1*m2.adjoint()).template selfadjointView() * (s2*rhs3).conjugate(), rhs13 = (s1*m1.adjoint()) * (s2*rhs3).conjugate()); @@ -92,6 +92,20 @@ template void symm(int size = Size, in VERIFY_IS_APPROX(rhs22 = (rhs2) * (m2).template selfadjointView(), rhs23 = (rhs2) * (m1)); VERIFY_IS_APPROX(rhs22 = (s2*rhs2) * (s1*m2).template selfadjointView(), rhs23 = (s2*rhs2) * (s1*m1)); + // destination with a non-default inner-stride + // see bug 1741 + { + typedef Matrix MatrixX; + MatrixX buffer(2*cols,2*othersize); + Map > map1(buffer.data(),cols,othersize,Stride(2*rows,2)); + buffer.setZero(); + VERIFY_IS_APPROX( map1.noalias() = (s1*m2).template selfadjointView() * (s2*rhs1), + rhs13 = (s1*m1) * (s2*rhs1)); + + Map > map2(buffer.data(),rhs22.rows(),rhs22.cols(),Stride(2*rhs22.outerStride(),2)); + buffer.setZero(); + VERIFY_IS_APPROX(map2 = (rhs2) * (m2).template selfadjointView(), rhs23 = (rhs2) * (m1)); + } } void test_product_symm() diff --git a/third_party/eigen/test/product_syrk.cpp b/third_party/eigen/test/product_syrk.cpp index 3ebbe14cad3..b8578215f99 100644 --- a/third_party/eigen/test/product_syrk.cpp +++ b/third_party/eigen/test/product_syrk.cpp @@ -115,6 +115,17 @@ template void syrk(const MatrixType& m) m2.setZero(); VERIFY_IS_APPROX((m2.template selfadjointView().rankUpdate(m1.row(c).adjoint(),s1)._expression()), ((s1 * m1.row(c).adjoint() * m1.row(c).adjoint().adjoint()).eval().template triangularView().toDenseMatrix())); + + // destination with a non-default inner-stride + // see bug 1741 + { + typedef Matrix MatrixX; + MatrixX buffer(2*rows,2*cols); + Map > map1(buffer.data(),rows,cols,Stride(2*rows,2)); + buffer.setZero(); + VERIFY_IS_APPROX((map1.template selfadjointView().rankUpdate(rhs2,s1)._expression()), + ((s1 * rhs2 * rhs2.adjoint()).eval().template triangularView().toDenseMatrix())); + } } void test_product_syrk() diff --git a/third_party/eigen/test/product_trmm.cpp b/third_party/eigen/test/product_trmm.cpp index e08d9f39f5c..ddcde9622ac 100644 --- a/third_party/eigen/test/product_trmm.cpp +++ b/third_party/eigen/test/product_trmm.cpp @@ -76,8 +76,18 @@ void trmm(int rows=get_random_size(), VERIFY_IS_APPROX( ge_xs = (s1*mat).adjoint().template triangularView() * ge_left.adjoint(), numext::conj(s1) * triTr.conjugate() * ge_left.adjoint()); VERIFY_IS_APPROX( ge_xs = (s1*mat).transpose().template triangularView() * ge_left.adjoint(), s1triTr * ge_left.adjoint()); - // TODO check with sub-matrix expressions ? + + // destination with a non-default inner-stride + // see bug 1741 + { + VERIFY_IS_APPROX( ge_xs.noalias() = mat.template triangularView() * ge_right, tri * ge_right); + typedef Matrix MatrixX; + MatrixX buffer(2*ge_xs.rows(),2*ge_xs.cols()); + Map > map1(buffer.data(),ge_xs.rows(),ge_xs.cols(),Stride(2*ge_xs.outerStride(),2)); + buffer.setZero(); + VERIFY_IS_APPROX( map1.noalias() = mat.template triangularView() * ge_right, tri * ge_right); + } } template diff --git a/third_party/eigen/test/product_trsolve.cpp b/third_party/eigen/test/product_trsolve.cpp index 4b97fa9d648..eaf62cb1139 100644 --- a/third_party/eigen/test/product_trsolve.cpp +++ b/third_party/eigen/test/product_trsolve.cpp @@ -71,6 +71,32 @@ template void trsolve(int size=Size,int cols int c = internal::random(0,cols-1); VERIFY_TRSM(rmLhs.template triangularView(), rmRhs.col(c)); VERIFY_TRSM(cmLhs.template triangularView(), rmRhs.col(c)); + + // destination with a non-default inner-stride + // see bug 1741 + { + typedef Matrix MatrixX; + MatrixX buffer(2*cmRhs.rows(),2*cmRhs.cols()); + Map,0,Stride > map1(buffer.data(),cmRhs.rows(),cmRhs.cols(),Stride(2*cmRhs.outerStride(),2)); + Map,0,Stride > map2(buffer.data(),rmRhs.rows(),rmRhs.cols(),Stride(2*rmRhs.outerStride(),2)); + buffer.setZero(); + VERIFY_TRSM(cmLhs.conjugate().template triangularView(), map1); + buffer.setZero(); + VERIFY_TRSM(cmLhs .template triangularView(), map2); + } + + if(Size==Dynamic) + { + cmLhs.resize(0,0); + cmRhs.resize(0,cmRhs.cols()); + Matrix res = cmLhs.template triangularView().solve(cmRhs); + VERIFY_IS_EQUAL(res.rows(),0); + VERIFY_IS_EQUAL(res.cols(),cmRhs.cols()); + res = cmRhs; + cmLhs.template triangularView().solveInPlace(res); + VERIFY_IS_EQUAL(res.rows(),0); + VERIFY_IS_EQUAL(res.cols(),cmRhs.cols()); + } } void test_product_trsolve() diff --git a/third_party/eigen/test/ref.cpp b/third_party/eigen/test/ref.cpp index 704495affa8..da399e287f1 100644 --- a/third_party/eigen/test/ref.cpp +++ b/third_party/eigen/test/ref.cpp @@ -102,10 +102,14 @@ template void ref_vector(const VectorType& m) Index i = internal::random(0,size-1); Index bsize = internal::random(1,size-i); - RefMat rm0 = v1; - VERIFY_IS_EQUAL(rm0, v1); - RefDynMat rv1 = v1; - VERIFY_IS_EQUAL(rv1, v1); + { RefMat rm0 = v1; VERIFY_IS_EQUAL(rm0, v1); } + { RefMat rm0 = v1.block(0,0,size,1); VERIFY_IS_EQUAL(rm0, v1); } + { RefDynMat rv1 = v1; VERIFY_IS_EQUAL(rv1, v1); } + { RefDynMat rv1 = v1.block(0,0,size,1); VERIFY_IS_EQUAL(rv1, v1); } + { VERIFY_RAISES_ASSERT( RefMat rm0 = v1.block(0, 0, size, 0); EIGEN_UNUSED_VARIABLE(rm0); ); } + if(VectorType::SizeAtCompileTime!=1) + { VERIFY_RAISES_ASSERT( RefDynMat rv1 = v1.block(0, 0, size, 0); EIGEN_UNUSED_VARIABLE(rv1); ); } + RefDynMat rv2 = v1.segment(i,bsize); VERIFY_IS_EQUAL(rv2, v1.segment(i,bsize)); rv2.setOnes(); diff --git a/third_party/eigen/test/rvalue_types.cpp b/third_party/eigen/test/rvalue_types.cpp index 8887f1b1b99..6a97dae340c 100644 --- a/third_party/eigen/test/rvalue_types.cpp +++ b/third_party/eigen/test/rvalue_types.cpp @@ -7,6 +7,8 @@ // Public License v. 2.0. If a copy of the MPL was not distributed // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. +#define EIGEN_RUNTIME_NO_MALLOC + #include "main.h" #include @@ -24,41 +26,85 @@ void rvalue_copyassign(const MatrixType& m) MatrixType tmp = m; UIntPtr src_address = reinterpret_cast(tmp.data()); + Eigen::internal::set_is_malloc_allowed(false); // moving from an rvalue reference shall never allocate // move the temporary to n MatrixType n = std::move(tmp); UIntPtr dst_address = reinterpret_cast(n.data()); - if (MatrixType::RowsAtCompileTime==Dynamic|| MatrixType::ColsAtCompileTime==Dynamic) { // verify that we actually moved the guts VERIFY_IS_EQUAL(src_address, dst_address); + VERIFY_IS_EQUAL(tmp.size(), 0); + VERIFY_IS_EQUAL(reinterpret_cast(tmp.data()), UIntPtr(0)); } // verify that the content did not change Scalar abs_diff = (m-n).array().abs().sum(); VERIFY_IS_EQUAL(abs_diff, Scalar(0)); + Eigen::internal::set_is_malloc_allowed(true); +} +template +void rvalue_transpositions(Index rows) +{ + typedef typename TranspositionsType::IndicesType PermutationVectorType; + + PermutationVectorType vec; + randomPermutationVector(vec, rows); + TranspositionsType t0(vec); + + Eigen::internal::set_is_malloc_allowed(false); // moving from an rvalue reference shall never allocate + + UIntPtr t0_address = reinterpret_cast(t0.indices().data()); + + // Move constructors: + TranspositionsType t1 = std::move(t0); + UIntPtr t1_address = reinterpret_cast(t1.indices().data()); + VERIFY_IS_EQUAL(t0_address, t1_address); + // t0 must be de-allocated: + VERIFY_IS_EQUAL(t0.size(), 0); + VERIFY_IS_EQUAL(reinterpret_cast(t0.indices().data()), UIntPtr(0)); + + + // Move assignment: + t0 = std::move(t1); + t0_address = reinterpret_cast(t0.indices().data()); + VERIFY_IS_EQUAL(t0_address, t1_address); + // t1 must be de-allocated: + VERIFY_IS_EQUAL(t1.size(), 0); + VERIFY_IS_EQUAL(reinterpret_cast(t1.indices().data()), UIntPtr(0)); + + Eigen::internal::set_is_malloc_allowed(true); } #else template void rvalue_copyassign(const MatrixType&) {} +template +void rvalue_transpositions(Index) {} #endif void test_rvalue_types() { - CALL_SUBTEST_1(rvalue_copyassign( MatrixXf::Random(50,50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( ArrayXXf::Random(50,50).eval() )); + for(int i = 0; i < g_repeat; i++) { + CALL_SUBTEST_1(rvalue_copyassign( MatrixXf::Random(50,50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( ArrayXXf::Random(50,50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); - - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + + CALL_SUBTEST_3((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); + CALL_SUBTEST_3((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); + CALL_SUBTEST_4((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); + CALL_SUBTEST_4((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); + } } diff --git a/third_party/eigen/test/sparse_basic.cpp b/third_party/eigen/test/sparse_basic.cpp index d0ef722b61c..43318da796c 100644 --- a/third_party/eigen/test/sparse_basic.cpp +++ b/third_party/eigen/test/sparse_basic.cpp @@ -612,6 +612,14 @@ template void sparse_basic(const SparseMatrixType& re iters[0] = IteratorType(m2,0); iters[1] = IteratorType(m2,m2.outerSize()-1); } + + // test reserve with empty rows/columns + { + SparseMatrixType m1(0,cols); + m1.reserve(ArrayXi::Constant(m1.outerSize(),1)); + SparseMatrixType m2(rows,0); + m2.reserve(ArrayXi::Constant(m2.outerSize(),1)); + } } diff --git a/third_party/eigen/test/stddeque.cpp b/third_party/eigen/test/stddeque.cpp index b511c4e6121..b6955f74758 100644 --- a/third_party/eigen/test/stddeque.cpp +++ b/third_party/eigen/test/stddeque.cpp @@ -18,7 +18,7 @@ void check_stddeque_matrix(const MatrixType& m) Index rows = m.rows(); Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::deque > v(10, MatrixType(rows,cols)), w(20, y); + std::deque > v(10, MatrixType::Zero(rows,cols)), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -33,7 +33,7 @@ void check_stddeque_matrix(const MatrixType& m) ++wi; } - v.resize(21); + v.resize(21,MatrixType::Zero(rows,cols)); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -46,8 +46,8 @@ template void check_stddeque_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()); - std::deque > v(10), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); + std::deque > v(10,ti), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -62,7 +62,7 @@ void check_stddeque_transform(const TransformType&) ++wi; } - v.resize(21); + v.resize(21,ti); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -75,8 +75,8 @@ template void check_stddeque_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::deque > v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::deque > v(10,qi), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -91,7 +91,7 @@ void check_stddeque_quaternion(const QuaternionType&) ++wi; } - v.resize(21); + v.resize(21,qi); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); diff --git a/third_party/eigen/test/stddeque_overload.cpp b/third_party/eigen/test/stddeque_overload.cpp index 4da618bbfc2..f495b5a04ab 100644 --- a/third_party/eigen/test/stddeque_overload.cpp +++ b/third_party/eigen/test/stddeque_overload.cpp @@ -31,7 +31,7 @@ void check_stddeque_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::deque v(10, MatrixType(rows,cols)), w(20, y); + std::deque v(10, MatrixType::Zero(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -64,8 +64,8 @@ template void check_stddeque_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()); - std::deque v(10), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); + std::deque v(10,ti), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -75,7 +75,7 @@ void check_stddeque_transform(const TransformType&) VERIFY_IS_APPROX(w[i], v[i]); } - v.resize(21); + v.resize(21,ti); v[20] = x; VERIFY_IS_APPROX(v[20], x); v.resize(22,y); @@ -98,8 +98,8 @@ template void check_stddeque_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::deque v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::deque v(10,qi), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -109,7 +109,7 @@ void check_stddeque_quaternion(const QuaternionType&) VERIFY_IS_APPROX(w[i], v[i]); } - v.resize(21); + v.resize(21,qi); v[20] = x; VERIFY_IS_APPROX(v[20], x); v.resize(22,y); diff --git a/third_party/eigen/test/stdlist.cpp b/third_party/eigen/test/stdlist.cpp index 23cbe903970..23b30ccaf1f 100644 --- a/third_party/eigen/test/stdlist.cpp +++ b/third_party/eigen/test/stdlist.cpp @@ -18,7 +18,7 @@ void check_stdlist_matrix(const MatrixType& m) Index rows = m.rows(); Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::list > v(10, MatrixType(rows,cols)), w(20, y); + std::list > v(10, MatrixType::Zero(rows,cols)), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -33,7 +33,7 @@ void check_stdlist_matrix(const MatrixType& m) ++wi; } - v.resize(21); + v.resize(21, MatrixType::Zero(rows,cols)); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -46,8 +46,8 @@ template void check_stdlist_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()); - std::list > v(10), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); + std::list > v(10,ti), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -62,7 +62,7 @@ void check_stdlist_transform(const TransformType&) ++wi; } - v.resize(21); + v.resize(21, ti); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -75,8 +75,8 @@ template void check_stdlist_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::list > v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::list > v(10,qi), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -91,7 +91,7 @@ void check_stdlist_quaternion(const QuaternionType&) ++wi; } - v.resize(21); + v.resize(21,qi); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); diff --git a/third_party/eigen/test/stdlist_overload.cpp b/third_party/eigen/test/stdlist_overload.cpp index bb910bd439f..aea7a28465c 100644 --- a/third_party/eigen/test/stdlist_overload.cpp +++ b/third_party/eigen/test/stdlist_overload.cpp @@ -47,7 +47,7 @@ void check_stdlist_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::list v(10, MatrixType(rows,cols)), w(20, y); + std::list v(10, MatrixType::Zero(rows,cols)), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -86,8 +86,8 @@ template void check_stdlist_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()); - std::list v(10), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); + std::list v(10,ti), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -103,7 +103,7 @@ void check_stdlist_transform(const TransformType&) ++itw; } - v.resize(21); + v.resize(21, ti); set(v, 20, x); VERIFY_IS_APPROX(*get(v, 20), x); v.resize(22,y); @@ -126,8 +126,8 @@ template void check_stdlist_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::list v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::list v(10,qi), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -143,7 +143,7 @@ void check_stdlist_quaternion(const QuaternionType&) ++itw; } - v.resize(21); + v.resize(21,qi); set(v, 20, x); VERIFY_IS_APPROX(*get(v, 20), x); v.resize(22,y); diff --git a/third_party/eigen/test/stdvector.cpp b/third_party/eigen/test/stdvector.cpp index fa928ea4fe9..383d9a5095e 100644 --- a/third_party/eigen/test/stdvector.cpp +++ b/third_party/eigen/test/stdvector.cpp @@ -17,7 +17,7 @@ void check_stdvector_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::vector > v(10, MatrixType(rows,cols)), w(20, y); + std::vector > v(10, MatrixType::Zero(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -86,8 +86,8 @@ template void check_stdvector_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::vector > v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::vector > v(10,qi), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); diff --git a/third_party/eigen/test/stdvector_overload.cpp b/third_party/eigen/test/stdvector_overload.cpp index 959665954e4..637e3ef52f3 100644 --- a/third_party/eigen/test/stdvector_overload.cpp +++ b/third_party/eigen/test/stdvector_overload.cpp @@ -31,7 +31,7 @@ void check_stdvector_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::vector v(10, MatrixType(rows,cols)), w(20, y); + std::vector v(10, MatrixType::Zero(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -100,8 +100,8 @@ template void check_stdvector_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); - std::vector v(10), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); + std::vector v(10,qi), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); diff --git a/third_party/eigen/test/vectorization_logic.cpp b/third_party/eigen/test/vectorization_logic.cpp index 37e7495f5b5..c2f77bfec21 100644 --- a/third_party/eigen/test/vectorization_logic.cpp +++ b/third_party/eigen/test/vectorization_logic.cpp @@ -22,6 +22,14 @@ #include "main.h" #include +// Disable "ignoring attributes on template argument" +// for packet_traits +// => The only workaround would be to wrap _m128 and the likes +// within wrappers. +#if EIGEN_GNUC_AT_LEAST(6,0) + #pragma GCC diagnostic ignored "-Wignored-attributes" +#endif + using internal::demangle_flags; using internal::demangle_traversal; using internal::demangle_unrolling; diff --git a/third_party/eigen/unsupported/Eigen/ArpackSupport b/third_party/eigen/unsupported/Eigen/ArpackSupport index 37a2799ef2f..a0d4820e125 100644 --- a/third_party/eigen/unsupported/Eigen/ArpackSupport +++ b/third_party/eigen/unsupported/Eigen/ArpackSupport @@ -11,8 +11,6 @@ #include -#include - /** \defgroup ArpackSupport_Module Arpack support module * * This module provides a wrapper to Arpack, a library for sparse eigenvalue decomposition. @@ -23,6 +21,8 @@ */ #include + +#include #include "src/Eigenvalues/ArpackSelfAdjointEigenSolver.h" #include diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h index 9b2cb3ff6bc..c28a10dd494 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h @@ -113,6 +113,7 @@ class SimpleTensorContractionMapper { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index computeIndex(Index row, Index col) const { const bool left = (side == Lhs); + EIGEN_UNUSED_VARIABLE(left); // annoying bug in g++8.1: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85963 Index nocontract_val = left ? row : col; Index linidx = 0; for (int i = static_cast(array_size::value) - 1; i > 0; i--) { @@ -151,6 +152,7 @@ class SimpleTensorContractionMapper { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE IndexPair computeIndexPair(Index row, Index col, const Index distance) const { const bool left = (side == Lhs); + EIGEN_UNUSED_VARIABLE(left); // annoying bug in g++8.1: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85963 Index nocontract_val[2] = {left ? row : col, left ? row + distance : col}; Index linidx[2] = {0, 0}; if (array_size::value > array_size::value) { diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h index 17f04665a12..a5e084a2409 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h @@ -31,7 +31,7 @@ class Barrier { eigen_assert(((count << 1) >> 1) == count); } ~Barrier() { - eigen_assert((state_>>1) == 0); + eigen_plain_assert((state_>>1) == 0); } void Notify() { diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h index 71d55552de6..4749d6240af 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h @@ -58,7 +58,7 @@ class EventCount { ~EventCount() { // Ensure there are no waiters. - eigen_assert((state_.load() & (kStackMask | kWaiterMask)) == kStackMask); + eigen_plain_assert((state_.load() & (kStackMask | kWaiterMask)) == kStackMask); } // Prewait prepares for waiting. diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h index 05ed76cbe49..6e505fc14c8 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h @@ -47,7 +47,7 @@ class RunQueue { array_[i].state.store(kEmpty, std::memory_order_relaxed); } - ~RunQueue() { eigen_assert(Size() == 0); } + ~RunQueue() { eigen_plain_assert(Size() == 0); } // PushFront inserts w at the beginning of the queue. // If queue is full returns w, otherwise returns default-constructed Work. diff --git a/third_party/eigen/unsupported/Eigen/Polynomials b/third_party/eigen/unsupported/Eigen/Polynomials index cece563374e..334b031427a 100644 --- a/third_party/eigen/unsupported/Eigen/Polynomials +++ b/third_party/eigen/unsupported/Eigen/Polynomials @@ -11,10 +11,10 @@ #include -#include - #include +#include + // Note that EIGEN_HIDE_HEAVY_CODE has to be defined per module #if (defined EIGEN_EXTERN_INSTANTIATIONS) && (EIGEN_EXTERN_INSTANTIATIONS>=2) #ifndef EIGEN_HIDE_HEAVY_CODE diff --git a/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h b/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h index 2f50e996805..58f3f3319d4 100755 --- a/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h +++ b/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h @@ -453,6 +453,24 @@ struct auto_diff_special_op<_DerType, false> void operator+() const; }; +template +void make_coherent_expression(CwiseBinaryOp xpr, const RefType &ref) +{ + make_coherent(xpr.const_cast_derived().lhs(), ref); + make_coherent(xpr.const_cast_derived().rhs(), ref); +} + +template +void make_coherent_expression(const CwiseUnaryOp &xpr, const RefType &ref) +{ + make_coherent(xpr.nestedExpression().const_cast_derived(), ref); +} + +// needed for compilation only +template +void make_coherent_expression(const CwiseNullaryOp &, const RefType &) +{} + template struct make_coherent_impl, B> { typedef Matrix A; @@ -462,6 +480,10 @@ struct make_coherent_impl struct make_coherent_impl, - Matrix > { + Matrix > { typedef Matrix A; typedef Matrix B; static void run(A& a, B& b) { diff --git a/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h b/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h index 866a8a46018..4170d26b6f6 100644 --- a/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h +++ b/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h @@ -3,24 +3,9 @@ // // Copyright (C) 2012 David Harmon // -// Eigen is free software; you can redistribute it and/or -// modify it under the terms of the GNU Lesser General Public -// License as published by the Free Software Foundation; either -// version 3 of the License, or (at your option) any later version. -// -// Alternatively, you can redistribute it and/or -// modify it under the terms of the GNU General Public License as -// published by the Free Software Foundation; either version 2 of -// the License, or (at your option) any later version. -// -// Eigen is distributed in the hope that it will be useful, but WITHOUT ANY -// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS -// FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License or the -// GNU General Public License for more details. -// -// You should have received a copy of the GNU Lesser General Public -// License and a copy of the GNU General Public License along with -// Eigen. If not, see . +// This Source Code Form is subject to the terms of the Mozilla +// Public License v. 2.0. If a copy of the MPL was not distributed +// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. #ifndef EIGEN_ARPACKGENERALIZEDSELFADJOINTEIGENSOLVER_H #define EIGEN_ARPACKGENERALIZEDSELFADJOINTEIGENSOLVER_H diff --git a/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h b/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h index d49aa17f510..7c1f716e2a8 100644 --- a/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h +++ b/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h @@ -231,6 +231,8 @@ namespace internal { protected: typedef fftw_plan PlanData; + typedef Eigen::numext::int64_t int64_t; + typedef std::map PlanMap; PlanMap m_plans; diff --git a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h index e5ebbcf2378..0b0ee6546a8 100644 --- a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h +++ b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h @@ -412,7 +412,7 @@ template struct MatrixExponentialReturnValue inline void evalTo(ResultType& result) const { const typename internal::nested_eval::type tmp(m_src); - internal::matrix_exp_compute(tmp, result, internal::is_exp_known_type()); + internal::matrix_exp_compute(tmp, result, internal::is_exp_known_type()); } Index rows() const { return m_src.rows(); } diff --git a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h index 2e5abda381b..9de0c3574eb 100644 --- a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h +++ b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h @@ -253,18 +253,19 @@ struct matrix_sqrt_compute template struct matrix_sqrt_compute { + typedef typename MatrixType::PlainObject PlainType; template static void run(const MatrixType &arg, ResultType &result) { eigen_assert(arg.rows() == arg.cols()); // Compute Schur decomposition of arg - const RealSchur schurOfA(arg); - const MatrixType& T = schurOfA.matrixT(); - const MatrixType& U = schurOfA.matrixU(); + const RealSchur schurOfA(arg); + const PlainType& T = schurOfA.matrixT(); + const PlainType& U = schurOfA.matrixU(); // Compute square root of T - MatrixType sqrtT = MatrixType::Zero(arg.rows(), arg.cols()); + PlainType sqrtT = PlainType::Zero(arg.rows(), arg.cols()); matrix_sqrt_quasi_triangular(T, sqrtT); // Compute square root of arg @@ -278,18 +279,19 @@ struct matrix_sqrt_compute template struct matrix_sqrt_compute { + typedef typename MatrixType::PlainObject PlainType; template static void run(const MatrixType &arg, ResultType &result) { eigen_assert(arg.rows() == arg.cols()); // Compute Schur decomposition of arg - const ComplexSchur schurOfA(arg); - const MatrixType& T = schurOfA.matrixT(); - const MatrixType& U = schurOfA.matrixU(); + const ComplexSchur schurOfA(arg); + const PlainType& T = schurOfA.matrixT(); + const PlainType& U = schurOfA.matrixU(); // Compute square root of T - MatrixType sqrtT; + PlainType sqrtT; matrix_sqrt_triangular(T, sqrtT); // Compute square root of arg diff --git a/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h b/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h index b515c29208e..359836cac80 100644 --- a/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h +++ b/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h @@ -75,8 +75,7 @@ class companion void setPolynomial( const VectorType& poly ) { const Index deg = poly.size()-1; - m_monic = -1/poly[deg] * poly.head(deg); - //m_bl_diag.setIdentity( deg-1 ); + m_monic = -poly.head(deg)/poly[deg]; m_bl_diag.setOnes(deg-1); } @@ -89,13 +88,13 @@ class companion { const Index deg = m_monic.size(); const Index deg_1 = deg-1; - DenseCompanionMatrixType companion(deg,deg); - companion << + DenseCompanionMatrixType companMat(deg,deg); + companMat << ( LeftBlock(deg,deg_1) << LeftBlockFirstRow::Zero(1,deg_1), BottomLeftBlock::Identity(deg-1,deg-1)*m_bl_diag.asDiagonal() ).finished() , m_monic; - return companion; + return companMat; } @@ -107,8 +106,8 @@ class companion * colB and rowB are repectively the multipliers for * the column and the row in order to balance them. * */ - bool balanced( Scalar colNorm, Scalar rowNorm, - bool& isBalanced, Scalar& colB, Scalar& rowB ); + bool balanced( RealScalar colNorm, RealScalar rowNorm, + bool& isBalanced, RealScalar& colB, RealScalar& rowB ); /** Helper function for the balancing algorithm. * \returns true if the row and the column, having colNorm and rowNorm @@ -116,8 +115,8 @@ class companion * colB and rowB are repectively the multipliers for * the column and the row in order to balance them. * */ - bool balancedR( Scalar colNorm, Scalar rowNorm, - bool& isBalanced, Scalar& colB, Scalar& rowB ); + bool balancedR( RealScalar colNorm, RealScalar rowNorm, + bool& isBalanced, RealScalar& colB, RealScalar& rowB ); public: /** @@ -139,10 +138,10 @@ class companion template< typename _Scalar, int _Deg > inline -bool companion<_Scalar,_Deg>::balanced( Scalar colNorm, Scalar rowNorm, - bool& isBalanced, Scalar& colB, Scalar& rowB ) +bool companion<_Scalar,_Deg>::balanced( RealScalar colNorm, RealScalar rowNorm, + bool& isBalanced, RealScalar& colB, RealScalar& rowB ) { - if( Scalar(0) == colNorm || Scalar(0) == rowNorm ){ return true; } + if( RealScalar(0) == colNorm || RealScalar(0) == rowNorm ){ return true; } else { //To find the balancing coefficients, if the radix is 2, @@ -150,29 +149,29 @@ bool companion<_Scalar,_Deg>::balanced( Scalar colNorm, Scalar rowNorm, // \f$ 2^{2\sigma-1} < rowNorm / colNorm \le 2^{2\sigma+1} \f$ // then the balancing coefficient for the row is \f$ 1/2^{\sigma} \f$ // and the balancing coefficient for the column is \f$ 2^{\sigma} \f$ - rowB = rowNorm / radix(); - colB = Scalar(1); - const Scalar s = colNorm + rowNorm; + rowB = rowNorm / radix(); + colB = RealScalar(1); + const RealScalar s = colNorm + rowNorm; while (colNorm < rowB) { - colB *= radix(); - colNorm *= radix2(); + colB *= radix(); + colNorm *= radix2(); } - rowB = rowNorm * radix(); + rowB = rowNorm * radix(); while (colNorm >= rowB) { - colB /= radix(); - colNorm /= radix2(); + colB /= radix(); + colNorm /= radix2(); } //This line is used to avoid insubstantial balancing - if ((rowNorm + colNorm) < Scalar(0.95) * s * colB) + if ((rowNorm + colNorm) < RealScalar(0.95) * s * colB) { isBalanced = false; - rowB = Scalar(1) / colB; + rowB = RealScalar(1) / colB; return false; } else{ @@ -182,21 +181,21 @@ bool companion<_Scalar,_Deg>::balanced( Scalar colNorm, Scalar rowNorm, template< typename _Scalar, int _Deg > inline -bool companion<_Scalar,_Deg>::balancedR( Scalar colNorm, Scalar rowNorm, - bool& isBalanced, Scalar& colB, Scalar& rowB ) +bool companion<_Scalar,_Deg>::balancedR( RealScalar colNorm, RealScalar rowNorm, + bool& isBalanced, RealScalar& colB, RealScalar& rowB ) { - if( Scalar(0) == colNorm || Scalar(0) == rowNorm ){ return true; } + if( RealScalar(0) == colNorm || RealScalar(0) == rowNorm ){ return true; } else { /** * Set the norm of the column and the row to the geometric mean * of the row and column norm */ - const _Scalar q = colNorm/rowNorm; + const RealScalar q = colNorm/rowNorm; if( !isApprox( q, _Scalar(1) ) ) { rowB = sqrt( colNorm/rowNorm ); - colB = Scalar(1)/rowB; + colB = RealScalar(1)/rowB; isBalanced = false; return false; @@ -219,8 +218,8 @@ void companion<_Scalar,_Deg>::balance() while( !hasConverged ) { hasConverged = true; - Scalar colNorm,rowNorm; - Scalar colB,rowB; + RealScalar colNorm,rowNorm; + RealScalar colB,rowB; //First row, first column excluding the diagonal //============================================== diff --git a/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h b/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h index 03198ec8ee2..5e0ecbb43b3 100644 --- a/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h +++ b/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h @@ -99,7 +99,7 @@ class PolynomialSolverBase */ inline const RootType& greatestRoot() const { - std::greater greater; + std::greater greater; return selectComplexRoot_withRespectToNorm( greater ); } @@ -108,7 +108,7 @@ class PolynomialSolverBase */ inline const RootType& smallestRoot() const { - std::less less; + std::less less; return selectComplexRoot_withRespectToNorm( less ); } @@ -126,7 +126,7 @@ class PolynomialSolverBase for( Index i=0; i::dummy_precision() ) const { - std::greater greater; + std::greater greater; return selectRealRoot_withRespectToAbsRealPart( greater, hasArealRoot, absImaginaryThreshold ); } @@ -236,7 +236,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::less less; + std::less less; return selectRealRoot_withRespectToAbsRealPart( less, hasArealRoot, absImaginaryThreshold ); } @@ -259,7 +259,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::greater greater; + std::greater greater; return selectRealRoot_withRespectToRealPart( greater, hasArealRoot, absImaginaryThreshold ); } @@ -282,7 +282,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::less less; + std::less less; return selectRealRoot_withRespectToRealPart( less, hasArealRoot, absImaginaryThreshold ); } @@ -327,7 +327,7 @@ class PolynomialSolverBase * However, almost always, correct accuracy is reached even in these cases for 64bit * (double) floating types and small polynomial degree (<20). */ -template< typename _Scalar, int _Deg > +template class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> { public: @@ -337,7 +337,10 @@ class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> EIGEN_POLYNOMIAL_SOLVER_BASE_INHERITED_TYPES( PS_Base ) typedef Matrix CompanionMatrixType; - typedef EigenSolver EigenSolverType; + typedef typename internal::conditional::IsComplex, + ComplexEigenSolver, + EigenSolver >::type EigenSolverType; + typedef typename internal::conditional::IsComplex, Scalar, std::complex >::type ComplexScalar; public: /** Computes the complex roots of a new polynomial. */ @@ -352,6 +355,25 @@ class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> companion.balance(); m_eigenSolver.compute( companion.denseMatrix() ); m_roots = m_eigenSolver.eigenvalues(); + // cleanup noise in imaginary part of real roots: + // if the imaginary part is rather small compared to the real part + // and that cancelling the imaginary part yield a smaller evaluation, + // then it's safe to keep the real part only. + RealScalar coarse_prec = RealScalar(std::pow(4,poly.size()+1))*NumTraits::epsilon(); + for(Index i = 0; i::Scalar u, DenseIndex degree, const typename SplineTraits::KnotVectorType& knots); diff --git a/third_party/eigen/unsupported/test/NonLinearOptimization.cpp b/third_party/eigen/unsupported/test/NonLinearOptimization.cpp index f0c336c15ef..dd93c21e9fe 100644 --- a/third_party/eigen/unsupported/test/NonLinearOptimization.cpp +++ b/third_party/eigen/unsupported/test/NonLinearOptimization.cpp @@ -15,6 +15,15 @@ // tolerance for chekcing number of iterations #define LM_EVAL_COUNT_TOL 4/3 +#define LM_CHECK_N_ITERS(SOLVER,NFEV,NJEV) { \ + ++g_test_level; \ + VERIFY_IS_EQUAL(SOLVER.nfev, NFEV); \ + VERIFY_IS_EQUAL(SOLVER.njev, NJEV); \ + --g_test_level; \ + VERIFY(SOLVER.nfev <= NFEV * LM_EVAL_COUNT_TOL); \ + VERIFY(SOLVER.njev <= NJEV * LM_EVAL_COUNT_TOL); \ + } + int fcn_chkder(const VectorXd &x, VectorXd &fvec, MatrixXd &fjac, int iflag) { /* subroutine fcn for chkder example. */ @@ -180,8 +189,7 @@ void testLmder1() // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 6); - VERIFY_IS_EQUAL(lm.njev, 5); + LM_CHECK_N_ITERS(lm, 6, 5); // check norm VERIFY_IS_APPROX(lm.fvec.blueNorm(), 0.09063596); @@ -209,8 +217,7 @@ void testLmder() // check return values VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 6); - VERIFY_IS_EQUAL(lm.njev, 5); + LM_CHECK_N_ITERS(lm, 6, 5); // check norm fnorm = lm.fvec.blueNorm(); @@ -294,8 +301,7 @@ void testHybrj1() // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(solver.nfev, 11); - VERIFY_IS_EQUAL(solver.njev, 1); + LM_CHECK_N_ITERS(solver, 11, 1); // check norm VERIFY_IS_APPROX(solver.fvec.blueNorm(), 1.192636e-08); @@ -329,8 +335,7 @@ void testHybrj() // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(solver.nfev, 11); - VERIFY_IS_EQUAL(solver.njev, 1); + LM_CHECK_N_ITERS(solver, 11, 1); // check norm VERIFY_IS_APPROX(solver.fvec.blueNorm(), 1.192636e-08); @@ -485,8 +490,7 @@ void testLmstr1() // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 6); - VERIFY_IS_EQUAL(lm.njev, 5); + LM_CHECK_N_ITERS(lm, 6, 5); // check norm VERIFY_IS_APPROX(lm.fvec.blueNorm(), 0.09063596); @@ -514,8 +518,7 @@ void testLmstr() // check return values VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 6); - VERIFY_IS_EQUAL(lm.njev, 5); + LM_CHECK_N_ITERS(lm, 6, 5); // check norm fnorm = lm.fvec.blueNorm(); @@ -686,8 +689,7 @@ void testNistChwirut2(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 10); - VERIFY_IS_EQUAL(lm.njev, 8); + LM_CHECK_N_ITERS(lm, 10, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.1304802941E+02); // check x @@ -707,8 +709,7 @@ void testNistChwirut2(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 7); - VERIFY_IS_EQUAL(lm.njev, 6); + LM_CHECK_N_ITERS(lm, 7, 6); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.1304802941E+02); // check x @@ -766,8 +767,7 @@ void testNistMisra1a(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 19); - VERIFY_IS_EQUAL(lm.njev, 15); + LM_CHECK_N_ITERS(lm, 19, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.2455138894E-01); // check x @@ -783,8 +783,7 @@ void testNistMisra1a(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 5); - VERIFY_IS_EQUAL(lm.njev, 4); + LM_CHECK_N_ITERS(lm, 5, 4); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.2455138894E-01); // check x @@ -856,8 +855,7 @@ void testNistHahn1(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 11); - VERIFY_IS_EQUAL(lm.njev, 10); + LM_CHECK_N_ITERS(lm, 11, 10); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.5324382854E+00); // check x @@ -878,8 +876,7 @@ void testNistHahn1(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 11); - VERIFY_IS_EQUAL(lm.njev, 10); + LM_CHECK_N_ITERS(lm, 11, 10); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.5324382854E+00); // check x @@ -942,8 +939,7 @@ void testNistMisra1d(void) // check return value VERIFY_IS_EQUAL(info, 3); - VERIFY_IS_EQUAL(lm.nfev, 9); - VERIFY_IS_EQUAL(lm.njev, 7); + LM_CHECK_N_ITERS(lm, 9, 7); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6419295283E-02); // check x @@ -959,8 +955,7 @@ void testNistMisra1d(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 4); - VERIFY_IS_EQUAL(lm.njev, 3); + LM_CHECK_N_ITERS(lm, 4, 3); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6419295283E-02); // check x @@ -1020,8 +1015,7 @@ void testNistLanczos1(void) // check return value VERIFY_IS_EQUAL(info, 2); - VERIFY_IS_EQUAL(lm.nfev, 79); - VERIFY_IS_EQUAL(lm.njev, 72); + LM_CHECK_N_ITERS(lm, 79, 72); // check norm^2 std::cout.precision(30); std::cout << lm.fvec.squaredNorm() << "\n"; @@ -1043,8 +1037,7 @@ void testNistLanczos1(void) // check return value VERIFY_IS_EQUAL(info, 2); - VERIFY_IS_EQUAL(lm.nfev, 9); - VERIFY_IS_EQUAL(lm.njev, 8); + LM_CHECK_N_ITERS(lm, 9, 8); // check norm^2 VERIFY(lm.fvec.squaredNorm() <= 1.4307867721E-25); // check x @@ -1108,8 +1101,7 @@ void testNistRat42(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 10); - VERIFY_IS_EQUAL(lm.njev, 8); + LM_CHECK_N_ITERS(lm, 10, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.0565229338E+00); // check x @@ -1126,8 +1118,7 @@ void testNistRat42(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 6); - VERIFY_IS_EQUAL(lm.njev, 5); + LM_CHECK_N_ITERS(lm, 6, 5); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.0565229338E+00); // check x @@ -1186,8 +1177,7 @@ void testNistMGH10(void) // check return value VERIFY_IS_EQUAL(info, 2); - VERIFY_IS_EQUAL(lm.nfev, 284 ); - VERIFY_IS_EQUAL(lm.njev, 249 ); + LM_CHECK_N_ITERS(lm, 284, 249); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7945855171E+01); // check x @@ -1204,8 +1194,7 @@ void testNistMGH10(void) // check return value VERIFY_IS_EQUAL(info, 3); - VERIFY_IS_EQUAL(lm.nfev, 126); - VERIFY_IS_EQUAL(lm.njev, 116); + LM_CHECK_N_ITERS(lm, 126, 116); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7945855171E+01); // check x @@ -1265,8 +1254,7 @@ void testNistBoxBOD(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY(lm.nfev < 31); // 31 - VERIFY(lm.njev < 25); // 25 + LM_CHECK_N_ITERS(lm, 31, 25); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.1680088766E+03); // check x @@ -1284,9 +1272,8 @@ void testNistBoxBOD(void) info = lm.minimize(x); // check return value - VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 15 ); - VERIFY_IS_EQUAL(lm.njev, 14 ); + VERIFY_IS_EQUAL(info, 1); + LM_CHECK_N_ITERS(lm, 15, 14); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.1680088766E+03); // check x @@ -1356,12 +1343,7 @@ void testNistMGH17(void) // check return value VERIFY_IS_EQUAL(info, 2); - ++g_test_level; - VERIFY_IS_EQUAL(lm.nfev, 602); // 602 - VERIFY_IS_EQUAL(lm.njev, 545); // 545 - --g_test_level; - VERIFY(lm.nfev < 602 * LM_EVAL_COUNT_TOL); - VERIFY(lm.njev < 545 * LM_EVAL_COUNT_TOL); + LM_CHECK_N_ITERS(lm, 602, 545); /* * Second try @@ -1373,8 +1355,7 @@ void testNistMGH17(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 18); - VERIFY_IS_EQUAL(lm.njev, 15); + LM_CHECK_N_ITERS(lm, 18, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.4648946975E-05); // check x @@ -1438,9 +1419,8 @@ void testNistMGH09(void) info = lm.minimize(x); // check return value - VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 490 ); - VERIFY_IS_EQUAL(lm.njev, 376 ); + VERIFY_IS_EQUAL(info, 1); + LM_CHECK_N_ITERS(lm, 490, 376); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 3.0750560385E-04); // check x @@ -1459,8 +1439,7 @@ void testNistMGH09(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 18); - VERIFY_IS_EQUAL(lm.njev, 16); + LM_CHECK_N_ITERS(lm, 18, 16); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 3.0750560385E-04); // check x @@ -1525,8 +1504,7 @@ void testNistBennett5(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 758); - VERIFY_IS_EQUAL(lm.njev, 744); + LM_CHECK_N_ITERS(lm, 758, 744); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.2404744073E-04); // check x @@ -1543,8 +1521,7 @@ void testNistBennett5(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 203); - VERIFY_IS_EQUAL(lm.njev, 192); + LM_CHECK_N_ITERS(lm, 203, 192); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.2404744073E-04); // check x @@ -1613,8 +1590,7 @@ void testNistThurber(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 39); - VERIFY_IS_EQUAL(lm.njev, 36); + LM_CHECK_N_ITERS(lm, 39,36); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6427082397E+03); // check x @@ -1638,8 +1614,7 @@ void testNistThurber(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 29); - VERIFY_IS_EQUAL(lm.njev, 28); + LM_CHECK_N_ITERS(lm, 29, 28); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6427082397E+03); // check x @@ -1705,8 +1680,7 @@ void testNistRat43(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 27); - VERIFY_IS_EQUAL(lm.njev, 20); + LM_CHECK_N_ITERS(lm, 27, 20); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7864049080E+03); // check x @@ -1727,8 +1701,7 @@ void testNistRat43(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 9); - VERIFY_IS_EQUAL(lm.njev, 8); + LM_CHECK_N_ITERS(lm, 9, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7864049080E+03); // check x @@ -1790,8 +1763,7 @@ void testNistEckerle4(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 18); - VERIFY_IS_EQUAL(lm.njev, 15); + LM_CHECK_N_ITERS(lm, 18, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.4635887487E-03); // check x @@ -1808,8 +1780,7 @@ void testNistEckerle4(void) // check return value VERIFY_IS_EQUAL(info, 1); - VERIFY_IS_EQUAL(lm.nfev, 7); - VERIFY_IS_EQUAL(lm.njev, 6); + LM_CHECK_N_ITERS(lm, 7, 6); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.4635887487E-03); // check x diff --git a/third_party/eigen/unsupported/test/autodiff.cpp b/third_party/eigen/unsupported/test/autodiff.cpp index 1c5e0dc6639..1d8c8b5fd71 100644 --- a/third_party/eigen/unsupported/test/autodiff.cpp +++ b/third_party/eigen/unsupported/test/autodiff.cpp @@ -352,6 +352,21 @@ double bug_1264() { return v2(0).value(); } +// check with expressions on constants +double bug_1281() { + int n = 2; + typedef AutoDiffScalar AD; + const AD c = 1.; + AD x0(2,n,0); + AD y1 = (AD(c)+AD(c))*x0; + y1 = x0 * (AD(c)+AD(c)); + AD y2 = (-AD(c))+x0; + y2 = x0+(-AD(c)); + AD y3 = (AD(c)*(-AD(c))+AD(c))*x0; + y3 = x0 * (AD(c)*(-AD(c))+AD(c)); + return (y1+y2+y3).value(); +} + #endif void test_autodiff() @@ -367,5 +382,6 @@ void test_autodiff() CALL_SUBTEST_5( bug_1223() ); CALL_SUBTEST_5( bug_1260() ); CALL_SUBTEST_5( bug_1261() ); + CALL_SUBTEST_5( bug_1281() ); } diff --git a/third_party/eigen/unsupported/test/matrix_function.cpp b/third_party/eigen/unsupported/test/matrix_function.cpp index 6a2b2194a08..005c9c15f35 100644 --- a/third_party/eigen/unsupported/test/matrix_function.cpp +++ b/third_party/eigen/unsupported/test/matrix_function.cpp @@ -177,6 +177,39 @@ void testMatrixType(const MatrixType& m) } } +template +void testMapRef(const MatrixType& A) +{ + // Test if passing Ref and Map objects is possible + // (Regression test for Bug #1796) + Index size = A.rows(); + MatrixType X; X.setRandom(size, size); + MatrixType Y(size,size); + Ref< MatrixType> R(Y); + Ref Rc(X); + Map< MatrixType> M(Y.data(), size, size); + Map Mc(X.data(), size, size); + + X = X*X; // make sure sqrt is possible + Y = X.sqrt(); + R = Rc.sqrt(); + M = Mc.sqrt(); + Y = X.exp(); + R = Rc.exp(); + M = Mc.exp(); + X = Y; // make sure log is possible + Y = X.log(); + R = Rc.log(); + M = Mc.log(); + + Y = X.cos() + Rc.cos() + Mc.cos(); + Y = X.sin() + Rc.sin() + Mc.sin(); + + Y = X.cosh() + Rc.cosh() + Mc.cosh(); + Y = X.sinh() + Rc.sinh() + Mc.sinh(); +} + + void test_matrix_function() { CALL_SUBTEST_1(testMatrixType(Matrix())); @@ -186,4 +219,9 @@ void test_matrix_function() CALL_SUBTEST_5(testMatrixType(Matrix())); CALL_SUBTEST_6(testMatrixType(Matrix4cd())); CALL_SUBTEST_7(testMatrixType(MatrixXd(13,13))); + + CALL_SUBTEST_1(testMapRef(Matrix())); + CALL_SUBTEST_2(testMapRef(Matrix3cf())); + CALL_SUBTEST_3(testMapRef(MatrixXf(8,8))); + CALL_SUBTEST_7(testMapRef(MatrixXd(13,13))); } diff --git a/third_party/eigen/unsupported/test/mpreal/mpreal.h b/third_party/eigen/unsupported/test/mpreal/mpreal.h index fe47bc3f88d..8404f1ff837 100644 --- a/third_party/eigen/unsupported/test/mpreal/mpreal.h +++ b/third_party/eigen/unsupported/test/mpreal/mpreal.h @@ -1,3104 +1,3104 @@ -/* - MPFR C++: Multi-precision floating point number class for C++. - Based on MPFR library: http://mpfr.org - - Project homepage: http://www.holoborodko.com/pavel/mpfr - Contact e-mail: pavel@holoborodko.com - - Copyright (c) 2008-2015 Pavel Holoborodko - - Contributors: - Dmitriy Gubanov, Konstantin Holoborodko, Brian Gladman, - Helmut Jarausch, Fokko Beekhof, Ulrich Mutze, Heinz van Saanen, - Pere Constans, Peter van Hoof, Gael Guennebaud, Tsai Chia Cheng, - Alexei Zubanov, Jauhien Piatlicki, Victor Berger, John Westwood, - Petr Aleksandrov, Orion Poplawski, Charles Karney, Arash Partow, - Rodney James, Jorge Leitao. - - Licensing: - (A) MPFR C++ is under GNU General Public License ("GPL"). - - (B) Non-free licenses may also be purchased from the author, for users who - do not want their programs protected by the GPL. - - The non-free licenses are for users that wish to use MPFR C++ in - their products but are unwilling to release their software - under the GPL (which would require them to release source code - and allow free redistribution). - - Such users can purchase an unlimited-use license from the author. - Contact us for more details. - - GNU General Public License ("GPL") copyright permissions statement: - ************************************************************************** - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . -*/ - -#ifndef __MPREAL_H__ -#define __MPREAL_H__ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -// Options -#define MPREAL_HAVE_MSVC_DEBUGVIEW // Enable Debugger Visualizer for "Debug" builds in MSVC. -#define MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS // Enable extended std::numeric_limits specialization. - // Meaning that "digits", "round_style" and similar members are defined as functions, not constants. - // See std::numeric_limits at the end of the file for more information. - -// Library version -#define MPREAL_VERSION_MAJOR 3 -#define MPREAL_VERSION_MINOR 6 -#define MPREAL_VERSION_PATCHLEVEL 2 -#define MPREAL_VERSION_STRING "3.6.2" - -// Detect compiler using signatures from http://predef.sourceforge.net/ -#if defined(__GNUC__) - #define IsInf(x) (isinf)(x) // GNU C++/Intel ICC compiler on Linux -#elif defined(_MSC_VER) // Microsoft Visual C++ - #define IsInf(x) (!_finite(x)) -#else - #define IsInf(x) (std::isinf)(x) // GNU C/C++ (and/or other compilers), just hope for C99 conformance -#endif - -// A Clang feature extension to determine compiler features. -#ifndef __has_feature - #define __has_feature(x) 0 -#endif - -// Detect support for r-value references (move semantic). Borrowed from Eigen. -#if (__has_feature(cxx_rvalue_references) || \ - defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L || \ - (defined(_MSC_VER) && _MSC_VER >= 1600)) - - #define MPREAL_HAVE_MOVE_SUPPORT - - // Use fields in mpfr_t structure to check if it was initialized / set dummy initialization - #define mpfr_is_initialized(x) (0 != (x)->_mpfr_d) - #define mpfr_set_uninitialized(x) ((x)->_mpfr_d = 0 ) -#endif - -// Detect support for explicit converters. -#if (__has_feature(cxx_explicit_conversions) || \ - (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GNUC_MINOR__ >= 5) || __cplusplus >= 201103L || \ - (defined(_MSC_VER) && _MSC_VER >= 1800)) - - #define MPREAL_HAVE_EXPLICIT_CONVERTERS -#endif - -#define MPFR_USE_INTMAX_T // Enable 64-bit integer types - should be defined before mpfr.h - -#if defined(MPREAL_HAVE_MSVC_DEBUGVIEW) && defined(_MSC_VER) && defined(_DEBUG) - #define MPREAL_MSVC_DEBUGVIEW_CODE DebugView = toString(); - #define MPREAL_MSVC_DEBUGVIEW_DATA std::string DebugView; -#else - #define MPREAL_MSVC_DEBUGVIEW_CODE - #define MPREAL_MSVC_DEBUGVIEW_DATA -#endif - -#include - -#if (MPFR_VERSION < MPFR_VERSION_NUM(3,0,0)) - #include // Needed for random() -#endif - -// Less important options -#define MPREAL_DOUBLE_BITS_OVERFLOW -1 // Triggers overflow exception during conversion to double if mpreal - // cannot fit in MPREAL_DOUBLE_BITS_OVERFLOW bits - // = -1 disables overflow checks (default) - -// Fast replacement for mpfr_set_zero(x, +1): -// (a) uses low-level data members, might not be compatible with new versions of MPFR -// (b) sign is not set, add (x)->_mpfr_sign = 1; -#define mpfr_set_zero_fast(x) ((x)->_mpfr_exp = __MPFR_EXP_ZERO) - -#if defined(__GNUC__) - #define MPREAL_PERMISSIVE_EXPR __extension__ -#else - #define MPREAL_PERMISSIVE_EXPR -#endif - -namespace mpfr { - -class mpreal { -private: - mpfr_t mp; - -public: - - // Get default rounding mode & precision - inline static mp_rnd_t get_default_rnd() { return (mp_rnd_t)(mpfr_get_default_rounding_mode()); } - inline static mp_prec_t get_default_prec() { return mpfr_get_default_prec(); } - - // Constructors && type conversions - mpreal(); - mpreal(const mpreal& u); - mpreal(const mpf_t u); - mpreal(const mpz_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const mpq_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - - // Construct mpreal from mpfr_t structure. - // shared = true allows to avoid deep copy, so that mpreal and 'u' share the same data & pointers. - mpreal(const mpfr_t u, bool shared = false); - - mpreal(const char* s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const std::string& s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); - - ~mpreal(); - -#ifdef MPREAL_HAVE_MOVE_SUPPORT - mpreal& operator=(mpreal&& v); - mpreal(mpreal&& u); -#endif - - // Operations - // = - // +, -, *, /, ++, --, <<, >> - // *=, +=, -=, /=, - // <, >, ==, <=, >= - - // = - mpreal& operator=(const mpreal& v); - mpreal& operator=(const mpf_t v); - mpreal& operator=(const mpz_t v); - mpreal& operator=(const mpq_t v); - mpreal& operator=(const long double v); - mpreal& operator=(const double v); - mpreal& operator=(const unsigned long int v); - mpreal& operator=(const unsigned long long int v); - mpreal& operator=(const long long int v); - mpreal& operator=(const unsigned int v); - mpreal& operator=(const long int v); - mpreal& operator=(const int v); - mpreal& operator=(const char* s); - mpreal& operator=(const std::string& s); - template mpreal& operator= (const std::complex& z); - - // + - mpreal& operator+=(const mpreal& v); - mpreal& operator+=(const mpf_t v); - mpreal& operator+=(const mpz_t v); - mpreal& operator+=(const mpq_t v); - mpreal& operator+=(const long double u); - mpreal& operator+=(const double u); - mpreal& operator+=(const unsigned long int u); - mpreal& operator+=(const unsigned int u); - mpreal& operator+=(const long int u); - mpreal& operator+=(const int u); - - mpreal& operator+=(const long long int u); - mpreal& operator+=(const unsigned long long int u); - mpreal& operator-=(const long long int u); - mpreal& operator-=(const unsigned long long int u); - mpreal& operator*=(const long long int u); - mpreal& operator*=(const unsigned long long int u); - mpreal& operator/=(const long long int u); - mpreal& operator/=(const unsigned long long int u); - - const mpreal operator+() const; - mpreal& operator++ (); - const mpreal operator++ (int); - - // - - mpreal& operator-=(const mpreal& v); - mpreal& operator-=(const mpz_t v); - mpreal& operator-=(const mpq_t v); - mpreal& operator-=(const long double u); - mpreal& operator-=(const double u); - mpreal& operator-=(const unsigned long int u); - mpreal& operator-=(const unsigned int u); - mpreal& operator-=(const long int u); - mpreal& operator-=(const int u); - const mpreal operator-() const; - friend const mpreal operator-(const unsigned long int b, const mpreal& a); - friend const mpreal operator-(const unsigned int b, const mpreal& a); - friend const mpreal operator-(const long int b, const mpreal& a); - friend const mpreal operator-(const int b, const mpreal& a); - friend const mpreal operator-(const double b, const mpreal& a); - mpreal& operator-- (); - const mpreal operator-- (int); - - // * - mpreal& operator*=(const mpreal& v); - mpreal& operator*=(const mpz_t v); - mpreal& operator*=(const mpq_t v); - mpreal& operator*=(const long double v); - mpreal& operator*=(const double v); - mpreal& operator*=(const unsigned long int v); - mpreal& operator*=(const unsigned int v); - mpreal& operator*=(const long int v); - mpreal& operator*=(const int v); - - // / - mpreal& operator/=(const mpreal& v); - mpreal& operator/=(const mpz_t v); - mpreal& operator/=(const mpq_t v); - mpreal& operator/=(const long double v); - mpreal& operator/=(const double v); - mpreal& operator/=(const unsigned long int v); - mpreal& operator/=(const unsigned int v); - mpreal& operator/=(const long int v); - mpreal& operator/=(const int v); - friend const mpreal operator/(const unsigned long int b, const mpreal& a); - friend const mpreal operator/(const unsigned int b, const mpreal& a); - friend const mpreal operator/(const long int b, const mpreal& a); - friend const mpreal operator/(const int b, const mpreal& a); - friend const mpreal operator/(const double b, const mpreal& a); - - //<<= Fast Multiplication by 2^u - mpreal& operator<<=(const unsigned long int u); - mpreal& operator<<=(const unsigned int u); - mpreal& operator<<=(const long int u); - mpreal& operator<<=(const int u); - - //>>= Fast Division by 2^u - mpreal& operator>>=(const unsigned long int u); - mpreal& operator>>=(const unsigned int u); - mpreal& operator>>=(const long int u); - mpreal& operator>>=(const int u); - - // Type Conversion operators - bool toBool ( ) const; - long toLong (mp_rnd_t mode = GMP_RNDZ) const; - unsigned long toULong (mp_rnd_t mode = GMP_RNDZ) const; - long long toLLong (mp_rnd_t mode = GMP_RNDZ) const; - unsigned long long toULLong (mp_rnd_t mode = GMP_RNDZ) const; - float toFloat (mp_rnd_t mode = GMP_RNDN) const; - double toDouble (mp_rnd_t mode = GMP_RNDN) const; - long double toLDouble (mp_rnd_t mode = GMP_RNDN) const; - -#if defined (MPREAL_HAVE_EXPLICIT_CONVERTERS) - explicit operator bool () const { return toBool(); } - explicit operator int () const { return int(toLong()); } - explicit operator long () const { return toLong(); } - explicit operator long long () const { return toLLong(); } - explicit operator unsigned () const { return unsigned(toULong()); } - explicit operator unsigned long () const { return toULong(); } - explicit operator unsigned long long () const { return toULLong(); } - explicit operator float () const { return toFloat(); } - explicit operator double () const { return toDouble(); } - explicit operator long double () const { return toLDouble(); } -#endif - - // Get raw pointers so that mpreal can be directly used in raw mpfr_* functions - ::mpfr_ptr mpfr_ptr(); - ::mpfr_srcptr mpfr_ptr() const; - ::mpfr_srcptr mpfr_srcptr() const; - - // Convert mpreal to string with n significant digits in base b - // n = -1 -> convert with the maximum available digits - std::string toString(int n = -1, int b = 10, mp_rnd_t mode = mpreal::get_default_rnd()) const; - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - std::string toString(const std::string& format) const; -#endif - - std::ostream& output(std::ostream& os) const; - - // Math Functions - friend const mpreal sqr (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sqrt(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sqrt(const unsigned long int v, mp_rnd_t rnd_mode); - friend const mpreal cbrt(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal root(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const long int b, mp_rnd_t rnd_mode); - friend const mpreal pow (const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode); - friend const mpreal pow (const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode); - friend const mpreal fabs(const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal abs(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); - friend inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); - friend inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); - friend int cmpabs(const mpreal& a,const mpreal& b); - - friend const mpreal log (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal logb (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log10(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp10(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log1p(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal expm1(const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal cos(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sin(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tan(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sec(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal csc(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal cot(const mpreal& v, mp_rnd_t rnd_mode); - friend int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal acos (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asin (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atan (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode); - friend const mpreal acot (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asec (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acsc (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal cosh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sinh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tanh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sech (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal csch (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal coth (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acosh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asinh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atanh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acoth (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asech (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acsch (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - - friend const mpreal fac_ui (unsigned long int v, mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal eint (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal gamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tgamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal lngamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal lgamma (const mpreal& v, int *signp, mp_rnd_t rnd_mode); - friend const mpreal zeta (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal erf (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal erfc (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselj0 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselj1 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besseljn (long n, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal bessely0 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal bessely1 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselyn (long n, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); - friend const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); - friend const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode); - friend const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t rnd_mode); - friend int sgn(const mpreal& v); // returns -1 or +1 - -// MPFR 2.4.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - friend int sinh_cosh (mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal li2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal rec_sqrt (const mpreal& v, mp_rnd_t rnd_mode); - - // MATLAB's semantic equivalents - friend const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Remainder after division - friend const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Modulus after division -#endif - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - friend const mpreal digamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal ai (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear -#endif - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - friend const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear - friend const mpreal grandom (unsigned int seed); -#endif - - // Uniformly distributed random number generation in [0,1] using - // Mersenne-Twister algorithm by default. - // Use parameter to setup seed, e.g.: random((unsigned)time(NULL)) - // Check urandom() for more precise control. - friend const mpreal random(unsigned int seed); - - // Splits mpreal value into fractional and integer parts. - // Returns fractional part and stores integer part in n. - friend const mpreal modf(const mpreal& v, mpreal& n); - - // Constants - // don't forget to call mpfr_free_cache() for every thread where you are using const-functions - friend const mpreal const_log2 (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_pi (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_euler (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_catalan (mp_prec_t prec, mp_rnd_t rnd_mode); - - // returns +inf iff sign>=0 otherwise -inf - friend const mpreal const_infinity(int sign, mp_prec_t prec); - - // Output/ Input - friend std::ostream& operator<<(std::ostream& os, const mpreal& v); - friend std::istream& operator>>(std::istream& is, mpreal& v); - - // Integer Related Functions - friend const mpreal rint (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal ceil (const mpreal& v); - friend const mpreal floor(const mpreal& v); - friend const mpreal round(const mpreal& v); - friend const mpreal trunc(const mpreal& v); - friend const mpreal rint_ceil (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_floor (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_round (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_trunc (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal frac (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal remainder ( const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - - // Miscellaneous Functions - friend const mpreal nexttoward (const mpreal& x, const mpreal& y); - friend const mpreal nextabove (const mpreal& x); - friend const mpreal nextbelow (const mpreal& x); - - // use gmp_randinit_default() to init state, gmp_randclear() to clear - friend const mpreal urandomb (gmp_randstate_t& state); - -// MPFR < 2.4.2 Specifics -#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) - friend const mpreal random2 (mp_size_t size, mp_exp_t exp); -#endif - - // Instance Checkers - friend bool (isnan) (const mpreal& v); - friend bool (isinf) (const mpreal& v); - friend bool (isfinite) (const mpreal& v); - - friend bool isnum (const mpreal& v); - friend bool iszero (const mpreal& v); - friend bool isint (const mpreal& v); - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - friend bool isregular(const mpreal& v); -#endif - - // Set/Get instance properties - inline mp_prec_t get_prec() const; - inline void set_prec(mp_prec_t prec, mp_rnd_t rnd_mode = get_default_rnd()); // Change precision with rounding mode - - // Aliases for get_prec(), set_prec() - needed for compatibility with std::complex interface - inline mpreal& setPrecision(int Precision, mp_rnd_t RoundingMode = get_default_rnd()); - inline int getPrecision() const; - - // Set mpreal to +/- inf, NaN, +/-0 - mpreal& setInf (int Sign = +1); - mpreal& setNan (); - mpreal& setZero (int Sign = +1); - mpreal& setSign (int Sign, mp_rnd_t RoundingMode = get_default_rnd()); - - //Exponent - mp_exp_t get_exp(); - int set_exp(mp_exp_t e); - int check_range (int t, mp_rnd_t rnd_mode = get_default_rnd()); - int subnormalize (int t, mp_rnd_t rnd_mode = get_default_rnd()); - - // Inexact conversion from float - inline bool fits_in_bits(double x, int n); - - // Set/Get global properties - static void set_default_prec(mp_prec_t prec); - static void set_default_rnd(mp_rnd_t rnd_mode); - - static mp_exp_t get_emin (void); - static mp_exp_t get_emax (void); - static mp_exp_t get_emin_min (void); - static mp_exp_t get_emin_max (void); - static mp_exp_t get_emax_min (void); - static mp_exp_t get_emax_max (void); - static int set_emin (mp_exp_t exp); - static int set_emax (mp_exp_t exp); - - // Efficient swapping of two mpreal values - needed for std algorithms - friend void swap(mpreal& x, mpreal& y); - - friend const mpreal fmax(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal fmin(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - -private: - // Human friendly Debug Preview in Visual Studio. - // Put one of these lines: - // - // mpfr::mpreal= ; Show value only - // mpfr::mpreal=, bits ; Show value & precision - // - // at the beginning of - // [Visual Studio Installation Folder]\Common7\Packages\Debugger\autoexp.dat - MPREAL_MSVC_DEBUGVIEW_DATA - - // "Smart" resources deallocation. Checks if instance initialized before deletion. - void clear(::mpfr_ptr); -}; - -////////////////////////////////////////////////////////////////////////// -// Exceptions -class conversion_overflow : public std::exception { -public: - std::string why() { return "inexact conversion from floating point"; } -}; - -////////////////////////////////////////////////////////////////////////// -// Constructors & converters -// Default constructor: creates mp number and initializes it to 0. -inline mpreal::mpreal() -{ - mpfr_init2(mpfr_ptr(), mpreal::get_default_prec()); - mpfr_set_zero_fast(mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpreal& u) -{ - mpfr_init2(mpfr_ptr(),mpfr_get_prec(u.mpfr_srcptr())); - mpfr_set (mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -#ifdef MPREAL_HAVE_MOVE_SUPPORT -inline mpreal::mpreal(mpreal&& other) -{ - mpfr_set_uninitialized(mpfr_ptr()); // make sure "other" holds no pointer to actual data - mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal& mpreal::operator=(mpreal&& other) -{ - mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} -#endif - -inline mpreal::mpreal(const mpfr_t u, bool shared) -{ - if(shared) - { - std::memcpy(mpfr_ptr(), u, sizeof(mpfr_t)); - } - else - { - mpfr_init2(mpfr_ptr(), mpfr_get_prec(u)); - mpfr_set (mpfr_ptr(), u, mpreal::get_default_rnd()); - } - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpf_t u) -{ - mpfr_init2(mpfr_ptr(),(mp_prec_t) mpf_get_prec(u)); // (gmp: mp_bitcnt_t) unsigned long -> long (mpfr: mp_prec_t) - mpfr_set_f(mpfr_ptr(),u,mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpz_t u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - mpfr_set_z(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpq_t u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - mpfr_set_q(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const double u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - -#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) - if(fits_in_bits(u, MPREAL_DOUBLE_BITS_OVERFLOW)) - { - mpfr_set_d(mpfr_ptr(), u, mode); - }else - throw conversion_overflow(); -#else - mpfr_set_d(mpfr_ptr(), u, mode); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long double u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ld(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned long long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_uj(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_sj(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ui(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ui(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_si(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_si(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_str(mpfr_ptr(), s, base, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const std::string& s, mp_prec_t prec, int base, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_str(mpfr_ptr(), s.c_str(), base, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline void mpreal::clear(::mpfr_ptr x) -{ -#ifdef MPREAL_HAVE_MOVE_SUPPORT - if(mpfr_is_initialized(x)) -#endif - mpfr_clear(x); -} - -inline mpreal::~mpreal() -{ - clear(mpfr_ptr()); -} - -// internal namespace needed for template magic -namespace internal{ - - // Use SFINAE to restrict arithmetic operations instantiation only for numeric types - // This is needed for smooth integration with libraries based on expression templates, like Eigen. - // TODO: Do the same for boolean operators. - template struct result_type {}; - - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; -} - -// + Addition -template -inline const typename internal::result_type::type - operator+(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) += rhs; } - -template -inline const typename internal::result_type::type - operator+(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) += lhs; } - -// - Subtraction -template -inline const typename internal::result_type::type - operator-(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) -= rhs; } - -template -inline const typename internal::result_type::type - operator-(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) -= rhs; } - -// * Multiplication -template -inline const typename internal::result_type::type - operator*(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) *= rhs; } - -template -inline const typename internal::result_type::type - operator*(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) *= lhs; } - -// / Division -template -inline const typename internal::result_type::type - operator/(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) /= rhs; } - -template -inline const typename internal::result_type::type - operator/(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) /= rhs; } - -////////////////////////////////////////////////////////////////////////// -// sqrt -const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const long int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const long double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -// abs -inline const mpreal abs(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()); - -////////////////////////////////////////////////////////////////////////// -// pow -const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -////////////////////////////////////////////////////////////////////////// -// Estimate machine epsilon for the given precision -// Returns smallest eps such that 1.0 + eps != 1.0 -inline mpreal machine_epsilon(mp_prec_t prec = mpreal::get_default_prec()); - -// Returns smallest eps such that x + eps != x (relative machine epsilon) -inline mpreal machine_epsilon(const mpreal& x); - -// Gives max & min values for the required precision, -// minval is 'safe' meaning 1 / minval does not overflow -// maxval is 'safe' meaning 1 / maxval does not underflow -inline mpreal minval(mp_prec_t prec = mpreal::get_default_prec()); -inline mpreal maxval(mp_prec_t prec = mpreal::get_default_prec()); - -// 'Dirty' equality check 1: |a-b| < min{|a|,|b|} * eps -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps); - -// 'Dirty' equality check 2: |a-b| < min{|a|,|b|} * eps( min{|a|,|b|} ) -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b); - -// 'Bitwise' equality check -// maxUlps - a and b can be apart by maxUlps binary numbers. -inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps); - -////////////////////////////////////////////////////////////////////////// -// Convert precision in 'bits' to decimal digits and vice versa. -// bits = ceil(digits*log[2](10)) -// digits = floor(bits*log[10](2)) - -inline mp_prec_t digits2bits(int d); -inline int bits2digits(mp_prec_t b); - -////////////////////////////////////////////////////////////////////////// -// min, max -const mpreal (max)(const mpreal& x, const mpreal& y); -const mpreal (min)(const mpreal& x, const mpreal& y); - -////////////////////////////////////////////////////////////////////////// -// Implementation -////////////////////////////////////////////////////////////////////////// - -////////////////////////////////////////////////////////////////////////// -// Operators - Assignment -inline mpreal& mpreal::operator=(const mpreal& v) -{ - if (this != &v) - { - mp_prec_t tp = mpfr_get_prec( mpfr_srcptr()); - mp_prec_t vp = mpfr_get_prec(v.mpfr_srcptr()); - - if(tp != vp){ - clear(mpfr_ptr()); - mpfr_init2(mpfr_ptr(), vp); - } - - mpfr_set(mpfr_ptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - } - return *this; -} - -inline mpreal& mpreal::operator=(const mpf_t v) -{ - mpfr_set_f(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const mpz_t v) -{ - mpfr_set_z(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const mpq_t v) -{ - mpfr_set_q(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long double v) -{ - mpfr_set_ld(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const double v) -{ -#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) - if(fits_in_bits(v, MPREAL_DOUBLE_BITS_OVERFLOW)) - { - mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); - }else - throw conversion_overflow(); -#else - mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned long int v) -{ - mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned int v) -{ - mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned long long int v) -{ - mpfr_set_uj(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long long int v) -{ - mpfr_set_sj(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long int v) -{ - mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const int v) -{ - mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const char* s) -{ - // Use other converters for more precise control on base & precision & rounding: - // - // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) - // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) - // - // Here we assume base = 10 and we use precision of target variable. - - mpfr_t t; - - mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); - - if(0 == mpfr_set_str(t, s, 10, mpreal::get_default_rnd())) - { - mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - } - - clear(t); - return *this; -} - -inline mpreal& mpreal::operator=(const std::string& s) -{ - // Use other converters for more precise control on base & precision & rounding: - // - // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) - // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) - // - // Here we assume base = 10 and we use precision of target variable. - - mpfr_t t; - - mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); - - if(0 == mpfr_set_str(t, s.c_str(), 10, mpreal::get_default_rnd())) - { - mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - } - - clear(t); - return *this; -} - -template -inline mpreal& mpreal::operator= (const std::complex& z) -{ - return *this = z.real(); -} - -////////////////////////////////////////////////////////////////////////// -// + Addition -inline mpreal& mpreal::operator+=(const mpreal& v) -{ - mpfr_add(mpfr_ptr(), mpfr_srcptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpf_t u) -{ - *this += mpreal(u); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpz_t u) -{ - mpfr_add_z(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpq_t u) -{ - mpfr_add_q(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+= (const long double u) -{ - *this += mpreal(u); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+= (const double u) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_add_d(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); -#else - *this += mpreal(u); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const unsigned long int u) -{ - mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const unsigned int u) -{ - mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const long int u) -{ - mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const int u) -{ - mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const long long int u) { *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator+=(const unsigned long long int u){ *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator-=(const long long int u) { *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator-=(const unsigned long long int u){ *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator*=(const long long int u) { *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator*=(const unsigned long long int u){ *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator/=(const long long int u) { *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator/=(const unsigned long long int u){ *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } - -inline const mpreal mpreal::operator+()const { return mpreal(*this); } - -inline const mpreal operator+(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_add(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline mpreal& mpreal::operator++() -{ - return *this += 1; -} - -inline const mpreal mpreal::operator++ (int) -{ - mpreal x(*this); - *this += 1; - return x; -} - -inline mpreal& mpreal::operator--() -{ - return *this -= 1; -} - -inline const mpreal mpreal::operator-- (int) -{ - mpreal x(*this); - *this -= 1; - return x; -} - -////////////////////////////////////////////////////////////////////////// -// - Subtraction -inline mpreal& mpreal::operator-=(const mpreal& v) -{ - mpfr_sub(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const mpz_t v) -{ - mpfr_sub_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const mpq_t v) -{ - mpfr_sub_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const long double v) -{ - *this -= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_sub_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this -= mpreal(v); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const unsigned long int v) -{ - mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const unsigned int v) -{ - mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const long int v) -{ - mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const int v) -{ - mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal mpreal::operator-()const -{ - mpreal u(*this); - mpfr_neg(u.mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); - return u; -} - -inline const mpreal operator-(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_sub(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline const mpreal operator-(const double b, const mpreal& a) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_d_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -#else - mpreal x(b, mpfr_get_prec(a.mpfr_ptr())); - x -= a; - return x; -#endif -} - -inline const mpreal operator-(const unsigned long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const unsigned int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -////////////////////////////////////////////////////////////////////////// -// * Multiplication -inline mpreal& mpreal::operator*= (const mpreal& v) -{ - mpfr_mul(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const mpz_t v) -{ - mpfr_mul_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const mpq_t v) -{ - mpfr_mul_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const long double v) -{ - *this *= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_mul_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this *= mpreal(v); -#endif - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const unsigned long int v) -{ - mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const unsigned int v) -{ - mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const long int v) -{ - mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const int v) -{ - mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator*(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_mul(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -////////////////////////////////////////////////////////////////////////// -// / Division -inline mpreal& mpreal::operator/=(const mpreal& v) -{ - mpfr_div(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const mpz_t v) -{ - mpfr_div_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const mpq_t v) -{ - mpfr_div_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const long double v) -{ - *this /= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_div_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this /= mpreal(v); -#endif - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const unsigned long int v) -{ - mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const unsigned int v) -{ - mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const long int v) -{ - mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const int v) -{ - mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator/(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_srcptr()), mpfr_get_prec(b.mpfr_srcptr()))); - mpfr_div(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline const mpreal operator/(const unsigned long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const unsigned int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const double b, const mpreal& a) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_d_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -#else - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - x /= a; - return x; -#endif -} - -////////////////////////////////////////////////////////////////////////// -// Shifts operators - Multiplication/Division by power of 2 -inline mpreal& mpreal::operator<<=(const unsigned long int u) -{ - mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const unsigned int u) -{ - mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const long int u) -{ - mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const int u) -{ - mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const unsigned long int u) -{ - mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const unsigned int u) -{ - mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const long int u) -{ - mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const int u) -{ - mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator<<(const mpreal& v, const unsigned long int k) -{ - return mul_2ui(v,k); -} - -inline const mpreal operator<<(const mpreal& v, const unsigned int k) -{ - return mul_2ui(v,static_cast(k)); -} - -inline const mpreal operator<<(const mpreal& v, const long int k) -{ - return mul_2si(v,k); -} - -inline const mpreal operator<<(const mpreal& v, const int k) -{ - return mul_2si(v,static_cast(k)); -} - -inline const mpreal operator>>(const mpreal& v, const unsigned long int k) -{ - return div_2ui(v,k); -} - -inline const mpreal operator>>(const mpreal& v, const long int k) -{ - return div_2si(v,k); -} - -inline const mpreal operator>>(const mpreal& v, const unsigned int k) -{ - return div_2ui(v,static_cast(k)); -} - -inline const mpreal operator>>(const mpreal& v, const int k) -{ - return div_2si(v,static_cast(k)); -} - -// mul_2ui -inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_mul_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -// mul_2si -inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_mul_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_div_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_div_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -////////////////////////////////////////////////////////////////////////// -//Relational operators - -// WARNING: -// -// Please note that following checks for double-NaN are guaranteed to work only in IEEE math mode: -// -// isnan(b) = (b != b) -// isnan(b) = !(b == b) (we use in code below) -// -// Be cautions if you use compiler options which break strict IEEE compliance (e.g. -ffast-math in GCC). -// Use std::isnan instead (C++11). - -inline bool operator > (const mpreal& a, const mpreal& b ){ return (mpfr_greater_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator > (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) > 0 ); } - -inline bool operator >= (const mpreal& a, const mpreal& b ){ return (mpfr_greaterequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator >= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } -// inline bool operator >= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (isnan()a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) >= 0 ); } - -inline bool operator < (const mpreal& a, const mpreal& b ){ return (mpfr_less_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator < (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) < 0 ); } - -inline bool operator <= (const mpreal& a, const mpreal& b ){ return (mpfr_lessequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator <= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) <= 0 ); } - -inline bool operator == (const mpreal& a, const mpreal& b ){ return (mpfr_equal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator == (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) == 0 ); } - -inline bool operator != (const mpreal& a, const mpreal& b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const unsigned long int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const unsigned int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const long int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const long double b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const double b ){ return !(a == b); } - -inline bool (isnan) (const mpreal& op){ return (mpfr_nan_p (op.mpfr_srcptr()) != 0 ); } -inline bool (isinf) (const mpreal& op){ return (mpfr_inf_p (op.mpfr_srcptr()) != 0 ); } -inline bool (isfinite) (const mpreal& op){ return (mpfr_number_p (op.mpfr_srcptr()) != 0 ); } -inline bool iszero (const mpreal& op){ return (mpfr_zero_p (op.mpfr_srcptr()) != 0 ); } -inline bool isint (const mpreal& op){ return (mpfr_integer_p(op.mpfr_srcptr()) != 0 ); } - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) -inline bool isregular(const mpreal& op){ return (mpfr_regular_p(op.mpfr_srcptr()));} -#endif - -////////////////////////////////////////////////////////////////////////// -// Type Converters -inline bool mpreal::toBool ( ) const { return mpfr_zero_p (mpfr_srcptr()) == 0; } -inline long mpreal::toLong (mp_rnd_t mode) const { return mpfr_get_si (mpfr_srcptr(), mode); } -inline unsigned long mpreal::toULong (mp_rnd_t mode) const { return mpfr_get_ui (mpfr_srcptr(), mode); } -inline float mpreal::toFloat (mp_rnd_t mode) const { return mpfr_get_flt(mpfr_srcptr(), mode); } -inline double mpreal::toDouble (mp_rnd_t mode) const { return mpfr_get_d (mpfr_srcptr(), mode); } -inline long double mpreal::toLDouble(mp_rnd_t mode) const { return mpfr_get_ld (mpfr_srcptr(), mode); } -inline long long mpreal::toLLong (mp_rnd_t mode) const { return mpfr_get_sj (mpfr_srcptr(), mode); } -inline unsigned long long mpreal::toULLong (mp_rnd_t mode) const { return mpfr_get_uj (mpfr_srcptr(), mode); } - -inline ::mpfr_ptr mpreal::mpfr_ptr() { return mp; } -inline ::mpfr_srcptr mpreal::mpfr_ptr() const { return mp; } -inline ::mpfr_srcptr mpreal::mpfr_srcptr() const { return mp; } - -template -inline std::string toString(T t, std::ios_base & (*f)(std::ios_base&)) -{ - std::ostringstream oss; - oss << f << t; - return oss.str(); -} - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - -inline std::string mpreal::toString(const std::string& format) const -{ - char *s = NULL; - std::string out; - - if( !format.empty() ) - { - if(!(mpfr_asprintf(&s, format.c_str(), mpfr_srcptr()) < 0)) - { - out = std::string(s); - - mpfr_free_str(s); - } - } - - return out; -} - -#endif - -inline std::string mpreal::toString(int n, int b, mp_rnd_t mode) const -{ - // TODO: Add extended format specification (f, e, rounding mode) as it done in output operator - (void)b; - (void)mode; - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - - std::ostringstream format; - - int digits = (n >= 0) ? n : 1 + bits2digits(mpfr_get_prec(mpfr_srcptr())); - - format << "%." << digits << "RNg"; - - return toString(format.str()); - -#else - - char *s, *ns = NULL; - size_t slen, nslen; - mp_exp_t exp; - std::string out; - - if(mpfr_inf_p(mp)) - { - if(mpfr_sgn(mp)>0) return "+Inf"; - else return "-Inf"; - } - - if(mpfr_zero_p(mp)) return "0"; - if(mpfr_nan_p(mp)) return "NaN"; - - s = mpfr_get_str(NULL, &exp, b, 0, mp, mode); - ns = mpfr_get_str(NULL, &exp, b, (std::max)(0,n), mp, mode); - - if(s!=NULL && ns!=NULL) - { - slen = strlen(s); - nslen = strlen(ns); - if(nslen<=slen) - { - mpfr_free_str(s); - s = ns; - slen = nslen; - } - else { - mpfr_free_str(ns); - } - - // Make human eye-friendly formatting if possible - if (exp>0 && static_cast(exp)s+exp) ptr--; - - if(ptr==s+exp) out = std::string(s,exp+1); - else out = std::string(s,exp+1)+'.'+std::string(s+exp+1,ptr-(s+exp+1)+1); - - //out = string(s,exp+1)+'.'+string(s+exp+1); - } - else - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s+exp-1) ptr--; - - if(ptr==s+exp-1) out = std::string(s,exp); - else out = std::string(s,exp)+'.'+std::string(s+exp,ptr-(s+exp)+1); - - //out = string(s,exp)+'.'+string(s+exp); - } - - }else{ // exp<0 || exp>slen - if(s[0]=='-') - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s+1) ptr--; - - if(ptr==s+1) out = std::string(s,2); - else out = std::string(s,2)+'.'+std::string(s+2,ptr-(s+2)+1); - - //out = string(s,2)+'.'+string(s+2); - } - else - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s) ptr--; - - if(ptr==s) out = std::string(s,1); - else out = std::string(s,1)+'.'+std::string(s+1,ptr-(s+1)+1); - - //out = string(s,1)+'.'+string(s+1); - } - - // Make final string - if(--exp) - { - if(exp>0) out += "e+"+mpfr::toString(exp,std::dec); - else out += "e"+mpfr::toString(exp,std::dec); - } - } - - mpfr_free_str(s); - return out; - }else{ - return "conversion error!"; - } -#endif -} - - -////////////////////////////////////////////////////////////////////////// -// I/O -inline std::ostream& mpreal::output(std::ostream& os) const -{ - std::ostringstream format; - const std::ios::fmtflags flags = os.flags(); - - format << ((flags & std::ios::showpos) ? "%+" : "%"); - if (os.precision() >= 0) - format << '.' << os.precision() << "R*" - << ((flags & std::ios::floatfield) == std::ios::fixed ? 'f' : - (flags & std::ios::floatfield) == std::ios::scientific ? 'e' : - 'g'); - else - format << "R*e"; - - char *s = NULL; - if(!(mpfr_asprintf(&s, format.str().c_str(), - mpfr::mpreal::get_default_rnd(), - mpfr_srcptr()) - < 0)) - { - os << std::string(s); - mpfr_free_str(s); - } - return os; -} - -inline std::ostream& operator<<(std::ostream& os, const mpreal& v) -{ - return v.output(os); -} - -inline std::istream& operator>>(std::istream &is, mpreal& v) -{ - // TODO: use cout::hexfloat and other flags to setup base - std::string tmp; - is >> tmp; - mpfr_set_str(v.mpfr_ptr(), tmp.c_str(), 10, mpreal::get_default_rnd()); - return is; -} - -////////////////////////////////////////////////////////////////////////// -// Bits - decimal digits relation -// bits = ceil(digits*log[2](10)) -// digits = floor(bits*log[10](2)) - -inline mp_prec_t digits2bits(int d) -{ - const double LOG2_10 = 3.3219280948873624; - - return mp_prec_t(std::ceil( d * LOG2_10 )); -} - -inline int bits2digits(mp_prec_t b) -{ - const double LOG10_2 = 0.30102999566398119; - - return int(std::floor( b * LOG10_2 )); -} - -////////////////////////////////////////////////////////////////////////// -// Set/Get number properties -inline int sgn(const mpreal& op) -{ - return mpfr_sgn(op.mpfr_srcptr()); -} - -inline mpreal& mpreal::setSign(int sign, mp_rnd_t RoundingMode) -{ - mpfr_setsign(mpfr_ptr(), mpfr_srcptr(), (sign < 0 ? 1 : 0), RoundingMode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline int mpreal::getPrecision() const -{ - return int(mpfr_get_prec(mpfr_srcptr())); -} - -inline mpreal& mpreal::setPrecision(int Precision, mp_rnd_t RoundingMode) -{ - mpfr_prec_round(mpfr_ptr(), Precision, RoundingMode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setInf(int sign) -{ - mpfr_set_inf(mpfr_ptr(), sign); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setNan() -{ - mpfr_set_nan(mpfr_ptr()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setZero(int sign) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - mpfr_set_zero(mpfr_ptr(), sign); -#else - mpfr_set_si(mpfr_ptr(), 0, (mpfr_get_default_rounding_mode)()); - setSign(sign); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mp_prec_t mpreal::get_prec() const -{ - return mpfr_get_prec(mpfr_srcptr()); -} - -inline void mpreal::set_prec(mp_prec_t prec, mp_rnd_t rnd_mode) -{ - mpfr_prec_round(mpfr_ptr(),prec,rnd_mode); - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mp_exp_t mpreal::get_exp () -{ - return mpfr_get_exp(mpfr_srcptr()); -} - -inline int mpreal::set_exp (mp_exp_t e) -{ - int x = mpfr_set_exp(mpfr_ptr(), e); - MPREAL_MSVC_DEBUGVIEW_CODE; - return x; -} - -inline const mpreal frexp(const mpreal& x, mp_exp_t* exp, mp_rnd_t mode = mpreal::get_default_rnd()) -{ - mpreal y(x); -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - mpfr_frexp(exp,y.mpfr_ptr(),x.mpfr_srcptr(),mode); -#else - *exp = mpfr_get_exp(y.mpfr_srcptr()); - mpfr_set_exp(y.mpfr_ptr(),0); -#endif - return y; -} - -inline const mpreal ldexp(const mpreal& v, mp_exp_t exp) -{ - mpreal x(v); - - // rounding is not important since we are just increasing the exponent (= exact operation) - mpfr_mul_2si(x.mpfr_ptr(), x.mpfr_srcptr(), exp, mpreal::get_default_rnd()); - return x; -} - -inline const mpreal scalbn(const mpreal& v, mp_exp_t exp) -{ - return ldexp(v, exp); -} - -inline mpreal machine_epsilon(mp_prec_t prec) -{ - /* the smallest eps such that 1 + eps != 1 */ - return machine_epsilon(mpreal(1, prec)); -} - -inline mpreal machine_epsilon(const mpreal& x) -{ - /* the smallest eps such that x + eps != x */ - if( x < 0) - { - return nextabove(-x) + x; - }else{ - return nextabove( x) - x; - } -} - -// minval is 'safe' meaning 1 / minval does not overflow -inline mpreal minval(mp_prec_t prec) -{ - /* min = 1/2 * 2^emin = 2^(emin - 1) */ - return mpreal(1, prec) << mpreal::get_emin()-1; -} - -// maxval is 'safe' meaning 1 / maxval does not underflow -inline mpreal maxval(mp_prec_t prec) -{ - /* max = (1 - eps) * 2^emax, eps is machine epsilon */ - return (mpreal(1, prec) - machine_epsilon(prec)) << mpreal::get_emax(); -} - -inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps) -{ - return abs(a - b) <= machine_epsilon((max)(abs(a), abs(b))) * maxUlps; -} - -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps) -{ - return abs(a - b) <= eps; -} - -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b) -{ - return isEqualFuzzy(a, b, machine_epsilon((max)(1, (min)(abs(a), abs(b))))); -} - -////////////////////////////////////////////////////////////////////////// -// C++11 sign functions. -inline mpreal copysign(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal rop(0, mpfr_get_prec(x.mpfr_ptr())); - mpfr_setsign(rop.mpfr_ptr(), x.mpfr_srcptr(), mpfr_signbit(y.mpfr_srcptr()), rnd_mode); - return rop; -} - -inline bool signbit(const mpreal& x) -{ - return mpfr_signbit(x.mpfr_srcptr()); -} - -inline const mpreal modf(const mpreal& v, mpreal& n) -{ - mpreal f(v); - - // rounding is not important since we are using the same number - mpfr_frac (f.mpfr_ptr(),f.mpfr_srcptr(),mpreal::get_default_rnd()); - mpfr_trunc(n.mpfr_ptr(),v.mpfr_srcptr()); - return f; -} - -inline int mpreal::check_range (int t, mp_rnd_t rnd_mode) -{ - return mpfr_check_range(mpfr_ptr(),t,rnd_mode); -} - -inline int mpreal::subnormalize (int t,mp_rnd_t rnd_mode) -{ - int r = mpfr_subnormalize(mpfr_ptr(),t,rnd_mode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return r; -} - -inline mp_exp_t mpreal::get_emin (void) -{ - return mpfr_get_emin(); -} - -inline int mpreal::set_emin (mp_exp_t exp) -{ - return mpfr_set_emin(exp); -} - -inline mp_exp_t mpreal::get_emax (void) -{ - return mpfr_get_emax(); -} - -inline int mpreal::set_emax (mp_exp_t exp) -{ - return mpfr_set_emax(exp); -} - -inline mp_exp_t mpreal::get_emin_min (void) -{ - return mpfr_get_emin_min(); -} - -inline mp_exp_t mpreal::get_emin_max (void) -{ - return mpfr_get_emin_max(); -} - -inline mp_exp_t mpreal::get_emax_min (void) -{ - return mpfr_get_emax_min(); -} - -inline mp_exp_t mpreal::get_emax_max (void) -{ - return mpfr_get_emax_max(); -} - -////////////////////////////////////////////////////////////////////////// -// Mathematical Functions -////////////////////////////////////////////////////////////////////////// -#define MPREAL_UNARY_MATH_FUNCTION_BODY(f) \ - mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); \ - mpfr_##f(y.mpfr_ptr(), x.mpfr_srcptr(), r); \ - return y; - -inline const mpreal sqr (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqr ); } - -inline const mpreal sqrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqrt); } - -inline const mpreal sqrt(const unsigned long int x, mp_rnd_t r) -{ - mpreal y; - mpfr_sqrt_ui(y.mpfr_ptr(), x, r); - return y; -} - -inline const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode) -{ - return sqrt(static_cast(v),rnd_mode); -} - -inline const mpreal sqrt(const long int v, mp_rnd_t rnd_mode) -{ - if (v>=0) return sqrt(static_cast(v),rnd_mode); - else return mpreal().setNan(); // NaN -} - -inline const mpreal sqrt(const int v, mp_rnd_t rnd_mode) -{ - if (v>=0) return sqrt(static_cast(v),rnd_mode); - else return mpreal().setNan(); // NaN -} - -inline const mpreal root(const mpreal& x, unsigned long int k, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); - mpfr_root(y.mpfr_ptr(), x.mpfr_srcptr(), k, r); - return y; -} - -inline const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_dim(y.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), r); - return y; -} - -inline int cmpabs(const mpreal& a,const mpreal& b) -{ - return mpfr_cmpabs(a.mpfr_ptr(), b.mpfr_srcptr()); -} - -inline int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - return mpfr_sin_cos(s.mpfr_ptr(), c.mpfr_ptr(), v.mpfr_srcptr(), rnd_mode); -} - -inline const mpreal sqrt (const long double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } -inline const mpreal sqrt (const double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } - -inline const mpreal cbrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cbrt ); } -inline const mpreal fabs (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } -inline const mpreal abs (const mpreal& x, mp_rnd_t r) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } -inline const mpreal log (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log ); } -inline const mpreal log2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log2 ); } -inline const mpreal log10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log10); } -inline const mpreal exp (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp ); } -inline const mpreal exp2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp2 ); } -inline const mpreal exp10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp10); } -inline const mpreal cos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cos ); } -inline const mpreal sin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sin ); } -inline const mpreal tan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tan ); } -inline const mpreal sec (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sec ); } -inline const mpreal csc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csc ); } -inline const mpreal cot (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cot ); } -inline const mpreal acos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acos ); } -inline const mpreal asin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asin ); } -inline const mpreal atan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atan ); } - -inline const mpreal logb (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { return log2 (abs(x),r); } - -inline const mpreal acot (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atan (1/v, r); } -inline const mpreal asec (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acos (1/v, r); } -inline const mpreal acsc (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asin (1/v, r); } -inline const mpreal acoth (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atanh(1/v, r); } -inline const mpreal asech (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acosh(1/v, r); } -inline const mpreal acsch (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asinh(1/v, r); } - -inline const mpreal cosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cosh ); } -inline const mpreal sinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sinh ); } -inline const mpreal tanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tanh ); } -inline const mpreal sech (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sech ); } -inline const mpreal csch (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csch ); } -inline const mpreal coth (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(coth ); } -inline const mpreal acosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acosh); } -inline const mpreal asinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asinh); } -inline const mpreal atanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atanh); } - -inline const mpreal log1p (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log1p ); } -inline const mpreal expm1 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(expm1 ); } -inline const mpreal eint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(eint ); } -inline const mpreal gamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } -inline const mpreal tgamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } -inline const mpreal lngamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(lngamma); } -inline const mpreal zeta (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(zeta ); } -inline const mpreal erf (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erf ); } -inline const mpreal erfc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erfc ); } -inline const mpreal besselj0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j0 ); } -inline const mpreal besselj1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j1 ); } -inline const mpreal bessely0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y0 ); } -inline const mpreal bessely1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y1 ); } - -inline const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_atan2(a.mpfr_ptr(), y.mpfr_srcptr(), x.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_hypot(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal remainder (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_remainder(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_remquo(a.mpfr_ptr(),q, x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal fac_ui (unsigned long int v, mp_prec_t prec = mpreal::get_default_prec(), - mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(0, prec); - mpfr_fac_ui(x.mpfr_ptr(),v,rnd_mode); - return x; -} - - -inline const mpreal lgamma (const mpreal& v, int *signp = 0, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(v); - int tsignp; - - if(signp) mpfr_lgamma(x.mpfr_ptr(), signp,v.mpfr_srcptr(),rnd_mode); - else mpfr_lgamma(x.mpfr_ptr(),&tsignp,v.mpfr_srcptr(),rnd_mode); - - return x; -} - - -inline const mpreal besseljn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, x.getPrecision()); - mpfr_jn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); - return y; -} - -inline const mpreal besselyn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, x.getPrecision()); - mpfr_yn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); - return y; -} - -inline const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2, p3; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - p3 = v3.get_prec(); - - a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); - - mpfr_fma(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); - return a; -} - -inline const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2, p3; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - p3 = v3.get_prec(); - - a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); - - mpfr_fms(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); - return a; -} - -inline const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - - a.set_prec(p1>p2?p1:p2); - - mpfr_agm(a.mp, v1.mp, v2.mp, rnd_mode); - - return a; -} - -inline const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t mode = mpreal::get_default_rnd()) -{ - mpfr_srcptr *p = new mpfr_srcptr[n]; - - for (unsigned long int i = 0; i < n; i++) - p[i] = tab[i].mpfr_srcptr(); - - mpreal x; - status = mpfr_sum(x.mpfr_ptr(), (mpfr_ptr*)p, n, mode); - - delete [] p; - return x; -} - -////////////////////////////////////////////////////////////////////////// -// MPFR 2.4.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - -inline int sinh_cosh(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - return mpfr_sinh_cosh(s.mp,c.mp,v.mp,rnd_mode); -} - -inline const mpreal li2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - MPREAL_UNARY_MATH_FUNCTION_BODY(li2); -} - -inline const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - /* R = rem(X,Y) if Y != 0, returns X - n * Y where n = trunc(X/Y). */ - return fmod(x, y, rnd_mode); -} - -inline const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - (void)rnd_mode; - - /* - - m = mod(x,y) if y != 0, returns x - n*y where n = floor(x/y) - - The following are true by convention: - - mod(x,0) is x - - mod(x,x) is 0 - - mod(x,y) for x != y and y != 0 has the same sign as y. - - */ - - if(iszero(y)) return x; - if(x == y) return 0; - - mpreal m = x - floor(x / y) * y; - - m.setSign(sgn(y)); // make sure result has the same sign as Y - - return m; -} - -inline const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t yp, xp; - - yp = y.get_prec(); - xp = x.get_prec(); - - a.set_prec(yp>xp?yp:xp); - - mpfr_fmod(a.mp, x.mp, y.mp, rnd_mode); - - return a; -} - -inline const mpreal rec_sqrt(const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(v); - mpfr_rec_sqrt(x.mp,v.mp,rnd_mode); - return x; -} -#endif // MPFR 2.4.0 Specifics - -////////////////////////////////////////////////////////////////////////// -// MPFR 3.0.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) -inline const mpreal digamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(digamma); } -inline const mpreal ai (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(ai); } -#endif // MPFR 3.0.0 Specifics - -////////////////////////////////////////////////////////////////////////// -// Constants -inline const mpreal const_log2 (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_log2(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_pi (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_pi(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_euler (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_euler(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_catalan (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_catalan(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_infinity (int sign = 1, mp_prec_t p = mpreal::get_default_prec()) -{ - mpreal x(0, p); - mpfr_set_inf(x.mpfr_ptr(), sign); - return x; -} - -////////////////////////////////////////////////////////////////////////// -// Integer Related Functions -inline const mpreal ceil(const mpreal& v) -{ - mpreal x(v); - mpfr_ceil(x.mp,v.mp); - return x; -} - -inline const mpreal floor(const mpreal& v) -{ - mpreal x(v); - mpfr_floor(x.mp,v.mp); - return x; -} - -inline const mpreal round(const mpreal& v) -{ - mpreal x(v); - mpfr_round(x.mp,v.mp); - return x; -} - -inline const mpreal trunc(const mpreal& v) -{ - mpreal x(v); - mpfr_trunc(x.mp,v.mp); - return x; -} - -inline const mpreal rint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint ); } -inline const mpreal rint_ceil (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_ceil ); } -inline const mpreal rint_floor (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_floor); } -inline const mpreal rint_round (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_round); } -inline const mpreal rint_trunc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_trunc); } -inline const mpreal frac (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(frac ); } - -////////////////////////////////////////////////////////////////////////// -// Miscellaneous Functions -inline void swap (mpreal& a, mpreal& b) { mpfr_swap(a.mp,b.mp); } -inline const mpreal (max)(const mpreal& x, const mpreal& y){ return (x>y?x:y); } -inline const mpreal (min)(const mpreal& x, const mpreal& y){ return (x= MPFR_VERSION_NUM(3,0,0)) -inline const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x; - mpfr_urandom(x.mpfr_ptr(), state, rnd_mode); - return x; -} -#endif - -#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) -inline const mpreal random2 (mp_size_t size, mp_exp_t exp) -{ - mpreal x; - mpfr_random2(x.mpfr_ptr(),size,exp); - return x; -} -#endif - -// Uniformly distributed random number generation -// a = random(seed); <- initialization & first random number generation -// a = random(); <- next random numbers generation -// seed != 0 -inline const mpreal random(unsigned int seed = 0) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - static gmp_randstate_t state; - static bool initialize = true; - - if(initialize) - { - gmp_randinit_default(state); - gmp_randseed_ui(state,0); - initialize = false; - } - - if(seed != 0) gmp_randseed_ui(state,seed); - - return mpfr::urandom(state); -#else - if(seed != 0) std::srand(seed); - return mpfr::mpreal(std::rand()/(double)RAND_MAX); -#endif - -} - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - -inline const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x; - mpfr_grandom(x.mpfr_ptr(), NULL, state, rnd_mode); - return x; -} - -inline const mpreal grandom(unsigned int seed = 0) -{ - static gmp_randstate_t state; - static bool initialize = true; - - if(initialize) - { - gmp_randinit_default(state); - gmp_randseed_ui(state,0); - initialize = false; - } - - if(seed != 0) gmp_randseed_ui(state,seed); - - return mpfr::grandom(state); -} -#endif - -////////////////////////////////////////////////////////////////////////// -// Set/Get global properties -inline void mpreal::set_default_prec(mp_prec_t prec) -{ - mpfr_set_default_prec(prec); -} - -inline void mpreal::set_default_rnd(mp_rnd_t rnd_mode) -{ - mpfr_set_default_rounding_mode(rnd_mode); -} - -inline bool mpreal::fits_in_bits(double x, int n) -{ - int i; - double t; - return IsInf(x) || (std::modf ( std::ldexp ( std::frexp ( x, &i ), n ), &t ) == 0.0); -} - -inline const mpreal pow(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow(x.mp,x.mp,b.mp,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_z(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_ui(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_si(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); -} - -inline const mpreal pow(const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_ui_pow(x.mp,a,b.mp,rnd_mode); - return x; -} - -inline const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),b,rnd_mode); -} - -inline const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),b,rnd_mode); - else return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),b,rnd_mode); - else return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); -} - -// pow unsigned long int -inline const mpreal pow(const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - mpreal x(a); - mpfr_ui_pow_ui(x.mp,a,b,rnd_mode); - return x; -} - -inline const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -// pow unsigned int -inline const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -// pow long int -inline const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -inline const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -// pow int -inline const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -inline const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -// pow long double -inline const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),mpreal(b),rnd_mode); -} - -inline const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),mpreal(b),rnd_mode); -} - -inline const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_ui -} - -inline const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_ui -} - -inline const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si -} -} // End of mpfr namespace - -// Explicit specialization of std::swap for mpreal numbers -// Thus standard algorithms will use efficient version of swap (due to Koenig lookup) -// Non-throwing swap C++ idiom: http://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Non-throwing_swap -namespace std -{ - // we are allowed to extend namespace std with specializations only - template <> - inline void swap(mpfr::mpreal& x, mpfr::mpreal& y) - { - return mpfr::swap(x, y); - } - - template<> - class numeric_limits - { - public: - static const bool is_specialized = true; - static const bool is_signed = true; - static const bool is_integer = false; - static const bool is_exact = false; - static const int radix = 2; - - static const bool has_infinity = true; - static const bool has_quiet_NaN = true; - static const bool has_signaling_NaN = true; - - static const bool is_iec559 = true; // = IEEE 754 - static const bool is_bounded = true; - static const bool is_modulo = false; - static const bool traps = true; - static const bool tinyness_before = true; - - static const float_denorm_style has_denorm = denorm_absent; - - inline static mpfr::mpreal (min) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::minval(precision); } - inline static mpfr::mpreal (max) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::maxval(precision); } - inline static mpfr::mpreal lowest (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return -mpfr::maxval(precision); } - - // Returns smallest eps such that 1 + eps != 1 (classic machine epsilon) - inline static mpfr::mpreal epsilon(mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::machine_epsilon(precision); } - - // Returns smallest eps such that x + eps != x (relative machine epsilon) - inline static mpfr::mpreal epsilon(const mpfr::mpreal& x) { return mpfr::machine_epsilon(x); } - - inline static mpfr::mpreal round_error(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - mp_rnd_t r = mpfr::mpreal::get_default_rnd(); - - if(r == GMP_RNDN) return mpfr::mpreal(0.5, precision); - else return mpfr::mpreal(1.0, precision); - } - - inline static const mpfr::mpreal infinity() { return mpfr::const_infinity(); } - inline static const mpfr::mpreal quiet_NaN() { return mpfr::mpreal().setNan(); } - inline static const mpfr::mpreal signaling_NaN() { return mpfr::mpreal().setNan(); } - inline static const mpfr::mpreal denorm_min() { return (min)(); } - - // Please note, exponent range is not fixed in MPFR - static const int min_exponent = MPFR_EMIN_DEFAULT; - static const int max_exponent = MPFR_EMAX_DEFAULT; - MPREAL_PERMISSIVE_EXPR static const int min_exponent10 = (int) (MPFR_EMIN_DEFAULT * 0.3010299956639811); - MPREAL_PERMISSIVE_EXPR static const int max_exponent10 = (int) (MPFR_EMAX_DEFAULT * 0.3010299956639811); - -#ifdef MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS - - // Following members should be constant according to standard, but they can be variable in MPFR - // So we define them as functions here. - // - // This is preferable way for std::numeric_limits specialization. - // But it is incompatible with standard std::numeric_limits and might not work with other libraries, e.g. boost. - // See below for compatible implementation. - inline static float_round_style round_style() - { - mp_rnd_t r = mpfr::mpreal::get_default_rnd(); - - switch (r) - { - case GMP_RNDN: return round_to_nearest; - case GMP_RNDZ: return round_toward_zero; - case GMP_RNDU: return round_toward_infinity; - case GMP_RNDD: return round_toward_neg_infinity; - default: return round_indeterminate; - } - } - - inline static int digits() { return int(mpfr::mpreal::get_default_prec()); } - inline static int digits(const mpfr::mpreal& x) { return x.getPrecision(); } - - inline static int digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - return mpfr::bits2digits(precision); - } - - inline static int digits10(const mpfr::mpreal& x) - { - return mpfr::bits2digits(x.getPrecision()); - } - - inline static int max_digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - return digits10(precision); - } -#else - // Digits and round_style are NOT constants when it comes to mpreal. - // If possible, please use functions digits() and round_style() defined above. - // - // These (default) values are preserved for compatibility with existing libraries, e.g. boost. - // Change them accordingly to your application. - // - // For example, if you use 256 bits of precision uniformly in your program, then: - // digits = 256 - // digits10 = 77 - // max_digits10 = 78 - // - // Approximate formula for decimal digits is: digits10 = floor(log10(2) * digits). See bits2digits() for more details. - - static const std::float_round_style round_style = round_to_nearest; - static const int digits = 53; - static const int digits10 = 15; - static const int max_digits10 = 16; -#endif - }; - -} - -#endif /* __MPREAL_H__ */ +/* + MPFR C++: Multi-precision floating point number class for C++. + Based on MPFR library: http://mpfr.org + + Project homepage: http://www.holoborodko.com/pavel/mpfr + Contact e-mail: pavel@holoborodko.com + + Copyright (c) 2008-2015 Pavel Holoborodko + + Contributors: + Dmitriy Gubanov, Konstantin Holoborodko, Brian Gladman, + Helmut Jarausch, Fokko Beekhof, Ulrich Mutze, Heinz van Saanen, + Pere Constans, Peter van Hoof, Gael Guennebaud, Tsai Chia Cheng, + Alexei Zubanov, Jauhien Piatlicki, Victor Berger, John Westwood, + Petr Aleksandrov, Orion Poplawski, Charles Karney, Arash Partow, + Rodney James, Jorge Leitao. + + Licensing: + (A) MPFR C++ is under GNU General Public License ("GPL"). + + (B) Non-free licenses may also be purchased from the author, for users who + do not want their programs protected by the GPL. + + The non-free licenses are for users that wish to use MPFR C++ in + their products but are unwilling to release their software + under the GPL (which would require them to release source code + and allow free redistribution). + + Such users can purchase an unlimited-use license from the author. + Contact us for more details. + + GNU General Public License ("GPL") copyright permissions statement: + ************************************************************************** + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +*/ + +#ifndef __MPREAL_H__ +#define __MPREAL_H__ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +// Options +#define MPREAL_HAVE_MSVC_DEBUGVIEW // Enable Debugger Visualizer for "Debug" builds in MSVC. +#define MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS // Enable extended std::numeric_limits specialization. + // Meaning that "digits", "round_style" and similar members are defined as functions, not constants. + // See std::numeric_limits at the end of the file for more information. + +// Library version +#define MPREAL_VERSION_MAJOR 3 +#define MPREAL_VERSION_MINOR 6 +#define MPREAL_VERSION_PATCHLEVEL 2 +#define MPREAL_VERSION_STRING "3.6.2" + +// Detect compiler using signatures from http://predef.sourceforge.net/ +#if defined(__GNUC__) + #define IsInf(x) (isinf)(x) // GNU C++/Intel ICC compiler on Linux +#elif defined(_MSC_VER) // Microsoft Visual C++ + #define IsInf(x) (!_finite(x)) +#else + #define IsInf(x) (std::isinf)(x) // GNU C/C++ (and/or other compilers), just hope for C99 conformance +#endif + +// A Clang feature extension to determine compiler features. +#ifndef __has_feature + #define __has_feature(x) 0 +#endif + +// Detect support for r-value references (move semantic). Borrowed from Eigen. +#if (__has_feature(cxx_rvalue_references) || \ + defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L || \ + (defined(_MSC_VER) && _MSC_VER >= 1600)) + + #define MPREAL_HAVE_MOVE_SUPPORT + + // Use fields in mpfr_t structure to check if it was initialized / set dummy initialization + #define mpfr_is_initialized(x) (0 != (x)->_mpfr_d) + #define mpfr_set_uninitialized(x) ((x)->_mpfr_d = 0 ) +#endif + +// Detect support for explicit converters. +#if (__has_feature(cxx_explicit_conversions) || \ + (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GNUC_MINOR__ >= 5) || __cplusplus >= 201103L || \ + (defined(_MSC_VER) && _MSC_VER >= 1800)) + + #define MPREAL_HAVE_EXPLICIT_CONVERTERS +#endif + +#define MPFR_USE_INTMAX_T // Enable 64-bit integer types - should be defined before mpfr.h + +#if defined(MPREAL_HAVE_MSVC_DEBUGVIEW) && defined(_MSC_VER) && defined(_DEBUG) + #define MPREAL_MSVC_DEBUGVIEW_CODE DebugView = toString(); + #define MPREAL_MSVC_DEBUGVIEW_DATA std::string DebugView; +#else + #define MPREAL_MSVC_DEBUGVIEW_CODE + #define MPREAL_MSVC_DEBUGVIEW_DATA +#endif + +#include + +#if (MPFR_VERSION < MPFR_VERSION_NUM(3,0,0)) + #include // Needed for random() +#endif + +// Less important options +#define MPREAL_DOUBLE_BITS_OVERFLOW -1 // Triggers overflow exception during conversion to double if mpreal + // cannot fit in MPREAL_DOUBLE_BITS_OVERFLOW bits + // = -1 disables overflow checks (default) + +// Fast replacement for mpfr_set_zero(x, +1): +// (a) uses low-level data members, might not be compatible with new versions of MPFR +// (b) sign is not set, add (x)->_mpfr_sign = 1; +#define mpfr_set_zero_fast(x) ((x)->_mpfr_exp = __MPFR_EXP_ZERO) + +#if defined(__GNUC__) + #define MPREAL_PERMISSIVE_EXPR __extension__ +#else + #define MPREAL_PERMISSIVE_EXPR +#endif + +namespace mpfr { + +class mpreal { +private: + mpfr_t mp; + +public: + + // Get default rounding mode & precision + inline static mp_rnd_t get_default_rnd() { return (mp_rnd_t)(mpfr_get_default_rounding_mode()); } + inline static mp_prec_t get_default_prec() { return mpfr_get_default_prec(); } + + // Constructors && type conversions + mpreal(); + mpreal(const mpreal& u); + mpreal(const mpf_t u); + mpreal(const mpz_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const mpq_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + + // Construct mpreal from mpfr_t structure. + // shared = true allows to avoid deep copy, so that mpreal and 'u' share the same data & pointers. + mpreal(const mpfr_t u, bool shared = false); + + mpreal(const char* s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const std::string& s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); + + ~mpreal(); + +#ifdef MPREAL_HAVE_MOVE_SUPPORT + mpreal& operator=(mpreal&& v); + mpreal(mpreal&& u); +#endif + + // Operations + // = + // +, -, *, /, ++, --, <<, >> + // *=, +=, -=, /=, + // <, >, ==, <=, >= + + // = + mpreal& operator=(const mpreal& v); + mpreal& operator=(const mpf_t v); + mpreal& operator=(const mpz_t v); + mpreal& operator=(const mpq_t v); + mpreal& operator=(const long double v); + mpreal& operator=(const double v); + mpreal& operator=(const unsigned long int v); + mpreal& operator=(const unsigned long long int v); + mpreal& operator=(const long long int v); + mpreal& operator=(const unsigned int v); + mpreal& operator=(const long int v); + mpreal& operator=(const int v); + mpreal& operator=(const char* s); + mpreal& operator=(const std::string& s); + template mpreal& operator= (const std::complex& z); + + // + + mpreal& operator+=(const mpreal& v); + mpreal& operator+=(const mpf_t v); + mpreal& operator+=(const mpz_t v); + mpreal& operator+=(const mpq_t v); + mpreal& operator+=(const long double u); + mpreal& operator+=(const double u); + mpreal& operator+=(const unsigned long int u); + mpreal& operator+=(const unsigned int u); + mpreal& operator+=(const long int u); + mpreal& operator+=(const int u); + + mpreal& operator+=(const long long int u); + mpreal& operator+=(const unsigned long long int u); + mpreal& operator-=(const long long int u); + mpreal& operator-=(const unsigned long long int u); + mpreal& operator*=(const long long int u); + mpreal& operator*=(const unsigned long long int u); + mpreal& operator/=(const long long int u); + mpreal& operator/=(const unsigned long long int u); + + const mpreal operator+() const; + mpreal& operator++ (); + const mpreal operator++ (int); + + // - + mpreal& operator-=(const mpreal& v); + mpreal& operator-=(const mpz_t v); + mpreal& operator-=(const mpq_t v); + mpreal& operator-=(const long double u); + mpreal& operator-=(const double u); + mpreal& operator-=(const unsigned long int u); + mpreal& operator-=(const unsigned int u); + mpreal& operator-=(const long int u); + mpreal& operator-=(const int u); + const mpreal operator-() const; + friend const mpreal operator-(const unsigned long int b, const mpreal& a); + friend const mpreal operator-(const unsigned int b, const mpreal& a); + friend const mpreal operator-(const long int b, const mpreal& a); + friend const mpreal operator-(const int b, const mpreal& a); + friend const mpreal operator-(const double b, const mpreal& a); + mpreal& operator-- (); + const mpreal operator-- (int); + + // * + mpreal& operator*=(const mpreal& v); + mpreal& operator*=(const mpz_t v); + mpreal& operator*=(const mpq_t v); + mpreal& operator*=(const long double v); + mpreal& operator*=(const double v); + mpreal& operator*=(const unsigned long int v); + mpreal& operator*=(const unsigned int v); + mpreal& operator*=(const long int v); + mpreal& operator*=(const int v); + + // / + mpreal& operator/=(const mpreal& v); + mpreal& operator/=(const mpz_t v); + mpreal& operator/=(const mpq_t v); + mpreal& operator/=(const long double v); + mpreal& operator/=(const double v); + mpreal& operator/=(const unsigned long int v); + mpreal& operator/=(const unsigned int v); + mpreal& operator/=(const long int v); + mpreal& operator/=(const int v); + friend const mpreal operator/(const unsigned long int b, const mpreal& a); + friend const mpreal operator/(const unsigned int b, const mpreal& a); + friend const mpreal operator/(const long int b, const mpreal& a); + friend const mpreal operator/(const int b, const mpreal& a); + friend const mpreal operator/(const double b, const mpreal& a); + + //<<= Fast Multiplication by 2^u + mpreal& operator<<=(const unsigned long int u); + mpreal& operator<<=(const unsigned int u); + mpreal& operator<<=(const long int u); + mpreal& operator<<=(const int u); + + //>>= Fast Division by 2^u + mpreal& operator>>=(const unsigned long int u); + mpreal& operator>>=(const unsigned int u); + mpreal& operator>>=(const long int u); + mpreal& operator>>=(const int u); + + // Type Conversion operators + bool toBool ( ) const; + long toLong (mp_rnd_t mode = GMP_RNDZ) const; + unsigned long toULong (mp_rnd_t mode = GMP_RNDZ) const; + long long toLLong (mp_rnd_t mode = GMP_RNDZ) const; + unsigned long long toULLong (mp_rnd_t mode = GMP_RNDZ) const; + float toFloat (mp_rnd_t mode = GMP_RNDN) const; + double toDouble (mp_rnd_t mode = GMP_RNDN) const; + long double toLDouble (mp_rnd_t mode = GMP_RNDN) const; + +#if defined (MPREAL_HAVE_EXPLICIT_CONVERTERS) + explicit operator bool () const { return toBool(); } + explicit operator int () const { return int(toLong()); } + explicit operator long () const { return toLong(); } + explicit operator long long () const { return toLLong(); } + explicit operator unsigned () const { return unsigned(toULong()); } + explicit operator unsigned long () const { return toULong(); } + explicit operator unsigned long long () const { return toULLong(); } + explicit operator float () const { return toFloat(); } + explicit operator double () const { return toDouble(); } + explicit operator long double () const { return toLDouble(); } +#endif + + // Get raw pointers so that mpreal can be directly used in raw mpfr_* functions + ::mpfr_ptr mpfr_ptr(); + ::mpfr_srcptr mpfr_ptr() const; + ::mpfr_srcptr mpfr_srcptr() const; + + // Convert mpreal to string with n significant digits in base b + // n = -1 -> convert with the maximum available digits + std::string toString(int n = -1, int b = 10, mp_rnd_t mode = mpreal::get_default_rnd()) const; + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + std::string toString(const std::string& format) const; +#endif + + std::ostream& output(std::ostream& os) const; + + // Math Functions + friend const mpreal sqr (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sqrt(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sqrt(const unsigned long int v, mp_rnd_t rnd_mode); + friend const mpreal cbrt(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal root(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const long int b, mp_rnd_t rnd_mode); + friend const mpreal pow (const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode); + friend const mpreal pow (const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode); + friend const mpreal fabs(const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal abs(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); + friend inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); + friend inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); + friend int cmpabs(const mpreal& a,const mpreal& b); + + friend const mpreal log (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal logb (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log10(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp10(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log1p(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal expm1(const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal cos(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sin(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tan(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sec(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal csc(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal cot(const mpreal& v, mp_rnd_t rnd_mode); + friend int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal acos (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asin (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atan (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode); + friend const mpreal acot (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asec (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acsc (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal cosh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sinh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tanh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sech (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal csch (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal coth (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acosh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asinh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atanh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acoth (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asech (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acsch (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + + friend const mpreal fac_ui (unsigned long int v, mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal eint (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal gamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tgamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal lngamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal lgamma (const mpreal& v, int *signp, mp_rnd_t rnd_mode); + friend const mpreal zeta (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal erf (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal erfc (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselj0 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselj1 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besseljn (long n, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal bessely0 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal bessely1 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselyn (long n, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); + friend const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); + friend const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode); + friend const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t rnd_mode); + friend int sgn(const mpreal& v); // returns -1 or +1 + +// MPFR 2.4.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + friend int sinh_cosh (mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal li2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal rec_sqrt (const mpreal& v, mp_rnd_t rnd_mode); + + // MATLAB's semantic equivalents + friend const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Remainder after division + friend const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Modulus after division +#endif + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + friend const mpreal digamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal ai (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear +#endif + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + friend const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear + friend const mpreal grandom (unsigned int seed); +#endif + + // Uniformly distributed random number generation in [0,1] using + // Mersenne-Twister algorithm by default. + // Use parameter to setup seed, e.g.: random((unsigned)time(NULL)) + // Check urandom() for more precise control. + friend const mpreal random(unsigned int seed); + + // Splits mpreal value into fractional and integer parts. + // Returns fractional part and stores integer part in n. + friend const mpreal modf(const mpreal& v, mpreal& n); + + // Constants + // don't forget to call mpfr_free_cache() for every thread where you are using const-functions + friend const mpreal const_log2 (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_pi (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_euler (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_catalan (mp_prec_t prec, mp_rnd_t rnd_mode); + + // returns +inf iff sign>=0 otherwise -inf + friend const mpreal const_infinity(int sign, mp_prec_t prec); + + // Output/ Input + friend std::ostream& operator<<(std::ostream& os, const mpreal& v); + friend std::istream& operator>>(std::istream& is, mpreal& v); + + // Integer Related Functions + friend const mpreal rint (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal ceil (const mpreal& v); + friend const mpreal floor(const mpreal& v); + friend const mpreal round(const mpreal& v); + friend const mpreal trunc(const mpreal& v); + friend const mpreal rint_ceil (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_floor (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_round (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_trunc (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal frac (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal remainder ( const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + + // Miscellaneous Functions + friend const mpreal nexttoward (const mpreal& x, const mpreal& y); + friend const mpreal nextabove (const mpreal& x); + friend const mpreal nextbelow (const mpreal& x); + + // use gmp_randinit_default() to init state, gmp_randclear() to clear + friend const mpreal urandomb (gmp_randstate_t& state); + +// MPFR < 2.4.2 Specifics +#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) + friend const mpreal random2 (mp_size_t size, mp_exp_t exp); +#endif + + // Instance Checkers + friend bool (isnan) (const mpreal& v); + friend bool (isinf) (const mpreal& v); + friend bool (isfinite) (const mpreal& v); + + friend bool isnum (const mpreal& v); + friend bool iszero (const mpreal& v); + friend bool isint (const mpreal& v); + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + friend bool isregular(const mpreal& v); +#endif + + // Set/Get instance properties + inline mp_prec_t get_prec() const; + inline void set_prec(mp_prec_t prec, mp_rnd_t rnd_mode = get_default_rnd()); // Change precision with rounding mode + + // Aliases for get_prec(), set_prec() - needed for compatibility with std::complex interface + inline mpreal& setPrecision(int Precision, mp_rnd_t RoundingMode = get_default_rnd()); + inline int getPrecision() const; + + // Set mpreal to +/- inf, NaN, +/-0 + mpreal& setInf (int Sign = +1); + mpreal& setNan (); + mpreal& setZero (int Sign = +1); + mpreal& setSign (int Sign, mp_rnd_t RoundingMode = get_default_rnd()); + + //Exponent + mp_exp_t get_exp(); + int set_exp(mp_exp_t e); + int check_range (int t, mp_rnd_t rnd_mode = get_default_rnd()); + int subnormalize (int t, mp_rnd_t rnd_mode = get_default_rnd()); + + // Inexact conversion from float + inline bool fits_in_bits(double x, int n); + + // Set/Get global properties + static void set_default_prec(mp_prec_t prec); + static void set_default_rnd(mp_rnd_t rnd_mode); + + static mp_exp_t get_emin (void); + static mp_exp_t get_emax (void); + static mp_exp_t get_emin_min (void); + static mp_exp_t get_emin_max (void); + static mp_exp_t get_emax_min (void); + static mp_exp_t get_emax_max (void); + static int set_emin (mp_exp_t exp); + static int set_emax (mp_exp_t exp); + + // Efficient swapping of two mpreal values - needed for std algorithms + friend void swap(mpreal& x, mpreal& y); + + friend const mpreal fmax(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal fmin(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + +private: + // Human friendly Debug Preview in Visual Studio. + // Put one of these lines: + // + // mpfr::mpreal= ; Show value only + // mpfr::mpreal=, bits ; Show value & precision + // + // at the beginning of + // [Visual Studio Installation Folder]\Common7\Packages\Debugger\autoexp.dat + MPREAL_MSVC_DEBUGVIEW_DATA + + // "Smart" resources deallocation. Checks if instance initialized before deletion. + void clear(::mpfr_ptr); +}; + +////////////////////////////////////////////////////////////////////////// +// Exceptions +class conversion_overflow : public std::exception { +public: + std::string why() { return "inexact conversion from floating point"; } +}; + +////////////////////////////////////////////////////////////////////////// +// Constructors & converters +// Default constructor: creates mp number and initializes it to 0. +inline mpreal::mpreal() +{ + mpfr_init2(mpfr_ptr(), mpreal::get_default_prec()); + mpfr_set_zero_fast(mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpreal& u) +{ + mpfr_init2(mpfr_ptr(),mpfr_get_prec(u.mpfr_srcptr())); + mpfr_set (mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +#ifdef MPREAL_HAVE_MOVE_SUPPORT +inline mpreal::mpreal(mpreal&& other) +{ + mpfr_set_uninitialized(mpfr_ptr()); // make sure "other" holds no pointer to actual data + mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal& mpreal::operator=(mpreal&& other) +{ + mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} +#endif + +inline mpreal::mpreal(const mpfr_t u, bool shared) +{ + if(shared) + { + std::memcpy(mpfr_ptr(), u, sizeof(mpfr_t)); + } + else + { + mpfr_init2(mpfr_ptr(), mpfr_get_prec(u)); + mpfr_set (mpfr_ptr(), u, mpreal::get_default_rnd()); + } + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpf_t u) +{ + mpfr_init2(mpfr_ptr(),(mp_prec_t) mpf_get_prec(u)); // (gmp: mp_bitcnt_t) unsigned long -> long (mpfr: mp_prec_t) + mpfr_set_f(mpfr_ptr(),u,mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpz_t u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + mpfr_set_z(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpq_t u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + mpfr_set_q(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const double u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + +#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) + if(fits_in_bits(u, MPREAL_DOUBLE_BITS_OVERFLOW)) + { + mpfr_set_d(mpfr_ptr(), u, mode); + }else + throw conversion_overflow(); +#else + mpfr_set_d(mpfr_ptr(), u, mode); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long double u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ld(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned long long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_uj(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_sj(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ui(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ui(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_si(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_si(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_str(mpfr_ptr(), s, base, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const std::string& s, mp_prec_t prec, int base, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_str(mpfr_ptr(), s.c_str(), base, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline void mpreal::clear(::mpfr_ptr x) +{ +#ifdef MPREAL_HAVE_MOVE_SUPPORT + if(mpfr_is_initialized(x)) +#endif + mpfr_clear(x); +} + +inline mpreal::~mpreal() +{ + clear(mpfr_ptr()); +} + +// internal namespace needed for template magic +namespace internal{ + + // Use SFINAE to restrict arithmetic operations instantiation only for numeric types + // This is needed for smooth integration with libraries based on expression templates, like Eigen. + // TODO: Do the same for boolean operators. + template struct result_type {}; + + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; +} + +// + Addition +template +inline const typename internal::result_type::type + operator+(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) += rhs; } + +template +inline const typename internal::result_type::type + operator+(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) += lhs; } + +// - Subtraction +template +inline const typename internal::result_type::type + operator-(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) -= rhs; } + +template +inline const typename internal::result_type::type + operator-(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) -= rhs; } + +// * Multiplication +template +inline const typename internal::result_type::type + operator*(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) *= rhs; } + +template +inline const typename internal::result_type::type + operator*(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) *= lhs; } + +// / Division +template +inline const typename internal::result_type::type + operator/(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) /= rhs; } + +template +inline const typename internal::result_type::type + operator/(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) /= rhs; } + +////////////////////////////////////////////////////////////////////////// +// sqrt +const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const long int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const long double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +// abs +inline const mpreal abs(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()); + +////////////////////////////////////////////////////////////////////////// +// pow +const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +////////////////////////////////////////////////////////////////////////// +// Estimate machine epsilon for the given precision +// Returns smallest eps such that 1.0 + eps != 1.0 +inline mpreal machine_epsilon(mp_prec_t prec = mpreal::get_default_prec()); + +// Returns smallest eps such that x + eps != x (relative machine epsilon) +inline mpreal machine_epsilon(const mpreal& x); + +// Gives max & min values for the required precision, +// minval is 'safe' meaning 1 / minval does not overflow +// maxval is 'safe' meaning 1 / maxval does not underflow +inline mpreal minval(mp_prec_t prec = mpreal::get_default_prec()); +inline mpreal maxval(mp_prec_t prec = mpreal::get_default_prec()); + +// 'Dirty' equality check 1: |a-b| < min{|a|,|b|} * eps +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps); + +// 'Dirty' equality check 2: |a-b| < min{|a|,|b|} * eps( min{|a|,|b|} ) +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b); + +// 'Bitwise' equality check +// maxUlps - a and b can be apart by maxUlps binary numbers. +inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps); + +////////////////////////////////////////////////////////////////////////// +// Convert precision in 'bits' to decimal digits and vice versa. +// bits = ceil(digits*log[2](10)) +// digits = floor(bits*log[10](2)) + +inline mp_prec_t digits2bits(int d); +inline int bits2digits(mp_prec_t b); + +////////////////////////////////////////////////////////////////////////// +// min, max +const mpreal (max)(const mpreal& x, const mpreal& y); +const mpreal (min)(const mpreal& x, const mpreal& y); + +////////////////////////////////////////////////////////////////////////// +// Implementation +////////////////////////////////////////////////////////////////////////// + +////////////////////////////////////////////////////////////////////////// +// Operators - Assignment +inline mpreal& mpreal::operator=(const mpreal& v) +{ + if (this != &v) + { + mp_prec_t tp = mpfr_get_prec( mpfr_srcptr()); + mp_prec_t vp = mpfr_get_prec(v.mpfr_srcptr()); + + if(tp != vp){ + clear(mpfr_ptr()); + mpfr_init2(mpfr_ptr(), vp); + } + + mpfr_set(mpfr_ptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + } + return *this; +} + +inline mpreal& mpreal::operator=(const mpf_t v) +{ + mpfr_set_f(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const mpz_t v) +{ + mpfr_set_z(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const mpq_t v) +{ + mpfr_set_q(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long double v) +{ + mpfr_set_ld(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const double v) +{ +#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) + if(fits_in_bits(v, MPREAL_DOUBLE_BITS_OVERFLOW)) + { + mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); + }else + throw conversion_overflow(); +#else + mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned long int v) +{ + mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned int v) +{ + mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned long long int v) +{ + mpfr_set_uj(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long long int v) +{ + mpfr_set_sj(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long int v) +{ + mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const int v) +{ + mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const char* s) +{ + // Use other converters for more precise control on base & precision & rounding: + // + // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) + // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) + // + // Here we assume base = 10 and we use precision of target variable. + + mpfr_t t; + + mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); + + if(0 == mpfr_set_str(t, s, 10, mpreal::get_default_rnd())) + { + mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + } + + clear(t); + return *this; +} + +inline mpreal& mpreal::operator=(const std::string& s) +{ + // Use other converters for more precise control on base & precision & rounding: + // + // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) + // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) + // + // Here we assume base = 10 and we use precision of target variable. + + mpfr_t t; + + mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); + + if(0 == mpfr_set_str(t, s.c_str(), 10, mpreal::get_default_rnd())) + { + mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + } + + clear(t); + return *this; +} + +template +inline mpreal& mpreal::operator= (const std::complex& z) +{ + return *this = z.real(); +} + +////////////////////////////////////////////////////////////////////////// +// + Addition +inline mpreal& mpreal::operator+=(const mpreal& v) +{ + mpfr_add(mpfr_ptr(), mpfr_srcptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpf_t u) +{ + *this += mpreal(u); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpz_t u) +{ + mpfr_add_z(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpq_t u) +{ + mpfr_add_q(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+= (const long double u) +{ + *this += mpreal(u); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+= (const double u) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_add_d(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); +#else + *this += mpreal(u); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const unsigned long int u) +{ + mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const unsigned int u) +{ + mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const long int u) +{ + mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const int u) +{ + mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const long long int u) { *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator+=(const unsigned long long int u){ *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator-=(const long long int u) { *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator-=(const unsigned long long int u){ *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator*=(const long long int u) { *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator*=(const unsigned long long int u){ *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator/=(const long long int u) { *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator/=(const unsigned long long int u){ *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } + +inline const mpreal mpreal::operator+()const { return mpreal(*this); } + +inline const mpreal operator+(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_add(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline mpreal& mpreal::operator++() +{ + return *this += 1; +} + +inline const mpreal mpreal::operator++ (int) +{ + mpreal x(*this); + *this += 1; + return x; +} + +inline mpreal& mpreal::operator--() +{ + return *this -= 1; +} + +inline const mpreal mpreal::operator-- (int) +{ + mpreal x(*this); + *this -= 1; + return x; +} + +////////////////////////////////////////////////////////////////////////// +// - Subtraction +inline mpreal& mpreal::operator-=(const mpreal& v) +{ + mpfr_sub(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const mpz_t v) +{ + mpfr_sub_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const mpq_t v) +{ + mpfr_sub_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const long double v) +{ + *this -= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_sub_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this -= mpreal(v); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const unsigned long int v) +{ + mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const unsigned int v) +{ + mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const long int v) +{ + mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const int v) +{ + mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal mpreal::operator-()const +{ + mpreal u(*this); + mpfr_neg(u.mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); + return u; +} + +inline const mpreal operator-(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_sub(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline const mpreal operator-(const double b, const mpreal& a) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_d_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +#else + mpreal x(b, mpfr_get_prec(a.mpfr_ptr())); + x -= a; + return x; +#endif +} + +inline const mpreal operator-(const unsigned long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const unsigned int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +////////////////////////////////////////////////////////////////////////// +// * Multiplication +inline mpreal& mpreal::operator*= (const mpreal& v) +{ + mpfr_mul(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const mpz_t v) +{ + mpfr_mul_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const mpq_t v) +{ + mpfr_mul_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const long double v) +{ + *this *= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_mul_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this *= mpreal(v); +#endif + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const unsigned long int v) +{ + mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const unsigned int v) +{ + mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const long int v) +{ + mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const int v) +{ + mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator*(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_mul(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +////////////////////////////////////////////////////////////////////////// +// / Division +inline mpreal& mpreal::operator/=(const mpreal& v) +{ + mpfr_div(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const mpz_t v) +{ + mpfr_div_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const mpq_t v) +{ + mpfr_div_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const long double v) +{ + *this /= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_div_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this /= mpreal(v); +#endif + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const unsigned long int v) +{ + mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const unsigned int v) +{ + mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const long int v) +{ + mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const int v) +{ + mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator/(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_srcptr()), mpfr_get_prec(b.mpfr_srcptr()))); + mpfr_div(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline const mpreal operator/(const unsigned long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const unsigned int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const double b, const mpreal& a) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_d_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +#else + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + x /= a; + return x; +#endif +} + +////////////////////////////////////////////////////////////////////////// +// Shifts operators - Multiplication/Division by power of 2 +inline mpreal& mpreal::operator<<=(const unsigned long int u) +{ + mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const unsigned int u) +{ + mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const long int u) +{ + mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const int u) +{ + mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const unsigned long int u) +{ + mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const unsigned int u) +{ + mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const long int u) +{ + mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const int u) +{ + mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator<<(const mpreal& v, const unsigned long int k) +{ + return mul_2ui(v,k); +} + +inline const mpreal operator<<(const mpreal& v, const unsigned int k) +{ + return mul_2ui(v,static_cast(k)); +} + +inline const mpreal operator<<(const mpreal& v, const long int k) +{ + return mul_2si(v,k); +} + +inline const mpreal operator<<(const mpreal& v, const int k) +{ + return mul_2si(v,static_cast(k)); +} + +inline const mpreal operator>>(const mpreal& v, const unsigned long int k) +{ + return div_2ui(v,k); +} + +inline const mpreal operator>>(const mpreal& v, const long int k) +{ + return div_2si(v,k); +} + +inline const mpreal operator>>(const mpreal& v, const unsigned int k) +{ + return div_2ui(v,static_cast(k)); +} + +inline const mpreal operator>>(const mpreal& v, const int k) +{ + return div_2si(v,static_cast(k)); +} + +// mul_2ui +inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_mul_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +// mul_2si +inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_mul_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_div_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_div_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +////////////////////////////////////////////////////////////////////////// +//Relational operators + +// WARNING: +// +// Please note that following checks for double-NaN are guaranteed to work only in IEEE math mode: +// +// isnan(b) = (b != b) +// isnan(b) = !(b == b) (we use in code below) +// +// Be cautions if you use compiler options which break strict IEEE compliance (e.g. -ffast-math in GCC). +// Use std::isnan instead (C++11). + +inline bool operator > (const mpreal& a, const mpreal& b ){ return (mpfr_greater_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator > (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) > 0 ); } + +inline bool operator >= (const mpreal& a, const mpreal& b ){ return (mpfr_greaterequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator >= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } +// inline bool operator >= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (isnan()a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) >= 0 ); } + +inline bool operator < (const mpreal& a, const mpreal& b ){ return (mpfr_less_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator < (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) < 0 ); } + +inline bool operator <= (const mpreal& a, const mpreal& b ){ return (mpfr_lessequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator <= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) <= 0 ); } + +inline bool operator == (const mpreal& a, const mpreal& b ){ return (mpfr_equal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator == (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) == 0 ); } + +inline bool operator != (const mpreal& a, const mpreal& b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const unsigned long int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const unsigned int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const long int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const long double b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const double b ){ return !(a == b); } + +inline bool (isnan) (const mpreal& op){ return (mpfr_nan_p (op.mpfr_srcptr()) != 0 ); } +inline bool (isinf) (const mpreal& op){ return (mpfr_inf_p (op.mpfr_srcptr()) != 0 ); } +inline bool (isfinite) (const mpreal& op){ return (mpfr_number_p (op.mpfr_srcptr()) != 0 ); } +inline bool iszero (const mpreal& op){ return (mpfr_zero_p (op.mpfr_srcptr()) != 0 ); } +inline bool isint (const mpreal& op){ return (mpfr_integer_p(op.mpfr_srcptr()) != 0 ); } + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) +inline bool isregular(const mpreal& op){ return (mpfr_regular_p(op.mpfr_srcptr()));} +#endif + +////////////////////////////////////////////////////////////////////////// +// Type Converters +inline bool mpreal::toBool ( ) const { return mpfr_zero_p (mpfr_srcptr()) == 0; } +inline long mpreal::toLong (mp_rnd_t mode) const { return mpfr_get_si (mpfr_srcptr(), mode); } +inline unsigned long mpreal::toULong (mp_rnd_t mode) const { return mpfr_get_ui (mpfr_srcptr(), mode); } +inline float mpreal::toFloat (mp_rnd_t mode) const { return mpfr_get_flt(mpfr_srcptr(), mode); } +inline double mpreal::toDouble (mp_rnd_t mode) const { return mpfr_get_d (mpfr_srcptr(), mode); } +inline long double mpreal::toLDouble(mp_rnd_t mode) const { return mpfr_get_ld (mpfr_srcptr(), mode); } +inline long long mpreal::toLLong (mp_rnd_t mode) const { return mpfr_get_sj (mpfr_srcptr(), mode); } +inline unsigned long long mpreal::toULLong (mp_rnd_t mode) const { return mpfr_get_uj (mpfr_srcptr(), mode); } + +inline ::mpfr_ptr mpreal::mpfr_ptr() { return mp; } +inline ::mpfr_srcptr mpreal::mpfr_ptr() const { return mp; } +inline ::mpfr_srcptr mpreal::mpfr_srcptr() const { return mp; } + +template +inline std::string toString(T t, std::ios_base & (*f)(std::ios_base&)) +{ + std::ostringstream oss; + oss << f << t; + return oss.str(); +} + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + +inline std::string mpreal::toString(const std::string& format) const +{ + char *s = NULL; + std::string out; + + if( !format.empty() ) + { + if(!(mpfr_asprintf(&s, format.c_str(), mpfr_srcptr()) < 0)) + { + out = std::string(s); + + mpfr_free_str(s); + } + } + + return out; +} + +#endif + +inline std::string mpreal::toString(int n, int b, mp_rnd_t mode) const +{ + // TODO: Add extended format specification (f, e, rounding mode) as it done in output operator + (void)b; + (void)mode; + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + + std::ostringstream format; + + int digits = (n >= 0) ? n : 1 + bits2digits(mpfr_get_prec(mpfr_srcptr())); + + format << "%." << digits << "RNg"; + + return toString(format.str()); + +#else + + char *s, *ns = NULL; + size_t slen, nslen; + mp_exp_t exp; + std::string out; + + if(mpfr_inf_p(mp)) + { + if(mpfr_sgn(mp)>0) return "+Inf"; + else return "-Inf"; + } + + if(mpfr_zero_p(mp)) return "0"; + if(mpfr_nan_p(mp)) return "NaN"; + + s = mpfr_get_str(NULL, &exp, b, 0, mp, mode); + ns = mpfr_get_str(NULL, &exp, b, (std::max)(0,n), mp, mode); + + if(s!=NULL && ns!=NULL) + { + slen = strlen(s); + nslen = strlen(ns); + if(nslen<=slen) + { + mpfr_free_str(s); + s = ns; + slen = nslen; + } + else { + mpfr_free_str(ns); + } + + // Make human eye-friendly formatting if possible + if (exp>0 && static_cast(exp)s+exp) ptr--; + + if(ptr==s+exp) out = std::string(s,exp+1); + else out = std::string(s,exp+1)+'.'+std::string(s+exp+1,ptr-(s+exp+1)+1); + + //out = string(s,exp+1)+'.'+string(s+exp+1); + } + else + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s+exp-1) ptr--; + + if(ptr==s+exp-1) out = std::string(s,exp); + else out = std::string(s,exp)+'.'+std::string(s+exp,ptr-(s+exp)+1); + + //out = string(s,exp)+'.'+string(s+exp); + } + + }else{ // exp<0 || exp>slen + if(s[0]=='-') + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s+1) ptr--; + + if(ptr==s+1) out = std::string(s,2); + else out = std::string(s,2)+'.'+std::string(s+2,ptr-(s+2)+1); + + //out = string(s,2)+'.'+string(s+2); + } + else + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s) ptr--; + + if(ptr==s) out = std::string(s,1); + else out = std::string(s,1)+'.'+std::string(s+1,ptr-(s+1)+1); + + //out = string(s,1)+'.'+string(s+1); + } + + // Make final string + if(--exp) + { + if(exp>0) out += "e+"+mpfr::toString(exp,std::dec); + else out += "e"+mpfr::toString(exp,std::dec); + } + } + + mpfr_free_str(s); + return out; + }else{ + return "conversion error!"; + } +#endif +} + + +////////////////////////////////////////////////////////////////////////// +// I/O +inline std::ostream& mpreal::output(std::ostream& os) const +{ + std::ostringstream format; + const std::ios::fmtflags flags = os.flags(); + + format << ((flags & std::ios::showpos) ? "%+" : "%"); + if (os.precision() >= 0) + format << '.' << os.precision() << "R*" + << ((flags & std::ios::floatfield) == std::ios::fixed ? 'f' : + (flags & std::ios::floatfield) == std::ios::scientific ? 'e' : + 'g'); + else + format << "R*e"; + + char *s = NULL; + if(!(mpfr_asprintf(&s, format.str().c_str(), + mpfr::mpreal::get_default_rnd(), + mpfr_srcptr()) + < 0)) + { + os << std::string(s); + mpfr_free_str(s); + } + return os; +} + +inline std::ostream& operator<<(std::ostream& os, const mpreal& v) +{ + return v.output(os); +} + +inline std::istream& operator>>(std::istream &is, mpreal& v) +{ + // TODO: use cout::hexfloat and other flags to setup base + std::string tmp; + is >> tmp; + mpfr_set_str(v.mpfr_ptr(), tmp.c_str(), 10, mpreal::get_default_rnd()); + return is; +} + +////////////////////////////////////////////////////////////////////////// +// Bits - decimal digits relation +// bits = ceil(digits*log[2](10)) +// digits = floor(bits*log[10](2)) + +inline mp_prec_t digits2bits(int d) +{ + const double LOG2_10 = 3.3219280948873624; + + return mp_prec_t(std::ceil( d * LOG2_10 )); +} + +inline int bits2digits(mp_prec_t b) +{ + const double LOG10_2 = 0.30102999566398119; + + return int(std::floor( b * LOG10_2 )); +} + +////////////////////////////////////////////////////////////////////////// +// Set/Get number properties +inline int sgn(const mpreal& op) +{ + return mpfr_sgn(op.mpfr_srcptr()); +} + +inline mpreal& mpreal::setSign(int sign, mp_rnd_t RoundingMode) +{ + mpfr_setsign(mpfr_ptr(), mpfr_srcptr(), (sign < 0 ? 1 : 0), RoundingMode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline int mpreal::getPrecision() const +{ + return int(mpfr_get_prec(mpfr_srcptr())); +} + +inline mpreal& mpreal::setPrecision(int Precision, mp_rnd_t RoundingMode) +{ + mpfr_prec_round(mpfr_ptr(), Precision, RoundingMode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setInf(int sign) +{ + mpfr_set_inf(mpfr_ptr(), sign); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setNan() +{ + mpfr_set_nan(mpfr_ptr()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setZero(int sign) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + mpfr_set_zero(mpfr_ptr(), sign); +#else + mpfr_set_si(mpfr_ptr(), 0, (mpfr_get_default_rounding_mode)()); + setSign(sign); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mp_prec_t mpreal::get_prec() const +{ + return mpfr_get_prec(mpfr_srcptr()); +} + +inline void mpreal::set_prec(mp_prec_t prec, mp_rnd_t rnd_mode) +{ + mpfr_prec_round(mpfr_ptr(),prec,rnd_mode); + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mp_exp_t mpreal::get_exp () +{ + return mpfr_get_exp(mpfr_srcptr()); +} + +inline int mpreal::set_exp (mp_exp_t e) +{ + int x = mpfr_set_exp(mpfr_ptr(), e); + MPREAL_MSVC_DEBUGVIEW_CODE; + return x; +} + +inline const mpreal frexp(const mpreal& x, mp_exp_t* exp, mp_rnd_t mode = mpreal::get_default_rnd()) +{ + mpreal y(x); +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + mpfr_frexp(exp,y.mpfr_ptr(),x.mpfr_srcptr(),mode); +#else + *exp = mpfr_get_exp(y.mpfr_srcptr()); + mpfr_set_exp(y.mpfr_ptr(),0); +#endif + return y; +} + +inline const mpreal ldexp(const mpreal& v, mp_exp_t exp) +{ + mpreal x(v); + + // rounding is not important since we are just increasing the exponent (= exact operation) + mpfr_mul_2si(x.mpfr_ptr(), x.mpfr_srcptr(), exp, mpreal::get_default_rnd()); + return x; +} + +inline const mpreal scalbn(const mpreal& v, mp_exp_t exp) +{ + return ldexp(v, exp); +} + +inline mpreal machine_epsilon(mp_prec_t prec) +{ + /* the smallest eps such that 1 + eps != 1 */ + return machine_epsilon(mpreal(1, prec)); +} + +inline mpreal machine_epsilon(const mpreal& x) +{ + /* the smallest eps such that x + eps != x */ + if( x < 0) + { + return nextabove(-x) + x; + }else{ + return nextabove( x) - x; + } +} + +// minval is 'safe' meaning 1 / minval does not overflow +inline mpreal minval(mp_prec_t prec) +{ + /* min = 1/2 * 2^emin = 2^(emin - 1) */ + return mpreal(1, prec) << mpreal::get_emin()-1; +} + +// maxval is 'safe' meaning 1 / maxval does not underflow +inline mpreal maxval(mp_prec_t prec) +{ + /* max = (1 - eps) * 2^emax, eps is machine epsilon */ + return (mpreal(1, prec) - machine_epsilon(prec)) << mpreal::get_emax(); +} + +inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps) +{ + return abs(a - b) <= machine_epsilon((max)(abs(a), abs(b))) * maxUlps; +} + +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps) +{ + return abs(a - b) <= eps; +} + +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b) +{ + return isEqualFuzzy(a, b, machine_epsilon((max)(1, (min)(abs(a), abs(b))))); +} + +////////////////////////////////////////////////////////////////////////// +// C++11 sign functions. +inline mpreal copysign(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal rop(0, mpfr_get_prec(x.mpfr_ptr())); + mpfr_setsign(rop.mpfr_ptr(), x.mpfr_srcptr(), mpfr_signbit(y.mpfr_srcptr()), rnd_mode); + return rop; +} + +inline bool signbit(const mpreal& x) +{ + return mpfr_signbit(x.mpfr_srcptr()); +} + +inline const mpreal modf(const mpreal& v, mpreal& n) +{ + mpreal f(v); + + // rounding is not important since we are using the same number + mpfr_frac (f.mpfr_ptr(),f.mpfr_srcptr(),mpreal::get_default_rnd()); + mpfr_trunc(n.mpfr_ptr(),v.mpfr_srcptr()); + return f; +} + +inline int mpreal::check_range (int t, mp_rnd_t rnd_mode) +{ + return mpfr_check_range(mpfr_ptr(),t,rnd_mode); +} + +inline int mpreal::subnormalize (int t,mp_rnd_t rnd_mode) +{ + int r = mpfr_subnormalize(mpfr_ptr(),t,rnd_mode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return r; +} + +inline mp_exp_t mpreal::get_emin (void) +{ + return mpfr_get_emin(); +} + +inline int mpreal::set_emin (mp_exp_t exp) +{ + return mpfr_set_emin(exp); +} + +inline mp_exp_t mpreal::get_emax (void) +{ + return mpfr_get_emax(); +} + +inline int mpreal::set_emax (mp_exp_t exp) +{ + return mpfr_set_emax(exp); +} + +inline mp_exp_t mpreal::get_emin_min (void) +{ + return mpfr_get_emin_min(); +} + +inline mp_exp_t mpreal::get_emin_max (void) +{ + return mpfr_get_emin_max(); +} + +inline mp_exp_t mpreal::get_emax_min (void) +{ + return mpfr_get_emax_min(); +} + +inline mp_exp_t mpreal::get_emax_max (void) +{ + return mpfr_get_emax_max(); +} + +////////////////////////////////////////////////////////////////////////// +// Mathematical Functions +////////////////////////////////////////////////////////////////////////// +#define MPREAL_UNARY_MATH_FUNCTION_BODY(f) \ + mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); \ + mpfr_##f(y.mpfr_ptr(), x.mpfr_srcptr(), r); \ + return y; + +inline const mpreal sqr (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqr ); } + +inline const mpreal sqrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqrt); } + +inline const mpreal sqrt(const unsigned long int x, mp_rnd_t r) +{ + mpreal y; + mpfr_sqrt_ui(y.mpfr_ptr(), x, r); + return y; +} + +inline const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode) +{ + return sqrt(static_cast(v),rnd_mode); +} + +inline const mpreal sqrt(const long int v, mp_rnd_t rnd_mode) +{ + if (v>=0) return sqrt(static_cast(v),rnd_mode); + else return mpreal().setNan(); // NaN +} + +inline const mpreal sqrt(const int v, mp_rnd_t rnd_mode) +{ + if (v>=0) return sqrt(static_cast(v),rnd_mode); + else return mpreal().setNan(); // NaN +} + +inline const mpreal root(const mpreal& x, unsigned long int k, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); + mpfr_root(y.mpfr_ptr(), x.mpfr_srcptr(), k, r); + return y; +} + +inline const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_dim(y.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), r); + return y; +} + +inline int cmpabs(const mpreal& a,const mpreal& b) +{ + return mpfr_cmpabs(a.mpfr_ptr(), b.mpfr_srcptr()); +} + +inline int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + return mpfr_sin_cos(s.mpfr_ptr(), c.mpfr_ptr(), v.mpfr_srcptr(), rnd_mode); +} + +inline const mpreal sqrt (const long double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } +inline const mpreal sqrt (const double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } + +inline const mpreal cbrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cbrt ); } +inline const mpreal fabs (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } +inline const mpreal abs (const mpreal& x, mp_rnd_t r) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } +inline const mpreal log (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log ); } +inline const mpreal log2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log2 ); } +inline const mpreal log10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log10); } +inline const mpreal exp (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp ); } +inline const mpreal exp2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp2 ); } +inline const mpreal exp10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp10); } +inline const mpreal cos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cos ); } +inline const mpreal sin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sin ); } +inline const mpreal tan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tan ); } +inline const mpreal sec (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sec ); } +inline const mpreal csc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csc ); } +inline const mpreal cot (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cot ); } +inline const mpreal acos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acos ); } +inline const mpreal asin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asin ); } +inline const mpreal atan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atan ); } + +inline const mpreal logb (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { return log2 (abs(x),r); } + +inline const mpreal acot (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atan (1/v, r); } +inline const mpreal asec (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acos (1/v, r); } +inline const mpreal acsc (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asin (1/v, r); } +inline const mpreal acoth (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atanh(1/v, r); } +inline const mpreal asech (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acosh(1/v, r); } +inline const mpreal acsch (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asinh(1/v, r); } + +inline const mpreal cosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cosh ); } +inline const mpreal sinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sinh ); } +inline const mpreal tanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tanh ); } +inline const mpreal sech (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sech ); } +inline const mpreal csch (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csch ); } +inline const mpreal coth (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(coth ); } +inline const mpreal acosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acosh); } +inline const mpreal asinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asinh); } +inline const mpreal atanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atanh); } + +inline const mpreal log1p (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log1p ); } +inline const mpreal expm1 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(expm1 ); } +inline const mpreal eint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(eint ); } +inline const mpreal gamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } +inline const mpreal tgamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } +inline const mpreal lngamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(lngamma); } +inline const mpreal zeta (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(zeta ); } +inline const mpreal erf (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erf ); } +inline const mpreal erfc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erfc ); } +inline const mpreal besselj0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j0 ); } +inline const mpreal besselj1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j1 ); } +inline const mpreal bessely0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y0 ); } +inline const mpreal bessely1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y1 ); } + +inline const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_atan2(a.mpfr_ptr(), y.mpfr_srcptr(), x.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_hypot(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal remainder (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_remainder(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_remquo(a.mpfr_ptr(),q, x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal fac_ui (unsigned long int v, mp_prec_t prec = mpreal::get_default_prec(), + mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(0, prec); + mpfr_fac_ui(x.mpfr_ptr(),v,rnd_mode); + return x; +} + + +inline const mpreal lgamma (const mpreal& v, int *signp = 0, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(v); + int tsignp; + + if(signp) mpfr_lgamma(x.mpfr_ptr(), signp,v.mpfr_srcptr(),rnd_mode); + else mpfr_lgamma(x.mpfr_ptr(),&tsignp,v.mpfr_srcptr(),rnd_mode); + + return x; +} + + +inline const mpreal besseljn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, x.getPrecision()); + mpfr_jn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); + return y; +} + +inline const mpreal besselyn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, x.getPrecision()); + mpfr_yn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); + return y; +} + +inline const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2, p3; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + p3 = v3.get_prec(); + + a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); + + mpfr_fma(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); + return a; +} + +inline const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2, p3; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + p3 = v3.get_prec(); + + a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); + + mpfr_fms(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); + return a; +} + +inline const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + + a.set_prec(p1>p2?p1:p2); + + mpfr_agm(a.mp, v1.mp, v2.mp, rnd_mode); + + return a; +} + +inline const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t mode = mpreal::get_default_rnd()) +{ + mpfr_srcptr *p = new mpfr_srcptr[n]; + + for (unsigned long int i = 0; i < n; i++) + p[i] = tab[i].mpfr_srcptr(); + + mpreal x; + status = mpfr_sum(x.mpfr_ptr(), (mpfr_ptr*)p, n, mode); + + delete [] p; + return x; +} + +////////////////////////////////////////////////////////////////////////// +// MPFR 2.4.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + +inline int sinh_cosh(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + return mpfr_sinh_cosh(s.mp,c.mp,v.mp,rnd_mode); +} + +inline const mpreal li2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + MPREAL_UNARY_MATH_FUNCTION_BODY(li2); +} + +inline const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + /* R = rem(X,Y) if Y != 0, returns X - n * Y where n = trunc(X/Y). */ + return fmod(x, y, rnd_mode); +} + +inline const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + (void)rnd_mode; + + /* + + m = mod(x,y) if y != 0, returns x - n*y where n = floor(x/y) + + The following are true by convention: + - mod(x,0) is x + - mod(x,x) is 0 + - mod(x,y) for x != y and y != 0 has the same sign as y. + + */ + + if(iszero(y)) return x; + if(x == y) return 0; + + mpreal m = x - floor(x / y) * y; + + m.setSign(sgn(y)); // make sure result has the same sign as Y + + return m; +} + +inline const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t yp, xp; + + yp = y.get_prec(); + xp = x.get_prec(); + + a.set_prec(yp>xp?yp:xp); + + mpfr_fmod(a.mp, x.mp, y.mp, rnd_mode); + + return a; +} + +inline const mpreal rec_sqrt(const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(v); + mpfr_rec_sqrt(x.mp,v.mp,rnd_mode); + return x; +} +#endif // MPFR 2.4.0 Specifics + +////////////////////////////////////////////////////////////////////////// +// MPFR 3.0.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) +inline const mpreal digamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(digamma); } +inline const mpreal ai (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(ai); } +#endif // MPFR 3.0.0 Specifics + +////////////////////////////////////////////////////////////////////////// +// Constants +inline const mpreal const_log2 (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_log2(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_pi (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_pi(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_euler (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_euler(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_catalan (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_catalan(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_infinity (int sign = 1, mp_prec_t p = mpreal::get_default_prec()) +{ + mpreal x(0, p); + mpfr_set_inf(x.mpfr_ptr(), sign); + return x; +} + +////////////////////////////////////////////////////////////////////////// +// Integer Related Functions +inline const mpreal ceil(const mpreal& v) +{ + mpreal x(v); + mpfr_ceil(x.mp,v.mp); + return x; +} + +inline const mpreal floor(const mpreal& v) +{ + mpreal x(v); + mpfr_floor(x.mp,v.mp); + return x; +} + +inline const mpreal round(const mpreal& v) +{ + mpreal x(v); + mpfr_round(x.mp,v.mp); + return x; +} + +inline const mpreal trunc(const mpreal& v) +{ + mpreal x(v); + mpfr_trunc(x.mp,v.mp); + return x; +} + +inline const mpreal rint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint ); } +inline const mpreal rint_ceil (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_ceil ); } +inline const mpreal rint_floor (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_floor); } +inline const mpreal rint_round (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_round); } +inline const mpreal rint_trunc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_trunc); } +inline const mpreal frac (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(frac ); } + +////////////////////////////////////////////////////////////////////////// +// Miscellaneous Functions +inline void swap (mpreal& a, mpreal& b) { mpfr_swap(a.mp,b.mp); } +inline const mpreal (max)(const mpreal& x, const mpreal& y){ return (x>y?x:y); } +inline const mpreal (min)(const mpreal& x, const mpreal& y){ return (x= MPFR_VERSION_NUM(3,0,0)) +inline const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x; + mpfr_urandom(x.mpfr_ptr(), state, rnd_mode); + return x; +} +#endif + +#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) +inline const mpreal random2 (mp_size_t size, mp_exp_t exp) +{ + mpreal x; + mpfr_random2(x.mpfr_ptr(),size,exp); + return x; +} +#endif + +// Uniformly distributed random number generation +// a = random(seed); <- initialization & first random number generation +// a = random(); <- next random numbers generation +// seed != 0 +inline const mpreal random(unsigned int seed = 0) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + static gmp_randstate_t state; + static bool initialize = true; + + if(initialize) + { + gmp_randinit_default(state); + gmp_randseed_ui(state,0); + initialize = false; + } + + if(seed != 0) gmp_randseed_ui(state,seed); + + return mpfr::urandom(state); +#else + if(seed != 0) std::srand(seed); + return mpfr::mpreal(std::rand()/(double)RAND_MAX); +#endif + +} + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + +inline const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x; + mpfr_grandom(x.mpfr_ptr(), NULL, state, rnd_mode); + return x; +} + +inline const mpreal grandom(unsigned int seed = 0) +{ + static gmp_randstate_t state; + static bool initialize = true; + + if(initialize) + { + gmp_randinit_default(state); + gmp_randseed_ui(state,0); + initialize = false; + } + + if(seed != 0) gmp_randseed_ui(state,seed); + + return mpfr::grandom(state); +} +#endif + +////////////////////////////////////////////////////////////////////////// +// Set/Get global properties +inline void mpreal::set_default_prec(mp_prec_t prec) +{ + mpfr_set_default_prec(prec); +} + +inline void mpreal::set_default_rnd(mp_rnd_t rnd_mode) +{ + mpfr_set_default_rounding_mode(rnd_mode); +} + +inline bool mpreal::fits_in_bits(double x, int n) +{ + int i; + double t; + return IsInf(x) || (std::modf ( std::ldexp ( std::frexp ( x, &i ), n ), &t ) == 0.0); +} + +inline const mpreal pow(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow(x.mp,x.mp,b.mp,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_z(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_ui(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_si(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); +} + +inline const mpreal pow(const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_ui_pow(x.mp,a,b.mp,rnd_mode); + return x; +} + +inline const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),b,rnd_mode); +} + +inline const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),b,rnd_mode); + else return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),b,rnd_mode); + else return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); +} + +// pow unsigned long int +inline const mpreal pow(const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + mpreal x(a); + mpfr_ui_pow_ui(x.mp,a,b,rnd_mode); + return x; +} + +inline const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +// pow unsigned int +inline const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +// pow long int +inline const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +inline const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +// pow int +inline const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +inline const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +// pow long double +inline const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),mpreal(b),rnd_mode); +} + +inline const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),mpreal(b),rnd_mode); +} + +inline const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_ui +} + +inline const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_ui +} + +inline const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si +} +} // End of mpfr namespace + +// Explicit specialization of std::swap for mpreal numbers +// Thus standard algorithms will use efficient version of swap (due to Koenig lookup) +// Non-throwing swap C++ idiom: http://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Non-throwing_swap +namespace std +{ + // we are allowed to extend namespace std with specializations only + template <> + inline void swap(mpfr::mpreal& x, mpfr::mpreal& y) + { + return mpfr::swap(x, y); + } + + template<> + class numeric_limits + { + public: + static const bool is_specialized = true; + static const bool is_signed = true; + static const bool is_integer = false; + static const bool is_exact = false; + static const int radix = 2; + + static const bool has_infinity = true; + static const bool has_quiet_NaN = true; + static const bool has_signaling_NaN = true; + + static const bool is_iec559 = true; // = IEEE 754 + static const bool is_bounded = true; + static const bool is_modulo = false; + static const bool traps = true; + static const bool tinyness_before = true; + + static const float_denorm_style has_denorm = denorm_absent; + + inline static mpfr::mpreal (min) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::minval(precision); } + inline static mpfr::mpreal (max) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::maxval(precision); } + inline static mpfr::mpreal lowest (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return -mpfr::maxval(precision); } + + // Returns smallest eps such that 1 + eps != 1 (classic machine epsilon) + inline static mpfr::mpreal epsilon(mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::machine_epsilon(precision); } + + // Returns smallest eps such that x + eps != x (relative machine epsilon) + inline static mpfr::mpreal epsilon(const mpfr::mpreal& x) { return mpfr::machine_epsilon(x); } + + inline static mpfr::mpreal round_error(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + mp_rnd_t r = mpfr::mpreal::get_default_rnd(); + + if(r == GMP_RNDN) return mpfr::mpreal(0.5, precision); + else return mpfr::mpreal(1.0, precision); + } + + inline static const mpfr::mpreal infinity() { return mpfr::const_infinity(); } + inline static const mpfr::mpreal quiet_NaN() { return mpfr::mpreal().setNan(); } + inline static const mpfr::mpreal signaling_NaN() { return mpfr::mpreal().setNan(); } + inline static const mpfr::mpreal denorm_min() { return (min)(); } + + // Please note, exponent range is not fixed in MPFR + static const int min_exponent = MPFR_EMIN_DEFAULT; + static const int max_exponent = MPFR_EMAX_DEFAULT; + MPREAL_PERMISSIVE_EXPR static const int min_exponent10 = (int) (MPFR_EMIN_DEFAULT * 0.3010299956639811); + MPREAL_PERMISSIVE_EXPR static const int max_exponent10 = (int) (MPFR_EMAX_DEFAULT * 0.3010299956639811); + +#ifdef MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS + + // Following members should be constant according to standard, but they can be variable in MPFR + // So we define them as functions here. + // + // This is preferable way for std::numeric_limits specialization. + // But it is incompatible with standard std::numeric_limits and might not work with other libraries, e.g. boost. + // See below for compatible implementation. + inline static float_round_style round_style() + { + mp_rnd_t r = mpfr::mpreal::get_default_rnd(); + + switch (r) + { + case GMP_RNDN: return round_to_nearest; + case GMP_RNDZ: return round_toward_zero; + case GMP_RNDU: return round_toward_infinity; + case GMP_RNDD: return round_toward_neg_infinity; + default: return round_indeterminate; + } + } + + inline static int digits() { return int(mpfr::mpreal::get_default_prec()); } + inline static int digits(const mpfr::mpreal& x) { return x.getPrecision(); } + + inline static int digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + return mpfr::bits2digits(precision); + } + + inline static int digits10(const mpfr::mpreal& x) + { + return mpfr::bits2digits(x.getPrecision()); + } + + inline static int max_digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + return digits10(precision); + } +#else + // Digits and round_style are NOT constants when it comes to mpreal. + // If possible, please use functions digits() and round_style() defined above. + // + // These (default) values are preserved for compatibility with existing libraries, e.g. boost. + // Change them accordingly to your application. + // + // For example, if you use 256 bits of precision uniformly in your program, then: + // digits = 256 + // digits10 = 77 + // max_digits10 = 78 + // + // Approximate formula for decimal digits is: digits10 = floor(log10(2) * digits). See bits2digits() for more details. + + static const std::float_round_style round_style = round_to_nearest; + static const int digits = 53; + static const int digits10 = 15; + static const int max_digits10 = 16; +#endif + }; + +} + +#endif /* __MPREAL_H__ */ diff --git a/third_party/eigen/unsupported/test/polynomialsolver.cpp b/third_party/eigen/unsupported/test/polynomialsolver.cpp index 4cfc46b41bf..db8ad7ddae3 100644 --- a/third_party/eigen/unsupported/test/polynomialsolver.cpp +++ b/third_party/eigen/unsupported/test/polynomialsolver.cpp @@ -26,14 +26,25 @@ struct increment_if_fixed_size } } +template +PolynomialType polyder(const PolynomialType& p) +{ + typedef typename PolynomialType::Scalar Scalar; + PolynomialType res(p.size()); + for(Index i=1; i bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) { typedef typename POLYNOMIAL::Scalar Scalar; + typedef typename POLYNOMIAL::RealScalar RealScalar; typedef typename SOLVER::RootsType RootsType; - typedef Matrix EvalRootsType; + typedef Matrix EvalRootsType; const Index deg = pols.size()-1; @@ -43,10 +54,17 @@ bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) psolve.compute( pols ); const RootsType& roots( psolve.roots() ); EvalRootsType evr( deg ); + POLYNOMIAL pols_der = polyder(pols); + EvalRootsType der( deg ); for( int i=0; i() ); + // we need to divide by the magnitude of the derivative because + // with a high derivative is very small error in the value of the root + // yiels a very large error in the polynomial evaluation. + bool evalToZero = (evr.cwiseQuotient(der)).isZero( test_precision() ); if( !evalToZero ) { cerr << "WRONG root: " << endl; @@ -56,7 +74,7 @@ bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) cerr << endl; } - std::vector rootModuli( roots.size() ); + std::vector rootModuli( roots.size() ); Map< EvalRootsType > aux( &rootModuli[0], roots.size() ); aux = roots.array().abs(); std::sort( rootModuli.begin(), rootModuli.end() ); @@ -82,7 +100,7 @@ void evalSolver( const POLYNOMIAL& pols ) { typedef typename POLYNOMIAL::Scalar Scalar; - typedef PolynomialSolver PolynomialSolverType; + typedef PolynomialSolver PolynomialSolverType; PolynomialSolverType psolve; aux_evalSolver( pols, psolve ); @@ -96,6 +114,7 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const { using std::sqrt; typedef typename POLYNOMIAL::Scalar Scalar; + typedef typename POLYNOMIAL::RealScalar RealScalar; typedef PolynomialSolver PolynomialSolverType; @@ -106,14 +125,12 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const // 1) the roots found are correct // 2) the roots have distinct moduli - typedef typename REAL_ROOTS::Scalar Real; - //Test realRoots - std::vector< Real > calc_realRoots; - psolve.realRoots( calc_realRoots ); - VERIFY( calc_realRoots.size() == (size_t)real_roots.size() ); + std::vector< RealScalar > calc_realRoots; + psolve.realRoots( calc_realRoots, test_precision()); + VERIFY_IS_EQUAL( calc_realRoots.size() , (size_t)real_roots.size() ); - const Scalar psPrec = sqrt( test_precision() ); + const RealScalar psPrec = sqrt( test_precision() ); for( size_t i=0; i 0 ) ); if( hasRealRoot ){ VERIFY( internal::isApprox( real_roots.array().abs().maxCoeff(), abs(r), psPrec ) ); } @@ -165,9 +182,11 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const template void polynomialsolver(int deg) { - typedef internal::increment_if_fixed_size<_Deg> Dim; + typedef typename NumTraits<_Scalar>::Real RealScalar; + typedef internal::increment_if_fixed_size<_Deg> Dim; typedef Matrix<_Scalar,Dim::ret,1> PolynomialType; typedef Matrix<_Scalar,_Deg,1> EvalRootsType; + typedef Matrix RealRootsType; cout << "Standard cases" << endl; PolynomialType pols = PolynomialType::Random(deg+1); @@ -180,15 +199,11 @@ void polynomialsolver(int deg) evalSolver<_Deg,PolynomialType>( pols ); cout << "Test sugar" << endl; - EvalRootsType realRoots = EvalRootsType::Random(deg); + RealRootsType realRoots = RealRootsType::Random(deg); roots_to_monicPolynomial( realRoots, pols ); evalSolverSugarFunction<_Deg>( pols, - realRoots.template cast < - std::complex< - typename NumTraits<_Scalar>::Real - > - >(), + realRoots.template cast >().eval(), realRoots ); } @@ -212,5 +227,6 @@ void test_polynomialsolver() internal::random(9,13) )) ); CALL_SUBTEST_11((polynomialsolver(1)) ); + CALL_SUBTEST_12((polynomialsolver,Dynamic>(internal::random(2,13))) ); } } From b5b304c7ae96c3bd0cb94a1e47fce6713bdb8f1e Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Mon, 16 Aug 2021 17:42:42 -0500 Subject: [PATCH 69/77] Bump cpgfunctionEP to 0.4.1 --- .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/.gitignore | 0 .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/CHANGELOG.md | 0 .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/CMakeLists.txt | 2 -- third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/LICENSE | 0 .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/README.md | 0 .../include/blas/blas.h | 0 .../include/cpgfunction/boreholes.h | 0 .../include/cpgfunction/coordinates.h | 0 .../include/cpgfunction/gfunction.h | 0 .../include/cpgfunction/heat_transfer.h | 0 .../include/cpgfunction/interpolation.h | 0 .../include/cpgfunction/segments.h | 0 .../include/cpgfunction/statistics.h | 0 .../include/cpgfunction/utilities.h | 0 .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/blas.cpp | 0 .../src/boreholes.cpp | 0 .../src/coordinates.cpp | 0 .../src/gfunction.cpp | 0 .../src/heat_transfer.cpp | 0 .../src/interpolation.cpp | 0 .../{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/segments.cpp | 0 .../src/statistics.cpp | 0 .../src/utilities.cpp | 0 .../third_party/qdt-master/LICENSE.txt | 0 .../third_party/qdt-master/README.md | 0 .../third_party/qdt-master/common/method.h | 0 .../third_party/qdt-master/methods/adaptive.h | 0 .../third_party/qdt-master/methods/gauss-kronrod.h | 0 .../third_party/qdt-master/qdt.h | 0 29 files changed, 2 deletions(-) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/.gitignore (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/CHANGELOG.md (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/CMakeLists.txt (97%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/LICENSE (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/README.md (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/blas/blas.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/boreholes.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/coordinates.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/gfunction.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/heat_transfer.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/interpolation.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/segments.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/statistics.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/include/cpgfunction/utilities.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/blas.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/boreholes.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/coordinates.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/gfunction.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/heat_transfer.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/interpolation.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/segments.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/statistics.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/src/utilities.cpp (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/LICENSE.txt (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/README.md (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/common/method.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/methods/adaptive.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/methods/gauss-kronrod.h (100%) rename third_party/{cpgfunctionEP-0.4 => cpgfunctionEP-0.4.1}/third_party/qdt-master/qdt.h (100%) diff --git a/third_party/cpgfunctionEP-0.4/.gitignore b/third_party/cpgfunctionEP-0.4.1/.gitignore similarity index 100% rename from third_party/cpgfunctionEP-0.4/.gitignore rename to third_party/cpgfunctionEP-0.4.1/.gitignore diff --git a/third_party/cpgfunctionEP-0.4/CHANGELOG.md b/third_party/cpgfunctionEP-0.4.1/CHANGELOG.md similarity index 100% rename from third_party/cpgfunctionEP-0.4/CHANGELOG.md rename to third_party/cpgfunctionEP-0.4.1/CHANGELOG.md diff --git a/third_party/cpgfunctionEP-0.4/CMakeLists.txt b/third_party/cpgfunctionEP-0.4.1/CMakeLists.txt similarity index 97% rename from third_party/cpgfunctionEP-0.4/CMakeLists.txt rename to third_party/cpgfunctionEP-0.4.1/CMakeLists.txt index 3ea7f4a94fb..d3f9372ed87 100644 --- a/third_party/cpgfunctionEP-0.4/CMakeLists.txt +++ b/third_party/cpgfunctionEP-0.4.1/CMakeLists.txt @@ -82,6 +82,4 @@ if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME) add_test(NAME RunTest6 COMMAND ${CMAKE_BINARY_DIR}/gFunction_tests) add_test(NAME RunTest7 COMMAND ${CMAKE_BINARY_DIR}/packed) add_test(NAME RunTest8 COMMAND ${CMAKE_BINARY_DIR}/gFunction_adaptive) -else() - include_directories(../third_party/eigen) # Eigen for LU decomposition endif() diff --git a/third_party/cpgfunctionEP-0.4/LICENSE b/third_party/cpgfunctionEP-0.4.1/LICENSE similarity index 100% rename from third_party/cpgfunctionEP-0.4/LICENSE rename to third_party/cpgfunctionEP-0.4.1/LICENSE diff --git a/third_party/cpgfunctionEP-0.4/README.md b/third_party/cpgfunctionEP-0.4.1/README.md similarity index 100% rename from third_party/cpgfunctionEP-0.4/README.md rename to third_party/cpgfunctionEP-0.4.1/README.md diff --git a/third_party/cpgfunctionEP-0.4/include/blas/blas.h b/third_party/cpgfunctionEP-0.4.1/include/blas/blas.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/blas/blas.h rename to third_party/cpgfunctionEP-0.4.1/include/blas/blas.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/boreholes.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/boreholes.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/boreholes.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/boreholes.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/coordinates.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/coordinates.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/coordinates.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/coordinates.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/gfunction.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/gfunction.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/gfunction.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/gfunction.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/heat_transfer.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/heat_transfer.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/heat_transfer.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/heat_transfer.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/interpolation.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/interpolation.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/interpolation.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/interpolation.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/segments.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/segments.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/segments.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/segments.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/statistics.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/statistics.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/statistics.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/statistics.h diff --git a/third_party/cpgfunctionEP-0.4/include/cpgfunction/utilities.h b/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/utilities.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/include/cpgfunction/utilities.h rename to third_party/cpgfunctionEP-0.4.1/include/cpgfunction/utilities.h diff --git a/third_party/cpgfunctionEP-0.4/src/blas.cpp b/third_party/cpgfunctionEP-0.4.1/src/blas.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/blas.cpp rename to third_party/cpgfunctionEP-0.4.1/src/blas.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/boreholes.cpp b/third_party/cpgfunctionEP-0.4.1/src/boreholes.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/boreholes.cpp rename to third_party/cpgfunctionEP-0.4.1/src/boreholes.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/coordinates.cpp b/third_party/cpgfunctionEP-0.4.1/src/coordinates.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/coordinates.cpp rename to third_party/cpgfunctionEP-0.4.1/src/coordinates.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/gfunction.cpp b/third_party/cpgfunctionEP-0.4.1/src/gfunction.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/gfunction.cpp rename to third_party/cpgfunctionEP-0.4.1/src/gfunction.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/heat_transfer.cpp b/third_party/cpgfunctionEP-0.4.1/src/heat_transfer.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/heat_transfer.cpp rename to third_party/cpgfunctionEP-0.4.1/src/heat_transfer.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/interpolation.cpp b/third_party/cpgfunctionEP-0.4.1/src/interpolation.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/interpolation.cpp rename to third_party/cpgfunctionEP-0.4.1/src/interpolation.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/segments.cpp b/third_party/cpgfunctionEP-0.4.1/src/segments.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/segments.cpp rename to third_party/cpgfunctionEP-0.4.1/src/segments.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/statistics.cpp b/third_party/cpgfunctionEP-0.4.1/src/statistics.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/statistics.cpp rename to third_party/cpgfunctionEP-0.4.1/src/statistics.cpp diff --git a/third_party/cpgfunctionEP-0.4/src/utilities.cpp b/third_party/cpgfunctionEP-0.4.1/src/utilities.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4/src/utilities.cpp rename to third_party/cpgfunctionEP-0.4.1/src/utilities.cpp diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/LICENSE.txt b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/LICENSE.txt similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/LICENSE.txt rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/LICENSE.txt diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/README.md b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/README.md similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/README.md rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/README.md diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/common/method.h b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/common/method.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/common/method.h rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/common/method.h diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/methods/adaptive.h b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/methods/adaptive.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/methods/adaptive.h rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/methods/adaptive.h diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/methods/gauss-kronrod.h b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/methods/gauss-kronrod.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/methods/gauss-kronrod.h rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/methods/gauss-kronrod.h diff --git a/third_party/cpgfunctionEP-0.4/third_party/qdt-master/qdt.h b/third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/qdt.h similarity index 100% rename from third_party/cpgfunctionEP-0.4/third_party/qdt-master/qdt.h rename to third_party/cpgfunctionEP-0.4.1/third_party/qdt-master/qdt.h From 32c6dac6d2f02197283b7dfd9fdd7fc1a34a550b Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Mon, 16 Aug 2021 18:57:26 -0500 Subject: [PATCH 70/77] Make cmake files link to cpgEP-0.4.1 --- CMakeLists.txt | 4 ++-- third_party/CMakeLists.txt | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/CMakeLists.txt b/CMakeLists.txt index d65d9f2a1f0..c21cadd247a 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -265,8 +265,8 @@ target_include_directories(project_options SYSTEM INTERFACE "${kiva_SOURCE_DIR}/ add_subdirectory(scripts/dev/generate_embeddable_epJSON_schema) set_target_properties(generate_embeddedable_epJSON_schema PROPERTIES FOLDER "Internal") -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4) -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4/include) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4.1) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4.1/include) if(OPENGL_FOUND) set(BUILD_PENUMBRA_TESTING diff --git a/third_party/CMakeLists.txt b/third_party/CMakeLists.txt index 895a89d7685..3fef930ba36 100644 --- a/third_party/CMakeLists.txt +++ b/third_party/CMakeLists.txt @@ -124,8 +124,8 @@ add_subdirectory(re2) set_target_properties(re2 PROPERTIES FOLDER ThirdParty/re2) add_subdirectory(fmt-6.1.2) set_target_properties(fmt PROPERTIES FOLDER ThirdParty/fmt) -add_subdirectory(cpgfunctionEP-0.4) -set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.4) +add_subdirectory(cpgfunctionEP-0.4.1) +set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.4.1) # Kiva include(cmake/kiva.cmake) From 378a2547af8ba9301e95df85c5f4edd095cb5fe8 Mon Sep 17 00:00:00 2001 From: Matt Mitchell Date: Tue, 17 Aug 2021 10:45:16 -0600 Subject: [PATCH 71/77] add -j CLI option and wire it up to the cpgfunction code. add notes and other descriptions describing limitations of new -j CLI option --- doc/essentials/src/title.tex | 138 ++++++++++---------- doc/man/energyplus.1 | 5 + doc/running-energyplus-from-command-line.md | 21 ++- src/EnergyPlus/CommandLineInterface.cc | 21 ++- src/EnergyPlus/DataGlobals.hh | 2 + src/EnergyPlus/GroundHeatExchangers.cc | 7 +- src/EnergyPlus/GroundHeatExchangers.hh | 2 +- 7 files changed, 118 insertions(+), 78 deletions(-) diff --git a/doc/essentials/src/title.tex b/doc/essentials/src/title.tex index 9a1346b3857..5934b2d7561 100644 --- a/doc/essentials/src/title.tex +++ b/doc/essentials/src/title.tex @@ -5,7 +5,7 @@ \chapter{Introduction} The following background on building energy modeling (BEM) will provide a foundation before learning the essentials of EnergyPlus. If you are already familiar with BEM, you may wish to skip to Section 1.4 -describing the EnergyPlus QuickStart Guide. +describing the EnergyPlus QuickStart Guide. \section{What is BEM?} @@ -82,7 +82,7 @@ \section{What is BEM?} in a building. For these reasons and more, what might first appear as something that can be calculated with just a few formulas in a spreadsheet is instead a software program and, in the case of EnergyPlus, -with over 500,000 lines of code. +with over 500,000 lines of code. \section{Questions that BEM can answer} @@ -190,7 +190,7 @@ \section{EnergyPlus Capabilities} \begin{itemize} \item ASCII text-based weather, input, and output files that include hourly or sub-hourly environmental conditions, and standard and user definable -reports, respectively. +reports, respectively. \item Transient heat conduction through building elements such as walls, roofs, floors, etc. using conduction transfer functions. \item Thermal comfort models based on activity, inside dry-bulb temperature, @@ -204,7 +204,7 @@ \section{EnergyPlus Capabilities} equipment and uses the heat balance method recommended in the \href{https://www.ashrae.org/technical-resources/ashrae-handbook}{ASHRAE Handbook Fundamentals}. Proper sizing of equipment without oversizing, generally saves energy as the equipment is operated nearer to optimal loads. -\item EnergyPlus runs on Windows, MacOS, and Linux computers. +\item EnergyPlus runs on Windows, MacOS, and Linux computers. \end{itemize} Integration of Loads, Systems, and Plants: One of the strong points of EnergyPlus is the integration of all aspects of the simulation---loads, @@ -213,7 +213,7 @@ \section{EnergyPlus Capabilities} first, then simulating systems and plants. The simulation is coupled allowing the designer to more accurately investigate the effect of undersizing fans and equipment and what impact that might have on -the thermal comfort of occupants within the building. +the thermal comfort of occupants within the building. \section{Open Source} @@ -225,7 +225,7 @@ \section{Open Source} code itself. \href{https://github.com/NREL/EnergyPlus/wiki/BuildingEnergyPlus}{Instructions to build the code} (compile the source code into an executable application) are available in the source code repository wiki if you see something that needs -to be enhanced or fixed, please also see the \href{https://energyplus.net/contributing}{contribution policy}. +to be enhanced or fixed, please also see the \href{https://energyplus.net/contributing}{contribution policy}. \section{Brief History} @@ -287,7 +287,7 @@ \section{Documentation} \item Using EnergyPlus for Compliance Guide: Contains information specific to using EnergyPlus in compliance and standard rating systems. \item Tips \& Tricks for Using EnergyPlus: Contains short tips and tricks -for using various parts of EnergyPlus. +for using various parts of EnergyPlus. \end{itemize} \section{Example Files } @@ -303,7 +303,7 @@ \section{Example Files } then the first three files that use that input object. It is possible that many other files also use a particular input object so if the first three files do not help, a text search of files in the ExampleFiles -folder may find more. +folder may find more. \chapter{The EnergyPlus Ecosystem } @@ -315,7 +315,7 @@ \section{Current Interfaces } found in section \ref{sec:Using-EnergyPlus}. In addition, EnergyPlus is often the simulation engine for graphical user interfaces. To see a list, see the \href{https://www.buildingenergysoftwaretools.com/}{BEST (Building Energy Software Tools) Directory} -that is operated by \href{https://www.ibpsa.us/}{IBPSA-USA}. +that is operated by \href{https://www.ibpsa.us/}{IBPSA-USA}. \section{Approaches to Implement Measures } @@ -352,7 +352,7 @@ \section{Approaches to Implement Measures } fixed values to change programmatically. EP-Macro is documented in the AuxiliaryPrograms documentation. A file with lines starting ``\#\#'' or with the ``imf'' file extension is likely to be a file using -EP-Macro commands. +EP-Macro commands. The EnergyPlus ParametricPreprocessor uses special input objects in EnergyPlus to set values for any field in any other input object for @@ -400,7 +400,7 @@ \section{Weather Files \label{subsec:Weather-Files}} calibration or verification studies where the simulation results are compared to actual utility bills or other measured data. -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/WeatherFileLocations.png} \caption{EnergyPlus.net Weather File Locations} @@ -424,9 +424,9 @@ \section{Co-Simulation and Linked Software} for linking the sub-systems at run-time for data-exchange. EnergyPlus implements three mechanisms to support co-simulation. \begin{itemize} -\item EnergyPlus implements the Building Controls Virtual Test Bed ( \href{https://www.tandfonline.com/doi/abs/10.1080/19401493.2010.518631}{BCVTB}) API. -This API leverages the BCVTB to enable the co-simulation of EnergyPlus with various simulation programs such as \href{http://www.trnsys.com/}{TRNSYS}, -\href{http://www.esru.strath.ac.uk/Programs/ESP-r.htm}{ESP-r}, \href{http://radsite.lbl.gov/radiance/HOME.html}{Radiance}, or +\item EnergyPlus implements the Building Controls Virtual Test Bed ( \href{https://www.tandfonline.com/doi/abs/10.1080/19401493.2010.518631}{BCVTB}) API. +This API leverages the BCVTB to enable the co-simulation of EnergyPlus with various simulation programs such as \href{http://www.trnsys.com/}{TRNSYS}, +\href{http://www.esru.strath.ac.uk/Programs/ESP-r.htm}{ESP-r}, \href{http://radsite.lbl.gov/radiance/HOME.html}{Radiance}, or \href{https://www.3ds.com/products-services/catia/products/dymola/}{DYMOLA}. \item EnergyPlus provides an interface which allows it to import, link, and exchange data with simulation models which implement the Functional @@ -490,7 +490,7 @@ \section{Running EnergyPlus} \item EP-Launch \end{itemize} In each case, a building model will be simulated in combination with -a weather file for the appropriate building location. +a weather file for the appropriate building location. \subsection*{Graphical User Interface} @@ -509,12 +509,12 @@ \subsection*{Command Line} in the QuickStart Guide. To learn more about the command line mode, you can type: \begin{verbatim} -energyplus --help +energyplus --help \end{verbatim} when in the EnergyPlus folder. This will give the following display of options: \begin{verbatim} -EnergyPlus, Version 9.0.1-bb7ca4f0da +EnergyPlus, Version 9.6.0-ec0190a2fc Usage: energyplus [options] [input-file] Options: -a, --annual Force annual simulation @@ -526,6 +526,8 @@ \subsection*{Command Line} -h, --help Display help information -i, --idd ARG Input data dictionary path (default: Energy+.idd in executable directory) + -j, --jobs ARG Multi-thread with N threads; 1 thread with no + arg. -m, --epmacro Run EPMacro prior to simulation -p, --output-prefix ARG Prefix for output file names (default: eplus) -r, --readvars Run ReadVarsESO after simulation @@ -537,6 +539,8 @@ \subsection*{Command Line} -w, --weather ARG Weather file path (default: in.epw in current directory) -x, --expandobjects Run ExpandObjects prior to simulation +--convert-only Only convert IDF->epJSON or epJSON->IDF, + dependent on input file type. No simulation Example: energyplus -w weather.epw -r input.idf \end{verbatim} EnergyPlus can be run by specifying a number of options followed by @@ -549,7 +553,7 @@ \subsection*{Command Line} \textquotedbl -{}-help\textquotedbl ). Several of these options are commonly used including the weather, output-prefix, expandobjects, and readvars options. The following are some examples of using the -command line options. +command line options. Pre-processing using EPMacro and ExpandObjects: \begin{verbatim} @@ -575,12 +579,12 @@ \subsection*{EP-Launch} open a text editor for the input and output files, open a spreadsheet for the result files, a web browser for the tabular results file, and start up a viewer for the selected drawing file. There are two -different versions of EP-Launch currently part of the EnergyPlus system. +different versions of EP-Launch currently part of the EnergyPlus system. The main screen of EP-Launch 2 is shown below: -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/eplaunch2.png} \caption{EP-Launch 2} @@ -591,11 +595,11 @@ \subsection*{EP-Launch} installation package when installing on Windows, so no additional steps are needed to run it. It is located in the main ``root'' folder of EnergyPlus, usually, a folder named EnergyPlusVx-x-x, where the -x's are the version number. +x's are the version number. In 2018, EP-Launch 3 was developed, and its main screen is shown below: -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/eplaunch3.png} \caption{EP-Launch 3} @@ -623,7 +627,7 @@ \section{IDF and JSON syntax} describe the building and system that is simulated. The file extensions for the two formats are IDF and epJSON. For both input files, the numeric inputs are in SI units (International System of Units often -called metric units). +called metric units). \subsection*{IDF} @@ -637,7 +641,7 @@ \subsection*{IDF} field per line. The names of each field are not required but are usually shown after the value and the comma or semicolon as a special comment using ``!-'' as an indicator. The input objects can be in any order. -An example of an input object in an IDF file is shown below: +An example of an input object in an IDF file is shown below: \begin{verbatim} Building, Simple One Zone, !- Name @@ -723,7 +727,7 @@ \section{Creating and Editing Input Files} Full details of the IDF Editor can be found in the Auxiliary Programs document under the ``Creating Input Files'' section. -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/idfeditor.png} \caption{IDF Editor} @@ -744,11 +748,11 @@ \section{Run-Check-Edit Repeat \label{subsec:Run-Check-Edit-Repeat}} Don't expect that an initial model is ever correct; it is probably not. Initially, errors are likely to exist. The .ERR file should be the first file checked each time EnergyPlus is run. The .ERR file -has several levels of messages: +has several levels of messages: \begin{itemize} \item Warning \item Severe -\item Fatal +\item Fatal \end{itemize} A Fatal error means that EnergyPlus has stopped during the simulation and the input file needs to be fixed before the simulation can be @@ -897,37 +901,37 @@ \section{What Are All These Folders?} The installation of EnergyPlus includes many different files in different folders: -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/energyplusfolder.png} \caption{EnergyPlus Installation Folders} \end{figure} Many of these folders include valuable resources for using and learning -EnergyPlus. +EnergyPlus. \begin{itemize} \item The main folder includes the EnergyPlus executable which can be used on the command line and EP-Launch 2, a program that makes it easier to use EnergyPlus and the Energy+.IDD that describes each possible EnergyPlus input object and the default, minimum, maximum, and options -for each field within each input object. +for each field within each input object. \item The Documentation folder includes this document as well as the InputOutputReference, EngineeringReference, AuxiliaryPrograms, OutputDetailsAndExamples which are very important to understand. If you haven't looked through -the documentation yet, take a few minutes and get familiar with it. +the documentation yet, take a few minutes and get familiar with it. \item The DataSets and MacroDataSets folders include files containing libraries of input objects that may be useful in constructing your own input files. The ASHRAE\_2005\_HOF\_Materials.idf and WindowConstructs.idf -files, for example, will help with defining walls and windows. +files, for example, will help with defining walls and windows. \item The ExampleFiles folder includes a huge number of example files that are indexed in the two HTML files in that folder or can be searched -through using most text editors. +through using most text editors. \item The Preprocess and PostProcess folders include many utilities that can be used directly or as part of EP-Launch that can aid in the setting up input files or extracting or converting results. The WeatherData folder includes a small sample of the many weather files that are available. For other weather files, please see the previous section -on \ref{subsec:Weather-Files}. +on \ref{subsec:Weather-Files}. \end{itemize} \section{What Are All These Output Files?} @@ -944,14 +948,14 @@ \section{What Are All These Output Files?} files. \item CSV, TAB, or TXT -- time series output from the Output:Variable input object in a comma, tab, or space delimited format (generated by the -ReadVarsESO postprocessor) +ReadVarsESO postprocessor) \item METER.CSV, METER.TAB, or METER.CSV File -- time series output from the Output:Meter input object in a comma, tab, or space delimited format (generated by the ReadVarsESO postprocessor) \item SQL - sqlite3 output database format \item EIO -- additional EnergyPlus results \item RDD -- list of output variables available from the run -\item MDD -- list of output meters available from the run +\item MDD -- list of output meters available from the run \item MTD -- list of meter component variables \item DXF -- drawing file in AutoCAD DXF format \item AUDIT -- input file echo with input processor errors and warnings @@ -1021,7 +1025,7 @@ \section{Versions and Updating} document. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/IDFVersionUpdater.png} \caption{IDFVersionUpdater} @@ -1080,7 +1084,7 @@ \section{Other Useful Utility Programs} and produces a Scalable Vector Graphics (SVG) file. Many web browsers and other drawing programs can open SVG files. This utility runs automatically with EP-Launch. More information on the HVAC Diagram program is found -in the Auxiliary Programs document. +in the Auxiliary Programs document. \item convertESOMTR - This simple post-processing program can be used seamlessly with EP-Launch to provide IP (inch-pound) unit output files rather than SI units. This program is described more fully in the Auxiliary @@ -1162,7 +1166,7 @@ \section{Location and Climate} can be used to speed up the simulation portion of the Run-Check-Edit cycle. \item RunPeriodControl:SpecialDays - allows specfication of holidays and -a good example can be seen in 5ZoneCostEst.idf. +a good example can be seen in 5ZoneCostEst.idf. \item RunPeriodControl:DaylightSavingTime - allows the specification of the start and ending period for daylight savings time. This will impact when schedules operate but please note that output reporting timesteps @@ -1224,7 +1228,7 @@ \section{Schedules} input specifically for shading. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/officeSchedules.png} \caption{Various Office Schedules} @@ -1279,7 +1283,7 @@ \section{Surface Construction Elements} input object. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/EnvelopeHierarchy.png} \caption{Envelope Component Hierarchy} @@ -1303,7 +1307,7 @@ \section{Surface Construction Elements} and blinds and screens, alternative ways of specifying fenestration such as equivalent layers or refraction extinction method or ASHWAT model or from a WINDOW program export/data file or specifying wavelength-by-wavelength -properties. +properties. The input objects described in this section are further explained in the InputOutputReference under the heading ``Group-Surface Construction @@ -1332,7 +1336,7 @@ \section{Thermal Zones and Surfaces} X with three dimensions. The X-axis points east, the Y-axis points north, and the Z-axis points up. -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/coordinatesystem.png} \caption{EnergyPlus Coordinate System} @@ -1361,7 +1365,7 @@ \section{Thermal Zones and Surfaces} (vertices). In addition, this input object defines what is on the other side of the surface from the zone, whether that is outside, another zone in the building, or the ground. Other inputs indicate -if the outside of the surface is exposed to the sun or the wind. +if the outside of the surface is exposed to the sun or the wind. \item FenestrationSurface:Detailed - describes windows, doors, and special daylighting tubes and it references the wall that it is part of. It also requires the specification of vertices (usually four but three @@ -1397,7 +1401,7 @@ \subsection*{Shading Related} as the length and width. For deciduous trees, and other situations that shading changes over time, the schedule for the transmittance can vary, otherwise, it should always be set to zero or leave the -schedule name blank. +schedule name blank. \item Shading:Zone:Detailed - describes an attachment to the building that casts a shadow such as an overhang or fin. It also includes a reference to a transmittance schedule. The wall that the fin or overhang is @@ -1463,7 +1467,7 @@ \subsection*{Other Related Input Objects} glazed area. See the PurchAirWithDaylighting.idf example file. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/window_frame_and_divider.png} \caption{A window with frame and divider} @@ -1506,22 +1510,22 @@ \section{Internal Gains } occurs for more than a few hours per year if at all and this is especially the case for retail stores, theaters, and sports complexes. Even office buildings when counting vacations and people out of the building for -meetings will rarely have peak occupancy. +meetings will rarely have peak occupancy. The most common internal gain input objects that are shown in almost -all the example files are: +all the example files are: \begin{itemize} \item People -- specifies not only the sensible, latent and radiant heat from people but also includes ways of reporting the comfort of occupants using a variety of thermal comfort models. The DynamicClothing.idf example file shows how to use the thermal comfort models. -\item Lights -- describes the heat related to lighting systems. +\item Lights -- describes the heat related to lighting systems. \item ElectricEquipment -- describes the heat related to electrical appliances, -office equipment, and other heat sources that are powered by electricity. +office equipment, and other heat sources that are powered by electricity. \item GasEquipment - specifies the heat related to cooking appliances and other equipment that uses natural gas. \end{itemize} -Less common internal gains input objects include: +Less common internal gains input objects include: \begin{itemize} \item OtherEquipment - describes any heat gain or loss (sensible, radiant, and/or latent) that impacts the space but does not consume utility @@ -1531,7 +1535,7 @@ \section{Internal Gains } example file. \item SwimmingPool:Indoor - see the 5ZoneSwimmingPool.idf example file. \end{itemize} -Other +Other \begin{itemize} \item ComfortViewFactorAngles -- allows the specification of how different surfaces impact the thermal comfort calculations for the occupants. @@ -1553,11 +1557,11 @@ \section{Daylighting } systems are a very common energy efficiency measure in buildings and are often required for new building designs depending on the energy code that applies to the building location. The most common input -objects related to daylighting are: +objects related to daylighting are: \begin{itemize} \item Daylighting:Controls -- specifies the algorithm used for daylighting, the dimming of lights is continuous or stepped, and how glare calculations -are performed. +are performed. \item Daylighting:ReferencePoint -- specifies the location of the sensors for the daylighting control system. \end{itemize} @@ -1565,14 +1569,14 @@ \section{Daylighting } input objects but also extra guidance on how they should be applied. The PurchAirWithDaylighting.idf contains examples of these input objects. -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/DaylightingContinuous.png} \caption{Daylighting with Continuous Dimming} \end{figure} -Three different devices can be used with daylighting: +Three different devices can be used with daylighting: \begin{itemize} \item DaylightingDevice:Tubular - see the DaylightingDeviceTubular.idf example file. @@ -1583,10 +1587,10 @@ \section{Daylighting } An input object called Daylighting:DELight:ComplexFenestration is used with one of the two control methods specified in the Daylighting:Controls input object when used in conjunction with complex fenestration systems -such as prismatic and holographic glass. +such as prismatic and holographic glass. Some flexibility is given to provide extra output related to daylighting -and includes: +and includes: \begin{itemize} \item Output:DaylightFactors -- creates a special report on the factors used in daylighting. See the ReportDaylightFactors.idf example file. @@ -1620,7 +1624,7 @@ \section{Advanced Construction, Surface, Zone Concepts} are not expected to have any heat transfer with the environment. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/kiva-2d-elements.png} \caption{Structural and insulation components of Foundation:Kiva input objects} @@ -1771,7 +1775,7 @@ \section{HVAC Templates} systems \end{itemize} The Input Output Reference includes a list of exactly which input -objects are needed for each of these configurations +objects are needed for each of these configurations \begin{itemize} \item HVACTemplate:Thermostat - describes the heating and cooling setpoints for a thermostat. This input object can be referenced by multiple @@ -1880,9 +1884,9 @@ \section{Detailed HVAC} feature in EnergyPlus. As a result, it is recommended that one of the first steps taken in defining an HVAC system in EnergyPlus be the definition of a node diagram or map. This is helpful for visualization -of the entire system. +of the entire system. -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/NodeDiagram.png} \caption{Example Node Diagram} @@ -1933,7 +1937,7 @@ \section{Detailed HVAC} a bypass. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/HvacHeatLoop.png} \caption{Detailed HVAC Heat Loop Line Diagram} @@ -2199,7 +2203,7 @@ \section{Economics} months, and this input object allows the specification of such a structure. \end{itemize} -\begin{figure}[hbtp] +\begin{figure}[hbtp] \centering \includegraphics[width=0.9\textwidth, height=0.9\textheight, keepaspectratio=true]{media/tariff-charges.png} \caption{Hierarchy of Economic Charges} @@ -2212,7 +2216,7 @@ \section{Economics} \item UtilityCost:Computation - provides the step-by-step specification of the steps needed to compute the tariff. The output report tariff summary includes the default computation steps for the input objects -specified. +specified. \end{itemize} Life-cycle costing allows the combination of initial and future costs (typically energy bills) to be understood using a single metric so @@ -2248,7 +2252,7 @@ \section{Economics} but note that this group is described in Chapter 3 ``EnergyPlus Economics'' while most of the other groups of input objects are described in Chapter 1. The OutputDetailsAndExamples documentation provides even more details -on the tabular outputs created in the section labeled ``eplustbl.''. +on the tabular outputs created in the section labeled ``eplustbl.''. \section{Other Groups of Input Objects} @@ -2287,7 +2291,7 @@ \section{Other Groups of Input Objects} of the building when used for compliance purposes. \item Parametrics - allows for simple parametric cases to be defined within an input file using input objects. Makes use of the ParametricPreprocessor -as described in the AuxiliaryPrograms document. +as described in the AuxiliaryPrograms document. \item Refrigeration - describes supermarket and warehouse refrigeration systems. \item Electric Load Center-Generator Specifications - allows the specification diff --git a/doc/man/energyplus.1 b/doc/man/energyplus.1 index ac202ae7514..6acc66b7cf6 100644 --- a/doc/man/energyplus.1 +++ b/doc/man/energyplus.1 @@ -54,6 +54,11 @@ Input data dictionary path (default: Energy+.idd in executable directory) .RS .RE .TP +.B \-j, \-\-jobs \f[I]ARG\f[] +Multi-thread with N threads; 1 thread with no arg. +.RS +.RE +.TP .B \-m, \-\-epmacro Run EPMacro prior to simulation .RS diff --git a/doc/running-energyplus-from-command-line.md b/doc/running-energyplus-from-command-line.md index 3d62166f32e..c50b28df64a 100644 --- a/doc/running-energyplus-from-command-line.md +++ b/doc/running-energyplus-from-command-line.md @@ -13,22 +13,27 @@ This will give the following display of options: -a, --annual Force annual simulation -c, --convert Output IDF->epJSON or epJSON->IDF, dependent on input file type - -d, --output-directory ARG Output directory path (default: current directory) + -d, --output-directory ARG Output directory path (default: current + directory) -D, --design-day Force design-day-only simulation -h, --help Display help information -i, --idd ARG Input data dictionary path (default: Energy+.idd in executable directory) + -j, --jobs ARG Multi-thread with N threads; 1 thread with no + arg. -m, --epmacro Run EPMacro prior to simulation -p, --output-prefix ARG Prefix for output file names (default: eplus) -r, --readvars Run ReadVarsESO after simulation -s, --output-suffix ARG Suffix style for output file names (default: L) - L: Legacy (e.g., eplustbl.csv) - C: Capital (e.g., eplusTable.csv) - D: Dash (e.g., eplus-table.csv) + L: Legacy (e.g., eplustbl.csv) + C: Capital (e.g., eplusTable.csv) + D: Dash (e.g., eplus-table.csv) -v, --version Display version information -w, --weather ARG Weather file path (default: in.epw in current - directory)) + directory) -x, --expandobjects Run ExpandObjects prior to simulation + --convert-only Only convert IDF->epJSON or epJSON->IDF, + dependent on input file type. No simulation Example: energyplus -w weather.epw -r input.idf EnergyPlus can be run by specifying a number of options followed by the path to the input file (`input-file`). The file itself is usually in IDF (Input Data File) format, but it may also be in IMF (Input Macro File) format to be run with EPMacro using the `--epmacro` option. @@ -53,6 +58,12 @@ The options generally fall into four categories: 4. Input override switches: - `annual` - `design-day` + - `jobs` + +Notes +----- + +As of V9.6, the `--jobs` option is only configured to increase thread usage for `GroundHeatExchanger:System` objects with the g-Function calculation method set to `UBHWTcalc`. As EnergyPlus continues to evolve, this may be adapted to control other features or models. Examples -------- diff --git a/src/EnergyPlus/CommandLineInterface.cc b/src/EnergyPlus/CommandLineInterface.cc index 14cca854552..55a48ef012f 100644 --- a/src/EnergyPlus/CommandLineInterface.cc +++ b/src/EnergyPlus/CommandLineInterface.cc @@ -45,6 +45,9 @@ // OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE // POSSIBILITY OF SUCH DAMAGE. +// C++ Headers +#include + // CLI Headers #include @@ -68,7 +71,7 @@ namespace CommandLineInterface { { typedef std::string::size_type size_type; - // Expand long-name options using "=" sign into two arguments + // Expand long-name options using "=" sign in to two arguments // and expand multiple short options into separate arguments std::vector arguments; @@ -149,6 +152,8 @@ namespace CommandLineInterface { opt.add("", false, 0, 0, "Display version information", "-v", "--version"); + opt.add("1", false, 1, 0, "Multi-thread with N threads; 1 thread with no arg.", "-j", "--jobs"); + opt.add("in.epw", false, 1, 0, "Weather file path (default: in.epw in current directory)", "-w", "--weather"); opt.add("", false, 0, 0, "Run ExpandObjects prior to simulation", "-x", "--expandobjects"); @@ -210,6 +215,18 @@ namespace CommandLineInterface { bool eplusRunningViaAPI = state.dataGlobal->eplusRunningViaAPI; + opt.get("-j")->getInt(state.dataGlobal->numThread); + + if (state.dataGlobal->numThread == 0) { + DisplayString(state, "Invalid value for -j arg. Defaulting to 1."); + state.dataGlobal->numThread = 1; + } else if (state.dataGlobal->numThread > (int)std::thread::hardware_concurrency()) { + DisplayString(state, + fmt::format("Invalid value for -j arg. Value exceeds num available. Defaulting to num available. -j {}", + (int)std::thread::hardware_concurrency())); + state.dataGlobal->numThread = (int)std::thread::hardware_concurrency(); + } + // Process standard arguments if (opt.isSet("-h")) { DisplayString(state, usage); @@ -529,7 +546,7 @@ namespace CommandLineInterface { } } - // This is a place holder in case there are required options in the future + // This is a placeholder in case there are required options in the future if (!opt.gotRequired(badOptions)) { for (size_type i = 0; i < badOptions.size(); ++i) { DisplayString(state, "ERROR: Missing required option " + badOptions[i]); diff --git a/src/EnergyPlus/DataGlobals.hh b/src/EnergyPlus/DataGlobals.hh index 2cb2557401d..3fd3c78cac2 100644 --- a/src/EnergyPlus/DataGlobals.hh +++ b/src/EnergyPlus/DataGlobals.hh @@ -168,6 +168,7 @@ struct DataGlobal : BaseGlobalStruct int FDnumIterYears = 0; bool printConsoleOutput = true; bool installRootOverride = false; + int numThread = 1; void clear_state() override { @@ -273,6 +274,7 @@ struct DataGlobal : BaseGlobalStruct this->FDnumIterYears = 0; this->printConsoleOutput = true; this->installRootOverride = false; + this->numThread = 1; } }; diff --git a/src/EnergyPlus/GroundHeatExchangers.cc b/src/EnergyPlus/GroundHeatExchangers.cc index 3c5368c8093..3ae5ae5a32c 100644 --- a/src/EnergyPlus/GroundHeatExchangers.cc +++ b/src/EnergyPlus/GroundHeatExchangers.cc @@ -979,7 +979,7 @@ void GLHEVert::calcLongTimestepGFunctions(EnergyPlusData &state) this->calcUniformHeatFluxGFunctions(state); break; case GFuncCalcMethod::UniformBoreholeWallTemp: - this->calcUniformBHWallTempGFunctions(); + this->calcUniformBHWallTempGFunctions(state); break; default: assert(false); @@ -988,7 +988,7 @@ void GLHEVert::calcLongTimestepGFunctions(EnergyPlusData &state) //****************************************************************************** -void GLHEVert::calcUniformBHWallTempGFunctions() +void GLHEVert::calcUniformBHWallTempGFunctions(EnergyPlusData &state) { // construct boreholes vector std::vector boreholes; @@ -1006,7 +1006,8 @@ void GLHEVert::calcUniformBHWallTempGFunctions() gt::segments::adaptive adptDisc; int nSegments = adptDisc.discretize(this->bhLength, this->totalTubeLength); - this->myRespFactors->GFNC = gt::gfunction::uniform_borehole_wall_temperature(boreholes, time, this->soil.diffusivity, nSegments); + this->myRespFactors->GFNC = + gt::gfunction::uniform_borehole_wall_temperature(boreholes, time, this->soil.diffusivity, nSegments, true, state.dataGlobal->numThread); } //****************************************************************************** diff --git a/src/EnergyPlus/GroundHeatExchangers.hh b/src/EnergyPlus/GroundHeatExchangers.hh index 3496d71f91d..917cd8c67ca 100644 --- a/src/EnergyPlus/GroundHeatExchangers.hh +++ b/src/EnergyPlus/GroundHeatExchangers.hh @@ -372,7 +372,7 @@ namespace GroundHeatExchangers { void calcUniformHeatFluxGFunctions(EnergyPlusData &state); - void calcUniformBHWallTempGFunctions(); + void calcUniformBHWallTempGFunctions(EnergyPlusData &state); Real64 calcHXResistance(EnergyPlusData &state) override; From 326a8e7b70bbcc42aee7fa63cba82337b3071a03 Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Tue, 17 Aug 2021 17:55:11 -0500 Subject: [PATCH 72/77] Revert "Bump Eigen library to 3.3.9" This reverts commit 33a28cd8b81991fadc680f34157ae704aea3b8ec. --- third_party/eigen/CMakeLists.txt | 53 +- third_party/eigen/COPYING.MINPACK | 104 +- third_party/eigen/CTestConfig.cmake | 6 +- third_party/eigen/Eigen/Core | 7 +- third_party/eigen/Eigen/Eigenvalues | 4 +- third_party/eigen/Eigen/Geometry | 4 +- third_party/eigen/Eigen/QR | 4 +- third_party/eigen/Eigen/SparseQR | 1 + third_party/eigen/Eigen/src/Core/ArrayBase.h | 4 +- .../eigen/Eigen/src/Core/CwiseUnaryView.h | 2 - third_party/eigen/Eigen/src/Core/DenseBase.h | 7 +- .../eigen/Eigen/src/Core/DenseStorage.h | 6 +- .../eigen/Eigen/src/Core/GenericPacketMath.h | 9 +- third_party/eigen/Eigen/src/Core/MapBase.h | 5 - .../eigen/Eigen/src/Core/MathFunctions.h | 16 +- third_party/eigen/Eigen/src/Core/MatrixBase.h | 3 +- .../eigen/Eigen/src/Core/PermutationMatrix.h | 28 + .../eigen/Eigen/src/Core/PlainObjectBase.h | 12 +- .../eigen/Eigen/src/Core/ProductEvaluators.h | 28 +- third_party/eigen/Eigen/src/Core/Ref.h | 5 +- .../eigen/Eigen/src/Core/SolveTriangular.h | 6 +- third_party/eigen/Eigen/src/Core/Transpose.h | 2 - .../eigen/Eigen/src/Core/Transpositions.h | 39 + .../eigen/Eigen/src/Core/TriangularMatrix.h | 8 +- .../src/Core/arch/AVX512/MathFunctions.h | 70 +- .../Eigen/src/Core/arch/AVX512/PacketMath.h | 717 +- .../eigen/Eigen/src/Core/arch/CUDA/Half.h | 1 - .../Eigen/src/Core/arch/CUDA/PacketMathHalf.h | 4 +- .../Eigen/src/Core/functors/UnaryFunctors.h | 2 +- .../Core/products/GeneralBlockPanelKernel.h | 7 +- .../src/Core/products/GeneralMatrixMatrix.h | 37 +- .../products/GeneralMatrixMatrixTriangular.h | 68 +- .../GeneralMatrixMatrixTriangular_BLAS.h | 10 +- .../Core/products/GeneralMatrixMatrix_BLAS.h | 6 +- .../Eigen/src/Core/products/Parallelizer.h | 9 +- .../Core/products/SelfadjointMatrixMatrix.h | 54 +- .../products/SelfadjointMatrixMatrix_BLAS.h | 24 +- .../src/Core/products/SelfadjointProduct.h | 4 +- .../Core/products/TriangularMatrixMatrix.h | 54 +- .../products/TriangularMatrixMatrix_BLAS.h | 26 +- .../Core/products/TriangularSolverMatrix.h | 62 +- .../products/TriangularSolverMatrix_BLAS.h | 12 +- .../eigen/Eigen/src/Core/util/BlasUtil.h | 115 +- .../src/Core/util/DisableStupidWarnings.h | 13 +- .../Eigen/src/Core/util/ForwardDeclarations.h | 6 +- .../eigen/Eigen/src/Core/util/Macros.h | 58 +- third_party/eigen/Eigen/src/Core/util/Meta.h | 34 - .../src/Core/util/ReenableStupidWarnings.h | 6 +- .../eigen/Eigen/src/Core/util/XprHelper.h | 17 - .../Eigen/src/Eigenvalues/ComplexSchur.h | 9 +- .../eigen/Eigen/src/Eigenvalues/RealSchur.h | 15 +- .../src/Eigenvalues/SelfAdjointEigenSolver.h | 7 +- .../eigen/Eigen/src/Geometry/Quaternion.h | 22 +- .../eigen/Eigen/src/Geometry/Scaling.h | 2 +- .../eigen/Eigen/src/Geometry/Transform.h | 4 +- .../eigen/Eigen/src/Geometry/Translation.h | 6 + .../eigen/Eigen/src/Geometry/Umeyama.h | 2 +- third_party/eigen/Eigen/src/LU/PartialPivLU.h | 5 +- .../eigen/Eigen/src/LU/arch/Inverse_SSE.h | 4 +- .../Eigen/src/PardisoSupport/PardisoSupport.h | 3 +- third_party/eigen/Eigen/src/SVD/BDCSVD.h | 61 +- third_party/eigen/Eigen/src/SVD/SVDBase.h | 2 +- .../src/SparseCholesky/SimplicialCholesky.h | 2 +- .../SparseCholesky/SimplicialCholesky_impl.h | 2 +- .../eigen/Eigen/src/SparseCore/AmbiVector.h | 7 +- .../Eigen/src/SparseCore/SparseCwiseUnaryOp.h | 2 - .../eigen/Eigen/src/SparseCore/SparseMatrix.h | 3 +- .../src/SparseCore/SparseSelfAdjointView.h | 4 +- .../eigen/Eigen/src/SparseCore/SparseView.h | 1 - .../eigen/Eigen/src/SparseLU/SparseLU.h | 4 +- .../eigen/Eigen/src/StlSupport/StdDeque.h | 6 +- .../Eigen/src/plugins/ArrayCwiseBinaryOps.h | 2 +- third_party/eigen/bench/bench_gemm.cpp | 5 +- third_party/eigen/blas/level3_impl.h | 182 +- third_party/eigen/cmake/EigenTesting.cmake | 2 - .../eigen/cmake/FindStandardMathLibrary.cmake | 7 +- third_party/eigen/debug/msvc/eigen.natvis | 470 +- .../eigen/debug/msvc/eigen_autoexp_part.dat | 590 +- third_party/eigen/doc/CMakeLists.txt | 9 +- .../doc/CustomizingEigen_CustomScalar.dox | 2 +- third_party/eigen/doc/Doxyfile.in | 17 +- third_party/eigen/doc/Pitfalls.dox | 84 +- .../eigen/doc/SparseQuickReference.dox | 2 +- third_party/eigen/doc/TopicLazyEvaluation.dox | 76 +- third_party/eigen/doc/TopicMultithreading.dox | 1 - third_party/eigen/doc/TutorialGeometry.dox | 4 +- third_party/eigen/doc/eigen_navtree_hacks.js | 5 +- .../eigen/doc/eigendoxy_footer.html.in | 13 + .../eigen/doc/eigendoxy_header.html.in | 3 - ...orial_BlockOperations_block_assignment.cpp | 2 +- third_party/eigen/lapack/CMakeLists.txt | 4 +- third_party/eigen/scripts/CMakeLists.txt | 8 +- third_party/eigen/test/CMakeLists.txt | 4 +- third_party/eigen/test/bdcsvd.cpp | 8 +- third_party/eigen/test/constructor.cpp | 14 - third_party/eigen/test/ctorleak.cpp | 20 +- .../eigen/test/eigensolver_generic.cpp | 2 +- third_party/eigen/test/exceptions.cpp | 4 +- third_party/eigen/test/fastmath.cpp | 34 +- third_party/eigen/test/geo_alignedbox.cpp | 3 +- third_party/eigen/test/geo_quaternion.cpp | 8 - .../eigen/test/geo_transformations.cpp | 61 +- third_party/eigen/test/inverse.cpp | 17 - third_party/eigen/test/main.h | 5 - third_party/eigen/test/numext.cpp | 5 +- third_party/eigen/test/packetmath.cpp | 17 +- third_party/eigen/test/product.h | 26 - third_party/eigen/test/product_large.cpp | 2 - third_party/eigen/test/product_mmtr.cpp | 10 - third_party/eigen/test/product_symm.cpp | 20 +- third_party/eigen/test/product_syrk.cpp | 11 - third_party/eigen/test/product_trmm.cpp | 12 +- third_party/eigen/test/product_trsolve.cpp | 26 - third_party/eigen/test/ref.cpp | 12 +- third_party/eigen/test/rvalue_types.cpp | 74 +- third_party/eigen/test/sparse_basic.cpp | 8 - third_party/eigen/test/stddeque.cpp | 16 +- third_party/eigen/test/stddeque_overload.cpp | 14 +- third_party/eigen/test/stdlist.cpp | 16 +- third_party/eigen/test/stdlist_overload.cpp | 14 +- third_party/eigen/test/stdvector.cpp | 6 +- third_party/eigen/test/stdvector_overload.cpp | 6 +- .../eigen/test/vectorization_logic.cpp | 8 - .../eigen/unsupported/Eigen/ArpackSupport | 4 +- .../src/Tensor/TensorContractionMapper.h | 2 - .../CXX11/src/Tensor/TensorDeviceThreadPool.h | 2 +- .../Eigen/CXX11/src/ThreadPool/EventCount.h | 2 +- .../Eigen/CXX11/src/ThreadPool/RunQueue.h | 2 +- .../eigen/unsupported/Eigen/Polynomials | 4 +- .../Eigen/src/AutoDiff/AutoDiffScalar.h | 28 +- .../ArpackSelfAdjointEigenSolver.h | 21 +- .../unsupported/Eigen/src/FFT/ei_fftw_impl.h | 2 - .../src/MatrixFunctions/MatrixExponential.h | 2 +- .../src/MatrixFunctions/MatrixSquareRoot.h | 18 +- .../Eigen/src/Polynomials/Companion.h | 57 +- .../Eigen/src/Polynomials/PolynomialSolver.h | 46 +- .../unsupported/Eigen/src/Splines/Spline.h | 2 +- .../test/NonLinearOptimization.cpp | 119 +- .../eigen/unsupported/test/autodiff.cpp | 16 - .../unsupported/test/matrix_function.cpp | 38 - .../eigen/unsupported/test/mpreal/mpreal.h | 6208 ++++++++--------- .../unsupported/test/polynomialsolver.cpp | 54 +- 142 files changed, 4884 insertions(+), 5666 deletions(-) diff --git a/third_party/eigen/CMakeLists.txt b/third_party/eigen/CMakeLists.txt index dbb9bcf220f..a3ed81c9dde 100644 --- a/third_party/eigen/CMakeLists.txt +++ b/third_party/eigen/CMakeLists.txt @@ -4,7 +4,7 @@ cmake_minimum_required(VERSION 2.8.5) # guard against in-source builds -if(${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_BINARY_DIR}) +if(${PROJECT_SOURCE_DIR} STREQUAL ${CMAKE_CURRENT_BINARY_DIR}) message(FATAL_ERROR "In-source builds not allowed. Please make a new directory (called a build directory) and run CMake from there. You may need to remove CMakeCache.txt. ") endif() @@ -42,11 +42,11 @@ set(EIGEN_MINOR_VERSION "${CMAKE_MATCH_1}") set(EIGEN_VERSION_NUMBER ${EIGEN_WORLD_VERSION}.${EIGEN_MAJOR_VERSION}.${EIGEN_MINOR_VERSION}) # if we are not in a mercurial clone -if(IS_DIRECTORY ${CMAKE_SOURCE_DIR}/.hg) +if(IS_DIRECTORY ${PROJECT_SOURCE_DIR}/.hg) # if the mercurial program is absent or this will leave the EIGEN_HG_CHANGESET string empty, # but won't stop CMake. - execute_process(COMMAND hg tip -R ${CMAKE_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_HGTIP_OUTPUT) - execute_process(COMMAND hg branch -R ${CMAKE_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_BRANCH_OUTPUT) + execute_process(COMMAND hg tip -R ${PROJECT_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_HGTIP_OUTPUT) + execute_process(COMMAND hg branch -R ${PROJECT_SOURCE_DIR} OUTPUT_VARIABLE EIGEN_BRANCH_OUTPUT) endif() # if this is the default (aka development) branch, extract the mercurial changeset number from the hg tip output... @@ -156,9 +156,9 @@ if(NOT MSVC) endif() ei_add_cxx_compiler_flag("-pedantic") ei_add_cxx_compiler_flag("-Wall") - ei_add_cxx_compiler_flag("-Wextra") + #ei_add_cxx_compiler_flag("-Wextra") #ei_add_cxx_compiler_flag("-Weverything") # clang - + ei_add_cxx_compiler_flag("-Wundef") ei_add_cxx_compiler_flag("-Wcast-align") ei_add_cxx_compiler_flag("-Wchar-subscripts") @@ -173,29 +173,29 @@ if(NOT MSVC) ei_add_cxx_compiler_flag("-Wc++11-extensions") ei_add_cxx_compiler_flag("-Wdouble-promotion") # ei_add_cxx_compiler_flag("-Wconversion") - + # -Wshadow is insanely too strict with gcc, hopefully it will become usable with gcc 6 # if(NOT CMAKE_COMPILER_IS_GNUCXX OR (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER "5.0.0")) if(NOT CMAKE_COMPILER_IS_GNUCXX) ei_add_cxx_compiler_flag("-Wshadow") endif() - + ei_add_cxx_compiler_flag("-Wno-psabi") ei_add_cxx_compiler_flag("-Wno-variadic-macros") ei_add_cxx_compiler_flag("-Wno-long-long") - + ei_add_cxx_compiler_flag("-fno-check-new") ei_add_cxx_compiler_flag("-fno-common") ei_add_cxx_compiler_flag("-fstrict-aliasing") ei_add_cxx_compiler_flag("-wd981") # disable ICC's "operands are evaluated in unspecified order" remark ei_add_cxx_compiler_flag("-wd2304") # disable ICC's "warning #2304: non-explicit constructor with single argument may cause implicit type conversion" produced by -Wnon-virtual-dtor - - + + # The -ansi flag must be added last, otherwise it is also used as a linker flag by check_cxx_compiler_flag making it fails # Moreover we should not set both -strict-ansi and -ansi check_cxx_compiler_flag("-strict-ansi" COMPILER_SUPPORT_STRICTANSI) ei_add_cxx_compiler_flag("-Qunused-arguments") # disable clang warning: argument unused during compilation: '-ansi' - + if(COMPILER_SUPPORT_STRICTANSI) set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -strict-ansi") else() @@ -206,7 +206,7 @@ if(NOT MSVC) ei_add_cxx_compiler_flag("-pie") ei_add_cxx_compiler_flag("-fPIE") endif() - + set(CMAKE_REQUIRED_FLAGS "") option(EIGEN_TEST_SSE2 "Enable/Disable SSE2 in tests/examples" OFF) @@ -391,27 +391,22 @@ endif() if(EIGEN_INCLUDE_INSTALL_DIR AND NOT INCLUDE_INSTALL_DIR) set(INCLUDE_INSTALL_DIR ${EIGEN_INCLUDE_INSTALL_DIR} - CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed") + CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed") else() set(INCLUDE_INSTALL_DIR "${CMAKE_INSTALL_INCLUDEDIR}/eigen3" - CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed" + CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen header files are installed" ) endif() set(CMAKEPACKAGE_INSTALL_DIR "${CMAKE_INSTALL_DATADIR}/eigen3/cmake" - CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where Eigen3Config.cmake is installed" + CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where Eigen3Config.cmake is installed" ) set(PKGCONFIG_INSTALL_DIR "${CMAKE_INSTALL_DATADIR}/pkgconfig" - CACHE STRING "The directory relative to CMAKE_PREFIX_PATH where eigen3.pc is installed" + CACHE PATH "The directory relative to CMAKE_PREFIX_PATH where eigen3.pc is installed" ) -foreach(var INCLUDE_INSTALL_DIR CMAKEPACKAGE_INSTALL_DIR PKGCONFIG_INSTALL_DIR) - if(IS_ABSOLUTE "${${var}}") - message(FATAL_ERROR "${var} must be relative to CMAKE_PREFIX_PATH. Got: ${${var}}") - endif() -endforeach() # similar to set_target_properties but append the property instead of overwriting it macro(ei_add_target_property target prop value) @@ -440,8 +435,8 @@ add_subdirectory(Eigen) add_subdirectory(doc EXCLUDE_FROM_ALL) -option(BUILD_TESTING "Enable creation of Eigen tests." ON) -if(BUILD_TESTING) +option(Eigen_BUILD_TESTING "Enable creation of Eigen tests." ON) +if(Eigen_BUILD_TESTING) include(EigenConfigureTesting) if(EIGEN_LEAVE_TEST_IN_ALL_TARGET) @@ -484,7 +479,7 @@ endif(NOT WIN32) configure_file(scripts/cdashtesting.cmake.in cdashtesting.cmake @ONLY) -if(BUILD_TESTING) +if(Eigen_BUILD_TESTING) ei_testing_print_summary() endif() @@ -581,11 +576,11 @@ if (NOT CMAKE_VERSION VERSION_LESS 3.0) else (NOT CMAKE_VERSION VERSION_LESS 3.0) # Fallback to legacy Eigen3Config.cmake without the imported target - + # If CMakePackageConfigHelpers module is available (CMake >= 2.8.8) - # create a relocatable Config file, otherwise leave the hardcoded paths + # create a relocatable Config file, otherwise leave the hardcoded paths include(CMakePackageConfigHelpers OPTIONAL RESULT_VARIABLE CPCH_PATH) - + if(CPCH_PATH) configure_package_config_file ( ${CMAKE_CURRENT_SOURCE_DIR}/cmake/Eigen3ConfigLegacy.cmake.in @@ -594,7 +589,7 @@ else (NOT CMAKE_VERSION VERSION_LESS 3.0) INSTALL_DESTINATION ${CMAKEPACKAGE_INSTALL_DIR} NO_CHECK_REQUIRED_COMPONENTS_MACRO # Eigen does not provide components ) - else() + else() # The PACKAGE_* variables are defined by the configure_package_config_file # but without it we define them manually to the hardcoded paths set(PACKAGE_INIT "") diff --git a/third_party/eigen/COPYING.MINPACK b/third_party/eigen/COPYING.MINPACK index 11d8a9a6c34..ae7984daec9 100644 --- a/third_party/eigen/COPYING.MINPACK +++ b/third_party/eigen/COPYING.MINPACK @@ -1,52 +1,52 @@ -Minpack Copyright Notice (1999) University of Chicago. All rights reserved - -Redistribution and use in source and binary forms, with or -without modification, are permitted provided that the -following conditions are met: - -1. Redistributions of source code must retain the above -copyright notice, this list of conditions and the following -disclaimer. - -2. Redistributions in binary form must reproduce the above -copyright notice, this list of conditions and the following -disclaimer in the documentation and/or other materials -provided with the distribution. - -3. The end-user documentation included with the -redistribution, if any, must include the following -acknowledgment: - - "This product includes software developed by the - University of Chicago, as Operator of Argonne National - Laboratory. - -Alternately, this acknowledgment may appear in the software -itself, if and wherever such third-party acknowledgments -normally appear. - -4. WARRANTY DISCLAIMER. THE SOFTWARE IS SUPPLIED "AS IS" -WITHOUT WARRANTY OF ANY KIND. THE COPYRIGHT HOLDER, THE -UNITED STATES, THE UNITED STATES DEPARTMENT OF ENERGY, AND -THEIR EMPLOYEES: (1) DISCLAIM ANY WARRANTIES, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO ANY IMPLIED WARRANTIES -OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE -OR NON-INFRINGEMENT, (2) DO NOT ASSUME ANY LEGAL LIABILITY -OR RESPONSIBILITY FOR THE ACCURACY, COMPLETENESS, OR -USEFULNESS OF THE SOFTWARE, (3) DO NOT REPRESENT THAT USE OF -THE SOFTWARE WOULD NOT INFRINGE PRIVATELY OWNED RIGHTS, (4) -DO NOT WARRANT THAT THE SOFTWARE WILL FUNCTION -UNINTERRUPTED, THAT IT IS ERROR-FREE OR THAT ANY ERRORS WILL -BE CORRECTED. - -5. LIMITATION OF LIABILITY. IN NO EVENT WILL THE COPYRIGHT -HOLDER, THE UNITED STATES, THE UNITED STATES DEPARTMENT OF -ENERGY, OR THEIR EMPLOYEES: BE LIABLE FOR ANY INDIRECT, -INCIDENTAL, CONSEQUENTIAL, SPECIAL OR PUNITIVE DAMAGES OF -ANY KIND OR NATURE, INCLUDING BUT NOT LIMITED TO LOSS OF -PROFITS OR LOSS OF DATA, FOR ANY REASON WHATSOEVER, WHETHER -SUCH LIABILITY IS ASSERTED ON THE BASIS OF CONTRACT, TORT -(INCLUDING NEGLIGENCE OR STRICT LIABILITY), OR OTHERWISE, -EVEN IF ANY OF SAID PARTIES HAS BEEN WARNED OF THE -POSSIBILITY OF SUCH LOSS OR DAMAGES. - +Minpack Copyright Notice (1999) University of Chicago. All rights reserved + +Redistribution and use in source and binary forms, with or +without modification, are permitted provided that the +following conditions are met: + +1. Redistributions of source code must retain the above +copyright notice, this list of conditions and the following +disclaimer. + +2. Redistributions in binary form must reproduce the above +copyright notice, this list of conditions and the following +disclaimer in the documentation and/or other materials +provided with the distribution. + +3. The end-user documentation included with the +redistribution, if any, must include the following +acknowledgment: + + "This product includes software developed by the + University of Chicago, as Operator of Argonne National + Laboratory. + +Alternately, this acknowledgment may appear in the software +itself, if and wherever such third-party acknowledgments +normally appear. + +4. WARRANTY DISCLAIMER. THE SOFTWARE IS SUPPLIED "AS IS" +WITHOUT WARRANTY OF ANY KIND. THE COPYRIGHT HOLDER, THE +UNITED STATES, THE UNITED STATES DEPARTMENT OF ENERGY, AND +THEIR EMPLOYEES: (1) DISCLAIM ANY WARRANTIES, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO ANY IMPLIED WARRANTIES +OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE +OR NON-INFRINGEMENT, (2) DO NOT ASSUME ANY LEGAL LIABILITY +OR RESPONSIBILITY FOR THE ACCURACY, COMPLETENESS, OR +USEFULNESS OF THE SOFTWARE, (3) DO NOT REPRESENT THAT USE OF +THE SOFTWARE WOULD NOT INFRINGE PRIVATELY OWNED RIGHTS, (4) +DO NOT WARRANT THAT THE SOFTWARE WILL FUNCTION +UNINTERRUPTED, THAT IT IS ERROR-FREE OR THAT ANY ERRORS WILL +BE CORRECTED. + +5. LIMITATION OF LIABILITY. IN NO EVENT WILL THE COPYRIGHT +HOLDER, THE UNITED STATES, THE UNITED STATES DEPARTMENT OF +ENERGY, OR THEIR EMPLOYEES: BE LIABLE FOR ANY INDIRECT, +INCIDENTAL, CONSEQUENTIAL, SPECIAL OR PUNITIVE DAMAGES OF +ANY KIND OR NATURE, INCLUDING BUT NOT LIMITED TO LOSS OF +PROFITS OR LOSS OF DATA, FOR ANY REASON WHATSOEVER, WHETHER +SUCH LIABILITY IS ASSERTED ON THE BASIS OF CONTRACT, TORT +(INCLUDING NEGLIGENCE OR STRICT LIABILITY), OR OTHERWISE, +EVEN IF ANY OF SAID PARTIES HAS BEEN WARNED OF THE +POSSIBILITY OF SUCH LOSS OR DAMAGES. + diff --git a/third_party/eigen/CTestConfig.cmake b/third_party/eigen/CTestConfig.cmake index 45de0e6fcaf..0039bf8acd8 100644 --- a/third_party/eigen/CTestConfig.cmake +++ b/third_party/eigen/CTestConfig.cmake @@ -4,10 +4,10 @@ ## # The following are required to uses Dart and the Cdash dashboard ## ENABLE_TESTING() ## INCLUDE(CTest) -set(CTEST_PROJECT_NAME "Eigen") +set(CTEST_PROJECT_NAME "Eigen 3.3") set(CTEST_NIGHTLY_START_TIME "00:00:00 UTC") set(CTEST_DROP_METHOD "http") -set(CTEST_DROP_SITE "my.cdash.org") -set(CTEST_DROP_LOCATION "/submit.php?project=Eigen") +set(CTEST_DROP_SITE "manao.inria.fr") +set(CTEST_DROP_LOCATION "/CDash/submit.php?project=Eigen+3.3") set(CTEST_DROP_SITE_CDASH TRUE) diff --git a/third_party/eigen/Eigen/Core b/third_party/eigen/Eigen/Core index ac7c5b3004f..b923b8c000c 100644 --- a/third_party/eigen/Eigen/Core +++ b/third_party/eigen/Eigen/Core @@ -279,10 +279,7 @@ #include #include #include -#include -#ifndef EIGEN_NO_IO - #include -#endif +#include #include #include #include @@ -378,9 +375,7 @@ using std::ptrdiff_t; #if defined EIGEN_VECTORIZE_AVX512 #include "src/Core/arch/SSE/PacketMath.h" - #include "src/Core/arch/SSE/MathFunctions.h" #include "src/Core/arch/AVX/PacketMath.h" - #include "src/Core/arch/AVX/MathFunctions.h" #include "src/Core/arch/AVX512/PacketMath.h" #include "src/Core/arch/AVX512/MathFunctions.h" #elif defined EIGEN_VECTORIZE_AVX diff --git a/third_party/eigen/Eigen/Eigenvalues b/third_party/eigen/Eigen/Eigenvalues index 7d6ac787bed..f3f661b074b 100644 --- a/third_party/eigen/Eigen/Eigenvalues +++ b/third_party/eigen/Eigen/Eigenvalues @@ -10,14 +10,14 @@ #include "Core" +#include "src/Core/util/DisableStupidWarnings.h" + #include "Cholesky" #include "Jacobi" #include "Householder" #include "LU" #include "Geometry" -#include "src/Core/util/DisableStupidWarnings.h" - /** \defgroup Eigenvalues_Module Eigenvalues module * * diff --git a/third_party/eigen/Eigen/Geometry b/third_party/eigen/Eigen/Geometry index da88c03bbf5..716d529529a 100644 --- a/third_party/eigen/Eigen/Geometry +++ b/third_party/eigen/Eigen/Geometry @@ -10,12 +10,12 @@ #include "Core" +#include "src/Core/util/DisableStupidWarnings.h" + #include "SVD" #include "LU" #include -#include "src/Core/util/DisableStupidWarnings.h" - /** \defgroup Geometry_Module Geometry module * * This module provides support for: diff --git a/third_party/eigen/Eigen/QR b/third_party/eigen/Eigen/QR index 1be1863a1d5..c7e91446994 100644 --- a/third_party/eigen/Eigen/QR +++ b/third_party/eigen/Eigen/QR @@ -10,12 +10,12 @@ #include "Core" +#include "src/Core/util/DisableStupidWarnings.h" + #include "Cholesky" #include "Jacobi" #include "Householder" -#include "src/Core/util/DisableStupidWarnings.h" - /** \defgroup QR_Module QR module * * diff --git a/third_party/eigen/Eigen/SparseQR b/third_party/eigen/Eigen/SparseQR index f5fc5fa7fee..a6f3b7f7d75 100644 --- a/third_party/eigen/Eigen/SparseQR +++ b/third_party/eigen/Eigen/SparseQR @@ -28,6 +28,7 @@ * */ +#include "OrderingMethods" #include "src/SparseCore/SparseColEtree.h" #include "src/SparseQR/SparseQR.h" diff --git a/third_party/eigen/Eigen/src/Core/ArrayBase.h b/third_party/eigen/Eigen/src/Core/ArrayBase.h index 33f644e2132..3dbc7084cda 100644 --- a/third_party/eigen/Eigen/src/Core/ArrayBase.h +++ b/third_party/eigen/Eigen/src/Core/ArrayBase.h @@ -153,8 +153,8 @@ template class ArrayBase // inline void evalTo(Dest& dst) const { dst = matrix(); } protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(ArrayBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(ArrayBase) + EIGEN_DEVICE_FUNC + ArrayBase() : Base() {} private: explicit ArrayBase(Index); diff --git a/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h b/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h index 5a30fa8df18..27103305629 100644 --- a/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h +++ b/third_party/eigen/Eigen/src/Core/CwiseUnaryView.h @@ -121,8 +121,6 @@ class CwiseUnaryViewImpl { return derived().nestedExpression().outerStride() * sizeof(typename internal::traits::Scalar) / sizeof(Scalar); } - protected: - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(CwiseUnaryViewImpl) }; } // end namespace Eigen diff --git a/third_party/eigen/Eigen/src/Core/DenseBase.h b/third_party/eigen/Eigen/src/Core/DenseBase.h index c55a68230cd..90066ae73fd 100644 --- a/third_party/eigen/Eigen/src/Core/DenseBase.h +++ b/third_party/eigen/Eigen/src/Core/DenseBase.h @@ -40,7 +40,7 @@ static inline void check_DenseIndex_is_signed() { */ template class DenseBase #ifndef EIGEN_PARSED_BY_DOXYGEN - : public DenseCoeffsBase::value> + : public DenseCoeffsBase #else : public DenseCoeffsBase #endif // not EIGEN_PARSED_BY_DOXYGEN @@ -71,7 +71,7 @@ template class DenseBase typedef Scalar value_type; typedef typename NumTraits::Real RealScalar; - typedef DenseCoeffsBase::value> Base; + typedef DenseCoeffsBase Base; using Base::derived; using Base::const_cast_derived; @@ -587,12 +587,11 @@ template class DenseBase } protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(DenseBase) /** Default constructor. Do nothing. */ EIGEN_DEVICE_FUNC DenseBase() { /* Just checks for self-consistency of the flags. - * Only do it when debugging Eigen, as this borders on paranoia and could slow compilation down + * Only do it when debugging Eigen, as this borders on paranoiac and could slow compilation down */ #ifdef EIGEN_INTERNAL_DEBUGGING EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, int(IsRowMajor)) diff --git a/third_party/eigen/Eigen/src/Core/DenseStorage.h b/third_party/eigen/Eigen/src/Core/DenseStorage.h index 7d6d4e66d4b..7958feeb9c0 100644 --- a/third_party/eigen/Eigen/src/Core/DenseStorage.h +++ b/third_party/eigen/Eigen/src/Core/DenseStorage.h @@ -404,7 +404,7 @@ template class DenseStorage(m_data, m_rows*m_cols); - if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + if (size) m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; @@ -479,7 +479,7 @@ template class DenseStorage(m_data, _Rows*m_cols); - if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + if (size) m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; @@ -553,7 +553,7 @@ template class DenseStorage(m_data, _Cols*m_rows); - if (size>0) // >0 and not simply !=0 to let the compiler knows that size cannot be negative + if (size) m_data = internal::conditional_aligned_new_auto(size); else m_data = 0; diff --git a/third_party/eigen/Eigen/src/Core/GenericPacketMath.h b/third_party/eigen/Eigen/src/Core/GenericPacketMath.h index e594437791b..029f8ac36fb 100644 --- a/third_party/eigen/Eigen/src/Core/GenericPacketMath.h +++ b/third_party/eigen/Eigen/src/Core/GenericPacketMath.h @@ -351,7 +351,10 @@ template EIGEN_DEVICE_FUNC inline Packet preverse(const Packet& /** \internal \returns \a a with real and imaginary part flipped (for complex type only) */ template EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet& a) { - return Packet(a.imag(),a.real()); + // FIXME: uncomment the following in case we drop the internal imag and real functions. +// using std::imag; +// using std::real; + return Packet(imag(a),real(a)); } /************************** @@ -521,10 +524,10 @@ inline void palign(PacketType& first, const PacketType& second) #ifndef __CUDACC__ template<> inline std::complex pmul(const std::complex& a, const std::complex& b) -{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } +{ return std::complex(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); } template<> inline std::complex pmul(const std::complex& a, const std::complex& b) -{ return std::complex(a.real()*b.real() - a.imag()*b.imag(), a.imag()*b.real() + a.real()*b.imag()); } +{ return std::complex(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); } #endif diff --git a/third_party/eigen/Eigen/src/Core/MapBase.h b/third_party/eigen/Eigen/src/Core/MapBase.h index 92c3b2818a4..668922ffccb 100644 --- a/third_party/eigen/Eigen/src/Core/MapBase.h +++ b/third_party/eigen/Eigen/src/Core/MapBase.h @@ -182,8 +182,6 @@ template class MapBase #endif protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) template EIGEN_DEVICE_FUNC @@ -296,9 +294,6 @@ template class MapBase // In theory we could simply refer to Base:Base::operator=, but MSVC does not like Base::Base, // see bugs 821 and 920. using ReadOnlyMapBase::Base::operator=; - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(MapBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MapBase) }; #undef EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS diff --git a/third_party/eigen/Eigen/src/Core/MathFunctions.h b/third_party/eigen/Eigen/src/Core/MathFunctions.h index 01736c2a060..b249ce0c8b0 100644 --- a/third_party/eigen/Eigen/src/Core/MathFunctions.h +++ b/third_party/eigen/Eigen/src/Core/MathFunctions.h @@ -287,7 +287,7 @@ struct abs2_impl_default // IsComplex EIGEN_DEVICE_FUNC static inline RealScalar run(const Scalar& x) { - return x.real()*x.real() + x.imag()*x.imag(); + return real(x)*real(x) + imag(x)*imag(x); } }; @@ -313,17 +313,14 @@ struct abs2_retval ****************************************************************************/ template -struct norm1_default_impl; - -template -struct norm1_default_impl +struct norm1_default_impl { typedef typename NumTraits::Real RealScalar; EIGEN_DEVICE_FUNC static inline RealScalar run(const Scalar& x) { EIGEN_USING_STD_MATH(abs); - return abs(x.real()) + abs(x.imag()); + return abs(real(x)) + abs(imag(x)); } }; @@ -665,8 +662,8 @@ struct random_default_impl { static inline Scalar run(const Scalar& x, const Scalar& y) { - return Scalar(random(x.real(), y.real()), - random(x.imag(), y.imag())); + return Scalar(random(real(x), real(y)), + random(imag(x), imag(y))); } static inline Scalar run() { @@ -919,9 +916,6 @@ inline EIGEN_MATHFUNC_RETVAL(abs2, Scalar) abs2(const Scalar& x) return EIGEN_MATHFUNC_IMPL(abs2, Scalar)::run(x); } -EIGEN_DEVICE_FUNC -inline bool abs2(bool x) { return x; } - template EIGEN_DEVICE_FUNC inline EIGEN_MATHFUNC_RETVAL(norm1, Scalar) norm1(const Scalar& x) diff --git a/third_party/eigen/Eigen/src/Core/MatrixBase.h b/third_party/eigen/Eigen/src/Core/MatrixBase.h index f8bcc8c6f56..e6c35907c39 100644 --- a/third_party/eigen/Eigen/src/Core/MatrixBase.h +++ b/third_party/eigen/Eigen/src/Core/MatrixBase.h @@ -464,8 +464,7 @@ template class MatrixBase EIGEN_MATRIX_FUNCTION_1(MatrixComplexPowerReturnValue, pow, power to \c p, const std::complex& p) protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(MatrixBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(MatrixBase) + EIGEN_DEVICE_FUNC MatrixBase() : Base() {} private: EIGEN_DEVICE_FUNC explicit MatrixBase(int); diff --git a/third_party/eigen/Eigen/src/Core/PermutationMatrix.h b/third_party/eigen/Eigen/src/Core/PermutationMatrix.h index 47c06ba7706..b1fb455b98c 100644 --- a/third_party/eigen/Eigen/src/Core/PermutationMatrix.h +++ b/third_party/eigen/Eigen/src/Core/PermutationMatrix.h @@ -87,6 +87,17 @@ class PermutationBase : public EigenBase return derived(); } + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + Derived& operator=(const PermutationBase& other) + { + indices() = other.indices(); + return derived(); + } + #endif + /** \returns the number of rows */ inline Index rows() const { return Index(indices().size()); } @@ -322,6 +333,12 @@ class PermutationMatrix : public PermutationBase& other) : m_indices(other.indices()) {} + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** Standard copy constructor. Defined only to prevent a default copy constructor + * from hiding the other templated constructor */ + inline PermutationMatrix(const PermutationMatrix& other) : m_indices(other.indices()) {} + #endif + /** Generic constructor from expression of the indices. The indices * array has the meaning that the permutations sends each integer i to indices[i]. * @@ -356,6 +373,17 @@ class PermutationMatrix : public PermutationBase::type EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if::type* = 0) { - const bool t0_is_integer_alike = internal::is_valid_index_type::value; - const bool t1_is_integer_alike = internal::is_valid_index_type::value; - EIGEN_STATIC_ASSERT(t0_is_integer_alike && - t1_is_integer_alike, + EIGEN_STATIC_ASSERT(bool(NumTraits::IsInteger) && + bool(NumTraits::IsInteger), FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) resize(rows,cols); } @@ -775,9 +773,9 @@ class PlainObjectBase : public internal::dense_xpr_base::type && ((!internal::is_same::XprKind,ArrayXpr>::value || Base::SizeAtCompileTime==Dynamic)),T>::type* = 0) { // NOTE MSVC 2008 complains if we directly put bool(NumTraits::IsInteger) as the EIGEN_STATIC_ASSERT argument. - const bool is_integer_alike = internal::is_valid_index_type::value; - EIGEN_UNUSED_VARIABLE(is_integer_alike); - EIGEN_STATIC_ASSERT(is_integer_alike, + const bool is_integer = NumTraits::IsInteger; + EIGEN_UNUSED_VARIABLE(is_integer); + EIGEN_STATIC_ASSERT(is_integer, FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED) resize(size); } diff --git a/third_party/eigen/Eigen/src/Core/ProductEvaluators.h b/third_party/eigen/Eigen/src/Core/ProductEvaluators.h index bce1310c96f..9b99bd76962 100644 --- a/third_party/eigen/Eigen/src/Core/ProductEvaluators.h +++ b/third_party/eigen/Eigen/src/Core/ProductEvaluators.h @@ -396,7 +396,7 @@ struct generic_product_impl // but easier on the compiler side call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::assign_op()); } - + template static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { @@ -410,32 +410,6 @@ struct generic_product_impl // dst.noalias() -= lhs.lazyProduct(rhs); call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::sub_assign_op()); } - - // Catch "dst {,+,-}= (s*A)*B" and evaluate it lazily by moving out the scalar factor: - // dst {,+,-}= s * (A.lazyProduct(B)) - // This is a huge benefit for heap-allocated matrix types as it save one costly allocation. - // For them, this strategy is also faster than simply by-passing the heap allocation through - // stack allocation. - // For fixed sizes matrices, this is less obvious, it is sometimes x2 faster, but sometimes x3 slower, - // and the behavior depends also a lot on the compiler... so let's be conservative and enable them for dynamic-size only, - // that is when coming from generic_product_impl<...,GemmProduct> in file GeneralMatrixMatrix.h - template - static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE - void eval_dynamic(Dst& dst, const CwiseBinaryOp, - const CwiseNullaryOp, Plain1>, Xpr2>& lhs, const Rhs& rhs, const Func &func) - { - call_assignment_no_alias(dst, lhs.lhs().functor().m_other * lhs.rhs().lazyProduct(rhs), func); - } - - // Here, we we always have LhsT==Lhs, but we need to make it a template type to make the above - // overload more specialized. - template - static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE - void eval_dynamic(Dst& dst, const LhsT& lhs, const Rhs& rhs, const Func &func) - { - call_assignment_no_alias(dst, lhs.lazyProduct(rhs), func); - } - // template // static inline void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha) diff --git a/third_party/eigen/Eigen/src/Core/Ref.h b/third_party/eigen/Eigen/src/Core/Ref.h index 17a1496b842..9c6e3c5d9b0 100644 --- a/third_party/eigen/Eigen/src/Core/Ref.h +++ b/third_party/eigen/Eigen/src/Core/Ref.h @@ -28,13 +28,12 @@ struct traits > template struct match { enum { - IsVectorAtCompileTime = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime, HasDirectAccess = internal::has_direct_access::ret, - StorageOrderMatch = IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), + StorageOrderMatch = PlainObjectType::IsVectorAtCompileTime || Derived::IsVectorAtCompileTime || ((PlainObjectType::Flags&RowMajorBit)==(Derived::Flags&RowMajorBit)), InnerStrideMatch = int(StrideType::InnerStrideAtCompileTime)==int(Dynamic) || int(StrideType::InnerStrideAtCompileTime)==int(Derived::InnerStrideAtCompileTime) || (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1), - OuterStrideMatch = IsVectorAtCompileTime + OuterStrideMatch = Derived::IsVectorAtCompileTime || int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime), // NOTE, this indirection of evaluator::Alignment is needed // to workaround a very strange bug in MSVC related to the instantiation diff --git a/third_party/eigen/Eigen/src/Core/SolveTriangular.h b/third_party/eigen/Eigen/src/Core/SolveTriangular.h index fd0acb1a58e..4652e2e19ff 100644 --- a/third_party/eigen/Eigen/src/Core/SolveTriangular.h +++ b/third_party/eigen/Eigen/src/Core/SolveTriangular.h @@ -19,7 +19,7 @@ namespace internal { template struct triangular_solve_vector; -template +template struct triangular_solve_matrix; // small helper struct extracting some traits on the underlying solver operation @@ -98,8 +98,8 @@ struct triangular_solver_selector BlockingType blocking(rhs.rows(), rhs.cols(), size, 1, false); triangular_solve_matrix - ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.innerStride(), rhs.outerStride(), blocking); + (Rhs::Flags&RowMajorBit) ? RowMajor : ColMajor> + ::run(size, othersize, &actualLhs.coeffRef(0,0), actualLhs.outerStride(), &rhs.coeffRef(0,0), rhs.outerStride(), blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/Transpose.h b/third_party/eigen/Eigen/src/Core/Transpose.h index 960dc4510c7..79b767bccae 100644 --- a/third_party/eigen/Eigen/src/Core/Transpose.h +++ b/third_party/eigen/Eigen/src/Core/Transpose.h @@ -146,8 +146,6 @@ template class TransposeImpl { return derived().nestedExpression().coeffRef(index); } - protected: - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TransposeImpl) }; /** \returns an expression of the transpose of *this. diff --git a/third_party/eigen/Eigen/src/Core/Transpositions.h b/third_party/eigen/Eigen/src/Core/Transpositions.h index 7718625e80b..86da5af5936 100644 --- a/third_party/eigen/Eigen/src/Core/Transpositions.h +++ b/third_party/eigen/Eigen/src/Core/Transpositions.h @@ -33,6 +33,17 @@ class TranspositionsBase indices() = other.indices(); return derived(); } + + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** This is a special case of the templated operator=. Its purpose is to + * prevent a default operator= from hiding the templated operator=. + */ + Derived& operator=(const TranspositionsBase& other) + { + indices() = other.indices(); + return derived(); + } + #endif /** \returns the number of transpositions */ Index size() const { return indices().size(); } @@ -160,6 +171,12 @@ class Transpositions : public TranspositionsBase& other) : m_indices(other.indices()) {} + #ifndef EIGEN_PARSED_BY_DOXYGEN + /** Standard copy constructor. Defined only to prevent a default copy constructor + * from hiding the other templated constructor */ + inline Transpositions(const Transpositions& other) : m_indices(other.indices()) {} + #endif + /** Generic constructor from expression of the transposition indices. */ template explicit inline Transpositions(const MatrixBase& indices) : m_indices(indices) @@ -172,6 +189,17 @@ class Transpositions : public TranspositionsBase class TriangularView explicit inline TriangularView(MatrixType& matrix) : m_matrix(matrix) {} - EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TriangularView) + using Base::operator=; + TriangularView& operator=(const TriangularView &other) + { return Base::operator=(other); } /** \copydoc EigenBase::rows() */ EIGEN_DEVICE_FUNC @@ -542,10 +544,6 @@ template class TriangularViewImpl<_Mat template EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TriangularViewType& _assignProduct(const ProductType& prod, const Scalar& alpha, bool beta); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(TriangularViewImpl) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(TriangularViewImpl) - }; /*************************************************************************** diff --git a/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h b/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h index b259c1e1f92..9c1717f76dd 100644 --- a/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h +++ b/third_party/eigen/Eigen/src/Core/arch/AVX512/MathFunctions.h @@ -29,7 +29,6 @@ namespace internal { #define _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(NAME, X) \ const Packet8d p8d_##NAME = _mm512_castsi512_pd(_mm512_set1_epi64(X)) - // Natural logarithm // Computes log(x) as log(2^e * m) = C*e + log(m), where the constant C =log(2) // and m is in the range [sqrt(1/2),sqrt(2)). In this range, the logarithm can @@ -48,7 +47,6 @@ plog(const Packet16f& _x) { // The smallest non denormalized float number. _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(min_norm_pos, 0x00800000); _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(minus_inf, 0xff800000); - _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(pos_inf, 0x7f800000); _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(nan, 0x7fc00000); // Polynomial coefficients. @@ -66,9 +64,11 @@ plog(const Packet16f& _x) { _EIGEN_DECLARE_CONST_Packet16f(cephes_log_q2, 0.693359375f); // invalid_mask is set to true when x is NaN - __mmask16 invalid_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_NGE_UQ); - __mmask16 iszero_mask = _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_EQ_OQ); - + __mmask16 invalid_mask = + _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_NGE_UQ); + __mmask16 iszero_mask = + _mm512_cmp_ps_mask(x, _mm512_setzero_ps(), _CMP_EQ_UQ); + // Truncate input values to the minimum positive normal. x = pmax(x, p16f_min_norm_pos); @@ -118,18 +118,11 @@ plog(const Packet16f& _x) { x = padd(x, y); x = padd(x, y2); - __mmask16 pos_inf_mask = _mm512_cmp_ps_mask(_x,p16f_pos_inf,_CMP_EQ_OQ); - // Filter out invalid inputs, i.e.: - // - negative arg will be NAN, - // - 0 will be -INF. - // - +INF will be +INF + // Filter out invalid inputs, i.e. negative arg will be NAN, 0 will be -INF. return _mm512_mask_blend_ps(iszero_mask, - _mm512_mask_blend_ps(invalid_mask, - _mm512_mask_blend_ps(pos_inf_mask,x,p16f_pos_inf), - p16f_nan), - p16f_minus_inf); + _mm512_mask_blend_ps(invalid_mask, x, p16f_nan), + p16f_minus_inf); } - #endif // Exponential function. Works by writing "x = m*log(2) + r" where @@ -265,39 +258,48 @@ pexp(const Packet8d& _x) { template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet16f psqrt(const Packet16f& _x) { - Packet16f neg_half = pmul(_x, pset1(-.5f)); - __mmask16 denormal_mask = _mm512_kand( - _mm512_cmp_ps_mask(_x, pset1((std::numeric_limits::min)()), - _CMP_LT_OQ), - _mm512_cmp_ps_mask(_x, _mm512_setzero_ps(), _CMP_GE_OQ)); + _EIGEN_DECLARE_CONST_Packet16f(one_point_five, 1.5f); + _EIGEN_DECLARE_CONST_Packet16f(minus_half, -0.5f); + _EIGEN_DECLARE_CONST_Packet16f_FROM_INT(flt_min, 0x00800000); - Packet16f x = _mm512_rsqrt14_ps(_x); + Packet16f neg_half = pmul(_x, p16f_minus_half); + + // select only the inverse sqrt of positive normal inputs (denormals are + // flushed to zero and cause infs as well). + __mmask16 non_zero_mask = _mm512_cmp_ps_mask(_x, p16f_flt_min, _CMP_GE_OQ); + Packet16f x = _mm512_mask_blend_ps(non_zero_mask, _mm512_setzero_ps(), _mm512_rsqrt14_ps(_x)); // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5f))); + x = pmul(x, pmadd(neg_half, pmul(x, x), p16f_one_point_five)); - // Flush results for denormals to zero. - return _mm512_mask_blend_ps(denormal_mask, pmul(_x,x), _mm512_setzero_ps()); + // Multiply the original _x by it's reciprocal square root to extract the + // square root. + return pmul(_x, x); } template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS EIGEN_UNUSED Packet8d psqrt(const Packet8d& _x) { - Packet8d neg_half = pmul(_x, pset1(-.5)); - __mmask16 denormal_mask = _mm512_kand( - _mm512_cmp_pd_mask(_x, pset1((std::numeric_limits::min)()), - _CMP_LT_OQ), - _mm512_cmp_pd_mask(_x, _mm512_setzero_pd(), _CMP_GE_OQ)); + _EIGEN_DECLARE_CONST_Packet8d(one_point_five, 1.5); + _EIGEN_DECLARE_CONST_Packet8d(minus_half, -0.5); + _EIGEN_DECLARE_CONST_Packet8d_FROM_INT64(dbl_min, 0x0010000000000000LL); - Packet8d x = _mm512_rsqrt14_pd(_x); + Packet8d neg_half = pmul(_x, p8d_minus_half); - // Do a single step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); + // select only the inverse sqrt of positive normal inputs (denormals are + // flushed to zero and cause infs as well). + __mmask8 non_zero_mask = _mm512_cmp_pd_mask(_x, p8d_dbl_min, _CMP_GE_OQ); + Packet8d x = _mm512_mask_blend_pd(non_zero_mask, _mm512_setzero_pd(), _mm512_rsqrt14_pd(_x)); + + // Do a first step of Newton's iteration. + x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); // Do a second step of Newton's iteration. - x = pmul(x, pmadd(neg_half, pmul(x, x), pset1(1.5))); + x = pmul(x, pmadd(neg_half, pmul(x, x), p8d_one_point_five)); - return _mm512_mask_blend_pd(denormal_mask, pmul(_x,x), _mm512_setzero_pd()); + // Multiply the original _x by it's reciprocal square root to extract the + // square root. + return pmul(_x, x); } #else template <> diff --git a/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h b/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h index 000b7762ff2..5adddc7aec4 100644 --- a/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h +++ b/third_party/eigen/Eigen/src/Core/arch/AVX512/PacketMath.h @@ -19,10 +19,10 @@ namespace internal { #endif #ifndef EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS -#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS 32 +#define EIGEN_ARCH_DEFAULT_NUMBER_OF_REGISTERS (2*sizeof(void*)) #endif -#ifdef EIGEN_VECTORIZE_FMA +#ifdef __FMA__ #ifndef EIGEN_HAS_SINGLE_INSTRUCTION_MADD #define EIGEN_HAS_SINGLE_INSTRUCTION_MADD #endif @@ -54,14 +54,13 @@ template<> struct packet_traits : default_packet_traits AlignedOnScalar = 1, size = 16, HasHalfPacket = 1, - HasBlend = 0, -#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) +#if EIGEN_GNUC_AT_LEAST(5, 3) #ifdef EIGEN_VECTORIZE_AVX512DQ HasLog = 1, #endif HasExp = 1, - HasSqrt = EIGEN_FAST_MATH, - HasRsqrt = EIGEN_FAST_MATH, + HasSqrt = 1, + HasRsqrt = 1, #endif HasDiv = 1 }; @@ -75,8 +74,8 @@ template<> struct packet_traits : default_packet_traits AlignedOnScalar = 1, size = 8, HasHalfPacket = 1, -#if EIGEN_GNUC_AT_LEAST(5, 3) || (!EIGEN_COMP_GNUC_STRICT) - HasSqrt = EIGEN_FAST_MATH, +#if EIGEN_GNUC_AT_LEAST(5, 3) + HasSqrt = 1, HasRsqrt = EIGEN_FAST_MATH, #endif HasDiv = 1 @@ -99,7 +98,6 @@ template <> struct unpacket_traits { typedef float type; typedef Packet8f half; - typedef Packet16i integer_packet; enum { size = 16, alignment=Aligned64 }; }; template <> @@ -134,7 +132,7 @@ EIGEN_STRONG_INLINE Packet16f pload1(const float* from) { } template <> EIGEN_STRONG_INLINE Packet8d pload1(const double* from) { - return _mm512_set1_pd(*from); + return _mm512_broadcastsd_pd(_mm_load_pd1(from)); } template <> @@ -160,11 +158,6 @@ EIGEN_STRONG_INLINE Packet8d padd(const Packet8d& a, const Packet8d& b) { return _mm512_add_pd(a, b); } -template <> -EIGEN_STRONG_INLINE Packet16i padd(const Packet16i& a, - const Packet16i& b) { - return _mm512_add_epi32(a, b); -} template <> EIGEN_STRONG_INLINE Packet16f psub(const Packet16f& a, @@ -176,11 +169,6 @@ EIGEN_STRONG_INLINE Packet8d psub(const Packet8d& a, const Packet8d& b) { return _mm512_sub_pd(a, b); } -template <> -EIGEN_STRONG_INLINE Packet16i psub(const Packet16i& a, - const Packet16i& b) { - return _mm512_sub_epi32(a, b); -} template <> EIGEN_STRONG_INLINE Packet16f pnegate(const Packet16f& a) { @@ -214,11 +202,6 @@ EIGEN_STRONG_INLINE Packet8d pmul(const Packet8d& a, const Packet8d& b) { return _mm512_mul_pd(a, b); } -template <> -EIGEN_STRONG_INLINE Packet16i pmul(const Packet16i& a, - const Packet16i& b) { - return _mm512_mul_epi32(a, b); -} template <> EIGEN_STRONG_INLINE Packet16f pdiv(const Packet16f& a, @@ -231,7 +214,7 @@ EIGEN_STRONG_INLINE Packet8d pdiv(const Packet8d& a, return _mm512_div_pd(a, b); } -#ifdef EIGEN_VECTORIZE_FMA +#ifdef __FMA__ template <> EIGEN_STRONG_INLINE Packet16f pmadd(const Packet16f& a, const Packet16f& b, const Packet16f& c) { @@ -247,73 +230,23 @@ EIGEN_STRONG_INLINE Packet8d pmadd(const Packet8d& a, const Packet8d& b, template <> EIGEN_STRONG_INLINE Packet16f pmin(const Packet16f& a, const Packet16f& b) { - // Arguments are reversed to match NaN propagation behavior of std::min. - return _mm512_min_ps(b, a); + return _mm512_min_ps(a, b); } template <> EIGEN_STRONG_INLINE Packet8d pmin(const Packet8d& a, const Packet8d& b) { - // Arguments are reversed to match NaN propagation behavior of std::min. - return _mm512_min_pd(b, a); + return _mm512_min_pd(a, b); } template <> EIGEN_STRONG_INLINE Packet16f pmax(const Packet16f& a, const Packet16f& b) { - // Arguments are reversed to match NaN propagation behavior of std::max. - return _mm512_max_ps(b, a); + return _mm512_max_ps(a, b); } template <> EIGEN_STRONG_INLINE Packet8d pmax(const Packet8d& a, const Packet8d& b) { - // Arguments are reversed to match NaN propagation behavior of std::max. - return _mm512_max_pd(b, a); -} - -#ifdef EIGEN_VECTORIZE_AVX512DQ -template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { return _mm512_extractf32x8_ps(x,I_); } -template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { return _mm512_extractf64x2_pd(x,I_); } -EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { return _mm512_insertf32x8(_mm512_castps256_ps512(a),b,1); } -#else -// AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 -template EIGEN_STRONG_INLINE Packet8f extract256(Packet16f x) { - return _mm256_castsi256_ps(_mm512_extracti64x4_epi64( _mm512_castps_si512(x),I_)); -} - -// AVX512F does not define _mm512_extractf64x2_pd to extract _m128 from _m512 -template EIGEN_STRONG_INLINE Packet2d extract128(Packet8d x) { - return _mm_castsi128_pd(_mm512_extracti32x4_epi32( _mm512_castpd_si512(x),I_)); -} - -EIGEN_STRONG_INLINE Packet16f cat256(Packet8f a, Packet8f b) { - return _mm512_castsi512_ps(_mm512_inserti64x4(_mm512_castsi256_si512(_mm256_castps_si256(a)), - _mm256_castps_si256(b),1)); -} -#endif - -// Helper function for bit packing snippet of low precision comparison. -// It packs the flags from 32x16 to 16x16. -EIGEN_STRONG_INLINE __m256i Pack32To16(Packet16f rf) { - // Split data into small pieces and handle with AVX instructions - // to guarantee internal order of vector. - // Operation: - // dst[15:0] := Saturate16(rf[31:0]) - // dst[31:16] := Saturate16(rf[63:32]) - // ... - // dst[255:240] := Saturate16(rf[255:224]) - __m256i lo = _mm256_castps_si256(extract256<0>(rf)); - __m256i hi = _mm256_castps_si256(extract256<1>(rf)); - __m128i result_lo = _mm_packs_epi32(_mm256_extractf128_si256(lo, 0), - _mm256_extractf128_si256(lo, 1)); - __m128i result_hi = _mm_packs_epi32(_mm256_extractf128_si256(hi, 0), - _mm256_extractf128_si256(hi, 1)); - return _mm256_insertf128_si256(_mm256_castsi128_si256(result_lo), result_hi, 1); -} - -template <> -EIGEN_STRONG_INLINE Packet16i pand(const Packet16i& a, - const Packet16i& b) { - return _mm512_and_si512(a,b); + return _mm512_max_pd(a, b); } template <> @@ -322,7 +255,24 @@ EIGEN_STRONG_INLINE Packet16f pand(const Packet16f& a, #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_and_ps(a, b); #else - return _mm512_castsi512_ps(pand(_mm512_castps_si512(a),_mm512_castps_si512(b))); + Packet16f res = _mm512_undefined_ps(); + Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); + Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); + res = _mm512_insertf32x4(res, _mm_and_ps(lane0_a, lane0_b), 0); + + Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); + Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); + res = _mm512_insertf32x4(res, _mm_and_ps(lane1_a, lane1_b), 1); + + Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); + Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); + res = _mm512_insertf32x4(res, _mm_and_ps(lane2_a, lane2_b), 2); + + Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); + Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); + res = _mm512_insertf32x4(res, _mm_and_ps(lane3_a, lane3_b), 3); + + return res; #endif } template <> @@ -338,21 +288,35 @@ EIGEN_STRONG_INLINE Packet8d pand(const Packet8d& a, Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); - return _mm512_insertf64x4(res, _mm256_and_pd(lane1_a, lane1_b), 1); -#endif -} + res = _mm512_insertf64x4(res, _mm256_and_pd(lane1_a, lane1_b), 1); -template <> -EIGEN_STRONG_INLINE Packet16i por(const Packet16i& a, const Packet16i& b) { - return _mm512_or_si512(a, b); + return res; +#endif } - template <> -EIGEN_STRONG_INLINE Packet16f por(const Packet16f& a, const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16f por(const Packet16f& a, + const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_or_ps(a, b); #else - return _mm512_castsi512_ps(por(_mm512_castps_si512(a),_mm512_castps_si512(b))); + Packet16f res = _mm512_undefined_ps(); + Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); + Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); + res = _mm512_insertf32x4(res, _mm_or_ps(lane0_a, lane0_b), 0); + + Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); + Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); + res = _mm512_insertf32x4(res, _mm_or_ps(lane1_a, lane1_b), 1); + + Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); + Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); + res = _mm512_insertf32x4(res, _mm_or_ps(lane2_a, lane2_b), 2); + + Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); + Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); + res = _mm512_insertf32x4(res, _mm_or_ps(lane3_a, lane3_b), 3); + + return res; #endif } @@ -362,65 +326,107 @@ EIGEN_STRONG_INLINE Packet8d por(const Packet8d& a, #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_or_pd(a, b); #else - return _mm512_castsi512_pd(por(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} + Packet8d res = _mm512_undefined_pd(); + Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); + Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); + res = _mm512_insertf64x4(res, _mm256_or_pd(lane0_a, lane0_b), 0); -template <> -EIGEN_STRONG_INLINE Packet16i pxor(const Packet16i& a, const Packet16i& b) { - return _mm512_xor_si512(a, b); + Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); + Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); + res = _mm512_insertf64x4(res, _mm256_or_pd(lane1_a, lane1_b), 1); + + return res; +#endif } template <> -EIGEN_STRONG_INLINE Packet16f pxor(const Packet16f& a, const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16f pxor(const Packet16f& a, + const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_xor_ps(a, b); #else - return _mm512_castsi512_ps(pxor(_mm512_castps_si512(a),_mm512_castps_si512(b))); + Packet16f res = _mm512_undefined_ps(); + Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); + Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); + res = _mm512_insertf32x4(res, _mm_xor_ps(lane0_a, lane0_b), 0); + + Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); + Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); + res = _mm512_insertf32x4(res, _mm_xor_ps(lane1_a, lane1_b), 1); + + Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); + Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); + res = _mm512_insertf32x4(res, _mm_xor_ps(lane2_a, lane2_b), 2); + + Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); + Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); + res = _mm512_insertf32x4(res, _mm_xor_ps(lane3_a, lane3_b), 3); + + return res; #endif } - template <> -EIGEN_STRONG_INLINE Packet8d pxor(const Packet8d& a, const Packet8d& b) { +EIGEN_STRONG_INLINE Packet8d pxor(const Packet8d& a, + const Packet8d& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ return _mm512_xor_pd(a, b); #else - return _mm512_castsi512_pd(pxor(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} + Packet8d res = _mm512_undefined_pd(); + Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); + Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); + res = _mm512_insertf64x4(res, _mm256_xor_pd(lane0_a, lane0_b), 0); -template <> -EIGEN_STRONG_INLINE Packet16i pandnot(const Packet16i& a, const Packet16i& b) { - return _mm512_andnot_si512(b, a); + Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); + Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); + res = _mm512_insertf64x4(res, _mm256_xor_pd(lane1_a, lane1_b), 1); + + return res; +#endif } template <> -EIGEN_STRONG_INLINE Packet16f pandnot(const Packet16f& a, const Packet16f& b) { +EIGEN_STRONG_INLINE Packet16f pandnot(const Packet16f& a, + const Packet16f& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_ps(b, a); + return _mm512_andnot_ps(a, b); #else - return _mm512_castsi512_ps(pandnot(_mm512_castps_si512(a),_mm512_castps_si512(b))); + Packet16f res = _mm512_undefined_ps(); + Packet4f lane0_a = _mm512_extractf32x4_ps(a, 0); + Packet4f lane0_b = _mm512_extractf32x4_ps(b, 0); + res = _mm512_insertf32x4(res, _mm_andnot_ps(lane0_a, lane0_b), 0); + + Packet4f lane1_a = _mm512_extractf32x4_ps(a, 1); + Packet4f lane1_b = _mm512_extractf32x4_ps(b, 1); + res = _mm512_insertf32x4(res, _mm_andnot_ps(lane1_a, lane1_b), 1); + + Packet4f lane2_a = _mm512_extractf32x4_ps(a, 2); + Packet4f lane2_b = _mm512_extractf32x4_ps(b, 2); + res = _mm512_insertf32x4(res, _mm_andnot_ps(lane2_a, lane2_b), 2); + + Packet4f lane3_a = _mm512_extractf32x4_ps(a, 3); + Packet4f lane3_b = _mm512_extractf32x4_ps(b, 3); + res = _mm512_insertf32x4(res, _mm_andnot_ps(lane3_a, lane3_b), 3); + + return res; #endif } template <> -EIGEN_STRONG_INLINE Packet8d pandnot(const Packet8d& a,const Packet8d& b) { +EIGEN_STRONG_INLINE Packet8d pandnot(const Packet8d& a, + const Packet8d& b) { #ifdef EIGEN_VECTORIZE_AVX512DQ - return _mm512_andnot_pd(b, a); + return _mm512_andnot_pd(a, b); #else - return _mm512_castsi512_pd(pandnot(_mm512_castpd_si512(a),_mm512_castpd_si512(b))); -#endif -} - -template EIGEN_STRONG_INLINE Packet16i parithmetic_shift_right(Packet16i a) { - return _mm512_srai_epi32(a, N); -} + Packet8d res = _mm512_undefined_pd(); + Packet4d lane0_a = _mm512_extractf64x4_pd(a, 0); + Packet4d lane0_b = _mm512_extractf64x4_pd(b, 0); + res = _mm512_insertf64x4(res, _mm256_andnot_pd(lane0_a, lane0_b), 0); -template EIGEN_STRONG_INLINE Packet16i plogical_shift_right(Packet16i a) { - return _mm512_srli_epi32(a, N); -} + Packet4d lane1_a = _mm512_extractf64x4_pd(a, 1); + Packet4d lane1_b = _mm512_extractf64x4_pd(b, 1); + res = _mm512_insertf64x4(res, _mm256_andnot_pd(lane1_a, lane1_b), 1); -template EIGEN_STRONG_INLINE Packet16i plogical_shift_left(Packet16i a) { - return _mm512_slli_epi32(a, N); + return res; +#endif } template <> @@ -455,55 +461,75 @@ EIGEN_STRONG_INLINE Packet16i ploadu(const int* from) { // {a0, a0 a1, a1, a2, a2, a3, a3, a4, a4, a5, a5, a6, a6, a7, a7} template <> EIGEN_STRONG_INLINE Packet16f ploaddup(const float* from) { - // an unaligned load is required here as there is no requirement - // on the alignment of input pointer 'from' - __m256i low_half = _mm256_loadu_si256(reinterpret_cast(from)); - __m512 even_elements = _mm512_castsi512_ps(_mm512_cvtepu32_epi64(low_half)); - __m512 pairs = _mm512_permute_ps(even_elements, _MM_SHUFFLE(2, 2, 0, 0)); - return pairs; -} + Packet8f lane0 = _mm256_broadcast_ps((const __m128*)(const void*)from); + // mimic an "inplace" permutation of the lower 128bits using a blend + lane0 = _mm256_blend_ps( + lane0, _mm256_castps128_ps256(_mm_permute_ps( + _mm256_castps256_ps128(lane0), _MM_SHUFFLE(1, 0, 1, 0))), + 15); + // then we can perform a consistent permutation on the global register to get + // everything in shape: + lane0 = _mm256_permute_ps(lane0, _MM_SHUFFLE(3, 3, 2, 2)); + + Packet8f lane1 = _mm256_broadcast_ps((const __m128*)(const void*)(from + 4)); + // mimic an "inplace" permutation of the lower 128bits using a blend + lane1 = _mm256_blend_ps( + lane1, _mm256_castps128_ps256(_mm_permute_ps( + _mm256_castps256_ps128(lane1), _MM_SHUFFLE(1, 0, 1, 0))), + 15); + // then we can perform a consistent permutation on the global register to get + // everything in shape: + lane1 = _mm256_permute_ps(lane1, _MM_SHUFFLE(3, 3, 2, 2)); #ifdef EIGEN_VECTORIZE_AVX512DQ -// FIXME: this does not look optimal, better load a Packet4d and shuffle... + Packet16f res = _mm512_undefined_ps(); + return _mm512_insertf32x8(res, lane0, 0); + return _mm512_insertf32x8(res, lane1, 1); + return res; +#else + Packet16f res = _mm512_undefined_ps(); + res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane0, 0), 0); + res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane0, 1), 1); + res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane1, 0), 2); + res = _mm512_insertf32x4(res, _mm256_extractf128_ps(lane1, 1), 3); + return res; +#endif +} // Loads 4 doubles from memory a returns the packet {a0, a0 a1, a1, a2, a2, a3, // a3} template <> EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { - __m512d x = _mm512_setzero_pd(); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[0]), 0); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[1]), 1); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[2]), 2); - x = _mm512_insertf64x2(x, _mm_loaddup_pd(&from[3]), 3); - return x; -} -#else -template <> -EIGEN_STRONG_INLINE Packet8d ploaddup(const double* from) { - __m512d x = _mm512_setzero_pd(); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<0, _mm_load_sd(from+0)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<2, _mm_load_sd(from+1)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<4, _mm_load_sd(from+2)); - x = _mm512_mask_broadcastsd_pd(x, 0x3<<6, _mm_load_sd(from+3)); - return x; + Packet4d lane0 = _mm256_broadcast_pd((const __m128d*)(const void*)from); + lane0 = _mm256_permute_pd(lane0, 3 << 2); + + Packet4d lane1 = _mm256_broadcast_pd((const __m128d*)(const void*)(from + 2)); + lane1 = _mm256_permute_pd(lane1, 3 << 2); + + Packet8d res = _mm512_undefined_pd(); + res = _mm512_insertf64x4(res, lane0, 0); + return _mm512_insertf64x4(res, lane1, 1); } -#endif // Loads 4 floats from memory a returns the packet // {a0, a0 a0, a0, a1, a1, a1, a1, a2, a2, a2, a2, a3, a3, a3, a3} template <> EIGEN_STRONG_INLINE Packet16f ploadquad(const float* from) { - Packet16f tmp = _mm512_castps128_ps512(ploadu(from)); - const Packet16i scatter_mask = _mm512_set_epi32(3,3,3,3, 2,2,2,2, 1,1,1,1, 0,0,0,0); - return _mm512_permutexvar_ps(scatter_mask, tmp); + Packet16f tmp = _mm512_undefined_ps(); + tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from), 0); + tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 1), 1); + tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 2), 2); + tmp = _mm512_insertf32x4(tmp, _mm_load_ps1(from + 3), 3); + return tmp; } - // Loads 2 doubles from memory a returns the packet // {a0, a0 a0, a0, a1, a1, a1, a1} template <> EIGEN_STRONG_INLINE Packet8d ploadquad(const double* from) { - __m256d lane0 = _mm256_set1_pd(*from); - __m256d lane1 = _mm256_set1_pd(*(from+1)); - __m512d tmp = _mm512_undefined_pd(); + Packet8d tmp = _mm512_undefined_pd(); + Packet2d tmp0 = _mm_load_pd1(from); + Packet2d tmp1 = _mm_load_pd1(from + 1); + Packet4d lane0 = _mm256_broadcastsd_pd(tmp0); + Packet4d lane1 = _mm256_broadcastsd_pd(tmp1); tmp = _mm512_insertf64x4(tmp, lane0, 0); return _mm512_insertf64x4(tmp, lane1, 1); } @@ -539,7 +565,7 @@ EIGEN_STRONG_INLINE void pstoreu(int* to, const Packet16i& from) { template <> EIGEN_DEVICE_FUNC inline Packet16f pgather(const float* from, Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); + Packet16i stride_vector = _mm512_set1_epi32(stride); Packet16i stride_multiplier = _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); @@ -549,7 +575,7 @@ EIGEN_DEVICE_FUNC inline Packet16f pgather(const float* from, template <> EIGEN_DEVICE_FUNC inline Packet8d pgather(const double* from, Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); + Packet8i stride_vector = _mm256_set1_epi32(stride); Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); @@ -560,7 +586,7 @@ template <> EIGEN_DEVICE_FUNC inline void pscatter(float* to, const Packet16f& from, Index stride) { - Packet16i stride_vector = _mm512_set1_epi32(convert_index(stride)); + Packet16i stride_vector = _mm512_set1_epi32(stride); Packet16i stride_multiplier = _mm512_set_epi32(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0); Packet16i indices = _mm512_mullo_epi32(stride_vector, stride_multiplier); @@ -570,7 +596,7 @@ template <> EIGEN_DEVICE_FUNC inline void pscatter(double* to, const Packet8d& from, Index stride) { - Packet8i stride_vector = _mm256_set1_epi32(convert_index(stride)); + Packet8i stride_vector = _mm256_set1_epi32(stride); Packet8i stride_multiplier = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); Packet8i indices = _mm256_mullo_epi32(stride_vector, stride_multiplier); _mm512_i32scatter_pd(to, indices, from, 8); @@ -634,8 +660,8 @@ EIGEN_STRONG_INLINE Packet8d pabs(const Packet8d& a) { #ifdef EIGEN_VECTORIZE_AVX512DQ // AVX512F does not define _mm512_extractf32x8_ps to extract _m256 from _m512 #define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ - __m256 OUTPUT##_0 = _mm512_extractf32x8_ps(INPUT, 0); \ - __m256 OUTPUT##_1 = _mm512_extractf32x8_ps(INPUT, 1) + __m256 OUTPUT##_0 = _mm512_extractf32x8_ps(INPUT, 0) __m256 OUTPUT##_1 = \ + _mm512_extractf32x8_ps(INPUT, 1) #else #define EIGEN_EXTRACT_8f_FROM_16f(INPUT, OUTPUT) \ __m256 OUTPUT##_0 = _mm256_insertf128_ps( \ @@ -648,136 +674,17 @@ EIGEN_STRONG_INLINE Packet8d pabs(const Packet8d& a) { #ifdef EIGEN_VECTORIZE_AVX512DQ #define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ - OUTPUT = _mm512_insertf32x8(_mm512_castps256_ps512(INPUTA), INPUTB, 1); + OUTPUT = _mm512_insertf32x8(OUTPUT, INPUTA, 0); \ + OUTPUT = _mm512_insertf32x8(OUTPUT, INPUTB, 1); #else #define EIGEN_INSERT_8f_INTO_16f(OUTPUT, INPUTA, INPUTB) \ - OUTPUT = _mm512_undefined_ps(); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 0), 0); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTA, 1), 1); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 0), 2); \ OUTPUT = _mm512_insertf32x4(OUTPUT, _mm256_extractf128_ps(INPUTB, 1), 3); #endif - -template <> -EIGEN_STRONG_INLINE float predux(const Packet16f& a) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - __m256 lane0 = _mm512_extractf32x8_ps(a, 0); - __m256 lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f x = _mm256_add_ps(lane0, lane1); - return predux(x); -#else - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 sum = _mm_add_ps(_mm_add_ps(lane0, lane1), _mm_add_ps(lane2, lane3)); - sum = _mm_hadd_ps(sum, sum); - sum = _mm_hadd_ps(sum, _mm_permute_ps(sum, 1)); - return _mm_cvtss_f32(sum); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d sum = _mm256_add_pd(lane0, lane1); - __m256d tmp0 = _mm256_hadd_pd(sum, _mm256_permute2f128_pd(sum, sum, 1)); - return _mm_cvtsd_f64(_mm256_castpd256_pd128(_mm256_hadd_pd(tmp0, tmp0))); -} - -template <> -EIGEN_STRONG_INLINE Packet8f predux_downto4(const Packet16f& a) { -#ifdef EIGEN_VECTORIZE_AVX512DQ - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - return padd(lane0, lane1); -#else - Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); - Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); - Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); - Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); - Packet4f sum0 = padd(lane0, lane2); - Packet4f sum1 = padd(lane1, lane3); - return _mm256_insertf128_ps(_mm256_castps128_ps256(sum0), sum1, 1); -#endif -} -template <> -EIGEN_STRONG_INLINE Packet4d predux_downto4(const Packet8d& a) { - Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); - Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); - Packet4d res = padd(lane0, lane1); - return res; -} - -template <> -EIGEN_STRONG_INLINE float predux_mul(const Packet16f& a) { -//#ifdef EIGEN_VECTORIZE_AVX512DQ -#if 0 - Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); - Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); - Packet8f res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_ps(res, res, 1)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#else - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = pmul(pmul(lane0, lane1), pmul(lane2, lane3)); - res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -#endif -} -template <> -EIGEN_STRONG_INLINE double predux_mul(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = pmul(lane0, lane1); - res = pmul(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(pmul(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_min(const Packet16f& a) { - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = _mm_min_ps(_mm_min_ps(lane0, lane1), _mm_min_ps(lane2, lane3)); - res = _mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} -template <> -EIGEN_STRONG_INLINE double predux_min(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = _mm256_min_pd(lane0, lane1); - res = _mm256_min_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_min_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template <> -EIGEN_STRONG_INLINE float predux_max(const Packet16f& a) { - __m128 lane0 = _mm512_extractf32x4_ps(a, 0); - __m128 lane1 = _mm512_extractf32x4_ps(a, 1); - __m128 lane2 = _mm512_extractf32x4_ps(a, 2); - __m128 lane3 = _mm512_extractf32x4_ps(a, 3); - __m128 res = _mm_max_ps(_mm_max_ps(lane0, lane1), _mm_max_ps(lane2, lane3)); - res = _mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); - return pfirst(_mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); -} - -template <> -EIGEN_STRONG_INLINE double predux_max(const Packet8d& a) { - __m256d lane0 = _mm512_extractf64x4_pd(a, 0); - __m256d lane1 = _mm512_extractf64x4_pd(a, 1); - __m256d res = _mm256_max_pd(lane0, lane1); - res = _mm256_max_pd(res, _mm256_permute2f128_pd(res, res, 1)); - return pfirst(_mm256_max_pd(res, _mm256_shuffle_pd(res, res, 1))); -} - -template<> EIGEN_STRONG_INLINE Packet16f preduxp(const Packet16f* vecs) +template<> EIGEN_STRONG_INLINE Packet16f preduxp(const Packet16f* +vecs) { EIGEN_EXTRACT_8f_FROM_16f(vecs[0], vecs0); EIGEN_EXTRACT_8f_FROM_16f(vecs[1], vecs1); @@ -966,7 +873,174 @@ template<> EIGEN_STRONG_INLINE Packet8d preduxp(const Packet8d* vecs) return _mm512_insertf64x4(final_output, final_1, 1); } - + +template <> +EIGEN_STRONG_INLINE float predux(const Packet16f& a) { + //#ifdef EIGEN_VECTORIZE_AVX512DQ +#if 0 + Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); + Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); + Packet8f sum = padd(lane0, lane1); + Packet8f tmp0 = _mm256_hadd_ps(sum, _mm256_permute2f128_ps(a, a, 1)); + tmp0 = _mm256_hadd_ps(tmp0, tmp0); + return pfirst(_mm256_hadd_ps(tmp0, tmp0)); +#else + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f sum = padd(padd(lane0, lane1), padd(lane2, lane3)); + sum = _mm_hadd_ps(sum, sum); + sum = _mm_hadd_ps(sum, _mm_permute_ps(sum, 1)); + return pfirst(sum); +#endif +} +template <> +EIGEN_STRONG_INLINE double predux(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d sum = padd(lane0, lane1); + Packet4d tmp0 = _mm256_hadd_pd(sum, _mm256_permute2f128_pd(sum, sum, 1)); + return pfirst(_mm256_hadd_pd(tmp0, tmp0)); +} + +template <> +EIGEN_STRONG_INLINE Packet8f predux_downto4(const Packet16f& a) { +#ifdef EIGEN_VECTORIZE_AVX512DQ + Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); + Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); + return padd(lane0, lane1); +#else + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f sum0 = padd(lane0, lane2); + Packet4f sum1 = padd(lane1, lane3); + return _mm256_insertf128_ps(_mm256_castps128_ps256(sum0), sum1, 1); +#endif +} +template <> +EIGEN_STRONG_INLINE Packet4d predux_downto4(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d res = padd(lane0, lane1); + return res; +} + +template <> +EIGEN_STRONG_INLINE float predux_mul(const Packet16f& a) { +//#ifdef EIGEN_VECTORIZE_AVX512DQ +#if 0 + Packet8f lane0 = _mm512_extractf32x8_ps(a, 0); + Packet8f lane1 = _mm512_extractf32x8_ps(a, 1); + Packet8f res = pmul(lane0, lane1); + res = pmul(res, _mm256_permute2f128_ps(res, res, 1)); + res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +#else + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f res = pmul(pmul(lane0, lane1), pmul(lane2, lane3)); + res = pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(pmul(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +#endif +} +template <> +EIGEN_STRONG_INLINE double predux_mul(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d res = pmul(lane0, lane1); + res = pmul(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(pmul(res, _mm256_shuffle_pd(res, res, 1))); +} + +template <> +EIGEN_STRONG_INLINE float predux_min(const Packet16f& a) { + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f res = _mm_min_ps(_mm_min_ps(lane0, lane1), _mm_min_ps(lane2, lane3)); + res = _mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(_mm_min_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +} +template <> +EIGEN_STRONG_INLINE double predux_min(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d res = _mm256_min_pd(lane0, lane1); + res = _mm256_min_pd(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(_mm256_min_pd(res, _mm256_shuffle_pd(res, res, 1))); +} + +template <> +EIGEN_STRONG_INLINE float predux_max(const Packet16f& a) { + Packet4f lane0 = _mm512_extractf32x4_ps(a, 0); + Packet4f lane1 = _mm512_extractf32x4_ps(a, 1); + Packet4f lane2 = _mm512_extractf32x4_ps(a, 2); + Packet4f lane3 = _mm512_extractf32x4_ps(a, 3); + Packet4f res = _mm_max_ps(_mm_max_ps(lane0, lane1), _mm_max_ps(lane2, lane3)); + res = _mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 3, 2))); + return pfirst(_mm_max_ps(res, _mm_permute_ps(res, _MM_SHUFFLE(0, 0, 0, 1)))); +} +template <> +EIGEN_STRONG_INLINE double predux_max(const Packet8d& a) { + Packet4d lane0 = _mm512_extractf64x4_pd(a, 0); + Packet4d lane1 = _mm512_extractf64x4_pd(a, 1); + Packet4d res = _mm256_max_pd(lane0, lane1); + res = _mm256_max_pd(res, _mm256_permute2f128_pd(res, res, 1)); + return pfirst(_mm256_max_pd(res, _mm256_shuffle_pd(res, res, 1))); +} + +template +struct palign_impl { + static EIGEN_STRONG_INLINE void run(Packet16f& first, + const Packet16f& second) { + if (Offset != 0) { + __m512i first_idx = _mm512_set_epi32( + Offset + 15, Offset + 14, Offset + 13, Offset + 12, Offset + 11, + Offset + 10, Offset + 9, Offset + 8, Offset + 7, Offset + 6, + Offset + 5, Offset + 4, Offset + 3, Offset + 2, Offset + 1, Offset); + + __m512i second_idx = + _mm512_set_epi32(Offset - 1, Offset - 2, Offset - 3, Offset - 4, + Offset - 5, Offset - 6, Offset - 7, Offset - 8, + Offset - 9, Offset - 10, Offset - 11, Offset - 12, + Offset - 13, Offset - 14, Offset - 15, Offset - 16); + + unsigned short mask = 0xFFFF; + mask <<= (16 - Offset); + + first = _mm512_permutexvar_ps(first_idx, first); + Packet16f tmp = _mm512_permutexvar_ps(second_idx, second); + first = _mm512_mask_blend_ps(mask, first, tmp); + } + } +}; +template +struct palign_impl { + static EIGEN_STRONG_INLINE void run(Packet8d& first, const Packet8d& second) { + if (Offset != 0) { + __m512i first_idx = _mm512_set_epi32( + 0, Offset + 7, 0, Offset + 6, 0, Offset + 5, 0, Offset + 4, 0, + Offset + 3, 0, Offset + 2, 0, Offset + 1, 0, Offset); + + __m512i second_idx = _mm512_set_epi32( + 0, Offset - 1, 0, Offset - 2, 0, Offset - 3, 0, Offset - 4, 0, + Offset - 5, 0, Offset - 6, 0, Offset - 7, 0, Offset - 8); + + unsigned char mask = 0xFF; + mask <<= (8 - Offset); + + first = _mm512_permutexvar_pd(first_idx, first); + Packet8d tmp = _mm512_permutexvar_pd(second_idx, second); + first = _mm512_mask_blend_pd(mask, first, tmp); + } + } +}; #define PACK_OUTPUT(OUTPUT, INPUT, INDEX, STRIDE) \ @@ -1228,76 +1302,13 @@ EIGEN_STRONG_INLINE Packet16f pblend(const Selector<16>& /*ifPacket*/, return Packet16f(); } template <> -EIGEN_STRONG_INLINE Packet8d pblend(const Selector<8>& ifPacket, - const Packet8d& thenPacket, - const Packet8d& elsePacket) { - __mmask8 m = (ifPacket.select[0] ) - | (ifPacket.select[1]<<1) - | (ifPacket.select[2]<<2) - | (ifPacket.select[3]<<3) - | (ifPacket.select[4]<<4) - | (ifPacket.select[5]<<5) - | (ifPacket.select[6]<<6) - | (ifPacket.select[7]<<7); - return _mm512_mask_blend_pd(m, elsePacket, thenPacket); -} - -template<> EIGEN_STRONG_INLINE Packet16i pcast(const Packet16f& a) { - return _mm512_cvttps_epi32(a); -} - -template<> EIGEN_STRONG_INLINE Packet16f pcast(const Packet16i& a) { - return _mm512_cvtepi32_ps(a); +EIGEN_STRONG_INLINE Packet8d pblend(const Selector<8>& /*ifPacket*/, + const Packet8d& /*thenPacket*/, + const Packet8d& /*elsePacket*/) { + assert(false && "To be implemented"); + return Packet8d(); } -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet16f& first, - const Packet16f& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - Offset + 15, Offset + 14, Offset + 13, Offset + 12, Offset + 11, - Offset + 10, Offset + 9, Offset + 8, Offset + 7, Offset + 6, - Offset + 5, Offset + 4, Offset + 3, Offset + 2, Offset + 1, Offset); - - __m512i second_idx = - _mm512_set_epi32(Offset - 1, Offset - 2, Offset - 3, Offset - 4, - Offset - 5, Offset - 6, Offset - 7, Offset - 8, - Offset - 9, Offset - 10, Offset - 11, Offset - 12, - Offset - 13, Offset - 14, Offset - 15, Offset - 16); - - unsigned short mask = 0xFFFF; - mask <<= (16 - Offset); - - first = _mm512_permutexvar_ps(first_idx, first); - Packet16f tmp = _mm512_permutexvar_ps(second_idx, second); - first = _mm512_mask_blend_ps(mask, first, tmp); - } - } -}; -template -struct palign_impl { - static EIGEN_STRONG_INLINE void run(Packet8d& first, const Packet8d& second) { - if (Offset != 0) { - __m512i first_idx = _mm512_set_epi32( - 0, Offset + 7, 0, Offset + 6, 0, Offset + 5, 0, Offset + 4, 0, - Offset + 3, 0, Offset + 2, 0, Offset + 1, 0, Offset); - - __m512i second_idx = _mm512_set_epi32( - 0, Offset - 1, 0, Offset - 2, 0, Offset - 3, 0, Offset - 4, 0, - Offset - 5, 0, Offset - 6, 0, Offset - 7, 0, Offset - 8); - - unsigned char mask = 0xFF; - mask <<= (8 - Offset); - - first = _mm512_permutexvar_pd(first_idx, first); - Packet8d tmp = _mm512_permutexvar_pd(second_idx, second); - first = _mm512_mask_blend_pd(mask, first, tmp); - } - } -}; - - } // end namespace internal } // end namespace Eigen diff --git a/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h b/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h index 59717b4fe6c..755e6209d1e 100644 --- a/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h +++ b/third_party/eigen/Eigen/src/Core/arch/CUDA/Half.h @@ -42,7 +42,6 @@ #define EIGEN_EXPLICIT_CAST(tgt_type) operator tgt_type() #endif -#include namespace Eigen { diff --git a/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h b/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h index f749c573ff6..c66d38469f8 100644 --- a/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +++ b/third_party/eigen/Eigen/src/Core/arch/CUDA/PacketMathHalf.h @@ -230,7 +230,7 @@ template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux(const half2& #else float a1 = __low2float(a); float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 + a2)); + return Eigen::half(half_impl::raw_uint16_to_half(__float2half_rn(a1 + a2))); #endif } @@ -264,7 +264,7 @@ template<> __device__ EIGEN_STRONG_INLINE Eigen::half predux_mul(const ha #else float a1 = __low2float(a); float a2 = __high2float(a); - return Eigen::half(__float2half_rn(a1 * a2)); + return Eigen::half(half_impl::raw_uint16_to_half(__float2half_rn(a1 * a2))); #endif } diff --git a/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h b/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h index b56e7afd2c5..2e6a00ffd1b 100644 --- a/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h +++ b/third_party/eigen/Eigen/src/Core/functors/UnaryFunctors.h @@ -768,7 +768,7 @@ struct scalar_sign_op { if (aa==real_type(0)) return Scalar(0); aa = real_type(1)/aa; - return Scalar(a.real()*aa, a.imag()*aa ); + return Scalar(real(a)*aa, imag(a)*aa ); } //TODO //template diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 681451cc309..e3980f6ffd4 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -115,8 +115,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // registers. However once the latency is hidden there is no point in // increasing the value of k, so we'll cap it at 320 (value determined // experimentally). - // To avoid that k vanishes, we make k_cache at least as big as kr - const Index k_cache = numext::maxi(kr, (numext::mini)((l1-ksub)/kdiv, 320)); + const Index k_cache = (numext::mini)((l1-ksub)/kdiv, 320); if (k_cache < k) { k = k_cache - (k_cache % kr); eigen_internal_assert(k > 0); @@ -649,8 +648,8 @@ class gebp_traits, std::complex, _ConjLhs, // Vectorized path EIGEN_STRONG_INLINE void loadRhs(const RhsScalar* b, DoublePacketType& dest) const { - dest.first = pset1(numext::real(*b)); - dest.second = pset1(numext::imag(*b)); + dest.first = pset1(real(*b)); + dest.second = pset1(imag(*b)); } EIGEN_STRONG_INLINE void loadRhsQuad(const RhsScalar* b, ResPacket& dest) const diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h index ed6234c378e..6440e1d09c7 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix.h @@ -20,9 +20,8 @@ template class level3_blocking; template< typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, - typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResInnerStride> -struct general_matrix_matrix_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs> +struct general_matrix_matrix_product { typedef gebp_traits Traits; @@ -31,7 +30,7 @@ struct general_matrix_matrix_product& blocking, GemmParallelInfo* info = 0) @@ -40,8 +39,8 @@ struct general_matrix_matrix_product - ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking,info); + ColMajor> + ::run(cols,rows,depth,rhs,rhsStride,lhs,lhsStride,res,resStride,alpha,blocking,info); } }; @@ -50,9 +49,8 @@ struct general_matrix_matrix_product -struct general_matrix_matrix_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs> +struct general_matrix_matrix_product { typedef gebp_traits Traits; @@ -61,17 +59,17 @@ typedef typename ScalarBinaryOpTraits::ReturnType ResScala static void run(Index rows, Index cols, Index depth, const LhsScalar* _lhs, Index lhsStride, const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, + ResScalar* _res, Index resStride, ResScalar alpha, level3_blocking& blocking, GemmParallelInfo* info = 0) { typedef const_blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; - LhsMapper lhs(_lhs, lhsStride); - RhsMapper rhs(_rhs, rhsStride); - ResMapper res(_res, resStride, resIncr); + typedef blas_data_mapper ResMapper; + LhsMapper lhs(_lhs,lhsStride); + RhsMapper rhs(_rhs,rhsStride); + ResMapper res(_res, resStride); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -228,7 +226,7 @@ struct gemm_functor Gemm::run(rows, cols, m_lhs.cols(), &m_lhs.coeffRef(row,0), m_lhs.outerStride(), &m_rhs.coeffRef(0,col), m_rhs.outerStride(), - (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.innerStride(), m_dest.outerStride(), + (Scalar*)&(m_dest.coeffRef(row,col)), m_dest.outerStride(), m_actualAlpha, m_blocking, info); } @@ -430,7 +428,7 @@ struct generic_product_impl static void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::assign_op()); + lazyproduct::evalTo(dst, lhs, rhs); else { dst.setZero(); @@ -442,7 +440,7 @@ struct generic_product_impl static void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::add_assign_op()); + lazyproduct::addTo(dst, lhs, rhs); else scaleAndAddTo(dst,lhs, rhs, Scalar(1)); } @@ -451,7 +449,7 @@ struct generic_product_impl static void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs) { if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0) - lazyproduct::eval_dynamic(dst, lhs, rhs, internal::sub_assign_op()); + lazyproduct::subTo(dst, lhs, rhs); else scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); } @@ -478,8 +476,7 @@ struct generic_product_impl Index, LhsScalar, (ActualLhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(LhsBlasTraits::NeedToConjugate), RhsScalar, (ActualRhsTypeCleaned::Flags&RowMajorBit) ? RowMajor : ColMajor, bool(RhsBlasTraits::NeedToConjugate), - (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime>, + (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor>, ActualLhsTypeCleaned, ActualRhsTypeCleaned, Dest, BlockingType> GemmFunctor; BlockingType blocking(dst.rows(), dst.cols(), lhs.cols(), 1, true); diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h index d68d2f96572..e844e37d16b 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h @@ -25,54 +25,51 @@ namespace internal { **********************************************************************/ // forward declarations (defined at the end of this file) -template +template struct tribb_kernel; /* Optimized matrix-matrix product evaluating only one triangular half */ template + int ResStorageOrder, int UpLo, int Version = Specialized> struct general_matrix_matrix_triangular_product; // as usual if the result is row major => we transpose the product template -struct general_matrix_matrix_triangular_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, int UpLo, int Version> +struct general_matrix_matrix_triangular_product { typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* lhs, Index lhsStride, - const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resIncr, Index resStride, + const RhsScalar* rhs, Index rhsStride, ResScalar* res, Index resStride, const ResScalar& alpha, level3_blocking& blocking) { general_matrix_matrix_triangular_product - ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resIncr,resStride,alpha,blocking); + ColMajor, UpLo==Lower?Upper:Lower> + ::run(size,depth,rhs,rhsStride,lhs,lhsStride,res,resStride,alpha,blocking); } }; template -struct general_matrix_matrix_triangular_product + typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, int UpLo, int Version> +struct general_matrix_matrix_triangular_product { typedef typename ScalarBinaryOpTraits::ReturnType ResScalar; static EIGEN_STRONG_INLINE void run(Index size, Index depth,const LhsScalar* _lhs, Index lhsStride, - const RhsScalar* _rhs, Index rhsStride, - ResScalar* _res, Index resIncr, Index resStride, + const RhsScalar* _rhs, Index rhsStride, ResScalar* _res, Index resStride, const ResScalar& alpha, level3_blocking& blocking) { typedef gebp_traits Traits; typedef const_blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); + ResMapper res(_res, resStride); Index kc = blocking.kc(); Index mc = (std::min)(size,blocking.mc()); @@ -90,7 +87,7 @@ struct general_matrix_matrix_triangular_product pack_lhs; gemm_pack_rhs pack_rhs; gebp_kernel gebp; - tribb_kernel sybb; + tribb_kernel sybb; for(Index k2=0; k2 +template struct tribb_kernel { typedef gebp_traits Traits; @@ -144,13 +142,11 @@ struct tribb_kernel enum { BlockSize = meta_least_common_multiple::ret }; - void operator()(ResScalar* _res, Index resIncr, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) + void operator()(ResScalar* _res, Index resStride, const LhsScalar* blockA, const RhsScalar* blockB, Index size, Index depth, const ResScalar& alpha) { - typedef blas_data_mapper ResMapper; - typedef blas_data_mapper BufferMapper; - ResMapper res(_res, resStride, resIncr); - gebp_kernel gebp_kernel1; - gebp_kernel gebp_kernel2; + typedef blas_data_mapper ResMapper; + ResMapper res(_res, resStride); + gebp_kernel gebp_kernel; Matrix buffer((internal::constructor_without_unaligned_array_assert())); @@ -162,32 +158,31 @@ struct tribb_kernel const RhsScalar* actual_b = blockB+j*depth; if(UpLo==Upper) - gebp_kernel1(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - + gebp_kernel(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha, + -1, -1, 0, 0); + // selfadjoint micro block { Index i = j; buffer.setZero(); // 1 - apply the kernel on the temporary buffer - gebp_kernel2(BufferMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, - -1, -1, 0, 0); - + gebp_kernel(ResMapper(buffer.data(), BlockSize), blockA+depth*i, actual_b, actualBlockSize, depth, actualBlockSize, alpha, + -1, -1, 0, 0); // 2 - triangular accumulation for(Index j1=0; j1 internal::general_matrix_matrix_triangular_product + IsRowMajor ? RowMajor : ColMajor, UpLo&(Lower|Upper)> ::run(size, depth, &actualLhs.coeffRef(SkipDiag&&(UpLo&Lower)==Lower ? 1 : 0,0), actualLhs.outerStride(), &actualRhs.coeffRef(0,SkipDiag&&(UpLo&Upper)==Upper ? 1 : 0), actualRhs.outerStride(), - mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? mat.innerStride() : mat.outerStride() ) : 0), - mat.innerStride(), mat.outerStride(), actualAlpha, blocking); + mat.data() + (SkipDiag ? (bool(IsRowMajor) != ((UpLo&Lower)==Lower) ? 1 : mat.outerStride() ) : 0), mat.outerStride(), actualAlpha, blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h index 691f95d6979..f6f9ebecae7 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h @@ -40,7 +40,7 @@ namespace internal { template struct general_matrix_matrix_rankupdate : general_matrix_matrix_triangular_product< - Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,1,UpLo,BuiltIn> {}; + Index,Scalar,AStorageOrder,ConjugateA,Scalar,AStorageOrder,ConjugateA,ResStorageOrder,UpLo,BuiltIn> {}; // try to go to BLAS specialization @@ -48,9 +48,9 @@ struct general_matrix_matrix_rankupdate : template \ struct general_matrix_matrix_triangular_product { \ + Scalar,RhsStorageOrder,ConjugateRhs,ColMajor,UpLo,Specialized> { \ static EIGEN_STRONG_INLINE void run(Index size, Index depth,const Scalar* lhs, Index lhsStride, \ - const Scalar* rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) \ + const Scalar* rhs, Index rhsStride, Scalar* res, Index resStride, Scalar alpha, level3_blocking& blocking) \ { \ if ( lhs==rhs && ((UpLo&(Lower|Upper))==UpLo) ) { \ general_matrix_matrix_rankupdate \ @@ -59,8 +59,8 @@ struct general_matrix_matrix_triangular_product \ - ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resIncr,resStride,alpha,blocking); \ + ColMajor, UpLo, BuiltIn> \ + ::run(size,depth,lhs,lhsStride,rhs,rhsStride,res,resStride,alpha,blocking); \ } \ } \ }; diff --git a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h index 71abf4013d1..b0f6b0d5b9e 100644 --- a/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/GeneralMatrixMatrix_BLAS.h @@ -51,22 +51,20 @@ template< \ typename Index, \ int LhsStorageOrder, bool ConjugateLhs, \ int RhsStorageOrder, bool ConjugateRhs> \ -struct general_matrix_matrix_product \ +struct general_matrix_matrix_product \ { \ typedef gebp_traits Traits; \ \ static void run(Index rows, Index cols, Index depth, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ + EIGTYPE* res, Index resStride, \ EIGTYPE alpha, \ level3_blocking& /*blocking*/, \ GemmParallelInfo* /*info = 0*/) \ { \ using std::conj; \ \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ char transa, transb; \ BlasIndex m, n, k, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ diff --git a/third_party/eigen/Eigen/src/Core/products/Parallelizer.h b/third_party/eigen/Eigen/src/Core/products/Parallelizer.h index a3cc05b77bb..c2f084c82cc 100644 --- a/third_party/eigen/Eigen/src/Core/products/Parallelizer.h +++ b/third_party/eigen/Eigen/src/Core/products/Parallelizer.h @@ -17,8 +17,7 @@ namespace internal { /** \internal */ inline void manage_multi_threading(Action action, int* v) { - static int m_maxThreads = -1; - EIGEN_UNUSED_VARIABLE(m_maxThreads); + static EIGEN_UNUSED int m_maxThreads = -1; if(action==SetAction) { @@ -151,10 +150,8 @@ void parallelize_gemm(const Functor& func, Index rows, Index cols, Index depth, info[i].lhs_start = r0; info[i].lhs_length = actualBlockRows; - if(transpose) - func(c0, actualBlockCols, 0, rows, info); - else - func(0, rows, c0, actualBlockCols, info); + if(transpose) func(c0, actualBlockCols, 0, rows, info); + else func(0, rows, c0, actualBlockCols, info); } #endif } diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h index 04c933480bd..da6f82abcd7 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix.h @@ -277,21 +277,20 @@ struct symm_pack_rhs template + int ResStorageOrder> struct product_selfadjoint_matrix; template -struct product_selfadjoint_matrix + int RhsStorageOrder, bool RhsSelfAdjoint, bool ConjugateRhs> +struct product_selfadjoint_matrix { static EIGEN_STRONG_INLINE void run( Index rows, Index cols, const Scalar* lhs, Index lhsStride, const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, + Scalar* res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { product_selfadjoint_matrix::IsComplex && EIGEN_LOGICAL_XOR(RhsSelfAdjoint,ConjugateRhs), EIGEN_LOGICAL_XOR(LhsSelfAdjoint,LhsStorageOrder==RowMajor) ? ColMajor : RowMajor, LhsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsSelfAdjoint,ConjugateLhs), - ColMajor,ResInnerStride> - ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); + ColMajor> + ::run(cols, rows, rhs, rhsStride, lhs, lhsStride, res, resStride, alpha, blocking); } }; template -struct product_selfadjoint_matrix + int RhsStorageOrder, bool ConjugateRhs> +struct product_selfadjoint_matrix { static EIGEN_DONT_INLINE void run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, + Scalar* res, Index resStride, const Scalar& alpha, level3_blocking& blocking); }; template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( + int RhsStorageOrder, bool ConjugateRhs> +EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, + Scalar* _res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { Index size = rows; @@ -337,11 +334,11 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix LhsMapper; typedef const_blas_data_mapper LhsTransposeMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); LhsTransposeMapper lhs_transpose(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); + ResMapper res(_res, resStride); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -401,28 +398,26 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix -struct product_selfadjoint_matrix + int RhsStorageOrder, bool ConjugateRhs> +struct product_selfadjoint_matrix { static EIGEN_DONT_INLINE void run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, + Scalar* res, Index resStride, const Scalar& alpha, level3_blocking& blocking); }; template -EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( + int RhsStorageOrder, bool ConjugateRhs> +EIGEN_DONT_INLINE void product_selfadjoint_matrix::run( Index rows, Index cols, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, + Scalar* _res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { Index size = cols; @@ -430,9 +425,9 @@ EIGEN_DONT_INLINE void product_selfadjoint_matrix Traits; typedef const_blas_data_mapper LhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); - ResMapper res(_res,resStride, resIncr); + ResMapper res(_res,resStride); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -508,13 +503,12 @@ struct selfadjoint_product_impl NumTraits::IsComplex && EIGEN_LOGICAL_XOR(LhsIsUpper,bool(LhsBlasTraits::NeedToConjugate)), EIGEN_LOGICAL_XOR(RhsIsUpper,internal::traits::Flags &RowMajorBit) ? RowMajor : ColMajor, RhsIsSelfAdjoint, NumTraits::IsComplex && EIGEN_LOGICAL_XOR(RhsIsUpper,bool(RhsBlasTraits::NeedToConjugate)), - internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor, - Dest::InnerStrideAtCompileTime> + internal::traits::Flags&RowMajorBit ? RowMajor : ColMajor> ::run( lhs.rows(), rhs.cols(), // sizes &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info + &dst.coeffRef(0,0), dst.outerStride(), // result info actualAlpha, blocking // alpha ); } diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h index 61396dbdf6f..9a5318507af 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointMatrixMatrix_BLAS.h @@ -44,18 +44,16 @@ namespace internal { template \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ \ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ + EIGTYPE* res, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ char side='L', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -93,17 +91,15 @@ struct product_selfadjoint_matrix \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ + EIGTYPE* res, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ char side='L', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -171,18 +167,16 @@ EIGEN_BLAS_HEMM_L(scomplex, float, cf, chemm_) template \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ \ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ + EIGTYPE* res, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ char side='R', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ @@ -219,17 +213,15 @@ struct product_selfadjoint_matrix \ -struct product_selfadjoint_matrix \ +struct product_selfadjoint_matrix \ {\ static void run( \ Index rows, Index cols, \ const EIGTYPE* _lhs, Index lhsStride, \ const EIGTYPE* _rhs, Index rhsStride, \ - EIGTYPE* res, Index resIncr, Index resStride, \ + EIGTYPE* res, Index resStride, \ EIGTYPE alpha, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ char side='R', uplo='L'; \ BlasIndex m, n, lda, ldb, ldc; \ const EIGTYPE *a, *b; \ diff --git a/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h b/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h index ef12c98f6cb..f038d686f5d 100644 --- a/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h +++ b/third_party/eigen/Eigen/src/Core/products/SelfadjointProduct.h @@ -109,10 +109,10 @@ struct selfadjoint_product_selector internal::general_matrix_matrix_triangular_product::IsComplex, Scalar, OtherIsRowMajor ? ColMajor : RowMajor, (!OtherBlasTraits::NeedToConjugate) && NumTraits::IsComplex, - IsRowMajor ? RowMajor : ColMajor, MatrixType::InnerStrideAtCompileTime, UpLo> + IsRowMajor ? RowMajor : ColMajor, UpLo> ::run(size, depth, &actualOther.coeffRef(0,0), actualOther.outerStride(), &actualOther.coeffRef(0,0), actualOther.outerStride(), - mat.data(), mat.innerStride(), mat.outerStride(), actualAlpha, blocking); + mat.data(), mat.outerStride(), actualAlpha, blocking); } }; diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h index 2fb408d1d73..f784507e77f 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix.h @@ -45,24 +45,22 @@ template + int ResStorageOrder, int Version = Specialized> struct product_triangular_matrix_matrix; template + int RhsStorageOrder, bool ConjugateRhs, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,RowMajor,Version> { static EIGEN_STRONG_INLINE void run( Index rows, Index cols, Index depth, const Scalar* lhs, Index lhsStride, const Scalar* rhs, Index rhsStride, - Scalar* res, Index resIncr, Index resStride, + Scalar* res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { product_triangular_matrix_matrix - ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resIncr, resStride, alpha, blocking); + ColMajor> + ::run(cols, rows, depth, rhs, rhsStride, lhs, lhsStride, res, resStride, alpha, blocking); } }; // implements col-major += alpha * op(triangular) * op(general) template + int RhsStorageOrder, bool ConjugateRhs, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,ColMajor,Version> { typedef gebp_traits Traits; @@ -98,21 +95,20 @@ struct product_triangular_matrix_matrix& blocking); }; template + int RhsStorageOrder, bool ConjugateRhs, int Version> EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( + RhsStorageOrder,ConjugateRhs,ColMajor,Version>::run( Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, + Scalar* _res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { // strip zeros @@ -123,10 +119,10 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); + ResMapper res(_res, resStride); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -239,11 +235,10 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix + int RhsStorageOrder, bool ConjugateRhs, int Version> struct product_triangular_matrix_matrix + RhsStorageOrder,ConjugateRhs,ColMajor,Version> { typedef gebp_traits Traits; enum { @@ -256,21 +251,20 @@ struct product_triangular_matrix_matrix& blocking); }; template + int RhsStorageOrder, bool ConjugateRhs, int Version> EIGEN_DONT_INLINE void product_triangular_matrix_matrix::run( + RhsStorageOrder,ConjugateRhs,ColMajor,Version>::run( Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride, const Scalar* _rhs, Index rhsStride, - Scalar* _res, Index resIncr, Index resStride, + Scalar* _res, Index resStride, const Scalar& alpha, level3_blocking& blocking) { const Index PacketBytes = packet_traits::size*sizeof(Scalar); @@ -282,10 +276,10 @@ EIGEN_DONT_INLINE void product_triangular_matrix_matrix LhsMapper; typedef const_blas_data_mapper RhsMapper; - typedef blas_data_mapper ResMapper; + typedef blas_data_mapper ResMapper; LhsMapper lhs(_lhs,lhsStride); RhsMapper rhs(_rhs,rhsStride); - ResMapper res(_res, resStride, resIncr); + ResMapper res(_res, resStride); Index kc = blocking.kc(); // cache block size along the K direction Index mc = (std::min)(rows,blocking.mc()); // cache block size along the M direction @@ -439,12 +433,12 @@ struct triangular_product_impl Mode, LhsIsTriangular, (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, LhsBlasTraits::NeedToConjugate, (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, RhsBlasTraits::NeedToConjugate, - (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor, Dest::InnerStrideAtCompileTime> + (internal::traits::Flags&RowMajorBit) ? RowMajor : ColMajor> ::run( stripedRows, stripedCols, stripedDepth, // sizes &lhs.coeffRef(0,0), lhs.outerStride(), // lhs info &rhs.coeffRef(0,0), rhs.outerStride(), // rhs info - &dst.coeffRef(0,0), dst.innerStride(), dst.outerStride(), // result info + &dst.coeffRef(0,0), dst.outerStride(), // result info actualAlpha, blocking ); diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h index a98d12e4ae1..a25197ab017 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularMatrixMatrix_BLAS.h @@ -46,7 +46,7 @@ template {}; + RhsStorageOrder, ConjugateRhs, ResStorageOrder, BuiltIn> {}; // try to go to BLAS specialization @@ -55,15 +55,13 @@ template \ struct product_triangular_matrix_matrix { \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder,ConjugateRhs,ColMajor,Specialized> { \ static inline void run(Index _rows, Index _cols, Index _depth, const Scalar* _lhs, Index lhsStride,\ - const Scalar* _rhs, Index rhsStride, Scalar* res, Index resIncr, Index resStride, Scalar alpha, level3_blocking& blocking) { \ - EIGEN_ONLY_USED_FOR_DEBUG(resIncr); \ - eigen_assert(resIncr == 1); \ + const Scalar* _rhs, Index rhsStride, Scalar* res, Index resStride, Scalar alpha, level3_blocking& blocking) { \ product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ } \ }; @@ -117,8 +115,8 @@ struct product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder, ConjugateRhs, ColMajor, BuiltIn>::run( \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ /*std::cout << "TRMM_L: A is not square! Go to Eigen TRMM implementation!\n";*/ \ } else { \ /* Make sense to call GEMM */ \ @@ -126,8 +124,8 @@ struct product_triangular_matrix_matrix_trmm(); \ BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, 1, resStride, alpha, gemm_blocking, 0); \ + general_matrix_matrix_product::run( \ + rows, cols, depth, aa_tmp.data(), aStride, _rhs, rhsStride, res, resStride, alpha, gemm_blocking, 0); \ \ /*std::cout << "TRMM_L: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ } \ @@ -234,8 +232,8 @@ struct product_triangular_matrix_matrix_trmm::run( \ - _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, 1, resStride, alpha, blocking); \ + LhsStorageOrder,ConjugateLhs, RhsStorageOrder, ConjugateRhs, ColMajor, BuiltIn>::run( \ + _rows, _cols, _depth, _lhs, lhsStride, _rhs, rhsStride, res, resStride, alpha, blocking); \ /*std::cout << "TRMM_R: A is not square! Go to Eigen TRMM implementation!\n";*/ \ } else { \ /* Make sense to call GEMM */ \ @@ -243,8 +241,8 @@ struct product_triangular_matrix_matrix_trmm(); \ BlasIndex aStride = convert_index(aa_tmp.outerStride()); \ gemm_blocking_space gemm_blocking(_rows,_cols,_depth, 1, true); \ - general_matrix_matrix_product::run( \ - rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, 1, resStride, alpha, gemm_blocking, 0); \ + general_matrix_matrix_product::run( \ + rows, cols, depth, _lhs, lhsStride, aa_tmp.data(), aStride, res, resStride, alpha, gemm_blocking, 0); \ \ /*std::cout << "TRMM_R: A is not square! Go to BLAS GEMM implementation! " << nthr<<" \n";*/ \ } \ diff --git a/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h b/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h index e3ed2cd19ea..223c38b8656 100644 --- a/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h +++ b/third_party/eigen/Eigen/src/Core/products/TriangularSolverMatrix.h @@ -15,48 +15,48 @@ namespace Eigen { namespace internal { // if the rhs is row major, let's transpose the product -template -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static void run( Index size, Index cols, const Scalar* tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, + Scalar* _other, Index otherStride, level3_blocking& blocking) { triangular_solve_matrix< Scalar, Index, Side==OnTheLeft?OnTheRight:OnTheLeft, (Mode&UnitDiag) | ((Mode&Upper) ? Lower : Upper), NumTraits::IsComplex && Conjugate, - TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor, OtherInnerStride> - ::run(size, cols, tri, triStride, _other, otherIncr, otherStride, blocking); + TriStorageOrder==RowMajor ? ColMajor : RowMajor, ColMajor> + ::run(size, cols, tri, triStride, _other, otherStride, blocking); } }; /* Optimized triangular solver with multiple right hand side and the triangular matrix on the left */ -template -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static EIGEN_DONT_INLINE void run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, + Scalar* _other, Index otherStride, level3_blocking& blocking); }; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( +template +EIGEN_DONT_INLINE void triangular_solve_matrix::run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, + Scalar* _other, Index otherStride, level3_blocking& blocking) { Index cols = otherSize; typedef const_blas_data_mapper TriMapper; - typedef blas_data_mapper OtherMapper; + typedef blas_data_mapper OtherMapper; TriMapper tri(_tri, triStride); - OtherMapper other(_other, otherStride, otherIncr); + OtherMapper other(_other, otherStride); typedef gebp_traits Traits; @@ -128,19 +128,19 @@ EIGEN_DONT_INLINE void triangular_solve_matrix -struct triangular_solve_matrix +template +struct triangular_solve_matrix { static EIGEN_DONT_INLINE void run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, + Scalar* _other, Index otherStride, level3_blocking& blocking); }; -template -EIGEN_DONT_INLINE void triangular_solve_matrix::run( +template +EIGEN_DONT_INLINE void triangular_solve_matrix::run( Index size, Index otherSize, const Scalar* _tri, Index triStride, - Scalar* _other, Index otherIncr, Index otherStride, + Scalar* _other, Index otherStride, level3_blocking& blocking) { Index rows = otherSize; typedef typename NumTraits::Real RealScalar; - typedef blas_data_mapper LhsMapper; + typedef blas_data_mapper LhsMapper; typedef const_blas_data_mapper RhsMapper; - LhsMapper lhs(_other, otherStride, otherIncr); + LhsMapper lhs(_other, otherStride); RhsMapper rhs(_tri, triStride); typedef gebp_traits Traits; @@ -297,24 +297,24 @@ EIGEN_DONT_INLINE void triangular_solve_matrix \ -struct triangular_solve_matrix \ +struct triangular_solve_matrix \ { \ enum { \ IsLower = (Mode&Lower) == Lower, \ @@ -51,10 +51,8 @@ struct triangular_solve_matrix& /*blocking*/) \ + EIGTYPE* _other, Index otherStride, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ BlasIndex m = convert_index(size), n = convert_index(otherSize), lda, ldb; \ char side = 'L', uplo, diag='N', transa; \ /* Set alpha_ */ \ @@ -101,7 +99,7 @@ EIGEN_BLAS_TRSM_L(scomplex, float, ctrsm_) // implements RightSide general * op(triangular)^-1 #define EIGEN_BLAS_TRSM_R(EIGTYPE, BLASTYPE, BLASFUNC) \ template \ -struct triangular_solve_matrix \ +struct triangular_solve_matrix \ { \ enum { \ IsLower = (Mode&Lower) == Lower, \ @@ -112,10 +110,8 @@ struct triangular_solve_matrix& /*blocking*/) \ + EIGTYPE* _other, Index otherStride, level3_blocking& /*blocking*/) \ { \ - EIGEN_ONLY_USED_FOR_DEBUG(otherIncr); \ - eigen_assert(otherIncr == 1); \ BlasIndex m = convert_index(otherSize), n = convert_index(size), lda, ldb; \ char side = 'R', uplo, diag='N', transa; \ /* Set alpha_ */ \ diff --git a/third_party/eigen/Eigen/src/Core/util/BlasUtil.h b/third_party/eigen/Eigen/src/Core/util/BlasUtil.h index 3dff9bc9b33..6e6ee119b64 100755 --- a/third_party/eigen/Eigen/src/Core/util/BlasUtil.h +++ b/third_party/eigen/Eigen/src/Core/util/BlasUtil.h @@ -31,7 +31,7 @@ template< typename Index, typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs, typename RhsScalar, int RhsStorageOrder, bool ConjugateRhs, - int ResStorageOrder, int ResInnerStride> + int ResStorageOrder> struct general_matrix_matrix_product; template -class BlasLinearMapper; - template -class BlasLinearMapper { +class BlasLinearMapper { public: typedef typename packet_traits::type Packet; typedef typename packet_traits::half HalfPacket; - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data, Index incr=1) - : m_data(data) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data) : m_data(data) {} EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { internal::prefetch(&operator()(i)); @@ -196,25 +188,16 @@ class BlasLinearMapper { }; // Lightweight helper class to access matrix coefficients. -template -class blas_data_mapper; - -template -class blas_data_mapper -{ -public: +template +class blas_data_mapper { + public: typedef typename packet_traits::type Packet; typedef typename packet_traits::half HalfPacket; typedef BlasLinearMapper LinearMapper; typedef BlasVectorMapper VectorMapper; - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr=1) - : m_data(data), m_stride(stride) - { - EIGEN_ONLY_USED_FOR_DEBUG(incr); - eigen_assert(incr==1); - } + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride) : m_data(data), m_stride(stride) {} EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper getSubMapper(Index i, Index j) const { @@ -268,90 +251,6 @@ class blas_data_mapper const Index m_stride; }; -// Implementation of non-natural increment (i.e. inner-stride != 1) -// The exposed API is not complete yet compared to the Incr==1 case -// because some features makes less sense in this case. -template -class BlasLinearMapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE BlasLinearMapper(Scalar *data,Index incr) : m_data(data), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void prefetch(int i) const { - internal::prefetch(&operator()(i)); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Scalar& operator()(Index i) const { - return m_data[i*m_incr.value()]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i) const { - return pgather(m_data + i*m_incr.value(), m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void storePacket(Index i, const PacketType &p) const { - pscatter(m_data + i*m_incr.value(), p, m_incr.value()); - } - -protected: - Scalar *m_data; - const internal::variable_if_dynamic m_incr; -}; - -template -class blas_data_mapper -{ -public: - typedef typename packet_traits::type Packet; - typedef typename packet_traits::half HalfPacket; - - typedef BlasLinearMapper LinearMapper; - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper(Scalar* data, Index stride, Index incr) : m_data(data), m_stride(stride), m_incr(incr) {} - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE blas_data_mapper - getSubMapper(Index i, Index j) const { - return blas_data_mapper(&operator()(i, j), m_stride, m_incr.value()); - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE LinearMapper getLinearMapper(Index i, Index j) const { - return LinearMapper(&operator()(i, j), m_incr.value()); - } - - EIGEN_DEVICE_FUNC - EIGEN_ALWAYS_INLINE Scalar& operator()(Index i, Index j) const { - return m_data[StorageOrder==RowMajor ? j*m_incr.value() + i*m_stride : i*m_incr.value() + j*m_stride]; - } - - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet loadPacket(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE PacketT load(Index i, Index j) const { - return pgather(&operator()(i, j),m_incr.value()); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void scatterPacket(Index i, Index j, const SubPacket &p) const { - pscatter(&operator()(i, j), p, m_stride); - } - - template - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE SubPacket gatherPacket(Index i, Index j) const { - return pgather(&operator()(i, j), m_stride); - } - -protected: - Scalar* EIGEN_RESTRICT m_data; - const Index m_stride; - const internal::variable_if_dynamic m_incr; -}; - // lightweight helper class to access matrix coefficients (const version) template class const_blas_data_mapper : public blas_data_mapper { diff --git a/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h b/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h index 74f74cc42b8..351bd6c6000 100755 --- a/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h +++ b/third_party/eigen/Eigen/src/Core/util/DisableStupidWarnings.h @@ -57,10 +57,7 @@ #if __GNUC__>=6 #pragma GCC diagnostic ignored "-Wignored-attributes" #endif - #if __GNUC__==7 - // See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89325 - #pragma GCC diagnostic ignored "-Wattributes" - #endif + #endif #if defined __NVCC__ @@ -83,12 +80,4 @@ #pragma diag_suppress 2737 #endif -#else -// warnings already disabled: -# ifndef EIGEN_WARNINGS_DISABLED_2 -# define EIGEN_WARNINGS_DISABLED_2 -# elif defined(EIGEN_INTERNAL_DEBUGGING) -# error "Do not include \"DisableStupidWarnings.h\" recursively more than twice!" -# endif - #endif // not EIGEN_WARNINGS_DISABLED diff --git a/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h b/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h index 134544f9643..ea107393a7d 100644 --- a/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h +++ b/third_party/eigen/Eigen/src/Core/util/ForwardDeclarations.h @@ -47,7 +47,11 @@ template struct NumTraits; template struct EigenBase; template class DenseBase; template class PlainObjectBase; -template class DenseCoeffsBase; + + +template::value > +class DenseCoeffsBase; templatex || (EIGEN_WORLD_VERSION>=x && \ (EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \ @@ -380,8 +380,7 @@ #if EIGEN_MAX_CPP_VER>=11 && \ ((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)) \ || (defined(__GNUC__) && defined(_GLIBCXX_USE_C99)) \ - || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER)) \ - || (EIGEN_COMP_MSVC >= 1900) ) + || (defined(_LIBCPP_VERSION) && !defined(_MSC_VER))) #define EIGEN_HAS_C99_MATH 1 #else #define EIGEN_HAS_C99_MATH 0 @@ -397,20 +396,6 @@ #endif #endif -// Does the compiler support type_traits? -// - full support of type traits was added only to GCC 5.1.0. -// - 20150626 corresponds to the last release of 4.x libstdc++ -#ifndef EIGEN_HAS_TYPE_TRAITS -#if EIGEN_MAX_CPP_VER>=11 && (EIGEN_HAS_CXX11 || EIGEN_COMP_MSVC >= 1700) \ - && ((!EIGEN_COMP_GNUC_STRICT) || EIGEN_GNUC_AT_LEAST(5, 1)) \ - && ((!defined(__GLIBCXX__)) || __GLIBCXX__ > 20150626) -#define EIGEN_HAS_TYPE_TRAITS 1 -#define EIGEN_INCLUDE_TYPE_TRAITS -#else -#define EIGEN_HAS_TYPE_TRAITS 0 -#endif -#endif - // Does the compiler support variadic templates? #ifndef EIGEN_HAS_VARIADIC_TEMPLATES #if EIGEN_MAX_CPP_VER>=11 && (__cplusplus > 199711L || EIGEN_COMP_MSVC >= 1900) \ @@ -850,48 +835,11 @@ namespace Eigen { #endif -/** - * \internal - * \brief Macro to explicitly define the default copy constructor. - * This is necessary, because the implicit definition is deprecated if the copy-assignment is overridden. - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) EIGEN_DEVICE_FUNC CLASS(const CLASS&) = default; -#else -#define EIGEN_DEFAULT_COPY_CONSTRUCTOR(CLASS) -#endif - - - /** \internal * \brief Macro to manually inherit assignment operators. * This is necessary, because the implicitly defined assignment operator gets deleted when a custom operator= is defined. - * With C++11 or later this also default-implements the copy-constructor */ -#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) \ - EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ - EIGEN_DEFAULT_COPY_CONSTRUCTOR(Derived) - -/** \internal - * \brief Macro to manually define default constructors and destructors. - * This is necessary when the copy constructor is re-defined. - * For empty helper classes this should usually be protected, to avoid accidentally creating empty objects. - * - * Hiding the default destructor lead to problems in C++03 mode together with boost::multiprecision - */ -#if EIGEN_HAS_CXX11 -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() = default; \ - EIGEN_DEVICE_FUNC ~Derived() = default; -#else -#define EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(Derived) \ - EIGEN_DEVICE_FUNC Derived() {}; \ - /* EIGEN_DEVICE_FUNC ~Derived() {}; */ -#endif - - - - +#define EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Derived) EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) /** * Just a side note. Commenting within defines works only by documenting diff --git a/third_party/eigen/Eigen/src/Core/util/Meta.h b/third_party/eigen/Eigen/src/Core/util/Meta.h index 9b61ff037aa..d31e9541122 100755 --- a/third_party/eigen/Eigen/src/Core/util/Meta.h +++ b/third_party/eigen/Eigen/src/Core/util/Meta.h @@ -97,9 +97,6 @@ template<> struct is_arithmetic { enum { value = true }; }; template<> struct is_arithmetic { enum { value = true }; }; template<> struct is_arithmetic { enum { value = true }; }; -#if EIGEN_HAS_CXX11 -using std::is_integral; -#else template struct is_integral { enum { value = false }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; @@ -111,11 +108,6 @@ template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; template<> struct is_integral { enum { value = true }; }; -#if EIGEN_COMP_MSVC -template<> struct is_integral { enum { value = true }; }; -template<> struct is_integral{ enum { value = true }; }; -#endif -#endif #if EIGEN_HAS_CXX11 using std::make_unsigned; @@ -539,30 +531,4 @@ bool not_equal_strict(const double& x,const double& y) { return std::not_equal_t } // end namespace Eigen -// Define portable (u)int{32,64} types -#if EIGEN_HAS_CXX11 -#include -namespace Eigen { -namespace numext { -typedef std::uint32_t uint32_t; -typedef std::int32_t int32_t; -typedef std::uint64_t uint64_t; -typedef std::int64_t int64_t; -} -} -#else -// Without c++11, all compilers able to compile Eigen also -// provides the C99 stdint.h header file. -#include -namespace Eigen { -namespace numext { -typedef ::uint32_t uint32_t; -typedef ::int32_t int32_t; -typedef ::uint64_t uint64_t; -typedef ::int64_t int64_t; -} -} -#endif - - #endif // EIGEN_META_H diff --git a/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h b/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h index 1ce6fd1b009..ecc82b7c8df 100644 --- a/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h +++ b/third_party/eigen/Eigen/src/Core/util/ReenableStupidWarnings.h @@ -1,8 +1,4 @@ -#ifdef EIGEN_WARNINGS_DISABLED_2 -// "DisableStupidWarnings.h" was included twice recursively: Do not reenable warnings yet! -# undef EIGEN_WARNINGS_DISABLED_2 - -#elif defined(EIGEN_WARNINGS_DISABLED) +#ifdef EIGEN_WARNINGS_DISABLED #undef EIGEN_WARNINGS_DISABLED #ifndef EIGEN_PERMANENTLY_DISABLE_STUPID_WARNINGS diff --git a/third_party/eigen/Eigen/src/Core/util/XprHelper.h b/third_party/eigen/Eigen/src/Core/util/XprHelper.h index 6bb4970828e..ba5bd186d29 100644 --- a/third_party/eigen/Eigen/src/Core/util/XprHelper.h +++ b/third_party/eigen/Eigen/src/Core/util/XprHelper.h @@ -34,20 +34,6 @@ inline IndexDest convert_index(const IndexSrc& idx) { return IndexDest(idx); } -// true if T can be considered as an integral index (i.e., and integral type or enum) -template struct is_valid_index_type -{ - enum { value = -#if EIGEN_HAS_TYPE_TRAITS - internal::is_integral::value || std::is_enum::value -#elif EIGEN_COMP_MSVC - internal::is_integral::value || __is_enum(T) -#else - // without C++11, we use is_convertible to Index instead of is_integral in order to treat enums as Index. - internal::is_convertible::value && !internal::is_same::value && !is_same::value -#endif - }; -}; // promote_scalar_arg is an helper used in operation between an expression and a scalar, like: // expression * scalar @@ -104,9 +90,6 @@ class no_assignment_operator { private: no_assignment_operator& operator=(const no_assignment_operator&); - protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(no_assignment_operator) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(no_assignment_operator) }; /** \internal return the index type with the largest number of bits */ diff --git a/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h b/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h index 4354e4018fd..7f38919f778 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/ComplexSchur.h @@ -300,13 +300,10 @@ typename ComplexSchur::ComplexScalar ComplexSchur::compu ComplexScalar trace = t.coeff(0,0) + t.coeff(1,1); ComplexScalar eival1 = (trace + disc) / RealScalar(2); ComplexScalar eival2 = (trace - disc) / RealScalar(2); - RealScalar eival1_norm = numext::norm1(eival1); - RealScalar eival2_norm = numext::norm1(eival2); - // A division by zero can only occur if eival1==eival2==0. - // In this case, det==0, and all we have to do is checking that eival2_norm!=0 - if(eival1_norm > eival2_norm) + + if(numext::norm1(eival1) > numext::norm1(eival2)) eival2 = det / eival1; - else if(eival2_norm!=RealScalar(0)) + else eival1 = det / eival2; // choose the eigenvalue closest to the bottom entry of the diagonal diff --git a/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h b/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h index 9191519abe9..17ea903f5f1 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/RealSchur.h @@ -236,7 +236,7 @@ template class RealSchur typedef Matrix Vector3s; Scalar computeNormOfT(); - Index findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero); + Index findSmallSubdiagEntry(Index iu); void splitOffTwoRows(Index iu, bool computeU, const Scalar& exshift); void computeShift(Index iu, Index iter, Scalar& exshift, Vector3s& shiftInfo); void initFrancisQRStep(Index il, Index iu, const Vector3s& shiftInfo, Index& im, Vector3s& firstHouseholderVector); @@ -302,16 +302,12 @@ RealSchur& RealSchur::computeFromHessenberg(const HessMa Index totalIter = 0; // iteration count for whole matrix Scalar exshift(0); // sum of exceptional shifts Scalar norm = computeNormOfT(); - // sub-diagonal entries smaller than considerAsZero will be treated as zero. - // We use eps^2 to enable more precision in small eigenvalues. - Scalar considerAsZero = numext::maxi( norm * numext::abs2(NumTraits::epsilon()), - (std::numeric_limits::min)() ); if(norm!=Scalar(0)) { while (iu >= 0) { - Index il = findSmallSubdiagEntry(iu,considerAsZero); + Index il = findSmallSubdiagEntry(iu); // Check for convergence if (il == iu) // One root found @@ -368,17 +364,14 @@ inline typename MatrixType::Scalar RealSchur::computeNormOfT() /** \internal Look for single small sub-diagonal element and returns its index */ template -inline Index RealSchur::findSmallSubdiagEntry(Index iu, const Scalar& considerAsZero) +inline Index RealSchur::findSmallSubdiagEntry(Index iu) { using std::abs; Index res = iu; while (res > 0) { Scalar s = abs(m_matT.coeff(res-1,res-1)) + abs(m_matT.coeff(res,res)); - - s = numext::maxi(s * NumTraits::epsilon(), considerAsZero); - - if (abs(m_matT.coeff(res,res-1)) <= s) + if (abs(m_matT.coeff(res,res-1)) <= NumTraits::epsilon() * s) break; res--; } diff --git a/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h b/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h index d37656fa207..9ddd553f2f1 100644 --- a/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h +++ b/third_party/eigen/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h @@ -605,8 +605,7 @@ template struct direct_selfadjoint_eigenvalues res, Ref representative) { - EIGEN_USING_STD_MATH(sqrt) - EIGEN_USING_STD_MATH(abs) + using std::abs; Index i0; // Find non-zero column i0 (by construction, there must exist a non zero coefficient on the diagonal): mat.diagonal().cwiseAbs().maxCoeff(&i0); @@ -617,8 +616,8 @@ template struct direct_selfadjoint_eigenvaluesn1) res = c0/sqrt(n0); - else res = c1/sqrt(n1); + if(n0>n1) res = c0/std::sqrt(n0); + else res = c1/std::sqrt(n1); return true; } diff --git a/third_party/eigen/Eigen/src/Geometry/Quaternion.h b/third_party/eigen/Eigen/src/Geometry/Quaternion.h index b818206568e..c3fd8c3e0f8 100644 --- a/third_party/eigen/Eigen/src/Geometry/Quaternion.h +++ b/third_party/eigen/Eigen/src/Geometry/Quaternion.h @@ -169,38 +169,20 @@ class QuaternionBase : public RotationBase /** return the result vector of \a v through the rotation*/ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Vector3 _transformVector(const Vector3& v) const; - #ifdef EIGEN_PARSED_BY_DOXYGEN /** \returns \c *this with scalar type casted to \a NewScalarType * * Note that if \a NewScalarType is equal to the current scalar type of \c *this * then this function smartly returns a const reference to \c *this. */ template - EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const; - - #else - - template - EIGEN_DEVICE_FUNC inline - typename internal::enable_if::value,const Derived&>::type cast() const + EIGEN_DEVICE_FUNC inline typename internal::cast_return_type >::type cast() const { - return derived(); + return typename internal::cast_return_type >::type(derived()); } - template - EIGEN_DEVICE_FUNC inline - typename internal::enable_if::value,Quaternion >::type cast() const - { - return Quaternion(coeffs().template cast()); - } - #endif - #ifdef EIGEN_QUATERNIONBASE_PLUGIN # include EIGEN_QUATERNIONBASE_PLUGIN #endif -protected: - EIGEN_DEFAULT_COPY_CONSTRUCTOR(QuaternionBase) - EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(QuaternionBase) }; /*************************************************************************** diff --git a/third_party/eigen/Eigen/src/Geometry/Scaling.h b/third_party/eigen/Eigen/src/Geometry/Scaling.h index 33eabd81a4c..f58ca03d94a 100755 --- a/third_party/eigen/Eigen/src/Geometry/Scaling.h +++ b/third_party/eigen/Eigen/src/Geometry/Scaling.h @@ -14,7 +14,7 @@ namespace Eigen { /** \geometry_module \ingroup Geometry_Module * - * \class UniformScaling + * \class Scaling * * \brief Represents a generic uniform scaling transformation * diff --git a/third_party/eigen/Eigen/src/Geometry/Transform.h b/third_party/eigen/Eigen/src/Geometry/Transform.h index c21d9e550e8..3f31ee45df9 100644 --- a/third_party/eigen/Eigen/src/Geometry/Transform.h +++ b/third_party/eigen/Eigen/src/Geometry/Transform.h @@ -252,11 +252,11 @@ class Transform public: /** Default constructor without initialization of the meaningful coefficients. - * If Mode==Affine or Mode==Isometry, then the last row is set to [0 ... 0 1] */ + * If Mode==Affine, then the last row is set to [0 ... 0 1] */ EIGEN_DEVICE_FUNC inline Transform() { check_template_params(); - internal::transform_make_affine<(int(Mode)==Affine || int(Mode)==Isometry) ? Affine : AffineCompact>::run(m_matrix); + internal::transform_make_affine<(int(Mode)==Affine) ? Affine : AffineCompact>::run(m_matrix); } EIGEN_DEVICE_FUNC inline Transform(const Transform& other) diff --git a/third_party/eigen/Eigen/src/Geometry/Translation.h b/third_party/eigen/Eigen/src/Geometry/Translation.h index 0e99ce68e26..51d9a82ebbd 100644 --- a/third_party/eigen/Eigen/src/Geometry/Translation.h +++ b/third_party/eigen/Eigen/src/Geometry/Translation.h @@ -138,6 +138,12 @@ class Translation /** \returns the inverse translation (opposite) */ Translation inverse() const { return Translation(-m_coeffs); } + Translation& operator=(const Translation& other) + { + m_coeffs = other.m_coeffs; + return *this; + } + static const Translation Identity() { return Translation(VectorType::Zero()); } /** \returns \c *this with scalar type casted to \a NewScalarType diff --git a/third_party/eigen/Eigen/src/Geometry/Umeyama.h b/third_party/eigen/Eigen/src/Geometry/Umeyama.h index 6b755008fdc..7e933fca13e 100644 --- a/third_party/eigen/Eigen/src/Geometry/Umeyama.h +++ b/third_party/eigen/Eigen/src/Geometry/Umeyama.h @@ -87,7 +87,7 @@ struct umeyama_transform_matrix_type * \f{align*} * T = \begin{bmatrix} c\mathbf{R} & \mathbf{t} \\ \mathbf{0} & 1 \end{bmatrix} * \f} -* minimizing the residual above. This transformation is always returned as an +* minimizing the resudiual above. This transformation is always returned as an * Eigen::Matrix. */ template diff --git a/third_party/eigen/Eigen/src/LU/PartialPivLU.h b/third_party/eigen/Eigen/src/LU/PartialPivLU.h index 6b10f39fab4..d439618879a 100644 --- a/third_party/eigen/Eigen/src/LU/PartialPivLU.h +++ b/third_party/eigen/Eigen/src/LU/PartialPivLU.h @@ -519,10 +519,7 @@ void PartialPivLU::compute() // the row permutation is stored as int indices, so just to be sure: eigen_assert(m_lu.rows()::highest()); - if(m_lu.cols()>0) - m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); - else - m_l1_norm = RealScalar(0); + m_l1_norm = m_lu.cwiseAbs().colwise().sum().maxCoeff(); eigen_assert(m_lu.rows() == m_lu.cols() && "PartialPivLU is only for square (and moreover invertible) matrices"); const Index size = m_lu.rows(); diff --git a/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h b/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h index 4dce2ef20ee..ebb64a62b02 100644 --- a/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h +++ b/third_party/eigen/Eigen/src/LU/arch/Inverse_SSE.h @@ -44,7 +44,7 @@ struct compute_inverse_size4 static void run(const MatrixType& mat, ResultType& result) { ActualMatrixType matrix(mat); - const Packet4f p4f_sign_PNNP = _mm_castsi128_ps(_mm_set_epi32(0x00000000, 0x80000000, 0x80000000, 0x00000000)); + EIGEN_ALIGN16 const unsigned int _Sign_PNNP[4] = { 0x00000000, 0x80000000, 0x80000000, 0x00000000 }; // Load the full matrix into registers __m128 _L1 = matrix.template packet( 0); @@ -139,7 +139,7 @@ struct compute_inverse_size4 iC = _mm_sub_ps(iC, _mm_mul_ps(_mm_shuffle_ps(A,A,0xB1), _mm_shuffle_ps(DC,DC,0x66))); rd = _mm_shuffle_ps(rd,rd,0); - rd = _mm_xor_ps(rd, p4f_sign_PNNP); + rd = _mm_xor_ps(rd, _mm_load_ps((float*)_Sign_PNNP)); // iB = C*|B| - D*B#*A iB = _mm_sub_ps(_mm_mul_ps(C,_mm_shuffle_ps(dB,dB,0)), iB); diff --git a/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h b/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h index 98d0e3f21ab..091c3970e81 100644 --- a/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h +++ b/third_party/eigen/Eigen/src/PardisoSupport/PardisoSupport.h @@ -192,8 +192,7 @@ class PardisoImpl : public SparseSolverBase void pardisoInit(int type) { m_type = type; - EIGEN_USING_STD_MATH(abs); - bool symmetric = abs(m_type) < 10; + bool symmetric = std::abs(m_type) < 10; m_iparm[0] = 1; // No solver default m_iparm[1] = 2; // use Metis for the ordering m_iparm[2] = 0; // Reserved. Set to zero. (??Numbers of processors, value of OMP_NUM_THREADS??) diff --git a/third_party/eigen/Eigen/src/SVD/BDCSVD.h b/third_party/eigen/Eigen/src/SVD/BDCSVD.h index a5b73f8f21f..1134d66e7e9 100644 --- a/third_party/eigen/Eigen/src/SVD/BDCSVD.h +++ b/third_party/eigen/Eigen/src/SVD/BDCSVD.h @@ -768,21 +768,6 @@ void BDCSVD::computeSingVals(const ArrayRef& col0, const ArrayRef& d // measure everything relative to shift Map diagShifted(m_workspace.data()+4*n, n); diagShifted = diag - shift; - - if(k!=actual_n-1) - { - // check that after the shift, f(mid) is still negative: - RealScalar midShifted = (right - left) / RealScalar(2); - if(shift==right) - midShifted = -midShifted; - RealScalar fMidShifted = secularEq(midShifted, col0, diag, perm, diagShifted, shift); - if(fMidShifted>0) - { - // fMid was erroneous, fix it: - shift = fMidShifted > Literal(0) ? left : right; - diagShifted = diag - shift; - } - } // initial guess RealScalar muPrev, muCur; @@ -860,13 +845,11 @@ void BDCSVD::computeSingVals(const ArrayRef& col0, const ArrayRef& d } RealScalar fLeft = secularEq(leftShifted, col0, diag, perm, diagShifted, shift); - eigen_internal_assert(fLeft::computeSingVals(const ArrayRef& col0, const ArrayRef& d } #endif eigen_internal_assert(fLeft * fRight < Literal(0)); - - if(fLeft Literal(2) * NumTraits::epsilon() * numext::maxi(abs(leftShifted), abs(rightShifted))) { - while (rightShifted - leftShifted > Literal(2) * NumTraits::epsilon() * numext::maxi(abs(leftShifted), abs(rightShifted))) + RealScalar midShifted = (leftShifted + rightShifted) / Literal(2); + fMid = secularEq(midShifted, col0, diag, perm, diagShifted, shift); + if (fLeft * fMid < Literal(0)) { - RealScalar midShifted = (leftShifted + rightShifted) / Literal(2); - fMid = secularEq(midShifted, col0, diag, perm, diagShifted, shift); - eigen_internal_assert((numext::isfinite)(fMid)); - - if (fLeft * fMid < Literal(0)) - { - rightShifted = midShifted; - } - else - { - leftShifted = midShifted; - fLeft = fMid; - } + rightShifted = midShifted; + } + else + { + leftShifted = midShifted; + fLeft = fMid; } - muCur = (leftShifted + rightShifted) / Literal(2); - } - else - { - // We have a problem as shifting on the left or right give either a positive or negative value - // at the middle of [left,right]... - // Instead fo abbording or entering an infinite loop, - // let's just use the middle as the estimated zero-crossing: - muCur = (right - left) * RealScalar(0.5); - if(shift == right) - muCur = -muCur; } + + muCur = (leftShifted + rightShifted) / Literal(2); } singVals[k] = shift + muCur; @@ -955,7 +924,7 @@ void BDCSVD::perturbCol0 Index j = i 0.9 ) + if(i!=k && std::abs(((singVals(j)+dk)*(mus(j)+(shifts(j)-dk)))/((diag(i)+dk)*(diag(i)-dk)) - 1) > 0.9 ) std::cout << " " << ((singVals(j)+dk)*(mus(j)+(shifts(j)-dk)))/((diag(i)+dk)*(diag(i)-dk)) << " == (" << (singVals(j)+dk) << " * " << (mus(j)+(shifts(j)-dk)) << ") / (" << (diag(i)+dk) << " * " << (diag(i)-dk) << ")\n"; #endif @@ -965,7 +934,7 @@ void BDCSVD::perturbCol0 std::cout << "zhat(" << k << ") = sqrt( " << prod << ") ; " << (singVals(last) + dk) << " * " << mus(last) + shifts(last) << " - " << dk << "\n"; #endif RealScalar tmp = sqrt(prod); - zhat(k) = col0(k) > Literal(0) ? RealScalar(tmp) : RealScalar(-tmp); + zhat(k) = col0(k) > Literal(0) ? tmp : -tmp; } } } diff --git a/third_party/eigen/Eigen/src/SVD/SVDBase.h b/third_party/eigen/Eigen/src/SVD/SVDBase.h index 53da2848883..3d1ef373ea7 100644 --- a/third_party/eigen/Eigen/src/SVD/SVDBase.h +++ b/third_party/eigen/Eigen/src/SVD/SVDBase.h @@ -183,7 +183,7 @@ class SVDBase // this temporary is needed to workaround a MSVC issue Index diagSize = (std::max)(1,m_diagSize); return m_usePrescribedThreshold ? m_prescribedThreshold - : RealScalar(diagSize)*NumTraits::epsilon(); + : diagSize*NumTraits::epsilon(); } /** \returns true if \a U (full or thin) is asked for in this SVD decomposition */ diff --git a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h index 369e6804afa..2907f65296b 100644 --- a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h +++ b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky.h @@ -608,7 +608,7 @@ template } if(Base::m_diag.size()>0) - dest = Base::m_diag.real().asDiagonal().inverse() * dest; + dest = Base::m_diag.asDiagonal().inverse() * dest; if (Base::m_matrix.nonZeros()>0) // otherwise I==I { diff --git a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h index 7b6183d08b1..31e06995b8a 100644 --- a/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h +++ b/third_party/eigen/Eigen/src/SparseCholesky/SimplicialCholesky_impl.h @@ -156,7 +156,7 @@ void SimplicialCholeskyBase::factorize_preordered(const CholMatrixType& /* the nonzero entry L(k,i) */ Scalar l_ki; if(DoLDLT) - l_ki = yi / numext::real(m_diag[i]); + l_ki = yi / m_diag[i]; else yi = l_ki = yi / Lx[Lp[i]]; diff --git a/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h b/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h index 2cb7747cc99..cec49d1841f 100644 --- a/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h +++ b/third_party/eigen/Eigen/src/SparseCore/AmbiVector.h @@ -28,7 +28,7 @@ class AmbiVector typedef typename NumTraits::Real RealScalar; explicit AmbiVector(Index size) - : m_buffer(0), m_zero(0), m_size(0), m_end(0), m_allocatedSize(0), m_allocatedElements(0), m_mode(-1) + : m_buffer(0), m_zero(0), m_size(0), m_allocatedSize(0), m_allocatedElements(0), m_mode(-1) { resize(size); } @@ -147,8 +147,7 @@ template void AmbiVector<_Scalar,_StorageIndex>::init(int mode) { m_mode = mode; - // This is only necessary in sparse mode, but we set these unconditionally to avoid some maybe-uninitialized warnings - // if (m_mode==IsSparse) + if (m_mode==IsSparse) { m_llSize = 0; m_llStart = -1; @@ -268,7 +267,7 @@ _Scalar& AmbiVector<_Scalar,_StorageIndex>::coeff(Index i) else { Index elid = m_llStart; - while (elid >= 0 && llElements[elid].index= 0 && (llElements[elid].index) class unary_evaluator, IteratorBased>::InnerIterator : public unary_evaluator, IteratorBased>::EvalIterator { - protected: typedef typename XprType::Scalar Scalar; typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; public: @@ -100,7 +99,6 @@ template class unary_evaluator, IteratorBased>::InnerIterator : public unary_evaluator, IteratorBased>::EvalIterator { - protected: typedef typename XprType::Scalar Scalar; typedef typename unary_evaluator, IteratorBased>::EvalIterator Base; public: diff --git a/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h b/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h index a5396538ba9..0a2490bcc3d 100644 --- a/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h +++ b/third_party/eigen/Eigen/src/SparseCore/SparseMatrix.h @@ -327,8 +327,7 @@ class SparseMatrix m_outerIndex[j] = newOuterIndex[j]; m_innerNonZeros[j] = innerNNZ; } - if(m_outerSize>0) - m_outerIndex[m_outerSize] = m_outerIndex[m_outerSize-1] + m_innerNonZeros[m_outerSize-1] + reserveSizes[m_outerSize-1]; + m_outerIndex[m_outerSize] = m_outerIndex[m_outerSize-1] + m_innerNonZeros[m_outerSize-1] + reserveSizes[m_outerSize-1]; m_data.resize(m_outerIndex[m_outerSize]); } diff --git a/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h b/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h index 76117a0107e..65611b3d4ca 100644 --- a/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h +++ b/third_party/eigen/Eigen/src/SparseCore/SparseSelfAdjointView.h @@ -453,7 +453,7 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix, IteratorBased> class InnerIterator : public EvalIterator { - protected: typedef typename XprType::Scalar Scalar; public: diff --git a/third_party/eigen/Eigen/src/SparseLU/SparseLU.h b/third_party/eigen/Eigen/src/SparseLU/SparseLU.h index 87f0efe37ed..7104831c03b 100644 --- a/third_party/eigen/Eigen/src/SparseLU/SparseLU.h +++ b/third_party/eigen/Eigen/src/SparseLU/SparseLU.h @@ -43,8 +43,8 @@ template struct SparseLUMatrixURetu * Simple example with key steps * \code * VectorXd x(n), b(n); - * SparseMatrix A; - * SparseLU, COLAMDOrdering > solver; + * SparseMatrix A; + * SparseLU, COLAMDOrdering > solver; * // fill A and b; * // Compute the ordering permutation vector from the structural pattern of A * solver.analyzePattern(A); diff --git a/third_party/eigen/Eigen/src/StlSupport/StdDeque.h b/third_party/eigen/Eigen/src/StlSupport/StdDeque.h index af158f425d4..cf1fedf9273 100644 --- a/third_party/eigen/Eigen/src/StlSupport/StdDeque.h +++ b/third_party/eigen/Eigen/src/StlSupport/StdDeque.h @@ -98,10 +98,8 @@ namespace std { { return deque_base::insert(position,x); } void insert(const_iterator position, size_type new_size, const value_type& x) { deque_base::insert(position, new_size, x); } -#elif defined(_GLIBCXX_DEQUE) && EIGEN_GNUC_AT_LEAST(4,2) && !EIGEN_GNUC_AT_LEAST(10, 1) +#elif defined(_GLIBCXX_DEQUE) && EIGEN_GNUC_AT_LEAST(4,2) // workaround GCC std::deque implementation - // GCC 10.1 doesn't let us access _Deque_impl _M_impl anymore and we have to - // fall-back to the default case void resize(size_type new_size, const value_type& x) { if (new_size < deque_base::size()) @@ -110,7 +108,7 @@ namespace std { deque_base::insert(deque_base::end(), new_size - deque_base::size(), x); } #else - // either non-GCC or GCC between 4.1 and 10.1 + // either GCC 4.1 or non-GCC // default implementation which should always work. void resize(size_type new_size, const value_type& x) { diff --git a/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h b/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h index 05a7449bc9e..1f8a531af59 100644 --- a/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h +++ b/third_party/eigen/Eigen/src/plugins/ArrayCwiseBinaryOps.h @@ -119,7 +119,7 @@ OP(const Scalar& s) const { \ return this->OP(Derived::PlainObject::Constant(rows(), cols(), s)); \ } \ EIGEN_DEVICE_FUNC friend EIGEN_STRONG_INLINE const RCmp ## COMPARATOR ## ReturnType \ -OP(const Scalar& s, const EIGEN_CURRENT_STORAGE_BASE_CLASS& d) { \ +OP(const Scalar& s, const Derived& d) { \ return Derived::PlainObject::Constant(d.rows(), d.cols(), s).OP(d); \ } diff --git a/third_party/eigen/bench/bench_gemm.cpp b/third_party/eigen/bench/bench_gemm.cpp index dccab96a8bb..8528c558742 100644 --- a/third_party/eigen/bench/bench_gemm.cpp +++ b/third_party/eigen/bench/bench_gemm.cpp @@ -112,7 +112,6 @@ void matlab_cplx_cplx(const M& ar, const M& ai, const M& br, const M& bi, M& cr, cr.noalias() -= ai * bi; ci.noalias() += ar * bi; ci.noalias() += ai * br; - // [cr ci] += [ar ai] * br + [-ai ar] * bi } void matlab_real_cplx(const M& a, const M& br, const M& bi, M& cr, M& ci) @@ -241,7 +240,7 @@ int main(int argc, char ** argv) blas_gemm(a,b,r); c.noalias() += a * b; if(!r.isApprox(c)) { - std::cout << (r - c).norm() << "\n"; + std::cout << r - c << "\n"; std::cerr << "Warning, your product is crap!\n\n"; } #else @@ -250,7 +249,7 @@ int main(int argc, char ** argv) gemm(a,b,c); r.noalias() += a.cast() .lazyProduct( b.cast() ); if(!r.isApprox(c)) { - std::cout << (r - c).norm() << "\n"; + std::cout << r - c << "\n"; std::cerr << "Warning, your product is crap!\n\n"; } } diff --git a/third_party/eigen/blas/level3_impl.h b/third_party/eigen/blas/level3_impl.h index 6dd6338b432..6c802cd5fdb 100644 --- a/third_party/eigen/blas/level3_impl.h +++ b/third_party/eigen/blas/level3_impl.h @@ -13,28 +13,28 @@ int EIGEN_BLAS_FUNC(gemm)(const char *opa, const char *opb, const int *m, const const RealScalar *pa, const int *lda, const RealScalar *pb, const int *ldb, const RealScalar *pbeta, RealScalar *pc, const int *ldc) { // std::cerr << "in gemm " << *opa << " " << *opb << " " << *m << " " << *n << " " << *k << " " << *lda << " " << *ldb << " " << *ldc << " " << *palpha << " " << *pbeta << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, Scalar, internal::level3_blocking&, Eigen::internal::GemmParallelInfo*); + typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, Scalar, internal::level3_blocking&, Eigen::internal::GemmParallelInfo*); static const functype func[12] = { // array index: NOTR | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (NOTR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0, // array index: NOTR | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (TR << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0, // array index: NOTR | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: TR | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), // array index: ADJ | (ADJ << 2) - (internal::general_matrix_matrix_product::run), + (internal::general_matrix_matrix_product::run), 0 }; @@ -71,7 +71,7 @@ int EIGEN_BLAS_FUNC(gemm)(const char *opa, const char *opb, const int *m, const internal::gemm_blocking_space blocking(*m,*n,*k,1,true); int code = OP(*opa) | (OP(*opb) << 2); - func[code](*m, *n, *k, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking, 0); + func[code](*m, *n, *k, a, *lda, b, *ldb, c, *ldc, alpha, blocking, 0); return 0; } @@ -79,63 +79,63 @@ int EIGEN_BLAS_FUNC(trsm)(const char *side, const char *uplo, const char *opa, c const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { // std::cerr << "in trsm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << "," << *n << " " << *palpha << " " << *lda << " " << *ldb<< "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, internal::level3_blocking&); static const functype func[32] = { // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run),\ + (internal::triangular_solve_matrix::run),\ 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::triangular_solve_matrix::run), + (internal::triangular_solve_matrix::run), 0 }; @@ -163,12 +163,12 @@ int EIGEN_BLAS_FUNC(trsm)(const char *side, const char *uplo, const char *opa, c if(SIDE(*side)==LEFT) { internal::gemm_blocking_space blocking(*m,*n,*m,1,false); - func[code](*m, *n, a, *lda, b, 1, *ldb, blocking); + func[code](*m, *n, a, *lda, b, *ldb, blocking); } else { internal::gemm_blocking_space blocking(*m,*n,*n,1,false); - func[code](*n, *m, a, *lda, b, 1, *ldb, blocking); + func[code](*n, *m, a, *lda, b, *ldb, blocking); } if(alpha!=Scalar(1)) @@ -184,63 +184,63 @@ int EIGEN_BLAS_FUNC(trmm)(const char *side, const char *uplo, const char *opa, c const RealScalar *palpha, const RealScalar *pa, const int *lda, RealScalar *pb, const int *ldb) { // std::cerr << "in trmm " << *side << " " << *uplo << " " << *opa << " " << *diag << " " << *m << " " << *n << " " << *lda << " " << *ldb << " " << *palpha << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[32] = { // array index: NOTR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (NUNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (UP << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (LEFT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0, // array index: NOTR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: TR | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), // array index: ADJ | (RIGHT << 2) | (LO << 3) | (UNIT << 4) - (internal::product_triangular_matrix_matrix::run), + (internal::product_triangular_matrix_matrix::run), 0 }; @@ -272,12 +272,12 @@ int EIGEN_BLAS_FUNC(trmm)(const char *side, const char *uplo, const char *opa, c if(SIDE(*side)==LEFT) { internal::gemm_blocking_space blocking(*m,*n,*m,1,false); - func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, 1, *ldb, alpha, blocking); + func[code](*m, *n, *m, a, *lda, tmp.data(), tmp.outerStride(), b, *ldb, alpha, blocking); } else { internal::gemm_blocking_space blocking(*m,*n,*n,1,false); - func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, 1, *ldb, alpha, blocking); + func[code](*m, *n, *n, tmp.data(), tmp.outerStride(), a, *lda, b, *ldb, alpha, blocking); } return 1; } @@ -338,12 +338,12 @@ int EIGEN_BLAS_FUNC(symm)(const char *side, const char *uplo, const int *m, cons internal::gemm_blocking_space blocking(*m,*n,size,1,false); if(SIDE(*side)==LEFT) - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); else return 0; else if(SIDE(*side)==RIGHT) - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); else return 0; else return 0; @@ -359,21 +359,21 @@ int EIGEN_BLAS_FUNC(syrk)(const char *uplo, const char *op, const int *n, const { // std::cerr << "in syrk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " << *pbeta << " " << *ldc << "\n"; #if !ISCOMPLEX - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[8] = { // array index: NOTR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: TR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: ADJ | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: NOTR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: TR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), // array index: ADJ | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0 }; #endif @@ -426,7 +426,7 @@ int EIGEN_BLAS_FUNC(syrk)(const char *uplo, const char *op, const int *n, const internal::gemm_blocking_space blocking(*n,*n,*k,1,false); int code = OP(*op) | (UPLO(*uplo) << 2); - func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); + func[code](*n, *k, a, *lda, a, *lda, c, *ldc, alpha, blocking); #endif return 0; @@ -537,18 +537,18 @@ int EIGEN_BLAS_FUNC(hemm)(const char *side, const char *uplo, const int *m, cons if(SIDE(*side)==LEFT) { - if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix - ::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix - ::run(*m, *n, a, *lda, b, *ldb, c, 1, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) internal::product_selfadjoint_matrix + ::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix + ::run(*m, *n, a, *lda, b, *ldb, c, *ldc, alpha, blocking); else return 0; } else if(SIDE(*side)==RIGHT) { - if(UPLO(*uplo)==UP) matrix(c,*m,*n,*ldc) += alpha * matrix(b,*m,*n,*ldb) * matrix(a,*n,*n,*lda).selfadjointView();/*internal::product_selfadjoint_matrix - ::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking);*/ - else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix - ::run(*m, *n, b, *ldb, a, *lda, c, 1, *ldc, alpha, blocking); + if(UPLO(*uplo)==UP) matrix(c,*m,*n,*ldc) += alpha * matrix(b,*m,*n,*ldb) * matrix(a,*n,*n,*lda).selfadjointView();/*internal::product_selfadjoint_matrix + ::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking);*/ + else if(UPLO(*uplo)==LO) internal::product_selfadjoint_matrix + ::run(*m, *n, b, *ldb, a, *lda, c, *ldc, alpha, blocking); else return 0; } else @@ -566,19 +566,19 @@ int EIGEN_BLAS_FUNC(herk)(const char *uplo, const char *op, const int *n, const { // std::cerr << "in herk " << *uplo << " " << *op << " " << *n << " " << *k << " " << *palpha << " " << *lda << " " << *pbeta << " " << *ldc << "\n"; - typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, DenseIndex, const Scalar&, internal::level3_blocking&); + typedef void (*functype)(DenseIndex, DenseIndex, const Scalar *, DenseIndex, const Scalar *, DenseIndex, Scalar *, DenseIndex, const Scalar&, internal::level3_blocking&); static const functype func[8] = { // array index: NOTR | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: ADJ | (UP << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: NOTR | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0, // array index: ADJ | (LO << 2) - (internal::general_matrix_matrix_triangular_product::run), + (internal::general_matrix_matrix_triangular_product::run), 0 }; @@ -620,7 +620,7 @@ int EIGEN_BLAS_FUNC(herk)(const char *uplo, const char *op, const int *n, const if(*k>0 && alpha!=RealScalar(0)) { internal::gemm_blocking_space blocking(*n,*n,*k,1,false); - func[code](*n, *k, a, *lda, a, *lda, c, 1, *ldc, alpha, blocking); + func[code](*n, *k, a, *lda, a, *lda, c, *ldc, alpha, blocking); matrix(c, *n, *n, *ldc).diagonal().imag().setZero(); } return 0; diff --git a/third_party/eigen/cmake/EigenTesting.cmake b/third_party/eigen/cmake/EigenTesting.cmake index 3d0074c71ad..a92a2978b0d 100644 --- a/third_party/eigen/cmake/EigenTesting.cmake +++ b/third_party/eigen/cmake/EigenTesting.cmake @@ -677,8 +677,6 @@ macro(ei_set_build_string) set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-cxx11) endif() - set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-v3.3) - if(EIGEN_BUILD_STRING_SUFFIX) set(TMP_BUILD_STRING ${TMP_BUILD_STRING}-${EIGEN_BUILD_STRING_SUFFIX}) endif() diff --git a/third_party/eigen/cmake/FindStandardMathLibrary.cmake b/third_party/eigen/cmake/FindStandardMathLibrary.cmake index 337f1b30469..711b0e4b4f1 100644 --- a/third_party/eigen/cmake/FindStandardMathLibrary.cmake +++ b/third_party/eigen/cmake/FindStandardMathLibrary.cmake @@ -19,11 +19,8 @@ include(CheckCXXSourceCompiles) # notice the std:: is required on some platforms such as QNX set(find_standard_math_library_test_program -" -#include -int main(int argc, char **){ - return int(std::sin(double(argc)) + std::log(double(argc))); -}") +"#include +int main() { std::sin(0.0); std::log(0.0f); }") # first try compiling/linking the test program without any linker flags diff --git a/third_party/eigen/debug/msvc/eigen.natvis b/third_party/eigen/debug/msvc/eigen.natvis index da898571797..22cf346576d 100644 --- a/third_party/eigen/debug/msvc/eigen.natvis +++ b/third_party/eigen/debug/msvc/eigen.natvis @@ -1,235 +1,235 @@ - - - - - - - - [{$T2}, {$T3}] (fixed matrix) - - - 2 - $i==0 ? $T2 : $T3 - m_storage.m_data.array - - - Backward - 2 - $i==0 ? $T2 : $T3 - m_storage.m_data.array - - - - - - - - [2, 2] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[2]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[3]}) - - - - - - - - [3, 3] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[3]}, {m_storage.m_data.array[6]}) - - - ({m_storage.m_data.array[3]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[7]}) - - - ({m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[8]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[8]}) - - - - - - - - [4, 4] (fixed matrix) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - - ({m_storage.m_data.array[0]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[8]}, {m_storage.m_data.array[12]}) - - - ({m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}) - - - ({m_storage.m_data.array[1]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[13]}) - - - ({m_storage.m_data.array[8]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[11]}) - - - ({m_storage.m_data.array[2]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[14]}) - - - ({m_storage.m_data.array[12]}, {m_storage.m_data.array[13]}, {m_storage.m_data.array[14]}, {m_storage.m_data.array[15]}) - - - ({m_storage.m_data.array[3]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[11]}, {m_storage.m_data.array[15]}) - - - - - - - - empty - [{m_storage.m_rows}, {m_storage.m_cols}] (dynamic matrix) - - - 2 - $i==0 ? m_storage.m_rows : m_storage.m_cols - m_storage.m_data - - - Backward - 2 - $i==0 ? m_storage.m_rows : m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{$T2}, {m_storage.m_cols}] (dynamic column matrix) - - - 2 - $i==0 ? $T2 : m_storage.m_cols - m_storage.m_data - - - Backward - 2 - $i==0 ? $T2 : m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{m_storage.m_rows}, {$T2}] (dynamic row matrix) - - - 2 - $i==0 ? m_storage.m_rows : $T2 - m_storage.m_data - - - Backward - 2 - $i==0 ? m_storage.m_rows : $T2 - m_storage.m_data - - - - - - - - empty - [{m_storage.m_cols}] (dynamic column vector) - - m_storage.m_cols - - m_storage.m_cols - m_storage.m_data - - - - - - - - empty - [{m_storage.m_rows}] (dynamic row vector) - - m_storage.m_rows - - m_storage.m_rows - m_storage.m_data - - - - - - - - [1] ({m_storage.m_data.array[0]}) - - m_storage.m_data.array[0] - - - - - - - - [2] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - - - - - - - - [3] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - m_storage.m_data.array[2] - - - - - - - - [4] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) - - m_storage.m_data.array[0] - m_storage.m_data.array[1] - m_storage.m_data.array[2] - m_storage.m_data.array[3] - - - - + + + + + + + + [{$T2}, {$T3}] (fixed matrix) + + + 2 + $i==0 ? $T2 : $T3 + m_storage.m_data.array + + + Backward + 2 + $i==0 ? $T2 : $T3 + m_storage.m_data.array + + + + + + + + [2, 2] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[2]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[3]}) + + + + + + + + [3, 3] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[3]}, {m_storage.m_data.array[6]}) + + + ({m_storage.m_data.array[3]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[7]}) + + + ({m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[8]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[8]}) + + + + + + + + [4, 4] (fixed matrix) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + + ({m_storage.m_data.array[0]}, {m_storage.m_data.array[4]}, {m_storage.m_data.array[8]}, {m_storage.m_data.array[12]}) + + + ({m_storage.m_data.array[4]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[7]}) + + + ({m_storage.m_data.array[1]}, {m_storage.m_data.array[5]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[13]}) + + + ({m_storage.m_data.array[8]}, {m_storage.m_data.array[9]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[11]}) + + + ({m_storage.m_data.array[2]}, {m_storage.m_data.array[6]}, {m_storage.m_data.array[10]}, {m_storage.m_data.array[14]}) + + + ({m_storage.m_data.array[12]}, {m_storage.m_data.array[13]}, {m_storage.m_data.array[14]}, {m_storage.m_data.array[15]}) + + + ({m_storage.m_data.array[3]}, {m_storage.m_data.array[7]}, {m_storage.m_data.array[11]}, {m_storage.m_data.array[15]}) + + + + + + + + empty + [{m_storage.m_rows}, {m_storage.m_cols}] (dynamic matrix) + + + 2 + $i==0 ? m_storage.m_rows : m_storage.m_cols + m_storage.m_data + + + Backward + 2 + $i==0 ? m_storage.m_rows : m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{$T2}, {m_storage.m_cols}] (dynamic column matrix) + + + 2 + $i==0 ? $T2 : m_storage.m_cols + m_storage.m_data + + + Backward + 2 + $i==0 ? $T2 : m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{m_storage.m_rows}, {$T2}] (dynamic row matrix) + + + 2 + $i==0 ? m_storage.m_rows : $T2 + m_storage.m_data + + + Backward + 2 + $i==0 ? m_storage.m_rows : $T2 + m_storage.m_data + + + + + + + + empty + [{m_storage.m_cols}] (dynamic column vector) + + m_storage.m_cols + + m_storage.m_cols + m_storage.m_data + + + + + + + + empty + [{m_storage.m_rows}] (dynamic row vector) + + m_storage.m_rows + + m_storage.m_rows + m_storage.m_data + + + + + + + + [1] ({m_storage.m_data.array[0]}) + + m_storage.m_data.array[0] + + + + + + + + [2] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + + + + + + + + [3] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + m_storage.m_data.array[2] + + + + + + + + [4] ({m_storage.m_data.array[0]}, {m_storage.m_data.array[1]}, {m_storage.m_data.array[2]}, {m_storage.m_data.array[3]}) + + m_storage.m_data.array[0] + m_storage.m_data.array[1] + m_storage.m_data.array[2] + m_storage.m_data.array[3] + + + + diff --git a/third_party/eigen/debug/msvc/eigen_autoexp_part.dat b/third_party/eigen/debug/msvc/eigen_autoexp_part.dat index 07aa43739e5..ba7eefc8e8a 100644 --- a/third_party/eigen/debug/msvc/eigen_autoexp_part.dat +++ b/third_party/eigen/debug/msvc/eigen_autoexp_part.dat @@ -1,295 +1,295 @@ -; *************************************************************** -; * Eigen Visualizer -; * -; * Author: Hauke Heibel -; * -; * Support the enhanced debugging of the following Eigen -; * types (*: any, +:fixed dimension) : -; * -; * - Eigen::Matrix<*,4,1,*,*,*> and Eigen::Matrix<*,1,4,*,*,*> -; * - Eigen::Matrix<*,3,1,*,*,*> and Eigen::Matrix<*,1,3,*,*,*> -; * - Eigen::Matrix<*,2,1,*,*,*> and Eigen::Matrix<*,1,2,*,*,*> -; * - Eigen::Matrix<*,-1,-1,*,*,*> -; * - Eigen::Matrix<*,+,-1,*,*,*> -; * - Eigen::Matrix<*,-1,+,*,*,*> -; * - Eigen::Matrix<*,+,+,*,*,*> -; * -; * Matrices are displayed properly independantly of the memory -; * alignment (RowMajor vs. ColMajor). -; * -; * This file is distributed WITHOUT ANY WARRANTY. Please ensure -; * that your original autoexp.dat file is copied to a safe -; * place before proceeding with its modification. -; *************************************************************** - -[Visualizer] - -; Fixed size 4-vectors -Eigen::Matrix<*,4,1,*,*,*>|Eigen::Matrix<*,1,4,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1], - z : ($c.m_storage.m_data.array)[2], - w : ($c.m_storage.m_data.array)[3] - ) - ) - - preview - ( - #( - "[", - 4, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 4), - ")" - ) - ) -} - -; Fixed size 3-vectors -Eigen::Matrix<*,3,1,*,*,*>|Eigen::Matrix<*,1,3,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1], - z : ($c.m_storage.m_data.array)[2] - ) - ) - - preview - ( - #( - "[", - 3, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 3), - ")" - ) - ) -} - -; Fixed size 2-vectors -Eigen::Matrix<*,2,1,*,*,*>|Eigen::Matrix<*,1,2,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0], - y : ($c.m_storage.m_data.array)[1] - ) - ) - - preview - ( - #( - "[", - 2, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 2), - ")" - ) - ) -} - -; Fixed size 1-vectors -Eigen::Matrix<*,1,1,*,*,*>|Eigen::Matrix<*,1,1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - x : ($c.m_storage.m_data.array)[0] - ) - ) - - preview - ( - #( - "[", - 1, - "](", - #array(expr: $e.m_storage.m_data.array[$i], size: 1), - ")" - ) - ) -} - -; Dynamic matrices (ColMajor and RowMajor support) -Eigen::Matrix<*,-1,-1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.m_storage.m_rows, - cols: $c.m_storage.m_cols, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.m_storage.m_cols + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.m_storage.m_rows, - ",", - $c.m_storage.m_cols, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.m_storage.m_rows*$c.m_storage.m_cols - ), - ")" - ) - ) -} - -; Fixed rows, dynamic columns matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,*,-1,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.RowsAtCompileTime, - cols: $c.m_storage.m_cols, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.RowsAtCompileTime)*$c.m_storage.m_cols + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.RowsAtCompileTime, - ",", - $c.m_storage.m_cols, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.RowsAtCompileTime*$c.m_storage.m_cols - ), - ")" - ) - ) -} - -; Dynamic rows, fixed columns matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,-1,*,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.m_storage.m_rows, - cols: $c.ColsAtCompileTime, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.ColsAtCompileTime + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data)[$i], - size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.m_storage.m_rows, - ",", - $c.ColsAtCompileTime, - "](", - #array( - expr : [($c.m_storage.m_data)[$i],g], - size : $c.m_storage.m_rows*$c.ColsAtCompileTime - ), - ")" - ) - ) -} - -; Fixed size matrix (ColMajor and RowMajor support) -Eigen::Matrix<*,*,*,*,*,*>{ - children - ( - #( - [internals]: [$c,!], - rows: $c.RowsAtCompileTime, - cols: $c.ColsAtCompileTime, - ; Check for RowMajorBit - #if ($c.Flags & 0x1) ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data.array)[($i % $c.RowsAtCompileTime)*$c.ColsAtCompileTime + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime - ) - ) #else ( - #array( - rank: 2, - base: 0, - expr: ($c.m_storage.m_data.array)[$i], - size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime - ) - ) - ) - ) - - preview - ( - #( - "[", - $c.RowsAtCompileTime, - ",", - $c.ColsAtCompileTime, - "](", - #array( - expr : [($c.m_storage.m_data.array)[$i],g], - size : $c.RowsAtCompileTime*$c.ColsAtCompileTime - ), - ")" - ) - ) -} +; *************************************************************** +; * Eigen Visualizer +; * +; * Author: Hauke Heibel +; * +; * Support the enhanced debugging of the following Eigen +; * types (*: any, +:fixed dimension) : +; * +; * - Eigen::Matrix<*,4,1,*,*,*> and Eigen::Matrix<*,1,4,*,*,*> +; * - Eigen::Matrix<*,3,1,*,*,*> and Eigen::Matrix<*,1,3,*,*,*> +; * - Eigen::Matrix<*,2,1,*,*,*> and Eigen::Matrix<*,1,2,*,*,*> +; * - Eigen::Matrix<*,-1,-1,*,*,*> +; * - Eigen::Matrix<*,+,-1,*,*,*> +; * - Eigen::Matrix<*,-1,+,*,*,*> +; * - Eigen::Matrix<*,+,+,*,*,*> +; * +; * Matrices are displayed properly independantly of the memory +; * alignment (RowMajor vs. ColMajor). +; * +; * This file is distributed WITHOUT ANY WARRANTY. Please ensure +; * that your original autoexp.dat file is copied to a safe +; * place before proceeding with its modification. +; *************************************************************** + +[Visualizer] + +; Fixed size 4-vectors +Eigen::Matrix<*,4,1,*,*,*>|Eigen::Matrix<*,1,4,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1], + z : ($c.m_storage.m_data.array)[2], + w : ($c.m_storage.m_data.array)[3] + ) + ) + + preview + ( + #( + "[", + 4, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 4), + ")" + ) + ) +} + +; Fixed size 3-vectors +Eigen::Matrix<*,3,1,*,*,*>|Eigen::Matrix<*,1,3,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1], + z : ($c.m_storage.m_data.array)[2] + ) + ) + + preview + ( + #( + "[", + 3, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 3), + ")" + ) + ) +} + +; Fixed size 2-vectors +Eigen::Matrix<*,2,1,*,*,*>|Eigen::Matrix<*,1,2,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0], + y : ($c.m_storage.m_data.array)[1] + ) + ) + + preview + ( + #( + "[", + 2, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 2), + ")" + ) + ) +} + +; Fixed size 1-vectors +Eigen::Matrix<*,1,1,*,*,*>|Eigen::Matrix<*,1,1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + x : ($c.m_storage.m_data.array)[0] + ) + ) + + preview + ( + #( + "[", + 1, + "](", + #array(expr: $e.m_storage.m_data.array[$i], size: 1), + ")" + ) + ) +} + +; Dynamic matrices (ColMajor and RowMajor support) +Eigen::Matrix<*,-1,-1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.m_storage.m_rows, + cols: $c.m_storage.m_cols, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.m_storage.m_cols + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.m_storage.m_cols + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.m_storage.m_rows, + ",", + $c.m_storage.m_cols, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.m_storage.m_rows*$c.m_storage.m_cols + ), + ")" + ) + ) +} + +; Fixed rows, dynamic columns matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,*,-1,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.RowsAtCompileTime, + cols: $c.m_storage.m_cols, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.RowsAtCompileTime)*$c.m_storage.m_cols + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.m_storage.m_cols + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.RowsAtCompileTime, + ",", + $c.m_storage.m_cols, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.RowsAtCompileTime*$c.m_storage.m_cols + ), + ")" + ) + ) +} + +; Dynamic rows, fixed columns matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,-1,*,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.m_storage.m_rows, + cols: $c.ColsAtCompileTime, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[($i % $c.m_storage.m_rows)*$c.ColsAtCompileTime + (($i- $i % $c.m_storage.m_rows)/$c.m_storage.m_rows)], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data)[$i], + size: ($r==1)*$c.m_storage.m_rows+($r==0)*$c.ColsAtCompileTime + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.m_storage.m_rows, + ",", + $c.ColsAtCompileTime, + "](", + #array( + expr : [($c.m_storage.m_data)[$i],g], + size : $c.m_storage.m_rows*$c.ColsAtCompileTime + ), + ")" + ) + ) +} + +; Fixed size matrix (ColMajor and RowMajor support) +Eigen::Matrix<*,*,*,*,*,*>{ + children + ( + #( + [internals]: [$c,!], + rows: $c.RowsAtCompileTime, + cols: $c.ColsAtCompileTime, + ; Check for RowMajorBit + #if ($c.Flags & 0x1) ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data.array)[($i % $c.RowsAtCompileTime)*$c.ColsAtCompileTime + (($i- $i % $c.RowsAtCompileTime)/$c.RowsAtCompileTime)], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime + ) + ) #else ( + #array( + rank: 2, + base: 0, + expr: ($c.m_storage.m_data.array)[$i], + size: ($r==1)*$c.RowsAtCompileTime+($r==0)*$c.ColsAtCompileTime + ) + ) + ) + ) + + preview + ( + #( + "[", + $c.RowsAtCompileTime, + ",", + $c.ColsAtCompileTime, + "](", + #array( + expr : [($c.m_storage.m_data.array)[$i],g], + size : $c.RowsAtCompileTime*$c.ColsAtCompileTime + ), + ")" + ) + ) +} diff --git a/third_party/eigen/doc/CMakeLists.txt b/third_party/eigen/doc/CMakeLists.txt index 179824dd149..8ff7559885e 100644 --- a/third_party/eigen/doc/CMakeLists.txt +++ b/third_party/eigen/doc/CMakeLists.txt @@ -11,7 +11,7 @@ if(CMAKE_COMPILER_IS_GNUCXX) endif(CMAKE_COMPILER_IS_GNUCXX) option(EIGEN_INTERNAL_DOCUMENTATION "Build internal documentation" OFF) -option(EIGEN_DOC_USE_MATHJAX "Use MathJax for rendering math in HTML docs" ON) + # Set some Doxygen flags set(EIGEN_DOXY_PROJECT_NAME "Eigen") @@ -19,19 +19,12 @@ set(EIGEN_DOXY_OUTPUT_DIRECTORY_SUFFIX "") set(EIGEN_DOXY_INPUT "\"${Eigen_SOURCE_DIR}/Eigen\" \"${Eigen_SOURCE_DIR}/doc\"") set(EIGEN_DOXY_HTML_COLORSTYLE_HUE "220") set(EIGEN_DOXY_TAGFILES "") - if(EIGEN_INTERNAL_DOCUMENTATION) set(EIGEN_DOXY_INTERNAL "YES") else(EIGEN_INTERNAL_DOCUMENTATION) set(EIGEN_DOXY_INTERNAL "NO") endif(EIGEN_INTERNAL_DOCUMENTATION) -if (EIGEN_DOC_USE_MATHJAX) - set(EIGEN_DOXY_USE_MATHJAX "YES") -else () - set(EIGEN_DOXY_USE_MATHJAX "NO") -endif() - configure_file( ${CMAKE_CURRENT_SOURCE_DIR}/Doxyfile.in ${CMAKE_CURRENT_BINARY_DIR}/Doxyfile diff --git a/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox b/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox index 24e5f563b8c..1ee78cbe5f8 100644 --- a/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox +++ b/third_party/eigen/doc/CustomizingEigen_CustomScalar.dox @@ -75,7 +75,7 @@ namespace Eigen { static inline Real epsilon() { return 0; } static inline Real dummy_precision() { return 0; } - static inline int digits10() { return 0; } + static inline Real digits10() { return 0; } enum { IsInteger = 0, diff --git a/third_party/eigen/doc/Doxyfile.in b/third_party/eigen/doc/Doxyfile.in index ac6eafcf927..37948a6122d 100644 --- a/third_party/eigen/doc/Doxyfile.in +++ b/third_party/eigen/doc/Doxyfile.in @@ -736,14 +736,6 @@ EXCLUDE = "${Eigen_SOURCE_DIR}/Eigen/src/Core/products" \ "${Eigen_SOURCE_DIR}/unsupported/doc/examples" \ "${Eigen_SOURCE_DIR}/unsupported/doc/snippets" -# Forward declarations of class templates cause the title of the main page for -# the class template to not contain the template signature. This only happens -# when the \class command is used to document the class. Possibly caused -# by https://github.com/doxygen/doxygen/issues/7698. Confirmed fixed by -# doxygen release 1.8.19. - -EXCLUDE += "${Eigen_SOURCE_DIR}/Eigen/src/Core/util/ForwardDeclarations.h" - # The EXCLUDE_SYMLINKS tag can be used to select whether or not files or # directories that are symbolic links (a Unix file system feature) are excluded # from the input. @@ -1253,7 +1245,7 @@ FORMULA_TRANSPARENT = YES # output. When enabled you may also need to install MathJax separately and # configure the path to it using the MATHJAX_RELPATH option. -USE_MATHJAX = @EIGEN_DOXY_USE_MATHJAX@ +USE_MATHJAX = NO # When MathJax is enabled you need to specify the location relative to the # HTML output directory using the MATHJAX_RELPATH option. The destination @@ -1265,12 +1257,12 @@ USE_MATHJAX = @EIGEN_DOXY_USE_MATHJAX@ # However, it is strongly recommended to install a local # copy of MathJax from http://www.mathjax.org before deployment. -MATHJAX_RELPATH = https://cdn.mathjax.org/mathjax/latest +MATHJAX_RELPATH = http://cdn.mathjax.org/mathjax/latest # The MATHJAX_EXTENSIONS tag can be used to specify one or MathJax extension # names that should be enabled during MathJax rendering. -MATHJAX_EXTENSIONS = TeX/AMSmath TeX/AMSsymbols +MATHJAX_EXTENSIONS = # When the SEARCHENGINE tag is enabled doxygen will generate a search box # for the HTML output. The underlying search engine uses javascript @@ -1617,9 +1609,6 @@ PREDEFINED = EIGEN_EMPTY_STRUCT \ EXPAND_AS_DEFINED = EIGEN_MAKE_TYPEDEFS \ EIGEN_MAKE_FIXED_TYPEDEFS \ EIGEN_MAKE_TYPEDEFS_ALL_SIZES \ - EIGEN_MAKE_ARRAY_TYPEDEFS \ - EIGEN_MAKE_ARRAY_FIXED_TYPEDEFS \ - EIGEN_MAKE_ARRAY_TYPEDEFS_ALL_SIZES \ EIGEN_CWISE_UNOP_RETURN_TYPE \ EIGEN_CWISE_BINOP_RETURN_TYPE \ EIGEN_CURRENT_STORAGE_BASE_CLASS \ diff --git a/third_party/eigen/doc/Pitfalls.dox b/third_party/eigen/doc/Pitfalls.dox index fda40257207..3f395053d09 100644 --- a/third_party/eigen/doc/Pitfalls.dox +++ b/third_party/eigen/doc/Pitfalls.dox @@ -7,30 +7,14 @@ namespace Eigen { See this \link TopicTemplateKeyword page \endlink. - \section TopicPitfalls_aliasing Aliasing Don't miss this \link TopicAliasing page \endlink on aliasing, especially if you got wrong results in statements where the destination appears on the right hand side of the expression. - -\section TopicPitfalls_alignment_issue Alignment Issues (runtime assertion) - -%Eigen does explicit vectorization, and while that is appreciated by many users, that also leads to some issues in special situations where data alignment is compromised. -Indeed, since C++17, C++ does not have quite good enough support for explicit data alignment. -In that case your program hits an assertion failure (that is, a "controlled crash") with a message that tells you to consult this page: -\code -http://eigen.tuxfamily.org/dox/group__TopicUnalignedArrayAssert.html -\endcode -Have a look at \link TopicUnalignedArrayAssert it \endlink and see for yourself if that's something that you can cope with. -It contains detailed information about how to deal with each known cause for that issue. - -Now what if you don't care about vectorization and so don't want to be annoyed with these alignment issues? Then read \link getrid how to get rid of them \endlink. - - \section TopicPitfalls_auto_keyword C++11 and the auto keyword -In short: do not use the auto keywords with %Eigen's expressions, unless you are 100% sure about what you are doing. In particular, do not use the auto keyword as a replacement for a \c Matrix<> type. Here is an example: +In short: do not use the auto keywords with Eigen's expressions, unless you are 100% sure about what you are doing. In particular, do not use the auto keyword as a replacement for a Matrix<> type. Here is an example: \code MatrixXd A, B; @@ -38,81 +22,23 @@ auto C = A*B; for(...) { ... w = C * v; ...} \endcode -In this example, the type of C is not a \c MatrixXd but an abstract expression representing a matrix product and storing references to \c A and \c B. -Therefore, the product of \c A*B will be carried out multiple times, once per iteration of the for loop. -Moreover, if the coefficients of A or B change during the iteration, then C will evaluate to different values. +In this example, the type of C is not a MatrixXd but an abstract expression representing a matrix product and storing references to A and B. Therefore, the product of A*B will be carried out multiple times, once per iteration of the for loop. Moreover, if the coefficients of A or B change during the iteration, then C will evaluate to different values. Here is another example leading to a segfault: \code auto C = ((A+B).eval()).transpose(); // do something with C \endcode -The problem is that \c eval() returns a temporary object (in this case a \c MatrixXd) which is then referenced by the \c Transpose<> expression. -However, this temporary is deleted right after the first line, and then the \c C expression references a dead object. -One possible fix consists in applying \c eval() on the whole expression: -\code -auto C = (A+B).transpose().eval(); -\endcode - -The same issue might occur when sub expressions are automatically evaluated by %Eigen as in the following example: +The problem is that eval() returns a temporary object (in this case a MatrixXd) which is then referenced by the Transpose<> expression. However, this temporary is deleted right after the first line, and there the C expression reference a dead object. The same issue might occur when sub expressions are automatically evaluated by Eigen as in the following example: \code VectorXd u, v; auto C = u + (A*v).normalized(); // do something with C \endcode -Here the \c normalized() method has to evaluate the expensive product \c A*v to avoid evaluating it twice. -Again, one possible fix is to call \c .eval() on the whole expression: +where the normalized() method has to evaluate the expensive product A*v to avoid evaluating it twice. On the other hand, the following example is perfectly fine: \code auto C = (u + (A*v).normalized()).eval(); \endcode -In this case, \c C will be a regular \c VectorXd object. -Note that DenseBase::eval() is smart enough to avoid copies when the underlying expression is already a plain \c Matrix<>. - - -\section TopicPitfalls_header_issues Header Issues (failure to compile) - -With all libraries, one must check the documentation for which header to include. -The same is true with %Eigen, but slightly worse: with %Eigen, a method in a class may require an additional #include over what the class itself requires! -For example, if you want to use the \c cross() method on a vector (it computes a cross-product) then you need to: -\code -#include -\endcode -We try to always document this, but do tell us if we forgot an occurrence. - - -\section TopicPitfalls_ternary_operator Ternary operator - -In short: avoid the use of the ternary operator (COND ? THEN : ELSE) with %Eigen's expressions for the \c THEN and \c ELSE statements. -To see why, let's consider the following example: -\code -Vector3f A; -A << 1, 2, 3; -Vector3f B = ((1 < 0) ? (A.reverse()) : A); -\endcode -This example will return B = 3, 2, 1. Do you see why? -The reason is that in c++ the type of the \c ELSE statement is inferred from the type of the \c THEN expression such that both match. -Since \c THEN is a Reverse, the \c ELSE statement A is converted to a Reverse, and the compiler thus generates: -\code -Vector3f B = ((1 < 0) ? (A.reverse()) : Reverse(A)); -\endcode -In this very particular case, a workaround would be to call A.reverse().eval() for the \c THEN statement, but the safest and fastest is really to avoid this ternary operator with %Eigen's expressions and use a if/else construct. - - -\section TopicPitfalls_pass_by_value Pass-by-value - -If you don't know why passing-by-value is wrong with %Eigen, read this \link TopicPassingByValue page \endlink first. - -While you may be extremely careful and use care to make sure that all of your code that explicitly uses %Eigen types is pass-by-reference you have to watch out for templates which define the argument types at compile time. - -If a template has a function that takes arguments pass-by-value, and the relevant template parameter ends up being an %Eigen type, then you will of course have the same alignment problems that you would in an explicitly defined function passing %Eigen types by reference. - -Using %Eigen types with other third party libraries or even the STL can present the same problem. -boost::bind for example uses pass-by-value to store arguments in the returned functor. -This will of course be a problem. - -There are at least two ways around this: - - If the value you are passing is guaranteed to be around for the life of the functor, you can use boost::ref() to wrap the value as you pass it to boost::bind. Generally this is not a solution for values on the stack as if the functor ever gets passed to a lower or independent scope, the object may be gone by the time it's attempted to be used. - - The other option is to make your functions take a reference counted pointer like boost::shared_ptr as the argument. This avoids needing to worry about managing the lifetime of the object being passed. - +In this case, C will be a regular VectorXd object. */ } diff --git a/third_party/eigen/doc/SparseQuickReference.dox b/third_party/eigen/doc/SparseQuickReference.dox index 653bf33ef84..a25622e800e 100644 --- a/third_party/eigen/doc/SparseQuickReference.dox +++ b/third_party/eigen/doc/SparseQuickReference.dox @@ -244,7 +244,7 @@ As stated earlier, for a read-write sub-matrix (RW), the evaluation can be done \code sm1.valuePtr(); // Pointer to the values -sm1.innerIndexPtr(); // Pointer to the indices. +sm1.innerIndextr(); // Pointer to the indices. sm1.outerIndexPtr(); // Pointer to the beginning of each inner vector \endcode diff --git a/third_party/eigen/doc/TopicLazyEvaluation.dox b/third_party/eigen/doc/TopicLazyEvaluation.dox index d2a704f132a..101ef8c72c2 100644 --- a/third_party/eigen/doc/TopicLazyEvaluation.dox +++ b/third_party/eigen/doc/TopicLazyEvaluation.dox @@ -2,95 +2,63 @@ namespace Eigen { /** \page TopicLazyEvaluation Lazy Evaluation and Aliasing -Executive summary: %Eigen has intelligent compile-time mechanisms to enable lazy evaluation and removing temporaries where appropriate. +Executive summary: Eigen has intelligent compile-time mechanisms to enable lazy evaluation and removing temporaries where appropriate. It will handle aliasing automatically in most cases, for example with matrix products. The automatic behavior can be overridden manually by using the MatrixBase::eval() and MatrixBase::noalias() methods. When you write a line of code involving a complex expression such as -\code mat1 = mat2 + mat3 * (mat4 + mat5); -\endcode +\code mat1 = mat2 + mat3 * (mat4 + mat5); \endcode -%Eigen determines automatically, for each sub-expression, whether to evaluate it into a temporary variable. Indeed, in certain cases it is better to evaluate a sub-expression into a temporary variable, while in other cases it is better to avoid that. +Eigen determines automatically, for each sub-expression, whether to evaluate it into a temporary variable. Indeed, in certain cases it is better to evaluate immediately a sub-expression into a temporary variable, while in other cases it is better to avoid that. A traditional math library without expression templates always evaluates all sub-expressions into temporaries. So with this code, -\code vec1 = vec2 + vec3; -\endcode +\code vec1 = vec2 + vec3; \endcode a traditional library would evaluate \c vec2 + vec3 into a temporary \c vec4 and then copy \c vec4 into \c vec1. This is of course inefficient: the arrays are traversed twice, so there are a lot of useless load/store operations. -Expression-templates-based libraries can avoid evaluating sub-expressions into temporaries, which in many cases results in large speed improvements. -This is called lazy evaluation as an expression is getting evaluated as late as possible. -In %Eigen all expressions are lazy-evaluated. -More precisely, an expression starts to be evaluated once it is assigned to a matrix. -Until then nothing happens beyond constructing the abstract expression tree. -In contrast to most other expression-templates-based libraries, however, %Eigen might choose to evaluate some sub-expressions into temporaries. -There are two reasons for that: first, pure lazy evaluation is not always a good choice for performance; second, pure lazy evaluation can be very dangerous, for example with matrix products: doing mat = mat*mat gives a wrong result if the matrix product is directly evaluated within the destination matrix, because of the way matrix product works. +Expression-templates-based libraries can avoid evaluating sub-expressions into temporaries, which in many cases results in large speed improvements. This is called lazy evaluation as an expression is getting evaluated as late as possible, instead of immediately. However, most other expression-templates-based libraries always choose lazy evaluation. There are two problems with that: first, lazy evaluation is not always a good choice for performance; second, lazy evaluation can be very dangerous, for example with matrix products: doing matrix = matrix*matrix gives a wrong result if the matrix product is lazy-evaluated, because of the way matrix product works. -For these reasons, %Eigen has intelligent compile-time mechanisms to determine automatically which sub-expression should be evaluated into a temporary variable. +For these reasons, Eigen has intelligent compile-time mechanisms to determine automatically when to use lazy evaluation, and when on the contrary it should evaluate immediately into a temporary variable. So in the basic example, -\code mat1 = mat2 + mat3; -\endcode +\code matrix1 = matrix2 + matrix3; \endcode -%Eigen chooses not to introduce any temporary. Thus the arrays are traversed only once, producing optimized code. -If you really want to force immediate evaluation, use \link MatrixBase::eval() eval()\endlink: +Eigen chooses lazy evaluation. Thus the arrays are traversed only once, producing optimized code. If you really want to force immediate evaluation, use \link MatrixBase::eval() eval()\endlink: -\code mat1 = (mat2 + mat3).eval(); -\endcode +\code matrix1 = (matrix2 + matrix3).eval(); \endcode Here is now a more involved example: -\code mat1 = -mat2 + mat3 + 5 * mat4; -\endcode +\code matrix1 = -matrix2 + matrix3 + 5 * matrix4; \endcode -Here again %Eigen won't introduce any temporary, thus producing a single fused evaluation loop, which is clearly the correct choice. +Eigen chooses lazy evaluation at every stage in that example, which is clearly the correct choice. In fact, lazy evaluation is the "default choice" and Eigen will choose it except in a few circumstances. -\section TopicLazyEvaluationWhichExpr Which sub-expressions are evaluated into temporaries? +The first circumstance in which Eigen chooses immediate evaluation, is when it sees an assignment a = b; and the expression \c b has the evaluate-before-assigning \link flags flag\endlink. The most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do -The default evaluation strategy is to fuse the operations in a single loop, and %Eigen will choose it except in a few circumstances. +\code matrix = matrix * matrix; \endcode -The first circumstance in which %Eigen chooses to evaluate a sub-expression is when it sees an assignment a = b; and the expression \c b has the evaluate-before-assigning \link flags flag\endlink. -The most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do - -\code mat = mat * mat; -\endcode - -%Eigen will evaluate mat * mat into a temporary matrix, and then copies it into the original \c mat. -This guarantees a correct result as we saw above that lazy evaluation gives wrong results with matrix products. -It also doesn't cost much, as the cost of the matrix product itself is much higher. -Note that this temporary is introduced at evaluation time only, that is, within operator= in this example. -The expression mat * mat still return a abstract product type. +Eigen first evaluates matrix * matrix into a temporary matrix, and then copies it into the original \c matrix. This guarantees a correct result as we saw above that lazy evaluation gives wrong results with matrix products. It also doesn't cost much, as the cost of the matrix product itself is much higher. What if you know that the result does no alias the operand of the product and want to force lazy evaluation? Then use \link MatrixBase::noalias() .noalias()\endlink instead. Here is an example: -\code mat1.noalias() = mat2 * mat2; -\endcode +\code matrix1.noalias() = matrix2 * matrix2; \endcode -Here, since we know that mat2 is not the same matrix as mat1, we know that lazy evaluation is not dangerous, so we may force lazy evaluation. Concretely, the effect of noalias() here is to bypass the evaluate-before-assigning \link flags flag\endlink. +Here, since we know that matrix2 is not the same matrix as matrix1, we know that lazy evaluation is not dangerous, so we may force lazy evaluation. Concretely, the effect of noalias() here is to bypass the evaluate-before-assigning \link flags flag\endlink. -The second circumstance in which %Eigen chooses to evaluate a sub-expression, is when it sees a nested expression such as a + b where \c b is already an expression having the evaluate-before-nesting \link flags flag\endlink. -Again, the most important example of such an expression is the \link Product matrix product expression\endlink. -For example, when you do +The second circumstance in which Eigen chooses immediate evaluation, is when it sees a nested expression such as a + b where \c b is already an expression having the evaluate-before-nesting \link flags flag\endlink. Again, the most important example of such an expression is the \link Product matrix product expression\endlink. For example, when you do -\code mat1 = mat2 * mat3 + mat4 * mat5; -\endcode +\code matrix1 = matrix2 + matrix3 * matrix4; \endcode -the products mat2 * mat3 and mat4 * mat5 gets evaluated separately into temporary matrices before being summed up in mat1. -Indeed, to be efficient matrix products need to be evaluated within a destination matrix at hand, and not as simple "dot products". -For small matrices, however, you might want to enforce a "dot-product" based lazy evaluation with lazyProduct(). -Again, it is important to understand that those temporaries are created at evaluation time only, that is in operator =. -See TopicPitfalls_auto_keyword for common pitfalls regarding this remark. +the product matrix3 * matrix4 gets evaluated immediately into a temporary matrix. Indeed, experiments showed that it is often beneficial for performance to evaluate immediately matrix products when they are nested into bigger expressions. -The third circumstance in which %Eigen chooses to evaluate a sub-expression, is when its cost model shows that the total cost of an operation is reduced if a sub-expression gets evaluated into a temporary. -Indeed, in certain cases, an intermediate result is sufficiently costly to compute and is reused sufficiently many times, that is worth "caching". Here is an example: +The third circumstance in which Eigen chooses immediate evaluation, is when its cost model shows that the total cost of an operation is reduced if a sub-expression gets evaluated into a temporary. Indeed, in certain cases, an intermediate result is sufficiently costly to compute and is reused sufficiently many times, that is worth "caching". Here is an example: -\code mat1 = mat2 * (mat3 + mat4); -\endcode +\code matrix1 = matrix2 * (matrix3 + matrix4); \endcode -Here, provided the matrices have at least 2 rows and 2 columns, each coefficient of the expression mat3 + mat4 is going to be used several times in the matrix product. Instead of computing the sum every time, it is much better to compute it once and store it in a temporary variable. %Eigen understands this and evaluates mat3 + mat4 into a temporary variable before evaluating the product. +Here, provided the matrices have at least 2 rows and 2 columns, each coefficienct of the expression matrix3 + matrix4 is going to be used several times in the matrix product. Instead of computing the sum everytime, it is much better to compute it once and store it in a temporary variable. Eigen understands this and evaluates matrix3 + matrix4 into a temporary variable before evaluating the product. */ diff --git a/third_party/eigen/doc/TopicMultithreading.dox b/third_party/eigen/doc/TopicMultithreading.dox index a2855745b60..47c9b261f89 100644 --- a/third_party/eigen/doc/TopicMultithreading.dox +++ b/third_party/eigen/doc/TopicMultithreading.dox @@ -49,7 +49,6 @@ int main(int argc, char** argv) In the case your application is parallelized with OpenMP, you might want to disable Eigen's own parallization as detailed in the previous section. -\warning Using OpenMP with custom scalar types that might throw exceptions can lead to unexpected behaviour in the event of throwing. */ } diff --git a/third_party/eigen/doc/TutorialGeometry.dox b/third_party/eigen/doc/TutorialGeometry.dox index 723f4dbcee9..2e1420f98c8 100644 --- a/third_party/eigen/doc/TutorialGeometry.dox +++ b/third_party/eigen/doc/TutorialGeometry.dox @@ -232,8 +232,8 @@ On the other hand, since there exist 24 different conventions, they are pretty c to create a rotation matrix according to the 2-1-2 convention.\code Matrix3f m; m = AngleAxisf(angle1, Vector3f::UnitZ()) - * * AngleAxisf(angle2, Vector3f::UnitY()) - * * AngleAxisf(angle3, Vector3f::UnitZ()); + * AngleAxisf(angle2, Vector3f::UnitY()) + * AngleAxisf(angle3, Vector3f::UnitZ()); \endcode diff --git a/third_party/eigen/doc/eigen_navtree_hacks.js b/third_party/eigen/doc/eigen_navtree_hacks.js index afb97edf5c7..a6f8c342821 100644 --- a/third_party/eigen/doc/eigen_navtree_hacks.js +++ b/third_party/eigen/doc/eigen_navtree_hacks.js @@ -5,7 +5,6 @@ function generate_autotoc() { if(headers.length > 1) { var toc = $("#side-nav").append(''); toc = $("#nav-toc"); - var footer = $("#nav-path"); var footerHeight = footer.height(); toc = toc.append('
      '); toc = toc.find('ul'); @@ -138,7 +137,7 @@ function initNavTree(toroot,relpath) } }) - $(window).on("load", showRoot); + $(window).load(showRoot); } // return false if the the node has no children at all, or has only section/subsection children @@ -242,6 +241,6 @@ $(document).ready(function() { } })(); - $(window).on("load", resizeHeight); + $(window).load(resizeHeight); }); diff --git a/third_party/eigen/doc/eigendoxy_footer.html.in b/third_party/eigen/doc/eigendoxy_footer.html.in index 126653589d5..9ac0596cb0a 100644 --- a/third_party/eigen/doc/eigendoxy_footer.html.in +++ b/third_party/eigen/doc/eigendoxy_footer.html.in @@ -17,6 +17,19 @@ $generatedby   + + + + diff --git a/third_party/eigen/doc/eigendoxy_header.html.in b/third_party/eigen/doc/eigendoxy_header.html.in index a6b1c1d081a..bb149f8f0f1 100644 --- a/third_party/eigen/doc/eigendoxy_header.html.in +++ b/third_party/eigen/doc/eigendoxy_header.html.in @@ -20,9 +20,6 @@ $mathjax - - -
      diff --git a/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp b/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp index 0b87313a1cf..76f49f2fbc6 100644 --- a/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp +++ b/third_party/eigen/doc/examples/Tutorial_BlockOperations_block_assignment.cpp @@ -14,5 +14,5 @@ int main() a.block<2,2>(1,1) = m; cout << "Here is now a with m copied into its central 2x2 block:" << endl << a << endl << endl; a.block(0,0,2,3) = a.block(2,1,2,3); - cout << "Here is now a with bottom-right 2x3 block copied into top-left 2x3 block:" << endl << a << endl << endl; + cout << "Here is now a with bottom-right 2x3 block copied into top-left 2x2 block:" << endl << a << endl << endl; } diff --git a/third_party/eigen/lapack/CMakeLists.txt b/third_party/eigen/lapack/CMakeLists.txt index fbecd662498..6df1fa9584b 100644 --- a/third_party/eigen/lapack/CMakeLists.txt +++ b/third_party/eigen/lapack/CMakeLists.txt @@ -133,14 +133,12 @@ if(EXISTS ${eigen_full_path_to_testing_lapack}) string(REGEX REPLACE "(.*)/STACK:(.*) (.*)" "\\1/STACK:900000000000000000 \\3" CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}") endif() - file(MAKE_DIRECTORY "${LAPACK_BINARY_DIR}/TESTING") add_subdirectory(testing/MATGEN) add_subdirectory(testing/LIN) add_subdirectory(testing/EIG) - cmake_policy(SET CMP0026 OLD) macro(add_lapack_test output input target) set(TEST_INPUT "${LAPACK_SOURCE_DIR}/testing/${input}") - set(TEST_OUTPUT "${LAPACK_BINARY_DIR}/TESTING/${output}") + set(TEST_OUTPUT "${LAPACK_BINARY_DIR}/testing/${output}") get_target_property(TEST_LOC ${target} LOCATION) string(REPLACE "." "_" input_name ${input}) set(testName "${target}_${input_name}") diff --git a/third_party/eigen/scripts/CMakeLists.txt b/third_party/eigen/scripts/CMakeLists.txt index 0d9a631a9dd..5f08da5b9f9 100644 --- a/third_party/eigen/scripts/CMakeLists.txt +++ b/third_party/eigen/scripts/CMakeLists.txt @@ -1,6 +1,6 @@ get_property(EIGEN_TESTS_LIST GLOBAL PROPERTY EIGEN_TESTS_LIST) -configure_file(buildtests.in ${CMAKE_BINARY_DIR}/buildtests.sh @ONLY) +configure_file(buildtests.in ${CMAKE_CURRENT_BINARY_DIR}/buildtests.sh @ONLY) -configure_file(check.in ${CMAKE_BINARY_DIR}/check.sh COPYONLY) -configure_file(debug.in ${CMAKE_BINARY_DIR}/debug.sh COPYONLY) -configure_file(release.in ${CMAKE_BINARY_DIR}/release.sh COPYONLY) +configure_file(check.in ${CMAKE_CURRENT_BINARY_DIR}/check.sh COPYONLY) +configure_file(debug.in ${CMAKE_CURRENT_BINARY_DIR}/debug.sh COPYONLY) +configure_file(release.in ${CMAKE_CURRENT_BINARY_DIR}/release.sh COPYONLY) diff --git a/third_party/eigen/test/CMakeLists.txt b/third_party/eigen/test/CMakeLists.txt index 47e6fee4b27..0747aa6cb65 100644 --- a/third_party/eigen/test/CMakeLists.txt +++ b/third_party/eigen/test/CMakeLists.txt @@ -163,7 +163,7 @@ ei_add_test(constructor) ei_add_test(linearstructure) ei_add_test(integer_types) ei_add_test(unalignedcount) -if(NOT EIGEN_TEST_NO_EXCEPTIONS AND NOT EIGEN_TEST_OPENMP) +if(NOT EIGEN_TEST_NO_EXCEPTIONS) ei_add_test(exceptions) endif() ei_add_test(redux) @@ -185,7 +185,7 @@ ei_add_test(smallvectors) ei_add_test(mapped_matrix) ei_add_test(mapstride) ei_add_test(mapstaticmethods) -ei_add_test(array_cwise) +ei_add_test(array) ei_add_test(array_for_matrix) ei_add_test(array_replicate) ei_add_test(array_reverse) diff --git a/third_party/eigen/test/bdcsvd.cpp b/third_party/eigen/test/bdcsvd.cpp index 3ca273635c4..6c7b09696cf 100644 --- a/third_party/eigen/test/bdcsvd.cpp +++ b/third_party/eigen/test/bdcsvd.cpp @@ -28,13 +28,9 @@ template void bdcsvd(const MatrixType& a = MatrixType(), bool pickrandom = true) { - MatrixType m; - if(pickrandom) { - m.resizeLike(a); + MatrixType m = a; + if(pickrandom) svd_fill_random(m); - } - else - m = a; CALL_SUBTEST(( svd_test_all_computation_options >(m, false) )); } diff --git a/third_party/eigen/test/constructor.cpp b/third_party/eigen/test/constructor.cpp index 9885399510a..eec9e219296 100644 --- a/third_party/eigen/test/constructor.cpp +++ b/third_party/eigen/test/constructor.cpp @@ -20,8 +20,6 @@ template struct Wrapper inline operator MatrixType& () { return m_mat; } }; -enum my_sizes { M = 12, N = 7}; - template void ctor_init1(const MatrixType& m) { // Check logic in PlainObjectBase::_init1 @@ -83,16 +81,4 @@ void test_constructor() Array a(123); VERIFY_IS_EQUAL(a(4), 123.f); } - { - MatrixXi m1(M,N); - VERIFY_IS_EQUAL(m1.rows(),M); - VERIFY_IS_EQUAL(m1.cols(),N); - ArrayXXi a1(M,N); - VERIFY_IS_EQUAL(a1.rows(),M); - VERIFY_IS_EQUAL(a1.cols(),N); - VectorXi v1(M); - VERIFY_IS_EQUAL(v1.size(),M); - ArrayXi a2(M); - VERIFY_IS_EQUAL(a2.size(),M); - } } diff --git a/third_party/eigen/test/ctorleak.cpp b/third_party/eigen/test/ctorleak.cpp index d73fecfe2c7..c158f5e4ee4 100644 --- a/third_party/eigen/test/ctorleak.cpp +++ b/third_party/eigen/test/ctorleak.cpp @@ -8,7 +8,7 @@ struct Foo static Index object_limit; int dummy; - Foo() : dummy(0) + Foo() { #ifdef EIGEN_EXCEPTIONS // TODO: Is this the correct way to handle this? @@ -37,33 +37,22 @@ void test_ctorleak() { typedef Matrix MatrixX; typedef Matrix VectorX; - Foo::object_count = 0; for(int i = 0; i < g_repeat; i++) { Index rows = internal::random(2,EIGEN_TEST_MAX_SIZE), cols = internal::random(2,EIGEN_TEST_MAX_SIZE); - Foo::object_limit = rows*cols; - { - MatrixX r(rows, cols); - Foo::object_limit = r.size()+internal::random(0, rows*cols - 2); + Foo::object_limit = internal::random(0, rows*cols - 2); std::cout << "object_limit =" << Foo::object_limit << std::endl; #ifdef EIGEN_EXCEPTIONS try { #endif - if(internal::random()) { - std::cout << "\nMatrixX m(" << rows << ", " << cols << ");\n"; - MatrixX m(rows, cols); - } - else { - std::cout << "\nMatrixX m(r);\n"; - MatrixX m(r); - } + std::cout << "\nMatrixX m(" << rows << ", " << cols << ");\n"; + MatrixX m(rows, cols); #ifdef EIGEN_EXCEPTIONS VERIFY(false); // not reached if exceptions are enabled } catch (const Foo::Fail&) { /* ignore */ } #endif - } VERIFY_IS_EQUAL(Index(0), Foo::object_count); { @@ -77,5 +66,4 @@ void test_ctorleak() } VERIFY_IS_EQUAL(Index(0), Foo::object_count); } - std::cout << "\n"; } diff --git a/third_party/eigen/test/eigensolver_generic.cpp b/third_party/eigen/test/eigensolver_generic.cpp index 5c13175690b..07bf65e0327 100644 --- a/third_party/eigen/test/eigensolver_generic.cpp +++ b/third_party/eigen/test/eigensolver_generic.cpp @@ -67,7 +67,7 @@ template void eigensolver(const MatrixType& m) // Test matrix with NaN a(0,0) = std::numeric_limits::quiet_NaN(); EigenSolver eiNaN(a); - VERIFY_IS_NOT_EQUAL(eiNaN.info(), Success); + VERIFY_IS_EQUAL(eiNaN.info(), NoConvergence); } // regression test for bug 1098 diff --git a/third_party/eigen/test/exceptions.cpp b/third_party/eigen/test/exceptions.cpp index 015b9fd33f3..b83fb82ba60 100644 --- a/third_party/eigen/test/exceptions.cpp +++ b/third_party/eigen/test/exceptions.cpp @@ -109,7 +109,5 @@ void memoryleak() void test_exceptions() { - EIGEN_TRY { - CALL_SUBTEST( memoryleak() ); - } EIGEN_CATCH(...) {} + CALL_SUBTEST( memoryleak() ); } diff --git a/third_party/eigen/test/fastmath.cpp b/third_party/eigen/test/fastmath.cpp index e84bdc97232..cc5db074637 100644 --- a/third_party/eigen/test/fastmath.cpp +++ b/third_party/eigen/test/fastmath.cpp @@ -43,11 +43,11 @@ void check_inf_nan(bool dryrun) { } else { - if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !(numext::isfinite)(m(3)) ); g_test_level=0; - if( (std::isinf) (m(3))) g_test_level=1; VERIFY( !(numext::isinf)(m(3)) ); g_test_level=0; - if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( (numext::isnan)(m(3)) ); g_test_level=0; - if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; - if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( m.hasNaN() ); g_test_level=0; + VERIFY( !(numext::isfinite)(m(3)) ); + VERIFY( !(numext::isinf)(m(3)) ); + VERIFY( (numext::isnan)(m(3)) ); + VERIFY( !m.allFinite() ); + VERIFY( m.hasNaN() ); } T hidden_zero = (std::numeric_limits::min)()*(std::numeric_limits::min)(); m(4) /= hidden_zero; @@ -62,29 +62,29 @@ void check_inf_nan(bool dryrun) { } else { - if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !(numext::isfinite)(m(4)) ); g_test_level=0; - if(!(std::isinf) (m(3))) g_test_level=1; VERIFY( (numext::isinf)(m(4)) ); g_test_level=0; - if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !(numext::isnan)(m(4)) ); g_test_level=0; - if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; - if(!(std::isnan) (m(3))) g_test_level=1; VERIFY( m.hasNaN() ); g_test_level=0; + VERIFY( !(numext::isfinite)(m(4)) ); + VERIFY( (numext::isinf)(m(4)) ); + VERIFY( !(numext::isnan)(m(4)) ); + VERIFY( !m.allFinite() ); + VERIFY( m.hasNaN() ); } m(3) = 0; if(dryrun) { std::cout << "std::isfinite(" << m(3) << ") = "; check((std::isfinite)(m(3)),true); std::cout << " ; numext::isfinite = "; check((numext::isfinite)(m(3)), true); std::cout << "\n"; - std::cout << "std::isinf(" << m(3) << ") = "; check((std::isinf)(m(3)),false); std::cout << " ; numext::isinf = "; check((numext::isinf)(m(3)), false); std::cout << "\n"; - std::cout << "std::isnan(" << m(3) << ") = "; check((std::isnan)(m(3)),false); std::cout << " ; numext::isnan = "; check((numext::isnan)(m(3)), false); std::cout << "\n"; + std::cout << "std::isinf(" << m(3) << ") = "; check((std::isinf)(m(3)),false); std::cout << " ; numext::isinf = "; check((numext::isinf)(m(3)), false); std::cout << "\n"; + std::cout << "std::isnan(" << m(3) << ") = "; check((std::isnan)(m(3)),false); std::cout << " ; numext::isnan = "; check((numext::isnan)(m(3)), false); std::cout << "\n"; std::cout << "allFinite: "; check(m.allFinite(), 0); std::cout << "\n"; std::cout << "hasNaN: "; check(m.hasNaN(), 0); std::cout << "\n"; std::cout << "\n\n"; } else { - if(!(std::isfinite)(m(3))) g_test_level=1; VERIFY( (numext::isfinite)(m(3)) ); g_test_level=0; - if( (std::isinf) (m(3))) g_test_level=1; VERIFY( !(numext::isinf)(m(3)) ); g_test_level=0; - if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !(numext::isnan)(m(3)) ); g_test_level=0; - if( (std::isfinite)(m(3))) g_test_level=1; VERIFY( !m.allFinite() ); g_test_level=0; - if( (std::isnan) (m(3))) g_test_level=1; VERIFY( !m.hasNaN() ); g_test_level=0; + VERIFY( (numext::isfinite)(m(3)) ); + VERIFY( !(numext::isinf)(m(3)) ); + VERIFY( !(numext::isnan)(m(3)) ); + VERIFY( !m.allFinite() ); + VERIFY( !m.hasNaN() ); } } diff --git a/third_party/eigen/test/geo_alignedbox.cpp b/third_party/eigen/test/geo_alignedbox.cpp index 4cf51aafb58..b64ea3bdc47 100644 --- a/third_party/eigen/test/geo_alignedbox.cpp +++ b/third_party/eigen/test/geo_alignedbox.cpp @@ -15,9 +15,8 @@ #include using namespace std; -// TODO not sure if this is actually still necessary anywhere ... template EIGEN_DONT_INLINE -void kill_extra_precision(T& ) { } +void kill_extra_precision(T& x) { eigen_assert((void*)(&x) != (void*)0); } template void alignedbox(const BoxType& _box) diff --git a/third_party/eigen/test/geo_quaternion.cpp b/third_party/eigen/test/geo_quaternion.cpp index 87680f1ccac..8ee8fdb27f1 100644 --- a/third_party/eigen/test/geo_quaternion.cpp +++ b/third_party/eigen/test/geo_quaternion.cpp @@ -244,14 +244,6 @@ template void mapQuaternion(void){ // is used to determine wether we can return a coeff by reference or not, which is not enough for Map. //const MCQuaternionUA& cmcq3(mcq3); //VERIFY( &cmcq3.x() == &mcq3.x() ); - - // test cast - { - Quaternion q1f = mq1.template cast(); - VERIFY_IS_APPROX(q1f.template cast(),mq1); - Quaternion q1d = mq1.template cast(); - VERIFY_IS_APPROX(q1d.template cast(),mq1); - } } template void quaternionAlignment(void){ diff --git a/third_party/eigen/test/geo_transformations.cpp b/third_party/eigen/test/geo_transformations.cpp index 8d064ddc3bc..278e527c252 100755 --- a/third_party/eigen/test/geo_transformations.cpp +++ b/third_party/eigen/test/geo_transformations.cpp @@ -612,62 +612,6 @@ template void transform_products() VERIFY_IS_APPROX((ac*p).matrix(), a_m*p_m); } -template void transformations_no_scale() -{ - /* this test covers the following files: - Cross.h Quaternion.h, Transform.h - */ - typedef Matrix Vector3; - typedef Matrix Vector4; - typedef Quaternion Quaternionx; - typedef AngleAxis AngleAxisx; - typedef Transform Transform3; - typedef Translation Translation3; - typedef Matrix Matrix4; - - Vector3 v0 = Vector3::Random(), - v1 = Vector3::Random(); - - Transform3 t0, t1, t2; - - Scalar a = internal::random(-Scalar(EIGEN_PI), Scalar(EIGEN_PI)); - - Quaternionx q1, q2; - - q1 = AngleAxisx(a, v0.normalized()); - - t0 = Transform3::Identity(); - VERIFY_IS_APPROX(t0.matrix(), Transform3::MatrixType::Identity()); - - t0.setIdentity(); - t1.setIdentity(); - v1 = Vector3::Ones(); - t0.linear() = q1.toRotationMatrix(); - t0.pretranslate(v0); - t1.linear() = q1.conjugate().toRotationMatrix(); - t1.translate(-v0); - - VERIFY((t0 * t1).matrix().isIdentity(test_precision())); - - t1.fromPositionOrientationScale(v0, q1, v1); - VERIFY_IS_APPROX(t1.matrix(), t0.matrix()); - VERIFY_IS_APPROX(t1*v1, t0*v1); - - // translation * vector - t0.setIdentity(); - t0.translate(v0); - VERIFY_IS_APPROX((t0 * v1).template head<3>(), Translation3(v0) * v1); - - // Conversion to matrix. - Transform3 t3; - t3.linear() = q1.toRotationMatrix(); - t3.translation() = v1; - Matrix4 m3 = t3.matrix(); - VERIFY((m3 * m3.inverse()).isIdentity(test_precision())); - // Verify implicit last row is initialized. - VERIFY_IS_APPROX(Vector4(m3.row(3)), Vector4(0.0, 0.0, 0.0, 1.0)); -} - void test_geo_transformations() { for(int i = 0; i < g_repeat; i++) { @@ -681,7 +625,7 @@ void test_geo_transformations() CALL_SUBTEST_3(( transformations() )); CALL_SUBTEST_3(( transformations() )); CALL_SUBTEST_3(( transform_alignment() )); - + CALL_SUBTEST_4(( transformations() )); CALL_SUBTEST_4(( non_projective_only() )); @@ -697,8 +641,5 @@ void test_geo_transformations() CALL_SUBTEST_8(( transform_associativity(Rotation2D(internal::random()*double(EIGEN_PI))) )); CALL_SUBTEST_8(( transform_associativity(Quaterniond::UnitRandom()) )); - - CALL_SUBTEST_9(( transformations_no_scale() )); - CALL_SUBTEST_9(( transformations_no_scale() )); } } diff --git a/third_party/eigen/test/inverse.cpp b/third_party/eigen/test/inverse.cpp index d81af26c1e6..be607cc8bea 100644 --- a/third_party/eigen/test/inverse.cpp +++ b/third_party/eigen/test/inverse.cpp @@ -92,22 +92,6 @@ template void inverse(const MatrixType& m) } } -template -void inverse_zerosized() -{ - Matrix A(0,0); - { - Matrix b, x; - x = A.inverse() * b; - } - { - Matrix b(0,1), x; - x = A.inverse() * b; - VERIFY_IS_EQUAL(x.rows(), 0); - VERIFY_IS_EQUAL(x.cols(), 1); - } -} - void test_inverse() { int s = 0; @@ -121,7 +105,6 @@ void test_inverse() s = internal::random(50,320); CALL_SUBTEST_5( inverse(MatrixXf(s,s)) ); TEST_SET_BUT_UNUSED_VARIABLE(s) - CALL_SUBTEST_5( inverse_zerosized() ); s = internal::random(25,100); CALL_SUBTEST_6( inverse(MatrixXcd(s,s)) ); diff --git a/third_party/eigen/test/main.h b/third_party/eigen/test/main.h index 18bb5c82550..8c868ee79e7 100644 --- a/third_party/eigen/test/main.h +++ b/third_party/eigen/test/main.h @@ -72,11 +72,6 @@ #define isnan(X) please_protect_your_isnan_with_parentheses #define isinf(X) please_protect_your_isinf_with_parentheses #define isfinite(X) please_protect_your_isfinite_with_parentheses - -// test possible conflicts -struct real {}; -struct imag {}; - #ifdef M_PI #undef M_PI #endif diff --git a/third_party/eigen/test/numext.cpp b/third_party/eigen/test/numext.cpp index beba9e911b3..3de33e2f918 100644 --- a/third_party/eigen/test/numext.cpp +++ b/third_party/eigen/test/numext.cpp @@ -12,7 +12,6 @@ template void check_abs() { typedef typename NumTraits::Real Real; - Real zero(0); if(NumTraits::IsSigned) VERIFY_IS_EQUAL(numext::abs(-T(1)), T(1)); @@ -27,9 +26,9 @@ void check_abs() { if(NumTraits::IsSigned) { VERIFY_IS_EQUAL(numext::abs(x), numext::abs(-x)); - VERIFY( numext::abs(-x) >= zero ); + VERIFY( numext::abs(-x) >= Real(0)); } - VERIFY( numext::abs(x) >= zero ); + VERIFY( numext::abs(x) >= Real(0)); VERIFY_IS_APPROX( numext::abs2(x), numext::abs2(numext::abs(x)) ); } } diff --git a/third_party/eigen/test/packetmath.cpp b/third_party/eigen/test/packetmath.cpp index 74ac435cf47..7821a1738df 100644 --- a/third_party/eigen/test/packetmath.cpp +++ b/third_party/eigen/test/packetmath.cpp @@ -16,6 +16,12 @@ #endif // using namespace Eigen; +#ifdef EIGEN_VECTORIZE_SSE +const bool g_vectorize_sse = true; +#else +const bool g_vectorize_sse = false; +#endif + namespace Eigen { namespace internal { template T negate(const T& x) { return -x; } @@ -242,13 +248,12 @@ template void packetmath() VERIFY(isApproxAbs(ref[0], internal::predux(internal::pload(data1)), refvalue) && "internal::predux"); { - int newsize = PacketSize>4?PacketSize/2:PacketSize; - for (int i=0; i(data1))); - VERIFY(areApprox(ref, data2, newsize) && "internal::predux_downto4"); + VERIFY(areApprox(ref, data2, PacketSize>4?PacketSize/2:PacketSize) && "internal::predux_downto4"); } ref[0] = 1; @@ -299,7 +304,7 @@ template void packetmath() } } - if (PacketTraits::HasBlend) { + if (PacketTraits::HasBlend || g_vectorize_sse) { // pinsertfirst for (int i=0; i void packetmath() VERIFY(areApprox(ref, data2, PacketSize) && "internal::pinsertfirst"); } - if (PacketTraits::HasBlend) { + if (PacketTraits::HasBlend || g_vectorize_sse) { // pinsertlast for (int i=0; i void product(const MatrixType& m) vcres.noalias() -= m1.transpose() * v1; VERIFY_IS_APPROX(vcres, vc2 - m1.transpose() * v1); - // test scaled products - res = square; - res.noalias() = s1 * m1 * m2.transpose(); - VERIFY_IS_APPROX(res, ((s1*m1).eval() * m2.transpose())); - res = square; - res.noalias() += s1 * m1 * m2.transpose(); - VERIFY_IS_APPROX(res, square + ((s1*m1).eval() * m2.transpose())); - res = square; - res.noalias() -= s1 * m1 * m2.transpose(); - VERIFY_IS_APPROX(res, square - ((s1*m1).eval() * m2.transpose())); - // test d ?= a+b*c rules res.noalias() = square + m1 * m2.transpose(); VERIFY_IS_APPROX(res, square + m1 * m2.transpose()); @@ -239,19 +228,4 @@ template void product(const MatrixType& m) VERIFY_IS_APPROX(square * (square*square).conjugate(), square * square.conjugate() * square.conjugate()); } - // destination with a non-default inner-stride - // see bug 1741 - if(!MatrixType::IsRowMajor) - { - typedef Matrix MatrixX; - MatrixX buffer(2*rows,2*rows); - Map > map1(buffer.data(),rows,rows,Stride(2*rows,2)); - buffer.setZero(); - VERIFY_IS_APPROX(map1 = m1 * m2.transpose(), (m1 * m2.transpose()).eval()); - buffer.setZero(); - VERIFY_IS_APPROX(map1.noalias() = m1 * m2.transpose(), (m1 * m2.transpose()).eval()); - buffer.setZero(); - VERIFY_IS_APPROX(map1.noalias() += m1 * m2.transpose(), (m1 * m2.transpose()).eval()); - } - } diff --git a/third_party/eigen/test/product_large.cpp b/third_party/eigen/test/product_large.cpp index 14a4f739d48..845cd40ca10 100644 --- a/third_party/eigen/test/product_large.cpp +++ b/third_party/eigen/test/product_large.cpp @@ -35,8 +35,6 @@ void test_product_large() for(int i = 0; i < g_repeat; i++) { CALL_SUBTEST_1( product(MatrixXf(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); CALL_SUBTEST_2( product(MatrixXd(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); - CALL_SUBTEST_2( product(MatrixXd(internal::random(1,10), internal::random(1,10))) ); - CALL_SUBTEST_3( product(MatrixXi(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); CALL_SUBTEST_4( product(MatrixXcf(internal::random(1,EIGEN_TEST_MAX_SIZE/2), internal::random(1,EIGEN_TEST_MAX_SIZE/2))) ); CALL_SUBTEST_5( product(Matrix(internal::random(1,EIGEN_TEST_MAX_SIZE), internal::random(1,EIGEN_TEST_MAX_SIZE))) ); diff --git a/third_party/eigen/test/product_mmtr.cpp b/third_party/eigen/test/product_mmtr.cpp index 35686460cdc..d3e24b012dd 100644 --- a/third_party/eigen/test/product_mmtr.cpp +++ b/third_party/eigen/test/product_mmtr.cpp @@ -82,16 +82,6 @@ template void mmtr(int size) ref2.template triangularView() = ref1.template triangularView(); matc.template triangularView() = sqc * matc * sqc.adjoint(); VERIFY_IS_APPROX(matc, ref2); - - // destination with a non-default inner-stride - // see bug 1741 - { - typedef Matrix MatrixX; - MatrixX buffer(2*size,2*size); - Map > map1(buffer.data(),size,size,Stride(2*size,2)); - buffer.setZero(); - CHECK_MMTR(map1, Lower, = s*soc*sor.adjoint()); - } } void test_product_mmtr() diff --git a/third_party/eigen/test/product_symm.cpp b/third_party/eigen/test/product_symm.cpp index 0ed027dff1a..7d1042a4f25 100644 --- a/third_party/eigen/test/product_symm.cpp +++ b/third_party/eigen/test/product_symm.cpp @@ -75,12 +75,12 @@ template void symm(int size = Size, in rhs13 = (s1*m1.adjoint()) * (s2*rhs2.adjoint())); // test row major = <...> - m2 = m1.template triangularView(); rhs32.setRandom(); rhs13 = rhs32; - VERIFY_IS_APPROX(rhs32.noalias() -= (s1*m2).template selfadjointView() * (s2*rhs3), + m2 = m1.template triangularView(); rhs12.setRandom(); rhs13 = rhs12; + VERIFY_IS_APPROX(rhs12 -= (s1*m2).template selfadjointView() * (s2*rhs3), rhs13 -= (s1*m1) * (s2 * rhs3)); m2 = m1.template triangularView(); - VERIFY_IS_APPROX(rhs32.noalias() = (s1*m2.adjoint()).template selfadjointView() * (s2*rhs3).conjugate(), + VERIFY_IS_APPROX(rhs12 = (s1*m2.adjoint()).template selfadjointView() * (s2*rhs3).conjugate(), rhs13 = (s1*m1.adjoint()) * (s2*rhs3).conjugate()); @@ -92,20 +92,6 @@ template void symm(int size = Size, in VERIFY_IS_APPROX(rhs22 = (rhs2) * (m2).template selfadjointView(), rhs23 = (rhs2) * (m1)); VERIFY_IS_APPROX(rhs22 = (s2*rhs2) * (s1*m2).template selfadjointView(), rhs23 = (s2*rhs2) * (s1*m1)); - // destination with a non-default inner-stride - // see bug 1741 - { - typedef Matrix MatrixX; - MatrixX buffer(2*cols,2*othersize); - Map > map1(buffer.data(),cols,othersize,Stride(2*rows,2)); - buffer.setZero(); - VERIFY_IS_APPROX( map1.noalias() = (s1*m2).template selfadjointView() * (s2*rhs1), - rhs13 = (s1*m1) * (s2*rhs1)); - - Map > map2(buffer.data(),rhs22.rows(),rhs22.cols(),Stride(2*rhs22.outerStride(),2)); - buffer.setZero(); - VERIFY_IS_APPROX(map2 = (rhs2) * (m2).template selfadjointView(), rhs23 = (rhs2) * (m1)); - } } void test_product_symm() diff --git a/third_party/eigen/test/product_syrk.cpp b/third_party/eigen/test/product_syrk.cpp index b8578215f99..3ebbe14cad3 100644 --- a/third_party/eigen/test/product_syrk.cpp +++ b/third_party/eigen/test/product_syrk.cpp @@ -115,17 +115,6 @@ template void syrk(const MatrixType& m) m2.setZero(); VERIFY_IS_APPROX((m2.template selfadjointView().rankUpdate(m1.row(c).adjoint(),s1)._expression()), ((s1 * m1.row(c).adjoint() * m1.row(c).adjoint().adjoint()).eval().template triangularView().toDenseMatrix())); - - // destination with a non-default inner-stride - // see bug 1741 - { - typedef Matrix MatrixX; - MatrixX buffer(2*rows,2*cols); - Map > map1(buffer.data(),rows,cols,Stride(2*rows,2)); - buffer.setZero(); - VERIFY_IS_APPROX((map1.template selfadjointView().rankUpdate(rhs2,s1)._expression()), - ((s1 * rhs2 * rhs2.adjoint()).eval().template triangularView().toDenseMatrix())); - } } void test_product_syrk() diff --git a/third_party/eigen/test/product_trmm.cpp b/third_party/eigen/test/product_trmm.cpp index ddcde9622ac..e08d9f39f5c 100644 --- a/third_party/eigen/test/product_trmm.cpp +++ b/third_party/eigen/test/product_trmm.cpp @@ -76,18 +76,8 @@ void trmm(int rows=get_random_size(), VERIFY_IS_APPROX( ge_xs = (s1*mat).adjoint().template triangularView() * ge_left.adjoint(), numext::conj(s1) * triTr.conjugate() * ge_left.adjoint()); VERIFY_IS_APPROX( ge_xs = (s1*mat).transpose().template triangularView() * ge_left.adjoint(), s1triTr * ge_left.adjoint()); + // TODO check with sub-matrix expressions ? - - // destination with a non-default inner-stride - // see bug 1741 - { - VERIFY_IS_APPROX( ge_xs.noalias() = mat.template triangularView() * ge_right, tri * ge_right); - typedef Matrix MatrixX; - MatrixX buffer(2*ge_xs.rows(),2*ge_xs.cols()); - Map > map1(buffer.data(),ge_xs.rows(),ge_xs.cols(),Stride(2*ge_xs.outerStride(),2)); - buffer.setZero(); - VERIFY_IS_APPROX( map1.noalias() = mat.template triangularView() * ge_right, tri * ge_right); - } } template diff --git a/third_party/eigen/test/product_trsolve.cpp b/third_party/eigen/test/product_trsolve.cpp index eaf62cb1139..4b97fa9d648 100644 --- a/third_party/eigen/test/product_trsolve.cpp +++ b/third_party/eigen/test/product_trsolve.cpp @@ -71,32 +71,6 @@ template void trsolve(int size=Size,int cols int c = internal::random(0,cols-1); VERIFY_TRSM(rmLhs.template triangularView(), rmRhs.col(c)); VERIFY_TRSM(cmLhs.template triangularView(), rmRhs.col(c)); - - // destination with a non-default inner-stride - // see bug 1741 - { - typedef Matrix MatrixX; - MatrixX buffer(2*cmRhs.rows(),2*cmRhs.cols()); - Map,0,Stride > map1(buffer.data(),cmRhs.rows(),cmRhs.cols(),Stride(2*cmRhs.outerStride(),2)); - Map,0,Stride > map2(buffer.data(),rmRhs.rows(),rmRhs.cols(),Stride(2*rmRhs.outerStride(),2)); - buffer.setZero(); - VERIFY_TRSM(cmLhs.conjugate().template triangularView(), map1); - buffer.setZero(); - VERIFY_TRSM(cmLhs .template triangularView(), map2); - } - - if(Size==Dynamic) - { - cmLhs.resize(0,0); - cmRhs.resize(0,cmRhs.cols()); - Matrix res = cmLhs.template triangularView().solve(cmRhs); - VERIFY_IS_EQUAL(res.rows(),0); - VERIFY_IS_EQUAL(res.cols(),cmRhs.cols()); - res = cmRhs; - cmLhs.template triangularView().solveInPlace(res); - VERIFY_IS_EQUAL(res.rows(),0); - VERIFY_IS_EQUAL(res.cols(),cmRhs.cols()); - } } void test_product_trsolve() diff --git a/third_party/eigen/test/ref.cpp b/third_party/eigen/test/ref.cpp index da399e287f1..704495affa8 100644 --- a/third_party/eigen/test/ref.cpp +++ b/third_party/eigen/test/ref.cpp @@ -102,14 +102,10 @@ template void ref_vector(const VectorType& m) Index i = internal::random(0,size-1); Index bsize = internal::random(1,size-i); - { RefMat rm0 = v1; VERIFY_IS_EQUAL(rm0, v1); } - { RefMat rm0 = v1.block(0,0,size,1); VERIFY_IS_EQUAL(rm0, v1); } - { RefDynMat rv1 = v1; VERIFY_IS_EQUAL(rv1, v1); } - { RefDynMat rv1 = v1.block(0,0,size,1); VERIFY_IS_EQUAL(rv1, v1); } - { VERIFY_RAISES_ASSERT( RefMat rm0 = v1.block(0, 0, size, 0); EIGEN_UNUSED_VARIABLE(rm0); ); } - if(VectorType::SizeAtCompileTime!=1) - { VERIFY_RAISES_ASSERT( RefDynMat rv1 = v1.block(0, 0, size, 0); EIGEN_UNUSED_VARIABLE(rv1); ); } - + RefMat rm0 = v1; + VERIFY_IS_EQUAL(rm0, v1); + RefDynMat rv1 = v1; + VERIFY_IS_EQUAL(rv1, v1); RefDynMat rv2 = v1.segment(i,bsize); VERIFY_IS_EQUAL(rv2, v1.segment(i,bsize)); rv2.setOnes(); diff --git a/third_party/eigen/test/rvalue_types.cpp b/third_party/eigen/test/rvalue_types.cpp index 6a97dae340c..8887f1b1b99 100644 --- a/third_party/eigen/test/rvalue_types.cpp +++ b/third_party/eigen/test/rvalue_types.cpp @@ -7,8 +7,6 @@ // Public License v. 2.0. If a copy of the MPL was not distributed // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. -#define EIGEN_RUNTIME_NO_MALLOC - #include "main.h" #include @@ -26,85 +24,41 @@ void rvalue_copyassign(const MatrixType& m) MatrixType tmp = m; UIntPtr src_address = reinterpret_cast(tmp.data()); - Eigen::internal::set_is_malloc_allowed(false); // moving from an rvalue reference shall never allocate // move the temporary to n MatrixType n = std::move(tmp); UIntPtr dst_address = reinterpret_cast(n.data()); + if (MatrixType::RowsAtCompileTime==Dynamic|| MatrixType::ColsAtCompileTime==Dynamic) { // verify that we actually moved the guts VERIFY_IS_EQUAL(src_address, dst_address); - VERIFY_IS_EQUAL(tmp.size(), 0); - VERIFY_IS_EQUAL(reinterpret_cast(tmp.data()), UIntPtr(0)); } // verify that the content did not change Scalar abs_diff = (m-n).array().abs().sum(); VERIFY_IS_EQUAL(abs_diff, Scalar(0)); - Eigen::internal::set_is_malloc_allowed(true); -} -template -void rvalue_transpositions(Index rows) -{ - typedef typename TranspositionsType::IndicesType PermutationVectorType; - - PermutationVectorType vec; - randomPermutationVector(vec, rows); - TranspositionsType t0(vec); - - Eigen::internal::set_is_malloc_allowed(false); // moving from an rvalue reference shall never allocate - - UIntPtr t0_address = reinterpret_cast(t0.indices().data()); - - // Move constructors: - TranspositionsType t1 = std::move(t0); - UIntPtr t1_address = reinterpret_cast(t1.indices().data()); - VERIFY_IS_EQUAL(t0_address, t1_address); - // t0 must be de-allocated: - VERIFY_IS_EQUAL(t0.size(), 0); - VERIFY_IS_EQUAL(reinterpret_cast(t0.indices().data()), UIntPtr(0)); - - - // Move assignment: - t0 = std::move(t1); - t0_address = reinterpret_cast(t0.indices().data()); - VERIFY_IS_EQUAL(t0_address, t1_address); - // t1 must be de-allocated: - VERIFY_IS_EQUAL(t1.size(), 0); - VERIFY_IS_EQUAL(reinterpret_cast(t1.indices().data()), UIntPtr(0)); - - Eigen::internal::set_is_malloc_allowed(true); } #else template void rvalue_copyassign(const MatrixType&) {} -template -void rvalue_transpositions(Index) {} #endif void test_rvalue_types() { - for(int i = 0; i < g_repeat; i++) { - CALL_SUBTEST_1(rvalue_copyassign( MatrixXf::Random(50,50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( ArrayXXf::Random(50,50).eval() )); - - CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( MatrixXf::Random(50,50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( ArrayXXf::Random(50,50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); - CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); - CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Matrix::Random(50).eval() )); + CALL_SUBTEST_1(rvalue_copyassign( Array::Random(50).eval() )); - CALL_SUBTEST_3((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); - CALL_SUBTEST_3((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); - CALL_SUBTEST_4((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); - CALL_SUBTEST_4((rvalue_transpositions >(internal::random(1,EIGEN_TEST_MAX_SIZE)))); - } + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); + CALL_SUBTEST_2(rvalue_copyassign( Array::Random().eval() )); } diff --git a/third_party/eigen/test/sparse_basic.cpp b/third_party/eigen/test/sparse_basic.cpp index 43318da796c..d0ef722b61c 100644 --- a/third_party/eigen/test/sparse_basic.cpp +++ b/third_party/eigen/test/sparse_basic.cpp @@ -612,14 +612,6 @@ template void sparse_basic(const SparseMatrixType& re iters[0] = IteratorType(m2,0); iters[1] = IteratorType(m2,m2.outerSize()-1); } - - // test reserve with empty rows/columns - { - SparseMatrixType m1(0,cols); - m1.reserve(ArrayXi::Constant(m1.outerSize(),1)); - SparseMatrixType m2(rows,0); - m2.reserve(ArrayXi::Constant(m2.outerSize(),1)); - } } diff --git a/third_party/eigen/test/stddeque.cpp b/third_party/eigen/test/stddeque.cpp index b6955f74758..b511c4e6121 100644 --- a/third_party/eigen/test/stddeque.cpp +++ b/third_party/eigen/test/stddeque.cpp @@ -18,7 +18,7 @@ void check_stddeque_matrix(const MatrixType& m) Index rows = m.rows(); Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::deque > v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::deque > v(10, MatrixType(rows,cols)), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -33,7 +33,7 @@ void check_stddeque_matrix(const MatrixType& m) ++wi; } - v.resize(21,MatrixType::Zero(rows,cols)); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -46,8 +46,8 @@ template void check_stddeque_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); - std::deque > v(10,ti), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()); + std::deque > v(10), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -62,7 +62,7 @@ void check_stddeque_transform(const TransformType&) ++wi; } - v.resize(21,ti); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -75,8 +75,8 @@ template void check_stddeque_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::deque > v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::deque > v(10), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -91,7 +91,7 @@ void check_stddeque_quaternion(const QuaternionType&) ++wi; } - v.resize(21,qi); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); diff --git a/third_party/eigen/test/stddeque_overload.cpp b/third_party/eigen/test/stddeque_overload.cpp index f495b5a04ab..4da618bbfc2 100644 --- a/third_party/eigen/test/stddeque_overload.cpp +++ b/third_party/eigen/test/stddeque_overload.cpp @@ -31,7 +31,7 @@ void check_stddeque_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::deque v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::deque v(10, MatrixType(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -64,8 +64,8 @@ template void check_stddeque_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); - std::deque v(10,ti), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()); + std::deque v(10), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -75,7 +75,7 @@ void check_stddeque_transform(const TransformType&) VERIFY_IS_APPROX(w[i], v[i]); } - v.resize(21,ti); + v.resize(21); v[20] = x; VERIFY_IS_APPROX(v[20], x); v.resize(22,y); @@ -98,8 +98,8 @@ template void check_stddeque_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::deque v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::deque v(10), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -109,7 +109,7 @@ void check_stddeque_quaternion(const QuaternionType&) VERIFY_IS_APPROX(w[i], v[i]); } - v.resize(21,qi); + v.resize(21); v[20] = x; VERIFY_IS_APPROX(v[20], x); v.resize(22,y); diff --git a/third_party/eigen/test/stdlist.cpp b/third_party/eigen/test/stdlist.cpp index 23b30ccaf1f..23cbe903970 100644 --- a/third_party/eigen/test/stdlist.cpp +++ b/third_party/eigen/test/stdlist.cpp @@ -18,7 +18,7 @@ void check_stdlist_matrix(const MatrixType& m) Index rows = m.rows(); Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::list > v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::list > v(10, MatrixType(rows,cols)), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -33,7 +33,7 @@ void check_stdlist_matrix(const MatrixType& m) ++wi; } - v.resize(21, MatrixType::Zero(rows,cols)); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -46,8 +46,8 @@ template void check_stdlist_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); - std::list > v(10,ti), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()); + std::list > v(10), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -62,7 +62,7 @@ void check_stdlist_transform(const TransformType&) ++wi; } - v.resize(21, ti); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); @@ -75,8 +75,8 @@ template void check_stdlist_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::list > v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::list > v(10), w(20, y); v.front() = x; w.front() = w.back(); VERIFY_IS_APPROX(w.front(), w.back()); @@ -91,7 +91,7 @@ void check_stdlist_quaternion(const QuaternionType&) ++wi; } - v.resize(21,qi); + v.resize(21); v.back() = x; VERIFY_IS_APPROX(v.back(), x); v.resize(22,y); diff --git a/third_party/eigen/test/stdlist_overload.cpp b/third_party/eigen/test/stdlist_overload.cpp index aea7a28465c..bb910bd439f 100644 --- a/third_party/eigen/test/stdlist_overload.cpp +++ b/third_party/eigen/test/stdlist_overload.cpp @@ -47,7 +47,7 @@ void check_stdlist_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::list v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::list v(10, MatrixType(rows,cols)), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -86,8 +86,8 @@ template void check_stdlist_transform(const TransformType&) { typedef typename TransformType::MatrixType MatrixType; - TransformType x(MatrixType::Random()), y(MatrixType::Random()), ti=TransformType::Identity(); - std::list v(10,ti), w(20, y); + TransformType x(MatrixType::Random()), y(MatrixType::Random()); + std::list v(10), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -103,7 +103,7 @@ void check_stdlist_transform(const TransformType&) ++itw; } - v.resize(21, ti); + v.resize(21); set(v, 20, x); VERIFY_IS_APPROX(*get(v, 20), x); v.resize(22,y); @@ -126,8 +126,8 @@ template void check_stdlist_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::list v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::list v(10), w(20, y); typename std::list::iterator itv = get(v, 5); typename std::list::iterator itw = get(w, 6); *itv = x; @@ -143,7 +143,7 @@ void check_stdlist_quaternion(const QuaternionType&) ++itw; } - v.resize(21,qi); + v.resize(21); set(v, 20, x); VERIFY_IS_APPROX(*get(v, 20), x); v.resize(22,y); diff --git a/third_party/eigen/test/stdvector.cpp b/third_party/eigen/test/stdvector.cpp index 383d9a5095e..fa928ea4fe9 100644 --- a/third_party/eigen/test/stdvector.cpp +++ b/third_party/eigen/test/stdvector.cpp @@ -17,7 +17,7 @@ void check_stdvector_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::vector > v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::vector > v(10, MatrixType(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -86,8 +86,8 @@ template void check_stdvector_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::vector > v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::vector > v(10), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); diff --git a/third_party/eigen/test/stdvector_overload.cpp b/third_party/eigen/test/stdvector_overload.cpp index 637e3ef52f3..959665954e4 100644 --- a/third_party/eigen/test/stdvector_overload.cpp +++ b/third_party/eigen/test/stdvector_overload.cpp @@ -31,7 +31,7 @@ void check_stdvector_matrix(const MatrixType& m) typename MatrixType::Index rows = m.rows(); typename MatrixType::Index cols = m.cols(); MatrixType x = MatrixType::Random(rows,cols), y = MatrixType::Random(rows,cols); - std::vector v(10, MatrixType::Zero(rows,cols)), w(20, y); + std::vector v(10, MatrixType(rows,cols)), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); @@ -100,8 +100,8 @@ template void check_stdvector_quaternion(const QuaternionType&) { typedef typename QuaternionType::Coefficients Coefficients; - QuaternionType x(Coefficients::Random()), y(Coefficients::Random()), qi=QuaternionType::Identity(); - std::vector v(10,qi), w(20, y); + QuaternionType x(Coefficients::Random()), y(Coefficients::Random()); + std::vector v(10), w(20, y); v[5] = x; w[6] = v[5]; VERIFY_IS_APPROX(w[6], v[5]); diff --git a/third_party/eigen/test/vectorization_logic.cpp b/third_party/eigen/test/vectorization_logic.cpp index c2f77bfec21..37e7495f5b5 100644 --- a/third_party/eigen/test/vectorization_logic.cpp +++ b/third_party/eigen/test/vectorization_logic.cpp @@ -22,14 +22,6 @@ #include "main.h" #include -// Disable "ignoring attributes on template argument" -// for packet_traits -// => The only workaround would be to wrap _m128 and the likes -// within wrappers. -#if EIGEN_GNUC_AT_LEAST(6,0) - #pragma GCC diagnostic ignored "-Wignored-attributes" -#endif - using internal::demangle_flags; using internal::demangle_traversal; using internal::demangle_unrolling; diff --git a/third_party/eigen/unsupported/Eigen/ArpackSupport b/third_party/eigen/unsupported/Eigen/ArpackSupport index a0d4820e125..37a2799ef2f 100644 --- a/third_party/eigen/unsupported/Eigen/ArpackSupport +++ b/third_party/eigen/unsupported/Eigen/ArpackSupport @@ -11,6 +11,8 @@ #include +#include + /** \defgroup ArpackSupport_Module Arpack support module * * This module provides a wrapper to Arpack, a library for sparse eigenvalue decomposition. @@ -21,8 +23,6 @@ */ #include - -#include #include "src/Eigenvalues/ArpackSelfAdjointEigenSolver.h" #include diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h index c28a10dd494..9b2cb3ff6bc 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorContractionMapper.h @@ -113,7 +113,6 @@ class SimpleTensorContractionMapper { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index computeIndex(Index row, Index col) const { const bool left = (side == Lhs); - EIGEN_UNUSED_VARIABLE(left); // annoying bug in g++8.1: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85963 Index nocontract_val = left ? row : col; Index linidx = 0; for (int i = static_cast(array_size::value) - 1; i > 0; i--) { @@ -152,7 +151,6 @@ class SimpleTensorContractionMapper { EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE IndexPair computeIndexPair(Index row, Index col, const Index distance) const { const bool left = (side == Lhs); - EIGEN_UNUSED_VARIABLE(left); // annoying bug in g++8.1: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85963 Index nocontract_val[2] = {left ? row : col, left ? row + distance : col}; Index linidx[2] = {0, 0}; if (array_size::value > array_size::value) { diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h index a5e084a2409..17f04665a12 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h @@ -31,7 +31,7 @@ class Barrier { eigen_assert(((count << 1) >> 1) == count); } ~Barrier() { - eigen_plain_assert((state_>>1) == 0); + eigen_assert((state_>>1) == 0); } void Notify() { diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h index 4749d6240af..71d55552de6 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/EventCount.h @@ -58,7 +58,7 @@ class EventCount { ~EventCount() { // Ensure there are no waiters. - eigen_plain_assert((state_.load() & (kStackMask | kWaiterMask)) == kStackMask); + eigen_assert((state_.load() & (kStackMask | kWaiterMask)) == kStackMask); } // Prewait prepares for waiting. diff --git a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h index 6e505fc14c8..05ed76cbe49 100644 --- a/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h +++ b/third_party/eigen/unsupported/Eigen/CXX11/src/ThreadPool/RunQueue.h @@ -47,7 +47,7 @@ class RunQueue { array_[i].state.store(kEmpty, std::memory_order_relaxed); } - ~RunQueue() { eigen_plain_assert(Size() == 0); } + ~RunQueue() { eigen_assert(Size() == 0); } // PushFront inserts w at the beginning of the queue. // If queue is full returns w, otherwise returns default-constructed Work. diff --git a/third_party/eigen/unsupported/Eigen/Polynomials b/third_party/eigen/unsupported/Eigen/Polynomials index 334b031427a..cece563374e 100644 --- a/third_party/eigen/unsupported/Eigen/Polynomials +++ b/third_party/eigen/unsupported/Eigen/Polynomials @@ -11,10 +11,10 @@ #include -#include - #include +#include + // Note that EIGEN_HIDE_HEAVY_CODE has to be defined per module #if (defined EIGEN_EXTERN_INSTANTIATIONS) && (EIGEN_EXTERN_INSTANTIATIONS>=2) #ifndef EIGEN_HIDE_HEAVY_CODE diff --git a/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h b/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h index 58f3f3319d4..2f50e996805 100755 --- a/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h +++ b/third_party/eigen/unsupported/Eigen/src/AutoDiff/AutoDiffScalar.h @@ -453,24 +453,6 @@ struct auto_diff_special_op<_DerType, false> void operator+() const; }; -template -void make_coherent_expression(CwiseBinaryOp xpr, const RefType &ref) -{ - make_coherent(xpr.const_cast_derived().lhs(), ref); - make_coherent(xpr.const_cast_derived().rhs(), ref); -} - -template -void make_coherent_expression(const CwiseUnaryOp &xpr, const RefType &ref) -{ - make_coherent(xpr.nestedExpression().const_cast_derived(), ref); -} - -// needed for compilation only -template -void make_coherent_expression(const CwiseNullaryOp &, const RefType &) -{} - template struct make_coherent_impl, B> { typedef Matrix A; @@ -480,10 +462,6 @@ struct make_coherent_impl struct make_coherent_impl, - Matrix > { + Matrix > { typedef Matrix A; typedef Matrix B; static void run(A& a, B& b) { diff --git a/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h b/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h index 4170d26b6f6..866a8a46018 100644 --- a/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h +++ b/third_party/eigen/unsupported/Eigen/src/Eigenvalues/ArpackSelfAdjointEigenSolver.h @@ -3,9 +3,24 @@ // // Copyright (C) 2012 David Harmon // -// This Source Code Form is subject to the terms of the Mozilla -// Public License v. 2.0. If a copy of the MPL was not distributed -// with this file, You can obtain one at http://mozilla.org/MPL/2.0/. +// Eigen is free software; you can redistribute it and/or +// modify it under the terms of the GNU Lesser General Public +// License as published by the Free Software Foundation; either +// version 3 of the License, or (at your option) any later version. +// +// Alternatively, you can redistribute it and/or +// modify it under the terms of the GNU General Public License as +// published by the Free Software Foundation; either version 2 of +// the License, or (at your option) any later version. +// +// Eigen is distributed in the hope that it will be useful, but WITHOUT ANY +// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS +// FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License or the +// GNU General Public License for more details. +// +// You should have received a copy of the GNU Lesser General Public +// License and a copy of the GNU General Public License along with +// Eigen. If not, see . #ifndef EIGEN_ARPACKGENERALIZEDSELFADJOINTEIGENSOLVER_H #define EIGEN_ARPACKGENERALIZEDSELFADJOINTEIGENSOLVER_H diff --git a/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h b/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h index 7c1f716e2a8..d49aa17f510 100644 --- a/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h +++ b/third_party/eigen/unsupported/Eigen/src/FFT/ei_fftw_impl.h @@ -231,8 +231,6 @@ namespace internal { protected: typedef fftw_plan PlanData; - typedef Eigen::numext::int64_t int64_t; - typedef std::map PlanMap; PlanMap m_plans; diff --git a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h index 0b0ee6546a8..e5ebbcf2378 100644 --- a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h +++ b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h @@ -412,7 +412,7 @@ template struct MatrixExponentialReturnValue inline void evalTo(ResultType& result) const { const typename internal::nested_eval::type tmp(m_src); - internal::matrix_exp_compute(tmp, result, internal::is_exp_known_type()); + internal::matrix_exp_compute(tmp, result, internal::is_exp_known_type()); } Index rows() const { return m_src.rows(); } diff --git a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h index 9de0c3574eb..2e5abda381b 100644 --- a/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h +++ b/third_party/eigen/unsupported/Eigen/src/MatrixFunctions/MatrixSquareRoot.h @@ -253,19 +253,18 @@ struct matrix_sqrt_compute template struct matrix_sqrt_compute { - typedef typename MatrixType::PlainObject PlainType; template static void run(const MatrixType &arg, ResultType &result) { eigen_assert(arg.rows() == arg.cols()); // Compute Schur decomposition of arg - const RealSchur schurOfA(arg); - const PlainType& T = schurOfA.matrixT(); - const PlainType& U = schurOfA.matrixU(); + const RealSchur schurOfA(arg); + const MatrixType& T = schurOfA.matrixT(); + const MatrixType& U = schurOfA.matrixU(); // Compute square root of T - PlainType sqrtT = PlainType::Zero(arg.rows(), arg.cols()); + MatrixType sqrtT = MatrixType::Zero(arg.rows(), arg.cols()); matrix_sqrt_quasi_triangular(T, sqrtT); // Compute square root of arg @@ -279,19 +278,18 @@ struct matrix_sqrt_compute template struct matrix_sqrt_compute { - typedef typename MatrixType::PlainObject PlainType; template static void run(const MatrixType &arg, ResultType &result) { eigen_assert(arg.rows() == arg.cols()); // Compute Schur decomposition of arg - const ComplexSchur schurOfA(arg); - const PlainType& T = schurOfA.matrixT(); - const PlainType& U = schurOfA.matrixU(); + const ComplexSchur schurOfA(arg); + const MatrixType& T = schurOfA.matrixT(); + const MatrixType& U = schurOfA.matrixU(); // Compute square root of T - PlainType sqrtT; + MatrixType sqrtT; matrix_sqrt_triangular(T, sqrtT); // Compute square root of arg diff --git a/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h b/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h index 359836cac80..b515c29208e 100644 --- a/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h +++ b/third_party/eigen/unsupported/Eigen/src/Polynomials/Companion.h @@ -75,7 +75,8 @@ class companion void setPolynomial( const VectorType& poly ) { const Index deg = poly.size()-1; - m_monic = -poly.head(deg)/poly[deg]; + m_monic = -1/poly[deg] * poly.head(deg); + //m_bl_diag.setIdentity( deg-1 ); m_bl_diag.setOnes(deg-1); } @@ -88,13 +89,13 @@ class companion { const Index deg = m_monic.size(); const Index deg_1 = deg-1; - DenseCompanionMatrixType companMat(deg,deg); - companMat << + DenseCompanionMatrixType companion(deg,deg); + companion << ( LeftBlock(deg,deg_1) << LeftBlockFirstRow::Zero(1,deg_1), BottomLeftBlock::Identity(deg-1,deg-1)*m_bl_diag.asDiagonal() ).finished() , m_monic; - return companMat; + return companion; } @@ -106,8 +107,8 @@ class companion * colB and rowB are repectively the multipliers for * the column and the row in order to balance them. * */ - bool balanced( RealScalar colNorm, RealScalar rowNorm, - bool& isBalanced, RealScalar& colB, RealScalar& rowB ); + bool balanced( Scalar colNorm, Scalar rowNorm, + bool& isBalanced, Scalar& colB, Scalar& rowB ); /** Helper function for the balancing algorithm. * \returns true if the row and the column, having colNorm and rowNorm @@ -115,8 +116,8 @@ class companion * colB and rowB are repectively the multipliers for * the column and the row in order to balance them. * */ - bool balancedR( RealScalar colNorm, RealScalar rowNorm, - bool& isBalanced, RealScalar& colB, RealScalar& rowB ); + bool balancedR( Scalar colNorm, Scalar rowNorm, + bool& isBalanced, Scalar& colB, Scalar& rowB ); public: /** @@ -138,10 +139,10 @@ class companion template< typename _Scalar, int _Deg > inline -bool companion<_Scalar,_Deg>::balanced( RealScalar colNorm, RealScalar rowNorm, - bool& isBalanced, RealScalar& colB, RealScalar& rowB ) +bool companion<_Scalar,_Deg>::balanced( Scalar colNorm, Scalar rowNorm, + bool& isBalanced, Scalar& colB, Scalar& rowB ) { - if( RealScalar(0) == colNorm || RealScalar(0) == rowNorm ){ return true; } + if( Scalar(0) == colNorm || Scalar(0) == rowNorm ){ return true; } else { //To find the balancing coefficients, if the radix is 2, @@ -149,29 +150,29 @@ bool companion<_Scalar,_Deg>::balanced( RealScalar colNorm, RealScalar rowNorm, // \f$ 2^{2\sigma-1} < rowNorm / colNorm \le 2^{2\sigma+1} \f$ // then the balancing coefficient for the row is \f$ 1/2^{\sigma} \f$ // and the balancing coefficient for the column is \f$ 2^{\sigma} \f$ - rowB = rowNorm / radix(); - colB = RealScalar(1); - const RealScalar s = colNorm + rowNorm; + rowB = rowNorm / radix(); + colB = Scalar(1); + const Scalar s = colNorm + rowNorm; while (colNorm < rowB) { - colB *= radix(); - colNorm *= radix2(); + colB *= radix(); + colNorm *= radix2(); } - rowB = rowNorm * radix(); + rowB = rowNorm * radix(); while (colNorm >= rowB) { - colB /= radix(); - colNorm /= radix2(); + colB /= radix(); + colNorm /= radix2(); } //This line is used to avoid insubstantial balancing - if ((rowNorm + colNorm) < RealScalar(0.95) * s * colB) + if ((rowNorm + colNorm) < Scalar(0.95) * s * colB) { isBalanced = false; - rowB = RealScalar(1) / colB; + rowB = Scalar(1) / colB; return false; } else{ @@ -181,21 +182,21 @@ bool companion<_Scalar,_Deg>::balanced( RealScalar colNorm, RealScalar rowNorm, template< typename _Scalar, int _Deg > inline -bool companion<_Scalar,_Deg>::balancedR( RealScalar colNorm, RealScalar rowNorm, - bool& isBalanced, RealScalar& colB, RealScalar& rowB ) +bool companion<_Scalar,_Deg>::balancedR( Scalar colNorm, Scalar rowNorm, + bool& isBalanced, Scalar& colB, Scalar& rowB ) { - if( RealScalar(0) == colNorm || RealScalar(0) == rowNorm ){ return true; } + if( Scalar(0) == colNorm || Scalar(0) == rowNorm ){ return true; } else { /** * Set the norm of the column and the row to the geometric mean * of the row and column norm */ - const RealScalar q = colNorm/rowNorm; + const _Scalar q = colNorm/rowNorm; if( !isApprox( q, _Scalar(1) ) ) { rowB = sqrt( colNorm/rowNorm ); - colB = RealScalar(1)/rowB; + colB = Scalar(1)/rowB; isBalanced = false; return false; @@ -218,8 +219,8 @@ void companion<_Scalar,_Deg>::balance() while( !hasConverged ) { hasConverged = true; - RealScalar colNorm,rowNorm; - RealScalar colB,rowB; + Scalar colNorm,rowNorm; + Scalar colB,rowB; //First row, first column excluding the diagonal //============================================== diff --git a/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h b/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h index 5e0ecbb43b3..03198ec8ee2 100644 --- a/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h +++ b/third_party/eigen/unsupported/Eigen/src/Polynomials/PolynomialSolver.h @@ -99,7 +99,7 @@ class PolynomialSolverBase */ inline const RootType& greatestRoot() const { - std::greater greater; + std::greater greater; return selectComplexRoot_withRespectToNorm( greater ); } @@ -108,7 +108,7 @@ class PolynomialSolverBase */ inline const RootType& smallestRoot() const { - std::less less; + std::less less; return selectComplexRoot_withRespectToNorm( less ); } @@ -126,7 +126,7 @@ class PolynomialSolverBase for( Index i=0; i::dummy_precision() ) const { - std::greater greater; + std::greater greater; return selectRealRoot_withRespectToAbsRealPart( greater, hasArealRoot, absImaginaryThreshold ); } @@ -236,7 +236,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::less less; + std::less less; return selectRealRoot_withRespectToAbsRealPart( less, hasArealRoot, absImaginaryThreshold ); } @@ -259,7 +259,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::greater greater; + std::greater greater; return selectRealRoot_withRespectToRealPart( greater, hasArealRoot, absImaginaryThreshold ); } @@ -282,7 +282,7 @@ class PolynomialSolverBase bool& hasArealRoot, const RealScalar& absImaginaryThreshold = NumTraits::dummy_precision() ) const { - std::less less; + std::less less; return selectRealRoot_withRespectToRealPart( less, hasArealRoot, absImaginaryThreshold ); } @@ -327,7 +327,7 @@ class PolynomialSolverBase * However, almost always, correct accuracy is reached even in these cases for 64bit * (double) floating types and small polynomial degree (<20). */ -template +template< typename _Scalar, int _Deg > class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> { public: @@ -337,10 +337,7 @@ class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> EIGEN_POLYNOMIAL_SOLVER_BASE_INHERITED_TYPES( PS_Base ) typedef Matrix CompanionMatrixType; - typedef typename internal::conditional::IsComplex, - ComplexEigenSolver, - EigenSolver >::type EigenSolverType; - typedef typename internal::conditional::IsComplex, Scalar, std::complex >::type ComplexScalar; + typedef EigenSolver EigenSolverType; public: /** Computes the complex roots of a new polynomial. */ @@ -355,25 +352,6 @@ class PolynomialSolver : public PolynomialSolverBase<_Scalar,_Deg> companion.balance(); m_eigenSolver.compute( companion.denseMatrix() ); m_roots = m_eigenSolver.eigenvalues(); - // cleanup noise in imaginary part of real roots: - // if the imaginary part is rather small compared to the real part - // and that cancelling the imaginary part yield a smaller evaluation, - // then it's safe to keep the real part only. - RealScalar coarse_prec = RealScalar(std::pow(4,poly.size()+1))*NumTraits::epsilon(); - for(Index i = 0; i::Scalar u, DenseIndex degree, const typename SplineTraits::KnotVectorType& knots); diff --git a/third_party/eigen/unsupported/test/NonLinearOptimization.cpp b/third_party/eigen/unsupported/test/NonLinearOptimization.cpp index dd93c21e9fe..f0c336c15ef 100644 --- a/third_party/eigen/unsupported/test/NonLinearOptimization.cpp +++ b/third_party/eigen/unsupported/test/NonLinearOptimization.cpp @@ -15,15 +15,6 @@ // tolerance for chekcing number of iterations #define LM_EVAL_COUNT_TOL 4/3 -#define LM_CHECK_N_ITERS(SOLVER,NFEV,NJEV) { \ - ++g_test_level; \ - VERIFY_IS_EQUAL(SOLVER.nfev, NFEV); \ - VERIFY_IS_EQUAL(SOLVER.njev, NJEV); \ - --g_test_level; \ - VERIFY(SOLVER.nfev <= NFEV * LM_EVAL_COUNT_TOL); \ - VERIFY(SOLVER.njev <= NJEV * LM_EVAL_COUNT_TOL); \ - } - int fcn_chkder(const VectorXd &x, VectorXd &fvec, MatrixXd &fjac, int iflag) { /* subroutine fcn for chkder example. */ @@ -189,7 +180,8 @@ void testLmder1() // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 6, 5); + VERIFY_IS_EQUAL(lm.nfev, 6); + VERIFY_IS_EQUAL(lm.njev, 5); // check norm VERIFY_IS_APPROX(lm.fvec.blueNorm(), 0.09063596); @@ -217,7 +209,8 @@ void testLmder() // check return values VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 6, 5); + VERIFY_IS_EQUAL(lm.nfev, 6); + VERIFY_IS_EQUAL(lm.njev, 5); // check norm fnorm = lm.fvec.blueNorm(); @@ -301,7 +294,8 @@ void testHybrj1() // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(solver, 11, 1); + VERIFY_IS_EQUAL(solver.nfev, 11); + VERIFY_IS_EQUAL(solver.njev, 1); // check norm VERIFY_IS_APPROX(solver.fvec.blueNorm(), 1.192636e-08); @@ -335,7 +329,8 @@ void testHybrj() // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(solver, 11, 1); + VERIFY_IS_EQUAL(solver.nfev, 11); + VERIFY_IS_EQUAL(solver.njev, 1); // check norm VERIFY_IS_APPROX(solver.fvec.blueNorm(), 1.192636e-08); @@ -490,7 +485,8 @@ void testLmstr1() // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 6, 5); + VERIFY_IS_EQUAL(lm.nfev, 6); + VERIFY_IS_EQUAL(lm.njev, 5); // check norm VERIFY_IS_APPROX(lm.fvec.blueNorm(), 0.09063596); @@ -518,7 +514,8 @@ void testLmstr() // check return values VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 6, 5); + VERIFY_IS_EQUAL(lm.nfev, 6); + VERIFY_IS_EQUAL(lm.njev, 5); // check norm fnorm = lm.fvec.blueNorm(); @@ -689,7 +686,8 @@ void testNistChwirut2(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 10, 8); + VERIFY_IS_EQUAL(lm.nfev, 10); + VERIFY_IS_EQUAL(lm.njev, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.1304802941E+02); // check x @@ -709,7 +707,8 @@ void testNistChwirut2(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 7, 6); + VERIFY_IS_EQUAL(lm.nfev, 7); + VERIFY_IS_EQUAL(lm.njev, 6); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.1304802941E+02); // check x @@ -767,7 +766,8 @@ void testNistMisra1a(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 19, 15); + VERIFY_IS_EQUAL(lm.nfev, 19); + VERIFY_IS_EQUAL(lm.njev, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.2455138894E-01); // check x @@ -783,7 +783,8 @@ void testNistMisra1a(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 5, 4); + VERIFY_IS_EQUAL(lm.nfev, 5); + VERIFY_IS_EQUAL(lm.njev, 4); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.2455138894E-01); // check x @@ -855,7 +856,8 @@ void testNistHahn1(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 11, 10); + VERIFY_IS_EQUAL(lm.nfev, 11); + VERIFY_IS_EQUAL(lm.njev, 10); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.5324382854E+00); // check x @@ -876,7 +878,8 @@ void testNistHahn1(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 11, 10); + VERIFY_IS_EQUAL(lm.nfev, 11); + VERIFY_IS_EQUAL(lm.njev, 10); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.5324382854E+00); // check x @@ -939,7 +942,8 @@ void testNistMisra1d(void) // check return value VERIFY_IS_EQUAL(info, 3); - LM_CHECK_N_ITERS(lm, 9, 7); + VERIFY_IS_EQUAL(lm.nfev, 9); + VERIFY_IS_EQUAL(lm.njev, 7); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6419295283E-02); // check x @@ -955,7 +959,8 @@ void testNistMisra1d(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 4, 3); + VERIFY_IS_EQUAL(lm.nfev, 4); + VERIFY_IS_EQUAL(lm.njev, 3); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6419295283E-02); // check x @@ -1015,7 +1020,8 @@ void testNistLanczos1(void) // check return value VERIFY_IS_EQUAL(info, 2); - LM_CHECK_N_ITERS(lm, 79, 72); + VERIFY_IS_EQUAL(lm.nfev, 79); + VERIFY_IS_EQUAL(lm.njev, 72); // check norm^2 std::cout.precision(30); std::cout << lm.fvec.squaredNorm() << "\n"; @@ -1037,7 +1043,8 @@ void testNistLanczos1(void) // check return value VERIFY_IS_EQUAL(info, 2); - LM_CHECK_N_ITERS(lm, 9, 8); + VERIFY_IS_EQUAL(lm.nfev, 9); + VERIFY_IS_EQUAL(lm.njev, 8); // check norm^2 VERIFY(lm.fvec.squaredNorm() <= 1.4307867721E-25); // check x @@ -1101,7 +1108,8 @@ void testNistRat42(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 10, 8); + VERIFY_IS_EQUAL(lm.nfev, 10); + VERIFY_IS_EQUAL(lm.njev, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.0565229338E+00); // check x @@ -1118,7 +1126,8 @@ void testNistRat42(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 6, 5); + VERIFY_IS_EQUAL(lm.nfev, 6); + VERIFY_IS_EQUAL(lm.njev, 5); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.0565229338E+00); // check x @@ -1177,7 +1186,8 @@ void testNistMGH10(void) // check return value VERIFY_IS_EQUAL(info, 2); - LM_CHECK_N_ITERS(lm, 284, 249); + VERIFY_IS_EQUAL(lm.nfev, 284 ); + VERIFY_IS_EQUAL(lm.njev, 249 ); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7945855171E+01); // check x @@ -1194,7 +1204,8 @@ void testNistMGH10(void) // check return value VERIFY_IS_EQUAL(info, 3); - LM_CHECK_N_ITERS(lm, 126, 116); + VERIFY_IS_EQUAL(lm.nfev, 126); + VERIFY_IS_EQUAL(lm.njev, 116); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7945855171E+01); // check x @@ -1254,7 +1265,8 @@ void testNistBoxBOD(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 31, 25); + VERIFY(lm.nfev < 31); // 31 + VERIFY(lm.njev < 25); // 25 // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.1680088766E+03); // check x @@ -1272,8 +1284,9 @@ void testNistBoxBOD(void) info = lm.minimize(x); // check return value - VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 15, 14); + VERIFY_IS_EQUAL(info, 1); + VERIFY_IS_EQUAL(lm.nfev, 15 ); + VERIFY_IS_EQUAL(lm.njev, 14 ); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.1680088766E+03); // check x @@ -1343,7 +1356,12 @@ void testNistMGH17(void) // check return value VERIFY_IS_EQUAL(info, 2); - LM_CHECK_N_ITERS(lm, 602, 545); + ++g_test_level; + VERIFY_IS_EQUAL(lm.nfev, 602); // 602 + VERIFY_IS_EQUAL(lm.njev, 545); // 545 + --g_test_level; + VERIFY(lm.nfev < 602 * LM_EVAL_COUNT_TOL); + VERIFY(lm.njev < 545 * LM_EVAL_COUNT_TOL); /* * Second try @@ -1355,7 +1373,8 @@ void testNistMGH17(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 18, 15); + VERIFY_IS_EQUAL(lm.nfev, 18); + VERIFY_IS_EQUAL(lm.njev, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.4648946975E-05); // check x @@ -1419,8 +1438,9 @@ void testNistMGH09(void) info = lm.minimize(x); // check return value - VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 490, 376); + VERIFY_IS_EQUAL(info, 1); + VERIFY_IS_EQUAL(lm.nfev, 490 ); + VERIFY_IS_EQUAL(lm.njev, 376 ); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 3.0750560385E-04); // check x @@ -1439,7 +1459,8 @@ void testNistMGH09(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 18, 16); + VERIFY_IS_EQUAL(lm.nfev, 18); + VERIFY_IS_EQUAL(lm.njev, 16); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 3.0750560385E-04); // check x @@ -1504,7 +1525,8 @@ void testNistBennett5(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 758, 744); + VERIFY_IS_EQUAL(lm.nfev, 758); + VERIFY_IS_EQUAL(lm.njev, 744); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.2404744073E-04); // check x @@ -1521,7 +1543,8 @@ void testNistBennett5(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 203, 192); + VERIFY_IS_EQUAL(lm.nfev, 203); + VERIFY_IS_EQUAL(lm.njev, 192); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.2404744073E-04); // check x @@ -1590,7 +1613,8 @@ void testNistThurber(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 39,36); + VERIFY_IS_EQUAL(lm.nfev, 39); + VERIFY_IS_EQUAL(lm.njev, 36); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6427082397E+03); // check x @@ -1614,7 +1638,8 @@ void testNistThurber(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 29, 28); + VERIFY_IS_EQUAL(lm.nfev, 29); + VERIFY_IS_EQUAL(lm.njev, 28); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 5.6427082397E+03); // check x @@ -1680,7 +1705,8 @@ void testNistRat43(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 27, 20); + VERIFY_IS_EQUAL(lm.nfev, 27); + VERIFY_IS_EQUAL(lm.njev, 20); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7864049080E+03); // check x @@ -1701,7 +1727,8 @@ void testNistRat43(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 9, 8); + VERIFY_IS_EQUAL(lm.nfev, 9); + VERIFY_IS_EQUAL(lm.njev, 8); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 8.7864049080E+03); // check x @@ -1763,7 +1790,8 @@ void testNistEckerle4(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 18, 15); + VERIFY_IS_EQUAL(lm.nfev, 18); + VERIFY_IS_EQUAL(lm.njev, 15); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.4635887487E-03); // check x @@ -1780,7 +1808,8 @@ void testNistEckerle4(void) // check return value VERIFY_IS_EQUAL(info, 1); - LM_CHECK_N_ITERS(lm, 7, 6); + VERIFY_IS_EQUAL(lm.nfev, 7); + VERIFY_IS_EQUAL(lm.njev, 6); // check norm^2 VERIFY_IS_APPROX(lm.fvec.squaredNorm(), 1.4635887487E-03); // check x diff --git a/third_party/eigen/unsupported/test/autodiff.cpp b/third_party/eigen/unsupported/test/autodiff.cpp index 1d8c8b5fd71..1c5e0dc6639 100644 --- a/third_party/eigen/unsupported/test/autodiff.cpp +++ b/third_party/eigen/unsupported/test/autodiff.cpp @@ -352,21 +352,6 @@ double bug_1264() { return v2(0).value(); } -// check with expressions on constants -double bug_1281() { - int n = 2; - typedef AutoDiffScalar AD; - const AD c = 1.; - AD x0(2,n,0); - AD y1 = (AD(c)+AD(c))*x0; - y1 = x0 * (AD(c)+AD(c)); - AD y2 = (-AD(c))+x0; - y2 = x0+(-AD(c)); - AD y3 = (AD(c)*(-AD(c))+AD(c))*x0; - y3 = x0 * (AD(c)*(-AD(c))+AD(c)); - return (y1+y2+y3).value(); -} - #endif void test_autodiff() @@ -382,6 +367,5 @@ void test_autodiff() CALL_SUBTEST_5( bug_1223() ); CALL_SUBTEST_5( bug_1260() ); CALL_SUBTEST_5( bug_1261() ); - CALL_SUBTEST_5( bug_1281() ); } diff --git a/third_party/eigen/unsupported/test/matrix_function.cpp b/third_party/eigen/unsupported/test/matrix_function.cpp index 005c9c15f35..6a2b2194a08 100644 --- a/third_party/eigen/unsupported/test/matrix_function.cpp +++ b/third_party/eigen/unsupported/test/matrix_function.cpp @@ -177,39 +177,6 @@ void testMatrixType(const MatrixType& m) } } -template -void testMapRef(const MatrixType& A) -{ - // Test if passing Ref and Map objects is possible - // (Regression test for Bug #1796) - Index size = A.rows(); - MatrixType X; X.setRandom(size, size); - MatrixType Y(size,size); - Ref< MatrixType> R(Y); - Ref Rc(X); - Map< MatrixType> M(Y.data(), size, size); - Map Mc(X.data(), size, size); - - X = X*X; // make sure sqrt is possible - Y = X.sqrt(); - R = Rc.sqrt(); - M = Mc.sqrt(); - Y = X.exp(); - R = Rc.exp(); - M = Mc.exp(); - X = Y; // make sure log is possible - Y = X.log(); - R = Rc.log(); - M = Mc.log(); - - Y = X.cos() + Rc.cos() + Mc.cos(); - Y = X.sin() + Rc.sin() + Mc.sin(); - - Y = X.cosh() + Rc.cosh() + Mc.cosh(); - Y = X.sinh() + Rc.sinh() + Mc.sinh(); -} - - void test_matrix_function() { CALL_SUBTEST_1(testMatrixType(Matrix())); @@ -219,9 +186,4 @@ void test_matrix_function() CALL_SUBTEST_5(testMatrixType(Matrix())); CALL_SUBTEST_6(testMatrixType(Matrix4cd())); CALL_SUBTEST_7(testMatrixType(MatrixXd(13,13))); - - CALL_SUBTEST_1(testMapRef(Matrix())); - CALL_SUBTEST_2(testMapRef(Matrix3cf())); - CALL_SUBTEST_3(testMapRef(MatrixXf(8,8))); - CALL_SUBTEST_7(testMapRef(MatrixXd(13,13))); } diff --git a/third_party/eigen/unsupported/test/mpreal/mpreal.h b/third_party/eigen/unsupported/test/mpreal/mpreal.h index 8404f1ff837..fe47bc3f88d 100644 --- a/third_party/eigen/unsupported/test/mpreal/mpreal.h +++ b/third_party/eigen/unsupported/test/mpreal/mpreal.h @@ -1,3104 +1,3104 @@ -/* - MPFR C++: Multi-precision floating point number class for C++. - Based on MPFR library: http://mpfr.org - - Project homepage: http://www.holoborodko.com/pavel/mpfr - Contact e-mail: pavel@holoborodko.com - - Copyright (c) 2008-2015 Pavel Holoborodko - - Contributors: - Dmitriy Gubanov, Konstantin Holoborodko, Brian Gladman, - Helmut Jarausch, Fokko Beekhof, Ulrich Mutze, Heinz van Saanen, - Pere Constans, Peter van Hoof, Gael Guennebaud, Tsai Chia Cheng, - Alexei Zubanov, Jauhien Piatlicki, Victor Berger, John Westwood, - Petr Aleksandrov, Orion Poplawski, Charles Karney, Arash Partow, - Rodney James, Jorge Leitao. - - Licensing: - (A) MPFR C++ is under GNU General Public License ("GPL"). - - (B) Non-free licenses may also be purchased from the author, for users who - do not want their programs protected by the GPL. - - The non-free licenses are for users that wish to use MPFR C++ in - their products but are unwilling to release their software - under the GPL (which would require them to release source code - and allow free redistribution). - - Such users can purchase an unlimited-use license from the author. - Contact us for more details. - - GNU General Public License ("GPL") copyright permissions statement: - ************************************************************************** - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . -*/ - -#ifndef __MPREAL_H__ -#define __MPREAL_H__ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -// Options -#define MPREAL_HAVE_MSVC_DEBUGVIEW // Enable Debugger Visualizer for "Debug" builds in MSVC. -#define MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS // Enable extended std::numeric_limits specialization. - // Meaning that "digits", "round_style" and similar members are defined as functions, not constants. - // See std::numeric_limits at the end of the file for more information. - -// Library version -#define MPREAL_VERSION_MAJOR 3 -#define MPREAL_VERSION_MINOR 6 -#define MPREAL_VERSION_PATCHLEVEL 2 -#define MPREAL_VERSION_STRING "3.6.2" - -// Detect compiler using signatures from http://predef.sourceforge.net/ -#if defined(__GNUC__) - #define IsInf(x) (isinf)(x) // GNU C++/Intel ICC compiler on Linux -#elif defined(_MSC_VER) // Microsoft Visual C++ - #define IsInf(x) (!_finite(x)) -#else - #define IsInf(x) (std::isinf)(x) // GNU C/C++ (and/or other compilers), just hope for C99 conformance -#endif - -// A Clang feature extension to determine compiler features. -#ifndef __has_feature - #define __has_feature(x) 0 -#endif - -// Detect support for r-value references (move semantic). Borrowed from Eigen. -#if (__has_feature(cxx_rvalue_references) || \ - defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L || \ - (defined(_MSC_VER) && _MSC_VER >= 1600)) - - #define MPREAL_HAVE_MOVE_SUPPORT - - // Use fields in mpfr_t structure to check if it was initialized / set dummy initialization - #define mpfr_is_initialized(x) (0 != (x)->_mpfr_d) - #define mpfr_set_uninitialized(x) ((x)->_mpfr_d = 0 ) -#endif - -// Detect support for explicit converters. -#if (__has_feature(cxx_explicit_conversions) || \ - (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GNUC_MINOR__ >= 5) || __cplusplus >= 201103L || \ - (defined(_MSC_VER) && _MSC_VER >= 1800)) - - #define MPREAL_HAVE_EXPLICIT_CONVERTERS -#endif - -#define MPFR_USE_INTMAX_T // Enable 64-bit integer types - should be defined before mpfr.h - -#if defined(MPREAL_HAVE_MSVC_DEBUGVIEW) && defined(_MSC_VER) && defined(_DEBUG) - #define MPREAL_MSVC_DEBUGVIEW_CODE DebugView = toString(); - #define MPREAL_MSVC_DEBUGVIEW_DATA std::string DebugView; -#else - #define MPREAL_MSVC_DEBUGVIEW_CODE - #define MPREAL_MSVC_DEBUGVIEW_DATA -#endif - -#include - -#if (MPFR_VERSION < MPFR_VERSION_NUM(3,0,0)) - #include // Needed for random() -#endif - -// Less important options -#define MPREAL_DOUBLE_BITS_OVERFLOW -1 // Triggers overflow exception during conversion to double if mpreal - // cannot fit in MPREAL_DOUBLE_BITS_OVERFLOW bits - // = -1 disables overflow checks (default) - -// Fast replacement for mpfr_set_zero(x, +1): -// (a) uses low-level data members, might not be compatible with new versions of MPFR -// (b) sign is not set, add (x)->_mpfr_sign = 1; -#define mpfr_set_zero_fast(x) ((x)->_mpfr_exp = __MPFR_EXP_ZERO) - -#if defined(__GNUC__) - #define MPREAL_PERMISSIVE_EXPR __extension__ -#else - #define MPREAL_PERMISSIVE_EXPR -#endif - -namespace mpfr { - -class mpreal { -private: - mpfr_t mp; - -public: - - // Get default rounding mode & precision - inline static mp_rnd_t get_default_rnd() { return (mp_rnd_t)(mpfr_get_default_rounding_mode()); } - inline static mp_prec_t get_default_prec() { return mpfr_get_default_prec(); } - - // Constructors && type conversions - mpreal(); - mpreal(const mpreal& u); - mpreal(const mpf_t u); - mpreal(const mpz_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const mpq_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const unsigned int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); - - // Construct mpreal from mpfr_t structure. - // shared = true allows to avoid deep copy, so that mpreal and 'u' share the same data & pointers. - mpreal(const mpfr_t u, bool shared = false); - - mpreal(const char* s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); - mpreal(const std::string& s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); - - ~mpreal(); - -#ifdef MPREAL_HAVE_MOVE_SUPPORT - mpreal& operator=(mpreal&& v); - mpreal(mpreal&& u); -#endif - - // Operations - // = - // +, -, *, /, ++, --, <<, >> - // *=, +=, -=, /=, - // <, >, ==, <=, >= - - // = - mpreal& operator=(const mpreal& v); - mpreal& operator=(const mpf_t v); - mpreal& operator=(const mpz_t v); - mpreal& operator=(const mpq_t v); - mpreal& operator=(const long double v); - mpreal& operator=(const double v); - mpreal& operator=(const unsigned long int v); - mpreal& operator=(const unsigned long long int v); - mpreal& operator=(const long long int v); - mpreal& operator=(const unsigned int v); - mpreal& operator=(const long int v); - mpreal& operator=(const int v); - mpreal& operator=(const char* s); - mpreal& operator=(const std::string& s); - template mpreal& operator= (const std::complex& z); - - // + - mpreal& operator+=(const mpreal& v); - mpreal& operator+=(const mpf_t v); - mpreal& operator+=(const mpz_t v); - mpreal& operator+=(const mpq_t v); - mpreal& operator+=(const long double u); - mpreal& operator+=(const double u); - mpreal& operator+=(const unsigned long int u); - mpreal& operator+=(const unsigned int u); - mpreal& operator+=(const long int u); - mpreal& operator+=(const int u); - - mpreal& operator+=(const long long int u); - mpreal& operator+=(const unsigned long long int u); - mpreal& operator-=(const long long int u); - mpreal& operator-=(const unsigned long long int u); - mpreal& operator*=(const long long int u); - mpreal& operator*=(const unsigned long long int u); - mpreal& operator/=(const long long int u); - mpreal& operator/=(const unsigned long long int u); - - const mpreal operator+() const; - mpreal& operator++ (); - const mpreal operator++ (int); - - // - - mpreal& operator-=(const mpreal& v); - mpreal& operator-=(const mpz_t v); - mpreal& operator-=(const mpq_t v); - mpreal& operator-=(const long double u); - mpreal& operator-=(const double u); - mpreal& operator-=(const unsigned long int u); - mpreal& operator-=(const unsigned int u); - mpreal& operator-=(const long int u); - mpreal& operator-=(const int u); - const mpreal operator-() const; - friend const mpreal operator-(const unsigned long int b, const mpreal& a); - friend const mpreal operator-(const unsigned int b, const mpreal& a); - friend const mpreal operator-(const long int b, const mpreal& a); - friend const mpreal operator-(const int b, const mpreal& a); - friend const mpreal operator-(const double b, const mpreal& a); - mpreal& operator-- (); - const mpreal operator-- (int); - - // * - mpreal& operator*=(const mpreal& v); - mpreal& operator*=(const mpz_t v); - mpreal& operator*=(const mpq_t v); - mpreal& operator*=(const long double v); - mpreal& operator*=(const double v); - mpreal& operator*=(const unsigned long int v); - mpreal& operator*=(const unsigned int v); - mpreal& operator*=(const long int v); - mpreal& operator*=(const int v); - - // / - mpreal& operator/=(const mpreal& v); - mpreal& operator/=(const mpz_t v); - mpreal& operator/=(const mpq_t v); - mpreal& operator/=(const long double v); - mpreal& operator/=(const double v); - mpreal& operator/=(const unsigned long int v); - mpreal& operator/=(const unsigned int v); - mpreal& operator/=(const long int v); - mpreal& operator/=(const int v); - friend const mpreal operator/(const unsigned long int b, const mpreal& a); - friend const mpreal operator/(const unsigned int b, const mpreal& a); - friend const mpreal operator/(const long int b, const mpreal& a); - friend const mpreal operator/(const int b, const mpreal& a); - friend const mpreal operator/(const double b, const mpreal& a); - - //<<= Fast Multiplication by 2^u - mpreal& operator<<=(const unsigned long int u); - mpreal& operator<<=(const unsigned int u); - mpreal& operator<<=(const long int u); - mpreal& operator<<=(const int u); - - //>>= Fast Division by 2^u - mpreal& operator>>=(const unsigned long int u); - mpreal& operator>>=(const unsigned int u); - mpreal& operator>>=(const long int u); - mpreal& operator>>=(const int u); - - // Type Conversion operators - bool toBool ( ) const; - long toLong (mp_rnd_t mode = GMP_RNDZ) const; - unsigned long toULong (mp_rnd_t mode = GMP_RNDZ) const; - long long toLLong (mp_rnd_t mode = GMP_RNDZ) const; - unsigned long long toULLong (mp_rnd_t mode = GMP_RNDZ) const; - float toFloat (mp_rnd_t mode = GMP_RNDN) const; - double toDouble (mp_rnd_t mode = GMP_RNDN) const; - long double toLDouble (mp_rnd_t mode = GMP_RNDN) const; - -#if defined (MPREAL_HAVE_EXPLICIT_CONVERTERS) - explicit operator bool () const { return toBool(); } - explicit operator int () const { return int(toLong()); } - explicit operator long () const { return toLong(); } - explicit operator long long () const { return toLLong(); } - explicit operator unsigned () const { return unsigned(toULong()); } - explicit operator unsigned long () const { return toULong(); } - explicit operator unsigned long long () const { return toULLong(); } - explicit operator float () const { return toFloat(); } - explicit operator double () const { return toDouble(); } - explicit operator long double () const { return toLDouble(); } -#endif - - // Get raw pointers so that mpreal can be directly used in raw mpfr_* functions - ::mpfr_ptr mpfr_ptr(); - ::mpfr_srcptr mpfr_ptr() const; - ::mpfr_srcptr mpfr_srcptr() const; - - // Convert mpreal to string with n significant digits in base b - // n = -1 -> convert with the maximum available digits - std::string toString(int n = -1, int b = 10, mp_rnd_t mode = mpreal::get_default_rnd()) const; - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - std::string toString(const std::string& format) const; -#endif - - std::ostream& output(std::ostream& os) const; - - // Math Functions - friend const mpreal sqr (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sqrt(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sqrt(const unsigned long int v, mp_rnd_t rnd_mode); - friend const mpreal cbrt(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal root(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode); - friend const mpreal pow (const mpreal& a, const long int b, mp_rnd_t rnd_mode); - friend const mpreal pow (const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode); - friend const mpreal pow (const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode); - friend const mpreal fabs(const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal abs(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); - friend inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); - friend inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); - friend inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); - friend int cmpabs(const mpreal& a,const mpreal& b); - - friend const mpreal log (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal logb (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log10(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal exp10(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal log1p(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal expm1(const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal cos(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sin(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tan(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sec(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal csc(const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal cot(const mpreal& v, mp_rnd_t rnd_mode); - friend int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal acos (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asin (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atan (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode); - friend const mpreal acot (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asec (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acsc (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal cosh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sinh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tanh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal sech (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal csch (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal coth (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acosh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asinh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal atanh (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acoth (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal asech (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal acsch (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - - friend const mpreal fac_ui (unsigned long int v, mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal eint (const mpreal& v, mp_rnd_t rnd_mode); - - friend const mpreal gamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal tgamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal lngamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal lgamma (const mpreal& v, int *signp, mp_rnd_t rnd_mode); - friend const mpreal zeta (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal erf (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal erfc (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselj0 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselj1 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besseljn (long n, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal bessely0 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal bessely1 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal besselyn (long n, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); - friend const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); - friend const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode); - friend const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t rnd_mode); - friend int sgn(const mpreal& v); // returns -1 or +1 - -// MPFR 2.4.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - friend int sinh_cosh (mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal li2 (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal rec_sqrt (const mpreal& v, mp_rnd_t rnd_mode); - - // MATLAB's semantic equivalents - friend const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Remainder after division - friend const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Modulus after division -#endif - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - friend const mpreal digamma (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal ai (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear -#endif - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - friend const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear - friend const mpreal grandom (unsigned int seed); -#endif - - // Uniformly distributed random number generation in [0,1] using - // Mersenne-Twister algorithm by default. - // Use parameter to setup seed, e.g.: random((unsigned)time(NULL)) - // Check urandom() for more precise control. - friend const mpreal random(unsigned int seed); - - // Splits mpreal value into fractional and integer parts. - // Returns fractional part and stores integer part in n. - friend const mpreal modf(const mpreal& v, mpreal& n); - - // Constants - // don't forget to call mpfr_free_cache() for every thread where you are using const-functions - friend const mpreal const_log2 (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_pi (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_euler (mp_prec_t prec, mp_rnd_t rnd_mode); - friend const mpreal const_catalan (mp_prec_t prec, mp_rnd_t rnd_mode); - - // returns +inf iff sign>=0 otherwise -inf - friend const mpreal const_infinity(int sign, mp_prec_t prec); - - // Output/ Input - friend std::ostream& operator<<(std::ostream& os, const mpreal& v); - friend std::istream& operator>>(std::istream& is, mpreal& v); - - // Integer Related Functions - friend const mpreal rint (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal ceil (const mpreal& v); - friend const mpreal floor(const mpreal& v); - friend const mpreal round(const mpreal& v); - friend const mpreal trunc(const mpreal& v); - friend const mpreal rint_ceil (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_floor (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_round (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal rint_trunc (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal frac (const mpreal& v, mp_rnd_t rnd_mode); - friend const mpreal remainder ( const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - - // Miscellaneous Functions - friend const mpreal nexttoward (const mpreal& x, const mpreal& y); - friend const mpreal nextabove (const mpreal& x); - friend const mpreal nextbelow (const mpreal& x); - - // use gmp_randinit_default() to init state, gmp_randclear() to clear - friend const mpreal urandomb (gmp_randstate_t& state); - -// MPFR < 2.4.2 Specifics -#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) - friend const mpreal random2 (mp_size_t size, mp_exp_t exp); -#endif - - // Instance Checkers - friend bool (isnan) (const mpreal& v); - friend bool (isinf) (const mpreal& v); - friend bool (isfinite) (const mpreal& v); - - friend bool isnum (const mpreal& v); - friend bool iszero (const mpreal& v); - friend bool isint (const mpreal& v); - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - friend bool isregular(const mpreal& v); -#endif - - // Set/Get instance properties - inline mp_prec_t get_prec() const; - inline void set_prec(mp_prec_t prec, mp_rnd_t rnd_mode = get_default_rnd()); // Change precision with rounding mode - - // Aliases for get_prec(), set_prec() - needed for compatibility with std::complex interface - inline mpreal& setPrecision(int Precision, mp_rnd_t RoundingMode = get_default_rnd()); - inline int getPrecision() const; - - // Set mpreal to +/- inf, NaN, +/-0 - mpreal& setInf (int Sign = +1); - mpreal& setNan (); - mpreal& setZero (int Sign = +1); - mpreal& setSign (int Sign, mp_rnd_t RoundingMode = get_default_rnd()); - - //Exponent - mp_exp_t get_exp(); - int set_exp(mp_exp_t e); - int check_range (int t, mp_rnd_t rnd_mode = get_default_rnd()); - int subnormalize (int t, mp_rnd_t rnd_mode = get_default_rnd()); - - // Inexact conversion from float - inline bool fits_in_bits(double x, int n); - - // Set/Get global properties - static void set_default_prec(mp_prec_t prec); - static void set_default_rnd(mp_rnd_t rnd_mode); - - static mp_exp_t get_emin (void); - static mp_exp_t get_emax (void); - static mp_exp_t get_emin_min (void); - static mp_exp_t get_emin_max (void); - static mp_exp_t get_emax_min (void); - static mp_exp_t get_emax_max (void); - static int set_emin (mp_exp_t exp); - static int set_emax (mp_exp_t exp); - - // Efficient swapping of two mpreal values - needed for std algorithms - friend void swap(mpreal& x, mpreal& y); - - friend const mpreal fmax(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - friend const mpreal fmin(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); - -private: - // Human friendly Debug Preview in Visual Studio. - // Put one of these lines: - // - // mpfr::mpreal= ; Show value only - // mpfr::mpreal=, bits ; Show value & precision - // - // at the beginning of - // [Visual Studio Installation Folder]\Common7\Packages\Debugger\autoexp.dat - MPREAL_MSVC_DEBUGVIEW_DATA - - // "Smart" resources deallocation. Checks if instance initialized before deletion. - void clear(::mpfr_ptr); -}; - -////////////////////////////////////////////////////////////////////////// -// Exceptions -class conversion_overflow : public std::exception { -public: - std::string why() { return "inexact conversion from floating point"; } -}; - -////////////////////////////////////////////////////////////////////////// -// Constructors & converters -// Default constructor: creates mp number and initializes it to 0. -inline mpreal::mpreal() -{ - mpfr_init2(mpfr_ptr(), mpreal::get_default_prec()); - mpfr_set_zero_fast(mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpreal& u) -{ - mpfr_init2(mpfr_ptr(),mpfr_get_prec(u.mpfr_srcptr())); - mpfr_set (mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -#ifdef MPREAL_HAVE_MOVE_SUPPORT -inline mpreal::mpreal(mpreal&& other) -{ - mpfr_set_uninitialized(mpfr_ptr()); // make sure "other" holds no pointer to actual data - mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal& mpreal::operator=(mpreal&& other) -{ - mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} -#endif - -inline mpreal::mpreal(const mpfr_t u, bool shared) -{ - if(shared) - { - std::memcpy(mpfr_ptr(), u, sizeof(mpfr_t)); - } - else - { - mpfr_init2(mpfr_ptr(), mpfr_get_prec(u)); - mpfr_set (mpfr_ptr(), u, mpreal::get_default_rnd()); - } - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpf_t u) -{ - mpfr_init2(mpfr_ptr(),(mp_prec_t) mpf_get_prec(u)); // (gmp: mp_bitcnt_t) unsigned long -> long (mpfr: mp_prec_t) - mpfr_set_f(mpfr_ptr(),u,mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpz_t u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - mpfr_set_z(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const mpq_t u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - mpfr_set_q(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const double u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2(mpfr_ptr(), prec); - -#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) - if(fits_in_bits(u, MPREAL_DOUBLE_BITS_OVERFLOW)) - { - mpfr_set_d(mpfr_ptr(), u, mode); - }else - throw conversion_overflow(); -#else - mpfr_set_d(mpfr_ptr(), u, mode); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long double u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ld(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned long long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_uj(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_sj(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ui(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const unsigned int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_ui(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const long int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_si(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const int u, mp_prec_t prec, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_si(mpfr_ptr(), u, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_str(mpfr_ptr(), s, base, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mpreal::mpreal(const std::string& s, mp_prec_t prec, int base, mp_rnd_t mode) -{ - mpfr_init2 (mpfr_ptr(), prec); - mpfr_set_str(mpfr_ptr(), s.c_str(), base, mode); - - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline void mpreal::clear(::mpfr_ptr x) -{ -#ifdef MPREAL_HAVE_MOVE_SUPPORT - if(mpfr_is_initialized(x)) -#endif - mpfr_clear(x); -} - -inline mpreal::~mpreal() -{ - clear(mpfr_ptr()); -} - -// internal namespace needed for template magic -namespace internal{ - - // Use SFINAE to restrict arithmetic operations instantiation only for numeric types - // This is needed for smooth integration with libraries based on expression templates, like Eigen. - // TODO: Do the same for boolean operators. - template struct result_type {}; - - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; - template <> struct result_type {typedef mpreal type;}; -} - -// + Addition -template -inline const typename internal::result_type::type - operator+(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) += rhs; } - -template -inline const typename internal::result_type::type - operator+(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) += lhs; } - -// - Subtraction -template -inline const typename internal::result_type::type - operator-(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) -= rhs; } - -template -inline const typename internal::result_type::type - operator-(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) -= rhs; } - -// * Multiplication -template -inline const typename internal::result_type::type - operator*(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) *= rhs; } - -template -inline const typename internal::result_type::type - operator*(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) *= lhs; } - -// / Division -template -inline const typename internal::result_type::type - operator/(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) /= rhs; } - -template -inline const typename internal::result_type::type - operator/(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) /= rhs; } - -////////////////////////////////////////////////////////////////////////// -// sqrt -const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const long int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const long double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal sqrt(const double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -// abs -inline const mpreal abs(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()); - -////////////////////////////////////////////////////////////////////////// -// pow -const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); -inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); - -////////////////////////////////////////////////////////////////////////// -// Estimate machine epsilon for the given precision -// Returns smallest eps such that 1.0 + eps != 1.0 -inline mpreal machine_epsilon(mp_prec_t prec = mpreal::get_default_prec()); - -// Returns smallest eps such that x + eps != x (relative machine epsilon) -inline mpreal machine_epsilon(const mpreal& x); - -// Gives max & min values for the required precision, -// minval is 'safe' meaning 1 / minval does not overflow -// maxval is 'safe' meaning 1 / maxval does not underflow -inline mpreal minval(mp_prec_t prec = mpreal::get_default_prec()); -inline mpreal maxval(mp_prec_t prec = mpreal::get_default_prec()); - -// 'Dirty' equality check 1: |a-b| < min{|a|,|b|} * eps -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps); - -// 'Dirty' equality check 2: |a-b| < min{|a|,|b|} * eps( min{|a|,|b|} ) -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b); - -// 'Bitwise' equality check -// maxUlps - a and b can be apart by maxUlps binary numbers. -inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps); - -////////////////////////////////////////////////////////////////////////// -// Convert precision in 'bits' to decimal digits and vice versa. -// bits = ceil(digits*log[2](10)) -// digits = floor(bits*log[10](2)) - -inline mp_prec_t digits2bits(int d); -inline int bits2digits(mp_prec_t b); - -////////////////////////////////////////////////////////////////////////// -// min, max -const mpreal (max)(const mpreal& x, const mpreal& y); -const mpreal (min)(const mpreal& x, const mpreal& y); - -////////////////////////////////////////////////////////////////////////// -// Implementation -////////////////////////////////////////////////////////////////////////// - -////////////////////////////////////////////////////////////////////////// -// Operators - Assignment -inline mpreal& mpreal::operator=(const mpreal& v) -{ - if (this != &v) - { - mp_prec_t tp = mpfr_get_prec( mpfr_srcptr()); - mp_prec_t vp = mpfr_get_prec(v.mpfr_srcptr()); - - if(tp != vp){ - clear(mpfr_ptr()); - mpfr_init2(mpfr_ptr(), vp); - } - - mpfr_set(mpfr_ptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - } - return *this; -} - -inline mpreal& mpreal::operator=(const mpf_t v) -{ - mpfr_set_f(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const mpz_t v) -{ - mpfr_set_z(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const mpq_t v) -{ - mpfr_set_q(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long double v) -{ - mpfr_set_ld(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const double v) -{ -#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) - if(fits_in_bits(v, MPREAL_DOUBLE_BITS_OVERFLOW)) - { - mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); - }else - throw conversion_overflow(); -#else - mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned long int v) -{ - mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned int v) -{ - mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const unsigned long long int v) -{ - mpfr_set_uj(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long long int v) -{ - mpfr_set_sj(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const long int v) -{ - mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const int v) -{ - mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator=(const char* s) -{ - // Use other converters for more precise control on base & precision & rounding: - // - // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) - // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) - // - // Here we assume base = 10 and we use precision of target variable. - - mpfr_t t; - - mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); - - if(0 == mpfr_set_str(t, s, 10, mpreal::get_default_rnd())) - { - mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - } - - clear(t); - return *this; -} - -inline mpreal& mpreal::operator=(const std::string& s) -{ - // Use other converters for more precise control on base & precision & rounding: - // - // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) - // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) - // - // Here we assume base = 10 and we use precision of target variable. - - mpfr_t t; - - mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); - - if(0 == mpfr_set_str(t, s.c_str(), 10, mpreal::get_default_rnd())) - { - mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - } - - clear(t); - return *this; -} - -template -inline mpreal& mpreal::operator= (const std::complex& z) -{ - return *this = z.real(); -} - -////////////////////////////////////////////////////////////////////////// -// + Addition -inline mpreal& mpreal::operator+=(const mpreal& v) -{ - mpfr_add(mpfr_ptr(), mpfr_srcptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpf_t u) -{ - *this += mpreal(u); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpz_t u) -{ - mpfr_add_z(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const mpq_t u) -{ - mpfr_add_q(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+= (const long double u) -{ - *this += mpreal(u); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+= (const double u) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_add_d(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); -#else - *this += mpreal(u); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const unsigned long int u) -{ - mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const unsigned int u) -{ - mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const long int u) -{ - mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const int u) -{ - mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator+=(const long long int u) { *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator+=(const unsigned long long int u){ *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator-=(const long long int u) { *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator-=(const unsigned long long int u){ *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator*=(const long long int u) { *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator*=(const unsigned long long int u){ *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator/=(const long long int u) { *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } -inline mpreal& mpreal::operator/=(const unsigned long long int u){ *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } - -inline const mpreal mpreal::operator+()const { return mpreal(*this); } - -inline const mpreal operator+(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_add(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline mpreal& mpreal::operator++() -{ - return *this += 1; -} - -inline const mpreal mpreal::operator++ (int) -{ - mpreal x(*this); - *this += 1; - return x; -} - -inline mpreal& mpreal::operator--() -{ - return *this -= 1; -} - -inline const mpreal mpreal::operator-- (int) -{ - mpreal x(*this); - *this -= 1; - return x; -} - -////////////////////////////////////////////////////////////////////////// -// - Subtraction -inline mpreal& mpreal::operator-=(const mpreal& v) -{ - mpfr_sub(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const mpz_t v) -{ - mpfr_sub_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const mpq_t v) -{ - mpfr_sub_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const long double v) -{ - *this -= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_sub_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this -= mpreal(v); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const unsigned long int v) -{ - mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const unsigned int v) -{ - mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const long int v) -{ - mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator-=(const int v) -{ - mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal mpreal::operator-()const -{ - mpreal u(*this); - mpfr_neg(u.mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); - return u; -} - -inline const mpreal operator-(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_sub(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline const mpreal operator-(const double b, const mpreal& a) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_d_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -#else - mpreal x(b, mpfr_get_prec(a.mpfr_ptr())); - x -= a; - return x; -#endif -} - -inline const mpreal operator-(const unsigned long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const unsigned int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator-(const int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -////////////////////////////////////////////////////////////////////////// -// * Multiplication -inline mpreal& mpreal::operator*= (const mpreal& v) -{ - mpfr_mul(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const mpz_t v) -{ - mpfr_mul_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const mpq_t v) -{ - mpfr_mul_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const long double v) -{ - *this *= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_mul_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this *= mpreal(v); -#endif - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const unsigned long int v) -{ - mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const unsigned int v) -{ - mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const long int v) -{ - mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator*=(const int v) -{ - mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator*(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); - mpfr_mul(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -////////////////////////////////////////////////////////////////////////// -// / Division -inline mpreal& mpreal::operator/=(const mpreal& v) -{ - mpfr_div(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const mpz_t v) -{ - mpfr_div_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const mpq_t v) -{ - mpfr_div_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const long double v) -{ - *this /= mpreal(v); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const double v) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpfr_div_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); -#else - *this /= mpreal(v); -#endif - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const unsigned long int v) -{ - mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const unsigned int v) -{ - mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const long int v) -{ - mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator/=(const int v) -{ - mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator/(const mpreal& a, const mpreal& b) -{ - mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_srcptr()), mpfr_get_prec(b.mpfr_srcptr()))); - mpfr_div(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); - return c; -} - -inline const mpreal operator/(const unsigned long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const unsigned int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const long int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const int b, const mpreal& a) -{ - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -} - -inline const mpreal operator/(const double b, const mpreal& a) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_d_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); - return x; -#else - mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); - x /= a; - return x; -#endif -} - -////////////////////////////////////////////////////////////////////////// -// Shifts operators - Multiplication/Division by power of 2 -inline mpreal& mpreal::operator<<=(const unsigned long int u) -{ - mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const unsigned int u) -{ - mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const long int u) -{ - mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator<<=(const int u) -{ - mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const unsigned long int u) -{ - mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const unsigned int u) -{ - mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const long int u) -{ - mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::operator>>=(const int u) -{ - mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline const mpreal operator<<(const mpreal& v, const unsigned long int k) -{ - return mul_2ui(v,k); -} - -inline const mpreal operator<<(const mpreal& v, const unsigned int k) -{ - return mul_2ui(v,static_cast(k)); -} - -inline const mpreal operator<<(const mpreal& v, const long int k) -{ - return mul_2si(v,k); -} - -inline const mpreal operator<<(const mpreal& v, const int k) -{ - return mul_2si(v,static_cast(k)); -} - -inline const mpreal operator>>(const mpreal& v, const unsigned long int k) -{ - return div_2ui(v,k); -} - -inline const mpreal operator>>(const mpreal& v, const long int k) -{ - return div_2si(v,k); -} - -inline const mpreal operator>>(const mpreal& v, const unsigned int k) -{ - return div_2ui(v,static_cast(k)); -} - -inline const mpreal operator>>(const mpreal& v, const int k) -{ - return div_2si(v,static_cast(k)); -} - -// mul_2ui -inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_mul_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -// mul_2si -inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_mul_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_div_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) -{ - mpreal x(v); - mpfr_div_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); - return x; -} - -////////////////////////////////////////////////////////////////////////// -//Relational operators - -// WARNING: -// -// Please note that following checks for double-NaN are guaranteed to work only in IEEE math mode: -// -// isnan(b) = (b != b) -// isnan(b) = !(b == b) (we use in code below) -// -// Be cautions if you use compiler options which break strict IEEE compliance (e.g. -ffast-math in GCC). -// Use std::isnan instead (C++11). - -inline bool operator > (const mpreal& a, const mpreal& b ){ return (mpfr_greater_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator > (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) > 0 ); } -inline bool operator > (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) > 0 ); } - -inline bool operator >= (const mpreal& a, const mpreal& b ){ return (mpfr_greaterequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator >= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } -// inline bool operator >= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (isnan()a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) >= 0 ); } -inline bool operator >= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) >= 0 ); } - -inline bool operator < (const mpreal& a, const mpreal& b ){ return (mpfr_less_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator < (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) < 0 ); } -inline bool operator < (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) < 0 ); } - -inline bool operator <= (const mpreal& a, const mpreal& b ){ return (mpfr_lessequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator <= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) <= 0 ); } -inline bool operator <= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) <= 0 ); } - -inline bool operator == (const mpreal& a, const mpreal& b ){ return (mpfr_equal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } -inline bool operator == (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) == 0 ); } -inline bool operator == (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) == 0 ); } - -inline bool operator != (const mpreal& a, const mpreal& b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const unsigned long int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const unsigned int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const long int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const int b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const long double b ){ return !(a == b); } -inline bool operator != (const mpreal& a, const double b ){ return !(a == b); } - -inline bool (isnan) (const mpreal& op){ return (mpfr_nan_p (op.mpfr_srcptr()) != 0 ); } -inline bool (isinf) (const mpreal& op){ return (mpfr_inf_p (op.mpfr_srcptr()) != 0 ); } -inline bool (isfinite) (const mpreal& op){ return (mpfr_number_p (op.mpfr_srcptr()) != 0 ); } -inline bool iszero (const mpreal& op){ return (mpfr_zero_p (op.mpfr_srcptr()) != 0 ); } -inline bool isint (const mpreal& op){ return (mpfr_integer_p(op.mpfr_srcptr()) != 0 ); } - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) -inline bool isregular(const mpreal& op){ return (mpfr_regular_p(op.mpfr_srcptr()));} -#endif - -////////////////////////////////////////////////////////////////////////// -// Type Converters -inline bool mpreal::toBool ( ) const { return mpfr_zero_p (mpfr_srcptr()) == 0; } -inline long mpreal::toLong (mp_rnd_t mode) const { return mpfr_get_si (mpfr_srcptr(), mode); } -inline unsigned long mpreal::toULong (mp_rnd_t mode) const { return mpfr_get_ui (mpfr_srcptr(), mode); } -inline float mpreal::toFloat (mp_rnd_t mode) const { return mpfr_get_flt(mpfr_srcptr(), mode); } -inline double mpreal::toDouble (mp_rnd_t mode) const { return mpfr_get_d (mpfr_srcptr(), mode); } -inline long double mpreal::toLDouble(mp_rnd_t mode) const { return mpfr_get_ld (mpfr_srcptr(), mode); } -inline long long mpreal::toLLong (mp_rnd_t mode) const { return mpfr_get_sj (mpfr_srcptr(), mode); } -inline unsigned long long mpreal::toULLong (mp_rnd_t mode) const { return mpfr_get_uj (mpfr_srcptr(), mode); } - -inline ::mpfr_ptr mpreal::mpfr_ptr() { return mp; } -inline ::mpfr_srcptr mpreal::mpfr_ptr() const { return mp; } -inline ::mpfr_srcptr mpreal::mpfr_srcptr() const { return mp; } - -template -inline std::string toString(T t, std::ios_base & (*f)(std::ios_base&)) -{ - std::ostringstream oss; - oss << f << t; - return oss.str(); -} - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - -inline std::string mpreal::toString(const std::string& format) const -{ - char *s = NULL; - std::string out; - - if( !format.empty() ) - { - if(!(mpfr_asprintf(&s, format.c_str(), mpfr_srcptr()) < 0)) - { - out = std::string(s); - - mpfr_free_str(s); - } - } - - return out; -} - -#endif - -inline std::string mpreal::toString(int n, int b, mp_rnd_t mode) const -{ - // TODO: Add extended format specification (f, e, rounding mode) as it done in output operator - (void)b; - (void)mode; - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - - std::ostringstream format; - - int digits = (n >= 0) ? n : 1 + bits2digits(mpfr_get_prec(mpfr_srcptr())); - - format << "%." << digits << "RNg"; - - return toString(format.str()); - -#else - - char *s, *ns = NULL; - size_t slen, nslen; - mp_exp_t exp; - std::string out; - - if(mpfr_inf_p(mp)) - { - if(mpfr_sgn(mp)>0) return "+Inf"; - else return "-Inf"; - } - - if(mpfr_zero_p(mp)) return "0"; - if(mpfr_nan_p(mp)) return "NaN"; - - s = mpfr_get_str(NULL, &exp, b, 0, mp, mode); - ns = mpfr_get_str(NULL, &exp, b, (std::max)(0,n), mp, mode); - - if(s!=NULL && ns!=NULL) - { - slen = strlen(s); - nslen = strlen(ns); - if(nslen<=slen) - { - mpfr_free_str(s); - s = ns; - slen = nslen; - } - else { - mpfr_free_str(ns); - } - - // Make human eye-friendly formatting if possible - if (exp>0 && static_cast(exp)s+exp) ptr--; - - if(ptr==s+exp) out = std::string(s,exp+1); - else out = std::string(s,exp+1)+'.'+std::string(s+exp+1,ptr-(s+exp+1)+1); - - //out = string(s,exp+1)+'.'+string(s+exp+1); - } - else - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s+exp-1) ptr--; - - if(ptr==s+exp-1) out = std::string(s,exp); - else out = std::string(s,exp)+'.'+std::string(s+exp,ptr-(s+exp)+1); - - //out = string(s,exp)+'.'+string(s+exp); - } - - }else{ // exp<0 || exp>slen - if(s[0]=='-') - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s+1) ptr--; - - if(ptr==s+1) out = std::string(s,2); - else out = std::string(s,2)+'.'+std::string(s+2,ptr-(s+2)+1); - - //out = string(s,2)+'.'+string(s+2); - } - else - { - // Remove zeros starting from right end - char* ptr = s+slen-1; - while (*ptr=='0' && ptr>s) ptr--; - - if(ptr==s) out = std::string(s,1); - else out = std::string(s,1)+'.'+std::string(s+1,ptr-(s+1)+1); - - //out = string(s,1)+'.'+string(s+1); - } - - // Make final string - if(--exp) - { - if(exp>0) out += "e+"+mpfr::toString(exp,std::dec); - else out += "e"+mpfr::toString(exp,std::dec); - } - } - - mpfr_free_str(s); - return out; - }else{ - return "conversion error!"; - } -#endif -} - - -////////////////////////////////////////////////////////////////////////// -// I/O -inline std::ostream& mpreal::output(std::ostream& os) const -{ - std::ostringstream format; - const std::ios::fmtflags flags = os.flags(); - - format << ((flags & std::ios::showpos) ? "%+" : "%"); - if (os.precision() >= 0) - format << '.' << os.precision() << "R*" - << ((flags & std::ios::floatfield) == std::ios::fixed ? 'f' : - (flags & std::ios::floatfield) == std::ios::scientific ? 'e' : - 'g'); - else - format << "R*e"; - - char *s = NULL; - if(!(mpfr_asprintf(&s, format.str().c_str(), - mpfr::mpreal::get_default_rnd(), - mpfr_srcptr()) - < 0)) - { - os << std::string(s); - mpfr_free_str(s); - } - return os; -} - -inline std::ostream& operator<<(std::ostream& os, const mpreal& v) -{ - return v.output(os); -} - -inline std::istream& operator>>(std::istream &is, mpreal& v) -{ - // TODO: use cout::hexfloat and other flags to setup base - std::string tmp; - is >> tmp; - mpfr_set_str(v.mpfr_ptr(), tmp.c_str(), 10, mpreal::get_default_rnd()); - return is; -} - -////////////////////////////////////////////////////////////////////////// -// Bits - decimal digits relation -// bits = ceil(digits*log[2](10)) -// digits = floor(bits*log[10](2)) - -inline mp_prec_t digits2bits(int d) -{ - const double LOG2_10 = 3.3219280948873624; - - return mp_prec_t(std::ceil( d * LOG2_10 )); -} - -inline int bits2digits(mp_prec_t b) -{ - const double LOG10_2 = 0.30102999566398119; - - return int(std::floor( b * LOG10_2 )); -} - -////////////////////////////////////////////////////////////////////////// -// Set/Get number properties -inline int sgn(const mpreal& op) -{ - return mpfr_sgn(op.mpfr_srcptr()); -} - -inline mpreal& mpreal::setSign(int sign, mp_rnd_t RoundingMode) -{ - mpfr_setsign(mpfr_ptr(), mpfr_srcptr(), (sign < 0 ? 1 : 0), RoundingMode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline int mpreal::getPrecision() const -{ - return int(mpfr_get_prec(mpfr_srcptr())); -} - -inline mpreal& mpreal::setPrecision(int Precision, mp_rnd_t RoundingMode) -{ - mpfr_prec_round(mpfr_ptr(), Precision, RoundingMode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setInf(int sign) -{ - mpfr_set_inf(mpfr_ptr(), sign); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setNan() -{ - mpfr_set_nan(mpfr_ptr()); - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mpreal& mpreal::setZero(int sign) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - mpfr_set_zero(mpfr_ptr(), sign); -#else - mpfr_set_si(mpfr_ptr(), 0, (mpfr_get_default_rounding_mode)()); - setSign(sign); -#endif - - MPREAL_MSVC_DEBUGVIEW_CODE; - return *this; -} - -inline mp_prec_t mpreal::get_prec() const -{ - return mpfr_get_prec(mpfr_srcptr()); -} - -inline void mpreal::set_prec(mp_prec_t prec, mp_rnd_t rnd_mode) -{ - mpfr_prec_round(mpfr_ptr(),prec,rnd_mode); - MPREAL_MSVC_DEBUGVIEW_CODE; -} - -inline mp_exp_t mpreal::get_exp () -{ - return mpfr_get_exp(mpfr_srcptr()); -} - -inline int mpreal::set_exp (mp_exp_t e) -{ - int x = mpfr_set_exp(mpfr_ptr(), e); - MPREAL_MSVC_DEBUGVIEW_CODE; - return x; -} - -inline const mpreal frexp(const mpreal& x, mp_exp_t* exp, mp_rnd_t mode = mpreal::get_default_rnd()) -{ - mpreal y(x); -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - mpfr_frexp(exp,y.mpfr_ptr(),x.mpfr_srcptr(),mode); -#else - *exp = mpfr_get_exp(y.mpfr_srcptr()); - mpfr_set_exp(y.mpfr_ptr(),0); -#endif - return y; -} - -inline const mpreal ldexp(const mpreal& v, mp_exp_t exp) -{ - mpreal x(v); - - // rounding is not important since we are just increasing the exponent (= exact operation) - mpfr_mul_2si(x.mpfr_ptr(), x.mpfr_srcptr(), exp, mpreal::get_default_rnd()); - return x; -} - -inline const mpreal scalbn(const mpreal& v, mp_exp_t exp) -{ - return ldexp(v, exp); -} - -inline mpreal machine_epsilon(mp_prec_t prec) -{ - /* the smallest eps such that 1 + eps != 1 */ - return machine_epsilon(mpreal(1, prec)); -} - -inline mpreal machine_epsilon(const mpreal& x) -{ - /* the smallest eps such that x + eps != x */ - if( x < 0) - { - return nextabove(-x) + x; - }else{ - return nextabove( x) - x; - } -} - -// minval is 'safe' meaning 1 / minval does not overflow -inline mpreal minval(mp_prec_t prec) -{ - /* min = 1/2 * 2^emin = 2^(emin - 1) */ - return mpreal(1, prec) << mpreal::get_emin()-1; -} - -// maxval is 'safe' meaning 1 / maxval does not underflow -inline mpreal maxval(mp_prec_t prec) -{ - /* max = (1 - eps) * 2^emax, eps is machine epsilon */ - return (mpreal(1, prec) - machine_epsilon(prec)) << mpreal::get_emax(); -} - -inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps) -{ - return abs(a - b) <= machine_epsilon((max)(abs(a), abs(b))) * maxUlps; -} - -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps) -{ - return abs(a - b) <= eps; -} - -inline bool isEqualFuzzy(const mpreal& a, const mpreal& b) -{ - return isEqualFuzzy(a, b, machine_epsilon((max)(1, (min)(abs(a), abs(b))))); -} - -////////////////////////////////////////////////////////////////////////// -// C++11 sign functions. -inline mpreal copysign(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal rop(0, mpfr_get_prec(x.mpfr_ptr())); - mpfr_setsign(rop.mpfr_ptr(), x.mpfr_srcptr(), mpfr_signbit(y.mpfr_srcptr()), rnd_mode); - return rop; -} - -inline bool signbit(const mpreal& x) -{ - return mpfr_signbit(x.mpfr_srcptr()); -} - -inline const mpreal modf(const mpreal& v, mpreal& n) -{ - mpreal f(v); - - // rounding is not important since we are using the same number - mpfr_frac (f.mpfr_ptr(),f.mpfr_srcptr(),mpreal::get_default_rnd()); - mpfr_trunc(n.mpfr_ptr(),v.mpfr_srcptr()); - return f; -} - -inline int mpreal::check_range (int t, mp_rnd_t rnd_mode) -{ - return mpfr_check_range(mpfr_ptr(),t,rnd_mode); -} - -inline int mpreal::subnormalize (int t,mp_rnd_t rnd_mode) -{ - int r = mpfr_subnormalize(mpfr_ptr(),t,rnd_mode); - MPREAL_MSVC_DEBUGVIEW_CODE; - return r; -} - -inline mp_exp_t mpreal::get_emin (void) -{ - return mpfr_get_emin(); -} - -inline int mpreal::set_emin (mp_exp_t exp) -{ - return mpfr_set_emin(exp); -} - -inline mp_exp_t mpreal::get_emax (void) -{ - return mpfr_get_emax(); -} - -inline int mpreal::set_emax (mp_exp_t exp) -{ - return mpfr_set_emax(exp); -} - -inline mp_exp_t mpreal::get_emin_min (void) -{ - return mpfr_get_emin_min(); -} - -inline mp_exp_t mpreal::get_emin_max (void) -{ - return mpfr_get_emin_max(); -} - -inline mp_exp_t mpreal::get_emax_min (void) -{ - return mpfr_get_emax_min(); -} - -inline mp_exp_t mpreal::get_emax_max (void) -{ - return mpfr_get_emax_max(); -} - -////////////////////////////////////////////////////////////////////////// -// Mathematical Functions -////////////////////////////////////////////////////////////////////////// -#define MPREAL_UNARY_MATH_FUNCTION_BODY(f) \ - mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); \ - mpfr_##f(y.mpfr_ptr(), x.mpfr_srcptr(), r); \ - return y; - -inline const mpreal sqr (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqr ); } - -inline const mpreal sqrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqrt); } - -inline const mpreal sqrt(const unsigned long int x, mp_rnd_t r) -{ - mpreal y; - mpfr_sqrt_ui(y.mpfr_ptr(), x, r); - return y; -} - -inline const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode) -{ - return sqrt(static_cast(v),rnd_mode); -} - -inline const mpreal sqrt(const long int v, mp_rnd_t rnd_mode) -{ - if (v>=0) return sqrt(static_cast(v),rnd_mode); - else return mpreal().setNan(); // NaN -} - -inline const mpreal sqrt(const int v, mp_rnd_t rnd_mode) -{ - if (v>=0) return sqrt(static_cast(v),rnd_mode); - else return mpreal().setNan(); // NaN -} - -inline const mpreal root(const mpreal& x, unsigned long int k, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); - mpfr_root(y.mpfr_ptr(), x.mpfr_srcptr(), k, r); - return y; -} - -inline const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, mpfr_get_prec(a.mpfr_srcptr())); - mpfr_dim(y.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), r); - return y; -} - -inline int cmpabs(const mpreal& a,const mpreal& b) -{ - return mpfr_cmpabs(a.mpfr_ptr(), b.mpfr_srcptr()); -} - -inline int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - return mpfr_sin_cos(s.mpfr_ptr(), c.mpfr_ptr(), v.mpfr_srcptr(), rnd_mode); -} - -inline const mpreal sqrt (const long double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } -inline const mpreal sqrt (const double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } - -inline const mpreal cbrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cbrt ); } -inline const mpreal fabs (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } -inline const mpreal abs (const mpreal& x, mp_rnd_t r) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } -inline const mpreal log (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log ); } -inline const mpreal log2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log2 ); } -inline const mpreal log10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log10); } -inline const mpreal exp (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp ); } -inline const mpreal exp2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp2 ); } -inline const mpreal exp10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp10); } -inline const mpreal cos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cos ); } -inline const mpreal sin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sin ); } -inline const mpreal tan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tan ); } -inline const mpreal sec (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sec ); } -inline const mpreal csc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csc ); } -inline const mpreal cot (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cot ); } -inline const mpreal acos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acos ); } -inline const mpreal asin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asin ); } -inline const mpreal atan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atan ); } - -inline const mpreal logb (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { return log2 (abs(x),r); } - -inline const mpreal acot (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atan (1/v, r); } -inline const mpreal asec (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acos (1/v, r); } -inline const mpreal acsc (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asin (1/v, r); } -inline const mpreal acoth (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atanh(1/v, r); } -inline const mpreal asech (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acosh(1/v, r); } -inline const mpreal acsch (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asinh(1/v, r); } - -inline const mpreal cosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cosh ); } -inline const mpreal sinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sinh ); } -inline const mpreal tanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tanh ); } -inline const mpreal sech (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sech ); } -inline const mpreal csch (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csch ); } -inline const mpreal coth (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(coth ); } -inline const mpreal acosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acosh); } -inline const mpreal asinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asinh); } -inline const mpreal atanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atanh); } - -inline const mpreal log1p (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log1p ); } -inline const mpreal expm1 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(expm1 ); } -inline const mpreal eint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(eint ); } -inline const mpreal gamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } -inline const mpreal tgamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } -inline const mpreal lngamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(lngamma); } -inline const mpreal zeta (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(zeta ); } -inline const mpreal erf (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erf ); } -inline const mpreal erfc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erfc ); } -inline const mpreal besselj0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j0 ); } -inline const mpreal besselj1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j1 ); } -inline const mpreal bessely0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y0 ); } -inline const mpreal bessely1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y1 ); } - -inline const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_atan2(a.mpfr_ptr(), y.mpfr_srcptr(), x.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_hypot(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal remainder (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_remainder(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); - mpfr_remquo(a.mpfr_ptr(),q, x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); - return a; -} - -inline const mpreal fac_ui (unsigned long int v, mp_prec_t prec = mpreal::get_default_prec(), - mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(0, prec); - mpfr_fac_ui(x.mpfr_ptr(),v,rnd_mode); - return x; -} - - -inline const mpreal lgamma (const mpreal& v, int *signp = 0, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(v); - int tsignp; - - if(signp) mpfr_lgamma(x.mpfr_ptr(), signp,v.mpfr_srcptr(),rnd_mode); - else mpfr_lgamma(x.mpfr_ptr(),&tsignp,v.mpfr_srcptr(),rnd_mode); - - return x; -} - - -inline const mpreal besseljn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, x.getPrecision()); - mpfr_jn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); - return y; -} - -inline const mpreal besselyn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal y(0, x.getPrecision()); - mpfr_yn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); - return y; -} - -inline const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2, p3; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - p3 = v3.get_prec(); - - a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); - - mpfr_fma(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); - return a; -} - -inline const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2, p3; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - p3 = v3.get_prec(); - - a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); - - mpfr_fms(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); - return a; -} - -inline const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t p1, p2; - - p1 = v1.get_prec(); - p2 = v2.get_prec(); - - a.set_prec(p1>p2?p1:p2); - - mpfr_agm(a.mp, v1.mp, v2.mp, rnd_mode); - - return a; -} - -inline const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t mode = mpreal::get_default_rnd()) -{ - mpfr_srcptr *p = new mpfr_srcptr[n]; - - for (unsigned long int i = 0; i < n; i++) - p[i] = tab[i].mpfr_srcptr(); - - mpreal x; - status = mpfr_sum(x.mpfr_ptr(), (mpfr_ptr*)p, n, mode); - - delete [] p; - return x; -} - -////////////////////////////////////////////////////////////////////////// -// MPFR 2.4.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) - -inline int sinh_cosh(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - return mpfr_sinh_cosh(s.mp,c.mp,v.mp,rnd_mode); -} - -inline const mpreal li2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) -{ - MPREAL_UNARY_MATH_FUNCTION_BODY(li2); -} - -inline const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - /* R = rem(X,Y) if Y != 0, returns X - n * Y where n = trunc(X/Y). */ - return fmod(x, y, rnd_mode); -} - -inline const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - (void)rnd_mode; - - /* - - m = mod(x,y) if y != 0, returns x - n*y where n = floor(x/y) - - The following are true by convention: - - mod(x,0) is x - - mod(x,x) is 0 - - mod(x,y) for x != y and y != 0 has the same sign as y. - - */ - - if(iszero(y)) return x; - if(x == y) return 0; - - mpreal m = x - floor(x / y) * y; - - m.setSign(sgn(y)); // make sure result has the same sign as Y - - return m; -} - -inline const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal a; - mp_prec_t yp, xp; - - yp = y.get_prec(); - xp = x.get_prec(); - - a.set_prec(yp>xp?yp:xp); - - mpfr_fmod(a.mp, x.mp, y.mp, rnd_mode); - - return a; -} - -inline const mpreal rec_sqrt(const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(v); - mpfr_rec_sqrt(x.mp,v.mp,rnd_mode); - return x; -} -#endif // MPFR 2.4.0 Specifics - -////////////////////////////////////////////////////////////////////////// -// MPFR 3.0.0 Specifics -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) -inline const mpreal digamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(digamma); } -inline const mpreal ai (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(ai); } -#endif // MPFR 3.0.0 Specifics - -////////////////////////////////////////////////////////////////////////// -// Constants -inline const mpreal const_log2 (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_log2(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_pi (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_pi(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_euler (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_euler(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_catalan (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) -{ - mpreal x(0, p); - mpfr_const_catalan(x.mpfr_ptr(), r); - return x; -} - -inline const mpreal const_infinity (int sign = 1, mp_prec_t p = mpreal::get_default_prec()) -{ - mpreal x(0, p); - mpfr_set_inf(x.mpfr_ptr(), sign); - return x; -} - -////////////////////////////////////////////////////////////////////////// -// Integer Related Functions -inline const mpreal ceil(const mpreal& v) -{ - mpreal x(v); - mpfr_ceil(x.mp,v.mp); - return x; -} - -inline const mpreal floor(const mpreal& v) -{ - mpreal x(v); - mpfr_floor(x.mp,v.mp); - return x; -} - -inline const mpreal round(const mpreal& v) -{ - mpreal x(v); - mpfr_round(x.mp,v.mp); - return x; -} - -inline const mpreal trunc(const mpreal& v) -{ - mpreal x(v); - mpfr_trunc(x.mp,v.mp); - return x; -} - -inline const mpreal rint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint ); } -inline const mpreal rint_ceil (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_ceil ); } -inline const mpreal rint_floor (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_floor); } -inline const mpreal rint_round (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_round); } -inline const mpreal rint_trunc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_trunc); } -inline const mpreal frac (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(frac ); } - -////////////////////////////////////////////////////////////////////////// -// Miscellaneous Functions -inline void swap (mpreal& a, mpreal& b) { mpfr_swap(a.mp,b.mp); } -inline const mpreal (max)(const mpreal& x, const mpreal& y){ return (x>y?x:y); } -inline const mpreal (min)(const mpreal& x, const mpreal& y){ return (x= MPFR_VERSION_NUM(3,0,0)) -inline const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x; - mpfr_urandom(x.mpfr_ptr(), state, rnd_mode); - return x; -} -#endif - -#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) -inline const mpreal random2 (mp_size_t size, mp_exp_t exp) -{ - mpreal x; - mpfr_random2(x.mpfr_ptr(),size,exp); - return x; -} -#endif - -// Uniformly distributed random number generation -// a = random(seed); <- initialization & first random number generation -// a = random(); <- next random numbers generation -// seed != 0 -inline const mpreal random(unsigned int seed = 0) -{ -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) - static gmp_randstate_t state; - static bool initialize = true; - - if(initialize) - { - gmp_randinit_default(state); - gmp_randseed_ui(state,0); - initialize = false; - } - - if(seed != 0) gmp_randseed_ui(state,seed); - - return mpfr::urandom(state); -#else - if(seed != 0) std::srand(seed); - return mpfr::mpreal(std::rand()/(double)RAND_MAX); -#endif - -} - -#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) - -inline const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x; - mpfr_grandom(x.mpfr_ptr(), NULL, state, rnd_mode); - return x; -} - -inline const mpreal grandom(unsigned int seed = 0) -{ - static gmp_randstate_t state; - static bool initialize = true; - - if(initialize) - { - gmp_randinit_default(state); - gmp_randseed_ui(state,0); - initialize = false; - } - - if(seed != 0) gmp_randseed_ui(state,seed); - - return mpfr::grandom(state); -} -#endif - -////////////////////////////////////////////////////////////////////////// -// Set/Get global properties -inline void mpreal::set_default_prec(mp_prec_t prec) -{ - mpfr_set_default_prec(prec); -} - -inline void mpreal::set_default_rnd(mp_rnd_t rnd_mode) -{ - mpfr_set_default_rounding_mode(rnd_mode); -} - -inline bool mpreal::fits_in_bits(double x, int n) -{ - int i; - double t; - return IsInf(x) || (std::modf ( std::ldexp ( std::frexp ( x, &i ), n ), &t ) == 0.0); -} - -inline const mpreal pow(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow(x.mp,x.mp,b.mp,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_z(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_ui(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_pow_si(x.mp,x.mp,b,rnd_mode); - return x; -} - -inline const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); -} - -inline const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); -} - -inline const mpreal pow(const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) -{ - mpreal x(a); - mpfr_ui_pow(x.mp,a,b.mp,rnd_mode); - return x; -} - -inline const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),b,rnd_mode); -} - -inline const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),b,rnd_mode); - else return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),b,rnd_mode); - else return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); -} - -inline const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); -} - -// pow unsigned long int -inline const mpreal pow(const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - mpreal x(a); - mpfr_ui_pow_ui(x.mp,a,b,rnd_mode); - return x; -} - -inline const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode) -{ - return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow -} - -// pow unsigned int -inline const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui -} - -inline const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode) -{ - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -inline const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode) -{ - return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow -} - -// pow long int -inline const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -inline const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -// pow int -inline const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode) -{ - if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode) -{ - if (a>0) - { - if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui - else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - }else{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si - } -} - -inline const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -inline const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode) -{ - if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow - else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow -} - -// pow long double -inline const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),mpreal(b),rnd_mode); -} - -inline const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui -} - -inline const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),mpreal(b),rnd_mode); -} - -inline const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_ui -} - -inline const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_ui -} - -inline const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si -} - -inline const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode) -{ - return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si -} -} // End of mpfr namespace - -// Explicit specialization of std::swap for mpreal numbers -// Thus standard algorithms will use efficient version of swap (due to Koenig lookup) -// Non-throwing swap C++ idiom: http://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Non-throwing_swap -namespace std -{ - // we are allowed to extend namespace std with specializations only - template <> - inline void swap(mpfr::mpreal& x, mpfr::mpreal& y) - { - return mpfr::swap(x, y); - } - - template<> - class numeric_limits - { - public: - static const bool is_specialized = true; - static const bool is_signed = true; - static const bool is_integer = false; - static const bool is_exact = false; - static const int radix = 2; - - static const bool has_infinity = true; - static const bool has_quiet_NaN = true; - static const bool has_signaling_NaN = true; - - static const bool is_iec559 = true; // = IEEE 754 - static const bool is_bounded = true; - static const bool is_modulo = false; - static const bool traps = true; - static const bool tinyness_before = true; - - static const float_denorm_style has_denorm = denorm_absent; - - inline static mpfr::mpreal (min) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::minval(precision); } - inline static mpfr::mpreal (max) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::maxval(precision); } - inline static mpfr::mpreal lowest (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return -mpfr::maxval(precision); } - - // Returns smallest eps such that 1 + eps != 1 (classic machine epsilon) - inline static mpfr::mpreal epsilon(mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::machine_epsilon(precision); } - - // Returns smallest eps such that x + eps != x (relative machine epsilon) - inline static mpfr::mpreal epsilon(const mpfr::mpreal& x) { return mpfr::machine_epsilon(x); } - - inline static mpfr::mpreal round_error(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - mp_rnd_t r = mpfr::mpreal::get_default_rnd(); - - if(r == GMP_RNDN) return mpfr::mpreal(0.5, precision); - else return mpfr::mpreal(1.0, precision); - } - - inline static const mpfr::mpreal infinity() { return mpfr::const_infinity(); } - inline static const mpfr::mpreal quiet_NaN() { return mpfr::mpreal().setNan(); } - inline static const mpfr::mpreal signaling_NaN() { return mpfr::mpreal().setNan(); } - inline static const mpfr::mpreal denorm_min() { return (min)(); } - - // Please note, exponent range is not fixed in MPFR - static const int min_exponent = MPFR_EMIN_DEFAULT; - static const int max_exponent = MPFR_EMAX_DEFAULT; - MPREAL_PERMISSIVE_EXPR static const int min_exponent10 = (int) (MPFR_EMIN_DEFAULT * 0.3010299956639811); - MPREAL_PERMISSIVE_EXPR static const int max_exponent10 = (int) (MPFR_EMAX_DEFAULT * 0.3010299956639811); - -#ifdef MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS - - // Following members should be constant according to standard, but they can be variable in MPFR - // So we define them as functions here. - // - // This is preferable way for std::numeric_limits specialization. - // But it is incompatible with standard std::numeric_limits and might not work with other libraries, e.g. boost. - // See below for compatible implementation. - inline static float_round_style round_style() - { - mp_rnd_t r = mpfr::mpreal::get_default_rnd(); - - switch (r) - { - case GMP_RNDN: return round_to_nearest; - case GMP_RNDZ: return round_toward_zero; - case GMP_RNDU: return round_toward_infinity; - case GMP_RNDD: return round_toward_neg_infinity; - default: return round_indeterminate; - } - } - - inline static int digits() { return int(mpfr::mpreal::get_default_prec()); } - inline static int digits(const mpfr::mpreal& x) { return x.getPrecision(); } - - inline static int digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - return mpfr::bits2digits(precision); - } - - inline static int digits10(const mpfr::mpreal& x) - { - return mpfr::bits2digits(x.getPrecision()); - } - - inline static int max_digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) - { - return digits10(precision); - } -#else - // Digits and round_style are NOT constants when it comes to mpreal. - // If possible, please use functions digits() and round_style() defined above. - // - // These (default) values are preserved for compatibility with existing libraries, e.g. boost. - // Change them accordingly to your application. - // - // For example, if you use 256 bits of precision uniformly in your program, then: - // digits = 256 - // digits10 = 77 - // max_digits10 = 78 - // - // Approximate formula for decimal digits is: digits10 = floor(log10(2) * digits). See bits2digits() for more details. - - static const std::float_round_style round_style = round_to_nearest; - static const int digits = 53; - static const int digits10 = 15; - static const int max_digits10 = 16; -#endif - }; - -} - -#endif /* __MPREAL_H__ */ +/* + MPFR C++: Multi-precision floating point number class for C++. + Based on MPFR library: http://mpfr.org + + Project homepage: http://www.holoborodko.com/pavel/mpfr + Contact e-mail: pavel@holoborodko.com + + Copyright (c) 2008-2015 Pavel Holoborodko + + Contributors: + Dmitriy Gubanov, Konstantin Holoborodko, Brian Gladman, + Helmut Jarausch, Fokko Beekhof, Ulrich Mutze, Heinz van Saanen, + Pere Constans, Peter van Hoof, Gael Guennebaud, Tsai Chia Cheng, + Alexei Zubanov, Jauhien Piatlicki, Victor Berger, John Westwood, + Petr Aleksandrov, Orion Poplawski, Charles Karney, Arash Partow, + Rodney James, Jorge Leitao. + + Licensing: + (A) MPFR C++ is under GNU General Public License ("GPL"). + + (B) Non-free licenses may also be purchased from the author, for users who + do not want their programs protected by the GPL. + + The non-free licenses are for users that wish to use MPFR C++ in + their products but are unwilling to release their software + under the GPL (which would require them to release source code + and allow free redistribution). + + Such users can purchase an unlimited-use license from the author. + Contact us for more details. + + GNU General Public License ("GPL") copyright permissions statement: + ************************************************************************** + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +*/ + +#ifndef __MPREAL_H__ +#define __MPREAL_H__ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +// Options +#define MPREAL_HAVE_MSVC_DEBUGVIEW // Enable Debugger Visualizer for "Debug" builds in MSVC. +#define MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS // Enable extended std::numeric_limits specialization. + // Meaning that "digits", "round_style" and similar members are defined as functions, not constants. + // See std::numeric_limits at the end of the file for more information. + +// Library version +#define MPREAL_VERSION_MAJOR 3 +#define MPREAL_VERSION_MINOR 6 +#define MPREAL_VERSION_PATCHLEVEL 2 +#define MPREAL_VERSION_STRING "3.6.2" + +// Detect compiler using signatures from http://predef.sourceforge.net/ +#if defined(__GNUC__) + #define IsInf(x) (isinf)(x) // GNU C++/Intel ICC compiler on Linux +#elif defined(_MSC_VER) // Microsoft Visual C++ + #define IsInf(x) (!_finite(x)) +#else + #define IsInf(x) (std::isinf)(x) // GNU C/C++ (and/or other compilers), just hope for C99 conformance +#endif + +// A Clang feature extension to determine compiler features. +#ifndef __has_feature + #define __has_feature(x) 0 +#endif + +// Detect support for r-value references (move semantic). Borrowed from Eigen. +#if (__has_feature(cxx_rvalue_references) || \ + defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L || \ + (defined(_MSC_VER) && _MSC_VER >= 1600)) + + #define MPREAL_HAVE_MOVE_SUPPORT + + // Use fields in mpfr_t structure to check if it was initialized / set dummy initialization + #define mpfr_is_initialized(x) (0 != (x)->_mpfr_d) + #define mpfr_set_uninitialized(x) ((x)->_mpfr_d = 0 ) +#endif + +// Detect support for explicit converters. +#if (__has_feature(cxx_explicit_conversions) || \ + (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GNUC_MINOR__ >= 5) || __cplusplus >= 201103L || \ + (defined(_MSC_VER) && _MSC_VER >= 1800)) + + #define MPREAL_HAVE_EXPLICIT_CONVERTERS +#endif + +#define MPFR_USE_INTMAX_T // Enable 64-bit integer types - should be defined before mpfr.h + +#if defined(MPREAL_HAVE_MSVC_DEBUGVIEW) && defined(_MSC_VER) && defined(_DEBUG) + #define MPREAL_MSVC_DEBUGVIEW_CODE DebugView = toString(); + #define MPREAL_MSVC_DEBUGVIEW_DATA std::string DebugView; +#else + #define MPREAL_MSVC_DEBUGVIEW_CODE + #define MPREAL_MSVC_DEBUGVIEW_DATA +#endif + +#include + +#if (MPFR_VERSION < MPFR_VERSION_NUM(3,0,0)) + #include // Needed for random() +#endif + +// Less important options +#define MPREAL_DOUBLE_BITS_OVERFLOW -1 // Triggers overflow exception during conversion to double if mpreal + // cannot fit in MPREAL_DOUBLE_BITS_OVERFLOW bits + // = -1 disables overflow checks (default) + +// Fast replacement for mpfr_set_zero(x, +1): +// (a) uses low-level data members, might not be compatible with new versions of MPFR +// (b) sign is not set, add (x)->_mpfr_sign = 1; +#define mpfr_set_zero_fast(x) ((x)->_mpfr_exp = __MPFR_EXP_ZERO) + +#if defined(__GNUC__) + #define MPREAL_PERMISSIVE_EXPR __extension__ +#else + #define MPREAL_PERMISSIVE_EXPR +#endif + +namespace mpfr { + +class mpreal { +private: + mpfr_t mp; + +public: + + // Get default rounding mode & precision + inline static mp_rnd_t get_default_rnd() { return (mp_rnd_t)(mpfr_get_default_rounding_mode()); } + inline static mp_prec_t get_default_prec() { return mpfr_get_default_prec(); } + + // Constructors && type conversions + mpreal(); + mpreal(const mpreal& u); + mpreal(const mpf_t u); + mpreal(const mpz_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const mpq_t u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long double u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const unsigned int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const long int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const int u, mp_prec_t prec = mpreal::get_default_prec(), mp_rnd_t mode = mpreal::get_default_rnd()); + + // Construct mpreal from mpfr_t structure. + // shared = true allows to avoid deep copy, so that mpreal and 'u' share the same data & pointers. + mpreal(const mpfr_t u, bool shared = false); + + mpreal(const char* s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); + mpreal(const std::string& s, mp_prec_t prec = mpreal::get_default_prec(), int base = 10, mp_rnd_t mode = mpreal::get_default_rnd()); + + ~mpreal(); + +#ifdef MPREAL_HAVE_MOVE_SUPPORT + mpreal& operator=(mpreal&& v); + mpreal(mpreal&& u); +#endif + + // Operations + // = + // +, -, *, /, ++, --, <<, >> + // *=, +=, -=, /=, + // <, >, ==, <=, >= + + // = + mpreal& operator=(const mpreal& v); + mpreal& operator=(const mpf_t v); + mpreal& operator=(const mpz_t v); + mpreal& operator=(const mpq_t v); + mpreal& operator=(const long double v); + mpreal& operator=(const double v); + mpreal& operator=(const unsigned long int v); + mpreal& operator=(const unsigned long long int v); + mpreal& operator=(const long long int v); + mpreal& operator=(const unsigned int v); + mpreal& operator=(const long int v); + mpreal& operator=(const int v); + mpreal& operator=(const char* s); + mpreal& operator=(const std::string& s); + template mpreal& operator= (const std::complex& z); + + // + + mpreal& operator+=(const mpreal& v); + mpreal& operator+=(const mpf_t v); + mpreal& operator+=(const mpz_t v); + mpreal& operator+=(const mpq_t v); + mpreal& operator+=(const long double u); + mpreal& operator+=(const double u); + mpreal& operator+=(const unsigned long int u); + mpreal& operator+=(const unsigned int u); + mpreal& operator+=(const long int u); + mpreal& operator+=(const int u); + + mpreal& operator+=(const long long int u); + mpreal& operator+=(const unsigned long long int u); + mpreal& operator-=(const long long int u); + mpreal& operator-=(const unsigned long long int u); + mpreal& operator*=(const long long int u); + mpreal& operator*=(const unsigned long long int u); + mpreal& operator/=(const long long int u); + mpreal& operator/=(const unsigned long long int u); + + const mpreal operator+() const; + mpreal& operator++ (); + const mpreal operator++ (int); + + // - + mpreal& operator-=(const mpreal& v); + mpreal& operator-=(const mpz_t v); + mpreal& operator-=(const mpq_t v); + mpreal& operator-=(const long double u); + mpreal& operator-=(const double u); + mpreal& operator-=(const unsigned long int u); + mpreal& operator-=(const unsigned int u); + mpreal& operator-=(const long int u); + mpreal& operator-=(const int u); + const mpreal operator-() const; + friend const mpreal operator-(const unsigned long int b, const mpreal& a); + friend const mpreal operator-(const unsigned int b, const mpreal& a); + friend const mpreal operator-(const long int b, const mpreal& a); + friend const mpreal operator-(const int b, const mpreal& a); + friend const mpreal operator-(const double b, const mpreal& a); + mpreal& operator-- (); + const mpreal operator-- (int); + + // * + mpreal& operator*=(const mpreal& v); + mpreal& operator*=(const mpz_t v); + mpreal& operator*=(const mpq_t v); + mpreal& operator*=(const long double v); + mpreal& operator*=(const double v); + mpreal& operator*=(const unsigned long int v); + mpreal& operator*=(const unsigned int v); + mpreal& operator*=(const long int v); + mpreal& operator*=(const int v); + + // / + mpreal& operator/=(const mpreal& v); + mpreal& operator/=(const mpz_t v); + mpreal& operator/=(const mpq_t v); + mpreal& operator/=(const long double v); + mpreal& operator/=(const double v); + mpreal& operator/=(const unsigned long int v); + mpreal& operator/=(const unsigned int v); + mpreal& operator/=(const long int v); + mpreal& operator/=(const int v); + friend const mpreal operator/(const unsigned long int b, const mpreal& a); + friend const mpreal operator/(const unsigned int b, const mpreal& a); + friend const mpreal operator/(const long int b, const mpreal& a); + friend const mpreal operator/(const int b, const mpreal& a); + friend const mpreal operator/(const double b, const mpreal& a); + + //<<= Fast Multiplication by 2^u + mpreal& operator<<=(const unsigned long int u); + mpreal& operator<<=(const unsigned int u); + mpreal& operator<<=(const long int u); + mpreal& operator<<=(const int u); + + //>>= Fast Division by 2^u + mpreal& operator>>=(const unsigned long int u); + mpreal& operator>>=(const unsigned int u); + mpreal& operator>>=(const long int u); + mpreal& operator>>=(const int u); + + // Type Conversion operators + bool toBool ( ) const; + long toLong (mp_rnd_t mode = GMP_RNDZ) const; + unsigned long toULong (mp_rnd_t mode = GMP_RNDZ) const; + long long toLLong (mp_rnd_t mode = GMP_RNDZ) const; + unsigned long long toULLong (mp_rnd_t mode = GMP_RNDZ) const; + float toFloat (mp_rnd_t mode = GMP_RNDN) const; + double toDouble (mp_rnd_t mode = GMP_RNDN) const; + long double toLDouble (mp_rnd_t mode = GMP_RNDN) const; + +#if defined (MPREAL_HAVE_EXPLICIT_CONVERTERS) + explicit operator bool () const { return toBool(); } + explicit operator int () const { return int(toLong()); } + explicit operator long () const { return toLong(); } + explicit operator long long () const { return toLLong(); } + explicit operator unsigned () const { return unsigned(toULong()); } + explicit operator unsigned long () const { return toULong(); } + explicit operator unsigned long long () const { return toULLong(); } + explicit operator float () const { return toFloat(); } + explicit operator double () const { return toDouble(); } + explicit operator long double () const { return toLDouble(); } +#endif + + // Get raw pointers so that mpreal can be directly used in raw mpfr_* functions + ::mpfr_ptr mpfr_ptr(); + ::mpfr_srcptr mpfr_ptr() const; + ::mpfr_srcptr mpfr_srcptr() const; + + // Convert mpreal to string with n significant digits in base b + // n = -1 -> convert with the maximum available digits + std::string toString(int n = -1, int b = 10, mp_rnd_t mode = mpreal::get_default_rnd()) const; + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + std::string toString(const std::string& format) const; +#endif + + std::ostream& output(std::ostream& os) const; + + // Math Functions + friend const mpreal sqr (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sqrt(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sqrt(const unsigned long int v, mp_rnd_t rnd_mode); + friend const mpreal cbrt(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal root(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode); + friend const mpreal pow (const mpreal& a, const long int b, mp_rnd_t rnd_mode); + friend const mpreal pow (const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode); + friend const mpreal pow (const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode); + friend const mpreal fabs(const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal abs(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode); + friend inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); + friend inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode); + friend inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode); + friend int cmpabs(const mpreal& a,const mpreal& b); + + friend const mpreal log (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal logb (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log10(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal exp10(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal log1p(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal expm1(const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal cos(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sin(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tan(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sec(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal csc(const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal cot(const mpreal& v, mp_rnd_t rnd_mode); + friend int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal acos (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asin (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atan (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode); + friend const mpreal acot (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asec (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acsc (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal cosh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sinh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tanh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal sech (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal csch (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal coth (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acosh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asinh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal atanh (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acoth (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal asech (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal acsch (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + + friend const mpreal fac_ui (unsigned long int v, mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal eint (const mpreal& v, mp_rnd_t rnd_mode); + + friend const mpreal gamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal tgamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal lngamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal lgamma (const mpreal& v, int *signp, mp_rnd_t rnd_mode); + friend const mpreal zeta (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal erf (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal erfc (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselj0 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselj1 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besseljn (long n, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal bessely0 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal bessely1 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal besselyn (long n, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); + friend const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode); + friend const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode); + friend const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t rnd_mode); + friend int sgn(const mpreal& v); // returns -1 or +1 + +// MPFR 2.4.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + friend int sinh_cosh (mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal li2 (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal rec_sqrt (const mpreal& v, mp_rnd_t rnd_mode); + + // MATLAB's semantic equivalents + friend const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Remainder after division + friend const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); // Modulus after division +#endif + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + friend const mpreal digamma (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal ai (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear +#endif + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + friend const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode); // use gmp_randinit_default() to init state, gmp_randclear() to clear + friend const mpreal grandom (unsigned int seed); +#endif + + // Uniformly distributed random number generation in [0,1] using + // Mersenne-Twister algorithm by default. + // Use parameter to setup seed, e.g.: random((unsigned)time(NULL)) + // Check urandom() for more precise control. + friend const mpreal random(unsigned int seed); + + // Splits mpreal value into fractional and integer parts. + // Returns fractional part and stores integer part in n. + friend const mpreal modf(const mpreal& v, mpreal& n); + + // Constants + // don't forget to call mpfr_free_cache() for every thread where you are using const-functions + friend const mpreal const_log2 (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_pi (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_euler (mp_prec_t prec, mp_rnd_t rnd_mode); + friend const mpreal const_catalan (mp_prec_t prec, mp_rnd_t rnd_mode); + + // returns +inf iff sign>=0 otherwise -inf + friend const mpreal const_infinity(int sign, mp_prec_t prec); + + // Output/ Input + friend std::ostream& operator<<(std::ostream& os, const mpreal& v); + friend std::istream& operator>>(std::istream& is, mpreal& v); + + // Integer Related Functions + friend const mpreal rint (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal ceil (const mpreal& v); + friend const mpreal floor(const mpreal& v); + friend const mpreal round(const mpreal& v); + friend const mpreal trunc(const mpreal& v); + friend const mpreal rint_ceil (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_floor (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_round (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal rint_trunc (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal frac (const mpreal& v, mp_rnd_t rnd_mode); + friend const mpreal remainder ( const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + + // Miscellaneous Functions + friend const mpreal nexttoward (const mpreal& x, const mpreal& y); + friend const mpreal nextabove (const mpreal& x); + friend const mpreal nextbelow (const mpreal& x); + + // use gmp_randinit_default() to init state, gmp_randclear() to clear + friend const mpreal urandomb (gmp_randstate_t& state); + +// MPFR < 2.4.2 Specifics +#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) + friend const mpreal random2 (mp_size_t size, mp_exp_t exp); +#endif + + // Instance Checkers + friend bool (isnan) (const mpreal& v); + friend bool (isinf) (const mpreal& v); + friend bool (isfinite) (const mpreal& v); + + friend bool isnum (const mpreal& v); + friend bool iszero (const mpreal& v); + friend bool isint (const mpreal& v); + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + friend bool isregular(const mpreal& v); +#endif + + // Set/Get instance properties + inline mp_prec_t get_prec() const; + inline void set_prec(mp_prec_t prec, mp_rnd_t rnd_mode = get_default_rnd()); // Change precision with rounding mode + + // Aliases for get_prec(), set_prec() - needed for compatibility with std::complex interface + inline mpreal& setPrecision(int Precision, mp_rnd_t RoundingMode = get_default_rnd()); + inline int getPrecision() const; + + // Set mpreal to +/- inf, NaN, +/-0 + mpreal& setInf (int Sign = +1); + mpreal& setNan (); + mpreal& setZero (int Sign = +1); + mpreal& setSign (int Sign, mp_rnd_t RoundingMode = get_default_rnd()); + + //Exponent + mp_exp_t get_exp(); + int set_exp(mp_exp_t e); + int check_range (int t, mp_rnd_t rnd_mode = get_default_rnd()); + int subnormalize (int t, mp_rnd_t rnd_mode = get_default_rnd()); + + // Inexact conversion from float + inline bool fits_in_bits(double x, int n); + + // Set/Get global properties + static void set_default_prec(mp_prec_t prec); + static void set_default_rnd(mp_rnd_t rnd_mode); + + static mp_exp_t get_emin (void); + static mp_exp_t get_emax (void); + static mp_exp_t get_emin_min (void); + static mp_exp_t get_emin_max (void); + static mp_exp_t get_emax_min (void); + static mp_exp_t get_emax_max (void); + static int set_emin (mp_exp_t exp); + static int set_emax (mp_exp_t exp); + + // Efficient swapping of two mpreal values - needed for std algorithms + friend void swap(mpreal& x, mpreal& y); + + friend const mpreal fmax(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + friend const mpreal fmin(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode); + +private: + // Human friendly Debug Preview in Visual Studio. + // Put one of these lines: + // + // mpfr::mpreal= ; Show value only + // mpfr::mpreal=, bits ; Show value & precision + // + // at the beginning of + // [Visual Studio Installation Folder]\Common7\Packages\Debugger\autoexp.dat + MPREAL_MSVC_DEBUGVIEW_DATA + + // "Smart" resources deallocation. Checks if instance initialized before deletion. + void clear(::mpfr_ptr); +}; + +////////////////////////////////////////////////////////////////////////// +// Exceptions +class conversion_overflow : public std::exception { +public: + std::string why() { return "inexact conversion from floating point"; } +}; + +////////////////////////////////////////////////////////////////////////// +// Constructors & converters +// Default constructor: creates mp number and initializes it to 0. +inline mpreal::mpreal() +{ + mpfr_init2(mpfr_ptr(), mpreal::get_default_prec()); + mpfr_set_zero_fast(mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpreal& u) +{ + mpfr_init2(mpfr_ptr(),mpfr_get_prec(u.mpfr_srcptr())); + mpfr_set (mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +#ifdef MPREAL_HAVE_MOVE_SUPPORT +inline mpreal::mpreal(mpreal&& other) +{ + mpfr_set_uninitialized(mpfr_ptr()); // make sure "other" holds no pointer to actual data + mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal& mpreal::operator=(mpreal&& other) +{ + mpfr_swap(mpfr_ptr(), other.mpfr_ptr()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} +#endif + +inline mpreal::mpreal(const mpfr_t u, bool shared) +{ + if(shared) + { + std::memcpy(mpfr_ptr(), u, sizeof(mpfr_t)); + } + else + { + mpfr_init2(mpfr_ptr(), mpfr_get_prec(u)); + mpfr_set (mpfr_ptr(), u, mpreal::get_default_rnd()); + } + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpf_t u) +{ + mpfr_init2(mpfr_ptr(),(mp_prec_t) mpf_get_prec(u)); // (gmp: mp_bitcnt_t) unsigned long -> long (mpfr: mp_prec_t) + mpfr_set_f(mpfr_ptr(),u,mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpz_t u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + mpfr_set_z(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const mpq_t u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + mpfr_set_q(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const double u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2(mpfr_ptr(), prec); + +#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) + if(fits_in_bits(u, MPREAL_DOUBLE_BITS_OVERFLOW)) + { + mpfr_set_d(mpfr_ptr(), u, mode); + }else + throw conversion_overflow(); +#else + mpfr_set_d(mpfr_ptr(), u, mode); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long double u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ld(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned long long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_uj(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_sj(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ui(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const unsigned int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_ui(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const long int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_si(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const int u, mp_prec_t prec, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_si(mpfr_ptr(), u, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_str(mpfr_ptr(), s, base, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mpreal::mpreal(const std::string& s, mp_prec_t prec, int base, mp_rnd_t mode) +{ + mpfr_init2 (mpfr_ptr(), prec); + mpfr_set_str(mpfr_ptr(), s.c_str(), base, mode); + + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline void mpreal::clear(::mpfr_ptr x) +{ +#ifdef MPREAL_HAVE_MOVE_SUPPORT + if(mpfr_is_initialized(x)) +#endif + mpfr_clear(x); +} + +inline mpreal::~mpreal() +{ + clear(mpfr_ptr()); +} + +// internal namespace needed for template magic +namespace internal{ + + // Use SFINAE to restrict arithmetic operations instantiation only for numeric types + // This is needed for smooth integration with libraries based on expression templates, like Eigen. + // TODO: Do the same for boolean operators. + template struct result_type {}; + + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; + template <> struct result_type {typedef mpreal type;}; +} + +// + Addition +template +inline const typename internal::result_type::type + operator+(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) += rhs; } + +template +inline const typename internal::result_type::type + operator+(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) += lhs; } + +// - Subtraction +template +inline const typename internal::result_type::type + operator-(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) -= rhs; } + +template +inline const typename internal::result_type::type + operator-(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) -= rhs; } + +// * Multiplication +template +inline const typename internal::result_type::type + operator*(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) *= rhs; } + +template +inline const typename internal::result_type::type + operator*(const Lhs& lhs, const mpreal& rhs){ return mpreal(rhs) *= lhs; } + +// / Division +template +inline const typename internal::result_type::type + operator/(const mpreal& lhs, const Rhs& rhs){ return mpreal(lhs) /= rhs; } + +template +inline const typename internal::result_type::type + operator/(const Lhs& lhs, const mpreal& rhs){ return mpreal(lhs) /= rhs; } + +////////////////////////////////////////////////////////////////////////// +// sqrt +const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const long int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const int v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const long double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal sqrt(const double v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +// abs +inline const mpreal abs(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()); + +////////////////////////////////////////////////////////////////////////// +// pow +const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); +inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode = mpreal::get_default_rnd()); + +////////////////////////////////////////////////////////////////////////// +// Estimate machine epsilon for the given precision +// Returns smallest eps such that 1.0 + eps != 1.0 +inline mpreal machine_epsilon(mp_prec_t prec = mpreal::get_default_prec()); + +// Returns smallest eps such that x + eps != x (relative machine epsilon) +inline mpreal machine_epsilon(const mpreal& x); + +// Gives max & min values for the required precision, +// minval is 'safe' meaning 1 / minval does not overflow +// maxval is 'safe' meaning 1 / maxval does not underflow +inline mpreal minval(mp_prec_t prec = mpreal::get_default_prec()); +inline mpreal maxval(mp_prec_t prec = mpreal::get_default_prec()); + +// 'Dirty' equality check 1: |a-b| < min{|a|,|b|} * eps +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps); + +// 'Dirty' equality check 2: |a-b| < min{|a|,|b|} * eps( min{|a|,|b|} ) +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b); + +// 'Bitwise' equality check +// maxUlps - a and b can be apart by maxUlps binary numbers. +inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps); + +////////////////////////////////////////////////////////////////////////// +// Convert precision in 'bits' to decimal digits and vice versa. +// bits = ceil(digits*log[2](10)) +// digits = floor(bits*log[10](2)) + +inline mp_prec_t digits2bits(int d); +inline int bits2digits(mp_prec_t b); + +////////////////////////////////////////////////////////////////////////// +// min, max +const mpreal (max)(const mpreal& x, const mpreal& y); +const mpreal (min)(const mpreal& x, const mpreal& y); + +////////////////////////////////////////////////////////////////////////// +// Implementation +////////////////////////////////////////////////////////////////////////// + +////////////////////////////////////////////////////////////////////////// +// Operators - Assignment +inline mpreal& mpreal::operator=(const mpreal& v) +{ + if (this != &v) + { + mp_prec_t tp = mpfr_get_prec( mpfr_srcptr()); + mp_prec_t vp = mpfr_get_prec(v.mpfr_srcptr()); + + if(tp != vp){ + clear(mpfr_ptr()); + mpfr_init2(mpfr_ptr(), vp); + } + + mpfr_set(mpfr_ptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + } + return *this; +} + +inline mpreal& mpreal::operator=(const mpf_t v) +{ + mpfr_set_f(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const mpz_t v) +{ + mpfr_set_z(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const mpq_t v) +{ + mpfr_set_q(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long double v) +{ + mpfr_set_ld(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const double v) +{ +#if (MPREAL_DOUBLE_BITS_OVERFLOW > -1) + if(fits_in_bits(v, MPREAL_DOUBLE_BITS_OVERFLOW)) + { + mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); + }else + throw conversion_overflow(); +#else + mpfr_set_d(mpfr_ptr(),v,mpreal::get_default_rnd()); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned long int v) +{ + mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned int v) +{ + mpfr_set_ui(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const unsigned long long int v) +{ + mpfr_set_uj(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long long int v) +{ + mpfr_set_sj(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const long int v) +{ + mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const int v) +{ + mpfr_set_si(mpfr_ptr(), v, mpreal::get_default_rnd()); + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator=(const char* s) +{ + // Use other converters for more precise control on base & precision & rounding: + // + // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) + // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) + // + // Here we assume base = 10 and we use precision of target variable. + + mpfr_t t; + + mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); + + if(0 == mpfr_set_str(t, s, 10, mpreal::get_default_rnd())) + { + mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + } + + clear(t); + return *this; +} + +inline mpreal& mpreal::operator=(const std::string& s) +{ + // Use other converters for more precise control on base & precision & rounding: + // + // mpreal(const char* s, mp_prec_t prec, int base, mp_rnd_t mode) + // mpreal(const std::string& s,mp_prec_t prec, int base, mp_rnd_t mode) + // + // Here we assume base = 10 and we use precision of target variable. + + mpfr_t t; + + mpfr_init2(t, mpfr_get_prec(mpfr_srcptr())); + + if(0 == mpfr_set_str(t, s.c_str(), 10, mpreal::get_default_rnd())) + { + mpfr_set(mpfr_ptr(), t, mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + } + + clear(t); + return *this; +} + +template +inline mpreal& mpreal::operator= (const std::complex& z) +{ + return *this = z.real(); +} + +////////////////////////////////////////////////////////////////////////// +// + Addition +inline mpreal& mpreal::operator+=(const mpreal& v) +{ + mpfr_add(mpfr_ptr(), mpfr_srcptr(), v.mpfr_srcptr(), mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpf_t u) +{ + *this += mpreal(u); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpz_t u) +{ + mpfr_add_z(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const mpq_t u) +{ + mpfr_add_q(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+= (const long double u) +{ + *this += mpreal(u); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+= (const double u) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_add_d(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); +#else + *this += mpreal(u); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const unsigned long int u) +{ + mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const unsigned int u) +{ + mpfr_add_ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const long int u) +{ + mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const int u) +{ + mpfr_add_si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator+=(const long long int u) { *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator+=(const unsigned long long int u){ *this += mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator-=(const long long int u) { *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator-=(const unsigned long long int u){ *this -= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator*=(const long long int u) { *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator*=(const unsigned long long int u){ *this *= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator/=(const long long int u) { *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } +inline mpreal& mpreal::operator/=(const unsigned long long int u){ *this /= mpreal(u); MPREAL_MSVC_DEBUGVIEW_CODE; return *this; } + +inline const mpreal mpreal::operator+()const { return mpreal(*this); } + +inline const mpreal operator+(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_add(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline mpreal& mpreal::operator++() +{ + return *this += 1; +} + +inline const mpreal mpreal::operator++ (int) +{ + mpreal x(*this); + *this += 1; + return x; +} + +inline mpreal& mpreal::operator--() +{ + return *this -= 1; +} + +inline const mpreal mpreal::operator-- (int) +{ + mpreal x(*this); + *this -= 1; + return x; +} + +////////////////////////////////////////////////////////////////////////// +// - Subtraction +inline mpreal& mpreal::operator-=(const mpreal& v) +{ + mpfr_sub(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const mpz_t v) +{ + mpfr_sub_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const mpq_t v) +{ + mpfr_sub_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const long double v) +{ + *this -= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_sub_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this -= mpreal(v); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const unsigned long int v) +{ + mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const unsigned int v) +{ + mpfr_sub_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const long int v) +{ + mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator-=(const int v) +{ + mpfr_sub_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal mpreal::operator-()const +{ + mpreal u(*this); + mpfr_neg(u.mpfr_ptr(),u.mpfr_srcptr(),mpreal::get_default_rnd()); + return u; +} + +inline const mpreal operator-(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_sub(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline const mpreal operator-(const double b, const mpreal& a) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_d_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +#else + mpreal x(b, mpfr_get_prec(a.mpfr_ptr())); + x -= a; + return x; +#endif +} + +inline const mpreal operator-(const unsigned long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const unsigned int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_ui_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator-(const int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + mpfr_si_sub(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +////////////////////////////////////////////////////////////////////////// +// * Multiplication +inline mpreal& mpreal::operator*= (const mpreal& v) +{ + mpfr_mul(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const mpz_t v) +{ + mpfr_mul_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const mpq_t v) +{ + mpfr_mul_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const long double v) +{ + *this *= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_mul_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this *= mpreal(v); +#endif + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const unsigned long int v) +{ + mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const unsigned int v) +{ + mpfr_mul_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const long int v) +{ + mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator*=(const int v) +{ + mpfr_mul_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator*(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_ptr()), mpfr_get_prec(b.mpfr_ptr()))); + mpfr_mul(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +////////////////////////////////////////////////////////////////////////// +// / Division +inline mpreal& mpreal::operator/=(const mpreal& v) +{ + mpfr_div(mpfr_ptr(),mpfr_srcptr(),v.mpfr_srcptr(),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const mpz_t v) +{ + mpfr_div_z(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const mpq_t v) +{ + mpfr_div_q(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const long double v) +{ + *this /= mpreal(v); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const double v) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpfr_div_d(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); +#else + *this /= mpreal(v); +#endif + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const unsigned long int v) +{ + mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const unsigned int v) +{ + mpfr_div_ui(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const long int v) +{ + mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator/=(const int v) +{ + mpfr_div_si(mpfr_ptr(),mpfr_srcptr(),v,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator/(const mpreal& a, const mpreal& b) +{ + mpreal c(0, (std::max)(mpfr_get_prec(a.mpfr_srcptr()), mpfr_get_prec(b.mpfr_srcptr()))); + mpfr_div(c.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), mpreal::get_default_rnd()); + return c; +} + +inline const mpreal operator/(const unsigned long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const unsigned int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_ui_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const long int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const int b, const mpreal& a) +{ + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_si_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +} + +inline const mpreal operator/(const double b, const mpreal& a) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + mpreal x(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_d_div(x.mpfr_ptr(), b, a.mpfr_srcptr(), mpreal::get_default_rnd()); + return x; +#else + mpreal x(0, mpfr_get_prec(a.mpfr_ptr())); + x /= a; + return x; +#endif +} + +////////////////////////////////////////////////////////////////////////// +// Shifts operators - Multiplication/Division by power of 2 +inline mpreal& mpreal::operator<<=(const unsigned long int u) +{ + mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const unsigned int u) +{ + mpfr_mul_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const long int u) +{ + mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator<<=(const int u) +{ + mpfr_mul_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const unsigned long int u) +{ + mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const unsigned int u) +{ + mpfr_div_2ui(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const long int u) +{ + mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),u,mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::operator>>=(const int u) +{ + mpfr_div_2si(mpfr_ptr(),mpfr_srcptr(),static_cast(u),mpreal::get_default_rnd()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline const mpreal operator<<(const mpreal& v, const unsigned long int k) +{ + return mul_2ui(v,k); +} + +inline const mpreal operator<<(const mpreal& v, const unsigned int k) +{ + return mul_2ui(v,static_cast(k)); +} + +inline const mpreal operator<<(const mpreal& v, const long int k) +{ + return mul_2si(v,k); +} + +inline const mpreal operator<<(const mpreal& v, const int k) +{ + return mul_2si(v,static_cast(k)); +} + +inline const mpreal operator>>(const mpreal& v, const unsigned long int k) +{ + return div_2ui(v,k); +} + +inline const mpreal operator>>(const mpreal& v, const long int k) +{ + return div_2si(v,k); +} + +inline const mpreal operator>>(const mpreal& v, const unsigned int k) +{ + return div_2ui(v,static_cast(k)); +} + +inline const mpreal operator>>(const mpreal& v, const int k) +{ + return div_2si(v,static_cast(k)); +} + +// mul_2ui +inline const mpreal mul_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_mul_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +// mul_2si +inline const mpreal mul_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_mul_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +inline const mpreal div_2ui(const mpreal& v, unsigned long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_div_2ui(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +inline const mpreal div_2si(const mpreal& v, long int k, mp_rnd_t rnd_mode) +{ + mpreal x(v); + mpfr_div_2si(x.mpfr_ptr(),v.mpfr_srcptr(),k,rnd_mode); + return x; +} + +////////////////////////////////////////////////////////////////////////// +//Relational operators + +// WARNING: +// +// Please note that following checks for double-NaN are guaranteed to work only in IEEE math mode: +// +// isnan(b) = (b != b) +// isnan(b) = !(b == b) (we use in code below) +// +// Be cautions if you use compiler options which break strict IEEE compliance (e.g. -ffast-math in GCC). +// Use std::isnan instead (C++11). + +inline bool operator > (const mpreal& a, const mpreal& b ){ return (mpfr_greater_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator > (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) > 0 ); } +inline bool operator > (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) > 0 ); } + +inline bool operator >= (const mpreal& a, const mpreal& b ){ return (mpfr_greaterequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator >= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } +// inline bool operator >= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (isnan()a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) >= 0 ); } +inline bool operator >= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) >= 0 ); } + +inline bool operator < (const mpreal& a, const mpreal& b ){ return (mpfr_less_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator < (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) < 0 ); } +inline bool operator < (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) < 0 ); } + +inline bool operator <= (const mpreal& a, const mpreal& b ){ return (mpfr_lessequal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator <= (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) <= 0 ); } +inline bool operator <= (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) <= 0 ); } + +inline bool operator == (const mpreal& a, const mpreal& b ){ return (mpfr_equal_p(a.mpfr_srcptr(),b.mpfr_srcptr()) != 0 ); } +inline bool operator == (const mpreal& a, const unsigned long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const unsigned int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_ui(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const long int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const int b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (mpfr_cmp_si(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const long double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_ld(a.mpfr_srcptr(),b) == 0 ); } +inline bool operator == (const mpreal& a, const double b ){ return !isnan EIGEN_NOT_A_MACRO (a) && (b == b) && (mpfr_cmp_d (a.mpfr_srcptr(),b) == 0 ); } + +inline bool operator != (const mpreal& a, const mpreal& b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const unsigned long int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const unsigned int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const long int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const int b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const long double b ){ return !(a == b); } +inline bool operator != (const mpreal& a, const double b ){ return !(a == b); } + +inline bool (isnan) (const mpreal& op){ return (mpfr_nan_p (op.mpfr_srcptr()) != 0 ); } +inline bool (isinf) (const mpreal& op){ return (mpfr_inf_p (op.mpfr_srcptr()) != 0 ); } +inline bool (isfinite) (const mpreal& op){ return (mpfr_number_p (op.mpfr_srcptr()) != 0 ); } +inline bool iszero (const mpreal& op){ return (mpfr_zero_p (op.mpfr_srcptr()) != 0 ); } +inline bool isint (const mpreal& op){ return (mpfr_integer_p(op.mpfr_srcptr()) != 0 ); } + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) +inline bool isregular(const mpreal& op){ return (mpfr_regular_p(op.mpfr_srcptr()));} +#endif + +////////////////////////////////////////////////////////////////////////// +// Type Converters +inline bool mpreal::toBool ( ) const { return mpfr_zero_p (mpfr_srcptr()) == 0; } +inline long mpreal::toLong (mp_rnd_t mode) const { return mpfr_get_si (mpfr_srcptr(), mode); } +inline unsigned long mpreal::toULong (mp_rnd_t mode) const { return mpfr_get_ui (mpfr_srcptr(), mode); } +inline float mpreal::toFloat (mp_rnd_t mode) const { return mpfr_get_flt(mpfr_srcptr(), mode); } +inline double mpreal::toDouble (mp_rnd_t mode) const { return mpfr_get_d (mpfr_srcptr(), mode); } +inline long double mpreal::toLDouble(mp_rnd_t mode) const { return mpfr_get_ld (mpfr_srcptr(), mode); } +inline long long mpreal::toLLong (mp_rnd_t mode) const { return mpfr_get_sj (mpfr_srcptr(), mode); } +inline unsigned long long mpreal::toULLong (mp_rnd_t mode) const { return mpfr_get_uj (mpfr_srcptr(), mode); } + +inline ::mpfr_ptr mpreal::mpfr_ptr() { return mp; } +inline ::mpfr_srcptr mpreal::mpfr_ptr() const { return mp; } +inline ::mpfr_srcptr mpreal::mpfr_srcptr() const { return mp; } + +template +inline std::string toString(T t, std::ios_base & (*f)(std::ios_base&)) +{ + std::ostringstream oss; + oss << f << t; + return oss.str(); +} + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + +inline std::string mpreal::toString(const std::string& format) const +{ + char *s = NULL; + std::string out; + + if( !format.empty() ) + { + if(!(mpfr_asprintf(&s, format.c_str(), mpfr_srcptr()) < 0)) + { + out = std::string(s); + + mpfr_free_str(s); + } + } + + return out; +} + +#endif + +inline std::string mpreal::toString(int n, int b, mp_rnd_t mode) const +{ + // TODO: Add extended format specification (f, e, rounding mode) as it done in output operator + (void)b; + (void)mode; + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + + std::ostringstream format; + + int digits = (n >= 0) ? n : 1 + bits2digits(mpfr_get_prec(mpfr_srcptr())); + + format << "%." << digits << "RNg"; + + return toString(format.str()); + +#else + + char *s, *ns = NULL; + size_t slen, nslen; + mp_exp_t exp; + std::string out; + + if(mpfr_inf_p(mp)) + { + if(mpfr_sgn(mp)>0) return "+Inf"; + else return "-Inf"; + } + + if(mpfr_zero_p(mp)) return "0"; + if(mpfr_nan_p(mp)) return "NaN"; + + s = mpfr_get_str(NULL, &exp, b, 0, mp, mode); + ns = mpfr_get_str(NULL, &exp, b, (std::max)(0,n), mp, mode); + + if(s!=NULL && ns!=NULL) + { + slen = strlen(s); + nslen = strlen(ns); + if(nslen<=slen) + { + mpfr_free_str(s); + s = ns; + slen = nslen; + } + else { + mpfr_free_str(ns); + } + + // Make human eye-friendly formatting if possible + if (exp>0 && static_cast(exp)s+exp) ptr--; + + if(ptr==s+exp) out = std::string(s,exp+1); + else out = std::string(s,exp+1)+'.'+std::string(s+exp+1,ptr-(s+exp+1)+1); + + //out = string(s,exp+1)+'.'+string(s+exp+1); + } + else + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s+exp-1) ptr--; + + if(ptr==s+exp-1) out = std::string(s,exp); + else out = std::string(s,exp)+'.'+std::string(s+exp,ptr-(s+exp)+1); + + //out = string(s,exp)+'.'+string(s+exp); + } + + }else{ // exp<0 || exp>slen + if(s[0]=='-') + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s+1) ptr--; + + if(ptr==s+1) out = std::string(s,2); + else out = std::string(s,2)+'.'+std::string(s+2,ptr-(s+2)+1); + + //out = string(s,2)+'.'+string(s+2); + } + else + { + // Remove zeros starting from right end + char* ptr = s+slen-1; + while (*ptr=='0' && ptr>s) ptr--; + + if(ptr==s) out = std::string(s,1); + else out = std::string(s,1)+'.'+std::string(s+1,ptr-(s+1)+1); + + //out = string(s,1)+'.'+string(s+1); + } + + // Make final string + if(--exp) + { + if(exp>0) out += "e+"+mpfr::toString(exp,std::dec); + else out += "e"+mpfr::toString(exp,std::dec); + } + } + + mpfr_free_str(s); + return out; + }else{ + return "conversion error!"; + } +#endif +} + + +////////////////////////////////////////////////////////////////////////// +// I/O +inline std::ostream& mpreal::output(std::ostream& os) const +{ + std::ostringstream format; + const std::ios::fmtflags flags = os.flags(); + + format << ((flags & std::ios::showpos) ? "%+" : "%"); + if (os.precision() >= 0) + format << '.' << os.precision() << "R*" + << ((flags & std::ios::floatfield) == std::ios::fixed ? 'f' : + (flags & std::ios::floatfield) == std::ios::scientific ? 'e' : + 'g'); + else + format << "R*e"; + + char *s = NULL; + if(!(mpfr_asprintf(&s, format.str().c_str(), + mpfr::mpreal::get_default_rnd(), + mpfr_srcptr()) + < 0)) + { + os << std::string(s); + mpfr_free_str(s); + } + return os; +} + +inline std::ostream& operator<<(std::ostream& os, const mpreal& v) +{ + return v.output(os); +} + +inline std::istream& operator>>(std::istream &is, mpreal& v) +{ + // TODO: use cout::hexfloat and other flags to setup base + std::string tmp; + is >> tmp; + mpfr_set_str(v.mpfr_ptr(), tmp.c_str(), 10, mpreal::get_default_rnd()); + return is; +} + +////////////////////////////////////////////////////////////////////////// +// Bits - decimal digits relation +// bits = ceil(digits*log[2](10)) +// digits = floor(bits*log[10](2)) + +inline mp_prec_t digits2bits(int d) +{ + const double LOG2_10 = 3.3219280948873624; + + return mp_prec_t(std::ceil( d * LOG2_10 )); +} + +inline int bits2digits(mp_prec_t b) +{ + const double LOG10_2 = 0.30102999566398119; + + return int(std::floor( b * LOG10_2 )); +} + +////////////////////////////////////////////////////////////////////////// +// Set/Get number properties +inline int sgn(const mpreal& op) +{ + return mpfr_sgn(op.mpfr_srcptr()); +} + +inline mpreal& mpreal::setSign(int sign, mp_rnd_t RoundingMode) +{ + mpfr_setsign(mpfr_ptr(), mpfr_srcptr(), (sign < 0 ? 1 : 0), RoundingMode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline int mpreal::getPrecision() const +{ + return int(mpfr_get_prec(mpfr_srcptr())); +} + +inline mpreal& mpreal::setPrecision(int Precision, mp_rnd_t RoundingMode) +{ + mpfr_prec_round(mpfr_ptr(), Precision, RoundingMode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setInf(int sign) +{ + mpfr_set_inf(mpfr_ptr(), sign); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setNan() +{ + mpfr_set_nan(mpfr_ptr()); + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mpreal& mpreal::setZero(int sign) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + mpfr_set_zero(mpfr_ptr(), sign); +#else + mpfr_set_si(mpfr_ptr(), 0, (mpfr_get_default_rounding_mode)()); + setSign(sign); +#endif + + MPREAL_MSVC_DEBUGVIEW_CODE; + return *this; +} + +inline mp_prec_t mpreal::get_prec() const +{ + return mpfr_get_prec(mpfr_srcptr()); +} + +inline void mpreal::set_prec(mp_prec_t prec, mp_rnd_t rnd_mode) +{ + mpfr_prec_round(mpfr_ptr(),prec,rnd_mode); + MPREAL_MSVC_DEBUGVIEW_CODE; +} + +inline mp_exp_t mpreal::get_exp () +{ + return mpfr_get_exp(mpfr_srcptr()); +} + +inline int mpreal::set_exp (mp_exp_t e) +{ + int x = mpfr_set_exp(mpfr_ptr(), e); + MPREAL_MSVC_DEBUGVIEW_CODE; + return x; +} + +inline const mpreal frexp(const mpreal& x, mp_exp_t* exp, mp_rnd_t mode = mpreal::get_default_rnd()) +{ + mpreal y(x); +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + mpfr_frexp(exp,y.mpfr_ptr(),x.mpfr_srcptr(),mode); +#else + *exp = mpfr_get_exp(y.mpfr_srcptr()); + mpfr_set_exp(y.mpfr_ptr(),0); +#endif + return y; +} + +inline const mpreal ldexp(const mpreal& v, mp_exp_t exp) +{ + mpreal x(v); + + // rounding is not important since we are just increasing the exponent (= exact operation) + mpfr_mul_2si(x.mpfr_ptr(), x.mpfr_srcptr(), exp, mpreal::get_default_rnd()); + return x; +} + +inline const mpreal scalbn(const mpreal& v, mp_exp_t exp) +{ + return ldexp(v, exp); +} + +inline mpreal machine_epsilon(mp_prec_t prec) +{ + /* the smallest eps such that 1 + eps != 1 */ + return machine_epsilon(mpreal(1, prec)); +} + +inline mpreal machine_epsilon(const mpreal& x) +{ + /* the smallest eps such that x + eps != x */ + if( x < 0) + { + return nextabove(-x) + x; + }else{ + return nextabove( x) - x; + } +} + +// minval is 'safe' meaning 1 / minval does not overflow +inline mpreal minval(mp_prec_t prec) +{ + /* min = 1/2 * 2^emin = 2^(emin - 1) */ + return mpreal(1, prec) << mpreal::get_emin()-1; +} + +// maxval is 'safe' meaning 1 / maxval does not underflow +inline mpreal maxval(mp_prec_t prec) +{ + /* max = (1 - eps) * 2^emax, eps is machine epsilon */ + return (mpreal(1, prec) - machine_epsilon(prec)) << mpreal::get_emax(); +} + +inline bool isEqualUlps(const mpreal& a, const mpreal& b, int maxUlps) +{ + return abs(a - b) <= machine_epsilon((max)(abs(a), abs(b))) * maxUlps; +} + +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b, const mpreal& eps) +{ + return abs(a - b) <= eps; +} + +inline bool isEqualFuzzy(const mpreal& a, const mpreal& b) +{ + return isEqualFuzzy(a, b, machine_epsilon((max)(1, (min)(abs(a), abs(b))))); +} + +////////////////////////////////////////////////////////////////////////// +// C++11 sign functions. +inline mpreal copysign(const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal rop(0, mpfr_get_prec(x.mpfr_ptr())); + mpfr_setsign(rop.mpfr_ptr(), x.mpfr_srcptr(), mpfr_signbit(y.mpfr_srcptr()), rnd_mode); + return rop; +} + +inline bool signbit(const mpreal& x) +{ + return mpfr_signbit(x.mpfr_srcptr()); +} + +inline const mpreal modf(const mpreal& v, mpreal& n) +{ + mpreal f(v); + + // rounding is not important since we are using the same number + mpfr_frac (f.mpfr_ptr(),f.mpfr_srcptr(),mpreal::get_default_rnd()); + mpfr_trunc(n.mpfr_ptr(),v.mpfr_srcptr()); + return f; +} + +inline int mpreal::check_range (int t, mp_rnd_t rnd_mode) +{ + return mpfr_check_range(mpfr_ptr(),t,rnd_mode); +} + +inline int mpreal::subnormalize (int t,mp_rnd_t rnd_mode) +{ + int r = mpfr_subnormalize(mpfr_ptr(),t,rnd_mode); + MPREAL_MSVC_DEBUGVIEW_CODE; + return r; +} + +inline mp_exp_t mpreal::get_emin (void) +{ + return mpfr_get_emin(); +} + +inline int mpreal::set_emin (mp_exp_t exp) +{ + return mpfr_set_emin(exp); +} + +inline mp_exp_t mpreal::get_emax (void) +{ + return mpfr_get_emax(); +} + +inline int mpreal::set_emax (mp_exp_t exp) +{ + return mpfr_set_emax(exp); +} + +inline mp_exp_t mpreal::get_emin_min (void) +{ + return mpfr_get_emin_min(); +} + +inline mp_exp_t mpreal::get_emin_max (void) +{ + return mpfr_get_emin_max(); +} + +inline mp_exp_t mpreal::get_emax_min (void) +{ + return mpfr_get_emax_min(); +} + +inline mp_exp_t mpreal::get_emax_max (void) +{ + return mpfr_get_emax_max(); +} + +////////////////////////////////////////////////////////////////////////// +// Mathematical Functions +////////////////////////////////////////////////////////////////////////// +#define MPREAL_UNARY_MATH_FUNCTION_BODY(f) \ + mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); \ + mpfr_##f(y.mpfr_ptr(), x.mpfr_srcptr(), r); \ + return y; + +inline const mpreal sqr (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqr ); } + +inline const mpreal sqrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ MPREAL_UNARY_MATH_FUNCTION_BODY(sqrt); } + +inline const mpreal sqrt(const unsigned long int x, mp_rnd_t r) +{ + mpreal y; + mpfr_sqrt_ui(y.mpfr_ptr(), x, r); + return y; +} + +inline const mpreal sqrt(const unsigned int v, mp_rnd_t rnd_mode) +{ + return sqrt(static_cast(v),rnd_mode); +} + +inline const mpreal sqrt(const long int v, mp_rnd_t rnd_mode) +{ + if (v>=0) return sqrt(static_cast(v),rnd_mode); + else return mpreal().setNan(); // NaN +} + +inline const mpreal sqrt(const int v, mp_rnd_t rnd_mode) +{ + if (v>=0) return sqrt(static_cast(v),rnd_mode); + else return mpreal().setNan(); // NaN +} + +inline const mpreal root(const mpreal& x, unsigned long int k, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, mpfr_get_prec(x.mpfr_srcptr())); + mpfr_root(y.mpfr_ptr(), x.mpfr_srcptr(), k, r); + return y; +} + +inline const mpreal dim(const mpreal& a, const mpreal& b, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, mpfr_get_prec(a.mpfr_srcptr())); + mpfr_dim(y.mpfr_ptr(), a.mpfr_srcptr(), b.mpfr_srcptr(), r); + return y; +} + +inline int cmpabs(const mpreal& a,const mpreal& b) +{ + return mpfr_cmpabs(a.mpfr_ptr(), b.mpfr_srcptr()); +} + +inline int sin_cos(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + return mpfr_sin_cos(s.mpfr_ptr(), c.mpfr_ptr(), v.mpfr_srcptr(), rnd_mode); +} + +inline const mpreal sqrt (const long double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } +inline const mpreal sqrt (const double v, mp_rnd_t rnd_mode) { return sqrt(mpreal(v),rnd_mode); } + +inline const mpreal cbrt (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cbrt ); } +inline const mpreal fabs (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } +inline const mpreal abs (const mpreal& x, mp_rnd_t r) { MPREAL_UNARY_MATH_FUNCTION_BODY(abs ); } +inline const mpreal log (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log ); } +inline const mpreal log2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log2 ); } +inline const mpreal log10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log10); } +inline const mpreal exp (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp ); } +inline const mpreal exp2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp2 ); } +inline const mpreal exp10 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(exp10); } +inline const mpreal cos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cos ); } +inline const mpreal sin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sin ); } +inline const mpreal tan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tan ); } +inline const mpreal sec (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sec ); } +inline const mpreal csc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csc ); } +inline const mpreal cot (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cot ); } +inline const mpreal acos (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acos ); } +inline const mpreal asin (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asin ); } +inline const mpreal atan (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atan ); } + +inline const mpreal logb (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { return log2 (abs(x),r); } + +inline const mpreal acot (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atan (1/v, r); } +inline const mpreal asec (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acos (1/v, r); } +inline const mpreal acsc (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asin (1/v, r); } +inline const mpreal acoth (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return atanh(1/v, r); } +inline const mpreal asech (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return acosh(1/v, r); } +inline const mpreal acsch (const mpreal& v, mp_rnd_t r = mpreal::get_default_rnd()) { return asinh(1/v, r); } + +inline const mpreal cosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(cosh ); } +inline const mpreal sinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sinh ); } +inline const mpreal tanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(tanh ); } +inline const mpreal sech (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(sech ); } +inline const mpreal csch (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(csch ); } +inline const mpreal coth (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(coth ); } +inline const mpreal acosh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(acosh); } +inline const mpreal asinh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(asinh); } +inline const mpreal atanh (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(atanh); } + +inline const mpreal log1p (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(log1p ); } +inline const mpreal expm1 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(expm1 ); } +inline const mpreal eint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(eint ); } +inline const mpreal gamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } +inline const mpreal tgamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(gamma ); } +inline const mpreal lngamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(lngamma); } +inline const mpreal zeta (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(zeta ); } +inline const mpreal erf (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erf ); } +inline const mpreal erfc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(erfc ); } +inline const mpreal besselj0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j0 ); } +inline const mpreal besselj1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(j1 ); } +inline const mpreal bessely0(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y0 ); } +inline const mpreal bessely1(const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(y1 ); } + +inline const mpreal atan2 (const mpreal& y, const mpreal& x, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_atan2(a.mpfr_ptr(), y.mpfr_srcptr(), x.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal hypot (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_hypot(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal remainder (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_remainder(a.mpfr_ptr(), x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal remquo (long* q, const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a(0,(std::max)(y.getPrecision(), x.getPrecision())); + mpfr_remquo(a.mpfr_ptr(),q, x.mpfr_srcptr(), y.mpfr_srcptr(), rnd_mode); + return a; +} + +inline const mpreal fac_ui (unsigned long int v, mp_prec_t prec = mpreal::get_default_prec(), + mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(0, prec); + mpfr_fac_ui(x.mpfr_ptr(),v,rnd_mode); + return x; +} + + +inline const mpreal lgamma (const mpreal& v, int *signp = 0, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(v); + int tsignp; + + if(signp) mpfr_lgamma(x.mpfr_ptr(), signp,v.mpfr_srcptr(),rnd_mode); + else mpfr_lgamma(x.mpfr_ptr(),&tsignp,v.mpfr_srcptr(),rnd_mode); + + return x; +} + + +inline const mpreal besseljn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, x.getPrecision()); + mpfr_jn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); + return y; +} + +inline const mpreal besselyn (long n, const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal y(0, x.getPrecision()); + mpfr_yn(y.mpfr_ptr(), n, x.mpfr_srcptr(), r); + return y; +} + +inline const mpreal fma (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2, p3; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + p3 = v3.get_prec(); + + a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); + + mpfr_fma(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); + return a; +} + +inline const mpreal fms (const mpreal& v1, const mpreal& v2, const mpreal& v3, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2, p3; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + p3 = v3.get_prec(); + + a.set_prec(p3>p2?(p3>p1?p3:p1):(p2>p1?p2:p1)); + + mpfr_fms(a.mp,v1.mp,v2.mp,v3.mp,rnd_mode); + return a; +} + +inline const mpreal agm (const mpreal& v1, const mpreal& v2, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t p1, p2; + + p1 = v1.get_prec(); + p2 = v2.get_prec(); + + a.set_prec(p1>p2?p1:p2); + + mpfr_agm(a.mp, v1.mp, v2.mp, rnd_mode); + + return a; +} + +inline const mpreal sum (const mpreal tab[], const unsigned long int n, int& status, mp_rnd_t mode = mpreal::get_default_rnd()) +{ + mpfr_srcptr *p = new mpfr_srcptr[n]; + + for (unsigned long int i = 0; i < n; i++) + p[i] = tab[i].mpfr_srcptr(); + + mpreal x; + status = mpfr_sum(x.mpfr_ptr(), (mpfr_ptr*)p, n, mode); + + delete [] p; + return x; +} + +////////////////////////////////////////////////////////////////////////// +// MPFR 2.4.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(2,4,0)) + +inline int sinh_cosh(mpreal& s, mpreal& c, const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + return mpfr_sinh_cosh(s.mp,c.mp,v.mp,rnd_mode); +} + +inline const mpreal li2 (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) +{ + MPREAL_UNARY_MATH_FUNCTION_BODY(li2); +} + +inline const mpreal rem (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + /* R = rem(X,Y) if Y != 0, returns X - n * Y where n = trunc(X/Y). */ + return fmod(x, y, rnd_mode); +} + +inline const mpreal mod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + (void)rnd_mode; + + /* + + m = mod(x,y) if y != 0, returns x - n*y where n = floor(x/y) + + The following are true by convention: + - mod(x,0) is x + - mod(x,x) is 0 + - mod(x,y) for x != y and y != 0 has the same sign as y. + + */ + + if(iszero(y)) return x; + if(x == y) return 0; + + mpreal m = x - floor(x / y) * y; + + m.setSign(sgn(y)); // make sure result has the same sign as Y + + return m; +} + +inline const mpreal fmod (const mpreal& x, const mpreal& y, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal a; + mp_prec_t yp, xp; + + yp = y.get_prec(); + xp = x.get_prec(); + + a.set_prec(yp>xp?yp:xp); + + mpfr_fmod(a.mp, x.mp, y.mp, rnd_mode); + + return a; +} + +inline const mpreal rec_sqrt(const mpreal& v, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(v); + mpfr_rec_sqrt(x.mp,v.mp,rnd_mode); + return x; +} +#endif // MPFR 2.4.0 Specifics + +////////////////////////////////////////////////////////////////////////// +// MPFR 3.0.0 Specifics +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) +inline const mpreal digamma (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(digamma); } +inline const mpreal ai (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(ai); } +#endif // MPFR 3.0.0 Specifics + +////////////////////////////////////////////////////////////////////////// +// Constants +inline const mpreal const_log2 (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_log2(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_pi (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_pi(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_euler (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_euler(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_catalan (mp_prec_t p = mpreal::get_default_prec(), mp_rnd_t r = mpreal::get_default_rnd()) +{ + mpreal x(0, p); + mpfr_const_catalan(x.mpfr_ptr(), r); + return x; +} + +inline const mpreal const_infinity (int sign = 1, mp_prec_t p = mpreal::get_default_prec()) +{ + mpreal x(0, p); + mpfr_set_inf(x.mpfr_ptr(), sign); + return x; +} + +////////////////////////////////////////////////////////////////////////// +// Integer Related Functions +inline const mpreal ceil(const mpreal& v) +{ + mpreal x(v); + mpfr_ceil(x.mp,v.mp); + return x; +} + +inline const mpreal floor(const mpreal& v) +{ + mpreal x(v); + mpfr_floor(x.mp,v.mp); + return x; +} + +inline const mpreal round(const mpreal& v) +{ + mpreal x(v); + mpfr_round(x.mp,v.mp); + return x; +} + +inline const mpreal trunc(const mpreal& v) +{ + mpreal x(v); + mpfr_trunc(x.mp,v.mp); + return x; +} + +inline const mpreal rint (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint ); } +inline const mpreal rint_ceil (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_ceil ); } +inline const mpreal rint_floor (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_floor); } +inline const mpreal rint_round (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_round); } +inline const mpreal rint_trunc (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(rint_trunc); } +inline const mpreal frac (const mpreal& x, mp_rnd_t r = mpreal::get_default_rnd()) { MPREAL_UNARY_MATH_FUNCTION_BODY(frac ); } + +////////////////////////////////////////////////////////////////////////// +// Miscellaneous Functions +inline void swap (mpreal& a, mpreal& b) { mpfr_swap(a.mp,b.mp); } +inline const mpreal (max)(const mpreal& x, const mpreal& y){ return (x>y?x:y); } +inline const mpreal (min)(const mpreal& x, const mpreal& y){ return (x= MPFR_VERSION_NUM(3,0,0)) +inline const mpreal urandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x; + mpfr_urandom(x.mpfr_ptr(), state, rnd_mode); + return x; +} +#endif + +#if (MPFR_VERSION <= MPFR_VERSION_NUM(2,4,2)) +inline const mpreal random2 (mp_size_t size, mp_exp_t exp) +{ + mpreal x; + mpfr_random2(x.mpfr_ptr(),size,exp); + return x; +} +#endif + +// Uniformly distributed random number generation +// a = random(seed); <- initialization & first random number generation +// a = random(); <- next random numbers generation +// seed != 0 +inline const mpreal random(unsigned int seed = 0) +{ +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,0,0)) + static gmp_randstate_t state; + static bool initialize = true; + + if(initialize) + { + gmp_randinit_default(state); + gmp_randseed_ui(state,0); + initialize = false; + } + + if(seed != 0) gmp_randseed_ui(state,seed); + + return mpfr::urandom(state); +#else + if(seed != 0) std::srand(seed); + return mpfr::mpreal(std::rand()/(double)RAND_MAX); +#endif + +} + +#if (MPFR_VERSION >= MPFR_VERSION_NUM(3,1,0)) + +inline const mpreal grandom (gmp_randstate_t& state, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x; + mpfr_grandom(x.mpfr_ptr(), NULL, state, rnd_mode); + return x; +} + +inline const mpreal grandom(unsigned int seed = 0) +{ + static gmp_randstate_t state; + static bool initialize = true; + + if(initialize) + { + gmp_randinit_default(state); + gmp_randseed_ui(state,0); + initialize = false; + } + + if(seed != 0) gmp_randseed_ui(state,seed); + + return mpfr::grandom(state); +} +#endif + +////////////////////////////////////////////////////////////////////////// +// Set/Get global properties +inline void mpreal::set_default_prec(mp_prec_t prec) +{ + mpfr_set_default_prec(prec); +} + +inline void mpreal::set_default_rnd(mp_rnd_t rnd_mode) +{ + mpfr_set_default_rounding_mode(rnd_mode); +} + +inline bool mpreal::fits_in_bits(double x, int n) +{ + int i; + double t; + return IsInf(x) || (std::modf ( std::ldexp ( std::frexp ( x, &i ), n ), &t ) == 0.0); +} + +inline const mpreal pow(const mpreal& a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow(x.mp,x.mp,b.mp,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const mpz_t b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_z(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const unsigned long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_ui(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const long int b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_pow_si(x.mp,x.mp,b,rnd_mode); + return x; +} + +inline const mpreal pow(const mpreal& a, const int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); +} + +inline const mpreal pow(const mpreal& a, const double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); +} + +inline const mpreal pow(const unsigned long int a, const mpreal& b, mp_rnd_t rnd_mode = mpreal::get_default_rnd()) +{ + mpreal x(a); + mpfr_ui_pow(x.mp,a,b.mp,rnd_mode); + return x; +} + +inline const mpreal pow(const unsigned int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),b,rnd_mode); +} + +inline const mpreal pow(const long int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),b,rnd_mode); + else return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const int a, const mpreal& b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),b,rnd_mode); + else return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const long double a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); +} + +inline const mpreal pow(const double a, const mpreal& b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); +} + +// pow unsigned long int +inline const mpreal pow(const unsigned long int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + mpreal x(a); + mpfr_ui_pow_ui(x.mp,a,b,rnd_mode); + return x; +} + +inline const mpreal pow(const unsigned long int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned long int a, const long int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(a,static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned long int a, const double b, mp_rnd_t rnd_mode) +{ + return pow(a,mpreal(b),rnd_mode); //mpfr_ui_pow +} + +// pow unsigned int +inline const mpreal pow(const unsigned int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui +} + +inline const mpreal pow(const unsigned int a, const long int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const int b, mp_rnd_t rnd_mode) +{ + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +inline const mpreal pow(const unsigned int a, const double b, mp_rnd_t rnd_mode) +{ + return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow +} + +// pow long int +inline const mpreal pow(const long int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long int a, const long int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const long int a, const int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const long int a, const long double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +inline const mpreal pow(const long int a, const double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +// pow int +inline const mpreal pow(const int a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),b,rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const int a, const unsigned int b, mp_rnd_t rnd_mode) +{ + if (a>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const int a, const long int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const int a, const int b, mp_rnd_t rnd_mode) +{ + if (a>0) + { + if(b>0) return pow(static_cast(a),static_cast(b),rnd_mode); //mpfr_ui_pow_ui + else return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + }else{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si + } +} + +inline const mpreal pow(const int a, const long double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +inline const mpreal pow(const int a, const double b, mp_rnd_t rnd_mode) +{ + if (a>=0) return pow(static_cast(a),mpreal(b),rnd_mode); //mpfr_ui_pow + else return pow(mpreal(a),mpreal(b),rnd_mode); //mpfr_pow +} + +// pow long double +inline const mpreal pow(const long double a, const long double b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),mpreal(b),rnd_mode); +} + +inline const mpreal pow(const long double a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long double a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); //mpfr_pow_ui +} + +inline const mpreal pow(const long double a, const long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const long double a, const int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const double a, const double b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),mpreal(b),rnd_mode); +} + +inline const mpreal pow(const double a, const unsigned long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_ui +} + +inline const mpreal pow(const double a, const unsigned int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_ui +} + +inline const mpreal pow(const double a, const long int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),b,rnd_mode); // mpfr_pow_si +} + +inline const mpreal pow(const double a, const int b, mp_rnd_t rnd_mode) +{ + return pow(mpreal(a),static_cast(b),rnd_mode); // mpfr_pow_si +} +} // End of mpfr namespace + +// Explicit specialization of std::swap for mpreal numbers +// Thus standard algorithms will use efficient version of swap (due to Koenig lookup) +// Non-throwing swap C++ idiom: http://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Non-throwing_swap +namespace std +{ + // we are allowed to extend namespace std with specializations only + template <> + inline void swap(mpfr::mpreal& x, mpfr::mpreal& y) + { + return mpfr::swap(x, y); + } + + template<> + class numeric_limits + { + public: + static const bool is_specialized = true; + static const bool is_signed = true; + static const bool is_integer = false; + static const bool is_exact = false; + static const int radix = 2; + + static const bool has_infinity = true; + static const bool has_quiet_NaN = true; + static const bool has_signaling_NaN = true; + + static const bool is_iec559 = true; // = IEEE 754 + static const bool is_bounded = true; + static const bool is_modulo = false; + static const bool traps = true; + static const bool tinyness_before = true; + + static const float_denorm_style has_denorm = denorm_absent; + + inline static mpfr::mpreal (min) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::minval(precision); } + inline static mpfr::mpreal (max) (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::maxval(precision); } + inline static mpfr::mpreal lowest (mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return -mpfr::maxval(precision); } + + // Returns smallest eps such that 1 + eps != 1 (classic machine epsilon) + inline static mpfr::mpreal epsilon(mp_prec_t precision = mpfr::mpreal::get_default_prec()) { return mpfr::machine_epsilon(precision); } + + // Returns smallest eps such that x + eps != x (relative machine epsilon) + inline static mpfr::mpreal epsilon(const mpfr::mpreal& x) { return mpfr::machine_epsilon(x); } + + inline static mpfr::mpreal round_error(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + mp_rnd_t r = mpfr::mpreal::get_default_rnd(); + + if(r == GMP_RNDN) return mpfr::mpreal(0.5, precision); + else return mpfr::mpreal(1.0, precision); + } + + inline static const mpfr::mpreal infinity() { return mpfr::const_infinity(); } + inline static const mpfr::mpreal quiet_NaN() { return mpfr::mpreal().setNan(); } + inline static const mpfr::mpreal signaling_NaN() { return mpfr::mpreal().setNan(); } + inline static const mpfr::mpreal denorm_min() { return (min)(); } + + // Please note, exponent range is not fixed in MPFR + static const int min_exponent = MPFR_EMIN_DEFAULT; + static const int max_exponent = MPFR_EMAX_DEFAULT; + MPREAL_PERMISSIVE_EXPR static const int min_exponent10 = (int) (MPFR_EMIN_DEFAULT * 0.3010299956639811); + MPREAL_PERMISSIVE_EXPR static const int max_exponent10 = (int) (MPFR_EMAX_DEFAULT * 0.3010299956639811); + +#ifdef MPREAL_HAVE_DYNAMIC_STD_NUMERIC_LIMITS + + // Following members should be constant according to standard, but they can be variable in MPFR + // So we define them as functions here. + // + // This is preferable way for std::numeric_limits specialization. + // But it is incompatible with standard std::numeric_limits and might not work with other libraries, e.g. boost. + // See below for compatible implementation. + inline static float_round_style round_style() + { + mp_rnd_t r = mpfr::mpreal::get_default_rnd(); + + switch (r) + { + case GMP_RNDN: return round_to_nearest; + case GMP_RNDZ: return round_toward_zero; + case GMP_RNDU: return round_toward_infinity; + case GMP_RNDD: return round_toward_neg_infinity; + default: return round_indeterminate; + } + } + + inline static int digits() { return int(mpfr::mpreal::get_default_prec()); } + inline static int digits(const mpfr::mpreal& x) { return x.getPrecision(); } + + inline static int digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + return mpfr::bits2digits(precision); + } + + inline static int digits10(const mpfr::mpreal& x) + { + return mpfr::bits2digits(x.getPrecision()); + } + + inline static int max_digits10(mp_prec_t precision = mpfr::mpreal::get_default_prec()) + { + return digits10(precision); + } +#else + // Digits and round_style are NOT constants when it comes to mpreal. + // If possible, please use functions digits() and round_style() defined above. + // + // These (default) values are preserved for compatibility with existing libraries, e.g. boost. + // Change them accordingly to your application. + // + // For example, if you use 256 bits of precision uniformly in your program, then: + // digits = 256 + // digits10 = 77 + // max_digits10 = 78 + // + // Approximate formula for decimal digits is: digits10 = floor(log10(2) * digits). See bits2digits() for more details. + + static const std::float_round_style round_style = round_to_nearest; + static const int digits = 53; + static const int digits10 = 15; + static const int max_digits10 = 16; +#endif + }; + +} + +#endif /* __MPREAL_H__ */ diff --git a/third_party/eigen/unsupported/test/polynomialsolver.cpp b/third_party/eigen/unsupported/test/polynomialsolver.cpp index db8ad7ddae3..4cfc46b41bf 100644 --- a/third_party/eigen/unsupported/test/polynomialsolver.cpp +++ b/third_party/eigen/unsupported/test/polynomialsolver.cpp @@ -26,25 +26,14 @@ struct increment_if_fixed_size } } -template -PolynomialType polyder(const PolynomialType& p) -{ - typedef typename PolynomialType::Scalar Scalar; - PolynomialType res(p.size()); - for(Index i=1; i bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) { typedef typename POLYNOMIAL::Scalar Scalar; - typedef typename POLYNOMIAL::RealScalar RealScalar; typedef typename SOLVER::RootsType RootsType; - typedef Matrix EvalRootsType; + typedef Matrix EvalRootsType; const Index deg = pols.size()-1; @@ -54,17 +43,10 @@ bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) psolve.compute( pols ); const RootsType& roots( psolve.roots() ); EvalRootsType evr( deg ); - POLYNOMIAL pols_der = polyder(pols); - EvalRootsType der( deg ); for( int i=0; i() ); + bool evalToZero = evr.isZero( test_precision() ); if( !evalToZero ) { cerr << "WRONG root: " << endl; @@ -74,7 +56,7 @@ bool aux_evalSolver( const POLYNOMIAL& pols, SOLVER& psolve ) cerr << endl; } - std::vector rootModuli( roots.size() ); + std::vector rootModuli( roots.size() ); Map< EvalRootsType > aux( &rootModuli[0], roots.size() ); aux = roots.array().abs(); std::sort( rootModuli.begin(), rootModuli.end() ); @@ -100,7 +82,7 @@ void evalSolver( const POLYNOMIAL& pols ) { typedef typename POLYNOMIAL::Scalar Scalar; - typedef PolynomialSolver PolynomialSolverType; + typedef PolynomialSolver PolynomialSolverType; PolynomialSolverType psolve; aux_evalSolver( pols, psolve ); @@ -114,7 +96,6 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const { using std::sqrt; typedef typename POLYNOMIAL::Scalar Scalar; - typedef typename POLYNOMIAL::RealScalar RealScalar; typedef PolynomialSolver PolynomialSolverType; @@ -125,12 +106,14 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const // 1) the roots found are correct // 2) the roots have distinct moduli + typedef typename REAL_ROOTS::Scalar Real; + //Test realRoots - std::vector< RealScalar > calc_realRoots; - psolve.realRoots( calc_realRoots, test_precision()); - VERIFY_IS_EQUAL( calc_realRoots.size() , (size_t)real_roots.size() ); + std::vector< Real > calc_realRoots; + psolve.realRoots( calc_realRoots ); + VERIFY( calc_realRoots.size() == (size_t)real_roots.size() ); - const RealScalar psPrec = sqrt( test_precision() ); + const Scalar psPrec = sqrt( test_precision() ); for( size_t i=0; i 0 ) ); if( hasRealRoot ){ VERIFY( internal::isApprox( real_roots.array().abs().maxCoeff(), abs(r), psPrec ) ); } @@ -182,11 +165,9 @@ void evalSolverSugarFunction( const POLYNOMIAL& pols, const ROOTS& roots, const template void polynomialsolver(int deg) { - typedef typename NumTraits<_Scalar>::Real RealScalar; - typedef internal::increment_if_fixed_size<_Deg> Dim; + typedef internal::increment_if_fixed_size<_Deg> Dim; typedef Matrix<_Scalar,Dim::ret,1> PolynomialType; typedef Matrix<_Scalar,_Deg,1> EvalRootsType; - typedef Matrix RealRootsType; cout << "Standard cases" << endl; PolynomialType pols = PolynomialType::Random(deg+1); @@ -199,11 +180,15 @@ void polynomialsolver(int deg) evalSolver<_Deg,PolynomialType>( pols ); cout << "Test sugar" << endl; - RealRootsType realRoots = RealRootsType::Random(deg); + EvalRootsType realRoots = EvalRootsType::Random(deg); roots_to_monicPolynomial( realRoots, pols ); evalSolverSugarFunction<_Deg>( pols, - realRoots.template cast >().eval(), + realRoots.template cast < + std::complex< + typename NumTraits<_Scalar>::Real + > + >(), realRoots ); } @@ -227,6 +212,5 @@ void test_polynomialsolver() internal::random(9,13) )) ); CALL_SUBTEST_11((polynomialsolver(1)) ); - CALL_SUBTEST_12((polynomialsolver,Dynamic>(internal::random(2,13))) ); } } From 816e4edcfd9679d3215e9ca632f5f45f7640e2a0 Mon Sep 17 00:00:00 2001 From: j-c-cook Date: Tue, 17 Aug 2021 17:58:59 -0500 Subject: [PATCH 73/77] Bump cpgfunctionEP to 0.5 --- .../.gitignore | 0 .../CHANGELOG.md | 10 +++ .../CMakeLists.txt | 4 + .../LICENSE | 0 .../README.md | 0 .../include/blas/blas.h | 0 .../include/cpgfunction/boreholes.h | 0 .../include/cpgfunction/coordinates.h | 0 .../include/cpgfunction/gfunction.h | 0 .../include/cpgfunction/heat_transfer.h | 0 .../include/cpgfunction/interpolation.h | 0 .../include/cpgfunction/segments.h | 0 .../include/cpgfunction/statistics.h | 0 .../include/cpgfunction/utilities.h | 0 .../src/blas.cpp | 0 .../src/boreholes.cpp | 0 .../src/coordinates.cpp | 0 .../src/gfunction.cpp | 34 ++++---- .../src/heat_transfer.cpp | 0 .../src/interpolation.cpp | 0 .../src/segments.cpp | 0 .../src/statistics.cpp | 0 .../src/utilities.cpp | 0 .../LU-Decomposition-0.1/CHANGELOG.md | 9 ++ .../LU-Decomposition-0.1/CMakeLists.txt | 20 +++++ .../third_party/LU-Decomposition-0.1/LICENSE | 29 +++++++ .../LU-Decomposition-0.1/README.md | 6 ++ .../include/LU-Decomposition/lu.h | 21 +++++ .../LU-Decomposition-0.1/src/lu.cpp | 85 +++++++++++++++++++ .../LU-Decomposition-0.1/tests/solve.cpp | 53 ++++++++++++ .../third_party/qdt-master/LICENSE.txt | 0 .../third_party/qdt-master/README.md | 0 .../third_party/qdt-master/common/method.h | 0 .../third_party/qdt-master/methods/adaptive.h | 0 .../qdt-master/methods/gauss-kronrod.h | 0 .../third_party/qdt-master/qdt.h | 0 36 files changed, 255 insertions(+), 16 deletions(-) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/.gitignore (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/CHANGELOG.md (95%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/CMakeLists.txt (95%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/LICENSE (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/README.md (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/blas/blas.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/boreholes.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/coordinates.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/gfunction.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/heat_transfer.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/interpolation.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/segments.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/statistics.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/include/cpgfunction/utilities.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/blas.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/boreholes.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/coordinates.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/gfunction.cpp (96%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/heat_transfer.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/interpolation.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/segments.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/statistics.cpp (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/src/utilities.cpp (100%) create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/CHANGELOG.md create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/CMakeLists.txt create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/LICENSE create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/README.md create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/include/LU-Decomposition/lu.h create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp create mode 100644 third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/tests/solve.cpp rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/LICENSE.txt (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/README.md (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/common/method.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/methods/adaptive.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/methods/gauss-kronrod.h (100%) rename third_party/{cpgfunctionEP-0.4.1 => cpgfunctionEP-0.5}/third_party/qdt-master/qdt.h (100%) diff --git a/third_party/cpgfunctionEP-0.4.1/.gitignore b/third_party/cpgfunctionEP-0.5/.gitignore similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/.gitignore rename to third_party/cpgfunctionEP-0.5/.gitignore diff --git a/third_party/cpgfunctionEP-0.4.1/CHANGELOG.md b/third_party/cpgfunctionEP-0.5/CHANGELOG.md similarity index 95% rename from third_party/cpgfunctionEP-0.4.1/CHANGELOG.md rename to third_party/cpgfunctionEP-0.5/CHANGELOG.md index dfc28804cf3..50f95e66535 100644 --- a/third_party/cpgfunctionEP-0.4.1/CHANGELOG.md +++ b/third_party/cpgfunctionEP-0.5/CHANGELOG.md @@ -1,5 +1,15 @@ # History of changes - cpgfunctionEP +## Version 0.5 (2021-08-17) + +### Changes + +* [Issue 46](https://github.com/j-c-cook/cpgfunctionEP/issues/46) - A library named LU-Decomposition is added. The library only includes minimal functions to perform LU-decomposition and solve a system of linear equations. + +### Removes + +* [Issue 44](https://github.com/j-c-cook/cpgfunctionEP/issues/44) - The dependency on the Eigen library is removed due to an unexplainable segment fault that occurred only during the Linux Ubuntu tests on the EnergyPlus CI machines. + ## Version 0.4 (2021-07-30) ### Build Changes diff --git a/third_party/cpgfunctionEP-0.4.1/CMakeLists.txt b/third_party/cpgfunctionEP-0.5/CMakeLists.txt similarity index 95% rename from third_party/cpgfunctionEP-0.4.1/CMakeLists.txt rename to third_party/cpgfunctionEP-0.5/CMakeLists.txt index d3f9372ed87..55718f63760 100644 --- a/third_party/cpgfunctionEP-0.4.1/CMakeLists.txt +++ b/third_party/cpgfunctionEP-0.5/CMakeLists.txt @@ -11,6 +11,8 @@ endif() include_directories(include) # Third party includes include_directories(third_party/qdt-master) +add_subdirectory(third_party/LU-Decomposition-0.1) +include_directories(third_party/LU-Decomposition-0.1/include) # Create a DLL with an API add_library(cpgfunctionEP src/boreholes.cpp @@ -23,6 +25,8 @@ add_library(cpgfunctionEP src/boreholes.cpp src/segments.cpp src/blas.cpp) +target_link_libraries(cpgfunctionEP INTERFACE LU-Decomposition) + # It depends on threads find_package(Threads REQUIRED) target_link_libraries(cpgfunctionEP INTERFACE Threads::Threads) diff --git a/third_party/cpgfunctionEP-0.4.1/LICENSE b/third_party/cpgfunctionEP-0.5/LICENSE similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/LICENSE rename to third_party/cpgfunctionEP-0.5/LICENSE diff --git a/third_party/cpgfunctionEP-0.4.1/README.md b/third_party/cpgfunctionEP-0.5/README.md similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/README.md rename to third_party/cpgfunctionEP-0.5/README.md diff --git a/third_party/cpgfunctionEP-0.4.1/include/blas/blas.h b/third_party/cpgfunctionEP-0.5/include/blas/blas.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/blas/blas.h rename to third_party/cpgfunctionEP-0.5/include/blas/blas.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/boreholes.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/boreholes.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/boreholes.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/boreholes.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/coordinates.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/coordinates.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/coordinates.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/coordinates.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/gfunction.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/gfunction.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/gfunction.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/gfunction.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/heat_transfer.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/heat_transfer.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/heat_transfer.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/heat_transfer.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/interpolation.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/interpolation.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/interpolation.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/interpolation.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/segments.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/segments.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/segments.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/segments.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/statistics.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/statistics.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/statistics.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/statistics.h diff --git a/third_party/cpgfunctionEP-0.4.1/include/cpgfunction/utilities.h b/third_party/cpgfunctionEP-0.5/include/cpgfunction/utilities.h similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/include/cpgfunction/utilities.h rename to third_party/cpgfunctionEP-0.5/include/cpgfunction/utilities.h diff --git a/third_party/cpgfunctionEP-0.4.1/src/blas.cpp b/third_party/cpgfunctionEP-0.5/src/blas.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/src/blas.cpp rename to third_party/cpgfunctionEP-0.5/src/blas.cpp diff --git a/third_party/cpgfunctionEP-0.4.1/src/boreholes.cpp b/third_party/cpgfunctionEP-0.5/src/boreholes.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/src/boreholes.cpp rename to third_party/cpgfunctionEP-0.5/src/boreholes.cpp diff --git a/third_party/cpgfunctionEP-0.4.1/src/coordinates.cpp b/third_party/cpgfunctionEP-0.5/src/coordinates.cpp similarity index 100% rename from third_party/cpgfunctionEP-0.4.1/src/coordinates.cpp rename to third_party/cpgfunctionEP-0.5/src/coordinates.cpp diff --git a/third_party/cpgfunctionEP-0.4.1/src/gfunction.cpp b/third_party/cpgfunctionEP-0.5/src/gfunction.cpp similarity index 96% rename from third_party/cpgfunctionEP-0.4.1/src/gfunction.cpp rename to third_party/cpgfunctionEP-0.5/src/gfunction.cpp index 225be989507..43312adfe69 100644 --- a/third_party/cpgfunctionEP-0.4.1/src/gfunction.cpp +++ b/third_party/cpgfunctionEP-0.5/src/gfunction.cpp @@ -6,7 +6,7 @@ #include #include -#include +#include #include #include #include @@ -187,9 +187,8 @@ namespace gt::gfunction { * **/ int SIZE = nSources + 1; - Eigen::MatrixXd A(SIZE, SIZE); - Eigen::MatrixXd B(SIZE, 1); - Eigen::MatrixXd X(SIZE, 1); + vector > A(SIZE, vector (SIZE, 0)); + vector B (SIZE, 0); // Fill A int n = SIZE - 1; @@ -231,19 +230,19 @@ namespace gt::gfunction { for (int j=0; j @@ -302,11 +301,14 @@ namespace gt::gfunction { // ----- LU decomposition ----- start = std::chrono::steady_clock::now(); - X = A.lu().solve(B); + vector indx(SIZE, 0); + double d; + jcc::decomposition(A, SIZE, indx, d); + jcc::back_substitution(A, SIZE, indx, B); - for (int i=0; i (end - start).count(); @@ -314,10 +316,10 @@ namespace gt::gfunction { // ---- Save Q's for next p --- for (int j=0; j +#include +#include +using namespace std; + +#ifndef LU_DECOMPOSITION_LU_H +#define LU_DECOMPOSITION_LU_H + +namespace jcc { + void decomposition(vector > &A, int &n, vector &indx, + double &d); + + void back_substitution(vector > &A, int &n, vector indx, + vector &b); +} + +#endif //LU_DECOMPOSITION_LU_H diff --git a/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp b/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp new file mode 100644 index 00000000000..52fb82d9ce4 --- /dev/null +++ b/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp @@ -0,0 +1,85 @@ +// +// Created by jackcook on 8/17/21. +// + +#include + + +void jcc::decomposition(vector > &A, int &n, vector &indx, + double &d) { + const double TINY = 1.0e-20; // a small number + int i, imax, j, k; + double big, dum, sum, temp; + + vector vv(n); // vv stores the implicit scaling of each row + d = 1.0; // No row interchanges yet + for (i=0; i big) big = fabs(A[i][j]); + } // next j + if (big == 0.0) throw invalid_argument("Singular matrix in routine " + "decomposition"); + // No nonzero biggest element + vv[i] = 1.0/big; // save the scaling + } // next i + + for (j=0; j= big) { + big = fabs(sum); + imax = i; + } // end if() + } // next i + if (j != imax) { + for (k=0; k > &A, int &n, vector indx, + vector &b) { + int i, ii=0, ip, j; + double sum; + + for (i=0; i=0;i--) { + sum=b[i]; + for (j=i+1;j + +#include +#include + +int main() { + + int n = 3; + + // Define 2D vector A of nxn full of zeros + vector > A(n, vector (n, 0)); + // {5., 9., 10.}, + // {2., 7., 3.}, + // {8., 2., 4.} + A[0][0] = 5.; + A[0][1] = 9.; + A[0][2] = 10.; + A[1][0] = 2.; + A[1][1] = 7.; + A[1][2] = 3.; + A[2][0] = 8.; + A[2][1] = 2.; + A[2][2] = 4.; + + vector b(n, 0); + b[0] = 22.; + b[1] = 13.; + b[2] = 17.; + + vector indx(n); + + double d; + + jcc::decomposition(A, n, indx, d); + + std::cout << "A matrix after decomposition function" << std::endl; + for (size_t i=0; i Date: Tue, 17 Aug 2021 18:00:35 -0500 Subject: [PATCH 74/77] Make EP's cmake point to cpgfunction-v0.5 --- CMakeLists.txt | 4 ++-- third_party/CMakeLists.txt | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/CMakeLists.txt b/CMakeLists.txt index c21cadd247a..2399e0048c6 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -265,8 +265,8 @@ target_include_directories(project_options SYSTEM INTERFACE "${kiva_SOURCE_DIR}/ add_subdirectory(scripts/dev/generate_embeddable_epJSON_schema) set_target_properties(generate_embeddedable_epJSON_schema PROPERTIES FOLDER "Internal") -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4.1) -target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.4.1/include) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.5) +target_include_directories(project_options INTERFACE ${PROJECT_SOURCE_DIR}/third_party/cpgfunctionEP-0.5/include) if(OPENGL_FOUND) set(BUILD_PENUMBRA_TESTING diff --git a/third_party/CMakeLists.txt b/third_party/CMakeLists.txt index 3fef930ba36..e247b96bc04 100644 --- a/third_party/CMakeLists.txt +++ b/third_party/CMakeLists.txt @@ -124,8 +124,8 @@ add_subdirectory(re2) set_target_properties(re2 PROPERTIES FOLDER ThirdParty/re2) add_subdirectory(fmt-6.1.2) set_target_properties(fmt PROPERTIES FOLDER ThirdParty/fmt) -add_subdirectory(cpgfunctionEP-0.4.1) -set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.4.1) +add_subdirectory(cpgfunctionEP-0.5) +set_target_properties(cpgfunctionEP PROPERTIES FOLDER ThirdParty/cpgfunctionEP-0.5) # Kiva include(cmake/kiva.cmake) From 22f3445b9ca16b6ee36a8acc3d6c8ec4ef7338ae Mon Sep 17 00:00:00 2001 From: Edwin Lee Date: Wed, 18 Aug 2021 15:52:41 -0500 Subject: [PATCH 75/77] Add clarification in cli usage --- src/EnergyPlus/CommandLineInterface.cc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/EnergyPlus/CommandLineInterface.cc b/src/EnergyPlus/CommandLineInterface.cc index 55a48ef012f..e2c3783510a 100644 --- a/src/EnergyPlus/CommandLineInterface.cc +++ b/src/EnergyPlus/CommandLineInterface.cc @@ -152,7 +152,7 @@ namespace CommandLineInterface { opt.add("", false, 0, 0, "Display version information", "-v", "--version"); - opt.add("1", false, 1, 0, "Multi-thread with N threads; 1 thread with no arg.", "-j", "--jobs"); + opt.add("1", false, 1, 0, "Multi-thread with N threads; 1 thread with no arg. (Currently only for G-Function generation)", "-j", "--jobs"); opt.add("in.epw", false, 1, 0, "Weather file path (default: in.epw in current directory)", "-w", "--weather"); From b7ff75b84e96514076155ac521b501f6df961640 Mon Sep 17 00:00:00 2001 From: Matt Mitchell Date: Thu, 19 Aug 2021 09:31:11 -0600 Subject: [PATCH 76/77] add ghe transition rules --- src/Transition/CreateNewIDFUsingRulesV9_6_0.f90 | 12 ++++++++++++ .../InputRulesFiles/Rules9-5-0-to-9-6-0.md | 10 +++++++++- 2 files changed, 21 insertions(+), 1 deletion(-) diff --git a/src/Transition/CreateNewIDFUsingRulesV9_6_0.f90 b/src/Transition/CreateNewIDFUsingRulesV9_6_0.f90 index 05cd013b992..b1623002a53 100644 --- a/src/Transition/CreateNewIDFUsingRulesV9_6_0.f90 +++ b/src/Transition/CreateNewIDFUsingRulesV9_6_0.f90 @@ -409,6 +409,18 @@ SUBROUTINE CreateNewIDFUsingRules(EndOfFile,DiffOnly,InLfn,AskForInput,InputFile ! If your original object starts with G, insert the rules here + CASE('GROUNDHEATEXCHANGER:SYSTEM') + CALL GetNewObjectDefInIDD(ObjectName,NwNumArgs,NwAorN,NwReqFld,NwObjMinFlds,NwFldNames,NwFldDefaults,NwFldUnits) + nodiff=.false. + OutArgs(1:9)=InArgs(1:9) + IF (CurArgs .gt. 9) THEN + OutArgs(11:CurArgs+1)=InArgs(10:CurArgs) + IF (InArgs(9) .eq. '') THEN + OutArgs(10)='UHFCALC' + ENDIF + CurArgs=CurArgs+1 + ENDIF + ! If your original object starts with H, insert the rules here ! If your original object starts with I, insert the rules here diff --git a/src/Transition/InputRulesFiles/Rules9-5-0-to-9-6-0.md b/src/Transition/InputRulesFiles/Rules9-5-0-to-9-6-0.md index c3c6f581b3a..44929de2046 100644 --- a/src/Transition/InputRulesFiles/Rules9-5-0-to-9-6-0.md +++ b/src/Transition/InputRulesFiles/Rules9-5-0-to-9-6-0.md @@ -24,9 +24,17 @@ See [8466](https://github.com/NREL/EnergyPlus/pull/8466) # Object Change: AirLoopHVAC:OutdoorAirSystem -Summary: The fourth field (Availability Manager name) was being ignored by EnergyPlus other than to verify that it was a valid availability manager (if it wasnt blank). It was not a required input and E+ wasnt doing anything with it. It was decided to remove this field to avoid any confusion. +Summary: The fourth field (Availability Manager name) was being ignored by EnergyPlus other than to verify that it was a valid availability manager (if it wasn't blank). It was not a required input and E+ wasn't doing anything with it. It was decided to remove this field to avoid any confusion. Field 1-3 remain the same. Field 4 has been eliminated/deleted and not replace with anything else. See [8884](https://github.com/NREL/EnergyPlus/pull/8884) + +# Object Change: GroundHeatExchanger:System + +Summary: A new field was added to specify the method used to compute the g-function values. + +Fields 1-9 remain the same. +New Field F10 (A7) g-Function Calculation Method. Leave blank or assign a value of 'UHFcalc' or 'UBHWTcalc'. Default value is 'UHFcalc'. +Remaining fields remain the same and are shifted. From 4c9f8326459a2fbfd2b71575024ff02dcafc12e2 Mon Sep 17 00:00:00 2001 From: Edwin Lee Date: Thu, 19 Aug 2021 12:58:55 -0500 Subject: [PATCH 77/77] Address Windows build warnings --- third_party/cpgfunctionEP-0.5/src/segments.cpp | 2 +- .../third_party/LU-Decomposition-0.1/src/lu.cpp | 2 +- .../cpgfunctionEP-0.5/third_party/qdt-master/methods/adaptive.h | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/third_party/cpgfunctionEP-0.5/src/segments.cpp b/third_party/cpgfunctionEP-0.5/src/segments.cpp index ce90f735856..c0f18937ba5 100644 --- a/third_party/cpgfunctionEP-0.5/src/segments.cpp +++ b/third_party/cpgfunctionEP-0.5/src/segments.cpp @@ -49,7 +49,7 @@ namespace gt::segments { nq = jcc::interpolation::interp1d(drilling_depth, drilling_depths[0], ideal_segment_lengths[0]); - } else if(heights[0] < height < heights[n1]) { + } else if(heights[0] < height && height < heights[n1]) { // do bilinear interpolation // find the lower and upper height values bool finished = false; diff --git a/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp b/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp index 52fb82d9ce4..507598fe9b9 100644 --- a/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp +++ b/third_party/cpgfunctionEP-0.5/third_party/LU-Decomposition-0.1/src/lu.cpp @@ -9,7 +9,7 @@ void jcc::decomposition(vector > &A, int &n, vector &indx, double &d) { const double TINY = 1.0e-20; // a small number int i, imax, j, k; - double big, dum, sum, temp; + double big, dum, sum; vector vv(n); // vv stores the implicit scaling of each row d = 1.0; // No row interchanges yet diff --git a/third_party/cpgfunctionEP-0.5/third_party/qdt-master/methods/adaptive.h b/third_party/cpgfunctionEP-0.5/third_party/qdt-master/methods/adaptive.h index 282161ebef3..44bd686a138 100644 --- a/third_party/cpgfunctionEP-0.5/third_party/qdt-master/methods/adaptive.h +++ b/third_party/cpgfunctionEP-0.5/third_party/qdt-master/methods/adaptive.h @@ -376,7 +376,7 @@ class Adaptive : template Adaptive adaptive(const BM& bm) { - return Adaptive(bm, 1.e-5, 1.e-10); + return Adaptive(bm, 1.e-5f, 1.e-10f); } template